Rac1 is deactivated at integrin activation sites through an IQGAP1–filamin-A–RacGAP1 pathway

Cell migration is a fundamental physiological process required for embryogenesis, immune surveillance and wound healing. Migration is also central to the regulation of disease pathogenesis by contributing to processes such as tumour invasion and leukocyte transmigration. Integrin receptors provide a physical link between the extracellular matrix (ECM) and the actin cytoskeleton, and serve as platforms for the assembly of signalling complexes that control cell phenotype, including migration (Hynes, 2002). Integrin function is regulated by a conformational switch that modulates extracellular ligand-binding affinity. Inactive integrins exist in a bent conformation and have a low affinity for their ligands, whereas primed or activated integrins are extended and have a high affinity for their ligands (Askari et al., 2009). Integrin-mediated adhesion modulates the activity of the small GTPase Rac1 (Del Pozo and Schwartz, 2007; Morgan et al., 2007; Price et al., 1998), and therefore dynamically regulates cell migration by promoting cytoskeletal re-organisation and membrane protrusion (Ridley, 2003; Ridley, 2006). Disruption of Rac1 signalling results in major migration defects, such as loss of cell directionality on fibrillar matrices (Bass et al., 2007; Pankov et al., 2005) and suppressed wound healing in vivo (Liu et al., 2009; Tscharntke et al., 2007). Productive cell migration requires coordinated activation and deactivation of Rac1, and accordingly a range of guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs) have been reported to be involved in integrin-dependent Rac1 regulation (Katoh and Negishi, 2003; Nishiya et al., 2005). However, the mechanism whereby integrin activation coordinates Rac1 activity is still only partially resolved.

In this study, which builds on published proteomic analyses of fibronectin (FN)-induced, integrin-associated complexes (Humphries et al., 2009; Kuo et al., 2011; Schiller et al., 2011), network analyses were used to identify filamin-A (FLNa) and IQ-motif-containing GTPase activating protein 1 (IQGAP1) as putative links between β1 integrin and Rac1. The hypothesis that FLNa and IQGAP1 modulate integrin-dependent Rac1 activation was tested and the mechanism elucidated. Specifically, FLNa and IQGAP1 are recruited to active integrins to constrain Rac1 activity via the recruitment of the GTPase-activating protein RacGAP1 (also known as MgcRacGAP and CYK4) in order to restrict protrusive activity during cell migration. These findings reveal a novel function for a FLNa–IQGAP1 complex in the regulation of Rac1 activity upon integrin activation.

To identify new mechanisms by which β1 integrin regulates Rac1 activity, data from three proteomic analyses of FN-induced, integrin-associated complexes (Humphries et al., 2009; Kuo et al., 2011; Schiller et al., 2011) were integrated with protein–protein interaction (PPI) databases, to generate a hypothetical FN-induced, integrin-associated PPI network. Analysis of the components connecting β1 integrin to Rac1 revealed FLNa and IQGAP1 as putative links between β1 integrin and Rac1 (Fig. 1A). Both FLNa and IQGAP1 were confidently identified in all three studies. Therefore, we tested the hypothesis that FLNa and IQGAP1 contribute to integrin-modulated Rac1 activity.

To assess the contribution of FLNa and IQGAP1 to Rac1 activation, mouse embryonic fibroblasts (MEFs) (Fig. 1B) and human U2OS osteosarcoma cells (Fig. 1C) were plated on FN, subjected to siRNA-mediated knockdown using different targeting oligos, and levels of GTP-Rac1 were measured by effector pull-down. Non-targeting-siRNA-treated MEFs exhibited a transient wave of Rac1 activity during spreading on FN, with a peak of activity observed at 45 minutes (Fig. 1D,E) (Bass et al., 2007; Humphries et al., 2009). By contrast, suppression of either FLNa or IQGAP1 expression resulted in enhanced and sustained Rac1 activation (Fig. 1D,E). Similarly, silencing of either FLNa or IQGAP1 expression in human osteosarcoma cells resulted in enhanced Rac1 activity during spreading on FN (Fig. 1F). These data indicate that both FLNa and IQGAP1 play an important role in suppressing Rac1 activity downstream of FN engagement.

