I. Introduction

Top Next References

Cell adhesion is critical for the genesis and maintenance of both three-dimensional structure and normal function in tissues. The biochemical entities mediating cell adhesion are multiprotein complexes comprising three broad classes of macromolecules: the adhesion receptors, the extracellular matrix molecules, and the adhesion plaque proteins (Gumbiner, 1996). Cell adhesion receptors are typically transmembrane glycoproteins that mediate binding to extracellular matrix (ECM)^b molecules or to counter-receptors on other cells; these molecules determine the specificity of cell-cell or cell-ECM interaction. The ECM proteins are usually fibrillar in nature and provide a complex structural and functional network that can interact simultaneously with multiple cell surface receptors. The intracellular plaque proteins (or peripheral membrane proteins) provide structural and functional linkages between adhesion receptors and the actin microfilaments, microtubules, and intermediate filaments of the cytoskeleton. An exciting concept that has emerged from recent cell biological research is that cell adhesion complexes are not simply static architectural entities. Rather, they are dynamic units that are capable of capturing and integrating signals from the extracellular environment (Rosales et al., 1995). Moreover, the functions of cell adhesion complexes are regulated precisely by biochemical events within the cell. Thus cell adhesion receptors are at a nexus of two-way signaling between the cell and its external environment. This review will examine current knowledge concerning signal transduction events initiated or modulated by cell adhesion receptors. The emphasis will be on the integrin family of receptors because their signaling functions have been studied more extensively than those of other adhesion receptor families. However, we also will examine signaling events involving the cadherin, immunoglobulin-cell adhesion molecule (Ig-CAM), and selectin families of adhesion receptors. Adhesion-mediated signaling influences several critical cellular processes including gene expression, cell cycle, and programmed cell death; these aspects will be explored. However, the major emphasis of this review will be on mechanisms of adhesion receptor signaling.

	II. Families of Cell Adhesion Receptors

Top Previous Next References

This section will review some of the key features of the integrin, cadherin, Ig-CAM, and selectin families of cell adhesion receptors. Because there are thousands of publications on these topics (Edelman, 1993), only a few research articles and reviews will be used here to introduce some key ideas about these molecules. A diagram of the structures of several types of adhesion receptors, their counter-receptors, and some of their associated cytoskeletal components is presented in fig. 1.

View larger version (30K): [in this window] [in a new window]   Fig. 1.   Families of cell adhesion receptors. This diagram presents basic information about the integrin, cadherin, Ig-CAM, and selectin families of cell adhesion receptors. It illustrates their approximate structures, their ligands or counter-receptors, and some of their associated cytoskeletal proteins when they are known. A more detailed description is given in the text. The molecules illustrated here are not drawn to scale.

The integrins are a family of cell-surface glycoproteins that act as receptors for ECM proteins, or for membrane-bound counter-receptors on other cells. Integrin-mediated cell-ECM adhesion sites are complex specialized structures termed focal contacts or focal adhesions (Jockush et al., 1995) and are discussed in more detail in Section III. Each integrin is a heterodimer that contains an  and a  subunit with each subunit having a large extracellular domain, a single membrane-spanning region, and in most cases (other than 4), a short cytoplasmic domain (Hynes, 1992; Rosales et al., 1992; Ruoslahti, 1991). The integrin receptor family of vertebrates includes at least 16 distinct  subunits and 8 or more  subunits which can associate to form more than 20 distinct integrins (Hynes, 1992; Rosales et al., 1995). The / pairings specify the ligand-binding abilities of the integrin heterodimers. Although the ligands for integrins are often large ECM proteins such as collagen, laminin, vitronectin, or fibronectin, some integrins recognize rather short peptide sequences within the larger protein, for example, the RGD (Arg-Gly-Asp) sequence found in fibronectin and vitronectin. Because of this, there has been considerable interest in the pharmaceutical industry in developing short peptides or peptidomimetics that can interdict integrin functions in a variety of disease processes including coagulation disorders, inflammation, and cancer (Arap and Pasqualini, 1998; Ruoslahti, 1996). In other cases, however, integrin-ligand recognition depends on the overall conformation of the ligand protein (Kuhn and Eble, 1994). For example, some integrins interact with members of other adhesion receptor families, including Ig-CAMs and cadherins, in a manner that does not involve RGD motifs (Felding-Habermann et al., 1997; Higgins et al., 1998; Piali et al., 1995). Some integrins, such as 51, the "classic" fibronectin receptor, bind to a single ECM protein. More generally, an individual integrin will recognize several distinct proteins (Hynes, 1992; Rosales and Juliano, 1995); for example, the v3 integrin has been reported to bind collagen VI, laminin, fibronectin, vitronectin, thrombospondin, Von Willebrand factor, and fibrinogen (Kuhn and Eble, 1994). Cells often display multiple integrins capable of interacting with a particular ECM protein, thus integrin expression is often apparently redundant, at least in terms of simple cell adhesion. Some integrin subunits undergo alternative splicing of their cytoplasmic domain regions in a tissue-type specific and developmentally regulated manner; this suggests that there are discrete intracellular functions for individual integrins, as we will discuss in much more detail below.

The relationships between integrin structure and the various functions of integrins are being actively investigated. In terms of ligand binding, it seems clear that both  and  subunit extracellular domains contribute to the formation of the binding site. The ligand binding regions of integrins have been explored with chemical cross-linking of ligands, monoclonal antibodies, mutation, and most recently, molecular modeling and X-ray crystallography (Loftus et al., 1994; Qu and Leahy, 1995; Springer, 1997). Three regions apparently are particularly important: (a) a series of seven repeats of approximately 60 amino acids in the N-terminal portion of the  chain, each containing a putative Ca²⁺ binding site; (b) an inserted domain (I-domain) of approximately 200 amino acids found in several  chains and containing a nucleotide binding fold and a divalent cation coordination site; (c) an I-domain like region of approximately 250 amino acids found in the N-terminal region of the  subunit (Loftus and Liddington, 1997). The seven-repeat sequences of the  chain have been predicted by molecular modeling to form a -propeller structure with the upper face of the propeller being the ligand binding region and the Ca²⁺ coordination motifs lying on the lower face (but probably essential to maintain the structure) (Loftus and Liddington, 1997; Springer, 1997). Propeller structures of this type have been found in a variety of other proteins, including the  subunit of trimeric G-proteins (Sondek et al., 1996), and usually are involved in protein-protein interactions. The structure of the I-domain region has been determined by X-ray crystallography (Lee et al., 1995; Qu and Leahy, 1995); the divalent ion binding motif has been designated the "metal ion-dependent adhesion site" and seems to play a critical role in ligand coordination for integrins with I-domains. Less information is available about the I-domain-like region in the  subunit; however, molecular modeling and mutagenesis studies support its ligand binding role (Loftus and Liddington, 1997). Integrins likely undergo dynamic structural changes as part of the ligand binding process including relative movements of subunits and of domains, and conformational changes within domains (Loftus and Liddington, 1997). As we will discuss in more detail below, integrins can exist in various affinity states for their ligands; structural changes that occur in moving from low- to high-affinity states (or vice versa) can be detected by certain monoclonal antibodies (Hughes et al., 1997).

Integrin cytoplasmic domains are a key nexus of interaction between the extracellular environment and intracellular structures and signaling cascades. Both the  and  subunit cytoplasmic domains make important contributions to various aspects of overall integrin function including cytoskeletal organization, cell motility, signal transduction, and modulation of integrin affinity for ligands ("activation"). Several cytoplasmic proteins including talin, -actinin, and possibly focal adhesion kinase (FAK) bind directly to the 1 cytoplasmic domain and contribute to integrin-cytoskeletal interactions (Burridge and Chrzanowska-Wodnicka, 1996; Yamada and Geiger, 1997). Studies using truncation, mutation, or "domain swaps" of cytoplasmic domains have delineated several important structure-function relationships for this region of integrins. The  cytoplasmic domain is critical for recruitment of integrins to focal contacts because its truncation/mutation impairs this process; in fact, the  cytoplasmic domain alone, expressed as a fusion chimera with another membrane protein, seems sufficient for focal contact localization (LaFlamme et al., 1992; Reszka et al., 1992; Rosales et al., 1992). A widely accepted current model is that the  subunit cytoplasmic domain inhibits certain functions of the  cytoplasmic domain (e.g., focal contact recruitment) but that binding of a ligand to the integrin relieves this inhibition, possibly by allowing the subunits to swing apart like a hinge (Burridge and Chrzanowska-Wodnicka, 1996; Hughes et al., 1996). The  cytoplasmic domain is also important in signal transduction, particularly integrin activation of FAK (discussed in detail below), whereas truncation/mutation of the  cytoplasmic domain has little effect on this process (Akiyama et al., 1994; Bauer et al., 1993; Tahiliani et al., 1997).  cytoplasmic domains have played a critical role in a variety of cellular processes including endocytosis (Van Nhieu et al., 1996), "cross-talk" between different integrins (Blystone et al., 1995), assembly of fibronectin fibrils (Wu et al., 1995), and cell motility (Pasqualini and Hemler, 1994), whereas  cytoplasmic domains also can strongly influence cell motility (Bauer et al., 1993; Chan et al., 1992)_.

Like other receptors, integrins can exist in different affinity states with respect to their ligands. Integrins can enter a high-affinity ("activated") state in response to certain agents that bind the extracellular domain and influence its conformation (divalent cations, antibodies); they also can respond to signals generated within the cell that presumably have an impact on the cytoplasmic domain ("inside-out signaling") (Humphries, 1996; Keely et al., 1998; O'Toole et al., 1994). The activation status of a particular integrin is cell-type dependent and usually critically depends on cellular energy metabolism. The roles of the cytoplasmic domains in integrin activation have been studied extensively, and the findings are quite complex. For example, partial truncation of 4, 2, or 2 cytoplasmic domains prevents integrin activation, whereas truncation of the IIb subunit activates the fibrinogen-binding integrin IIb3 (O'Toole et al., 1994). Recently, a coherent model of how cytoplasmic domain structure relates to integrin activation has been presented (Hughes et al., 1996). The membrane proximal regions of all  and  subunits are highly conserved, whereas the remainder of the subunits are quite divergent. The conserved  sequence is GFFKR, whereas the conserved  sequence is LLv-iHDR. Deletion of either of these conserved sequences activates the integrin, essentially "locking" it in a high-affinity conformation independent of cellular energy metabolism, whereas mutation at sites C-terminal to these conserved sequences can affect energy-dependent activation (Hughes et al., 1996; O'Toole et al., 1994). The conserved membrane proximal sequences may interact through salt bridges between the subunits, normally keeping the integrin in a low-affinity state. Cytoplasmic events, that is "inside-out signaling," can disrupt the / association allowing the "hinge" to swing and opening up the extracellular ligand binding site (see fig. 2).

View larger version (18K): [in this window] [in a new window]   Fig. 2.   Activation of integrins. Both high-affinity and low-affinity binding states exist for integrins. This may be because of conformational changes that occur when regulatory proteins bind to integrin cytoplasmic tails. The conformational changes are depicted here as integrin subunits swinging on a "hinge" and exposing ligand binding sites (I, I and propeller "prop"). The activating process is regulated by G-protein-coupled receptors and by small GTPases, where R-Ras activates and H-Ras inhibits high-affinity binding. Integrin ligands bind to the  and  subunits with the I domain and I-like domain being particularly important. Adapted from Loftus and Liddington (1997).

Recently, some of the biochemical pathways underlying "inside-out signaling" have begun to come to light. As first elucidated in platelets with regard to the IIb3 integrin (Smyth et al., 1993), several agonists (usually acting through heterotrimeric G-protein-coupled receptors) can activate integrins in various cell types. It also seems clear that signaling pathways coupled to Ras-related guanosine 5'-triphosphatases (GTPases) impinge on integrin activation and function. For example, chemokines such as formyl peptide or interleukin (IL)-8 which enhance 1 and 2 integrin-mediated adhesion in lymphoid cells seem to act via the Rho GTPase (Laudanna et al., 1996); in this case, it is not entirely clear whether this represents is a change in integrin affinity (interaction of a single ligand with its receptor) or in avidity (concerted interaction of several receptors with a polyvalent ligand). Transfection of myeloid cells with a constitutively activated form of the R-Ras GTPase clearly causes an increase in the affinity of several integrins, including v3, 41, and 51, whereas coexpressed IIb3 can be activated by R-Ras in Chinese hamster ovary (CHO) cells (Zhang et al., 1996). By contrast, transfection of activated H-Ras, or of its downstream kinase Raf-1, inhibited the activation of coexpressed IIb3 (Hughes et al., 1997). Thus, different Ras-related small GTPases and their downstream effectors can either positively or negatively modulate the ligand binding affinity of integrins. Presumably integrin activation through inside-out signaling ultimately involves intracellular proteins that directly interact with integrin cytoplasmic domains; a discussion of proteins that interact directly with integrins is presented further on in this review (see Section III.).

In summary, integrins are a particularly complex family of cell adhesion receptors. These heterodimeric cell membrane glycoproteins can bind to a variety of ECM ligands or cellular counter-receptors. They engage with elements of the cytoskeleton and can thus influence cell shape, intracellular architecture, and cell motility. Integrins exist in either high- or low-affinity states with respect to their ligands, and they can engage in both "inside-out" and "outside-in" signal transduction via pathways that just now are being elucidated.

The cadherins comprise a family of transmembrane proteins that share an extracellular domain consisting of multiple repeats of a cadherin-specific motif (Suzuki, 1996) (see fig. 1). Members of the "classic" cadherin subfamily are calcium-dependent homotypic cell-cell adhesion molecules. This subfamily includes the N-, P-, R-, B-, and E-cadherins as well approximately 10 other members (Takeichi, 1995). These molecules localize in specialized sites of cell-to-cell adhesion that are termed adherence junctions; at these sites cadherins can establish linkages with the actin-containing cytoskeleton. The classic cadherins play a key role in developmental processes. For example, "knockout" of the gene for E-cadherin (which mediates epithelial interactions) results in an embryonic lethal that arrests before blastocyst formation (Takeichi, 1995). Another important subfamily of cadherins involved in adhesion is represented by the desmogleins and desmocollins, a group of desmosome-associated cadherins that form intracellular linkages to intermediate filaments rather than actin filaments (Cowin and Burke, 1996).

The structure of a typical classic cadherin consists of an amino-terminal external domain having five tandem repeats, a single transmembrane segment, and a cytoplasmic carboxy-terminal domain of approximately 150 amino acids (fig. 1). The binding functions of the cadherin are localized in the amino-terminal tandem repeat, whereas the other repeats are bridged by calcium binding sites that impart rigidity to the molecule (Aberle et al., 1996). A recent model based on the X-ray structure of N-cadherin (Shapiro et al., 1995) suggests that cell-cell adhesion mediated by cadherins involves a "zipper" type of organization. Cadherins on one cell surface form a series of rigid dimers that presents the first (N-terminal) cadherin repeat to equivalent dimers on the opposing cells (Aberle et al., 1996; Gumbiner, 1996). Lateral motion of these complexes allows the cell junction site to "zip up."

The cytoplasmic domains of cadherins interact strongly with a group of intracellular proteins known as catenins; these proteins are essential for cadherin function, because truncation of the cadherin cytoplasmic domain to delete catenin binding sites leads to a loss of cadherin-mediated adhesion (Gumbiner, 1996; Takeichi, 1995). Because there is considerable homology among their cytoplasmic domains, different classic cadherins can compete for the same pool of catenins (Kinter, 1992). The catenins were described initially as a set of three proteins, -, -, and -catenin (also termed plakoglobin). -Catenin binds directly to the cadherin cytoplasmic domain; subsequently, -catenin binds to -catenin and links the complex to the actin cytoskeleton by direct interaction with actin and by binding -actinin, an actin-bundling protein (Cowin and Burke, 1996). The structure of -catenin shows substantial homology to the protein vinculin, which binds -actinin and talin and is critical for cytoskeletal assembly at integrin-mediated focal adhesion sites (see Section III. below) (Aberle et al., 1996). In some cases plakoglobin replaces -catenin in mediating cadherin-cytoskeletal complexes.

The cadherins play a critical role in the development of tissue organization during ontogeny and in maintenance of normal tissue structures in adult organisms (Huber et al., 1996). For example, a variety of cadherins have been implicated in generating the complex architecture of the brain (Redies and Takeichi, 1996). In adult organisms, loss of cadherin expression or function can lead to serious pathophysiological consequences. Thus progression toward an invasive, malignant phenotype in epithelial tumors is associated with loss or mutation of E-cadherin, or with the disruption of cadherin-catenin complexes. Phosphorylation of -catenin by v-Src or other tyrosine kinases can lead to a breakdown of cadherin-cytoskeleton associations and loss of the adherence junctions that maintain normal epithelial organization (Behrens and Birchmeier, 1994; Birchmeier, 1995).

