Genetic Analyses of Integrin Signaling

Abstract

The development of multicellular organisms, as well as maintenance of organ architecture and function, requires robust regulation of cell fates. This is in part achieved by conserved signaling pathways through which cells process extracellular information and translate this information into changes in proliferation, differentiation, migration, and cell shape. Gene deletion studies in higher eukaryotes have assigned critical roles for components of the extracellular matrix (ECM) and their cellular receptors in a vast number of developmental processes, indicating that a large proportion of this signaling is regulated by cell-ECM interactions. In addition, genetic alterations in components of this signaling axis play causative roles in several human diseases. This review will discuss what genetic analyses in mice and lower organisms have taught us about adhesion signaling in development and disease.

Integrins anchor cells to the extracellular matrix but also transmit “inside-out” and “outside-in” signals. These are relayed by proteins such as talin and kindlin, mutations which lead to disorders such as Kindler syndrome.

Almost all cells in multicellular organisms are surrounded by a three-dimensional organized meshwork of macromolecules that constitute the extracellular matrix (ECM). The ECM is a dynamic structure that is generated and constantly remodeled by cells that secrete and manipulate its components into a precise configuration. It functions as a structural framework that provides cells with positional and environmental information, but also forms specialized structures such as cartilage, tendons, basement membranes (BM), bone, and teeth. In addition to its structural properties, the ECM acts as a signaling platform that regulates a large number of cellular functions. It is capable of binding growth factors, chemokines, and cytokines thereby modulating their bioavailability and activity. On the other hand, the ECM is recognized by multiple cell surface receptors that transmit information from the extracellular environment by propagating intracellular signals (for a review, see Hynes 2009).

The major cell surface receptors that recognize and assemble the ECM are integrins. Integrins are heterodimeric transmembrane proteins composed of α and β subunits. Eighteen α subunits and eight β subunits can assemble in 24 different combinations with overlapping substrate specificity and cell-type-specific expression patterns (Hynes 2002; Humphries et al. 2006). This, together with the ability of different heterodimers to assemble specific intracellular signaling complexes, provides multiple layers of signaling specificity to these receptors. Conversely, the integrin expression profile of a given cell type determines which ECM components it can bind. Signals arising from integrins regulate virtually all aspects of cell behavior, including cell migration, survival, cell cycle progression, and differentiation.

Genetics has proven to be a powerful tool to dissect the functions of ECM–cell interactions in complex organisms. To date, all of the integrin subunits and their major ligands have been deleted in mice. Given the large variety of cellular processes regulated by adhesion signaling, it is not surprising that a significant subset of these proteins has proven to be essential for embryonic development and/or tissue maintenance. However, in addition to underlining the importance of cell-ECM interactions in development, genetic studies also revealed critical roles for tissue- and cell-type-specific modes of adhesion signaling and provided important insights into human disease.

THE INTEGRIN-INTERMEDIATE FILAMENT AXIS

Basement membranes (BMs) are dense sheets of extracellular matrix that function as structural barriers that separate epithelial and endothelial cells as well as peripheral nerve axons, fat cells, and muscle cells from the underlying tissue stroma. BMs provide structural support, separate tissues into compartments, and regulate cell behavior. All cell types are known to produce components of BMs, which include type IV collagen, laminin, fibronectin, heparan sulfate proteoglycans, and nidogen/entactin as well as several minor components (Erickson and Couchman 2000; Yurchenco 2010). The molecular composition of the BM varies among different tissues, conferring signaling specificity important for defining the specialized functions of epithelial and endothelial cells in different organs.

All BMs contain laminins, a family of 16 heterotrimeric glycoproteins generated by the combination of 5α, 4β, and 3γ chains. When present in sufficient concentrations, laminin networks can self-assemble into polymers that interact with other ECM components, as well as cell surface receptors such as integrins and dystroglycan. The laminin isoforms present in BMs are regulated in a tissue-specific manner, as well as temporally during development (Miner 2008; Yurchenco 2010). The central role of laminins in BM assembly is illustrated by gene deletion studies in mice. Laminin-111 (containing α1, β1, and γ1 chains) and laminin-511 (containing α5, β1, and γ1 chains) are the central constituents of peri-implantation stage BMs. Mice deficient in γ1 or β1 subunits are unable to generate laminin trimers, and therefore lack BMs (Smyth et al. 1999; Miner et al. 2004).

Laminin-332 (containing α3, β3, and γ2 chains; previously termed laminin-5) is a component of epithelial BMs and thus present in skin, stratified squamous mucosa, amnion, and cornea. Its main task is to maintain epithelial integrity and epithelial-mesenchymal cohesion in tissues exposed to high mechanical forces. This function is facilitated by the unique ability of laminin-322 to interact with two distinct integrin heterodimers. Its interaction with α3β1 integrin results in the assembly of canonical focal adhesions (FA), whereas the interaction with α6β4 integrin results in the formation of specialized adhesion complexes termed hemidesmosomes (Tsuruta et al. 2008).

Hemidesmosomes—Structure and Assembly

Hemidesmosomes were first characterized at the ultrastructural level as electron-dense clusters at the plasma membrane-ECM interphase (Weiss and Ferris 1954). Further studies identified them as multiprotein adhesion complexes present in stratified and simple epithelia. Two types of hemidesmosomes can be distinguished based on their molecular composition. Type I or classical hemidesmosomes are found in stratified epithelium such as the skin and contain three transmembrane proteins: α6β4 integrin, tetraspanin CD151, and type XVII collagen (also called bullous pemphigoid [BP] antigen 180) (Fig. 1). Type II hemidesmosomes are found in simple epithelia such as intestine, and contain only integrin α6β4 (Uematsu et al. 1994; reviewed in Litjens et al. 2006).

Figure 1.

Molecular architecture of type I hemidesmosomes. Schematic depiction of a type I hemidesmosome found in stratified epithelia such as in basal skin keratinocytes. The core component is α6β4 integrin, which binds the basement membrane (BM) component laminin (LM)-322. α6β4 integrin recruits the plakin protein plectin through multiple interactions with the β4 integrin cytoplasmic tail, which initiates the formation of hemidesmosomes. This is followed by the recruitment of collagen XVII, which interacts both with β4 integrin and plectin as well as with LM 322. Collagen XVII in turn mediates the recruitment of another plakin, bullous pemphigoid antigen 230 (BP 230), which together with plectin provides the connection to intermediate filaments (IF). This linkage is essential to stabilize the hemidesmosome and to provide stable adhesion of the basal keratinocyte to the BM. Also, the transmembrane protein tetraspanin CD151 that interacts with α6 integrin is found in type I hemidesmosomes. Phosphorylation of serines (S) 1356, 1360, and 1364 on the cytoplasmic tail of β4 integrin by growth factors induces disassembly of hemidesmosomes, which promotes cell migration and signaling. The molecules are not drawn to scale.