Integrins undergo conformational regulation that determines their activation state and ligand-binding competency. Proteomic analyses suggested that FLNa and IQGAP1 were specifically enriched to adhesion complexes upon FN engagement (Fig. 1A). As both FLNa and IQGAP1 have been reported to co-immunoprecipitate with β1 integrin (Calderwood et al., 2001; Nakajima et al., 2005), we tested the requirement for ligand engagement in the recruitment of FLNa and IQGAP1 to β1 integrin. FLNa and IQGAP1 did not co-precipitate with β1 integrin in cells in suspension, but were found in complexes with β1 integrin when cells adhered and spread on FN (Fig. 1G). The recruitment of FLNa and IQGAP1 to β1 integrin increased with time as cells spread on FN (Fig. 1G), and correlated with Rac1 inactivation (Fig. 1D,E). As the kinetics of FLNa and IQGAP1 recruitment to β1 integrin correlated with the inactivation of Rac1 (Fig. 1G), and as suppression of FLNa and IQGAP1 expression resulted in elevated and sustained Rac1 activity (Fig. 1D,E), we hypothesised that FLNa and IQGAP1 might cooperate to constrain Rac1 activity following integrin engagement.

To test if FLNa and IQGAP1 cooperate to constrain Rac1 activity, we initially assessed whether they were part of the same protein complex in cells. Reciprocal co-immunoprecipitation experiments demonstrated an association between FLNa and IQGAP1 (Fig. 2A). By contrast CD98hc, a protein enriched in FN-induced adhesion complexes that links β1 integrin to another small GTPase, Arf6 (Humphries et al., 2009), did not co-immunoprecipitate with either FLNa or IQGAP1 (Fig. 2A). To complement immunoprecipitation analyses, the subcellular distribution of FLNa and IQGAP1 was assessed. FLNa and IQGAP1 demonstrated substantial co-localisation with each other (Fig. 2B,C) and with the adhesion complex protein paxillin in cells plated on FN (Fig. 2C). These data suggest that FLNa and IQGAP1 associate, either directly or indirectly, at the leading edge to regulate Rac1 activity.

As FLNa and IQGAP1 co-purified with β1 integrin upon FN engagement (Fig. 1G), we studied the role of integrin activation in the formation of the FLNa–IQGAP1 complex. Anti-β1 integrin antibodies that induce different integrin activation states (Byron et al., 2009) were used to dictate the conformation of integrins on the surface of cells. MEFs expressing human β1 integrin–GFP (Fig. 2D) and U2OS cells expressing FLNa–RFP and IQGAP1–GFP (Fig. 2E) were allowed to spread on immobilised 12G10 (a mouse anti-human monoclonal antibody that induces a high affinity, activated conformation of β1 integrin) or 4B4 (a mouse anti-human monoclonal antibody that induces a low affinity, inactive conformation of β1 integrin). Extensive co-localisation of FLNa and IQGAP1 was observed in the lamellipodia of cells with active β1 integrins but not inactive β1 integrins (Fig. 2D,E). This finding suggests that integrin activation regulates the subcellular distribution of FLNa and IQGAP1.

To analyse further the relationship between integrin activation and FLNa–IQGAP1 complex formation, β1 integrin was immunoprecipitated using activation-state-specific anti-β1 integrin antibodies (Fig. 3A). FLNa, IQGAP1 and talin, a protein that associates with high affinity integrin (Tadokoro et al., 2003), co-purified with active β1 integrin but not with inactive β1 integrin or transferrin receptor (Fig. 3A). Given that FLNa and IQGAP1 co-localise when cells are plated on activating anti-integrin antibodies and co-purify with active β1 integrin, we tested whether integrin activation controls the localisation of the FLNa–IQGAP1 complex. Micropatterned surfaces were fabricated to immobilise conformation-specific anti-β1 integrin antibodies and therefore spatially constrain integrin conformation. Both IQGAP1–GFP and FLNa–RFP were specifically recruited and enriched to sites of integrin activation but not to regions of integrin inactivation (Fig. 3B). To test if either FLNa or IQGAP1 were necessary to recruit the complex to active β1 integrin, assays were performed with cells depleted of FLNa or IQGAP1. Suppression of either FLNa or IQGAP1 expression was sufficient to substantially reduce the recruitment of both proteins to anti-active β1 integrin antibody-coated patches (Fig. 3B) and to reduce co-purification with active β1 integrin (Fig. 3C). These data indicate that both FLNa and IQGAP1 are necessary to recruit the complex to active β1 integrin. Taken together, these findings led us to postulate that integrin engagement initially triggers the activation of Rac1 followed by the dynamic recruitment of a complex containing FLNa and IQGAP1 that serves to reduce Rac1 activity and prevent dysregulated Rac1 signalling. This is consistent with the fact that the FLNa–IQGAP1 complex does not localise to focal adhesions, which are sites of integrin activation (Askari et al., 2010), but instead to lamellipodia, at early adhesion structures (Fig. 2C), which are sites of dynamic Rac1 activation (Machacek et al., 2009).