Recent evidence has indicated that cadherins and catenins also play an important role in signal transduction. As discussed in more detail below (Section VIII.), -catenin not only interacts with cadherins but also with components of the wingless/Wnt signaling pathway (Peifer, 1996). Thus the cell-cell adherence junction, maintained by cadherins and their associated catenins, is a key element in both tissue organization and in regulation of signaling cascades.

A diverse array of cell adhesion receptors are included in the immunoglobulin superfamily of cell adhesion molecules (Ig-CAMs). Proteins of this family are defined by the presence of one or more copies of the Ig fold, a compact structure with two cysteine residues separated by 55 to 75 amino acids arranged as two antiparallel  sheets (Vaughn and Bjorkman, 1996). In many (but not all) cases, CAMs in the Ig superfamily also contain one or more copies of a fibronectin type III repeat domain. Ig family adhesion receptors typically have a large amino-terminal extracellular domain, a single transmembrane helical segment, and a cytoplasmic tail (fig. 1). Members of the Ig-CAM family function in a wide variety of cell types and are involved in many different biological processes.

One of the most important contexts for Ig-CAMs is the developing nervous system, where many different members of this superfamily are involved in axon guidance and in the establishment and maintenance of neural connections (Baldwin et al., 1996; Tessier-Lavigne and Goodman, 1996). The "classic" example of an Ig superfamily adhesion receptor is NCAM, which contains five Ig folds in its extracellular portion; there are three forms of NCAM, two with transmembrane domains and one having a glycosylphosphatidylinositol (GPI) link to the membrane (Edelman and Crossin, 1991). NCAM functions as a homotypic calcium-independent adhesion receptor and seems to be almost universally present in the nervous system. Adhesive interactions mediated by NCAMs are known to be regulated by both the abundance of receptor and its degree of polysialyation (Edelman and Crossin, 1991; Tessier-Lavigne and Goodman, 1996). Several other homotypic neural cell adhesion molecules belong to the Ig superfamily, including L1, TAG1, contactin, and Drosophila fasiculin II (Baldwin et al., 1996; Tessier-Lavigne and Goodman, 1996). Another group of Ig-CAMs important in neural development are the netrin receptors. Netrins are ECM matrix proteins, distantly related to laminins, that provide guidance cues to migrating axons (Tessier-Lavigne and Goodman, 1996). Vertebrate netrin receptors belong to the Ig family and include DCC, a protein originally identified as the product of a key tumor suppressor gene for colon cancer (Keino-Masu et al., 1996). Yet another group of key adhesion receptors involved in neural development is the dozen or so members of the Eph subfamily of transmembrane tyrosine kinases; these proteins contain an Ig fold in their extracellular domains and are thus Ig-CAMs (Tessier-Lavigne and Goodman, 1996). The Eph kinases and their cell-bound ligands seem to be involved primarily in axon-target cell interactions. Thus neural Ig-CAMs can be involved in either homotypic (NCAM) or heterotypic (DCC, Eph kinases) adhesive interactions. There is relatively little known about the interactions of neural Ig-CAMs with cytoskeletal proteins. One report has suggested an association between the 180 kDa form of NCAM and spectrin (which can link to the actin cytoskeleton) (Pollerberg et al., 1997). However, it seems likely that other interactions exist and contribute to the functions of neural Ig-CAMs. Presumably many of the interactions between neural adhesion molecules result in signal generation, as well as in establishment of adhesive contacts between cells; however, with a few exceptions (see Section VIII.), the mechanistic details of the signaling pathways are poorly understood at present.

Another important biological context for Ig-CAMs lies in the immune system. In fact, integrins, selectins, and Ig-CAMs are all critically involved in multiple aspects of immune function (Dustin and Springer, 1991; Rosales and Juliano, 1995; Springer, 1995). T lymphocytes express several Ig superfamily receptors including CD2, CD4, or CD8, ICAMs 1 and 2, and the T-cell receptor (TCR) itself. These receptors play important roles in antigen recognition, cytotoxic T-cell functions, and lymphocyte recirculation. In contrast to many of the neural Ig-CAMs, which are often homotypic receptors, Ig family proteins involved in the immune system primarily engage in heterotypic interactions. For example, CD2 on T cells interacts with LFA-3 (another Ig-CAM) expressed on target cells, the TCR interacts with major histocompatibility complex (MHC) class II proteins on antigen presenting cells (both Ig superfamily), whereas ICAMs on endothelial cells are recognized by 2 integrins on leukocytes. Other Ig-CAM family adhesion receptors are found on vascular endothelial cells and play an important role in leukocyte trafficking to inflamed tissue sites. For example, vascular cell adhesion molecule-1 (VCAM-1) is an endothelial cell counter-receptor for the integrin 41 found on leukocytes. Platelet endothelial cell adhesion molecule-1 (PECAM-1) is an Ig-family cell-cell adhesion molecule that can engage in both homotypic and heterotypic interactions; one of its roles seems to be maintaining tight contacts between adjacent vascular endothelial cells (DeLisser et al., 1994).

Recently, X-ray crystal structure was obtained for the extracellular domains of several Ig-CAMs important in immune function including intercellular adhesion molecule (ICAM)-2, vascular cell adhesion molecule-1 and CD2 (Casanovas et al., 1997). This has allowed a more detailed understanding of how these molecules recognize their ligands and support cell-cell contact. As with the cadherins, the intracellular domains of Ig-CAMs may also be important in regulating the adhesive functions of these receptors (DeLisser et al., 1994; Doherty et al., 1992).

Immune system Ig-CAMs are critically involved in the key signal transduction processes leading to activation of T cells and B cells by antigens (Weiss and Littman, 1994). A detailed account of these complex events is beyond the scope of this review. In essence, however, Ig-CAM and integrin-mediated contacts are established between the antigen-presenting cell and the T cell such that the TCR recognizes antigen bound to an MHC protein on the presenting cell. This triggers the activation of intracellular tyrosine kinases associated with the TCR and with accessory receptors (Crabtree and Clipstone, 1994). In B cells, the B-cell receptor (also an Ig-CAM) can recognize either soluble or particulate antigen, and it also can activate intracellular tyrosine kinases upon ligation. Both the Src family and the spleen tyrosine kinase (SYK)/ZAP-70 family of tyrosine kinases have been implicated in antigen-induced T- and B-lymphocyte signaling (Weiss and Littman, 1994). Further on in this review, we will return to the theme of adhesion receptors activating intracellular tyrosine kinases in great detail as we consider adhesion receptor signaling in fibroblasts, epithelial cells, and tumor cells.

Another important group of signaling proteins that overlaps the Ig-CAM superfamily are the receptor protein tyrosine phosphatases (RPTPs) (Neel and Tonks, 1997). Many of these transmembrane enzymes have extracellular domains that include Ig folds (and often fibronectin type III repeats as well). The RPTPs typically have a large external domain, a single transmembrane helix, and a cytoplasmic domain containing two signature tyrosine phosphatase domains flanked by a variety of noncatalytic sequences. RPTPs seem to function conversely to receptor tyrosine kinases (RTKs); that is, ligand binding results in dimerization of RPTPs but inhibits enzyme activity (rather than activating it as with RTKs). It has become clear recently that several RPTPs can engage in homotypic or heterotypic cell adhesion through their extracellular domains. For example, Drosophila LAR, a Drosophila Ig-CAM-RPTP, plays an important role in axon migration and innervation of muscle targets (Neel and Tonks, 1997).

Thus Ig-CAMs play multiple roles in the developing embryo and in the adult organism. In addition to mediating adhesive contacts that are important in tissue organization, or in cellular trafficking in the immune system, many Ig-CAMs function in key signal transduction processes as well.

The selectins are a small family of lectin-like adhesion receptors composed of three members, L-, E-, and P-selectin (Lasky, 1995; Rosen and Bertozzi, 1994; Tedder et al., 1995). The structure of a selectin includes an amino-terminal domain that is homologous to calcium-dependent animal lectins, followed by an epidermal growth factor (EGF)-type domain, two to nine complement regulatory protein repeats, a transmembrane helical segment, and a short cytoplasmic tail (fig. 1). Selectins mediate heterotypic cell-cell interactions through calcium-dependent recognition of sialyated glycans. The best defined physiological role for selectins concerns leukocyte adherence to endothelial cells and platelets during inflammatory processes (Rosales and Juliano, 1995; Springer, 1995). The expression and function of selectins is tightly regulated so as to come into play only when leukocytes need to stick to the vessel wall as part of normal immune system cellular trafficking or during inflammation. Thus P-selectin is present in latent form in endothelial cells and platelets and it is rapidly translocated from secretory granules to the cell surface upon cell activation by thrombin or other agonists. E-selectin is synthesized and expressed on endothelial cells in response to inflammatory cytokines such as tumor necrosis factor (TNF) or IL-1. L-selectin is expressed constitutively on leukocytes, but its presentation at the cell surface may be regulated.

The precise identities of the ligands for the three currently known selectins are being pursued actively and are a matter of some controversy (Varki, 1997). For all three selectins, a key component of the binding ligand is tetrasaccharide residues of the sialyl-Lewis^X type (previously defined as a blood group antigen); such motifs appear on glycolipids as well as on glycoproteins. However, the binding affinities of selectins for isolated sialyl-Lewis^X saccharides are very poor, although selectins clearly are responsible for high-affinity cell-cell binding. Thus physiological high-affinity ligands for selectins likely must include sialyl-Lewis^X saccharides in the context of a macromolecular scaffold. The best documented high-affinity counter-receptor for a selectin is P-selectin glycoprotein ligand-1 (PSGL-1), an O-glycosylated mucin-like transmembrane glycoprotein found on leukocytes and lymphoid cells (Norman et al., 1995). Glycoproteins of this type likely have a rigid, rod-like structure and may be able to present multiple copies of sialyl-Lewis^X to endothelial cell selectins. Two potential counter-receptors for L-selectin have been identified; GlyCAM-1 which is a small, secreted mucin-like glycoprotein, and CD34 a transmembrane mucin-like protein found on endothelial cells. A putative counter-receptor for E-selectin has been termed ESL-1 and is a glycoprotein bearing N-linked carbohydrate residues. However, the physiological importance of GlyCAM-1, CD34, and ESL-1 in leukocyte to endothelial adherence and leukocyte trafficking is still unclear (Varki, 1997).

There is much less known about selectin-mediated signal transduction than is the case for integrins, cadherins, and Ig-CAMs. However, it is rapidly becoming clear that selectins play an important role in signaling processes that regulate leukocyte-endothelial cell interactions (Zimmerman et al., 1996b).

	III. The Components of Adhesive Junctions

Top Previous Next References

Signal transduction by cell adhesion receptors primarily occurs in the context of highly organized supramolecular complexes that are assembled at sites of cell-to-cell or cell-to-ECM adhesion. There are several distinct types of adhesive junctions. However, our present understanding of the role of adhesive junctions in signaling extends primarily to one type of junction, the integrin-containing focal contact (also termed focal adhesion) and thus we will discuss these structures in some detail. Another growing body of information concerns cadherin-containing adherence junctions, whereas less is known about the signaling properties of other types of adhesive junctions.

Focal contacts are specialized sites where cells attach to the ECM (Burridge and Chrzanowska-Wodnicka, 1996; Jockush et al., 1995). At focal adhesions, clusters of integrins bind externally to ECM proteins and internally to several specialized cytoplasmic proteins that in turn bind to actin filaments. It is becoming clear that focal adhesions are dynamic structures that change in size (and likely in composition) as the cell adhesion process progresses. Initially, rather small punctate structures are formed at sites of cell-to-substratum contact. As the focal contact matures, the actin filaments extend and bundle to form prominent structures termed stress fibers. Integrins can interact with numerous proteins at focal contacts and at other sites in the cell. In addition, there is a complex network of interactions among the specialized cytoplasmic focal contact proteins. In this section of the review we will introduce some of the key components of focal contacts. We also will discuss transmembrane proteins that directly interact with integrins. Finally, we will introduce several newly discovered cytoplasmic proteins that bind directly to integrin  or  subunit cytoplasmic tails and which may play important roles in integrin signaling. The known interactions of most of the proteins to be discussed below are depicted in fig. 3.

View larger version (21K): [in this window] [in a new window]   Fig. 3.   Proteins associated with integrin-containing focal contacts. (A) Focal contact proteins and their interactions. This diagram illustrates the patterns of binding interactions of key structural proteins of the focal contact including talin (Tal), vinculin (Vinc), -actinin (-act), tensin (Tens), and paxillin (Pax). See text for additional details. (B) Proteins interacting directly with integrins. This diagram recapitulates current information concerning proteins that bind directly to integrins. In many cases it is clear that the interaction occurs with a particular  or  subunit or with an individual integrin heterodimer, whereas in other cases the interaction may involve binding to many different integrins. Clearly only some of these interactions will take place in any given type of cell, with the pattern of interaction determined by the expression of the various binding proteins, of various integrins, and by the physiological status of the cell.

1. Cytoskeletal proteins The cytoplasmic structural proteins of the focal contact that directly bind to integrins include talin and -actinin, which in turn bind to other structural proteins including vinculin, paxillin, and tensin, ultimately leading to the recruitment of actin filaments. Integrin-mediated phosphorylation of several cytoskeletal proteins seems to have a role in the organization of stress fibers at focal adhesions. It has been shown that several kinases, including FAK, play a role in the organization of cytoskeleton (see Section IV. below). The structure and function of several of the key structural proteins of the focal contact is described below (fig. 3A).

b. -ACTININ. -Actinin is an actin-bundling protein that comprises two identical polypeptides, each of which is approximately 104 kDa in size. The protein can be divided into three regions: the N-terminal actin binding site, a central region with four -helical motifs, and a carboxy-terminal region (Jockush et al., 1995

d. VINCULIN. Vinculin is a 116 kDa polypeptide enriched in focal contacts. Vinculin binds to actin, -actinin, talin, and paxillin (reviewed in Jockush et al., 1995

2. Transmembrane proteins: tetraspanners, integrin-associated protein 50, caveolin As discussed above, integrins interact via their cytoplasmic tails with an assemblage of cytoplasmic proteins. However, the external and transmembrane domains of integrins also can participate in protein-protein interactions; in some cases these interactions may occur at sites away from focal contacts (fig. 3B).