The unique feature of hemidesmosomes is that they connect the ECM to the intermediate filament (IF) network. This interaction is established by two plakin proteins, plectins and BP230 (also called BPAG1), of which plectin is present both in type I and II hemidesmosomes. Plectins are large cytoplasmic proteins, which at their carboxyl terminus contain six plakin repeats. A stretch of basic residues linking the fourth and fifth plakin repeat mediates the interaction of plectin with IFs. The amino terminus contains two calponin-homology (CH) domains that constitute an actin-binding domain. Through the actin-binding domain, plectins can associate with either the cytoplasmic domain of the β4 integrin subunit or actin filaments in a mutually exclusive manner, which explains the absence of actin in hemidesmosomes (Rezniczek et al. 1998; Geerts et al. 1999; Litjens et al. 2003). The interaction of plectin with the cytoplasmic tail of β4 integrin is considered as the initial step of hemidesmosome assembly (Geerts et al. 1999; Koster et al. 2004). The cytoplasmic tail of β4 integrin is unusually large and shares no homology with other β integrin subunits. Its interaction with plectin has been shown to induce a conformational change in the integrin tail (de Pereda et al. 2009). Because the recruitment of type XVII collagen and the plakin BP230 to hemidesmosomes requires prior interaction of plectin with β4, this conformational change might facilitate the interactions of both proteins with the integrin (Koster et al. 2004).

Deletion of α6β4 leads to the loss of hemidesmosomes and epithelial detachment in mice (Dowling et al. 1996; van der Neut et al. 1996), indicating that BP180, despite its ability to bind laminin-332 and plectin in vitro, is not sufficient to maintain adhesion of cells to laminin-332 in the absence of α6β4 (Tasanen et al. 2004) and identifying α6β4 as the core component of hemidesmosomes. Another critical event in hemidesmosome assembly is the interaction between β4 integrin tails and plectin. Blocking this interaction compromises hemidesmosome assembly in vitro (Geerts et al. 1999; Koster et al. 2001). In plectin-deficient mice hemidesmosomes appear ultrastructurally normal, but their number and mechanical stability are reduced (Andra et al. 1997). The fact that type II hemidesmosomes lack both type XVII collagen and BP230 further shows that α6β4 and plectin alone are sufficient to initiate the formation of these structures and to maintain their integrity.

Although α6β4 is the only integrin found in hemidesmosomes, other integrins can indirectly contribute to their assembly. Integrin α3β1 has been shown to cluster in “pre-hemidesmosomal” structures on the basal cell surface of human keratinocytes together with CD151. However, as hemidesmosomes mature, α3β1 integrins become recruited to cell-cell or focal contacts, whereas CD151 remains in the hemidesmosomes (Sterk et al. 2000). The initial α3β1-containing adhesions have been proposed to contribute to the recruitment of α6β4 to the plasma membrane, and thereby increase the efficacy with which hemidesmosomes are formed. This is supported by the observation that β1-deficient mice have reduced numbers of hemidesmosomes in the skin, although it should be noted that the expression level of α6β4 is also reduced in these mice, suggesting that the mechanism could also be more indirect (Brakebusch et al. 2000). Polarity proteins have also been shown to play a role in hemidesmosome assembly through the regulation of polarized targeting of integrins. In the larval epidermis of the zebrafish, the basolateral identity protein Lgl2 is required for hemidesmosome formation through recruitment of β4 integrin to the basal plasma membrane (Sonawane et al. 2005, 2009). Taken together, the formation of hemidesmosomes requires the establishment of a basolateral membrane identity through polarity proteins and possibly also other integrin-based adhesion structures to recruit α6β4 integrin to these sites. α6β4 integrin then interacts with BM laminin and together with plectin nucleates stable, IF-bound hemidesmosomes. Interestingly, however, keratinocytes expressing a β4 integrin subunit unable to bind laminin can still form structures that contain all components of type I hemidesmosomes, which implies that ligand binding is not necessary for their formation (Nievers et al. 1998, 2000). These complexes, however, differ in their morphology, density, and dynamics from normal hemidesmosomes, suggesting that the integrin-laminin interaction plays a role in the stability and structural organization of hemidesmosomes (Geuijen and Sonnenberg 2002).

Hemidesmosomes—Guardians of Epidermal Integrity

A central function of the skin is to act as a bidirectional barrier to prevent dehydration and to protect against injury and pathogens. In addition, it is exposed to constant mechanical stress. On the other hand, this tissue is continuously renewed through a terminal differentiation program, in which basal keratinocytes detach from the BM and move to the upper layers of the epithelium where they slough off. Finally, in case of injury, keratinocytes need to rapidly acquire a motile phenotype and migrate to close the wound. These functions require mechanically stable but dynamic modes of cell adhesion. Hemidesmosomes are found in the basal plasma membrane of basal keratinocytes, where they function to attach these cells firmly to the BM separating the epidermis from the dermis (Fig. 1). Interestingly, α6β4 integrins and thus also hemidesmosomes are not required for skin morphogenesis or developmental homeostasis (DiPersio et al. 2000). The importance of hemidesmosomes in maintaining epidermal integrity is, however, shown by multiple lines of genetic evidence. Ablation of genes encoding α6 integrin, β4 integrin, or plectin in mice results in severe blistering of the skin causing neonatal death because of a severe epithelial barrier defect (Dowling et al. 1996; Georges-Labouesse et al. 1996a; van der Neut et al. 1996; Andra et al. 1997). In line with their dispensable role in hemidesmosome assembly, mice lacking type XVII collagen or BP230 display only mild forms of skin blistering (Guo et al. 1995; Nishie et al. 2007), whereas deletion of CD151 is dispensable for both hemidesmosome stability and the integrity of skin (Wright et al. 2004).

