Accumulating evidence indicates that there is extensive crosstalk between integrins and TGF‐β signalling. TGF‐β affects integrin‐mediated cell adhesion and migration by regulating the expression of integrins, their ligands and integrin‐associated proteins. Conversely, several integrins directly control TGF‐β activation. In addition, a number of integrins can interfere with both Smad‐dependent and Smad‐independent TGF‐β signalling in different ways, including the regulation of the expression of TGF‐β signalling pathway components, the physical association of integrins with TGF‐β receptors and the modulation of downstream effectors. Reciprocal TGF‐β–integrin signalling is implicated in normal physiology, as well as in a variety of pathological processes including systemic sclerosis, idiopathic pulmonary fibrosis, chronic obstructive pulmonary disease and cancer; thus, integrins could provide attractive therapeutic targets to interfere with TGF‐β signalling in these processes.
See Glossary for abbreviations used in this article.
Integrins—which consist of an α‐subunit and a β‐subunit—constitute a family of transmembrane receptors that bind extracellularly to the ECM and intracellularly to the cytoskeleton, thereby ‘integrating’ the extracellular environment with the cell interior (Hynes, 2002; Legate et al, 2009). Integrins transduce signals from the outside into the cell and vice versa to regulate cell adhesion and cell spreading, as well as migration, proliferation, differentiation and remodelling of the ECM. In addition, integrins can modulate the signalling cascade elicited by several growth factors, including TGF‐β. The TGF‐β isoforms TGF‐β1, TGF‐β2 and TGF‐β3 are pleiotropic cytokines that mediate a variety of effects on a range of cell types. TGF‐βs bind to a heterodimeric serine/threonine kinase receptor complex—which consists of TGF‐βRI and TGF‐βRII—leading to the recruitment and phosphorylation of the intracellular effector proteins Smad2 and Smad3. Phosphorylated Smad2 and Smad3 subsequently bind to Smad4 and translocate to the nucleus to initiate gene expression. TGF‐β signalling is negatively regulated by inhibitory Smads, including Smad6 and Smad7 (Massagué & Chen, 2000). In addition, TGF‐β can affect numerous signal transduction pathways in a Smad‐independent manner. Although the effect of TGF‐β signalling depends on the context and cell type, TGF‐β clearly controls a vast number of transcriptional targets, many of which are integrins and their ligands. The connection between integrins and TGF‐β is therefore bidirectional, and it is becoming increasingly clear that it is relevant in many physiological and pathological phenomena. Here, we discuss the integrin–TGF‐β interplay and highlight its importance in fibrosis, cancer and wound repair.
Integrin regulation by TGF‐β
TGF‐β controls the transcription of genes that encode numerous integrins (Table 1) in several cell types and tissues, as well as in various human cancers. Although the downregulation of integrin expression—mostly laminin receptors—has also been reported, in most cases TGF‐β stimulates integrin expression. Intriguingly, the induction of integrin expression by TGF‐β can be driven by cooperative signalling between the integrin and TGF‐β, thereby creating a feedforward loop (Pechkovsky et al, 2008). TGF‐β not only regulates the expression of integrin ligands—including βig‐h3, tenascin, vitronectin, fibronectin, and several members of the laminin and collagen families—but also stimulates the expression of integrin‐associated proteins—including disabled 2, ILK, kindlin 1, paxillin and PINCH—which could increase integrin activation. Therefore, the transcriptional control exerted by TGF‐β can strongly affect integrin‐mediated processes. Finally, TGF‐β could also directly regulate integrin activation, by a still unidentified ‘inside–out’ mechanism (Fransvea et al, 2009).