3. Cytoplasmic integrin-binding proteins The cytoplasmic domains of integrins have been implicated in the bi-directional transfer of information across the plasma membrane. Integrin cytoplasmic tails interact with several proteins that may participate in integrin-mediated functions including signal transduction. Novel proteins have been identified that interact with either the  or  cytoplasmic domains of integrins (fig. 3B).

a. SUBUNIT-ASSOCIATED PROTEINS.

i. 3 endonexin. A novel 111 amino acid protein, 3 endonexin, was cloned by using the integrin 3 cytoplasmic tail as bait in the yeast two-hybrid system (Shattil et al., 1995). This interaction is very specific to 3, because a mutation of serine to proline at position 752 in the 3 cytoplasmic tail reduced the binding by 64%; this mutation is critical for IIb3 integrin-mediated function (Chen et al., 1994b). Because this 3 mutation abolishes binding to fibrinogen (Chen et al., 1992b) and cell spreading (Chen et al., 1994b), it suggests that 3 endonexin is involved in the selective modulation of 3 integrin function. In addition, it has been proposed that 3 endonexin has a function independent of integrins because endonexin is expressed in a wide variety of tissues. An NITY motif at positions 756 to 759 of 3 is critical for the interaction with endonexin (Eigenthaler et al., 1997); this motif is highly conserved in 3 integrins of different species. Another study from the same group (Kashiwagi et al., 1997) demonstrated that 3 endonexin bound to IIb3 in mammalian cells and was able to modulate the affinity state of IIb3, leading to fibrinogen-dependent aggregation. Only a weak staining of 3 endonexin was observed in focal adhesions, suggesting that 3 endonexin interacts with integrins transiently, so that other interactions with the integrin tails can follow after endonexin dissociation (Kashiwagi et al., 1997). A role of 3 endonexin in outside-in signaling events has yet to be demonstrated.

ii. Integrin-linked kinase. A yeast two-hybrid approach also was used to identify another  subunit-binding protein termed integrin-linked kinase (ILK) (Hannigan et al., 1996). ILK is a 59 kDa serine-threonine kinase that can associate with 1, 2, and 3 cytoplasmic domains and has been shown to phosphorylate a 1 integrin cytoplasmic domain peptide in vitro (Dedhar and Hannigan, 1996). Cell attachment to fibronectin reduced ILK kinase activity, whereas overexpression of ILK induced anchorage-independent growth and reduced cell adhesion to laminin, fibronectin, and vitronectin (Hannigan et al., 1996). ILK and 1 integrins colocalize in focal adhesions, suggesting a possible role in cytoskeletal assembly. Recently, it has been shown that ILK overexpression induces anchorage-independent cell cycle progression (Radeva et al., 1997). As discussed in detail in Section X. below, upon integrin-mediated cell adhesion, levels of cyclins D1 and A are up-regulated and cyclin-dependent kinase (CDK) inhibitor levels are down-regulated. Normal cells placed in suspension have down-regulated expression of cyclin D1 and cyclin A and hypophosphorylated retinoblastoma (Rb), whereas suspension cells overexpressing ILK have higher expression of cyclin D1 and cyclin A and hyperphosphorylated Rb (Radeva et al., 1997). Thus ILK may be a key link between integrins and downstream signaling events that influence the cell cycle. In addition, very recent evidence suggests that ILK is involved in epithelial cell differentiation and gene expression (see Section XI.).

iii. Cytohesin 1. Using the 2 cytoplasmic domain as bait in the yeast two-hybrid system, Kolanus et al. (1996) identified a protein they termed cytohesin-1. In lymphoid cells cytohesin-1 binds to 2; further, there seems to be a significant role for cytohesin-1 in the regulation of integrin activation, as measured by L2 adhesion to ICAM1. Cytohesin-1 has a plekstrin homology (PH) domain and a domain that resembles the yeast SEC7 protein. The SEC7 domain alone induces L2 binding to ICAM1, whereas the PH domain inhibits the process, seemingly acting in a dominant negative role. It has been postulated that PH domains function like SH2 and SH3 domains, mediating protein-protein associations in signal transduction pathways and/or localizing proteins to the plasma membrane (Pawson, 1995). Cytohesin-1 recently has been shown to function as a guanine nucleotide exchange factor (GNEF) probably via its SEC7 domain (Meacci et al., 1997). Moreover, cytohesin-1 also binds via its PH domain to phosphoinositol 3,4,5-triphosphate, suggesting a link with PI-3K pathways (Klarlund et al., 1997). Thus cytohesin-1 may play several important roles in integrin activation and in intracellular signaling.

iv. Integrin cytoplasmic domain-associated protein 1. Recently, another novel 1 integrin cytoplasmic domain-associated protein (ICAP1) was isolated using a yeast two-hybrid approach. Two isoforms of ICAP1 that are derived from alternatively spliced mRNA have been reported; the longer isoform has 200 amino acids (ICAP1) and the shorter isoform has 150 amino acids (ICAP1). The interaction of ICAP1 with 1 integrins has been shown in vitro and in vivo. ICAP is a phosphoprotein, and its phosphorylation is regulated during cell-matrix interactions. Cell attachment to fibronectin resulted in enhancement of ICAP1 phosphorylation, whereas expression of Rho A that disrupts cell-matrix interactions resulted in dephosphorylation of ICAP1. These data indicate that ICAP1 has a role in integrin-mediated cell adhesion and spreading (Chang et al., 1997).

b. SUBUNIT INTERACTING PROTEINS.

i. Calreticulin. A conserved stretch of seven amino acids (KXGFFKR) exists in all integrins and subunits. To understand the role of such a highly conserved region, investigators tried to identify proteins that interact with this region. Rojiani et al. (1991) reported a 60 kDa protein that specifically binds to the KXGFFKR sequence. This protein had significant homologies to rabbit calreticulin (Fliegel et al., 1989) and the Ro/SS-A antigen, its human equivalent (McCauliffe et al., 1990). An in vivo interaction of calreticulin with the active form of the 21 integrin has been demonstrated (Coppolino et al., 1995). Further, inhibition of expression of calreticulin by antisense oligonucleotides resulted in inhibition of cell adhesion and less cell spreading, indicating a role for calreticulin in cell-ECM interactions (Leung-Hagesteijn et al., 1994). Recently, a double knockout of calreticulin in embryonic stem cells suggested that calreticulin is essential for the function of integrins including "outside-in" signaling (Coppolino et al., 1997).

ii. Calcium- and integrin-binding protein. A novel calcium- and integrin-binding protein (CIB) of 25 kDa that interacts with IIb was identified using the yeast two-hybrid system. This protein specifically binds to IIb and has similarities to the calcium-binding proteins calmodulin and calcineurin. CIB interacts with IIb3 in vitro and is expressed in platelets, which suggests that it may function as a regulatory molecule for IIb3 (Naik et al., 1997).

iii. Integrin-binding protein 1. Using a yeast two-hybrid approach, we identified a novel protein that interacts with the 5 subunit cytoplasmic tail (Alahari and Juliano, unpublished results). This protein does not have significant homologies to known proteins, and henceforth was tentatively named integrin-binding protein 1 (IBP1). IBP1 binds to human 51 in vitro, but its physiological role is currently unknown.

a. INSULIN RECEPTOR SUBSTRATE 1 AND 190 KDA. Two cytoplasmic proteins have been shown to associate with v3 integrin in response to growth factor stimulation. A 190 kDa tyrosine-phosphorylated protein associates with v3 in PDGF-stimulated fibroblasts (Bartfeld et al., 1993

b. PROTEASES. Some recent observations have indicated the existence of interactions between integrins and proteases or protease receptors that may be very important in the regulation of inflammation and of tumor cell invasion and metastasis. Thus the v3 integrin has been shown to associate with both the matrix metallo protease (MMP2) (Brooks et al., 1996

In addition to integrin-containing focal adhesions, there are several other types of structures that link cells to each other or to the extracellular matrix. These include: cadherin-based cell-cell adherence junctions, and cell adhesion junctions containing certain proteoglycans and ezrin/radixin/moesin (ERM) proteins, both of which associate with the actin cytoskeleton; desmosomes and hemidemosomes that interdigitate with intermediate filaments; tight junctions that serve a barrier function in epithelial sheets; gap junctions that play a role in cell-cell ionic communication; transient cell-cell junctions that occur in immune recognition; and specialized junctions such as nerve synapses (Staehlin and Hull, 1978). In most cases little is known about the role of junctional proteins in signal transduction. Thus we will mention only briefly two types of actin-based junctions where some understanding of signaling processes is beginning to emerge.

1. Cadherin-mediated adherence junctions As discussed in Section II. above, the classic cadherins bind to a group of proteins termed catenins which in turn can link to the actin containing cytoskeleton forming a structure termed an adherence junction. Structures of this type are particularly important to the biology of epithelial cells, where "belts" of adherence junctions in neighboring cells are joined by dense bundles of actin filaments thus stabilizing the epithelial sheet. A more detailed picture of the structure of cadherin-based adherence junctions can be found elsewhere (Cowin and Burke, 1996; Geiger et al., 1995); whereas the role of cadherin-based junctions in signaling will be described in detail in Section VIII.

2. Junctions containing ezrin/radixin/moesin proteins The ERM family of linker proteins seems to play a vital role in some aspects of cytoskeletal assembly (Tsukita et al., 1997). These proteins share a highly homologous COOH-terminal actin binding domain and an equally well conserved N-terminal membrane binding domain. The membrane molecules that are the binding partners of ERM proteins have not been fully worked out; the polymorphic transmembrane glycoprotein/proteoglycan CD44 is clearly one of them, but the Na/K ATPase and the Ig-superfamily member ICAM-1 may also interact with ERM proteins. The ability of ERM proteins to link membrane molecules to actin seems to be closely regulated by the small GTPase Rho. However, evidence has been developed recently that moesin may in turn play a role in regulating the activity of Rho family GTPases (Takahashi et al., 1997). In this manner, ERM proteins might affect the assembly of a variety of cytoskeletal structures thought to be regulated by Rho family proteins (see Section VI. for more detail). Thus, the precise biological role of ERM proteins has yet to be defined, but it seems likely that assembly/disassembly of adhesive junctions will be one aspect of their function.

	IV. Direct Signal Transduction by Integrins: Activation of Tyrosine Kinases

Top Previous Next References

The concept that integrins function not merely as receptors for cell adhesion, but also elicit signal transduction events, emerged from several types of observations. For example, when 31 integrins were clustered on the surface of human carcinoma cells, the tyrosine phosphorylation level of a 115 to 130 kDa subset of proteins was enhanced (Kornberg et al., 1991). Essentially the same subset of proteins was tyrosine phosphorylated in NIH3T3 and rat embryo fibroblasts (REFs) spreading on fibronectin-coated surfaces (Burridge et al., 1992; Guan et al., 1991). These effects were deemed to be integrin-specific through the use of antibodies to non-integrin cell surface molecules or by means of nonspecific attachment to polylysine-coated surfaces (Burridge et al., 1992; Guan et al., 1991; Kornberg et al., 1991). Simultaneously, a similar pattern of tyrosine phosphorylation was observed in v-Src-transformed chicken embryo fibroblasts (CEFs) (Kanner et al., 1990) and upon treatment of quiescent Swiss 3T3 fibroblasts with the G-protein-coupled receptor agonists bombesin, vasopressin, and endothelin (Zachary et al., 1991). These data helped to consolidate the idea that integrins have the ability to convey signals directly from the ECM to the inside of the cell. In this section and the next, we will review the identified key components, mechanistic models, and significance of these direct integrin-mediated effects on signaling, concentrating here on tyrosine phosphorylation events.

B. Focal Adhesion Kinase-Mediated Events

1. Linkage to integrin signaling Of the tyrosine phosphorylation events described above, the most prominent effect is observed in proteins in the molecular weight range of 110 to 130 kDa. It is now clear that the cytoplasmic tyrosine kinase, FAK accounts for a large proportion of the tyrosine phosphorylation in this region. Furthermore, tyrosine phosphorylation of FAK is regarded not only as an important event in integrin-mediated signaling, but as a common theme in multiple signal transduction pathways (see table 1) (Zachary et al., 1992). FAK was cloned independently by the use of monoclonal antibodies raised to tyrosine-phosphorylated proteins in v-Src transformed CEFs (Schaller et al., 1992) and by using a homology-based strategy to identify novel protein tyrosine kinases in a mouse cDNA library (Hanks et al., 1992). FAK is a unique protein tyrosine kinase of approximately 125 kDa; it contains a central consensus kinase domain, a C-terminal domain having two proline-rich sequence, and a region required for focal adhesion targeting termed the "FAT" sequence. FAK has no transmembrane, SH2, or SH3 domains (fig. 4) (Hanks et al., 1992; Hildebrand et al., 1993; Schaller et al., 1992). FAK now also has been cloned from humans (Andre and Becker, 1993) and Xenopus (Zhang et al., 1995b) and exhibits a high degree of conservation between species.

View larger version (21K): [in this window] [in a new window]   Fig. 4.   The structural domains of integrin-responsive tyrosine kinases. The domain structure of FAK, c-Src, c-Abl, and Syk are shown. The position of the kinase domain and Src homology 2 (SH2) and 3 (SH3) domains, which mediate protein-protein interactions, are indicated. In FAK, the C-terminal proline-rich (PR) and focal adhesion targeting (FAT) domains are shown. Similarly, in c-Abl the position of the DNA and F-actin binding domains are illustrated.

2. Mechanism of activation In growing fibroblasts, FAK is phosphorylated on both tyrosine and serine residues (Calalb et al., 1995; Schlaepfer and Hunter, 1996). All phosphotyrosine residues are dephosphorylated upon cell detachment, whereas phosphorylation at serine residues is largely unaffected (Calalb et al., 1995). Presumably a protein tyrosine phosphatase is activated upon cell detachment, but the identity of this enzyme(s) remains elusive. Although it is worth remembering that serine phosphorylation can control protein function and protein-protein interactions (Muslin et al., 1996), relatively little is known about the sites and function of FAK serine phosphorylation, and therefore this review will focus on the importance of tyrosine-phosphorylated residues. As observed for many receptor tyrosine kinases, it is the autophosphorylation of FAK that is the trigger for the regulation its downstream functional activities. FAK autophosphorylation within the sequence Y³⁹⁷AEI occurs in vitro in a bacterial-fusion protein and in immune complex kinase reactions of FAK expressed in fibroblasts (Calalb et al., 1995; Schaller et al., 1994). In some cell types, autophosphorylation of FAK in trans may also occur, as exogenously expressed kinase-dead FAK can be tyrosine phosphorylated (Cary et al., 1996; Eide et al., 1995; Schlaepfer and Hunter, 1996). FAK additionally is phosphorylated at Tyr407, Tyr576, Tyr577, Tyr861, and Tyr925 in growing fibroblasts (Calalb et al., 1995, 1996; Schlaepfer et al., 1994; Schlaepfer and Hunter, 1996). Phosphorylation at Tyr576 and Tyr577 in the catalytic domain of FAK is important for maximal kinase activity (Calalb et al., 1995). Tyr925 is within a consensus site, YENV, for binding the SH2 domain of the adapter protein, growth factor receptor-binding protein-2 (GRB2) (Schlaepfer et al., 1994; Songyang et al., 1993). The roles of P-Tyr407 and P-Tyr861 as yet remain elusive.

3. The role of Src in focal adhesion kinase signaling Evidence of Src playing a pivotal role in FAK signaling events is manifest by augmented tyrosine phosphorylation of FAK in v-Src-transformed fibroblasts (Calalb et al., 1995; Guan and Shalloway, 1992). The FAK autophosphorylation motif (YAEI) resembles the YEEI consensus site for Src-SH2 binding (Songyang et al., 1993) and indeed Src binds FAK at this site after integrin-mediated FAK autophosphorylation in fibroblasts and in v-Src-transformed 3T3 fibroblasts (Schaller et al., 1994; Schlaepfer et al., 1994). This association is transient in adhering fibroblasts and is absent upon serum-starvation despite unaffected phosphorylation of FAK at Tyr397. However, in v-Src-transformed cells the association is stable and independent of serum and adhesion (Cobb et al., 1994a; Schaller et al., 1994; Schlaepfer et al., 1994; Schlaepfer and Hunter, 1996). Thus, additional factors regulate the ability of Src to bind FAK, even when FAK is in a "primed" state. The FAK residues Tyr407, Tyr576, Tyr577, Tyr861, and Tyr925 all show enhanced phosphorylation upon v-Src transformation and can be phosphorylated directly by Src in vitro (Calalb et al., 1995, 1996; Schlaepfer and Hunter, 1996). Src interaction with Tyr397 and subsequent phosphorylation of FAK results in GRB2 binding to Tyr925; GRB2 binding to FAK is a putative link between FAK tyrosine phosphorylation and the integrin-mediated activation of mitogen-activated protein kinase (MAPK) (see Section V.) (Schlaepfer and Hunter, 1996, 1997). FAK-Src interaction also enhances the association of the adapter protein, p130^CAS with FAK signaling complexes in cells adhering to fibronectin (Schlaepfer and Hunter, 1997).

View larger version (17K): [in this window] [in a new window]   Fig. 5.   Mechanism of integrin-mediated activation of FAK and signaling complex formation. Cytoplasmic FAK (1) is associated constitutively with paxillin and talin. Upon integrin engagement (2), FAK localizes to focal adhesions, possibly through an interaction with the integrin  subunit, resulting in autophosphorylation at Tyr397. This creates a binding site for the SH2 domain of c-Src, which on binding phosphorylates at the sites Tyr407, Tyr576, Tyr577, Tyr861, and Tyr925. Src phosphorylation of FAK enhances binding of p130CAS and GRB2, the latter through binding to Tyr925 (3). Thus, a multimeric signaling complex may be formed (4), which can control downstream events and which contains a negative feedback mechanism through the binding of GRAF. The interaction of PI-3K with FAK has not been illustrated here.

4. Other focal adhesion kinase-binding proteins Formation of complexes through SH2 and SH3 domain-mediated protein-protein interactions is often essential for the instigation of a signaling cascade. FAK interacts with a variety of SH2/SH3 domain containing proteins within the focal adhesion. Some of these interactions are constitutive, whereas some depend on the tyrosine phosphorylation state of FAK. Src-family kinases are not the only possible occupants of the Tyr397 binding site. The p85 subunit of PI-3K interacts with Tyr397 upon cell adhesion or PDGF treatment of NIH3T3 cells, despite the fact that YAEI is not an optimal binding site for the p85 SH2 domain (Chen et al., 1994a, 1996a; Songyang et al., 1993). In thrombin-activated platelets, p85 binds FAK, but interactions are directed by the SH3 domain of p85 and the second proline-rich region of FAK (Guinebault et al., 1995). Addition of the relevant FAK-derived peptides to p85 immunoprecipitates activates PI-3K activity in each case (Chen et al., 1996a; Guinebault et al., 1995).