No experimental data exist on the dynamic behavior of hemidesmosomes in vivo, but it has been analyzed in keratinocyte culture, in which hemidesmosomes have been shown to display a certain dynamics also in nonmigratory cells (Tsuruta et al. 2003). When cells are stimulated to migrate, α6β4 integrin is mobilized from hemidesmosomes to actin-based structures such as lamellipodia and filopodia to promote cell motility, resulting in hemidesmosome disassembly (Geuijen and Sonnenberg 2002; Tsuruta et al. 2003). This type of relocalization has also been visualized in human wound edges, where β4 integrins can be seen in trailing-edge hemidesmosomes as well as in lamellipodia of the leading edge (Underwood et al. 2009). In vitro analyses have shown that the mobilization of α6β4 integrin from hemidesmosomes is facilitated by growth factor-mediated phosphorylation of multiple residues in the cytoplasmic tail of β4. Both serines and tyrosines can be phosphorylated, but the exact sites and their relevance remain controversial. The current data suggest that serine phosphorylation might be more relevant under physiological conditions, leading to hemidesmosome disassembly through the release of plectin from the integrin (Margadant et al. 2008) (Fig. 1)

α6β4 Integrin and Cancer

β4 integrin was originally identified as a “tumor-specific” protein (Tumor Surface Protein 180) up-regulated in metastatic variants of mouse lung carcinoma and melanoma cell lines (Kennel et al. 1981), and has later been shown to be up-regulated in several human cancers, suggesting that the expression of β4 integrin is beneficial for tumor cells (Giancotti 2007). The role of α6β4 integrin signaling in tumorigenesis has been studied in mice carrying a truncated cytoplasmic tail of β4 lacking the tyrosine phosphorylation sites as well as other potential interaction motifs. In contrast to mice lacking the entire β4 cytoplasmic domain, which display extensive skin blistering because of the absence of hemidesmosomes (Murgia et al. 1998), mice with a more restricted truncation do not show these defects. However, they show delayed wound healing and defective neoangiogenesis in tumor xenografts (Nikolopoulos et al. 2004, 2005). In addition, when crossed to the MMTV-Neu mice that carry an activated form of epidermal growth factor receptor-2 (ErbB2) driven by a mouse mammary tumor virus (MMTV) promoter leading to mammary tumors, the β4 mutant mice display delayed tumor onset, impaired tumor growth and decreased metastatic potential (Guo et al. 2006). Whether this applies to other tumor types as well remains to be shown.

The tumor-promoting properties of α6β4 integrin seem to result from both promigratory and signaling functions. β4 integrin has been shown to cross-talk with several receptor tyrosine kinases (RTK), including ErbB2, epidermal growth factor receptor (EGF-R), Met and Ron (Mariotti et al. 2001; Trusolino et al. 2001; Santoro et al. 2003; Guo et al. 2006). Following stimulation, these RTKs induce activation of phosphatidylinositol 3-kinase (PI3-K) or Src family kinases (SFK), leading to phosphorylation of the β4 integrin tail, disassembly of hemidesmosomes, and induction of cell motility (Mariotti et al. 2001; Santoro et al. 2003). On the other hand, β4 integrin has been shown to promote SFK-dependent phosphorylation of the catalytic sites of RTKs and their substrates (Bertotti et al. 2006; Guo et al. 2006), thereby amplifying their signaling ability. Taken together, α6β4 integrin seems to have a proinvasive and proangiogenic activity in tumors, but additional genetic studies are needed to determine whether this role is limited to malignancies involving hyperactivation of specific RTKs. In addition, several groups have failed to detect significant tyrosine phosphorylation of β4 integrin in keratinocytes or transformed cell lines in response to stimulation with EGF and hemidesmosome disassembly, making the relevance of tyrosine phosphorylation events controversial (Rabinovitz et al. 1999, 2004; Alt et al. 2001; Wilhelmsen et al. 2007). Knockin mice carrying mutations in specific phosphorylation sites of the β4 integrin subunit would provide more conclusive information on the in vivo significance of the various phosphorylation events in α6β4 integrin signaling, hemidesmosome turnover, and invasion.

THE INTEGRIN-ACTIN AXIS

In contrast to α6β4, most integrins engage the actin cytoskeleton following ligand binding. A central ligand of actin-associated integrins is FN, which is recognized by 11 integrin heterodimers in mice and humans, and it will be used in this section as an example to discuss the principles of ECM-integrin interactions leading to engagement and remodeling of the actin cytoskeleton. FN is a large, modular glycoprotein that exists in two forms; cellular FN which is present in tissues where it is assembled into a fibrillar matrix, and plasma FN, which is produced by hepatocytes and is secreted into the blood where it remains in a nonfibrillar, soluble form, or following entry into tissues becomes incorporated into the ECM (Leiss et al. 2008). FN is found only in vertebrates, and it has co-evolved together with the cardiovascular system. Consistently, FN is critical for the development of the vasculature, where it localizes between the endothelium and perivascular cells. Deletion of FN in mice results in embryonic lethal cardiovascular defects, which vary depending on the genetic background. FN-null embryos from 129S4 mice are unable to form the dorsal aorta, indicative of an early defect already in vasculogenesis, whereas this structure is present in C57BL/6 derived embryos that display defects in vascular lumen formation (George et al. 1993, 1997; Georges-Labouesse et al. 1996b).

FN is secreted as a disulfide-bonded dimer, and its deposition into a fibrillar matrix is a cell-driven process that critically depends on integrins. So far α5β1, αvβ3, α4β1, and αIIbβ3 integrins have been shown to induce fibrillogenesis in vitro. These integrins bind FN that is secreted as a compact globular structure where binding sites for other FN molecules are buried within the protein (see Schwarzbauer and DeSimone 2011). Binding is followed by the activation of integrin signaling, leading to the recruitment of cytoplasmic proteins to form focal complexes (FCs). Several FC components are actin-binding and modulatory proteins, allowing the recruitment and reorganization of the actin cytoskeleton at these sites and their maturation into FAs (Geiger and Yamada 2011). The major FN-fibril-forming integrin α5β1 leaves FAs and moves along F-actin to the cell center to form fibrillar adhesions and to facilitate the generation of mechanical tension via the actin cytoskeleton, leading to stretching of the bound FN molecule, unraveling of the cryptic self-association sites, and finally the binding to other FN molecules resulting in fibril formation (Mao and Schwarzbauer 2005; Leiss et al. 2008; Schwarzbauer and DeSimone 2011).