Regulation of TGF‐β activation by integrins
TGF‐β is secreted in an inactive (latent) form in a complex with two proteins—LAP and LTBP. Its activation requires the dissociation from the complex, which occurs at low pH or through the action of reactive oxygen species, proteases, thrombospondin 1 or several integrins. The LAPs of TGF‐β1 and TGF‐β3—but not of TGF‐β2—contain an RGD motif that can potentially be bound by the five αv‐containing integrins and αIIbβ3, α5β1 and α8β1. Integrin binding to LAP has been demonstrated formally for α8β1 and all αv‐integrins, although binding of α8β1 does not seem to lead to activation, and whether αvβ1 can activate TGF‐β is also unclear (Table 2; Munger et al, 1999; Lu et al, 2002; Ludbrook et al, 2003). Integrin‐mediated TGF‐β activation seems to be possible in a protease‐dependent or protease‐independent manner. Protease‐dependent TGF‐β activation has only been demonstrated for αvβ8 and depends on the binding of the integrin to the RGD site in LAP and simultaneous recruitment of MMP14, which then releases TGF‐β by proteolytic cleavage (Fig 1A; Mu et al, 2002). This mode of activation does not require that the activating cell and target cell be in close proximity. Interestingly, αvβ3 can be a docking site for MMP2 and MMP9 (Brooks et al, 1996; Rolli et al, 2003), although whether this also leads to proteolytic activation of TGF‐β remains to be seen. Notably, the genes for these MMPs are TGF‐β targets and, therefore, a self‐amplifying TGF‐β feedforward loop could be envisioned. Non‐proteolytic TGF‐β activation occurs through cell traction forces exerted by the actin cytoskeleton. These forces are translated by integrins into a conformational change of the TGF‐β–LAP–LTBP complex, leading to the presentation of active TGF‐β to its receptor (Annes et al, 2004; Fontana et al, 2005; Wipff et al, 2007; Wipff & Hinz, 2008). Hence, non‐proteolytic activation requires cytoskeletal integrity, the connection of the β‐tail of the integrin to the cytoskeleton, a mechanically resistant matrix, the interaction between LAP and the ECM through LTBP (Fig 1B), and that the target cell be in the direct vicinity of the activating cell. Non‐proteolytic activation has been demonstrated in vitro for αvβ3, αvβ5 and αvβ6, as well as for a β1‐integrin with a still unidentified α‐subunit (Wipff et al, 2007). Whether or not the activation of TGF‐β by a β1‐integrin is relevant physiologically remains controversial.
The activation of TGF‐β by integrins can also be initiated by G‐protein‐coupled receptors. For example, the stimulation of PAR1 with thrombin leads to RhoA‐dependent and ROCK‐dependent TGF‐β activation by integrin αvβ6 in vitro and in vivo (Jenkins et al, 2006). Similarly, PAR1 stimulation with coagulation factor X induces αvβ5‐regulated TGF‐β activation through ROCK signalling (Scotton et al, 2009). Furthermore, αvβ6‐mediated TGF‐β activation can be induced by lysophosphatidic acid signalling to RhoA and ROCK, through the lysophosphatidic acid receptor coupled to small G protein Gαq (Fig 1B; Xu et al, 2009). Whether other integrins mediate TGF‐β activation through similar signalling pathways remains to be established.
The importance of integrin‐mediated activation of TGF‐β in vivo is evident, as mutation of the RGD site of LAP leads to defects similar to those observed in TGF‐β1‐null mice (Yang et al, 2007). In addition, genetic ablation of the β6‐subunit, or conditional deletion of αv or β8 from dendritic cells, causes exaggerated inflammation as a result of impaired TGF‐β signalling (Lacy‐Hulbert et al, 2007; Travis et al, 2007). The phenotype of mice lacking both the αvβ6 and αvβ8 integrins recapitulates the abnormalities observed in TGF‐β1 and TGF‐β3—but not in TGF‐β2—knockout mice, indicating that the integrins αvβ6 and αvβ8 can account for the full activation of TGF‐β1 and TGF‐β3 in vivo (Aluwihare et al, 2009). Indeed, mice lacking β3, β5, or both do not develop abnormalities similar to those due to deficient TGF‐β signalling (Hodivala‐Dilke et al, 1999; Huang et al, 2000; Reynolds et al, 2002). Nevertheless, αvβ3‐mediated or αvβ5‐mediated TGF‐β activation could be important in pathological conditions, as increased expression of both of these integrins is observed in the dermis of scleroderma patients, and these integrins elicit autocrine TGF‐β signalling in patient fibroblasts in vitro (Asano et al, 2005a, 2006a). In addition, TGF‐β activation by αvβ5 is important in pulmonary fibrosis, as discussed below. However, a causal effect of αvβ3‐mediated TGF‐β activation in human pathology has not yet been established.
