Networks and crosstalk: integrin signalling spreads

Multicellular organisms must coordinate signals from adhesion receptors with those from other signalling receptors (for example, growth factor receptors). Here, we briefly review paradigms of integrin–adhesion-receptor signalling. We discuss how adhesive signalling is coordinately regulated through intersecting networks. We also examine some examples of how some forms of integrin crosstalk may lead to unforeseen and potentially deleterious responses.

The coordination of cell–cell and cell–matrix adhesion is essential for the development and proper functioning of all multicellular organisms. In the last 20 years, several families of cell-surface adhesion receptors and their ligands have been defined. Major new challenges have arisen from the appreciation that adhesion receptors are bona fide signalling receptors that function as 'proprioceptors', which inform the cell about the physical and chemical nature of its surrounding environment. The coordination of adhesion-receptor signals with those that come from more traditional signalling receptors, such as growth factor receptors, is essential to how organismal body plans are specified — perturbing this coordination can lead to events such as malignant transformation. Here we briefly review paradigms in adhesion-receptor signalling using examples from the integrin receptor family. We then discuss the coordinate regulation of adhesion signalling, describing general patterns of adhesion-receptor signalling networks. Finally, we examine how certain forms of crosstalk — defined in the Merriam Webster Collegiate Dictionary as “unwanted signals in a communication channel (as in telephone, radio or computer) caused by transfer of energy from another circuit (as by leakage or coupling)” — between integrins or between integrins and growth-factor receptors may lead to unforeseen and potentially deleterious biological responses.

Signalling by integrins

Integrin engagement during cell adhesion regulates gene expression, and cell growth, differentiation and survival¹. These events are controlled by biochemical signals generated by ligand-occupied and clustered integrins. Because many of the biochemical elements of integrin signalling pathways are shared with growth-factor receptors and other classes of receptors, integrins are genuine signalling receptors. But integrins have three additional features that dramatically expand their range of functionality.

First, integrin ligands are generally immobilized. Consequently, integrin signalling is usually localized to topographically discrete regions of the plasma membrane. Second, integrin cytoplasmic domains provide critical anchors for the actin cytoskeleton, and therefore provide a physical connection between internal and external filament systems. Third, the capacity of many integrins to bind their ligands is regulated by cellular signalling mechanisms through a process often called integrin activation or inside-out signal transduction. An increase in ligand-binding affinity can arise because of a change in the integrin conformation. Alternatively, because integrin ligands are usually multivalent, integrin clustering can also increase affinity as a result of cooperative binding, sometimes referred to as changes in receptor avidity. There are clear-cut examples of both types of regulation. Integrins are, therefore, signalling receptors that transmit information in both directions across the plasma membrane.

Consequently, integrins provide an intersection where mechanical forces, cytoskeletal organization, biochemical signals and adhesion can meet. Mechanical stresses transmitted by integrins influence the cytoskeleton, the cytoskeleton affects integrin activity and cell adhesion, and cell adhesion regulates the cytoskeleton and the mechanical forces that cells exert on their surrounding extracellular matrix (ECM). All of these systems affect and are affected by biochemical signalling pathways. This intersection is therefore a critical element of morphogenesis in which specification of cell shape, polarity and motility are tied to gene expression and cell function.

Integrating downstream signals

Because integrins and growth factors share many common elements in their signalling pathways, it is clear that there are many opportunities for integrin signals to modulate growth-factor signals and vice versa. But, in fact, integrins enable growth-factor signalling in many cases — that is, normal growth-factor signalling does not occur unless cells are adherent to the ECM or other cells through integrins. The mitogen-activated protein (MAP) kinase pathway provides the best characterized example of this principle, because here a number of integrin and growth-factor signals converge at multiple points. In this pathway, activation of Ras leads to sequential stimulation of the protein kinases Raf, MEK and finally the MAP kinases Erk1 and Erk2.