Interactions via the RGD Motif

Cell adhesion to FN critically depends on the Arg-Gly-Asp (RGD) motif of FN, which is recognized by α5β1, the αv family, α8β1, α9β1, and the platelet-specific αIIbβ3 integrins (Fig. 2). α5β1 is considered the major integrin responsible for FN assembly, and its interaction with the RGD motif is required for this function. However, although deletion of α5 integrin in mice also leads to embryonic lethality and vascular defects, these mice develop significantly further than the FN-null mice (Yang et al. 1993). This is attributable to the ability of cells to assemble FN using other integrins, mainly of the αv subfamily. The deletion of all five αv heterodimers also leads to late embryonic defects in vascular development mainly in the placenta. A subset of animals can even proceed through embryonic development and die after birth because of hemorrhages, indicating that, unlike α5, αv integrins are dispensable for vasculogenesis and partly also for angiogenesis (Bader et al. 1998). Only a double knockout of the αv and α5 integrin genes results in a loss of FN fibrillogenesis (Yang et al. 1999). Interestingly, although the RGD motif is central for the interaction of FN with α5β1 and αvβ3, the inactivation of this motif by a RGD to RGE point mutation also allows FN fibrillogenesis in vivo. The knockin mice carrying this mutation display a phenotype closely resembling that of the α5-null mice (Takahashi et al. 2007). This highlights the importance of this motif in FN signaling through α5β1, but also shows the ability of other integrin interaction sites on FN to take over the role of RGD in FN fibrillogenesis. Whether these sites play a role in FN assembly when the RGD sequence is intact, remains open. Together, these results indicate that FN fibrillogenesis is not sufficient for it to carry out its functions during development. They further illustrate that the majority of FN signaling in embryogenesis occurs through α5β1 and that αv and α5β1 integrins relay distinct signals on FN binding. The severity of the FN-null phenotype in contrast to the integrin deletions also suggests that the functions of FN during development extend beyond the FN-integrin signaling axis. These functions could be related to its mechanical role in the ECM, or its ability to induce integrin-independent signaling, for example, through sequestering growth factors such as vascular endothelial growth factor (VEGF) or transforming growth factor-β (TGF-β).

Figure 2.

Developmental functions of fibronectin-integrin interactions. Fibronectin (FN) is a dimeric glycyoprotein consisting of type I, type II, and type III modular repeats. Dimerization is achieved by disulfide bonds mediated by the two cysteines (S) in the COOH-terminus of the protein. The binding domains and motifs for integrins, as well as mouse developmental functions reported to depend on the particular interaction are indicated. Integrins that have been shown to mediate FN fibrillogenesis in vitro are marked by gray boxes.

Non-RGD Binding Integrins

The FN gene can be alternatively spliced allowing the expression of up to 20 possible monomeric isoforms in humans and up to 12 in mouse, potentially giving rise to a larger variety of FN dimers (Pankov and Yamada 2002). Some of these splice variants can generate additional integrin interaction sites on FN. One of them is the variable region (v-region) that can interact with two non-RGD-binding integrins, α4β1 and α4β7 (Wayner et al. 1989; Guan and Hynes 1990; Fig. 2). These integrins are mainly expressed by hematopoietic cells, although α4β1 is also found in several other cell types, such as neural crest cells and cellular components of the cardiovascular and peripheral nervous systems. The two α4 integrins have many important roles during development. Deletion of the α4 integrin gene results in embryonic lethality because of early placental and cardiovascular defects (Yang et al. 1995). However, these defects are probably not attributed to the FN- α4β1/α4β7 integrin interaction, as these two integrins also bind cell counter receptors such as vascular-cell adhesion molecule-1 (VCAM-1) and MadCAM, respectively. The developmental defects in the α4 integrin-null mice are very similar to those found in mice lacking VCAM-1, indicating that the observed defects are caused by an abrogated interaction between α4β1 integrin and VCAM-1 (Kwee et al. 1995; Yang et al. 1995). From a clinical point of view, it is particularly interesting that the interaction between α4β1 and VCAM-1 is important for the homing of autoreactive T cells to the central nervous system during the pathogenesis of an autoimmune disease called multiple sclerosis (Yednock et al. 1992; Vajkoczy et al. 2001; Bauer et al. 2009). This disease is characterized by axon demyelination and damage, leading to numbness, weakness, paresis as well as cognitive problems. Blocking antibodies against α4 integrin have proven to be an effective treatment of multiple sclerosis (Steinman 2009), and genetic analysis in mice showed that α4β1 is required for the arrest of T cells on the brain endothelial surface (Bauer et al. 2009).

α4β1 integrins can also interact with the alternative spliced extra domain A (EDA) of FN (Fig. 2), which is highly expressed during development and during pathological conditions such as tumorigenesis. This “reactivation” of the embryonic splice pattern is interesting, as α4β1 has recently been shown to play a critical role in tumor lymphangiogenesis in a FN-dependent manner. Both α4β1 and FN are highly expressed in lymphatic endothelial cells of tumors, whereas other integrin subunits seem not to be up-regulated (Garmy-Susini et al. 2010). Whether this function really depends on the EDA domain was, however, not analyzed. In addition, it is not clear whether α4β1 integrin is also regulating developmental lymphangiogenesis, as the constitutive knockout mice die before the initiation of this process. The EDA domain itself, however, plays a role in developmental lymphangiogenesis through integrin α9β1. Interestingly, this integrin is highly expressed in lymphatic but not blood endothelial cells (Huang et al. 2000). Deletion of α9 integrin gene leads to a severe chylothorax caused by defects in lymphatic valve morphogenesis and subsequent respiratory failure (Huang et al. 2000; Bazigou et al. 2009). This phenotype was partially recapitulated by deletion of EDA. Furthermore, both α9β1 integrin as well as the EDA domain were shown to be critical for FN assembly in lymphatic endothelial cells in vitro (Bazigou et al. 2009), in contrast with previous studies showing that EDA is not required for FN assembly or mouse development because of compensation by EDB (Tan et al. 2004; Astrof et al. 2007). This study suggests that matrix assembly might be regulated by integrins that are expressed in a tissue-specific manner to allow more stringent control of distinct anatomical structures.

Linkage to the Actin Cytoskeleton

In addition to catalyzing matrix assembly, the binding of integrins to FN or other ligands leads to the induction of multiple modes of intracellular signaling. These include activation of classical signaling pathways through tyrosine and serine phosphorylation of specific substrates, resulting ultimately in the regulation of cell survival, growth and differentiation. This signaling operates in close collaboration with growth factor signaling, making it difficult to dissect which signals specifically emanate from ligated integrins (Legate et al. 2009; Ivaska and Heino 2010). The other central signaling mode is the recruitment of filamentous (F-) actin to integrin adhesion sites, a process tightly coupled to the active remodeling of the actin cytoskeleton. The integrin-actin linkage is central to integrin function in several respects. First, active actin polymerization, crosslinking, and subsequent force generation are critical for the maturation of nascent and short-lived integrin adhesions into signaling-competent and stable FAs. Second, it is required for the precise spatiotemporal control of cell protrusion and retraction during cell migration. Third, it allows cells to adopt or change their characteristic shape, for example to polarize. Finally, it is critical for FN matrix assembly. Integrins do not bind actin directly, but regulate this linkage by recruiting a large number of actin-binding and regulatory proteins. A recent study based on database mining identified 156 signaling, structural and adaptor molecules that can be found in integrin adhesions (Zaidel-Bar et al. 2007). Only a small subset of these proteins binds directly to integrins, and the large majority is recruited through protein-protein interactions between the various scaffold proteins (for reviews, see Legate and Fässler 2009; Geiger and Yamada 2011). A large body of structural, biochemical, and genetic evidence points to talin, kindlin, and the integrin-linked kinase (ILK)-pinch-parvin (IPP) complex as some of the central FA components regulating the integrin-actin linkage. All of these proteins are ubiquitously expressed and regulate a wide range of integrin heterodimers. Deletion of any of these proteins leads to early embryonic lethality, which results from compromised integrin function and defects in their connection to the cytoskeleton. On the cell-biological level these defects span the entire range of functions attributed to the integrin-actin linkage. Hence, they can be viewed as global regulators of integrin function. The precise functions of these proteins will be discussed in the next section.