Regulation of TGF‐β signalling by integrins
In addition to the direct activation of TGF‐β, several integrins seem to influence TGF‐β‐induced signal transduction (Table 2). The effect is almost exclusively an amplification of the signal, that is, increased activation of signalling proteins and/or increased expression of TGF‐β target genes. The regulation of TGF‐β signalling by integrins occurs at several levels. Integrin‐mediated adhesion can potentiate TGF‐β‐induced signalling and gene expression, in an analogous way to how integrins regulate growth factor signalling through receptor tyrosine kinases. Indeed, TGF‐β‐induced collagen expression through p42/p44 MAPK requires integrin‐mediated FAK activation in mesangial cells (Hayashida et al, 2007). Furthermore, β1‐integrins induce TGF‐β‐dependent p38 MAPK activity during EMT in mammary epithelial cells, and TGF‐β‐stimulated MMP9 expression in keratinocytes is enhanced by the integrin α3β1 (Bhowmick et al, 2001; Lamar et al, 2008).
Integrins can also indirectly control the expression of components of the TGF‐β pathway. For example, the ectopic expression of the integrin α5‐subunit induces TGF‐βRII expression, which is potentiated further by α5β1 ligation to fibronectin, rendering cells responsive to TGF‐β (Wang et al, 1999). In fibroblasts deficient in the integrin β3‐subunit, TGF‐β signalling is enhanced owing to an increased expression of both TGF‐βRI and TGF‐βRII, suggesting that the expression of these receptors is repressed by αvβ3 (Reynolds et al, 2005). In addition, TGF‐β signalling is repressed in α3‐deficient keratinocytes due to an elevated expression of the inhibitory Smad7, which could mean that α3β1 can downregulate Smad7 to enhance TGF‐β signalling (Reynolds et al, 2008).
Integrins might also regulate TGF‐β signalling synergistically, through their physical interaction with TGF‐βRs. For example, TGF‐β stimulation induces the association of integrin αvβ3 with TGF‐βRII in both breast cancer cells and lung fibroblasts, initiating cooperative signalling to c‐Src and MAPKs (Scaffidi et al, 2004; Galliher & Schiemann, 2006). Similarly, TGF‐βRII associates with αvβ5 in sclerodermal fibroblasts, and integrin signalling through FAK is necessary for TGF‐β‐induced myofibroblastic differentiation (Asano et al, 2006b). Furthermore, α3β1 association with E‐cadherin and TGF‐βRs mediates the TGF‐β‐stimulated phosphorylation of β‐catenin and its association with phosphorylated Smad2, as well as the subsequent nuclear translocation of the Smad2–β‐catenin complex. Interestingly, both phenomena are independent of ligand binding by α3β1 (Kim et al, 2009a,b). Finally, in mammary epithelial cells overexpressing HER2, TGF‐β stimulates integrin clustering with HER2 and their association with the cytoskeleton, leading to PI(3)K signalling through c‐Src and FAK (Wang et al, 2009).
In conclusion, integrins can control TGF‐β signalling directly by TGF‐β activation, or indirectly by affecting Smad‐dependent and Smad‐independent signalling pathways at various levels (Table 2). Although the physiological relevance of some of the proposed mechanisms needs to be clarified, others are clearly important in the context of EMT, cancer, fibrosis and wound healing, as will be described below.
Integrin–TGF‐β crosstalk in fibrosis
Fibrosis results from an aberrant response to organ injury and is characterized by the proliferation of fibroblasts, their differentiation into myofibroblasts, and excessive ECM production and deposition; these processes are all mediated by TGF‐β. Fibrosis can ultimately lead to major organ failure and even death. Increasing evidence points to the integrin–TGF‐β crosstalk as crucial for the development and pathogenesis of fibrosis. TGF‐β induces the expression of the integrins α1β1 and α2β1, which mediate collagen remodelling and myofibroblast contraction (Fig 2A). Furthermore, the integrins α3β1, αvβ5 and—most notably—αvβ6 control TGF‐β activity or signalling in fibrosis.