Adding soluble mitogens to cells in suspension triggers weak or transient activation of the MAP kinases Erk1 and Erk2, compared with the strong and sustained Erk activity observed in adherent cells (reviewed in refs 2 and 3). In vivo, endothelial cells treated with integrin α_vβ₃ antagonists also show diminished long-term Erk activation, which supports the biological relevance of the in vitro results⁴. Nearly all the published studies agree that activation of growth-factor receptors per se is unaffected or modestly affected by adhesion. Beyond that, however, there is a curious lack of consensus. These disagreements suggest that multiple steps in the MAP kinase cascade have integrin-dependent components, but that different steps are rate limiting depending on the precise cell types and conditions.

Loss of integrin-mediated adhesion can affect the phosphorylation of a specific tyrosine residue in the platelet-derived growth factor receptor (PDGFR), which leads to increased recruitment of p120Ras-GTPase activating protein (GAP) and downregulation of Ras activity⁵. Other studies have reported that activation of Ras is unchanged in nonadherant cells, but that Raf or MEK activation are affected. Raf activation was found to be downstream of integrin-regulated p21-activated kinase (PAK), whereas MEK activation is downstream of integrin-regulated focal adhesion kinase (FAK)^3,6. At another level of regulation, even under conditions where Erk is activated in suspended cells, Erk fails to translocate to the nucleus or phosphorylate the transcription factor targets that are critical for cell-cycle progression⁷. Interestingly, treating cells with cytochalasin to disrupt actin filaments and focal adhesions has effects very similar to loss of adhesion, including decreased Erk, PAK and FAK activity. Though these mechanistic details may vary, Erk is often a key mediator of effects on the cell cycle, primarily because it is required for cyclin D expression². These observations enforce the view that integrins connect to growth factor pathways through a web of potential interactions, the actual usage of which varies depending on the cell type and environmental conditions. It is clear, however, that the ultimate outcome is affected by the strong synergy between soluble stimuli and cell adhesion.

The physical side of cell proliferation

Integrins may also be involved in mediating the integratation of mechanical stresses with cellular responses to soluble regulatory factors. In many cases, mechanical stresses are transmitted to cells by changes in the rigidity of the ECM, sensed by integrins. Studies dating back to the early 1980s showed that detaching collagen gels from the dish allowed their contraction by the cells into small rafts (so-called floating collagen gels). This release of tension promoted epithelial cell differentiation and inhibited mitogenesis (reviewed in ref. 8). Conversely, attached collagen gels that resist contraction inhibit differentiation and promote proliferation. Recent studies using fibroblasts showed that release of tension by detaching collagen gels resulted in altered responses to growth factors, specifically decreased PDGF-induced activation of the MAP kinase pathway⁹. These changes occurred without any change in the amount of growth factor that was binding to cells, similar to effects seen in detached cells.

Cells sense tension through their ECM contacts and increase contractile force in response to either externally applied force or rigid substrates that resist deformation by cell-generated forces^10,11. This force increase is accompanied by accumulation of focal adhesion proteins at the site of attachment. Therefore, cells on rigid substrates or those in vivo exposed to strong mechanical forces make focal adhesions¹². Additionally, a myosin-dependent force is required to assemble a fibronectin (FN) matrix outside the cell, which also promotes accumulation of focal adhesions. Conversely, on malleable matrices that do not resist tension, cell-generated forces and focal adhesions are decreased. This regulatory loop has the characteristics of a feedback mechanism that leads to homeostasis.

These considerations lead us to propose a general model. First, we know that rigid matrices promote high contractility, leading to assembly of focal adhesions and highly clustered integrins. These conditions may also promote assembly of an ECM, that through increased rigidity or attachment to other structural elements, resists contraction. Second, we know that phosphorylation of FAK and perhaps other focal adhesion proteins will be high and will facilitate growth-factor activation of Erks to promote proliferation. Malleable matrices, in contrast, will induce a low contractility state. Cells will assemble a distinct type of ECM that further stabilizes the low contractility state. This condition will be accompanied by a distinct organization of integrins, focal adhesion proteins and the cytoskeleton characterized by lower FAK phosphorylation and activity. Therefore, Erk activity stimulated by growth factor will be lower, and differentiation will be favoured over proliferation. This hypothetical model may explain how integrins may integrate adhesive, humoral and mechanical aspects of the environment, allowing cells to respond accordingly.