IN VIVO MECHANISMS OF INTEGRIN SIGNALING THROUGH INTRACELLULAR EFFECTORS AND SCAFFOLDS

Talin and kindlin are two FA proteins that directly bind integrins and thereby are involved in the very early events of integrin signaling. They are distinguished from the large group of FA proteins by their ability to regulate both integrin activation (inside-out signaling) and intracellular signaling downstream of ligand binding (outside-in signaling). Talin is present in all multicellular eukaryotes; lower eukaryotes encode only a single talin isoform, whereas vertebrates have two talin isoforms, talin-1 and talin-2. Talin consists of a large carboxy-terminal rod and an amino-terminal head domain composed of a FERM domain with four subdomains: F0, F1, F2, and F3. The F3 subdomain contains a phosphotyrosine-binding motif, which harbors a high affinity binding site for β-integrin tails. The rod domain contains multiple binding sites for the actin–binding protein vinculin as well as for actin itself. In addition it contains a second integrin-binding site (Critchley and Gingras 2008). The talin head has been shown to be sufficient for integrin binding and activation, whereas the rod is required for the scaffolding function of talin in outside-in signaling (Garcia-Alvarez et al. 2003; Tanentzapf and Brown 2006; Tanentzapf et al. 2006).

Kindlins are an evolutionarily conserved family of multidomain proteins, which in mammals consists of three members: kindlin-1 (also known as FERMT1), kindlin-2 (also known as FERMT2 or MIG-2), and kindlin-3 (also known as FERMT3 or URP2). Although encoded by separate genes, they show identical domain architecture and high sequence similarity (Ussar et al. 2006). Like talin, kindlins contain a FERM domain through which they interact with β1, β2, and β3 integrins, but this FERM domain is unique in that it is interrupted by a pleckstrin homology (PH) domain, which provides a putative binding motif for membrane lipids.

Consequences of Loss of Integrin Inside-Out Signaling

Integrins are present on the plasma membrane in low-, intermediate-, and high-affinity states. Structural studies suggest that a low-affinity state is characterized by a bent, “closed” conformation of the extracellular domains and a high-affinity state by an extended, “open” conformation (Campbell and Humphries 2011). The extended conformation is thought to be achieved by separation of the α and β cytoplasmic tails through binding of talin to the β subunit (Moser et al. 2009b; Shattil et al. 2010). Studies in cultured cells have shown that the binding of talin-1 to the cytoplasmic domain of the β-integrin subunit is a common step in β1 and β3 integrin activation in vitro (Tadokoro et al. 2003). In contrast, kindlin alone is not sufficient to activate these integrins, but its co-operation with talin is required to generate fully active integrins (Montanez et al. 2008; Moser et al. 2008, 2009a). The molecular details of this cooperation are not clear.

Regulation of the ligand binding affinity is fundamental for various cellular functions. For example, the platelet integrin αIIbβ3 needs to be maintained inactive to prevent it from binding fibrinogen present in the blood stream thus prohibiting platelet aggregation and thrombus formation. On the other hand, this integrin needs to be rapidly activated to mediate platelet adhesion on vascular injury to prevent bleeding (Fig. 3). Similarly, leukocytes require integrin activation to adhere to and migrate across the endothelium to combat pathogens. The importance of this process is illustrated by the phenotypes of talin- and kindlin-deficient mice. Deletion of talin-1 or kindlin-3 in platelets blocks activation of β1 and β3 integrins, leading to the inability of platelets to bind fibrinogen and to form clots. As a result, these mice suffer from severe bleeding (Nieswandt et al. 2007; Petrich et al. 2007; Moser et al. 2008). Similarly, neutrophils lacking kindlin-3 are unable to activate β2 integrins, resulting in loss of neutrophil adhesion to activated endothelial cells (Moser et al. 2009a). Talin has also been shown to bind and activate neutrophil integrin β2 in vitro (Simonson et al. 2006). The importance of integrin affinity regulation in adherent cell types is not as clear. Although deletion of talin and kindlin in mesenchymal or epithelial cells leads to defects in integrin activation, these cells are still capable of adhering, albeit to a reduced extent (Ussar et al. 2008; Zhang et al. 2008), suggesting that under conditions of less stringent requirements for rapid induction of adhesion, high concentrations of ligands, and the absence of high shear flow, avidity regulation might be the predominant mechanism of integrin regulation. The fact that both talin and kindlin are also important in regulation of outside-in signaling makes the in vivo dissection of these two processes difficult.

Figure 3.

Talin and kindlin regulate bidirectional integrin signaling. Bidirectional integrin signaling is essential for platelets (tan) to seal the injured blood vessel endothelium (red) and stop bleeding. Integrins in circulating, resting platelets exist in a low affinity state indicated by a bent confirmation (1). Following vessel injury, von Willebrand factor (vWF) and collagen are exposed to bind their receptors GPIb and GPVI that are expressed on the surface of platelets. Together with locally produced thrombin these receptors trigger the activation of αIIbβ3 integrin. This is achieved by promoting the association of talin-1 and kindlin-3 with the cytoplasmic tail of β3 integrin, facilitating a conformational change in the integrin (inside-out signaling) (2). The conformational change allows integrins to bind fibrinogen, vWF, and fibronectin with high affinity. As a result, platelets adhere to the vessel wall. Integrin ligation subsequently initiates signaling through kindlin and talin (outside-in signaling) (3) resulting in the recruitment of adaptor proteins and rearrangement of the cytoskeleton to promote cell spreading (4). This, together with integrin-mediated binding of soluble fibrinogen, results in platelet aggregation and formation of a stable clot (5). The molecules are not drawn to scale.

