Induction of membrane circular dorsal ruffles requires co-signalling of integrin–ILK-complex and EGF receptor

Cells are exposed to a wide variety of mechanical and chemical stimuli that must be integrated at the molecular level to achieve an appropriate biological response. The integration of distinct signalling pathways from different cell surface receptors into a common downstream response is referred to as signalling crosstalk. Such crosstalk occurs between integrins and receptor tyrosine kinases (RTKs) to control important biological processes such as cell differentiation, proliferation, survival, migration, innate immune response and angiogenesis (Cabodi et al., 2004; Chan et al., 2006; King et al., 2011; Loubaki et al., 2010; McCall-Culbreath et al., 2008; Ross, 2004; Somanath et al., 2009). However, the molecular details of how distinct signalling pathways arising from integrins and RTKs such as epidermal growth factor receptor (EGFR) can converge to regulate these processes remain largely unknown.

Integrins are heterodimeric transmembrane proteins that interact with extracellular matrix molecules to trigger intracellular signal transduction cascades leading to the reorganization of the actin cytoskeleton and activation of downstream signalling pathways (Hynes, 2002; Legate et al., 2009; Wiesner et al., 2005). Integrins assemble in different α/β-subunit combinations that confer substrate and signalling specificity. Because integrins have short cytoplasmic domains that lack enzymatic and actin-binding activity, they depend on the assembly of adaptor proteins onto their cytoplasmic tails for signal transduction. More than 180 signalling and scaffolding molecules have been identified that can be recruited to large integrin-based signalling hubs called focal adhesions (FAs) (Legate and Fässler, 2009; ’Schiller et al., 2011; Kuo et al., 2011; Zaidel-Bar and Geiger, 2010). Among these molecules, integrin-linked kinase (ILK) is a key player that directly binds the β1 and β3 integrin cytoplasmic tails (Hannigan et al., 1996; Pasquet et al., 2002). ILK is a multifunctional protein that regulates various cellular processes by associating with regulatory and adaptor proteins such as Pinch, α- and β-parvins, IQGAP1 and paxillin (Bottcher et al., 2009; Lange et al., 2009; Wickstrom et al., 2011). The analysis of constitutive and conditional deletion of the Ilk gene in mice, Drosophila melanogaster and Caenorhabditis elegans revealed that ILK controls the organization of the F-actin cytoskeleton, cell polarity, differentiation and proliferation (Esfandiarei et al., 2010; Grashoff et al., 2003; Hannigan et al., 1996; Legate and Fässler, 2009; Lorenz et al., 2007; Mackinnon et al., 2002; Sakai et al., 2003; Wang et al., 2008; Zervas et al., 2001).

FAs serve as a signalling nexus to condense and direct numerous signalling molecules, including kinases. The proto-oncogene Src is one of the kinases that localizes to FAs. Src activity is regulated by both integrin and RTK signalling (Huveneers and Danen, 2009; Yeatman, 2004), and precise spatiotemporal activation is important for its biological functions, including the regulation of FA stability, turnover and integrity (Fincham and Frame, 1998; Zou et al., 2002). Src also regulates F-actin cytoskeleton remodelling through activation of various effector proteins, including small GTPases (Huveneers and Danen, 2009; Timpson et al., 2001), kinases such as Abl as well as p120-catenin and cortactin (Castano et al., 2007; Chang et al., 1995; Plattner et al., 1999).

Circular dorsal ruffles or waves [also known as dorsal ruffles (DRs) or actin ribbons] are dynamic actin-based structures that assemble on the dorsal plasma membrane in response to a variety of growth factors (Abercrombie et al., 1970; Buccione et al., 2004; King et al., 2011; Schliwa et al., 1984). Growth factor stimulation activates a signalling cascade that starts with activation of master kinases such as Src and ends with transient cytoskeletal rearrangements regulated by cortical actin polymerization (Buccione et al., 2004). The exact function of DRs is still unclear, but they have been proposed to be important for macropinocytosis, trafficking of β3 integrin, sequestration and internalization of RTKs after ligand stimulation, and fast remodelling of the actin cytoskeleton during cell migration and invasion (Abella et al., 2010; Buccione et al., 2004; Dowrick et al., 1993; Gu et al., 2011; Krueger et al., 2003; Orth et al., 2006; Suetsugu et al., 2003).

