Integrins: versatile receptors controlling melanocyte adhesion, migration and proliferation

Authors

  • Perrine Pinon,

    1. Department of Cell Physiology and Metabolism, Centre Médical Universitaire, University of Geneva, Medical School, Geneva 4, Switzerland
    Search for more papers by this author
  • Bernhard Wehrle-Haller

    1. Department of Cell Physiology and Metabolism, Centre Médical Universitaire, University of Geneva, Medical School, Geneva 4, Switzerland
    Search for more papers by this author

B. Wehrle-Haller, e-mail: Bernhard.Wehrle-Haller@unige.ch

Summary

From the onset of melanocyte specification from the neural crest, throughout their migration during embryogenesis and until they reside in their niche in the basal keratinocyte layer, melanocytes interact in dynamic ways with the extracellular environment of the growing embryo. To recognize and to adhere to their environment, melanocytes depend on heterodimeric cell surface receptors of the family of integrins. In addition to the control of adhesive interactions between melanocytes and the extracellular matrix scaffold secreted by fibroblasts and keratinocytes, the integrin receptors allow cells also to sense the mechanical condition of the extracellular environment, responding by intracellular signaling, triggering cell survival, proliferation or migration events. In this review, we summarize the recently emerged concepts that explain integrin-dependent adhesion and how this adhesion system interfaces with integrin-dependent signaling events. The gained information will help to understand melanocyte behavior in pathological situations such as melanoma growth and metastasis formation.

Introduction

In contrast to fibroblasts or epithelial cells, which express an integrin repertoire to be able to adhere to their own secreted extracellular matrix such as collagen fibrils of the interstitial matrix or laminin containing basement membranes, respectively, melanocytes do not secrete but mainly respond to previously deposited extracellular matrix (ECM). This requires neural crest cells and their derivatives such as melanocytes to express an integrin repertoire enabling adhesion to different types of ECM components (Delannet et al., 1994; Scott et al., 1992). After completion of their migration into the skin, melanocytes assume a dendritic morphology adhering to the surrounding keratinocytes via E-cadherin-dependent adherence junctions and to the basement membrane of the epidermis with laminin-binding integrins such as α6β1 (Hara et al., 1994; Tang et al., 1994). Because melanocytes do not express the β4-integrin subunit or desmosomal proteins, they are mechanically uncoupled from the keratin network that maintains adhesion and cohesion of the stratified epithelium. This in turn makes them dependent on survival signals from adjacent keratinocytes as well as on integrin-dependent adhesion to the laminin-containing basement membrane (Mortarini et al., 1995; Scott et al., 1994, 1997). This dual stimulation creates synergies in intracellular signaling (Miyamoto et al., 1996) that are critical for adhesion and survival of melanocytes in the epidermis.

Because adhesion to the basement membrane is essential for the homeostasis of melanocytes, changes in integrin expression levels are often associated with altered behavior of melanocytes. For example, UV-irradiation down-regulates cell surface expression of α6β1 integrins, which can cause detachment of melanocytes from the basement membrane (Krengel et al., 2005). In contrast, the de novo expression of the αvβ3 integrin in melanoma has been associated with the ability to convert a non-invasive radial growth phase melanoma into an invasive, vertical growth phase melanoma (Hsu et al., 1998; Seftor et al., 1992, 1999). Furthermore, changes in the balance between the laminin-binding integrins α6β1 and α7β1 can determine the invasive or a non-invasive behavior of melanoma (Ziober et al., 1999).

The ground-breaking observation that the augmented expression of the αvβ3 integrin in melanoma or endothelial cells is associated with increased invasive capacity has boosted an interest in developing highly specific anti-integrin therapeutics in an attempt to block integrin-dependent adhesion and invasion (Albelda et al., 1990; Brooks et al., 1994a,b; Felding-Habermann et al., 1992). In addition, the β3-integrin receptor was directly implicated in metastasis formation, since the expression of an activated form (D723R mutation) increased the formation of lung metastasis (Felding-Habermann et al., 2001). Accordingly, treatment of mice with αvβ3-integrin ligand-mimetics revealed inhibition of tumor-induced angiogenesis at high doses (Brooks et al., 1994b). However, the treatment of mice with low doses of αvβ3-integrin ligand-mimetics stimulated tumor-mediated angiogenesis (Reynolds et al., 2009). In addition, the knockout of β3 and β5 integrins resulted in increased angiogenesis, which indicated that the αvβ3 integrin plays mainly a regulatory role in integrin-mediated morphogenetic events (Reynolds et al., 2002). That such differential behavior is not limited only to the αvβ3 integrin receptor has been demonstrated by the ligation of α6β1 integrins with ligand mimetic peptides. This treatment stimulated melanoma invasion (Nakahara et al., 1996), suggesting that the morphological consequences of integrin–ligand binding are strongly context-dependent and are modulated by the mechanical tethering of ligands.