Consequences of Loss of Outside-In Signaling

A central piece of evidence that talin has important functions besides integrin activation comes from studies performed with talin-null flies, which show that in the absence of talin, integrins are able to associate with the ECM, but are unable to connect to the cytoskeleton, leading to muscle detachment at the integrin-actin interphase (Brown et al. 2002). Consistent with this finding, mouse fibroblasts lacking both talin-1 and talin-2 are able to adhere, but not to link integrins to the cytoskeleton (Priddle et al. 1998; Zhang et al. 2008). This function requires the talin rod domain, which contains binding sites for F-actin and vinculin indicating that talin provides an essential linkage to the cytoskeleton independent of its ability to activate integrins. Kindlins do not bind actin directly, but connect to the actin cytoskeleton via a migfilin-filamin interaction as well as through the IPP complex (Montanez et al. 2008). Platelets lacking kindlin-3 expression are unable to organize their cytoskeleton or to establish stable lamellipodia (Moser et al. 2008), which probably contributes to their inability to spread and form a crosslinked clot (see Fig. 3). Loss of kindlin-2 impairs actin organization and FA formation in endoderm cells, which together with defective integrin activation leads to early embryonic lethality. Interestingly, the rudimentary FAs in kindlin-2-deficient cells do not contain ILK, suggesting that kindlin acts as a nucleator of FA formation (Montanez et al. 2008).

Despite the central role of talin and kindlin in outside-in signaling, they are not sufficient to establish a stable integrin-actin connection. An important scaffold also required for this function is the IPP complex, which is a constituent of at least β1 and β3 integrin-containing adhesion sites. The core component of this complex is ILK, a pseudokinase that is also capable of binding these integrins directly, at least in vitro (Wickström et al. 2010). Whether the binding is actually relevant for the function of this complex, or whether it occurs in vivo, has not been conclusively shown. The other two members of this complex are pinch and parvin. Pinch, which interacts with the amino-terminal ankyrin repeats of ILK, is a family of proteins containing only LIM domains and consisting of two members, pinch-1 and pinch-2 (Tu et al. 1999, 2001; Chiswell et al. 2008). The parvins, which interact with the carboxy-terminal kinase-homology domain of ILK, are a family of CH-domain-containing proteins with three members; the ubiquitously expressed α-parvin (also known as actopaxin or CH-ILKBP), β-parvin (also known as affixin), which is highly expressed in heart and skeletal muscle, and γ-parvin, which is restricted to the hematopoietic system (Nikolopoulos and Turner 2000; Olski et al. 2001; Tu et al. 2001; Yamaji et al. 2001; Chu et al. 2006). The IPP complex is thought to be preassembled in the cytoplasm and is subsequently recruited to integrin adhesions in an ILK-dependent manner (Zhang et al. 2002). The stability of each individual component depends on the assembly of the complex, as depletion of ILK or pinch leads to a decrease in the protein levels of the other two complex members (Fukuda et al. 2003; Li et al. 2005).

The central biological function of the IPP complex is the regulation of the cytoskeleton downstream of integrins, although a number of additional functions for this complex as well as its individual components have been assigned by in vitro studies. The importance of these proteins in the integrin-actin connection is highlighted by deletion studies in mice, where deletion of ILK or pinch-1 results in peri-implantation lethality caused by severe defects in F-actin organization at adhesion sites, leading to a failure in epiblast polarization (Sakai et al. 2003; Li et al. 2005). The role of parvins is more complex because of the presence of three structurally very similar isoforms. Mice lacking β- or γ-parvin show no obvious phenotypes, whereas α-parvin-null mice survive up to embryonic day 14.5 and die because of defects in cardiovascular development (Chu et al. 2006; Montanez et al. 2009; I Thievessen and R Fässler, unpubl.). The α/β-parvin double knock-out results in early embryonic lethality, suggesting that parvins can compensate for each other during early development (E. Montanez and R. Fässler, unpubl.). Caenorhabditis elegans expresses only single orthologs for pinch and parvin, and is thus a suitable model organism for a more straightforward analysis of the IPP complex. Studies in C. elegans have shown that ILK (PAT-4), pinch (UNC-97), and parvin (PAT-6) colocalize with β integrin (PAT-3) at muscle attachment sites, where a robust connection of cells to the cytoskeleton is required during muscle contraction. Deletion of β integrin or any member of the IPP complex leads to detachment of muscles from the body wall and embryonic lethality (Mackinnon et al. 2002; Lin et al. 2003; Norman et al. 2007).

On the cellular level, ILK-deficiency leads to compromised cell adhesion, spreading and migration. This is because of defective recruitment of the cytoskeleton to adhesion sites, resulting in defects in FA maturation and cytoskeletal remodeling. The defective maturation of ILK-deficient FAs into fibrillar adhesions subsequently leads to impaired deposition of the FN matrix (Sakai et al. 2003; Stanchi et al. 2009). The precise molecular mechanism by which ILK regulates the cytoskeleton is not clear. Parvins are capable of interacting with F-actin through their two CH domains, but as these domains also regulate the interaction of parvins with ILK and paxillin, they might not be available for actin binding. Therefore, it is likely that the IPP complex requires additional downstream partners for the regulation of the actin cytoskeleton, and the identification of these proteins is a central focus of future research. It is, however, clear that the presence of parvin in the IPP complex is critical for the regulation of the actin cytoskeleton. The absence of α-parvin in mice causes impaired migration of vascular smooth muscle cells (vSMC) toward developing vessels resulting in defective stabilization of the vasculature and subsequent dilation of vessels, formation of microaneurysms and vessel rupture. The migration defect is caused by aberrant actomyosin contractility (Montanez et al. 2009), and can be phenocopied by vSMC-specific ablation of ILK (Kogata et al. 2009). However, this function seems to be restricted to certain cell types, suggesting that although the central role of the IPP complex in integrin outside-in signaling is ubiquitous, the precise molecular mechanisms might be cell type-specific to accommodate the specialized needs of various cell and tissue types.

GENETIC DISEASES OF CELL–MATRIX INTERACTIONS

Aberrant cell–matrix interactions are involved in a large number of pathological conditions such as cancer and various inflammatory diseases. Altered function of this signaling axis is however rarely the actual cause of these disorders. In contrast, recent studies have identified a panel of genetic diseases whose etiology can be pinpointed to a mutation in a component of the adhesive machinery. The high degree of conservation in the components of this pathway has allowed the generation of mouse models for these diseases that can be analyzed to generate mechanistic insights and therapeutic modalities for clinical applications.