In this study, we show that DRs are the result of cooperative signals emanating from integrin and RTK signalling pathways. We found that ILK is an essential component in the DR signalling cascade downstream of fibronectin (FN)–α5β1 integrins. ILK regulates the spatiotemporal activation of Src at FAs, which is required for tyrosine phosphorylation of the tumour-suppressor Cyld and the formation of DRs. The implications of these findings are discussed.

We generated ILK-floxed (ILK^f/f) and ILK-deficient (ILK^−/−) fibroblasts to investigate the consequence of ILK deletion in vitro (Sakai et al., 2003). During our experiments we realized that stimulation of starved ILK^f/f cells with media containing 10% fetal calf serum induced DRs in approximately 30% of ILK^f/f cells, whereas ILK^−/− cells very rarely formed DRs (Fig. 1A). To study this effect under defined conditions in the presence of specific growth factors, we measured epidermal growth factor (EGF)-triggered DR formation in serum-starved ILK^f/f and ILK^−/− fibroblasts that were seeded on FN-coated surfaces. Consistent with our observation using 10% fetal calf serum, about 25% of ILK^f/f cells formed DRs after EGF stimulation, whereas ILK^−/− cells showed very few ruffles (Fig. 1B,C). Similarly, ILK^−/− cells formed fewer DRs in response to platelet-derived growth factor (PDGF) stimulation (supplementary material Fig. S1A). This reduction in DR formation was not a clonal artifact because we consistently found a significant reduction of DRs in all ILK^−/− clones compared with their ILK^f/f counterparts (supplementary material Fig. S1B). Immunostaining of ILK showed no localization to DRs after EGF stimulation (supplementary material Fig. S1C). The reduced DR frequency in ILK^−/− cells was not due to reduced EGFR phosphorylation or ERK1/2 activation because their relative levels were similar in ILK^f/f and ILK^−/− cells with the exception of phosphorylation of EGFR Tyr992, which was increased in ILK^−/− cells (Fig. 1D, supplementary material Fig. S1D). Similarly, EGF-induced Rac1 activation was similar in ILK^−/− and ILK^f/f cells, although the activation was prolonged in ILK^−/− cells but the differences were not statistically significant (supplementary material Fig. S1E,F). ILK^−/− cells show spreading defect, raising the possibility that impaired DR formation is a consequence of the reduced spread area of these cells. However, when ILK^−/− cells were allowed to spread for longer time periods of up to 2 days they did not show a significantly increased frequency of DR formation despite a normal spread area (Fig. 1A; data not shown). Moreover, stable re-expression of FLAG-tagged ILK (Fig. 1E,F) or ILK–EGFP (supplementary material Fig. S1G) fully rescued the DR defect of ILK^−/− cells.

It has been suggested that ILK interconnects integrins with growth factor pathways through Pinch1. Additionally, ILK and Pinch1 are components of the ILK–Pinch–Parvin (IPP) complex, whose members depend on complex formation for maintaining their stability (Legate et al., 2006). Western blot analysis showed that the level of Pinch1 expression is strongly reduced in ILK^−/− cells (Fig. 1E). To test whether the DR formation defect in ILK^−/− cells is caused by the diminished Pinch1 protein level, we stably re-expressed FLAG-tagged N-terminal ANK-repeats of ILK (ANK–FLAG) in ILK^−/− cells. The presence of ANK–FLAG stabilized Pinch1 expression to wild-type levels, but cells were still not able to form DRs (Fig. 1E,F). Hence, the reduced DR frequency in ILK^−/− cells was not due to reduced Pinch1 protein levels. Expression of ANK–FLAG had no effect on DR formation in ILK^f/f cells (supplementary material Fig. S1H, I). Conversely, Pinch1^−/− cells showed strongly reduced ILK protein levels (Stanchi et al., 2009) and decreased DR formation (supplementary material Fig. S1J). Together, these data demonstrate that ILK plays an essential role in the induction of DRs.

The precise biological function of DRs is still uncertain, but various reports ascribe macropinocytosis, large scale actin reorganizations prior to migration and growth factor receptor internalization as downstream consequences of DR formation (King et al., 2011; Orth et al., 2006). We found that ILK^−/− cells migrated towards a source of EGF or PDGF less efficiently than ILK^f/f cells or ILK–FLAG-rescued ILK^−/− cells (Fig. 2A). Furthermore, the internalization of activated EGFR was significantly reduced in ILK^−/− cells, whereas the internalization of transferrin receptor remained unchanged (Fig. 2B, supplementary material Fig. S2). Reduced internalization of EGFR is expected to result in prolonged signalling, and indeed the relative level of phosphorylation of EGFR in ILK^−/− cells was increased, perhaps as a result of impaired downregulation through internalization (supplementary material, Fig S1D). Therefore, ILK^−/− cells displayed phenotypic differences that are consistent with a reduction in the formation and number of DRs.