These examples show that the physiological functions of integrins at the tissue level are still not completely understood. Therefore, it is still difficult or impossible to predict how integrin inhibitors react in vivo. For example, when αvβ3 inhibitor treatment was combined with radiation therapy for the treatment of experimental rat brain glioma, no therapeutic benefit was obtained when the inhibitor was injected at the time of irradiation. However, irradiation between 4 and 12 h after injection of the integrin inhibitor caused complete remission of the tumor (Mikkelsen et al., 2009). This demonstrates that the integrin inhibitor treatment profoundly changed the physiology of the glioma or its tissue environment, in as yet unknown ways but with an important therapeutic benefit.

Again, these unexpected results illustrate the lack of information concerning the function and regulation of integrins. Before using integrin inhibitors in the clinic, it is therefore paramount to understand how this family of receptor is regulated and involved in the mechanical coupling to the actin cytoskeleton and how these receptors generate or affect intracellular signaling.

Affinity versus valency regulation of integrin-dependent adhesion

Two different concepts have been developed to explain integrin-dependent adhesion. The first concept is based on results obtained with the platelet-specific integrin receptor αIIbβ3 and hematopoietic specific β2-integrin receptors (e.g. αLβ2). These receptors switch rapidly from an inactive (low affinity) to an adhesive (high affinity) conformational state (Takagi and Springer, 2002; Takagi et al., 2002; Xiao et al., 2004) when brought in contact with fibrinogen or activated endothelial cells. This rapid switch in adhesion allows leukocytes and platelets to adhere to walls of lesioned blood vessels even under conditions of high shear forces. Opposing this affinity-switching model is the concept that increased integrin adhesion is mainly due to the modulation of the lateral clustering of integrin receptors. This ‘valency’ regulation predicts an increase in cell adhesion, when integrin receptors are stabilized in large clusters, for example due to cytoskeletal rearrangements (Bazzoni and Hemler, 1998). As often the case with such controversies (Carman and Springer, 2003), varying the contribution of these two concepts creates diversity among integrin-dependent adhesion (Kim et al., 2004). For example, elegant work by Hughes and coworkers (Hughes et al., 1996) has shown an inhibitory role of a membrane proximal salt-bridge in β3-integrins, the mutation of which switches integrins from a low to a high affinity state, increasing integrin-dependent adhesion via ‘affinity’ regulation. However, the same mutation in this integrin also increases its clustering, as in ‘valency’ regulation (Cluzel et al., 2005). Interestingly, the same ‘activating’ mutation in β1-integrin was without physiological consequences (Czuchra et al., 2006), suggesting that β1-integrins behave differently, which would be consistent with their use in cellular tasks that do not require rapid affinity switching. The choice between affinity or valency regulation could also depend on the cell type. For example in fibroblasts, a strong correlation between the strength of integrin-dependent adhesion and the size (surface area = valency) of focal adhesions has been established (Balaban et al., 2001; Ballestrem et al., 2001). Furthermore, in fibroblasts, rapid changes in valency can be induced by applying external tension or RhoA GTPase signaling, which suggests that ‘valency’ regulations are intimately coupled to intracellular signaling (Riveline et al., 2001; Wehrle-Haller, 2005; Wehrle-Haller and Imhof, 2002).

In addition, the degree of integrin valency can be further modulated by changing the integrin densities (number of integrins per adhesion site surface) and dynamics (how many integrins are substrate bound at a given time), which makes it possible to create focal adhesions with different functionalities (Wehrle-Haller, 2005). Whether integrin-dependent adhesion is mainly regulated by ‘affinity’ or ‘valency’ modulations most likely depends on the specific cell and integrin type involved. Whereas platelet adhesion after injury or chemokine-induced leukocyte adhesion requires rapid increases in integrin affinity, slowly crawling fibroblasts rely on valency changes for remodeling and the maintenance of the tensional state of the ECM. Nevertheless, both concepts are indispensable for integrin and cellular function, while affinity changes control the physical coupling between the ECM and the actin cytoskeleton. In turn, valency changes modulate the number of adhesive bonds per adhesion site, as well as providing a link to intracellular signaling (see below).

In this review, we would like to summarize recent advances in the understanding of the regulation of integrin-based adhesion. In addition, we will use evolutionary and functional arguments to define the respective roles of specific integrin parts as well as the role of cytoplasmic integrin adapter proteins and their regulated association with the plasma membrane and actin cytoskeleton. What emerges is an allosteric regulated transmembrane receptor system that can convert mechanical states into chemical and intracellular signals, putting it at the origin of the regulation of melanocyte adhesion as well as melanoma growth and invasion.