Integrin Activation Diseases

As discussed earlier, leukocyte and platelet adhesion are cellular events in which integrin affinity regulation is of key importance. Inherited human diseases with defects in these processes have been described more than 25 years ago, but only recently, after the discovery of the central role of talin and kindlin in integrin activation, the causes for these diseases have been identified.

There are currently three distinct syndromes that together constitute the leukocyte adhesion deficiency (LAD) family of diseases. They affect distinct phases of the leukocyte adhesion cascade and therefore cause symptoms of variable severity. LAD I, a disease that affects several hundreds of patients worldwide, results from impaired firm adhesion of leukocytes, leading to recurrent severe infections and impaired wound healing. This disease is caused by a range of mutations in the β2 integrin (ITGB2) gene, including deletions, truncations, substitutions, frame shifts, and intronic mutations, resulting in loss of protein expression or expression of a truncated protein. LAD III (also known as LAD I/variant), which has been more recently described, is characterized by similar symptoms as LAD I. Interestingly, these patients also suffer from a bleeding tendency, indicating that additional β subunits are involved. Indeed, these patients show defects in β1, β2, and β3 integrin activation (Kuijpers et al. 1997; Alon and Etzioni 2003). In contrast to LAD I, no mutations in integrin genes have been identified, and the cause of this disease remained unknown until the role of kindlin in integrin activation was discovered. Subsequently, genetic sequencing of LAD III patients has revealed mutations in the kindlin 3 (FRMT3) gene in all patients tested so far, leading to expression of a truncated protein, or reduction or loss of protein expression (Mory et al. 2008; Kuijpers et al. 2009; Malinin et al. 2009; Svensson et al. 2009). Transfection of patient’s lymphocytes with wild-type kindlin-3 was shown to restore integrin activation (Malinin et al. 2009; Svensson et al. 2009). In addition, kindlin-3-deficient mice recapitulate all symptoms of LAD III, firmly identifying defective kindlin-3 as the cause of LAD III (Moser et al. 2008; Moser et al. 2009a). In this respect, reconstitution of kindlin-3 expression provides a possible therapeutic modality for this disease. Interestingly, the naturally occurring mutations have also provided information on the structure-function aspects of kindlin-3. A homozygous stop codon identified in three patients occurs distal to the PTB-containing integrin-binding subdomain of kindlin-3, suggesting that the carboxyl terminus of kindlin is required for its function in integrin activation (Mory et al. 2008). A point mutation leading to a truncation in the middle of the PH-domain was further shown to impair membrane localization of kindlin-3 and thereby block both lymphocyte adhesion and migration, whereas a point mutation in the F2 subdomain inhibited only the migration of these cells, demonstrating that the functions of kindlin in inside-out and outside-in signaling can be uncoupled (McDowall et al. 2010).

The bleeding tendency of LAD III patients closely resembles that of Glanzmann’s thrombasthenia (GT), a rare, autosomal recessive bleeding disorder caused by mutations leading to quantitative or qualitative defects in platelet αIIbβ3 integrin. The genetic background is very heterogeneous: more than 100 different mutations in either the αIIb (ITGA2B) or β3 integrin (ITGB3) genes have been reported so far (Kannan and Saxena 2009). Genetic ablation of the β3 integrin in mice almost fully recapitulates the human disease, including gastrointestinal and cutaneous hemorrhage, increased bleeding time, reduced platelet aggregation, and clot retraction (Hodivala-Dilke et al. 1999). However, the heterogeneity in the mutation spectrum of the integrins found in GT patients also applies to the symptoms; siblings sharing the same mutation can display symptoms of varying severity, suggesting that polymorphisms in other genes involved in hemostasis could influence the phenotype. In addition, several GT patients lacking a mutation in the αIIbβ3 integrin have been identified (Kannan and Saxena 2009). As the bleeding phenotype of the kindlin-3-deficient mice and the LADIII patients closely resemble GT, it is possible that altered function of this protein might play a role in this disease as well. However, no polymorphisms in the kindlin-3 gene in GT patients have been reported so far. Taken together, integrin activation plays a central role in immune function and hemostasis. The prominent role of kindlin-3 in these processes together with its restricted expression pattern makes it a promising candidate for anti-inflammatory and antithrombotic therapies.

Adhesion Strengthening Diseases

Another “class” of diseases involving impaired cell–matrix interactions can be found affecting tissues that are subjected to high levels of mechanical stress such as the skin and skeletal muscle. The common denominator of these disorders, namely muscular dystrophies and skin-blistering diseases, is that the symptoms are caused by weakened interactions of cells with their extracellular environment, leading to tissue dysfunction and disruption. Another common feature is that causative mutations for a single disorder can be found in either components of the ECM, their integrin receptors, or adhesion-strengthening scaffold proteins. In the case of congenital muscular dystrophies, although the dystrophin-dystroglycan-laminin/agrin axis is a common target (Yurchenco 2010), other mutations affect the laminin-211/α7β1 integrin/plectin axis, leading to defects in organization and migration of myoblasts, subsequent myofiber degeneration and progressive muscle weakness (for more detailed review see Kanagawa and Toda 2006; Mendell et al. 2006). Clinical features, cellular defects and their corresponding causative mutations of muscular dystrophies resulting from defective cell-ECM interaction are summarized in Table 1.

Table 1.

Selected mutations causing muscular dystrophy and their corresponding mouse models

Human disease	Central clinical features	Affected gene (protein)	Cellular defect	Mouse model	Reference
Limb-Girdle muscular dystrophy	Early adulthood onset proximal muscle weakness, nasal pattern of speech	MYOT (myotilin)	Sarcomeric organization	N/A	–
	Severe, early onset muscle weakness	SGCA (α sarcoglycan)	Muscle membrane integrity	KO	(Duclos et al. 1998)
		SGCB (β sarcoglycan)		KO	(Durbeej et al. 2000)
		SGCG (γ sarcoglycan)		KO	(Sasaoka et al. 2003)
		SGCD (δ sarcoglycan)		KO	(Coral-Vazquez et al. 1999)
	Late onset muscle weakness and atrophy, cardiomyopathy	TTN (titin)	Sarcomere contraction /relaxation	N/A	–
	Relatively mild muscle weakness, variable clinical features	TCAP (telethonin)	Sarcomere stability	N/A	–
Congenital muscular dystrophy	Early onset with variable symptoms, hypotonia, torticollis	ITGA7 (integrin α7)	Impaired muscle cell–matrix interactions	KO	(Mayer et al. 1997)
	Early onset hypotonia and muscle weakness	COL6A1 COL6A2 COL6A3 (type VI collagen)	Stability of extracellular matrix	KOa	(Bonaldo et al. 1998)
	Severe, early onset muscle weakness, white matter hypodensity, mental retardation	LAMA2 (laminin 211)	Attachment, migration and organization of myoblasts	Dy/Dy (naturally occurring mutant)	(Sunada et al. 1994)
				Dy2j/Dy2j (naturally occurring mutant)	(Xu et al. 1994)
				DyPas/DyPas (naturally occurring null allele)	(Besse et al. 2003)
				Dy3K/Dy3K (KO)	(Miyagoe et al. 1997)
				DyW/DyW (hypomorph)	(Kuang et al. 1998)
Duchenne muscular dystrophy	Severe, early onset muscle weakness with respiratory, orthopedic and cardiac complications	DMD (dystrophin)	Stabilization of membrane during contraction	Mdx (naturally occurring null allele)	(Bulfield et al. 1984)
				Mdx 2Cv, Mdx 3Cv Mdx 4Cv, Mdx 5Cvb	(Chapman et al. 1989)
				Mdx 52 (deletion of exon 52)	(Araki et al. 1997)