The first clue that the integrin–TGF‐β interplay was important in fibrosis came from the observation that mice lacking the β6‐subunit are protected from bleomycin‐induced pulmonary fibrosis (Munger et al, 1999). The importance of αvβ6 for fibrogenesis has been demonstrated subsequently in several models; αvβ6 is not normally expressed in healthy epithelia but its expression is induced in many human fibrotic disorders in the kidney (such as diabetes mellitus, progressive fibrosing glomerulonephritis and Alport syndrome), the liver (acute biliary fibrosis) and the lung (sclerosis and idiopathic pulmonary fibrosis (IPF)). In mice, the constitutive expression of αvβ6 in the basal layer of the epidermis leads to elevated TGF‐β1 activation and the development of spontaneous chronic ulcers with severe fibrosis (Häkkinen et al, 2004). Conversely, β6 knockout mice are partly or completely protected from pulmonary fibrosis induced by radiation, tubulointerstitial fibrosis as a response to kidney obstruction, or acute biliary fibrosis caused by bile duct ligation. In wild‐type mice, fibrosis can be equally inhibited by treatment with antagonists of TGF‐β signalling or by using a blocking antibody against αvβ6 (Ma et al, 2003; Hahm et al, 2007; Wang et al, 2007; Patsenker et al, 2008). In fact, given that blocking the TGF‐β pathway has serious adverse effects—such as the development of autoimmunity—the specific inhibition of αvβ6‐induced TGF‐β activation at sites of injury is a promising therapeutic tool to combat TGF‐β‐mediated fibrosis. Indeed, low doses of antibodies against αvβ6 prevent radiation‐induced or bleomycin‐induced pulmonary fibrosis in mice, without causing inflammation (Puthawala et al, 2008; Horan et al, 2007).
Observations suggest that the integrins αvβ3, αvβ5 and αvβ8 provide additional therapeutic targets for this pathology. As mentioned above, αvβ3 and αvβ5 are thought to contribute to the pathogenesis of systemic sclerosis and scleroderma through TGF‐β activation (Asano et al, 2005b, 2006a). In human fibrotic lungs, epithelial cells expressing αvβ5 and PAR1 co‐localize with myofibroblasts, and TGF‐β‐mediated pulmonary fibrosis is reduced by the blockade of αvβ5 in a mouse model (Scotton et al, 2009). Furthermore, TGF‐β activation by αvβ8 can induce the differentiation of airway fibroblasts into myofibroblasts, and the expression of αvβ8 is increased in the airways of chronic obstructive pulmonary disease patients, correlating with the severity of the obstruction (Araya et al, 2006, 2007). However, the importance of αvβ8 in this process has not been corroborated by knockout or targeting studies. Finally, α3β1 also contributes to the development of IPF through a β‐catenin–Smad2‐dependent mechanism, as described above (Fig 2B). In IPF, a subset of differentiating fibroblasts is derived initially from alveolar epithelial cells by EMT (Kim et al, 2006). The lung‐specific deletion of the α3‐subunit in a mouse model of IPF reduces myofibroblast accumulation, collagen deposition, expression of EMT‐associated genes and progression to fibrosis, suggesting that blocking α3β1 could also be effective against fibrosis (Kim et al, 2009a,b).
Collectively, these results show that several integrins aggravate TGF‐β‐mediated fibrotic disorders, either by direct activation of TGF‐β, or by affecting downstream signalling. Thus, targeting these integrins could prove to be a valuable anti‐fibrotic therapy in humans. Alternatively, integrin‐associated proteins might represent targets for therapeutic intervention. For example, ILK is essential for TGF‐β‐induced kidney and liver fibrosis, although whether this depends on the modulation of integrin activity, or is an integrin‐independent effect of ILK, is unknown (Li et al, 2003).