The structure and dynamics of integrins

The ectodomain

Thanks to crystallization, NMR and molecular dynamics analysis, a clear model of the integrin receptor has emerged (Figure 1A). One hallmark is the ability of the integrin ectodomain to bend or fold like a ‘Swiss army’ knife, bringing the ligand binding site close to the insertion of the transmembrane domains (Xiong et al., 2001). Based on antibody mapping studies, it has been proposed that the bent state represents the inactive and the extended state the active conformation of the integrin (Askari et al., 2010; Honda et al., 1995; Zhu et al., 2008). Despite the attractiveness of this model, accounting for the low and high affinity states of integrins, so far the use of FRET/FLIM-based methods to observe changes in head-domain/membrane distances has not been able to conclusively demonstrate transitions from bent to extended conformations of integrins in living cells (Askari et al., 2010; Xiong et al., 2009).

Figure 1.

 Integrin structure and allosteric movements of the ectodomain: (A) Crystal structure and color-coded cartoon of the bent form of the ectodomain of the αvβ3 integrin [PDB: 3IJE, (Xiong et al., 2009)]. The double line indicates the position of the plasma membrane and the dashed line the proposed rotational axis leading to the extended form of the integrin. (B) Close-up of the PSI/hybrid/β-I domain module in the low affinity position (from A). The dashed line indicates the rotational axis leading to the extended, high affinity conformation shown here for the PSI/hybrid/β-I domain of αIIbβ3 integrin [PDB: 3FCU, (Zhu et al., 2008), C]. (B’,B”) Cartoon of the extended low affinity conformation with parallel and twisted legs. (C’,C”) Cartoon of the high affinity conformation with separated and connected legs. Note the Ca2+ ion in the AdMIDAS position in B and the two Ca2+ (magenta) and Mg2+ (blue) ions in the SyMBS, AdMIDAS and MIDAS sites, respectively, in C (Zhu et al., 2008). (D) Two different NMR-based models of the clasped αIIbβ3 integrin transmembrane domains indicating the inhibitory cytoplasmic salt bridge (D723-R995) between the β- (magenta) and α- (orange) subunits (left side: PDB: 2K9J, (Lau et al., 2009); right side: PDB: 2KNC (Yang et al., 2009)). The model on the right hand side shows the two conserved tyrosine residues in β3-integrin (Y747 and Y759) that are binding sites for talin and kindlin respectively (Moser et al., 2008). The conserved acidic residues E726 and E733 that are critical for talin binding to the membrane proximal integrin domain are hidden at the α-β-subunit interface (Saltel et al., 2009).

Recent studies narrow the critical conformational modulations that are directly responsible for an increase in integrin ligand affinity to the β-I (green) and hybrid (blue) domain located at the end of a flexible stalk (Figure 1B,C). The affinity change in the integrin receptor is linked to the rearrangements of the coordination pattern of three metal ions located at the top of the β-I domain (Xiao et al., 2004; Zhu et al., 2008). A downward movement of α-helix 7 of the β-I domain modifies the metal ion coordination and thereby increases the electrophilicity of the central magnesium ion, which increases the electrostatic attraction to negatively charged integrin ligands such as the Arg-Gly-Asp peptide (Zhu et al., 2008). Because α-helix 7 is directly connected to the hybrid domain, the change in ion coordination is coupled to an opening of the angle between the β-I and hybrid domain, also termed hybrid domain swing-out (Xiao et al., 2004) (Figure 1B,C). Compared to this defined angular movement, other parts of the integrin receptor appear to be very flexible. For example, the conformational analysis of purified integrin receptors clasped at the level of their transmembrane domains revealed that 32% of all receptors are in an extended state, showing straight or twisted conformations of the integrin legs (Zhu et al., 2008) (Figure 1B,C). However, when a cysteine bridge is introduced into the integrin ectodomain that stabilizes the bent conformation, no extended integrins can be identified (Zhu et al., 2008). This suggests that during integrin processing and transport during the ER and Golgi, integrins must change their conformational states. Because of this apparent flexibility, it is likely that integrins exist in multiple interchangeable conformations at the cell surface. In turn, the binding to an extracellular ligand will transiently stabilize the high-affinity ion coordination as well as the open angle between the β-I and hybrid domain. As both immobilized or soluble integrin ligands will stabilize the high affinity or extended state of integrins, modification of the conformational pool of integrins is expected if cells are treated with ligand mimetic drugs.

Transmembrane and cytoplasmic domains

In addition to the extracellular domain, integrins exhibit highly conserved transmembrane and cytoplasmic domains. The transmembrane domain of the α-integrin subunit contains a conserved glycine-repeat motif that creates a notch in the helix accommodating the β-transmembrane domain, enabling close transmembrane domain association (Lau et al., 2009; Luo et al., 2004; Yang et al., 2009; Zhu et al., 2009). To further stabilize this association of the two transmembrane domains, a conserved aromatic and positively charged (GFFKR) motif of the α-subunit comes into contact with acidic juxtamembrane residues of the β-subunit, clasping the integrin cytoplasmic domains and thereby preventing the access of adapter proteins to conserved motifs of the β-cytoplasmic domain (see below).