The finding that ILK plays a crucial role in DR formation suggested a requirement for integrin signalling in the formation of these structures. To test whether integrin engagement is necessary for DR formation, we monitored EGF-induced DRs in ILK^f/f cells seeded on FN or poly-L-lysine (PLL). Whereas FN can be recognized by many integrin receptors, most notably α5β1 and αvβ3, PLL-mediated adhesion is integrin-independent. Only FN-seeded ILK^f/f cells formed DRs (Fig. 3A), and the rate of DR formation in ILK^f/f cells increased with the FN concentration (Fig. 3B).

To examine whether the formation of DRs depends on a specific integrin heterodimer, we evaluated DR assembly in a FN-free system by seeding serum-starved FN-null (FN^−/−) fibroblasts on FN, vitronectin (VN), collagen1 (Col1), or PLL (Fig. 3C–E). FN^−/− cells established a distinct morphology on each substratum and formed paxillin-rich focal adhesions on FN, VN and Col1 but not on PLL (Fig. 3E). Whereas about 25% of FN-seeded FN^−/− cells formed DRs, cells adherent to PLL, VN or Col1 formed significantly fewer DRs (<5%, Fig. 3F). The spread areas of FN^−/− cells on FN, VN and Col1 were comparable, indicating that differences in spreading do not contribute to altered DR formation (supplementary material Fig. S3A). In addition, when we limited the spreading time of FN-seeded FN^−/− cells to 30 minutes, so that they covered the same spread area as PLL-attached cells, they still formed DRs normally (supplementary material Fig. S3B). Moreover, DR formation in FN^−/− cells plated on FN was also dependent on ILK (supplementary material Fig. S3C,D).

Integrin-mediated cell adhesion to FN is mainly achieved through α5β1 and αVβ3 integrins, whereas VN is bound by αVβ3 but not α5β1 integrin (Hynes, 2002). Therefore, our results suggest that only α5β1 integrin signals trigger formation of DRs. To confirm this, we examined EGF-induced DRs in serum-starved FN-seeded integrin β1^f/f and β1^−/− fibroblasts (which lack α5β1 integrin but express αVβ3 integrin). In agreement with the previous experiments, about 30% of β1^f/f cells formed DRs, whereas β1^−/− cells showed significantly reduced DR frequency (Fig. 3G). Re-expression of β1 integrin in β1^−/− cells rescued DR formation in these cells (supplementary material Fig. S3E,F). These results suggest that DRs are the consequence of FN–α5β1-integrin–ILK and EGFR co-signalling.

Both integrin and RTK signalling stimulate Src tyrosine kinase activity, which is known to play a central role in DR formation (Chang et al., 1995; Huveneers and Danen, 2009). In line with these previous reports, ILK^f/f cells pretreated with a Src inhibitor (PP1) failed to form DRs (supplementary material Fig. S4A). Therefore we decided to investigate the role of Src in more detail. First, we investigated whether Src activation is impaired in ILK^−/− cells. Immunostaining with antibody against Tyr416-phosphorylated Src (pY416-Src) showed that active Src levels were dramatically reduced in FAs of FN-seeded ILK^−/− cells before EGF stimulation and remained reduced after EGF stimulation (Fig. 4A,B). Re-expression of ILK–EGFP in ILK^−/− cells rescued the level of active Src in FAs (supplementary material Fig. S4B–D). Western blot analysis showed that total Src levels were similar in ILK^f/f and ILK^−/− cells (Fig. 4C) and that non-adherent ILK^f/f and ILK^−/− cells showed a similar (twofold) increase in Src activity after EGF treatment (Fig. 4C,D). Plating cells on FN caused a basal increase in Src phosphorylation in ILK^f/f cells that did not manifest in ILK^−/− cells, but EGF treatment induced a similar activation of Src in both cell lines, resulting in a net decrease in active Src in ILK^−/− cells of about 20% (Fig. 4C,D). Co-immunoprecipitation of ILK with antibody against Src in ILK–FLAG-rescued ILK^−/− cells (Fig. 4E), and of Src with antibody against GFP in ILK–GFP-rescued ILK^−/− cells (Fig. 4F) indicated that Src and ILK form a complex in our fibroblast cell lines. However, a complex between Src and endogenous ILK was not easily detectable in our cells (data not shown).