Integrin–TGF‐β crosstalk in carcinoma progression
TGF‐β has a dual role in the development and progression of epithelial tumours: initially, it acts as a tumour suppressor for epithelial cells, but at a later stage can also promote growth, invasion and metastasis. The ability of TGF‐β to promote or suppress carcinoma progression is at least partly dependent on the tumour microenvironment (Bierie & Moses, 2006; Massagué, 2008). The interactions between TGF‐β and integrins can affect tumorigenesis and malignant progression in several ways. For example, an inappropriate suprabasal expression of α6β4 in stratified squamous epithelia inhibits TGF‐β signalling, thereby enhancing tumorigenesis by relieving the inhibitory effects of TGF‐β on epithelial proliferation (Owens et al, 2003). In addition, squamous cell carcinomas develop in stratified epithelia after the abrogation of TGF‐β signalling, which could be associated with enhanced integrin activity and would suggest that, under normal circumstances, TGF‐β has a suppressive effect on integrins (Guasch et al, 2007). However, it should be noted that most studies support a role for TGF‐β in inducing the de novo expression of several integrins that are not normally expressed in epithelial cells—such as α5β1, αvβ3, αvβ5 and αvβ6—thereby enhancing the migratory and invasive behaviour of carcinoma cells, particularly in conjunction with newly expressed MMPs and ECM components such as fibronectin (Fig 2C). Indeed, antagonizing the TGF‐β pathway blocks the induction of the expression of these integrins, as well as TGF‐β‐mediated invasion and metastasis, without affecting the growth of the primary tumour, suggesting that inhibiting integrin upregulation by TGF‐β is sufficient to block metastasis (Bandyopadhyay et al, 2006; Kawajiri et al, 2008).
As in fibrosis, αvβ6 seems to have a crucial role in the TGF‐β–integrin crosstalk in carcinomas. TGF‐β induces the expression of αvβ6 during EMT in vitro and in vivo, and αvβ6 is upregulated at the tumour–stromal interface of several aggressive squamous cell carcinomas—including cervical, colorectal, esophageal, head and neck, and skin carcinomas—and its upregulation is a prognostic factor for decreased survival (Bates et al, 2005; Hazelbag et al, 2007; van Aarsen et al, 2008; Marsh et al, 2008). αvβ6 can mediate migration and invasion, but can also establish a self‐amplifying loop by activating TGF‐β; the interruption of this feedforward mechanism could be an important step to arrest malignant progression. Although the blockade of αvβ6 had no effect on TGF‐β‐mediated proliferation of tumour cells in vitro, it did successfully inhibit the growth of xenograft tumours in vivo, suggesting that the tumour microenvironment has an important regulatory role (van Aarsen et al, 2008). Indeed, αvβ6‐mediated TGF‐β activation in an organotypic culture system for basal cell carcinoma induced differentiation of fibroblasts into myofibroblasts, which subsequently induced tumour cell invasion by the secretion of hepatocyte growth factor. Interestingly, the stroma of high‐risk basal cell carcinomas is rich in myofibroblasts that express hepatocyte growth factor, and its receptor—c‐Met—is expressed on the tumour cells, suggesting that a similar tumour–stromal interaction can occur in patients (Marsh et al, 2008). Therefore, although the blockade of several TGF‐β‐induced integrins might inhibit the migratory and invasive behaviour of tumour cells, antagonizing αvβ6 could also be important for interfering with self‐amplifying, TGF‐β‐mediated tumour–stromal interactions. This approach could ultimately become an effective treatment for various carcinomas.
Integrin–TGF‐β crosstalk during wound healing
The repair of cutaneous wounds is achieved through the concerted efforts of many cell types (Fig 3A; Singer & Clark, 1999). TGF‐β is involved in every phase of wound repair and is released by platelets, neutrophils, macrophages, fibroblasts and migrating keratinocytes. TGF‐β suppresses the inflammatory response and promotes the formation of granulation tissue by inducing fibroblast proliferation and differentiation, the expression of integrins and deposition of ECM proteins by fibroblasts, and endothelial cell migration and angiogenesis (Fig 3B; Werner & Grose, 2003). However, there are conflicting results as to the role of TGF‐β during re‐epithelialization. On the one hand, TGF‐β stimulates the expression of fibronectin and the integrins α5β1, αvβ5 and αvβ6 in keratinocytes, thereby inducing a migratory phenotype (Fig 3C). On the other hand, TGF‐β inhibits keratinocyte proliferation, and there is evidence to indicate that the net result of TGF‐β signalling on re‐epithelialization is inhibitory. For example, re‐epithelialization is delayed in mice that overexpress TGF‐β1 in the basal layer of the epidermis (Yang et al, 2001; Chan et al, 2002; Tredget et al, 2005), whereas it is accelerated and keratinocyte proliferation is increased in mice that express a dominant negative TGF‐βRII in basal keratinocytes or those that lack TGF‐βRII (Amendt et al, 2002; Guasch et al, 2007). In addition, re‐epithelialization is accelerated in Smad3 knockout mice (Ashcroft et al, 1999; Falanga et al, 2004).