Integrin adapter proteins such as kindlin and talin have been demonstrated to bind to the distal and membrane proximal NPXY motifs, respectively (Garcia-Alvarez et al., 2003; Moser et al., 2008). In addition, talin exhibits binding sites for membrane proximal acidic and aromatic residues (Saltel et al., 2009; Wegener et al., 2007), which interaction allows talin to unclasp the transmembrane domain association and thereby ‘activate’ the integrin receptor.

The dynamics of integrin receptors

Based on this information, it becomes apparent that conformational changes of the integrin receptor are induced by cytoplasmic adapter protein binding as well as extracellular integrin ligands. Nevertheless, recent work demonstrated that the clustering of integrins into a functional adhesion site requires the simultaneous interactions at the extracellular as well as intracellular sides (Saltel et al., 2009). Despite the low affinity of integrin/ligand or integrin/adapter interactions, focal adhesions appear to be stable over time. However, it should be underlined that within these stable adhesions, individual αvβ3 integrins are bound only transiently and exchange very rapidly. This highly dynamic binding mode is critical for the proper function of an integrin-dependent adhesion site and is directly linked to the conformational flexibility of the integrin ectodomain. When the integrin ectodomain is stabilized in the ligand-binding conformation by introduction of a glycan wedge between the β-I and hybrid domain (Luo et al., 2003), integrin exchange rates within focal adhesions slow down by about fivefold, which can result in the accidental shedding of substrate-bound integrins when cells decide to move (Cluzel et al., 2005). In contrast, increasing the number of clustered integrins by either mutational unclasping at the juxtamembrane level, or by manganese-induced switching to the high-affinity metal coordination state, does not change integrin dynamics within adhesion sites (Cluzel et al., 2005).

These data indicate that despite detailed structural information and precise models of integrin unclasping, the mechanisms and the regulation of integrin-dependent adhesion are not yet completely understood. In the next chapter, we will analyze the integrin receptor from an evolutionary point of view, hoping to identify critical concepts relevant for its function.

The evolution of integrins

When comparing the primary sequence of metazoan integrins, all of the above mentioned structural features are conserved. For example, the β-integrin subunit from the placozoan Trichoplax adhaerens, one of the most primitive multicellular organisms demonstrates perfect conserved metal ion coordination sites in its β-I domain, as well as an N-terminal PSI-domain and EGF-repeats. In addition, the transmembrane and cytoplasmic domains bear the critical residues to bind to the equally conserved talin and kindlin adapter proteins (Srivastava et al., 2008). This evolutionary conservation among metazoans strongly suggests that the critical functions of integrins are conserved among all multicellular organisms (Engler et al., 2009). This conservation means that we need either to identify integrin precursors among proteins that carry structurally similar domains or to analyze organisms further down the evolutionary tree.

When comparing the different structural motifs found in integrin receptors, common motifs such as the EGF-repeat, β-propeller or β-sheet bundles are found in other multi-domain proteins. One notable exception is, however, the PSI, hybrid and β-I domains in the β-integrin subunit. These three domains are arranged in such a way that the β-I domain is inserted within the β-sheet bundles of the hybrid domain, which C-terminus is disulfide-bridged to the N-terminal PSI-domain. The only other known proteins with a similar folding topology are the intracellular COPII adapter proteins Sec23 and Sec24 (Figure 2) (Xiong et al., 2001).

Figure 2.

 Structure-based alignment and evolutionary comparison of the β-integrin sequence. (A) Structure-based alignment of the hybrid and β-I domain of human (hs Itgb3) and Trichoplax adhaerens integrins (ta Itgb) with sibA from Dictyostelium (dd sibA) and I-domain from human αM-integrin (hs ItgaM). β-sheets (blocks) and α-helices (wavy lines) are color coded as in Figures 1B and 2C. The MIDAS, AdMIDAS and SyMBS sites are boxed in red, magenta and purple respectively. The conserved GW motif and the GFG-loop in the β-I and hybrid domains, respectively are boxed in dark green. (B) Alignment of transmembrane and cytoplasmic domains of human, Trichoplax adhaerens and sibA and sibB proteins reveal a highly conserved transmembrane domain and proximal and distal NPXY motifs. In sib’s, the α-helical structure of the membrane proximal domain is interrupted by a double proline replacing the acidic amino acid involved in contacting the α-subunit in metazoan integrins (boxed in red). C,D,E color coded models of the PSI/hybrid/β-I module of bent αvβ3 integrin (PDB: 3IJE, (Xiong et al., 2009)) compared to Sec23 (PDB: 2QTV, (Bi et al., 2007)) and the I-domain of αM-integrin (PDB:1IDO, (Lee et al., 1995). The GF-loop, specificity domain and α-subunit contacting domains are indicated in C. (F) Cartoon of a catch bond mechanism in the integrin hybrid/β-I module. At a weak mechanical load, the hybrid/β-I angle is flexible and can spontaneously close, switching the integrin into the low affinity conformation. At a heavy load, closing of the hybrid/β-I angle is prevented, keeping the integrin in the high-affinity conformation.