5. Inhibition of focal adhesion kinase activity The adhesion-induced phosphorylation of FAK can be inhibited by agents that disrupt the cytoskeleton or delay cell spreading. Tyrosine phosphorylation of FAK can be potently blocked by treatment with either the actin depolymerizing agent, cytochalasin, or the tyrosine kinase inhibitor, herbimycin (Burridge et al., 1992; Sinnett-Smith et al., 1993). Extracellular stimulation by insulin causes transient dephosphorylation of FAK and paxillin in rat 1 fibroblasts and in CHO cell lines overexpressing the insulin receptor (Knight et al., 1995; Konstantpolulos and Clark, 1996; Ouwens et al., 1996; Pillay et al., 1995). This may be caused by insulin-mediated effects on the cytoskeleton or through phosphotyrosine phosphatase (PTP) 1D activity (Knight et al., 1995; Ouwens et al., 1996). FAK is dephosphorylated rapidly upon cell detachment (Calalb et al., 1995), indicating a PTP is activated under these conditions. Recently the leukocyte common antigen related (LAR) PTP was found to localize partially to focal adhesions in human breast adenocarcinoma cells, although its specificity for proteins, such as FAK and paxillin, and therefore its role in focal adhesion disassembly, are presently unclear (Serra-Pages et al., 1995). Additionally, a proteolytic mechanism may be used in activated platelets. FAK, activated on IIb3 engagement and platelet aggregation, can be proteolyzed by the protease calpain to produce fragments that exhibit reduced autokinase activity and dissociate from the cytoskeleton, an action that may be involved in cytoskeletal changes associated with blood clot retraction (Cooray et al., 1996).

6. Focal adhesion kinase substrates Because of the complex associations between kinases and structural proteins, unveiling biologically relevant in vivo substrates is an arduous task. For example, paxillin, tensin, and p130^CAS tyrosine phosphorylation closely correlate with that of FAK upon cell adhesion. Additionally, FAK phosphorylates recombinant paxillin in immunoprecipitation kinase assays, whereas FRNK overexpression reduces the tyrosine phosphorylation of paxillin and tensin (Bellis et al., 1995; Burridge et al., 1992; Harte et al., 1996; Richardson and Parsons, 1996); all this suggests that these proteins may be FAK substrates. On the other hand, Src is capable of interacting with paxillin (fig. 3A) (Salgia et al., 1995; Weng et al., 1993)_, and integrin-stimulated phosphorylation of p130^CAS likely is directed by c-Src (Bockholt and Burridge, 1995; Hamasaki et al., 1996; Vuori et al., 1996); further, phosphorylation of paxillin and tensin is not altered significantly in FAK / fibroblasts (Ilic et al., 1995). Thus the precise cause of tyrosine phosphorylation events in focal contacts remains obscure.

7. Focal adhesion kinase function Analysis of fibroblast lines derived from FAK / mice downplays the importance of FAK in the assembly of focal adhesions during cell adhesion. FAK-deficient fibroblasts retain the ability to spread on a fibronectin substratum and have normal levels of tyrosine phosphorylation on paxillin and tensin (Ilic et al., 1995). The cells, however, are morphologically dissimilar from wild-type controls in that they are more rounded, exhibit abundant ventral focal adhesions, and have numerous peripheral stress fibers and microspikes. Further, these cells show decreased mobility on fibronectin (Ilic et al., 1995), which indicates a possible defect in filopodia formation/extension and focal adhesion turnover. Relevant to this concept is the knowledge that FAK tyrosine phosphorylation is increased during neurite outgrowth of differentiating neuroblastoma cells, a process that involves growth cone filopodia extension (Leventhal and Feldman, 1996).

Recently, the identification of proline-rich tyrosine kinase 2 (PYK2), also known as cell adhesion kinase  (CAK), related adhesion focal tyrosine kinase (RAFTK), and calcium-dependent tyrosine kinase (CADTK), has established the existence of FAK family members (Avraham et al., 1995; Lev et al., 1995; Sasaki et al., 1995; Yu et al., 1996). PYK2 shows 60% identity to FAK in the kinase domain, with approximately 40% identity in the N- and C-terminal regions (Avraham et al., 1995; Sasaki et al., 1995). An autophosphorylation site (Tyr402) and the ability to associate with c-Src, p130^CAS, GRB2, and paxillin are retained in PYK2, suggesting that PYK2 and FAK may act through similar downstream mechanisms (Astier et al., 1997; Dikic et al., 1996; Lev et al., 1995; Li and Earp, 1997). However, PYK2 shows a more limited tissue distribution than FAK (Avraham et al., 1995; Sasaki et al., 1995) and is regulated differently, becoming tyrosine phosphorylated in response to agents that increase intracellular calcium under conditions that do not influence FAK (Lev et al., 1995; Siciliano et al., 1996; Tokiwa et al., 1996; Yu et al., 1996). Further, recent studies have suggested that FAK and PYK2 have opposing functions in terms of cell survival (Xiong and Parsons, 1997). Thus the biology of FAK and PYK2 may be quite distinct even though the two proteins are structurally closely related. In certain cell types integrin signaling may impinge on the PYK2 activation. For example, PYK2 is tyrosine phosphorylated upon integrin-mediated adhesion of human B cells and megakaryocytic cells (Astier et al., 1997; Li et al., 1996), and when overexpressed in CEFs, is localized to focal adhesions in a fraction of the cells (Schaller and Terukatsu, 1997). However, PYK2 seems to show no response to ECM-mediated signals in other cell types (Sasaki et al., 1995). Downstream signaling from PYK2 can lead to the activation of both the Jun N-terminal kinase (JNK) and ERK MAPK pathways (Dikic et al., 1996; Tokiwa et al., 1996). It is presently unclear whether PYK2 is capable of partially compensating for FAK in integrin signaling pathways. In particular, this may be relevant when studying adhesion-mediated signaling in fibroblasts derived from FAK / mice.

D. Src Family Members

1. Activation of Src family members downstream of 1 integrins Because of their ability to bind FAK and augment FAK signaling, the role of Src family kinases in integrin-mediated signaling events warrants further discussion. c-Src and its other family members (e.g., Yes, Fyn, Fgr, and Hck) share common structural features; an N-terminal myristylation and membrane localization signal, one SH2 and one SH3 domain, the catalytic domain, and a short C-terminal region containing a conserved critical tyrosine residue (see fig. 4), as described in a recent comprehensive review (Brown and Cooper, 1996). c-Src can be activated by many extracellular stimuli, including the ECM through integrins, growth factors acting via receptor tyrosine kinases (e.g., PDGF and EGF), G-protein coupled receptor agonists (e.g., thrombin and LPA), signals involved in T- and B-cell activation, and ultraviolet irradiation (Erpel and Courtneidge, 1995; Parsons and Parsons, 1997). c-Src acts as a molecular switch whereby activation, either by dephosphorylation of Tyr527 in the C-terminus or possibly through competition for the SH2 domain, results in a conformational change that breaks an intramolecular interaction between the SH2 domain and P-Tyr527 and in turn exposes the kinase domain (Matsuda et al., 1990). Src autophosphorylation at Tyr416 within the kinase domain is necessary for full activity but is overridden by phosphorylation at Tyr527 (Stover et al., 1994). The SH3 domain is also important in maintaining the inactive closed conformation of Src. This model recently was supported with structural evidence from a recombinant inactive fragment of human c-Src showing interaction of the SH2 domain with the phosphorylated Tyr527 residue and association of the SH3 domain with the linker region between the SH2 and kinase domains (Xu et al., 1997).

2. Activation of Src family members downstream of 2 integrins The studies mentioned thus far in this section mainly concern signaling through 1 integrins. Interestingly, signaling through other  subunits shows discrete differences in the subset of proteins tyrosine phosphorylated compared with those proteins phosphorylated upon 1 engagement (Berton et al., 1994; Graham et al., 1994). Differential expression of alternative Src kinases may account for at least some of these observations. A pertinent example here is polymorphonuclear leukocytes (PMNs), in which 2-integrins are the dominant functional adhesion receptors important for attachment and migration into sites of inflammation, and in which the major expressed Src family kinases are Fgr, Hck, and Lyn (Tsygankov and Bolen, 1993). In PMNs, TNF activation of Fgr depends on the engagement of 2 integrins (Berton et al., 1994). Enhancement of Fgr activity may result in the tyrosine phosphorylation of other proteins, such as paxillin and the guanine nucleotide exchange factor, Vav, which also displays enhanced tyrosine phosphorylation levels in a 2-dependent manner in PMNs (Graham et al., 1994; Zheng et al., 1996a). These responses are specific to 2 because PMNs derived from patients with leukocyte adhesion deficiency, which lack 2 expression, do not activate Fgr or phosphorylate paxillin in response to TNF (Berton et al., 1994; Graham et al., 1994). Other studies implicate a role for Fgr and Hck in the spreading and respiratory bursts of PMNs adhering to ECM components (Lowell et al., 1996). Interestingly, 2 antibody-mediated clustering does not activate FAK, despite FAK expression in neutrophils (Fuortes et al., 1994). Thus, several of the Src family kinases apparently perform physiological functions in integrin adhesion-dependent events in numerous cell types, and some degree of subunit specificity exists in the connections between integrins and cytoplasmic tyrosine kinases.

E. Other Tyrosine Kinases Activated by Integrins

1. C-Abl A more recently identified noncytoplasmic tyrosine kinase that is regulated by integrins is encoded by the c-Abl protooncogene. The kinase domain of c-Abl is flanked by N-terminal SH2 and SH3 domains, whereas the F-actin and DNA binding domains are at the extreme C-terminus of the protein (fig. 4) (Wang, 1993). Both the cytoplasmic and nuclear pools of c-Abl become activated upon adhesion of fibroblasts to fibronectin and a pool of the enzyme transiently localizes to focal adhesions (Lewis et al., 1996). It is unclear at present where c-Abl activation is sequentially on the integrin signaling pathway. A constitutively active form of c-Abl, the Bcr-Abl found in human patients with chronic myelogenous leukemia, causes enhanced phosphorylation of paxillin in myeloid cells (Salgia et al., 1995). Thus, c-Abl may participate in integrin-mediated events by phosphorylating focal adhesion proteins. Bcr-Abl also induces anchorage-independent, but growth factor-dependent, proliferation in fibroblasts, placing Abl downstream of integrins but upstream of the point of convergence with growth factor signaling pathways (Renshaw et al., 1995). In addition, c-Abl, through interactions with the C-terminal F-actin and DNA binding domains, may function in integrin-dependent regulation of the structure of actin filaments and in cell cycle progression.

2. Spleen tyrosine kinase In hematopoietic cell types such as monocytes, neutrophils, and platelets, tyrosine phosphorylation and activation of the 76 kDa SYK apparently are a common link in the integrin signaling chain. SYK is a member of a family, also containing ZAP-70, which is characterized by the presence of two tandem N-terminal SH2 domains and a C-terminal kinase domain (fig. 4) (Law et al., 1994). The SH2 domains of SYK are capable of binding phosphotyrosine residues in B- and T-cell receptors, leading to phosphorylation of SYK and activation of SYK during T- and B-cell stimulation by antigens (Couture et al., 1994; Qian and Weiss, 1997).

	V. Direct Signal Transduction by Integrins: Activation of the Mitogen-Activated Protein Kinase Cascade

Top Previous Next References

Activation of the MAPK cascade is a common event in the responses to many diverse extracellular stimuli (Cobb et al., 1994b; Gotoh and Nishida, 1995; Guan, 1994). These stimuli vary widely in both their chemical nature (e.g., peptide growth factors, phorbol esters, okadaic acid) and in their biological activities [e.g., nerve growth factor (NGF) induces neuronal outgrowth in PC12 cells, whereas EGF induces cell division]. Several groups now have firmly established that integrin-mediated cell adhesion, in the absence of soluble factors, can lead directly to the activation of the MAPK cascade (Chen et al., 1994a; Clark and Hynes, 1996; Miyamoto et al., 1996; Morino et al., 1995; Papkoff et al., 1994; Renshaw et al., 1996b; Schlaepfer et al., 1994; Takahashi and Berk, 1996; Wary et al., 1996; Zhu and Assoian, 1995). Thus, the insoluble proteins of the ECM join the many factors that can activate this cascade. Since the discovery of the MAPKs, an enormous literature has accrued which describes in detail both the mechanism(s) through which soluble factors activate MAPK and the consequences of this activation. Currently, the activation of MAPK by cell adhesion is being similarly dissected. In this section, we will review the biochemical characteristics, possible mechanisms, and significance of integrin-mediated MAPK activation.

The general experimental paradigm used to demonstrate direct integrin signaling involves detaching adherent cells from tissue culture plates and replating them onto surfaces coated with integrin ligand, typically a purified ECM protein such as fibronectin, vitronectin, or laminin. The supposition that MAPK activation is caused specifically by adhesion mediated by integrins, rather than by heterologous cell surface receptors or through nonspecific binding, has been confirmed using three experimental approaches. First, surfaces coated with polylysine support firm, nonspecific attachment of cells, but this attachment does not significantly activate MAPK (Chen et al., 1994a; Mainiero et al., 1997; Miyamoto et al., 1995b; Morino et al., 1995; Schlaepfer et al., 1994; Wary et al., 1996; Zhu and Assoian, 1995). Second, several groups have reported activation of MAPK when cells are plated on surfaces or incubated with beads coated with antibodies specific for various integrins but not with antibodies against other cell surfaces proteins, such as CD44 or MHC class I (Chen et al., 1996b; Lin et al., 1997a; Mainiero et al., 1997; Miyamoto et al., 1995b; Morino et al., 1995; Wary et al., 1996). Third, immobilized recombinant fragments of fibronectin and fibronectin-derived RGD peptides have been used to mediate cell attachment and subsequent MAPK activation (Chen et al., 1994a; Lin et al., 1997a), and soluble RGD peptides have been used to compete integrin-mediated adhesion to fibronectin and block MAPK activation (Zhu and Assoian, 1995).

Using this paradigm, integrin-mediated MAPK activation has been demonstrated in a wide variety of cell types. The predominant cell type in these studies is the fibroblast, with the source species typically being rodent (NIH or Swiss 3T3, REF52), human (WI38, HDF), and occasionally avian (CEF, primary quail embryo fibroblasts). Other cell types that exhibit integrin-mediated MAPK activation include human epithelial cells (293, HeLa), keratinocytes, and endothelial cells (human umbilical vein endothelial cells). All cell lines listed thus far are derived from solid tissues and are grown as adherent cultures. However, monocytic cells (THP-1) and platelets are nonadherent cells that exhibit integrin-mediated MAPK activation (Clemetson, 1995; Wahl et al., 1996). The activation of MAPK in response to cell adhesion is not an occurrence restricted to mammalian cells, or even to metazoans. The unicellular parasite Entamoeba histolytica, which expresses an integrin-like collagen receptor, shows activation of a MAPK homolog upon attachment to collagen (Perez et al., 1996). This suggests that integrin-mediated MAPK activation is likely to be an important biochemical event in many cell types.

In NIH3T3 cells, MAPK activity peaks 10 to 15 min after plating on fibronectin, whereas lower, but significant, activity persists 40 to 60 min after plating (Chen et al., 1994a; Clark and Hynes, 1996; Lin et al., 1997a; Renshaw et al., 1996b; Zhu and Assoian, 1995). Similar time courses are seen in other rodent (Swiss 3T3, REF52) fibroblast lines (Chen et al., 1994a) as well as in human dermal (Miyamoto et al., 1995b; Morino et al., 1995) and lung (Chen et al., 1994a) fibroblasts. Swiss 3T3 and REF52 cells also show activation of MAPK 10 min after adhesion to laminin-coated surfaces (Chen et al., 1994a). In primary human keratinocytes plated on laminin, peak MAPK activation occurs 30 min after plating (Mainiero et al., 1997), slightly slower than in fibroblasts on fibronectin. Conversely, HUVECs plated on fibronectin show peak levels of MAPK activity within 10 min of plating (Takahashi and Berk, 1996), slightly faster than fibroblasts on the same matrix protein. From its peak, MAPK activity undergoes a gradual decline (Chen et al., 1994a, 1996b; Clark and Hynes, 1996; Lin et al., 1997a; Renshaw et al., 1996a) or a delayed, but relatively rapid decrease (Mainiero et al., 1997; Miyamoto et al., 1996; Morino et al., 1995; Zhu and Assoian, 1995). As a result, integrin-mediated MAPK activity may persist at 60 to 100% of maximum levels for periods ranging from 15 min to 2 h or more. Thus the time course of integrin-mediated MAPK activation is likely to vary based on the type of cell, integrin, and matrix involved.