Importantly, transient expression of constitutively active EGFP-tagged SrcY527A mutant was localized to FAs and rescued DR formation in ILK^−/− cells (supplementary material Fig. S4E–G). Furthermore, a decreased level of active Src at FAs significantly correlated with decreased DR frequency, whereas Src activity and the number of DRs concomitantly increased in ILK^f/f cells when seeded on increasing FN concentrations (supplementary material Fig. S4H; compare with Fig. 3B). Together, these experiments suggest that ILK affects DR formation through control of Src activity at FAs.

The kinase signalling cascade leading to DRs is mediated by β1 integrin–ILK and EGFR co-signalling, which activates Src in FAs. To identify potential ILK-dependent substrates for EGFR and Src, which are involved in DR formation, we compared the phosphoproteome of ILK^f/f and ILK^−/− cells treated with EGF for 30 seconds or 2 minutes by combining phosphorylated tyrosine immunoprecipitation and SILAC-based mass spectrometry (Fig. 5A). Candidate proteins involved in DR formation induced by β1 integrin and EGFR were defined as those that displayed increased phosphorylation upon EGF stimulation in ILK^f/f cells, but not in ILK^−/− cells. Our analyses identified and quantified more than 2000 proteins and 140 specific phosphorylation sites (supplementary material Tables S1 and S2) after excluding proteins that are expressed at different levels in ILK^f/f and ILK^−/− cells, identified in whole proteome SILAC-based mass spectrometry experiments (data not shown). The majority of proteins identified 2 minutes after EGF stimulation had the same SILAC ratio in ILK^f/f and ILK^−/− cells, although certain proteins were upregulated in an ILK-dependent manner (Fig. 5B).

To test the involvement of these proteins in DR formation, we performed short hairpin RNA (shRNA)- or short interfering (siRNA)-mediated knockdown in ILK^f/f cells of candidate proteins that were consistently represented in four independent SILAC screens (Fig. 5C). Knockdown efficiency was evaluated using western blotting and quantitative RT-PCR (supplementary material Fig. S5A,B). DR frequency decreased significantly in Cyld- and Asap2-depleted cells (Fig. 5D,E). We decided to further analyse the involvement of Cyld in the β1 integrin–ILK and EGFR co-signalling pathway.

The tumour suppressor protein Cyld tunes several signal transduction pathways including NFkB, JNK and Wnt–β-catenin through its deubiquitylating (DUB) activity (Massoumi, 2010). We prepared Cyld^−/− fibroblasts from Cyld-deficient mice (Massoumi et al., 2006) to corroborate the crucial role of Cyld in DR formation (Fig. 6A). Re-expression of FLAG–Cyld normalized DR formation in Cyld^−/− fibroblasts (Fig. 6A; supplementary material Fig. S6A). Similarly, re-expression of a catalytically inactive Cyld mutant (Cyld C>S) (Brummelkamp et al., 2003; Trompouki et al., 2003) restored DR formation in Cyld^−/− cells (Fig. 6A). Immunostaining of endogenous Cyld in wild-type cells (Fig. 6B) and overexpression of GFP–Cyld in Cyld^−/− cells revealed that Cyld was recruited to DRs (supplementary material Fig. S6B–D). Rac1 activation was not impaired in Cyld^−/− cells (supplementary material Fig. S6E,F), which was similar to the results using ILK^−/− cells (supplementary material Fig. S1).

Because our phosphoproteomics screen enriched for tyrosine-phosphorylated proteins, we tested whether Cyld becomes tyrosine-phosphorylated when serum-starved cells are seeded on FN and stimulated with EGF. Fig. 6C shows that Cyld became tyrosine-phosphorylated within 2–4 minutes after EGF stimulation. This phosphorylation was abrogated when cells were treated with the selective EGFR inhibitor Gefitinib (Iressa), indicating that Cyld phosphorylation is indeed downstream of EGFR signalling (Fig. 6D). Cyld became tyrosine-phosphorylated in response to EGF stimulation when cells were plated on FN but not when they were seeded on PLL, VN or Col1 (Fig. 6E,F). Consistent with our phosphoproteomics data, Cyld phosphorylation was reduced in FLAG–Cyld-rescued Cyld^−/− cells in which ILK was depleted (Fig. 7A). Similarly, Cyld phosphorylation was diminished in ILK^−/− cells expressing FLAG–Cyld as compared with control ILK^f/f cells (Fig. 7B), whereas localization of endogenous Cyld was not changed in ILK^−/− cells upon EGF stimulation (supplementary material Fig. S7A). Finally, Src inhibition with PP1 abolished EGF-induced Cyld tyrosine phosphorylation in FLAG–Cyld-rescued Cyld^−/− cells, indicating that EGF-induced Cyld tyrosine phosphorylation occurs downstream of Src (Fig. 7C). Immunostaining of active Src in Cyld^−/− cells expressing mCherry–Cyld suggested that cytoplasmic Cyld was not localized to FAs (Fig. 7D). Although phosphorylation of Cyld on serine has recently been reported (Hutti et al., 2009), this is the first report demonstrating Cyld tyrosine phosphorylation.