Integrin ligands are bound by the β-I domain which exhibits a von Willebrand factor A domain fold (VWA), which is normally characterized by only one metal ion binding site (MIDAS), which is responsible for ligand binding very similar to that of the β-I-domain (Takagi and Springer, 2002; Whittaker and Hynes, 2002; Zhu et al., 2008). In more modern integrin α-subunits, a VWA domain is inserted into the β-propeller, creating an integrin receptor that transmits the structural changes of ligand binding to the α-subunit, via a conserved glutamic acid residue to the associated β-I domain (Takagi and Springer, 2002). The benefit of such a structural arrangement is not yet understood; in addition to positioning the ligand binding site further away from the cell membrane, it appears to duplicate a ligand-induced conformational change as shown in the β-I-hybrid domain module. This could increase the number of different conformational changes of the receptor or increase the energy barrier between the low and high affinity conformations, for example preventing the spontaneous activation of leukocyte-expressed integrins.

When stepping down the evolutionary tree, a recent mutational analysis of adhesion and phagocytosis in dictyostelium has revealed a new adhesion receptor in which some of the features of β-integrin subunits are conserved (Cornillon et al., 2006). The deletion of this receptor, termed sib (similar to β-integrin), affects both adhesion and phagocytosis, suggesting that this receptor is used to adhere to different types of surfaces as well as catching prey. Sequence comparison reveals homology to β-I as well as the hybrid domain of β-integrins (Figure 2). Its VWA domain exhibits one or two potential metal ion binding sites but is missing the specificity loop as well as the α-subunit binding motif, suggesting that sibs are not regulated by an α-subunit. When comparing critical sequence motifs in the VWA domain, sibs exhibit a conserved GW motif at the base of β-sheet 4 that is found in the β-1-domain of integrins but not in other VWA domains (Figure 2). The presence of a putative hybrid domain is also intriguing, since the conservation is strongest in a small residue-bearing loop (GP-loop) that faces the β-I-domain. As this small loop assures the unobstructed movement and closing of the β-I-hybrid domain angle, it is tempting to speculate that ligand binding-induced conformational changes are also relevant in sibs.

This notion is further supported by the conservation of a double NPXY motif in the cytoplasmic domain and the ability to bind to the adapter protein talin (Cornillon et al., 2006). Moreover, the membrane proximal domain is rich in aromatic and acidic residues, potentially allowing talin binding at this site as well. Thus, we are confronted with an integrin-like receptor that can be regulated by affinity and valency changes. In addition, the conservation of a flexible hinge between the β-I and hybrid domain may suggest that sibs behave like integrins as catch bonds (Kong et al., 2009). Catch bonds increase in strength when ligands are pulled away from the receptor. From an evolutionary point of view, it would make sense if hunting organisms such as amoebas have developed a way to capture prey based on catch bonds, a concept also found in bacterial proteins responsible for interactions with host cells (Tchesnokova et al., 2008). Recent molecular dynamics modeling propose that the catch bond in integrins is due to a direct correlation between high-affinity binding and an open angle between the β-I and hybrid domain, which is also the configuration which would be favored under mechanical tension along the β-I-hybrid domain axis (Figure 2F) (Puklin-Faucher and Vogel, 2009; Puklin-Faucher et al., 2006).

Although the mechanical influence on the capacity of integrin-dependent binding needs to be further evaluated, the instructive role of the mechanical properties of the cellular environment has emerged as one of the most relevant aspects influencing normal and pathological cell behavior (Engler et al., 2009) (see below). However, before discussing the critical role of integrins in this process, we need to introduce the basic concepts that control integrin-dependent adhesion and signaling, respectively.

Integrin clustering and its connectionto signaling

When melanocytes or melanoma cells are plated on adhesive extracellular matrix, they develop integrin-containing focal adhesion sites to which the actin cytoskeleton is anchored. In addition, in the presence of structural proteins such as talin and vinculin, which couple integrins to the actin network, these focal adhesions recruit focal adhesion kinase (FAK) and phospho-tyrosine-containing proteins, such as paxillin. As a consequence, focal adhesions are not only sites that allow mechanical coupling to the environment, but they constitute an important source of survival signals, which led to the creation of the term adhesome (Zaidel-Bar et al., 2007).