The durable activation of MAPK by integrins is in stark contrast to that seen in response to many soluble mitogens such as peptide growth factors (e.g., PDGF), which typically elicit a sharp peak of activity within 2 to 5 min (Gotoh and Nishida, 1995; Zhu and Assoian, 1995). It is clear that integrin-mediated MAPK activation alone is insufficient to promote induction of DNA synthesis (Mainiero et al., 1997; Zhu and Assoian, 1995). This is despite the fact that integrin-mediated adhesion, like growth factor stimulation, results in nuclear translocation of MAPK (Chen et al., 1994a; Davis, 1995; Zhu and Assoian, 1995) and activation of MAPK-responsive transcription factors (Davis, 1995; Hipskind et al., 1994; Janknecht et al., 1993; Mainiero et al., 1997; Wary et al., 1996). It has been suggested that the duration of MAPK activity can affect the nature of the cellular response (Marshall, 1995); in PC12 cells, for example, transient or prolonged MAPK activity induces cell division or differentiation, respectively. How the duration of integrin-mediated MAPK activity may affect downstream biological events is not yet understood.

The failure of cells to enter S phase in response to integrin-mediated MAPK activation may be caused by the level of activity induced by adhesion. In most cases, the average induction of MAPK activity by adhesion is three- to four-fold (Chen et al., 1994a; Clark and Hynes, 1996; Lin et al., 1997a; Morino et al., 1995; Renshaw et al., 1996b; Zhu and Assoian, 1995), although some groups have reported more robust responses (from six- to ten-fold) in fibroblasts stimulated with ligand- or antibody-coated beads (Miyamoto et al., 1995b), in HUVECs transiently transfected with a MAPK expression plasmid (Wary et al., 1996), and in primary keratinocytes (Mainiero et al., 1997). Nonetheless, the average level of integrin-induced MAPK activity is significantly less than the 20- to 50-fold activation induced in adherent cells by classical mitogens such as PDGF (Kyriakis et al., 1993; Reuter et al., 1995; Zhu and Assoian, 1995), EGF (Chen et al., 1996b; Zheng et al., 1994a) and phorbol esters (Marquardt et al., 1994; Renshaw et al., 1996b; Ueda et al., 1996). It is likely that a certain threshold of MAPK activity must be reached or exceeded to deliver a cell into S phase, whereas a subthreshold level of activity, regardless of duration, will lend itself to an altogether different biochemical event(s). The possible nature of this event in the case of integrin-mediated MAPK activation will be discussed in detail below. It should be noted that the robust activation of MAPK by polypeptide growth factors occurs in adherent cells, whereas responses to growth factors are markedly attenuated in cells in suspension. Thus, as discussed in detail in Section IX., both soluble growth factors and integrin-mediated cell anchorage seem to be required for efficient mitogenic signaling and cell cycle traverse.

1. Mechanism of growth factor activation of mitogen-activated protein kinase Upon treatment of cells with peptide mitogens (e.g., PDGF, EGF), the respective receptor tyrosine kinase (RTK) homodimerizes and undergoes autophosphorylation on various tyrosine residues. These modifications create binding sites for proteins containing SH2 domains (Pawson and Gish, 1992). Some of these proteins are enzymes, such as kinases (Src, p85^PI-3K, CSK), phosphatases (SHP), GTPase activators (p120Ras^GAP), and regulators of phospholipid turnover [phospholipase C (PLC) ], whereas others are adapter proteins (Shc, Nck, GRB2) that serve to relocate specific target proteins to sites of tyrosine phosphorylation (Schlessinger, 1994). One of these target proteins, SOS, is a Ras guanine nucleotide exchange factor (GNEF) and is recruited to activated RTKs through its interaction with GRB2. Once at the membrane, SOS promotes conversion of Ras to an activated, GTP-bound form (Downward, 1996). Activated Ras binds the amino-terminal domain of Raf-1 (Van Aelst et al., 1993; Vojtek et al., 1993) forming a stable complex (Finney and Herrera, 1995; Hallberg et al., 1994) and thereby localizing Raf to the plasma membrane. Subsequent to this localization, Raf is activated (Leevers et al., 1994; Stokoe et al., 1994) by a mechanism that remains unclear (Cutler and Morrison, 1997), but is likely to involve phosphorylation of tyrosine (Fabian et al., 1993; Marais et al., 1995) or serine (Morrison et al., 1993) residues, and may include interactions with 14-3-3 proteins (Fantl et al., 1994; Fu et al., 1994; Irie et al., 1994; Michaud et al., 1995) or phospholipid (Ghosh et al., 1996; Mott et al., 1996). Additional complexity comes from observations that various isoforms of protein kinase C (PKC) can directly phosphorylate and activate Raf in a Ras-independent manner (Cai et al., 1997; Carroll and May, 1994; Kolch et al., 1993; Ueda et al., 1996). Once activated, Raf phosphorylates and activates MAPK/ERK kinase (MEK), the dual-specificity (threonine/tyrosine) kinase directly responsible for activation of MAPK (Graves et al., 1995).

2. Mechanism of integrin-mediated activation of mitogen-activated protein kinase Tracing the pathway of integrin-mediated MAPK activation backward from MAPK offered no surprises, initially. To date, the only way to activate MAPK is through direct phosphorylation by MEK (Ahn et al., 1992), and auspiciously, MEK is activated by integrin-mediated adhesion with kinetics similar to those of MAPK activation (Chen et al., 1996b). Moreover, activation of MEK by adhesion is required for the activation of MAPK, as pharmacological inhibition of MEK activation ablates integrin-mediated MAPK activation in fibroblasts (Chen et al., 1996b) as well as human monocytes (McGilvray et al., 1997). It is safe to assume that, as no other mechanisms for activation of MAPK besides phosphorylation by MEK have been discovered, that integrin-mediated MAPK activation will behave no differently in this regard.

View larger version (32K): [in this window] [in a new window]   Fig. 6.   Possible mechanisms for integrin-mediated activation of the MAPK pathway. Currently, the best characterized mechanisms involve activation through FAK- and Ras-dependent pathways (A), although substantial evidence suggests that other distinct mechanisms exist. Integrin-mediated adhesion regulates several activities that can contribute to MAPK activation, especially through activation of Raf. Among these activities are phospholipases C and D (B), protein kinase C (B), the integrin-linked kinase (ILK) (C), and Src-family tyrosine kinases (C). Activation of Raf coincides with its localization to the plasma membrane and cytoskeleton, raising the possibility that integrin-mediated adhesion may directly recruit Raf to the membrane or may regulate a membrane-associated Raf activator (X in D). Other regulators of various components of the MAPK pathway have been described (D), but their regulation by adhesion has not yet been determined. See text for details.

c. EVIDENCE FOR FOCAL ADHESION KINASE-INDEPENDENT MECHANISMS. Similarly, reports have suggested that integrin-mediated MAPK activation can occur independently of FAK function. The first report demonstrates that whereas FAK tyrosine phosphorylation requires integrin  subunits, the recruitment and activation (by tyrosine phosphorylation) of the Shc adapter protein apparently is specified by the transmembrane and extracellular juxtamembrane regions of  subunits in a mechanism involving interaction with caveolin (Wary et al., 1996

3. The role of focal contact hierarchies and cytoskeletal organization in integrin-mediated mitogen-activated protein kinase activation As might be expected, there is an intricate relationship between integrin-mediated focal contact assembly and integrin-triggered signal transduction. Recent studies (Miyamoto et al., 1995a,b) have illustrated a hierarchical order in the assembly of cytoskeletal and signaling complexes that depends both on receptor occupancy and on lateral aggregation of integrins in the membrane. Thus, clustering of integrins with antibodies was sufficient to recruit the focal contact proteins FAK and tensin to the cytoplasmic face of the integrin cluster; in the absence of tyrosine kinase inhibitors, FAK was activated and numerous signaling transducing proteins were recruited. Further, integrin clustering by antibodies was sufficient to activate both MAPK and JNK, but these processes were blocked when tyrosine kinase inhibitors were used. However, integrin clustering, without occupancy of the integrin's ligand binding site, did not support additional recruitment of focal contact proteins such as vinculin, -actinin, and talin. By contrast, these three proteins were recruited when both ligand binding and integrin clustering took place, even in the presence of tyrosine kinase inhibitors. As discussed in more detail below, the use of cytochalasin D disrupted the recruitment of most signaling components. Thus, development of a mature focal contact assembly and its associated signaling complexes requires integrin aggregation, integrin occupancy, tyrosine kinase activity, and actin cytoskeletal integrity in a coordinated and hierarchical manner.

Although the picture is presently incomplete, it seems that integrin-mediated cell adhesion can regulate the SAPK or JNK pathway, a kinase cascade which closely resembles the MAPK pathway (Kyriakis et al., 1995; Waskiewicz and Cooper, 1995). Surprisingly, JNK activity has been reported to be both stimulated (Mainiero et al., 1997; Miyamoto et al., 1995a,b; Wary et al., 1996) and suppressed (Cardone et al., 1997; Frisch et al., 1996a) by integrin-mediated adhesion, but the reason for this contradiction is not clear. Integrin-mediated activation of JNK may contribute to adhesion-regulated gene expression through phosphorylation and activation of c-Jun (Minden et al., 1995). Further, integrin-mediated suppression of JNK activity may prevent apoptosis; this topic is discussed at length in Section X.

The biological role of integrin-mediated MAPK activation is not known. Adhesion to the ECM is required for several cellular functions. Although integrin-mediated MAPK activation is durable (lasting tens of minutes to hours), its duration is shorter than the time span during which adhesion mediates its effects on cell growth and differentiation. Thus, if integrin-mediated MAPK activation is important for regulating anchorage-dependent events, it must establish some type of "permanent record" to allow adhesion-dependent signaling events to occur properly. This may be through alteration of the cellular architecture, regulation of gene expression, or a complex commixture of both. Alternatively, integrin-mediated MAPK activation may serve a more immediate function in regulating new contacts with the ECM, such as promotion of cell spreading or cell motility. Insight into the function of integrin-mediated MAPK activation is likely to come from investigation of the known effectors of MAPK and their regulation by adhesion. There is no lack of candidates, because MAPK has diverse targets. However, as discussed earlier, the kinetics of adhesion-mediated MAPK activity differs substantially from the kinetics of mitogen-stimulated MAPK activity. Thus, the subset of potential targets that are affected, as well as the degree to which those targets are affected, may very well differ between adhesion-mediated and mitogenic MAPK activities. An overview of integrin signaling to the MAPK cascade is given in fig. 7.

View larger version (47K): [in this window] [in a new window]   Fig. 7.   Hypothetical model for collaboration between integrin-mediated signaling events. The function of integrin-mediated mitogen-activated protein kinase (MAPK) activation is not fully understood but possibilities include activation of transcription factors, PLA2, and myosin light chain kinase (MLCK). Rho family GTPases (Cdc42, Rac, Rho) have many effectors, some of which mediate their effects on actin cytoskeletal organization. Among those known for Rho in this regard are PI4-5K, a lipid kinase important for PIP2 synthesis, and Rho kinase (RhoK). Synthesis of PIP2 by PI4-5K could provide a substrate for MAPK-activated PLA2. Also, RhoK can directly phosphorylate the myosin light chain (MLC) and can inhibit MLCP. These events may synergize with activation of MLCK by MAPK to increase MLC phosphorylation, which can lead to contractility, focal adhesion formation, and other morphological effects. See text for details.

1. Transcriptional regulation One of the hallmark functions for MAPK is the regulation of various transcription factors, including c-Myc, ATF-2, Elk-1, SAP-1, and AP-1 (Davis, 1995; Karin, 1995; Treisman, 1996). As mentioned previously, MAPK translocates to the nucleus after integrin-mediated adhesion (Chen et al., 1994a; Zhu and Assoian, 1995). Integrin-mediated adhesion can activate transcription from the c-fos serum response element (SRE) (Mainiero et al., 1997; Wary et al., 1996), a MAPK-responsive promoter element that is regulated, in part, by Elk-1 and SAP-1 (Davis, 1995; Denhardt, 1996; Karin, 1995). Furthermore, it has been known for several years that adhesion to fibronectin rapidly induces the expression of both c-fos and c-jun (Dike and Farmer, 1988). Whereas any promoter that contains an SRE may potentially be affected by integrin-mediated MAPK activation, it is particularly interesting to note that the vinculin promoter contains an SRE (Moiseyeva et al., 1993) and that expression of vinculin, as well as actin, -actinin, the 1 integrin subunit, and fibronectin, is induced with immediate-early kinetics (Ryseck et al., 1989). To date, however, physiological transcriptional targets for integrin-mediated MAPK activity have not been identified. Nonetheless, it is of considerable importance that the role of the MAPK pathway in integrin-induced gene expression be investigated.

2. Cytoplasmic targets In addition to transcription factors, potential targets for adhesion-activated MAPK also exist in the cytoplasm. One such target is pp90^RSK (a.k.a. MAPKAPK-1) (Denhardt, 1996; Papkoff et al., 1994); MAPK can phosphorylate and activate pp90^RSK, which can then move into the nucleus (Chen et al., 1992a) and contribute to the regulation of transcription by AP-1 and cAMP-responsive element binding protein (CREB) (Denhardt, 1996). Another candidate for integrin-mediated MAPK activation is not a single enzyme, but a cytoarchitectural system, the microtubule network. MAPK is associated physically with and has enzymatic activity toward microtubule components (Reszka et al., 1995) (MAPK once stood for microtubule-associated protein-2 kinase). Two recent reports have implicated microtubules and microtubule-associated MAPK activity in the regulation of integrin-mediated signaling events and the formation of focal adhesions and stress fibers (Bershadsky et al., 1996; Reszka et al., 1997). Finally, a very attractive cytosolic target for integrin-mediated MAPK activation is PLA₂ (Clark and Hynes, 1996). PLA₂ is responsible for liberating arachidonic acid (AA) from glycerolphospholipids such as PI(4,5)P₂ (Divecha and Irvine, 1995), an activity important for optimal integrin-mediated cell-substrate interactions in some cell types (e.g., HeLa) (Auer and Jacobson, 1995; Chun and Jacobson, 1993) but apparently not in others (e.g., NIH3T3) (Clark and Hynes, 1996).

3. Cell adhesion and motility Two recent reports provide a highly satisfying function for integrin-mediated MAPK activation by suggesting that MAPK activated by integrins may "feedback" and contribute to the regulation of cell adhesion and motility. The first demonstrated that constitutively active mutants of MAPK pathway components, namely Ras and Raf, can suppress the activation of integrins (Hughes et al., 1997). Also, activation of Raf inhibits fibronectin matrix assembly and induces cell rounding. Suppression of integrin activation correlates with activation of MAPK, but is not caused by direct phosphorylation of the integrin and is independent of protein and mRNA synthesis. Inactivation of integrins by MAPK may represent a negative feedback loop for the regulation of integrin function which may contribute to the dynamic modulation of ligand binding affinity. Moreover, the efficacy of this loop in a given cell type may be a determinant of the duration of integrin-mediated MAPK activation in that cell type. The second report demonstrated that expression of a constitutively active MEK mutant stimulates cell migration on collagen and leads to phosphorylation and activation of myosin light chain kinase (MLCK), leading to increased phosphorylation of the myosin light chain (MLC) (Klemke et al., 1997). Inhibition of MAPK activity suppresses both cell migration and phosphorylation of MLCK and MLC, but not cell adhesion or in situ cell spreading, suggesting that integrin-mediated MAPK activation is not required for these initial events in this system. In vitro experiments showed that MAPK can phosphorylate and activate MLCK directly. Maximal phosphorylation and activation of MLCK by MAPK in vitro required prolonged incubation (approximately 40 min). Hypothetically, a similar requirement in vivo (should it exist) may be met by the extended duration of MAPK activity after cell adhesion. Whereas the evidence linking MAPK and MLCK is compelling, it is clear that other mechanisms may contribute to regulation of MLC phosphorylation and subsequent control of actinomyosin dynamics. Most notable are the phosphorylation of MLC and inactivation of MLC phosphatase by Rho-kinase (Amano et al., 1996b; Burridge and Chrzanowska-Wodnicka, 1996; Kimura et al., 1996), which may govern cell contractility and subsequent signal transduction and cytoskeletal organization (Burridge and Chrzanowska-Wodnicka, 1996; Chrzanowska-Wodnicka and Burridge, 1996; Tapon and Hall, 1997) (see fig. 7 and Section VI.). The relative contributions of MAPK and other regulators of MLCK function still must be determined. Nonetheless, these recent observations have placed MAPK in a central locus for the regulation of cell adhesion, contractility, and locomotion.