FLNa and IQGAP1 have been reported to bind directly to Rac1 (Kuroda et al., 1996; Ohta et al., 1999); however, these proteins lack intrinsic GAP activity (Ho et al., 1999). Therefore, we hypothesised that FLNa and IQGAP1 could modulate Rac1 activation through the recruitment of additional GTPase activity modulators. To identify new FLNa and IQGAP1 binding partners, GFP pull-downs were performed in cells transiently expressing FLNa–GFP, IQGAP1–GFP, Rac1–GFP, RCC2–GFP or GFP alone and co-purifying associated proteins were identified by mass spectrometry (MS) (Fig. 4A–C). GFP and RCC2–GFP, a protein previously reported to regulate Rac1 activity (Humphries et al., 2009), were used as controls. In total, 1371 proteins were detected in the five pull-downs (supplementary material Table S1). Following normalisation and specificity validation against the GFP and RCC2–GFP controls, 141, 122 and 408 proteins were identified as specifically enriched to FLNa–GFP (supplementary material Table S2), IQGAP1–GFP (supplementary material Table S3) and Rac1–GFP (supplementary material Table S4), respectively. These datasets were integrated with PPI databases to construct a network-based view of the FLNa–IQGAP1–Rac1 complex (Fig. 4C). As a validation of this approach, many known binding partners of the bait proteins were identified (Fig. 4C). For example, refilin B (Gay et al., 2011) and R-Ras (Griffiths et al., 2011) were identified in the FLNa–GFP pull-down; calmodulin (Ho et al., 1999), Rac1 and Cdc42 (Bashour et al., 1997) in the IQGAP1–GFP pull-down; and RhoGDI (Chuang et al., 1993) in the Rac1–GFP pull-down (supplementary material Tables S2–S4). Interestingly, FLNa was found to be weakly enriched in the IQGAP1–GFP pull-downs and IQGAP1 was detected but not enriched in the FLNa–GFP pull-downs (supplementary material Table S1). These observations might suggest that the IQGAP1–FLNa association is not direct.

To highlight proteins detected by MS that may have putative roles in the regulation of Rac1, Gene Ontology analysis was used to identify proteins annotated with the term ‘GTPase regulator activity’, and these proteins were hierarchically clustered using normalised spectral counts as a measure of relative protein abundance (Fig. 4B). Using this approach, two Rac1 GAPs, RacGAP1 (MgcRacGAP or CYK4) and srGAP2, were found to be specifically enriched in the IQGAP1–GFP and Rac1–GFP pull-downs, but only RacGAP1 was robustly identified in all three biological repeats (Fig. 4B). In addition, analysis of the sub-network connecting Rac1, FLNa and IQGAP1, using an algorithm that clusters nodes within a network as a function of their connectivity, revealed that RacGAP1 had a central position in this network (Fig. 4C). Surprisingly, no known Rac1 GAPs were identified in the FLNa–GFP pull-down even though such interactions have been reported previously (Ohta et al., 2006). We therefore focused on the potential interaction between IQGAP1 and RacGAP1 as a possible mechanism by which the FLNa–IQGAP1 complex might suppress Rac1 activity.