The importance of integrin-dependent adhesion as a survival signal for melanoma cells in ectopic places has also been suggested by Hoek and coworkers, who  identified different molecular signatures linking slow proliferation and invasive phenotypes of melanoma cells to an adhesion-mediating ‘TGF-β-like’ signature. In contrast, rapidly proliferating, weakly adhering and non-invasive melanomas demonstrate a ‘Wnt’ signaling signature with up-regulated expression of the transcription factor MITF (Hoek et al., 2006). Interestingly, both phenotypes can coexist in melanoma, potentially reflecting the migratory and proliferative capacities of embryonic melanocytes (Hoek et al., 2008). The important role of MITF in this phenotype switching is intriguing; high MITF levels correlate with an increase in Dia1, which orchestrates actin polymerization downstream of RhoA activity and mechanical stresses (Riveline et al., 2001). Suppression of MITF leads to a loss of Dia1, which provokes a change in the actin cytoskeleton, resulting in increased melanoma cell invasion (Carreira et al., 2006).

These data demonstrate the emergence of ‘inside-out’ regulatory pathways that link changes in transcription factors with altered F-actin polymerization and integrin-dependent adhesion and migration. In contrast,‘outside-in’ pathways, in which the signal originates at the level of the integrin receptors, are less well understood, although they are caused by the clustering of integrin receptors. When integrins are clustered on the cell surface by non-activating integrin-antibody-coated beads, FAK and other non-receptor tyrosine kinases are recruited and activated by the clustered integrins. Importantly, these beads do not recruit talin or establish a link to the actin cytoskeleton. However, after addition of soluble ligands (RGD peptides), talin-binding and F-actin coupling is induced, concomitant with a further increase in FAK activity (Miyamoto et al., 1995) (Figure 3A). This experiment demonstrates (i) the correlation between integrin clustering and integrin signaling and (ii) the link between ligand-induced conformational changes and the recruitment of the adapter protein talin (Figure 3A).

Figure 3.

 Different model systems to analyze integrin functions. (A) Experimental set-up used to monitor the recruitment of integrin adapters in living cells. Adherent cells were incubated with non-activating anti-integrin antibody-coated beads. After bead attachment to the dorsal plasma membrane, beads and cells were fixed and adapter protein recruitment to the bead/cell interface was scored with fluorescence microscopy. In the absence of integrin ligands, beads recruit signaling adapter proteins such as FAK. However, beads only recruit talin and connect with the F-actin network when soluble ligand (RGD-peptide) is added (Miyamoto et al., 1995, 1996). (B) αIIbβ3 integrin activation assay using the PAC-1 antibody. Ligand-mimetic PAC-1 antibody binding is low in the absence of talin or integrin activation. After inside-out activation of integrins (e.g. via activated talin that binds to the appropriate cytoplasmic motif), increased binding of PAC-1 antibody can be detected by FACS (Tadokoro et al., 2003). It has recently been suggested that this system measures not only integrin affinity and conformational changes but also integrin clustering (Bunch, 2010). (C) αvβ3-GFP integrin clustering in B16F1 melanoma cells after affinity switching by Mn2+ treatment. When plated on vitronectin, fibronectin or serum, only a few integrins are recruited into focal adhesions, while others remain in their low affinity state in the plasma membrane (left side). Upon Mn2+ addition, extended integrins start to form clusters in contact with substrate-bound ligands in response to interaction with talin and PI(4,5)P2-containing membranes (Cluzel et al., 2005; Saltel et al., 2009). Importantly, these de novo integrin clusters are not associated with the actin cytoskeleton, which allows separating integrin and F-actin-dependent mechanisms of focal adhesion formation.

Another classical assay that has advanced the understanding of talin recruitment to ligand-bound integrins has been developed based on a ligand-mimetic IgM antibody, PAC-1. At the surface of integrin-transfected CHO cells, this antibody detects the high affinity conformation of integrin receptors by FACS, notably induced by mutations at the juxtamembrane region, or by the expression of the head domain of talin (Tadokoro et al., 2003) (Figure 3B). Due to the multiple binding sites of this antibody, it is likely that not only changes in integrin affinity but also integrin valency (integrin clustering) are measured (Bunch, 2010). Valency changes in integrin receptors can also be recorded by measuring the degree of clustering of fluorescently tagged integrin receptors bound to immobilized ligands (Figure 3C). In contrast to the FACS assay, which is performed with cells in suspension, here B16F1 melanoma cells are analyzed in their normal substrate-adherent state. In response to the addition of Mn2+, which switches integrins into their high affinity conformation, talin-dependent, but F-actin independent, clustering can be measured (Cluzel et al., 2005). Using this system, the talin–integrin as well as the talin–lipid interactions were analyzed in intact cells, demonstrating that integrin clustering depends on protein–protein as well as protein–lipid interactions (Saltel et al., 2009) (see below).