It is also important to realize that integrin signaling may impinge on effectors other than MAPK. For example, integrin-mediated activation of Raf may have downstream effects that are independent of MAPK. Raf is implicated in the transcriptional activation of NF-B (Finco and Baldwin, 1993; Janosch et al., 1996; Li and Sedivy, 1993), the regulation of apoptosis (Troppmair and Rapp, 1997) through interactions with the antiapoptotic proteins BCL-2 (Wang et al., 1994, 1996b) and BAG-1 (a BCL-2 interacting protein) (Wang et al., 1996b), and phosphorylation of the pro-apoptotic protein Bcl-XL/Bcl-2-associated death promotor (BAD) (Wang et al., 1996a). Raf also can phosphorylate the p53 tumor suppressor protein (Jamal and Ziff, 1995) and cdc25, a dual specificity phosphatase that activates cyclin-dependent kinases (Galaktionov et al., 1995). Thus, it is important to remember that the activation of MAPK pathway components is only one of many signaling events triggered by integrin-mediated adhesion (see fig. 7).

	VI. Integrins and Rho Family Guanosine 5'-Triphosphatases

Top Previous Next References

An important consideration in the field of integrin-mediated signaling is what structure is actually eliciting the signals. As discussed in Section V. above, the formation of signaling complexes is coordinated closely with the assembly of focal contact/cytoskeletal structures. Further, it is well documented that focal adhesions and other integrin-associated structures "mature" with increasing time after adhesion (Burridge et al., 1988; Craig and Johnson, 1996; Gilmore and Burridge, 1996a; Jockush et al., 1995). Some integrin-mediated signals simply may promote focal adhesion assembly thus setting the stage for other events, some may require early or intermediate stages of maturation of adhesion complexes, whereas others may be generated only by a fully formed, mature focal adhesion complex and its associated linkages to the actin cytoskeleton.

The Rho family of small GTPases, including various forms of Cdc42, Rac, and Rho, are involved intimately in the regulation of the actin cytoskeleton. Several excellent recent reviews describe practically every aspect of the Rho family (Denhardt, 1996; Lim et al., 1996; Narumiya, 1996; Ridley, 1996; Symons, 1996; Tapon and Hall, 1997), and an exhaustive analysis will not be attempted here, but an introduction is in order. Rho family GTPases couple extracellular signals to the formation and/or organization of higher-order actin structures. Specifically, Rho, Rac, and Cdc42 mediate the formation of stress fibers, lamellipodia, and filopodia, respectively (Nobes and Hall, 1995a; Ridley and Hall, 1992; Ridley et al., 1992). Distinct extracellular signals activate individual members of the Rho family through specific receptors, either RTKs or G-protein-coupled receptors. In this way, LPA, bombesin, and bradykinin stimulate the activity of Rho, Rac, and Cdc42, respectively, to induce formation of the respective actin structures mentioned above. Cdc42 activity can lead to Rac activation, and Rac activity can lead to Rho activation, suggesting a GTPase cascade (Chant and Stowers, 1995). In support of this, LPA induces stress fibers within 2 to 5 min, whereas induction of stress fibers by bombesin and bradykinin (as well as PDGF, EGF, and insulin) occurs during a longer time course of 20 to 30 min (Nobes and Hall, 1995b). However, the regulation of this cascade is not without its intricacies, as the formation of filopodia by Cdc42 correlates with the dissolution of stress fibers (Kozma et al., 1995), and inhibition of Rho can enhance Cdc42 activity (Nobes and Hall, 1995a). An overview of Rho family GTPases in relation to integrin signaling and the cytoskeleton is provided in fig. 7.

The pathways between the various receptors and Rho family proteins have not been fully charted, and several candidate intermediaries have been suggested including PI-3K, PKC, and a tyrphostin-sensitive tyrosine kinase (Chrzanowska-Wodnicka and Burridge, 1994; Nobes et al., 1995). However, because Rho family proteins are active when bound to GTP and inactive when bound to guanosine 5'-diphosphate (GDP), the pathways that regulate their function will involve proteins which control the state of the bound nucleotide, namely GAPs and GNEFs (Denhardt, 1996; Lamarche and Hall, 1994; Lim et al., 1996). GAPs enhance the hydrolysis of GTP and thus promote inactivation of Ras-related proteins, whereas GNEFs promote the release of GDP, allowing GTP to bind and activate the proteins. GAPs specific for Rho family members include Ral-BP1 (a.k.a. Cdc42GAP), breakpoint cluster region (BCR) and active, breakpoint cluster region-related (ABR), p190Rho^GAP, chimaerins (, , and n) (Herrera and Shivers, 1994; Kozma et al., 1996) and an atypical myosin (myr5) (Reinhard et al., 1995). Rho family GNEFs include diffuse B-cell lymphoma (DBL), Ost (from osteosarcoma), Lbc, Vav, Tiam1, the faciogenital dysplasia gene product (FGD1) (Zheng et al., 1996b), and small G-protein GTP dissociation stimulator (smgGDS). Ost acts as an effector for Rac and stimulates nucleotide exchange and activation of Cdc42 and Rho (Horii et al., 1994), and therefore may play a role in mediating a GTPase cascade. Significantly, many Rho family GNEFs contain a PH domain, which can bind to subunits of heterotrimeric G-proteins, suggesting a possible direct link between G-protein-coupled receptors and Rho family proteins. PH domains can also bind PI(4,5)P₂, which may serve to localize Rho GNEFs to the plasma membrane (Lemmon et al., 1996; Michiels et al., 1997). Another class of GTPase regulators is represented by Rho guanine nucleotide dissociation inhibitor (GDI), which inhibits GDP dissociation (and in some instances GTP hydrolysis), sequesters GDP-bound GTPases in the cytoplasm, prevents the interaction of Rho GTPases with GAPs, and also may be involved in membrane relocation of Rho proteins (Denhardt, 1996).

Although serum components clearly can activate Rho and its relatives, the question of whether cell adhesion also can stimulate their activation remains unanswered. As mentioned previously, formation of focal adhesions and stress fibers is a Rho-dependent process; in Swiss 3T3 cells, this process requires both attachment to the ECM and serum components (Hotchin and Hall, 1995; Burridge and Chrzanowska-Wodnicka, 1996). However, some cell types can form stress fibers and focal adhesions under serum-free conditions and can maintain these structures after the removal of serum (Chrzanowska-Wodnicka and Burridge, 1996); it is not clear whether this is because of persistence of Rho activity after serum withdrawal or persistence of cytoskeletal structures after Rho inactivation. This suggests that in these cells, either Rho (and stress fiber/focal adhesion formation) is not required continually for the maintenance of cell adhesion, or Rho is not fully inactivated by serum withdrawal. If signaling events do depend on the formation of mature focal adhesions, and focal adhesion formation depends on Rho, an elementary step in integrin-mediated signaling must be the activation of Rho; this also implies regulation of Rho GNEFs and Rho GAPs. Thus, an important avenue of future experimentation should be whether (and how) Rho GNEFs and GAPs are regulated by cell adhesion. In support of this, there is certainly emerging evidence that suggests that Rho proteins are mediators of integrin signaling (Barry et al., 1997; Nakahara et al., 1998; Schwartz et al., 1996).

The list of proteins with which Rho family members interact is growing rapidly (table 2) (Ridley, 1996; Tapon and Hall, 1997). With the knowledge that Rho GTPases instigate complex and sometimes diverse cellular events (see below), the notion that these proteins may have many targets makes sense. Some of these target proteins mediate cytoskeletal organization, whereas others are likely to mediate noncytoskeletal effects, and still others are currently without discernible function. The assignment of effectors to downstream events is still in the early stages, and although substantial progress has been made, it is complicated by several regulatory intricacies. For instance, several effectors are shared by more than one GTPase: the serine/threonine kinase p65^PAK is activated by both Cdc42 and Rac, and both Rac and Rho associate with PI4-5K lipid kinase activity (see table 2). Further complexity is added by communication between effectors. A prime example of this is the direct regulation by Cdc42 and Rac of both PI-3K and its effector p70^S6K.

In addition to their functions in reorganization of the actin cytoskeleton (see below), Rho, Rac, and Cdc42 are required for the serum-induced progression of fibroblasts into S phase (Olson et al., 1995; Yamamoto et al., 1993), although recent evidence casts doubt on the role of Cdc42 in this regard (Molnar et al., 1997). Furthermore, Rac is required for induction of cell division by the oncogenic tyrosine kinase v-Abl (Renshaw et al., 1996a) and both Rac and Rho are required for efficient transformation by oncogenic Ras (Khosravi-Far et al., 1995; Qiu et al., 1995a,b) and can synergize with a weakly transforming allele of Raf to promote efficient transformation (Khosravi-Far et al., 1995). Further affirmation for the role of Rho family proteins in cell cycle progression comes from the observation that several Rho family GNEFs, namely DBL, Vav, and Lbc have the potential to oncogenically transform fibroblasts (Hunter, 1997; Khosravi-Far et al., 1994; Zheng et al., 1995).

One mechanism through which Rho family proteins could regulate cell cycle progression is by activation of transcription factors. All three Rho family GTPases can activate NF-B, and Cdc42 and Rho are required for activation of this factor by TNF (Perona et al., 1997). Another target for Rho family transcriptional regulation is serum-response factor (SRF). SRF, alone or in a complex with ternary complex factor (TCF) proteins (e.g., Elk-1, SAP-1), regulates transcription from promoters containing an SRE (Treisman, 1995). Activated forms of Cdc42, Rac, and Rho can all induce SRF-dependent expression from the SRE, and Rho is required for activation of the SRE by several stimuli (e.g., LPA, PDGF, phorbol ester) (Hill et al., 1995). Recently it has been demonstrated that activated Rho-kinase (see below) can stimulate transcriptional activity from the c-fos SRE (Chihara et al., 1997). The ability of Rac to regulate SRF activity is separable from its cytoskeletal functions (Westwick et al., 1997), and there is evidence to suggest that Rho also can regulate transcription in the absence of actin effects (Thornburn et al., 1997). Expression of a constitutively active form of PI-3K (a putative effector for Cdc42 and Rac) can induce the formation of Rac- and Rho-dependent actin structures (i.e., lamellipodia and stress fibers), but cannot stimulate Rac- and Rho-mediated transcriptional activation (Reif et al., 1996), emphasizing the separation of these functions. The activation of SRF transcriptional activity raises the interesting possibility that Cdc42/Rac/Rho-stimulated SRF may combine with TCF proteins phosphorylated by integrin-mediated MAPK activity to induce transcription of genes containing SREs, such as c-fos or vinculin (see Section V.). The induction of c-fos expression is particularly interesting in light of the fact that Cdc42 and Rac can also activate JNK (Coso et al., 1995; Minden et al., 1995), which phosphorylates and activates c-Jun. Jun dimerizes with c-Fos to form the AP-1 complex, which is implicated in the response to integrin-mediated cell adhesion in several systems (Dike and Farmer, 1988; Fan et al., 1995; Tremble et al., 1994) (see fig. 6).

Activation of the JNK pathway by Cdc42 and Rac originally was thought to be mediated by p65^PAK because of its interaction with the two GTPases and its homology to Ste20p, the upstream activator of a yeast JNK/SAPK homolog (Waskiewicz and Cooper, 1995). However, it seems clear that p21-activated kinase (PAK) is not involved in Cdc42- and Rac-mediated JNK activation (Teramoto et al., 1996a,b; Westwick et al., 1997; Zhang et al., 1995a), although it may mediate the activation of the related p38/RK pathway (Zhang et al., 1995a). Instead, the serine/threonine kinase MLK3, a mixed-lineage kinase that can interact with Cdc42 and Rac and whose overexpression can activate the JNK pathway, seems to be involved in mediating this effect (Teramoto et al., 1996a). Finally, direct interaction between Cdc42 or Rac and MEKK1, a known upstream activator of the JNK pathway, has been described recently (Fanger et al., 1997) and provides another highly plausible mechanism for JNK activation by Rho family GTPases.

All three Rho family GTPases have profound effects on cell morphology and microfilament organization (Tapon and Hall, 1997), but the mechanisms by which they exert their effects are not fully understood. All three proteins cycle on and off the plasma membrane (Adamson et al., 1992; Bokoch et al., 1994; Takeichi, 1995), although the role that this plays in the regulation of their function is unclear. Of the known effectors (see above), the best candidate for directly mediating Cdc42-induced actin effects is Wiskott-Aldrich Syndrome protein (WASP). Overexpression of WASP in Jurkat cells induces the formation of large cytoplasmic clusters of polymerized actin which colocalized with WASP (Symons et al., 1996). However, WASP expression is restricted to hematopoietic cells, and therefore cannot explain the induction of filopodia by Cdc42 in fibroblasts, although the existence of WASP isoforms in other cells types is a possibility. The Rac-interacting protein POR-1 clearly is involved in Rac-mediated lamellipodia formation (truncated POR-1 interferes with Rac-induced membrane ruffling) (Van Aelst et al., 1996), although it has no known catalytic activity and only a leucine-zipper region to suggest a possible function. Finally, the serine/threonine kinase p65^PAK has been implicated recently as a possible effector for both Cdc42 and Rac in rearrangement of the actin cytoskeleton (Sells et al., 1997), although this is controversial (Westwick et al., 1997).

1. Lipid metabolism PI4P-5K, a kinase regulated by Rac and Rho, is the major enzyme responsible for generating PI(4,5)P₂, a lipid of critical importance in numerous cellular functions. It has been known for several years that the synthesis of this lipid is induced by integrin-mediated cell adhesion (McNamee et al., 1993). By their activation of PI4P-5K, Rac and Rho provide a molecule that can impinge on a wide variety of biochemical pathways. PI(4,5)P₂ is a substrate for phospholipases A, C, and D, which generate AA, DAG plus IP₃, and phosphatidic acid (PA), respectively. Integrin-mediated adhesion induces PLA₂ activity and a concomitant increase in AA which precedes cell spreading (Auer and Jacobson, 1995; Chun and Jacobson, 1992; Cybulsky et al., 1993). It is tempting to speculate that this activity may be attributable to integrin-mediated MAPK activation, but this has not been tested formally. AA release and subsequent generation of leukotrienes may be both necessary and sufficient to activate Rho (through an unknown mechanism) and induce stress fiber formation in a Rho-dependent manner (Peppelenbosch et al., 1995), thus possibly linking stress fiber assembly to the initial cell adhesion event in a positive feedback manner.

2. Rho, mitogen-activated protein kinase, and contractility In addition to the effects on transcription and on actin filaments mentioned earlier, another intriguing collaboration between signals from the Rho family proteins and integrin-activated MAPK might occur in the regulation of cell contractility (fig. 7). As mentioned above, Rho uses Rho-K as an effector to regulate MLC phosphorylation. Specifically, Rho-K phosphorylates and deactivates the myosin binding subunit of myosin light chain phosphatase (MLCP) (Kimura et al., 1996). Furthermore, Rho-K is able to directly phosphorylate myosin at the same site targeted by MLCK (Amano et al., 1996a). Also, as stated earlier, MAPK activated by integrins can phosphorylate and activate MLCK (Klemke et al., 1997). The result of each of these events is the net increase in MLC phosphorylation, which activates myosin and stimulates contractility (Burridge and Chrzanowska-Wodnicka, 1996). Although highly plausible, synergy between Rho and integrin-mediated MAPK activation in this regard has not yet been demonstrated. It has been proposed that contractility is the driving force behind the Rho-mediated formation of stress fibers and focal adhesions (Burridge and Chrzanowska-Wodnicka, 1996; Chrzanowska-Wodnicka and Burridge, 1996). Several other events mediated by Rho family proteins (discussed above) also may function in this regard. Increased intracellular PI(4,5)P₂ induced by Rac and Rho can stimulate actin polymerization and attachment of filaments to the membrane in primordial focal-adhesion-like structures. MLC phosphorylation is increased through inhibition of MLCP by Rho-K, and possibly by activation of MLCK by MAPK and/or calmodulin. This can stimulate cellular contractility and promote formation of stress fibers and focal adhesions.