Integrin ligation and angiogenesis

In addition to promoting cell growth, cell adhesion can act to suppress signalling through growth-factor pathways. This phenomenon is best understood by looking at the interactions of cadherins and the Wnt pathways, in which the cytoplasmic levels of free β-catenin are regulated in opposing directions by cadherins and Wnt. There are also examples of integrin-dependent growth suppression¹³. Recent work provides strong evidence that, under some circumstances, occupancy of integrins may suppress the signalling pathways that regulate angiogenesis. Integrin α_vβ₃ antagonists can inhibit angiogenesis, a result that suggests that these receptors are crucial in angiogenesis¹⁴. Furthermore, several natural inhibitors of angiogenesis have proved to be ligands for these integrins^15,16,17. But genetic deletion of α_v integrins has only modest effects on angiogenesis¹⁸. Two recent papers have provided provocative new insights into this apparent paradox. Stupack et al.¹⁹ described 'integrin-mediated death' (IMD) whereby an unligated integrin promotes apoptosis of cells. The proposed mechanism for IMD involves recruitment of caspase-8 to unligated integrins, thereby initiating a novel apoptotic pathway. Reynolds et al.²⁰ reported the surprising finding that genetically deleting integrins β₃ and β₅ enhances normal and pathological angiogenesis. This compelling study clearly suggests that endothelial signalling reactions might be suppressed by α_v integrins. One possibility is that the absence of β₃ and β₅ integrins reduces IMD. Alternatively²⁰, the absence of β₃ and β₅ integrins could upregulate signalling through Flk-2, a receptor for vascular endothelial growth factor (VEGF). A third possibility is a form of crosstalk called trans-dominant integrin inhibition.

Trans-dominant integrin inhibition is defined as suppression of the function of one integrin as a consequence of ligand binding to a second integrin (the suppressive integrin). Soluble integrin-specific ligands are particularly liable to cause trans-dominant suppression of other integrins when the suppressive integrin is abundant²¹. Angiogenic endothelial cells are greatly enriched in integrin α_vβ₃ (ref. 22). If β₃ integrins are occupied, the functions of integrin α₅β₁ can be suppressed²³, including the capacity of α₅β₁ to block apoptosis²¹. In contrast to α_v integrins, integrin α₅β₁ is essential for angiogenesis²⁴. Tumstatin, an angiogenesis inhibitor that binds to α_vβ₃, blocks adhesion-dependent activation of Akt²⁵, an important cell-survival signal. This inhibition seems to depend on tumstatin binding to α_vβ₃, suggesting that the survival signal is the result of α_vβ₃-dependent trans-dominant inhibition of integrin signalling. These recent data support the idea that the abundance of α_vβ₃ on angiogenic endothelial cells may account for their unusual susceptibility to α_vβ₃-ligand-induced trans-dominant inhibition and consequent apoptosis²¹. Thus, trans-dominant integrin inhibition, a form of crosstalk, may account for some of the effects of α_vβ₃ ligands and antagonists on angiogenesis.

Coordinate regulation

The function of adhesion receptors can be regulated in response to growth factors and other agonists. One intriguing example is the regulation of one adhesion receptor by another. This sort of networking may permit a cell to fine-tune its behaviour to the topographical patterning of adhesive targets. Integrins initiate many biochemical signals that can regulate the function of other integrins. There is, therefore, an almost unlimited number of possibilities by which one adhesion receptor could regulate the function of another.