Details of talin-dependent integrin clustering

Talin is an auto-inhibited F-actin and integrin-binding adapter protein of the FERM (band F1, ezrin, radixin, moesin) family (Critchley, 2009). The integrin-binding talin N-terminal FERM domain is followed by a long C-terminal tail domain formed of multiple helical bundles. The F-actin-binding domain is localized at the extreme C-terminus and requires dimerization (Gingras et al., 2008). Within these helical bundles, many vinculin-binding motifs have been identified that can be exposed upon mechanical stretching applied between the N-terminal integrin and C-terminal F-actin binding site (Hytonen and Vogel, 2008; del Rio et al., 2009). Based on this experiment, talin and vinculin are stress-sensitive integrin adapter proteins able to transform mechanical stress between extracellular ligands and the actin cytoskeleton into the remodeling of adhesion sites (Grashoff et al., 2010; Vogel and Sheetz, 2009).

Similar to radixin and ezrin, the auto-inhibition of talin can also be relieved by the interaction of PI(4,5)P2 lipids with the N-terminal FERM domain (Goksoy et al., 2008; Goult et al., 2009; Hamada et al., 2000; Niggli, 2005; Saltel et al., 2009). In addition to activation, PI(4,5)P2 lipid binding also orients these adapters at the plasma membrane to allow defined interactions with cytoplasmic tails of transmembrane receptors.

When radixin binding to ICAM-2 (Hamada et al., 2003) is compared with the integrin-tail binding of talin (Saltel et al., 2009), several differences as well as similarities are apparent (Figure 4). First, the PI(4,5)P2 lipid binding site is rotated with respect to the conserved ligand binding pocket of the PTB domain (F3), compensating the different length of the juxtamembrane domains of the respective receptor. Whereas the membrane proximal domain of integrins binds as a helix, ICAM-2 forms an antiparallel β-sheet.

Figure 4.

 Comparison of lipid and receptor binding between talin and radixin. (A,D) Cartoon of the association of the talin (A) and radixin (D) FERM domains with the plasma membrane (blue surface) and the cytoplasmic tail of β3-integrin (A) and ICAM-2 receptor (D) (magenta), respectively. (B,C,E,F) Color-coded Pymol models of the cartoons seen in A and D, respectively. Basic residues involved in PI(4,5)P2 binding in talin (B,C) and the IP3 molecule, representing the PI(4,5)P2 head domain in radixin (E,F), are shown as spheres. The putative membrane proximal salt-bridge is circled in red. The PTB (F3 domain) binding residue in integrin (Tyr) and ICAM-2 (Trp) are represented as yellow sticks. Note the conservation of the position of the plasma membrane in B and E and the conservation of the position of the F3 domain in C and F. The talin/integrin model is from Saltel et al. (2009). The radixin model is generated from overlapping PDB structures 1GC6 and 1J19 (Hamada et al., 2000, 2003). A recent structure-based model of an internal deletion mutant of the talin-head domain suggests that the F0–F1 sub-domains can detach from the F2–F3 domains. Further functional data are required to validate this model as it exhibits the solvent exposure of a highly conserved Trp side chain, otherwise buried within the F1-F2-F3 interface in all other FERM domains (Elliott et al., 2010).

An interesting similarity concerns also a juxtamembrane salt-bridge between the adapter protein and the cytoplasmic tail. Due to the proximity of the highly charged phospholipid head groups, it is not clear how much these salt-bridges contribute to the affinity of the adapter protein binding. However, in the case of β3-integrin the indicated aspartic acid is an important regulatory residue, which is involved either in clasping the α-integrin subunit in the closed integrin conformation or in binding to talin in the high-affinity conformation (Saltel et al., 2009). This double role of the aspartic acid in the cytoplasmic tail of integrins, as well as the auto-inhibition controlling PI(4,5)P2-binding site in talin, ensures that the formation of the integrin/talin/PI(4,5)P2 complex is perfectly reversible as well as tightly controlled by the microenvironment of the plasma membrane at the focal adhesion site (Saltel et al., 2009).

As the extended talin conformation is further stabilized by vinculin and F-actin association, the talin/integrin/lipid complex ensures the lateral clustering of integrins as well as regulation by mechanical stresses via the actin cytoskeleton. This creates a complex that can integrate a number of inputs, consisting of the dynamic aspects of the actin cytoskeleton, the chemical modification of plasma membrane lipids, as well as the physical properties and mechanical anchoring of the integrin ligand (Figure 5). As the clustering of integrins is responsible for the recruitment of the signaling adapter proteins FAK and paxillin, integrin-dependent intracellular signaling also depends on the above-mentioned parameter.

Figure 5.