Integrins mediate adhesion and migration during re‐epithelialization (Grose et al, 2002), and emerging evidence suggests that several can modulate TGF‐β signalling during wound healing, although the precise mechanisms are controversial and poorly understood. Re‐epithelialization is accelerated in β3‐null mice, which is accompanied by enhanced fibroblast infiltration, fibronectin deposition and neo‐angiogenesis, and elevated TGF‐β levels in the granulation tissue, suggesting that αvβ3 suppresses TGF‐β signalling (Reynolds et al, 2005). However, this is inconsistent with both the activation of TGF‐β by αvβ3 and the inhibitory effects of TGF‐β on re‐epithelialization. In addition, although the targeted deletion of the β6‐subunit does not affect wound healing, abnormal wound healing is observed in β6‐null mice when TGF‐β signalling is disturbed—for example, in the presence of glucocorticoids (Huang et al, 1996, 2000; AlDahlawi et al, 2006; Xie et al, 2009)—suggesting that rather than maintaining adhesion and mediating migration, αvβ6 functions as a safeguard in wounds, ensuring sufficient supply of TGF‐β when required. The activation of TGF‐β by αvβ8 has also been seen to delay the closure of scratch wounds in vitro, although whether it has a physiological role during re‐epithelialization in vivo is unknown (Fjellbirkeland et al, 2003; Neurohr et al, 2006). Finally, delayed wound re‐epithelialization has been observed in full‐thickness skin explants from α3‐null mice, supposedly owing to repressed TGF‐β signalling caused by an upregulation of Smad7 in the absence of integrin α3β1 (Fig 3C; Reynolds et al, 2008). However, these data are controversial in the light of the evidence that TGF‐β signalling inhibits re‐epithelialization, and because the targeted deletion of α3 from the basal layer of the epidermis has been recently shown not to inhibit re‐epithelialization (Margadant et al, 2009; Mitchell et al, 2009). Therefore, although regulation of TGF‐β signalling by integrins is potentially important in many aspects of the wound healing process, it is not fully understood. Future studies should shed more light on the exact mechanisms involved.
Extensive interactions undoubtedly exist between integrins and the TGF‐β pathway. Although our knowledge of the wide implications of this crosstalk and the underlying mechanisms has increased greatly in recent years, there are still several outstanding questions to address (Sidebar A). Clarification of these issues is important as it will not only increase our understanding of integrin signalling, TGF‐β signalling and integrin–TGF‐β crosstalk, but—importantly—could also lead to new treatment strategies for several human pathologies.
Sidebar A | In need of answers
(i) Do integrins αIIbβ3 and α5β1 interact with LAP TGF‐β and therefore have a role in TGF‐β activation?
(ii) Do αvβ1 and α8β1 activate TGF‐β?
(iii) What is the exact nature and function of αvβ3–TGF‐β crosstalk, and is it important in vivo—for example, in wound healing?
(iv) Can αvβ6 be a therapeutic target in cancer and fibrosis?
(v) Will antagonism of αvβ8 be effective against chronic obstructive pulmonary disease?
(vi) Are the effects of ILK on pulmonary fibrosis dependent on integrins?
We thank Rabab Charafeddine for stimulating discussions. We apologize to colleagues whose work might have been omitted owing to space constraints. This study was supported by grants from the Dystrophic Epidermolysis Bullosa Research Association (Crowthorne, UK) and the Dutch Cancer Society.
- TGF‐β‐inducible gene‐h3
- extracellular matrix
- epithelial‐to‐mesenchymal transition
- focal adhesion kinase
- human epidermal growth factor receptor 2
- integrin‐linked kinase
- latency‐associated protein
- latent TGF‐β binding protein
- mitogen‐activated protein kinase
- matrix metalloproteinase
- protease‐activated receptor 1
- Rho‐associated kinase
- transforming growth factor‐β
- transforming growth factor‐β type I receptor
- transforming growth factor‐β type II receptor
- Copyright © 2010 European Molecular Biology Organization