A RacGAP1–IQGAP1 association was supported by immunoblotting of GFP pull-downs of FLNa–GFP, IQGAP1–GFP and Rac1–GFP (Fig. 4D), and further validated by co-immunoprecipitation of endogenous RacGAP1 with IQGAP1 (Fig. 5A). In addition, proximity ligation assays between endogenous IQGAP1 and endogenous RacGAP1 indicated that both were in close proximity within cells (theoretical maximum distance of 30–40 nm) suggesting that these two proteins are likely to interact directly (Fig. 5B). Furthermore, although the subcellular localisation of endogenous RacGAP1 and RacGAP1–FLAG were mostly nuclear [consistent with the reported role of RacGAP1 in cell cycle regulation (Minoshima et al., 2003)], both endogenous RacGAP1 and RacGAP1–FLAG displayed substantial co-localisation with IQGAP1 in membrane ruffles (Fig. 5C). The membrane localisation of endogenous RacGAP1 was specific and was confirmed by the loss of membrane RacGAP1 staining in RacGAP1-knockdown cells (Fig. 5C). Importantly, suppression of IQGAP1 or FLNa expression triggered the loss of RacGAP1 localisation from membrane ruffles (Fig. 5C). All together, these data indicate that the FLNa–IQGAP1 complex recruits RacGAP1 to the cell periphery. In addition, knockdown of RacGAP1 expression triggered an increase in Rac1 activity following integrin–FN engagement (Fig. 5D). Furthermore, immunoprecipitation using activation-state-specific anti-β1 integrin antibodies revealed that RacGAP1 co-purifies only with active integrins (Fig. 5E). Since RacGAP1 exhibits functional GAP activity toward Rac1/2/3 (Bastos et al., 2012; Touré et al., 1998), these findings are consistent with a mechanism whereby the formation of a FLNa–IQGAP1 complex, at sites of integrin activation, functions to recruit RacGAP1 and constrain Rac1 activity.

As FLNa and IQGAP1 co-purified with β1 integrin upon FN engagement (Fig. 1G), and as the FLNa–IQGAP1 complex was found to recruit RacGAP1 to the cell periphery, we tested the requirement for ligand engagement in the assembly of the FLNa–IQGAP1–RacGAP1 complex (Fig. 5F). FLNa and RacGAP1 co-precipitated with IQGAP1 only when cells were adhered and spread on FN (Fig. 5F). The recruitment of FLNa and RacGAP1 to IQGAP1 increased as cells spread on FN (Fig. 5F), correlating with Rac1 inactivation (Fig. 1D,E) and with FLNa and IQGAP1 recruitment to β1 integrin (Fig. 1G). Taken together, these data demonstrate that a FLNa–IQGAP1–RacGAP1 complex assembles at β1 integrin activation sites, during cell spreading on FN, to suppress Rac1 activity.

FLNa, IQGAP1 and RacGAP1 were demonstrated to modulate Rac1 activity during cell spreading on FN by using an effector pull-down approach (Fig. 1; Fig. 5D). To validate these findings further, the activation state of Rac1 was monitored using a Rac1-Raichu probe by fluorescence lifetime microscopy (FLIM), in live cells previously knocked down for FLNa, IQGAP1 or RacGAP1 (Fig. 6A,C,E) (Itoh et al., 2002). To rule out possible off-target effects, these experiments were performed using three individual siRNA oligos targeting FLNa and IQGAP1 and two individual siRNA oligos targeting RacGAP1 (Fig. 6B,D). In comparison with control-treated cells, suppression of FLNa, IQGAP1 or RacGAP1 expression resulted in lower lifetime of the Rac1-Raichu probe, in both human and mouse cell lines, indicating higher Rac1 activation (Fig. 6A,C,E). These data are consistent with results obtained previously, using effector pull-down assays (Fig. 1D,E; Fig. 5D), and demonstrate that FLNa, IQGAP1 and RacGAP1 are involved in suppressing Rac1 activity during cell spreading on FN.

Rac1 has a central role in regulating cell membrane protrusion during directional cell migration (Pankov et al., 2005; Wu et al., 2009). Having identified a mechanism whereby FLNa, IQGAP1 and RacGAP1 cooperate to regulate Rac1 activity, we investigated the role of these molecules in regulating cellular protrusion and migration. To assess actin dynamics and membrane protrusive activity by live-cell microscopy, FLNa, IQGAP1 or RacGAP1 expression was knocked down in U2OS cells transiently transfected with Lifeact-mEGFP (supplementary material Movies 1–4). For each condition, cell outlines were tracked (Fig. 6F) and membrane velocity measured (Fig. 6G) and quantified (Fig. 6H) using QuimP11 (Tyson et al., 2010). Consistent with increased Rac1 activity, suppression of FLNa, IQGAP1 or RacGAP1 expression enhanced membrane protrusive activity (Fig. 6F–H and supplementary material Movies 1–4) compared with control cells.