To identify which tyrosine residue(s) are phosphorylated in response to EGF, we conducted a mutational analysis. Mutation of four tyrosine residues to alanine (FLAG–Cyld-4×) identified in other phosphoproteomics experiments (data not shown; locations of these tyrosines are given in supplementary material Table S3) reduced neither EGF-induced tyrosine phosphorylation of Cyld nor DR formation (Fig. 6A; supplementary material Fig. S7B). The substitutions of additional tyrosine residues with alanine (FLAG–Cyld-9× and FLAG–Cyld-18× mutants, supplementary material Table S3) led to a significant decrease in EGF-induced tyrosine phosphorylation and DR formation (Fig. 6A, Fig. 7E).

Together, these data demonstrate that Cyld lies downstream of an integrin–ILK and EGFR co-signalling pathway leading to the formation of DRs in fibroblasts. Although the DUB activity of Cyld is dispensable for DR formation, tyrosine phosphorylation of Cyld is required, and this phosphorylation lies downstream of EGF-mediated and ILK-dependent Src activation.

Previous studies have shown that dynamic, transient actin-based DRs form in response to a variety of growth factors, including EGF, PDGF and hepatocyte growth factor (HGF) (Buccione et al., 2004). In the present study, we report that integrin and RTK signalling pathways cooperatively control the formation of DRs. The integrin-based signalling leading to DR formation emanates specifically from α5β1 integrin through a signalling module containing ILK, Src and Cyld.

The specificity of the involvement of α5β1 integrin in DR formation can be explained by the differential assembly of specific FA signalling complexes at the integrin tails that confers distinct signalling specificities to different α/β integrin subunit combinations (Humphries et al., 2009). For example, α5β1 and αVβ3 integrins have distinct effects on actin cytoskeletal regulation through different modulation of Rho GTPases (Danen et al., 2005; Huveneers et al., 2008). Whereas adhesion to FN by α5β1 integrin causes high levels of RhoA activity and low levels of Rac activity, adhesion via αvβ3 integrin induces low levels of RhoA activity (Danen et al., 2002; Huveneers et al., 2008). Additional actin modulators, such as VASP, are also regulated differently by β1 and β3 integrins, leading to changes in actin-dependent processes such as migratory behaviour (Worth et al., 2010). By plating cells on different substrates and making use of β1^−/− cells, we show that the specific signalling complex that assembles on α5β1 integrin tails supports DR formation, whereas the signals emanating from β3 integrins or collagen-binding integrins do not. This is consistent with a recent study demonstrating that β1 integrin is essential for both PDGF-induced DRs and chemotaxis in fibroblasts (King et al., 2011). We also show that ILK is a key component of the DR signalling complex downstream of α5β1 integrin. ILK^−/− cells formed only a few DRs, whereas ILK^f/f cells as well as ILK^−/− cells rescued with ILK–EGFP or ILK–FLAG formed a full complement of DRs. This effect of ILK is downstream of β1 integrin because β1^−/− cells are also unable to support DRs despite ample levels of ILK expression.

ILK has been intensively studied as a FA adaptor molecule that is involved in integrin-mediated actin cytoskeletal rearrangements (Bottcher et al., 2009; Legate et al., 2006; Legate et al., 2009). We found that ILK affects DR formation by regulating the levels of active Src in FAs in the absence of growth factors. Src is a tyrosine kinase that mediates signalling pathways involved in actin reorganization and DR formation, and is activated downstream of RTK and integrin signalling (Huveneers and Danen, 2009). There is increasing evidence that Src activation and its biological functions are tightly controlled by its subcellular localization. Whereas Src at FAs inhibits Rho GTPase by inducing p190GAP activation, it activates Rho GTPase when localized to podosomes (Arthur et al., 2000; Bass et al., 2008; Berdeaux et al., 2004). In addition, the PDGF receptor uses different pools of Src to initiate distinct pathways. Whereas PDGF activates caveolae-associated Src for mitogenesis, PDGF-activated Src outside of caveolae affects F-actin assembly leading to DR formation (Veracini et al., 2006). The absence of ILK strongly reduces the levels of active Src in FAs, but rescuing ILK^−/− cells with ILK–FLAG or ILK–GFP restores its presence in FAs, and induces an ILK–Src complex. It has been suggested that an ILK–Src complex regulates actin polymerization by phosphorylating cofilin (Kim et al., 2008). The ILK–Src interaction seems to be important for Src activation in FAs, but does not appear to play a role in Src phosphorylation in response to EGF stimulation. Therefore, the localization of Src to FAs via associating with ILK is necessary for DR formation, whereas activation of Src in other subcellular compartments is not sufficient for this process. However, the precise mechanism by which ILK affects active Src levels in FAs is still unclear.