^aAblation of the col6a1 gene.

^bMutants recovered from ENU chemical mutagenesis screen.

Abbreviations: N/A, not analyzed; KO, knockout mouse; Dy, Dystrophin; mdx, X chromosome-linked muscular dystrophy.

The same principles are true for epidermolysis bullosa (EB), the prototype of human skin-blistering diseases (Uitto 2009). EB is a group of highly variable disorders characterized by fragility of skin leading to blistering and erosions on mechanical trauma. Genetic studies on different variants of EB have so far shown mutations in 12 distinct genes encoding structural components of the BM and adhesion proteins (Table 2). Interestingly, some of the EB patients also develop muscular dystrophy, underlining the pathophysiological link between these two diseases. As the histopathology of these diseases is very complex, the identification of the causative mutation has become a central diagnostic tool of EB (Nagy and McGrath 2010). Genetic mouse models for the different EB variants are available, and they have been used to work out therapeutic strategies for these diseases (Natsuga et al. 2010). As a result, gene delivery, ectopic protein replacement as well as cellular therapies have already been tested on EB patients (Uitto 2009).

Table 2.

Mutations causing epidermolysis bullosa (EB) and their corresponding mouse models

Human disease	Site of tissue separation	Affected gene (protein)	Mouse model	Mouse skin phenotype	Reference
Epidermolysis bullosa simplex (EBS)	Intraepidermal	PKP1 (plakophilin-1)	N/A	–	–
		DSP (desmoplakin)	N/A	–	–
		KRT5 (keratin-5)	KOa	Severe blistering, loss of keratin filaments	(Peters et al. 2001)
		KRT14 (keratin-14)	KO	Blistering, decreased keratin filaments	(Lloyd et al. 1995)
			K14-R131C KIa	Severe blistering	(Cao et al. 2001)
			K14-ΔCT TG	Blistering	(Vassar et al. 1991)
		PLEC1 (plectin)	KOa	Blistering, reduced stability and number of hemidesmosomes	(Andra et al. 1997)
			Conditional KO	Blistering, skin fragility	(Ackerl et al. 2007)
Junctional epidermolysis bullosa (JBS)	Lamina lucida	ITGA6 (integrin α6)	KOa	Severe blistering, loss of hemidesmosomes	(Georges-Labouesse et al. 1996a)
		ITGB4 (integrin β4)	KOa	Severe blistering, loss of hemidesmosomes	(Dowling et al. 1996; van der Neut et al. 1996)
			ITGB4-ΔCT KIa	Severe blistering, loss of hemidesmosomes, hypoproliferation	(Murgia et al. 1998)
			Conditional KO	Blistering, loss of hemidesmosomes	(Raymond et al. 2005)
		LAMA3 (laminin-332)	KOa	Severe blistering, abnormal hemidesmosomes, reduced cell survival	(Ryan et al. 1999)
		LAMB3 (laminin-332)	Spontaneous null allele	Blistering, abnormal hemidesmosomes	(Kuster et al. 1997)
		LAMC2 (laminin-332)	KOa	Blistering, rudimentary hemidesmosomes, apoptosis	(Meng et al. 2003)
			Spontaneous hypomorphic allele	Blistering, ulcerations, hyperkeratosis, rudimentary hemidesmosomes	(Bubier et al. 2010)
Dystrophic epidermolysis bullosa (DBS)	Sub-lamina densa	COL7A1 (type VII collagen)	KOa	Severe blistering, absence of anchoring fibrils	(Heinonen et al. 1999)
			COL7 hypomorph	Severe blistering, decreased anchoring fibrils	(Fritsch et al. 2008)
Kindler syndrome	Primarily lamina lucida	FERMT1 (kindlin-1)	KOa	Skin atrophy	(Ussar et al. 2008)

^aLeads to early postnatal lethality.

Abbreviations used: N/A, not analyzed; KO, knock-out mouse; K14-R131C, Arginine-131 to Cysteine substitution in the Keratin-14 gene; KI, knock in mouse; K14-ΔCT, carboxy-terminal truncation of Keratin-14; ITGB4-ΔCT, deletion of the entire cytoplamsic domain of integrin β4; TG, transgenic mouse.

CONCLUDING REMARKS

Two decades of genetic studies on cell-ECM interactions have illustrated the importance of adhesion signaling in the development of multicellular organisms as well as in disease. It is now clear that these interactions not only provide structural support and positional cues to guide morphogenesis, but also act as signaling platforms to regulate cell fate in a highly tissue-specific manner. It is also evident that the relationship between ECM proteins and their integrin receptors is complex: Integrins transmit information from the ECM to the cell, but also regulate the deposition and remodeling of the matrix itself. In addition, the ECM functions as a reservoir for growth factors, whereas integrins cross talk with growth factor signaling on various levels. Because of this complexity, mechanistic interpretations of the various phenotypes caused by genetic ablation of individual components of the adhesion signaling machinery have proven to be difficult. Thus, a major challenge for the future is to understand which types of intracellular signals directly emanate from integrin adhesions, and how these signals are propagated. This requires generation of more sophisticated mouse models, such as knock-in mice carrying point mutations that only partially disrupt protein function together with reporter mice for specific signaling events, transcriptional changes, or changes in protein-protein interactions or conformations. In addition, a large majority of the cell-biological analyses on integrin signaling are still performed on rigid 2D-surfaces, which very poorly recapitulate the more compliant and complex 3D environment of tissues. As our understanding of the biophysical properties of the ECM in tissues steadily increases, a major goal of future research is to translate this knowledge into generation of more relevant in vitro models for cell biological studies of integrin signaling.