There is clearly an intimate as well as intricate relationship between members of the Rho GTPase family and integrins. Rho family members seem to have two basic but separable functions, the first being signaling to nuclear transcription factors, and the second being control of cytoskeletal assembly. Integrins cooperate with Rho proteins in the organization and regulation of the actinomyosin containing cytoskeleton, because in the absence of integrin-mediated anchorage appropriate vectorial assembly of structures such as actin stress fibers cannot occur. As discussed above, Rho family proteins can influence cytoskeletal organization by several different means, including activation of kinases that regulate myosin function, and by the generation of lipids, such as PI(4,5)P₂, that can affect the function of key focal contact proteins. Conversely, Rho-dependent cytoskeletal structures may be essential "scaffolds" for integrin signaling, bringing together molecules that are involved both in direct integrin signaling and in integrin-mitogen collaboration (see Section IX.). Integrin-mediated signaling, for example through the MAPK pathway, then potentially could collaborate with direct Rho-family signaling via the JNK pathway or other pathways to provide a coordinated activation of key transcription factors involved in cell cycle progression. One aspect of our current understanding that is still obscure is whether integrin-mediated adhesion can influence directly the GAPs and GNEFs that regulate Cdc42, Rac, and Rho, although recent work suggests the existence of a link between integrins and Cdc42 (Schwartz, 1997).

	VII. Other Aspects of Integrin Signaling

Top Previous Next References

The integrin signaling processes described in Sections IV. and V. above, that is, FAK activation and activation of the MAPK cascade, can be initiated by several different integrins. However, a good deal of the biology of integrins seems to imply specific actions of different integrin subunits Thus, it seems likely that individual integrins participate in specific signaling events that thus far are delineated very poorly. Along these lines, it is interesting to speculate that several recently discovered proteins that bind selectively to the cytoplasmic tails of individual integrin subunits (described in Section III.) may play an important role in the specificity of integrin-mediated events within cells. Several integrin-specific actions on the cell cycle and on apoptosis are discussed in Section X. Here we will simply tabulate several examples of what are apparently specific biological actions of individual members of the integrin family that cannot be substituted by other members of the family (table 3).

In this review we have emphasized integrin signaling to downstream processes presumably involved in growth control. However, there is another aspect of integrin signaling that should be mentioned briefly. Several observations indicate that integrins signal to each other; this usually takes the form of one integrin regulating the ability of another integrin family member to support cell adhesion, motility, or phagocytosis. There are also a few instances of this sort of "cross-talk" between integrins and non-integrin adhesion receptors. Some examples of adhesion receptor cross-talk are given in table 4. In most cases the mechanistic basis of receptor cross-talk is unknown. However, as we proceed with the elucidation of integrin signaling pathways, described in Sections IV. through VI. above, several likely possibilities come to mind, and the issue of adhesion receptor interactions becomes less mysterious. First, recent observations (Hughes et al., 1997) have shown that elements of the MAPK cascade can modulate the ligand binding affinity of integrins. Because integrins can also activate the MAPK cascade (Section V.), this is one obvious way in which one integrin might cross-talk with another. Second, many integrin-mediated activities, such as cell adhesion or cell motility, require engagement with the cytoskeleton. Because of the interplay between integrins, Rho family GTPases, and the cytoskeleton (Section VI.), this suggests another way in which signals generated by one integrin can affect functions mediated by another. Third, it is now clear that several integrins can share interactions with transmembrane or cytoplasmic binding proteins (Section III.), thus hinting at yet another mechanism for cross-talk. With respect to cross-talk between integrins and other adhesion receptors, several interesting examples have emerged recently (table 4); the mechanisms are far from clear, but the biological significance may be substantial.

One interesting hypothesis concerning signaling by cell adhesion receptors is a mechanochemical one; the integrins or other receptors essentially would play the role of a physical link between the ECM and the cytoskelton, whereas the organization and mechanical tension of the cytoskeleton consequently would influence events in the cytoplasm and nucleus that are involved in cell division or cell differentiation. This "tensegrity" concept has been developed elegantly by D. Ingber and his colleagues (Ingber, 1993, 1997). In this model, overall cell shape and the bulk mechanical properties of cells would be as important or more important than discrete biochemical reactions. Although somewhat unsatisfying to the biochemically inclined, there is a remarkable amount of support for the idea that cell shape is a key regulator of cell growth and differentiation. For example, by use of powerful engineering techniques similar to those used in microchip manufacturing, very precise microdomains of adhesive material were placed on surfaces. When endothelial cells attached to those surfaces there was an excellent correlation between cell shape and the cell's ability to enter S phase (Chen et al., 1997). Another example of the importance of cell shape in controlling biological processes concerns studies with breast epithelial cells conducted by M. Bissell and colleagues. Here a substantial amount of evidence suggests that appropriate patterns of gene expression and cell differentiation only occur when the cells have both the correct biochemical signals and the correct cell shape (Boudreau et al., 1996; Roskelley et al., 1995). Although at first glance these observations seem at variance with mainstream concepts of biochemically based signaling pathways, this may not be so. As illustrated in this review, the more we learn about signaling pathways the clearer it becomes that the three-dimensional architecture of signaling complexes is a key aspect of efficient signaling. Thus, in the long run, a more detailed analysis of cell signaling may well resolve any conflicts between "tensegrity" models and more conventional biochemical concepts of signaling.

One of the major unresolved issues regarding pathways of integrin signaling concerns the relationships between FAK and MAPK. Under most circumstances there are close parallels in the circumstances and kinetics of FAK activation and MAPK activation. Further, overexpression of FAK has been reported to increase MAPK activation (Schlaepfer and Hunter, 1997). However, the preponderance of evidence suggests that FAK is not required for integrin-mediated MAPK activation (Frisch et al., 1996a; Lin et al., 1997a; Wary et al., 1996). Despite the numerous studies on FAK, the precise biological role of this kinase remains open to question. Many investigators emphasize a key role for FAK in regulating the formation or dissolution of focal adhesion structures and thus in controlling cell motility. Another alternative is that FAK plays a vital role in adhesion regulation of apoptosis (Frisch et al., 1996a); however, other mechanisms may be equally or more important in this phenomenon (Cardone et al., 1997) (Section X.).

Likewise, although many investigations have been concerned with activation of the MAPK pathway by integrin-mediated adhesion, our biological insights into this process remain limited. One important possibility is that integrin signaling to MAPK helps regulate cytoskeletal assembly through PLA₂ and arachidonate metabolites (Auer and Jacobson, 1995; Clark and Hynes, 1996) and contractility of actinomyosin through MLCK (Klemke et al., 1997). However, one should not discount the possibility that durable activation of MAPK via cell adhesion contributes to cell cycle traverse or cell differentiation mediated primarily by soluble growth or differentiation factors. We will return to this theme in Section IX.

In addition to events involving kinases, integrins have been reported to elicit a variety of other signaling events in cells, particularly ones involving mono- and divalent cations. Through regulation of ionic transients, integrin-mediated cell adhesion can affect the intracellular environment which has the potential to affect myriad signaling pathways. Integrin-mediated increases in intracellular calcium play an important role in the regulation of cell adhesion and can also regulate contractility through calmodulin-dependent regulation of MLCK. Furthermore, calcium transients may contribute to the regulation of PKC activity and may negatively regulate FAK-mediated signaling through calpain-mediated proteolysis (Cooray et al., 1996; Schwartz, 1993a; Sjasstad and Nelson, 1997). Integrin-mediated adhesion also elevates intracellular pH (pH_i), largely through activation of the Na⁺/H⁺ antiporter (Ingber et al., 1990; Schwartz et al., 1989, 1991a,b). Regulation of pH_i through antiporter activity plays an important, but not well understood, role in many aspects of cell growth and division (Bianchini and Pouyssegur, 1994; L' Allemain et al., 1984a,b; Pouyssegur et al., 1985; Vairo et al., 1992). One particularly relevant observation is that inhibition of antiporter function (pharmacologically or by genetic deficiency) inhibits Rho-mediated formation of stress fibers (Vexler et al., 1996). MAPK is thought to play a major role in regulation of antiporter function (Bianchini et al., 1997), raising the possibility that integrin-mediated cell adhesion may activate the Na⁺/H⁺ antiporter through a mechanism involving MAPK activation.

A confounding degree of cross-talk exists between various integrin-associated signaling events. One manifestation of this is the convergence of two or more regulators on a single effector. Several of the signaling events discussed previously can be regulated by multiple adhesion-related proteins. A good example is the regulation of PI-3K activity in response to adhesion. PI-3K is implicated as an effector for a host of regulators including Cdc42 and Rac, FAK, and Abl (see Sections IV. and V.), yet the relative contributions of each of these proteins in the adhesion-mediated regulation of PI-3K activity is not fully understood. Finally, the possible regulation of MLCK by integrin-mediated MAPK activation and of MLCP by Rho-kinase illustrates that signaling processes can use separate pathways to achieve the same result, in this case the phosphorylation of the MLC. Another characteristic of the intricacy of integrin signaling is that a given regulator and effector often can be connected through multiple pathways. The multiple potential mechanisms of integrin-mediated MAPK activation exemplify this. Another example is provided by the integrin regulation of intracellular calcium, which can be achieved through IP₃-mediated release of intracellular stores or activation of plasma membrane ion channels (Sjasstad and Nelson, 1997).

In summary, integrin-mediated cell adhesion regulates numerous biochemical activities that may impinge on an even greater number of cellular pathways and events. These activities may appear tortuously interwoven and at least a bit redundant. Clearly not all activities will occur in all cells under all adhesive conditions. Indeed, it is most likely a combination of (at least) the species of integrin and the type of matrix that will determine the panel of responses for any given cell type. Furthermore, despite the recent flurry of activity, the field of integrin signaling is still in its infancy. The consequences of integrin-mediated signals, their actual contributions to cell function, have been speculated upon more often than investigated. A simplistic, but functional hypothesis is that signals generated by the interaction of a cell with a physiologically relevant matrix will be permissive for or causal to that cell's function, whereas interaction of a cell with an inappropriate matrix will generate signals that suppress cell function or growth, or perhaps lead to apoptosis. Permissiveness for cell function can include regulation of cell morphology and/or motility, induction of specific acute responses (e.g., in immune or inflammatory cells), or communication of the state of proper attachment to other biochemical pathways to allow appropriate response to other extracellular signals such as differentiation and growth factors. This last aspect is of particular importance and will be discussed in detail in Section IX.

	VIII. Signaling by Cadherins, Selectins, and Immunoglobulin-Cell Adhesion Molecules

Top Previous Next References

In this section we will deal with signaling mediated by non-integrin cell adhesion receptors, particularly the cadherins, selectins, and Ig-CAMs. Considerable information is available concerning cadherin signaling, whereas less is known about signaling mediated by the other families of adhesion receptors. As we shall see, there are both similarities and differences in signal transduction pathways mediated by integrins and by non-integrin adhesion receptors.

A. Cadherin Signaling

1. Cadherins, catenins, adenomatous polyposis coli, and the Wnt pathway Current understanding of a major aspect of signal transduction involving cadherins comes through a felicitous congruence of research from several different fields. As described above, investigators interested in cadherin-mediated adhesion initially described the catenins as a family of peripheral membrane proteins involved in cadherin function and in the structure of cell-cell adherence junctions. At approximately the same time, developmental biologists were delineating the Wingless signaling pathway in Drosophila that is vital for establishing segment polarity in fly embryos and adults. Genetic evidence implicated the Armadillo protein as a key player in Wingless signaling; it turns out that Armadillo is the fly homolog of mammalian -catenin and has approximately 70% amino acid identity to the mammalian protein. Wingless is a secreted glycoprotein that is a member of the Wnt family of growth factors that control developmental patterning in both vertebrates and insects. Thus, Armadillo/-catenin is not only a structural protein of cell adhesion junctions, but also a key player in a vital developmental signal transduction cascade. To make matters even more interesting, tumor biologists identified a tumor suppressor gene called adenomatous polyposis coli (APC) whose loss or inactivation is responsible for a familial predisposition to colon cancer. The protein product of APC can bind to -catenin and regulate its stability, abundance, and ability to participate in signaling events. At this point the reader can appreciate that cadherins, catenins, APC, and Wnt proteins are all implicated in a complex signaling and growth control network. The past several years have seen a coalescence of our understanding of this network, as we now will discuss.

c. ARMADILLO/-CATENIN. Both Armadillo and -catenin (781 amino acids) contain 13 consecutive partially conserved 42 amino acid "armadillo" repeats flanked by N-terminal and C-terminal domains (Cowin and Burke, 1996

d. ADENOMATOUS POLYPOSIS COLI. The APC gene originally was identified as a key tumor suppressor gene in colorectal carcinogenesis. Individuals with inherited mutations in APC develop numerous colon polyps and are predisposed to the development of colon cancer. Human APC is a very large protein (2843 amino acids) with several distinct domains. APC interacts with -catenin, -catenin, GSK3, tubulin, as well as several other proteins (Huber et al., 1996

3. A model for the Wnt-cadherin signaling pathway Here we present a consensus model of the Wnt-Cadherin signal transduction cascade (see fig. 8) (Gumbiner, 1996; Huber et al., 1996; Peifer, 1996). The key feature of this model is the existence of discrete but interchangeable pools of -catenin/Armadillo which can function both in cell adhesion and in regulation of transcription. One pool exists at the adherence junction where -catenin can interact with cadherins and, via -catenin, with the actin cytoskeleton. Another pool of -catenin exists in the cytoplasm. Molecules from this pool can bind to LEF-1/TCF and the resulting binary complex can migrate into the nucleus forming a ternary complex with DNA (and thus a third pool of -catenin) that selectively regulates transcription of LEF-1 responsive genes. In the consensus model, the size of the cytoplasmic pool of -catenin determines the extent of binary complex formation and the degree of transcriptional activation. A fourth pool of -catenin associates with APC, which targets the -catenin for degradation by an as yet undetermined mechanism. The Wnt/Wingless signaling pathway regulates the size of the cytoplasmic pool of -catenin. Thus binding of Wingless to Frizzled (or Wnt-1 to a vertebrate frizzled homolog) sends a signal to Disheveled which in turn negatively regulates the kinase ZW3 (or its vertebrate homolog GSK3). Active ZW3/GSK3 kinase binds the APC/-catenin complex, phosphorylates APC, and enhances its association with -catenin (Rubinfeld et al., 1996). Wingless/Wnt signals inhibit ZW3/GSK3 thus permitting dissociation of the APC/-catenin complex and increasing cytoplasmic pools of free -catenin. The cadherins of the adherence junction compete for -catenin binding and thus can also influence the pool of free cytoplasmic -catenin.

View larger version (27K): [in this window] [in a new window]   Fig. 8.   The cadherin-Wnt signaling pathway. The drawing illustrates the concept that both cadherins and the Wnt/Wingless signaling pathway contribute to the control of cell growth and differentiation by regulating the pool of intracellular -catenin/Armadillo. Four interchangeable pools of -catenin/Armadillo are shown: (1) at the cadherin-containing adherence junction; (2) free in the cytoplasm; (3) associated with the Lef-1/TCF DNA binding protein in the nucleus; (4) associated with APC and targeted for degradation. Signaling through the Wnt/Wngless pathway regulates APC--catenin association via the GSK3/ZW3 kinase.

4. Other aspects of cadherin signaling In addition to modulating the Wnt pathway, cadherins have been implicated in other signal transduction events. Cadherin-mediated cell-cell association itself can trigger intracellular tyrosine phosphorylation events (Kinch et al., 1997); the proteins which are tyrosine phosphorylated are quite distinct from those that are phosphorylated in response to integrin-mediated adhesion to the extracellular matrix. Cadherins have been reported to associate with other types of receptors and signaling molecules in the context of the adherence junction. For example, -catenin seems to mediate an association between the EGF-receptor and cadherins (Hoschuetzky et al., 1994), whereas the cytoplasmic kinase FER is associated with p120cas, another structural component of the adherence junction (Kim and Wong, 1995). Shc, an adapter protein implicated in Ras pathway signal transduction, recently has been reported to bind to the cytoplasmic domain of cadherins via the Shc SH2 domain (Xu et al., 1997). A early report that cadherin/catenin complexes interact with RPTPµ, an Ig-CAM family receptor tyrosine phosphatase (Brady-Kalnay et al., 1995), has been challenged recently (Zondag et al., 1996). However, it seems likely that members of other adhesion receptor families will "talk" to cadherins; for example, an integrin-cadherin cross-signaling interaction has been reported recently (Monier-Gavelle and Duband, 1997). Along these lines, an exciting recent study indicates a direct link between integrin signaling and regulation of the cadherin/-catenin pathway. Thus, transfection of epithelial cells with ILK (see Section III.A.3.a. above) results in an epithelial to mesenchymal conversion, with a disruption of cadherin-based adherence junctions and the production of a more motile fibroblastic phenotype (Novak et al., 1998). This seems to result from ILK phosphorylation and inhibition of ZW3/GSK3, leading to accumulation of -catenin and activation of mesenchymal genes responsive to the -catenin/LEF-1 complex. These results offer important new insights into the complex relationships between cadherin-mediated cell-cell interactions and integrin-mediated cell-matrix interactions.