The mechanism by which Src, and in particular FA-associated Src, induces DR formation is largely unknown. Src at FAs activates Rac1 locally through phosphorylation of specific GEFs, and thereby induces lamellipodia (Huveneers and Danen, 2009). In concert with RTK signals, active Rac could participate in DR induction (Buccione et al., 2004). In addition, several Src substrates have been implicated in DR formation, such as cortactin (Lai et al., 2009), Abl (Plattner et al., 1999) and Cbl (Sirvent et al, 2008). In this study we used a SILAC-based phosphoproteomics screen to detect proteins differentially phosphorylated by EGF in the presence or absence of functional integrin–ILK signalling. We confirmed the reliability of our SILAC list by monitoring DR formation in cell lines depleted of selected candidate proteins by RNA interference. In this way, we identified Cyld as a new player in DR formation downstream of Src. This screen also identified additional candidates that could potentially serve key roles in integrin–RTK crosstalk.

The function of Cyld in the regulation of signalling pathways has previously been linked to its Lys63 DUB activity (Brummelkamp et al., 2003; Kovalenko et al., 2003; Massoumi et al., 2006; Reiley et al., 2004), which can be controlled by phosphorylation on Ser418 by IKKε (Hutti et al., 2009). However, the role of Cyld in DR formation is independent of its DUB activity. Rather, EGF stimulation resulted in Cyld tyrosine phosphorylation, which is necessary for DR formation. Cyld phosphorylation is dependent on Src activity and occurs downstream of a cooperative EGFR and integrin signalling network involving FN, α5β1 integrin and ILK. Because ILK does not localize to DRs and Cyld does not localize to FAs we propose a model whereby, upon EGF stimulation, activated Src localizes to ILK-containing FAs, where it activates substrates that either directly or indirectly phosphorylate Cyld, causing it to redistribute to DRs to exert its specific function.

Although we have identified Cyld as an important intermediary for DR formation, the precise function of Cyld tyrosine phosphorylation in this process has yet to be elucidated. Tyrosine phosphorylation might be required for the interaction of Cyld with as-yet-unknown binding partners, including proteins that can directly regulate actin dynamics. On the other hand, Cyld associates with α-tubulin and microtubules via its CAP-Gly domains and increases the levels of acetylated tubulin through an inhibitory interaction with histone deacetylase-6 (Gao et al., 2008; Wickstrom et al., 2010). Cyld tyrosine phosphorylation might control DR assembly by affecting the ability of Cyld to bind to microtubules and influence their dynamic instability, thereby controlling actin–microtubule crosstalk. We are currently addressing these possibilities to understand the role of Cyld in DR formation more precisely.

In conclusion, our work has identified Cyld as a key member of an integrin–ILK and EGFR co-signalling pathway. Interestingly, deregulation of each of these molecules has been implicated in cancer progression (Cabodi et al., 2010; Demchenko et al., 2010; Grandal and Madshus, 2008; Massoumi, 2010). Although the biological function of DRs is unknown, proposed functions such as RTK endocytosis and sites of localized matrix degradation could be important for tumourigenesis and metastatic behaviour. Future work to more precisely define how integrin–ILK and EGFR collaborate to activate Cyld, and how Cyld functions to enable rapid actin reorganizations leading to DRs, could provide novel insights into how deregulation of these signalling pathways promotes the formation and spread of cancer.