A particularly provocative example of such trans-regulation is in the interplay between leukocyte integrins α₄β₁ and α_Lβ₂. Both of these integrins are central to the transport of leukocytes from the bloodstream into the peripheral tissues²⁶. ICAM-1 and ICAM-2, ligands for β₂ integrins, are constitutively expressed on many endothelial cells. In contrast, VCAM-1, a major α₄β₁ ligand, is not expressed on most endothelial cells but is upregulated by a variety of inflammatory events. Very small quantities of VCAM-1 on an ICAM-1 substrate dramatically enhances β₂-integrin-dependent cell migration and adhesion^27,28,29. In this setting, α₄β₁ may act as a signalling receptor in response to its agonist, VCAM-1. Furthermore, the affinity of α₄β₁ regulates the sensitivity of cells to different doses of VCAM-1 (ref. 30). Thus, when α₄β₁ is activated, there is a profound increase in the sensitivity of leukocytes to VCAM-1 stimulation of leukocyte migration. This type of regulation could contribute to the specificity of leukocyte transmigration at different vascular sites. For example, natural killer cells express constitutively active integrin α₄β₁ (ref. 31) and migrate into early inflammatory lesions where endothelial VCAM-1 expression is low. This example of coordinated integrin regulation defines a paradigm in which inside-out integrin signalling controls outside-in signalling.

Recent work has shown that mechanical stimuli can also influence integrin activity. Fluid shear stress applied to endothelial cells results in rapid conformational activation of integrin α_vβ₃ (ref. 32). These newly activated integrins subsequently bind to available ECM proteins and initiate signals characteristic of new adhesions such as transient downregulation of Rho³² and phosphorylation of Shc and FAK³³. These integrin-dependent signals account for a significant subset of the responses to shear stress including cell and cytoskeletal alignment.

Does integrin crosstalk cause cancer?

We have defined crosstalk as unwanted signals in a communication channel caused by transfer of energy from another circuit. In biological signalling networks one might consider adding a less stringent term, such as 'unanticipated', to this definition. The trans-activation of growth-factor receptors by integrins provides an example of signals that can lead to unexpected consequences, and therefore represent potential examples of authentic crosstalk.

Cell adhesion through integrins leads to ligand-independent phosphorylation of several different growth-factor receptors including those for PDGF, fibroblast growth factor (FGF), VEGF, hepatocyte growth factor (HGF) and epidermal growth factor (EGF)³⁴. This trans-activation of growth-factor receptors is often transient, and of a smaller magnitude than that initiated by the cognate growth factor. This kind of activation is associated with the formation of complexes containing both integrins and growth-factor receptors. Furthermore, in certain cases including cell adhesion, such activation is necessary for maximal growth-factor-dependent receptor phosphorylation. Integrin ligation may, therefore, be required for optimal growth-factor-receptor activation and this mechanism may be important in the regulation of cell growth by integrins. Indeed, Moro et al.^35,36 showed that an EGF receptor (EGFR) kinase inhibitor can block adhesion-dependent activation of the Erk MAP kinase in cells that express high levels of the EGFR. This is true in EGFR-transfected 3T3 cells. Surprisingly, adhesion of untransfected 3T3 cells also resulted in Erk activation, but in this case activation was independent of EGFR kinase activity. These contrasting results highlight the dependence of signalling networks on cell type, and underscore the importance of context in interpreting experimental data.

What is the biological significance of trans-activation of growth factor receptors by integrins? One possibility is that it represents one of the adhesion checkpoints in growth-factor signalling. But in most nonadherent cells, growth factors can activate their receptors, which leads to phosphorylation and recruitment of relevant downstream effectors. Secondly, Src kinase family members have been implicated in trans-activation³⁵. Despite this, even in the absence of the three most abundant Src kinases, PDGF can elicit activation, downstream signalling and mitogenic responses from the PDGFR³⁷. The purpose of integrin trans-activation may, therefore, be to modulate, rather than control growth-factor signalling.

A fascinating implication of these results is that a significant fraction of integrin signalling may be mediated by trans-activation of growth-factor receptors. Indeed, such a model is strongly supported by the dependence of adhesion-induced Erk activation on the activity of the EGFR in certain cells. But activation of Src kinases and pp125FAK are independent of EGFR³⁵. Furthermore, trans-activation of EGFR by integrins depends on Src kinases, and the absence of Src kinases profoundly suppresses integrin signalling³⁷. But Src kinases are also involved in integrin activation of Erk through the EGFR-independent pp125^FAK and Shc pathways^34,38. Thus, the relative impact of trans-activation of growth-factor receptors on integrin signalling needs additional analysis that takes into account cellular context. The insightful dissection of the connections between integrins and EGFR signalling by Moro et al.³⁵ should provide starting points for experimental approaches to probe this question.