 Regulation of integrin clustering by interfacing with the acto/myosin and PI-lipid system. Graphical representation of the regulatory pathways involved in integrin clustering and formation of focal adhesions. Individual state changes are shown as equilibrium reactions that are influenced by the microenvironment of the integrin receptors or focal adhesions. (I) Integrin bending or extending occurs spontaneously or can be induced by integrin ligands or adapter protein binding. (II) The extended conformation of integrins is stabilized by kindlin and talin (Ma et al., 2008; Montanez et al., 2008), the latter regulated by PI(4,5)P2 binding (Goksoy et al., 2008; Goult et al., 2009; Saltel et al., 2009). However, in the absence of immobilized ligand such extended integrins remain unstable precursors of focal complexes or adhesions (dashed box). (III) The polymerization of F-actin together with the recruitment of additional adapter proteins such as vinculin (see text) assures the establishment of a mechanical link that is used for substrate sensing and intracellular signaling (e.g. via FAK recruitment). (IV) Acto/myosin or external stresses will induce tension-dependent maturation of focal complexes to focal adhesion (Ballestrem et al., 2001), thereby increasing adhesion signaling and causing anchorage-dependent cell survival (Wehrle-Haller and Imhof, 2002). In parallel with the remodeling of the actin cytoskeleton (blue), the PI-lipids are modified in trans-Golgi and plasma membranes (red). PIPKI-α and PIPKI-γ661 are involved in PI(4,5)P2 synthesis in lamellipodia (Chao et al., 2010) and in focal adhesions in a talin-dependent manner, respectively (Di Paolo et al., 2002; Ling et al., 2002; de Pereda et al., 2005). In turn, PI(4,5)P2 regulates talin activation as well as F-actin polymerization and, in addition, serves as a substrate for PI3 kinases to produce the mobility- and survival-inducing PI(3,4,5)P3 lipid. For clarity, only a subset of PI lipid modifications are shown.

Integrins as sensors and transducersof the mechanical state of the tissue environment

Due to the role of integrins as sensors of the mechanical properties of the environment, it appears plausible that physical changes in the extracellular matrix influence cell behavior and migration. For example, when fibroblasts are plated on a polyacrylamide gel, exhibiting a gradient in stiffness, cells will always turn to and migrate onto the stiffer surface (Pelham and Wang, 1997). This sensing capacity indicates that cells can transform mechanical perception into intracellular signaling that modifies cell adhesion and migration (Wehrle-Haller and Imhof, 2002).

Consistent with increased signaling in response to stiffer extracellular matrix, tumor stroma have been recognized to have altered mechanical properties compared to normal tissue (Paszek et al., 2005). In addition, augmenting collagen cross-linking induces the growth of tumors and increases integrin-dependent signaling (Levental et al., 2009).

Another aspect of stiffness-dependent signaling operates during differentiation of stem cells. Whereas neuronal differentiation is favored on soft substrates, increasing mechanical resistance will cause the induction of the myogenic lineage. On even stiffer substrates, osteogenic differentiation is induced (Discher et al., 2005; Engler et al., 2006). This demonstrates that intracellular signaling in response to integrin-dependent adhesion requires delicate tuning of the mechanical properties of the extracellular environment to allow appropriate morphogenesis (Engler et al., 2009).

It is this last point that could be used to explain why the expression of αvβ3 integrins and not any other integrin is increased in melanoma and correlates with induction of the vertical growth phase (Albelda et al., 1990; Hsu et al., 1998). Integrins expressed in neurons as well as in differentiated melanocytes such as α6β1 or α3β1 may only allow mechanical coupling to the ECM under low mechanical stresses. In contrast, α5β1 integrins, which are involved in the synthesis of fibronectin fibrils, bind simultaneously to the synergy as well as the RGD sites in fibronectin. This additional binding functionality of the α5β1 integrin ectodomain (absent in the RGD-binding αvβ3 integrin), may create outside-in signaling that can modify conformational flexibility of the integrin, thereby influencing cell adhesion and migration (Friedland et al., 2009).

The highly dynamic association of αvβ3 with protruding or retracting adhesive structures (Cluzel et al., 2005), its ability to recognize a large panel of extracellular matrix components, as well as to associate constitutively with the non-receptor tyrosine kinase c-src (Arias-Salgado et al., 2003), makes αvβ3 a unique integrin to support cell migration, proliferation and invasion events during embryogenesis as well as pathological situations such as melanoma growth and invasion (Desgrosellier et al., 2009).

To conclude, these last examples demonstrate that integrin receptors play specific and critical roles in sensing and perceiving changes in the physical state of the tissue environment. Although we do not yet understand all of the relevant signaling pathways involved, it is evident that melanocytes and melanoma cells are embedded in a complex tissue formed by different cell types. This complex environment will influence melanocyte behavior and needs to be considered when developing strategies for melanocyte regeneration or melanoma prevention. Due to their origin from the neural crest and their extensive migration capacities, melanocytes have a unique ability to adapt and to react to changes in their environment, making them one of the most fascinating cells in our bodies.

Acknowledgements

We like to thank Dr. Pierre Cosson for discussion and Nevus Outreach, the Swiss Foundation for Research on Myopathies and the Swiss National Science foundation (31003A-130742) for financial support.

Ancillary