Human recombinant EGF and PDGF-BB were from Millipore; Boyden chambers were from BD Bioscience; PP1 inhibitor was from Cell Signaling Technology. Gefitinib (Iressa) was supplied by Selleck. The following antibodies were used: ILK, Rac1 and paxillin (BD Bioscience); EGFR, Tyr1173-phosphorylated EGFR, Tyr1068-phosphorylated EGFR, Tyr992-phosphorylated EGFR, Tyr845-phosphorylated EGFR and Tyr416-phosphorylated Src (Cell Signaling Technology); tyrosine-phosphorylated 4G10 and Lasp-1 (Millipore); β1 integrin and Cyld antibodies were homemade antibodies raised in rabbit; anti-FLAG, SHC2, and vinculin (Sigma); CDC42BPB, Dock4 (ABNOVA); anti-mouse horseradish peroxidase (HRP), anti-rabbit HRP and anti-rat HRP (BioRad); phalloidin–Alexa-Fluor-488 (Invitrogen); anti-rabbit Cy3 and anti-mouse Cy3 (Jackson ImmunoResearch); and Pinch1 (Li et al., 2007).

Cells were cultured on glass cover slips coated with 10 μg/ml FN (Calbiochem). For staining, cells were fixed in 2–4% paraformaldehyde in PBS for 15 minutes and permeabilized for 10 minutes with 0.2% Triton X-100 in PBS. The cells were blocked with 3% BSA in PBS for 1 hour and incubated with the primary antibody for 1 hour at room temperature or overnight at 4°C. Secondary antibodies were incubated for 1 hour at room temperature. Images were collected using a confocal microscope (DMIRE2; Leica, Bensheim, Germany) equipped with 63× NA 1.4 or 100× NA1.4 oil objectives and the Leica Confocal Software (version 2.5, build 1227), or collected with a AxioImager Z1 microscope (Zeiss, Germany) with the 63× NA 1.4 oil objective, using Metamorph software.

Cell lysates were prepared by quickly washing cells in ice-cold PBS prior to addition of lysis buffer [50 mM Tris-HCl pH 7.6, 150 mM NaCl, 1% Triton X-100, protease inhibitor cocktail (Roche) and phosphatase inhibitor cocktail (Sigma-Aldrich)]. FLAG-tagged protein immunoprecipitation was performed according to the manufacturer's instructions (cat#A2220, Sigma). For other immunoprecipitations, cell lysates at 0.5–1 mg/ml were pre-cleared by centrifugation for 1 hour at 4°C and incubated with antibody for 3 hours or overnight at 4°C. Protein complexes were captured using protein A or G agarose beads for 1 hour at 4°C, washed three times with lysis buffer, eluted with SDS loading buffer and analyzed by SDS-PAGE. GTP–Rac1 was pulled down by PAK-CRIB peptide from EGF-triggered cell lysates and blotted with Rac1 antibody.

cDNA constructs of FLAG–Cyld, FLAG–ILK, FLAG–ANK ILK and phosphorylated EGFP–ILK were amplified by PCR. The DUB-dead FLAG–Cyld C>S construct was kindly donated by Rene Bernards (NKI Amsterdam, The Netherlands). FLAG–Cyld mutant constructs with different numbers of tyrosine residues substituted by alanine were generated using the QuikChange Site-Directed Mutagenesis kit (#200518 Stratagene). Positions of mutations in FLAG–Cyld-9× and FLAG–Cyld-18× constructs are summarized in supplementary material Table S3. Mission siRNAs were ordered from Sigma. shRNAs were supplied from a shRNAmir30 library (Thermo Scientific). For shRNA and siRNA sequences, see supplementary material Table S3.

Cells were transiently transfected with Lipofectamine 2000 (Invitrogen). To generate stable cell lines cDNAs were cloned into pCLMFG retroviral vectors and transiently transfected into human embryonic kidney (HEK293T) cells; viral particles were used for infection of ILK^−/− fibroblasts as previously described (Pfeifer et al., 2000). For stable knockdowns Phoenix viral packaging cells were used to generate virus that was subsequently used to infect ILK^f/f fibroblasts (Pear et al., 1993). Integrin β1^f/f fibroblasts were isolated from murine kidney and immortalized by SV40 Large T-antigen. β1^−/− cells were generated by adenoviral infection of β1^f/f cells with GFP–Cre recombinase, followed by flow cytometry cell sorting of GFP-positive cells. β1^−/− cells were rescued by retroviral infection with β1 integrin cDNA. For DR experiments, fibroblasts were serum-starved overnight, trypsinized, seeded on dishes coated with FN, VN, Col1 or PLL, stimulated with 50 ng/ml EGF and monitored for DR formation using a Zeiss Axiovert 200M microscope (Zeiss, Germany) at 37°C. Time-lapse images were captured for 15 minutes with an interval time of 90 seconds. We isolated Cyld^−/− fibroblasts from kidneys of Cyld^−/− mice (Massoumi et al., 2006).