A third possibility suggested by this phenomenon is that trans-activation of growth-factor receptors by integrins can lead to unwanted consequences such as tumorigenesis. In quiescent, adherent cells, integrins are ligated by binding to the substrate. Furthermore, when such cells are subjected to mechanical stimulation, trans-activation of growth-factor receptors occurs³⁹. Cells could, therefore, mistake signals from integrins as signals from growth factors. Such confusion could lead to inappropriate cell growth. Indeed, innovative studies by Wang et al.⁴⁰ showed that cell adhesion, presumably through integrins, could activate human c-Met, the receptor for HGF. Furthermore, hepatocyte-specific overexpression of c-Met resulted in hepatocellular carcinoma in mice. Importantly, this effect of c-Met is independent of the binding of HGF, suggesting that integrin-dependent trans-activation could be responsible for tumorigenesis. Thus trans-activation of growth factor receptors by integrins may lead to dysregulated cellular growth.

How might the cell protect itself from tumorigenesis arising from trans-activation of growth-factor receptors by integrins? One possibility is that there may be mechanisms to limit such trans-activation. Integrin trans-activation may lead to phosphorylation of a different subset of tyrosine residues on the growth-factor receptor³⁵ with consequent differences in downstream signalling. But because trans-activation uses the endogenous kinase activity of the receptor, the reported differences in tyrosine phosphorylation may reflect quantitative rather than qualitative differences in signalling. Secondly, trans-activation in response to adhesion is brief in duration unless the growth-factor receptors are overexpressed. The brevity of the response could prevent inappropriate growth, and the mechanisms terminating the response are unclear. An intriguing possibility is that specific inhibitors of trans-activation exist. Loss-of-function of gene products responsible for limiting trans-activation could lead to tumour formation. Thus, additional analysis of trans-activation and of its regulation may lead to new insights into malignant transformation and the discovery of new tumour suppressor genes.

Perspectives

Recent studies have provided numerous instances where integrin signals influence the function of other cell-surface receptors. Integrins can either inhibit or activate other integins, in some cases this results in local modulation of cell adhesiveness. These effects may account for the potency of integrin antagonists for inhibiting angiogenesis. Integrins most often synergize with growth-factor pathways to enhance their activity. Detailed analysis of the MAP kinase pathway has revealed multiple points of intersection, suggesting a complex network of interactions. The importance of cytoskeletal organization and mechanical tension for integrin signalling thereby provides a mechanism by which these considerations can influence the responses of cells to growth factors as well, providing a convergence point for different classes of extracellular cues. Integrins can also cause ligand-independent trans-activation of tyrosine-kinase growth-factor receptors. Effects of this type have the potential to broaden the range of integrin signals but, if uncontrolled, may result in tumorigenesis.

The picture emerging from these studies is that integrins function as nodes within webs of signalling, adhesive and cytoskeletal pathways. These networks enable integrin-mediated adhesion to participate in highly regulated processes such as migration, growth control and morphogenesis. But like all systems, greater complexity also affords greater opportunities for subversion. Understanding these circuits in detail should shed light on a variety of pathological states, as well as revealing the intrinsic robustness of homeostasis.

Figure 1: Integrins can intersect the MAP kinase pathway at multiple points.

Growth-factor receptors initially trigger recruitment of the Grb2/Sos complex to the plasma membrane leading to sequential activation of Ras, Raf, MEK and Erk, which translocates to the nucleus to phosphorylate Elk-1 and other transcription factors. Integrin-mediated adhesion enhances the transmission of this signal at the level of recruitment of RasGAP and downregulation of Ras activity, activation of Raf, activation of MEK and nuclear translocation of Erk. See text for details.