B. Signaling by Immunoglobulin-Cell Adhesion Molecules

1. Neural immunoglobulin-cell adhesion molecules NCAM and L1 are neural Ig-CAM family members that play an important role in neurite outgrowth (Tessier-Lavigne and Goodman, 1996). Several years ago it was established that a signal transduction process involving tyrosine kinases, AA metabolites, and calcium fluxes was responsible for NCAM/L1-mediated growth cone activation and neurite extension (Baldwin et al., 1996). During the past few years two distinct models have been proposed to explain this phenomenon. One model predicates a key role for an fibroblast growth factor (FGF) receptor, suggesting that NCAM (or L1) can interact directly with FGF receptor via a peptide sequence conserved between the two types of molecules, thus activating tyrosine kinase activity and subsequently triggering PLC-mediated calcium responses (Saffell et al., 1997). This model is discussed further in Section IX. of this review concerning adhesion receptor-growth factor receptor interactions.

2. Other signaling by immunoglobulin-cell adhesion molecules A vast literature on the role of Ig-CAMs in signal transduction in the immune system is available. Indeed both the T-cell receptor and the B-cell receptor are members of the Ig-CAM family, as are several other key receptors in lymphoid cells. Because of the specialized nature of this extensive literature we cannot pursue it in the context of this article. For further information on Ig-CAM signaling and the immune response the reader is referred to excellent reviews on these topics (Crabtree and Clipstone, 1994; Weiss and Littman, 1994).

Relatively little is known currently about signaling by members of the selectin family of adhesion receptors; however, this area is developing very rapidly. One very important aspect of selectin signaling concerns the role of selectins as cosignaling agents in leukocyte activation by vascular endothelial cells (Zimmerman et al., 1996b). Thus, at sites of inflammation, endothelial cells both up-regulate surface expression of P-selectin and also express factors [platelet-activating factor (PAF), C-X-C and C-C family chemokines] act through G-protein-coupled receptors to activate 2 integrins on leukocytes. The activating factors such as PAF act in a juxtacrine fashion and require P-selectin-mediated interaction for full effect. This topic is discussed in more detail in Section IX. below, which deals with cooperation between adhesion receptors and other receptors.

Here we will deal with the limited information currently available on direct signaling by selectins or their mucin-type counter receptors. Downey and colleagues provided some of the earliest observation on selectin signaling, showing that engagement of L-selectin contributed to the oxidative burst in neutrophils (Waddell et al., 1994), and later showing that the MAPK pathway in these cells could be triggered through L-selectin (Waddell et al., 1995). In a more detailed analysis, Brenner et al. (1996) showed that antibody engagement of L-selectin in wild-type Jurkat T-lymphoid cells led to tyrosine phosphorylation of the selectin, association with GRB2/SOS, GTP-loading of Ras, activation of MAPK, and a Ras/Rac-dependent oxidative burst; however, in Jurkat cells deficient in p56^lck, a Src family member, none of these events occurred. Thus, L-selectin signaling requires the Lck kinase as an early upstream signaling pathway component. In contrast to integrin signaling, L-selectin signaling to tyrosine kinases and MAPK does not seem to require extensive cross-linking, because use of a secondary antibody does not enhance the effects of primary anti-L-selectin. In contrast, a pathway leading from L-selectin to activation of 2 integrins in T cells seems to require use of a secondary antibody to promote cross-linking, or use of GlyCAM-1, a physiological ligand for L-selectin that is polyvalent (Hwang et al., 1996). The mucin-like counter receptors for selectins also seem to be able to signal. An early study (Celi et al., 1994) showed that adhesion of monocytes to P-selectin could induce the expression of tissue factor mRNA in these cells; the adhesion presumably is mediated through PSGL-1, the P-selectin counter receptor. In a similar vein, the binding of T cells to P-selectin has been reported to trigger tyrosine phosphorylation of FAK and other cellular proteins (Haller et al., 1997). Thus both selectins and transmembrane counter-receptors for selectins (such as PSGL-1) seem to be able to generate various types of signaling events in leukocytes. This area is still in an early phase of development and much remains to be learned about the biochemistry and biological implications of signaling involving the selectin family.

An important new development in adhesion signaling is the identification of the "orphan" tyrosine kinases DDR1 and DDR2 as collagen binding receptor tyrosine kinases (Shrivastava et al., 1997; Vogel et al., 1997). These molecules have an extracellular domain that is related to discoidin 1, a lectin-like adhesion protein from Dictyostelium discoideum, whereas the intracellular domains have typical tyrosine kinase structures. DDR1 and DDR2 bind to native triple helical collagens, whereas collagen-associated carbohydrates also may play a role in binding and stimulation of tyrosine kinase activity. Because integrins (particularly 21) also interact with collagens, these novel kinases may well play some sort of cosignaling role with integrins. Activation of DDR1 and 2 proceeds relatively slowly, does not seem to trigger MAPK, but enhances the expression of matrix metalloproteases and thus may play a role in tumor invasion.

	IX. Adhesion Modulation of Signaling by Soluble Mitogens and Differentiation Factors

Top Previous Next References

It has become increasingly clear that, in addition to generating their own brand of signals, cell adhesion receptors can regulate the cellular response to other extracellular stimuli, such as soluble growth factors and differentiation-inducing agents. Perhaps the best known incarnation of this is anchorage-dependent growth, a phenomenon which has been studied for nearly 30 years (Stoker et al., 1968). Recently many aspects of anchorage control of cell growth have been ascribed to integrin-mediated interactions with ECM proteins, thus placing anchorage dependence in a more biochemical context. At one level, the reason for adhesion-mediated regulation of signal-induced events is simple; not everything a cell can do should happen wherever and whenever a cell happens to be. Thus, the regulation of cellular events, both reversible (e.g., synthesis/secretion of specific proteins) and irreversible (e.g., cell division, differentiation, apoptosis), is through the coordinated effects of positive and negative signals, including ones from soluble factors, as well as ones from the ECM and from adjacent cells. It is this summation of signals conveying both chemical and positional information which tells a cell when the time and place is right to conduct a particular activity. In this section we will deal with cooperation or collaboration between soluble factors and cell adhesion receptors. Once again the emphasis is on integrins, but we will deal with other adhesion receptors as well.

Integrin regulation of signaling by soluble growth factors has been documented in many types of adherent cells. These include endothelial cells (Short et al., 1998), epithelial cells (Streuli, 1993), fibroblasts (Lin et al., 1997b; Miyamoto et al., 1996), keratinocytes (Mainiero et al., 1997), hepatocytes (Liu et al., 1991), chondrocytes (Arner and Tortorella, 1995), myoblasts (Sastry et al., 1996), astrocytes (Cazaubon et al., 1997), and neurons (Schmidt and Kater, 1995).

There are several ways in which integrins and growth factors may cooperate in mediating cellular events (see fig. 9A). For example, there is compelling evidence that ECM and soluble factors can synergize to regulate the intracellular ionic environment. Specifically, whereas both bFGF and adhesion to fibronectin can independently activate the Na⁺/H⁺ antiporter and raise pH_i in endothelial cells, growth factor stimulation of adherent cells is more efficient in the process (Ingber et al., 1990). Similarly, in fibroblasts, PKC-dependent activation of the antiporter (and elevation of pH_i) by PDGF requires cell adhesion (Schwartz and Lechene, 1992). Also, PDGF-stimulated Ca²⁺ mobilization, another ion transient required for cell cycle progression in murine fibroblasts, does not occur in cells in suspension but readily occurs in cells adherent to fibronectin (Tucker et al., 1990). The mechanism underlying this regulation is among the most intellectually satisfying in the field. In adherent cells, PDGF stimulates tyrosine phosphorylation and activation of PLC, which hydrolyzes PI(4,5)P₂ into DAG and IP₃, which in turn activate PKC and increase Ca²⁺_i. In nonadherent cells, the activation of PLC by PDGF still occurs, but there is a dramatic decrease in the level of its substrate, PI(4,5)P₂, and therefore no appreciable generation of DAG and IP₃ can occur (McNamee et al., 1993). Presumably, the adhesion-dependent synthesis of PI(4,5)P₂ is largely caused by the activation of PI4-5K activity by Rac and Rho (as discussed in Section VI.).

View larger version (42K): [in this window] [in a new window]   Fig. 9.   Collaboration between cell adhesion molecules and soluble factors. (A) Mechanisms for collaboration between integrins and soluble factors. Integrin-mediated adhesion and signaling may down-regulate a factor that inhibits receptor signaling [e.g., a protein tyrosine phosphatase (PTPase)] or may provide a necessary substrate (e.g., PIP2) or cofactor for a growth factor-mediated activity (e.g., PLC). Also, integrin-mediated adhesion may establish "signaling scaffolds" that juxtapose cytoskeleton-associated coactivators (X) with growth factor-regulated activities (e.g., Ras/Raf), or which physically interact with growth factor receptors themselves. (B) Signal transduction events stimulated by N-cadherin, L1, and NCAM leading to neurite outgrowth. Homophilically-bound N-cadherin, L1, or NCAM (CAM) molecules are postulated to recruit the FGF receptor (FGFR) via binding motifs to the CAM homology domain (filled triangle). FGFR recruitment causes autophosphorylation of the FGFR at Tyr766 in the cytoplasmic domain creating a binding site for the SH2 domain of phospholipase C gamma (PLC). Activation of PLC triggers a cascade through the utilization of phosphatidylinositol 4,5-bisphosphate (PIP2) to yield diacylglycerol (DAG) and inositol trisphospate (IP3), conversation of DAG to arachidonic acid (AA) culminates in the AA-dependent influx of calcium ions through N- and L-type calcium channels. Diagram adapted from Doherty and Walsh, 1996. (C) Collaboration between P-selectin and platelet-activating factor (PAF) in the activation of leukocytes. Upon activation of endothelial cells P-selectin and PAF are expressed rapidly on the endothelial cell surface. P-selectin is capable of binding carbohydrate moieties such as PSGL-1 on the surface of leukocytes, such as neutrophils. Tethering of neutrophils to the vascular vessel wall promotes binding of PAF with its receptor on the neutrophil membrane. Resultant activation of neutrophils leads to increases in intracellular calcium, cell polarization and activation of 2 integrins by an inside-out signaling mechanism. Diagram adapted from Zimmerman et al. (1996).

Perhaps the most direct mechanism whereby integrins can collaborate with soluble factors is the physical interaction of integrins and associated proteins with one or more components of a given growth factor signaling cascade. Assembly of these "signaling scaffolds" would localize, or rather juxtapose normally diffusible elements of the signaling cascade and thereby allow transduction of the signal to occur more efficiently. Receptor-mediated signals are generated at the membrane and propagated through the cytoplasm to their appropriate targets. Within focal adhesions, integrins physically bridge the ECM to the network of cytoplasmic actin microfilaments, a situation that seems particularly well suited for providing an appropriate molecular scaffold for signaling components, as discussed in Section III. above. One of the most direct examples of physical coupling of adhesion and growth factor pathways is the insulin-stimulated association of v3 with the IRS-1 (Vuori and Ruoslahti, 1994), and in a recent report, with the insulin receptor itself (Schneller et al., 1997). IRS-1 is heavily tyrosine phosphorylated after insulin stimulation and is required for the transduction of signals initiated at the insulin receptor. Significantly, insulin-stimulated DNA synthesis in v3-expressing cells is enhanced when the cells are plated on vitronectin (an v3 ligand) versus other substrates, whereas no enhancement is seen in cells expressing the v5 vitronectin receptor (Vuori and Ruoslahti, 1994). There are other examples of scaffolding occurring directly at the level of the growth factor receptor. Thus the EGF receptor directly associates with microfilaments, and its close relative, p185^neu, is found in a microfilament- and membrane-associated, high-molecular weight glycoprotein complex that contains elements of several signal transduction cascades, including the Ras-MAPK pathway (Carraway and Carraway, 1995). In epithelial cells, the association of the EGFR with the actin cytoskeleton appears to be regulated by cell density, being far less stable in subconfluent versus confluent cells, and this affects the efficiency of EGF-induced signal transduction (Bedrin et al., 1997). Recently, it has been reported that a highly tyrosine-phosphorylated subfraction of PDGF receptor from stimulated cells coimmunoprecipitates with the v3 integrin, and significantly, engagement of v3 by adhesion to vitronectin can enhance PDGF-induced mitogenesis (Schneller et al., 1997).

In another report of integrin-mediated scaffolding, cells incubated with microbeads coated with integrin ligands (fibronectin or RGD peptides) form focal contact-like structures at the cell-bead interface, and these structures are enriched for several signaling molecules, including Src, FAK, PLC, PI-3K, the Na⁺/H⁺ antiporter, and the high-affinity FGF receptor flg (Plopper and Ingber, 1993; Plopper et al., 1995). Significantly, the bead-associated complexes, after isolation, retained tyrosine kinase and inositol lipid-metabolizing activities. As discussed in Section V. above, other experiments, using ligand- or anti-integrin antibody-coated beads, have identified more than 20 signaling molecules associated with focal contacts (Miyamoto et al., 1995b, 1996), including several Src family kinases, the PTP-1D tyrosine phosphatase, PLC, and PI-3K, as well as Rho family GTPases Rac and Rho, several components of the MAPK and JNK cascades, and most recently, receptors for bFGF, PDGF, and EGF. The association of RTKs with integrin/focal adhesion complexes is transient (lasting approximately 60 min after bead addition) and independent of growth factor binding, but it seems to require both integrin aggregation and occupancy (Miyamoto et al., 1996). Integrin engagement apparently is sufficient to induce ligand-independent RTK tyrosine phosphorylation in some systems (Sundberg and Rubin, 1996), but not in others (Lin et al., 1997a; Miyamoto et al., 1996).

Recent findings hint at several different mechanisms whereby integrins may regulate growth factor signal transduction. Several reports have described cooperation between adhesion and growth factors in the activation of MAPK (Cybulsky and McTavish, 1997; Inoue et al., 1996; Lin et al., 1997b; Miyamoto et al., 1996; Renshaw et al., 1997). In most instances, growth factor stimulation of nonadherent cells or cells attached nonspecifically to polylysine or plastic has little or no effect on MAPK activity [but not so in some systems (Hotchin and Hall, 1995)], and in all cases, engagement of integrins through adhesion to ligand-coated substrata or beads significantly enhances growth factor activation of MAPK. The adhesion dependence of growth factor activation of MAPK is ablated in cells transformed by viral oncogenes, including v-ras, v-src, and v-mos (Inoue et al., 1996). Dissection of the pathway between growth factor receptors and MAPK suggests that in nonadherent fibroblasts, the early stages of growth factor signaling, including RTK autophosphorylation (Chen et al., 1996b) and Ras GTP-loading, are intact, whereas the activation of downstream kinases (i.e., Raf, MEK, and MAPK) is attenuated significantly, with Raf activation being particularly affected (Lin et al., 1997b). Observations in a recent report partially reiterate these results, but suggest that attenuation of mitogenic MAPK activation in suspended cells may occur at the level of MEK (Renshaw et al., 1997). However, effects of adherence also have been observed on receptor autophosphorylation in fibroblasts and epithelial cells (Cybulsky and McTavish, 1997; Miyamoto et al., 1996). In both cases, integrin engagement through either adhesion to ligand-coated plates (Cybulsky and McTavish, 1997) or incubation with antibody- or ligand-coated beads (Miyamoto et al., 1996) was shown to synergize with growth factors to allow efficient tyrosine phosphorylation of the growth factor receptor. Thus, there is convincing evidence in several cell types that integrin-mediated cell adhesion can increase the efficiency of growth factor signaling to the MAPK pathway. Whether the locus of adhesion modulation of growth factor signaling is at the level of the receptor tyrosine kinase or at a downstream step may vary among cell types or be influenced by experimental conditions.

C. Signal Modulation by Other Cell Adhesion Molecules

1. Neuronal cell adhesion molecules Neuronal axon guidance and outgrowth involve the integration of signals emanating from growth factors and guidance molecules, and from interactions of cell adhesion receptors with ligands on surrounding cells and the ECM. As mentioned in Section VIII., there are currently two theories concerning the mechanism of signaling by neuronal adhesion molecules involved in neurite outgrowth. One version suggests direct intracytoplasmic interactions between the cytoplasmic domains of adhesion receptors such as NCAM and Src-family kinases. Another theory, now also supported by compelling data, is that in the neuronal growth cone, homotypic cell adhesion receptors such as N-cadherin, NCAM and L1 promote axon extension through their ability to modulate the fibroblast growth factor receptor (FGFR) (reviewed in Doherty and Walsh, 1996).