ILK^f/f and ILK^−/− fibroblasts were serum-starved for 4 hours at 37°C. Medium was exchanged for ice-cold medium containing 25 ng/ml Alexa-Fluor-488-conjugated EGF (Invitrogen) and cells were incubated on ice for 30 minutes to allow for full ligand binding. Plates were washed twice with cold PBS, and pre-warmed DMEM was added to induce internalization of EGFR. Plates were incubated at 37°C for the indicated times, and placed on ice to stop internalization. Cells were washed for 5 minutes with 0.2 M acetic acid containing 500 mM NaCl, pH 2.8, washed three times with cold PBS and scraped into tubes for counting by fluorescence-activated cell sorting (FACS). Negative and 100% binding controls were kept on ice and either acid-washed or directly scraped into tubes, respectively. Fluorescence intensity was measured by FACS, and the mean values of the peaks were normalized against 100% binding controls to obtain percentage internalization.

Chemotactic migration assays were performed as previously described (Legate et al., 2011). Briefly, 30,000 cells were re-suspended in serum-free conditions in Cell Culture Inserts (cat#353097; BD Biosciences) that were pre-coated on the underside with 5 μg/ml FN. Cells were allowed to migrate for 4 hours at 37°C towards the lower chamber containing 50 ng/ml PDGF-BB, 50 ng/ml EGF or BSA control. Cells were fixed and stained with 0.1% Crystal Violet in 20% methanol for 5 minutes at room temperature, and non-migratory cells were manually removed with a cotton swab. Four random fields were captured by a Zeiss Axioskop microscope equipped with a LEICA DC 500 digital camera, and cells were counted manually.

SILAC-based proteomics was performed as described (Mann, 2006). Briefly, cells were metabolically labelled with SILAC medium, serum-starved overnight, trypsinized, seeded on FN for 90 minutes and stimulated with EGF (50 ng/ml) for 30 seconds or 2 minutes. Cells lysates were prepared, mixed 1:1 and subjected to immunoprecipitation using antibodies against phosphorylated tyrosine (4G10 from Millipore and anti-pY100 from Cell Signaling). The immunoprecipitated proteins were subjected to in-gel digestion; peptides were concentrated and desalted using the micropurification system StageTips, separated by online reverse phase nanoscale capillary liquid chromatography and analyzed by electrospray ionization coupled with tandem mass spectrometry (ES MS/MS) on a linear trap quadrupole (LTQ)-Orbitrap mass spectrometer (Thermo Fisher Scientific). Mass spectra were processed with the software MaxQuant in combination with the Mascot search engine.

Images of live cells were acquired at 37°C and 5% CO₂ on a Zeiss Axiovert 200M microscope with a 10× 1.6 objective; the microscope was equipped with a motorized stage (Märzhäuser, Wetzlar, Germany), an environment chamber (EMBL Precision Engineering, Heidelberg, Germany) and a cooled CCD camera (Roper Scientific, Princeton, NJ). Microscope control, image acquisition and post-acquisition analysis were carried out using MetaMorph software (Molecular Devices, Downington, PA). To monitor dynamics of DR formation in different cell types, serum-starved cells were seeded on FN for 90 minutes, stimulated with EGF (50 ng/ml) or PDGF-BB (50 ng/ml) and time-lapse movies captured for 17 minutes in the case of EGF stimulation (40 and 60 minutes for PDGF and serum stimulation, respectively) with 90-second time intervals. DRs were classified as transient actin-rich structures that appeared after growth factor stimulation. To measure active Src levels at FAs, serum-starved cells were seeded on FN for 90 minutes, stimulated with EGF, fixed and immunostained with antibodies against Tyr416-phosphorylated Src and vinculin. Images were captured with a DMIRE2 confocal microscope (Leica, Bensheim, Germany) equipped with 63× NA 1.4 oil objectives. Images were then processed by MetaMorph software to calculate the intensity of active Src and vinculin at FAs. The averages of active Src intensities normalized to vinculin intensities are presented in histograms (means ± s.d.). The intensity of western blot bands was quantified by MultiGauge software (Fujifilm).

Results are expressed as means ± s.d. or means ± s.e.m. Statistical analysis was performed using GraphPad Prism (version 5.00, GraphPad Software) or Excel software. ANOVA or Student's t-test were used for comparisons between different data sets.

We thank Rene Bernards for providing the DUB-dead Cyld C>S construct.