Given that FLNa, IQGAP1 and RacGAP1 suppress membrane protrusion, we next investigated whether these proteins also regulate cell migration. FLNa, IQGAP1 and RacGAP1 expression was knocked down in U2OS cells (Fig. 7A–D), and FLNa and IQGAP1 were knocked down either individually or together in MEFs (Fig. 7E,F), and migration on cell-derived matrices (CDMs) was assessed. Non-targeting-siRNA-treated cells migrated in a directionally persistent manner on CDMs (Fig. 7C,E) (Pankov et al., 2005). However, siRNA-mediated suppression of FLNa, IQGAP1 or RacGAP1 expression, in both MEFs and U2OS cells, triggered a decrease in directional migration (Fig. 7C,E) accompanied by a concomitant increase in cell speed (Fig. 7D,F). The loss of persistent migration is consistent with the observed elevation of Rac1 activity (Fig. 1D–F) and the consequent increase in protrusive activity in cells plated on FN (Fig. 6A–C) or CDMs (Fig. 7A) (Bass et al., 2007; Pankov et al., 2005). Combined suppression of both FLNa and IQGAP1 expression did not further decrease directionality or increase cell speed, suggesting that these proteins cooperate in regulating cell migration (Fig. 7E,F). These data indicate that FLNa, IQGAP1 and RacGAP1 play a fundamental role in regulating cell migration by regulating both cell directionality and cell speed.

We have identified a novel mechanism that coordinates Rac1 activity downstream of β1 integrin activation. This conclusion is based on the following major findings: (1) FLNa and IQGAP1 are part of a complex that forms downstream of β1 integrin activation; (2) FLNa and IQGAP1 suppress Rac1 activity following FN-integrin engagement; (3) IQGAP1 recruits RacGAP1 to suppress Rac1 activity; (4) Knockdown of FLNa, IQGAP1 and RacGAP1 disrupts membrane protrusion and directional cell migration. Thus, our data suggest a model whereby integrin engagement, followed by FLNa, IQGAP1 and RacGAP1 recruitment, maintains tight control over Rac1 activity and thereby coordinates directional cell migration (Fig. 8).

Accumulating evidence indicates that IQGAP1 can function to regulate the activity of small GTPases both positively and negatively (Bashour et al., 1997; Casteel et al., 2012; Jeong et al., 2007). To date, studies of IQGAP1-dependent Rac1 activity modulation have focused primarily on the response triggered by growth factors or cytokines (Brown and Sacks, 2006; David et al., 2011; Hu et al., 2009). This is the first report to describe the essential role that cell–ECM engagement plays in IQGAP1-dependent modulation of Rac1 activity, and in contrast to stimulation with soluble extracellular cues, we show that this mechanism suppresses Rac1 activity.

IQGAP1 contains an inactive RasGAP domain and does not act directly as either a GEF or a GAP (Hart et al., 1996; Ho et al., 1999), so the precise mechanism by which IQGAP1 regulates Rac1 activity is unclear. It has been suggested that binding of Cdc42 to IQGAP1 inhibits Cdc42 intrinsic GTPase activity in vitro, prolonging Cdc42 activation (Hart et al., 1996; Ho et al., 1999). Thus, it was proposed that IQGAP1-mediated potentiation of Rac1 activity could occur through a similar mechanism (Brown and Sacks, 2006). Here, we report, for the first time, that IQGAP1 can also suppress Rac1 activity and we have identified RacGAP1 as the negative regulator of Rac1 that is recruited by IQGAP1. Consistent with our model, others have reported that IQGAP1 binds preferentially to GTP-Rac1 (Kuroda et al., 1996), suggesting that IQGAP1 might immobilise active Rac1 at sites of integrin activation and subsequently recruit RacGAP1 to promote GTP hydrolysis. Consistent with IQGAP1 suppressing Rac1 activity, a recent study reported that IQGAP1 promoted RhoA activation (Casteel et al., 2012), a well-established antagonist of Rac1 activity (Guilluy et al., 2011). Proteomic analyses of IQGAP1-associated proteins identified a putative RhoA GEF, namely pleckstrin homology domain-containing family G member 3, which could be involved in IQGAP1-mediated activation of RhoA (Fig. 4B and supplementary material Table S3). In addition to Rac1 and RhoA, there is evidence that IQGAP1 can recruit and regulate other small GTPases including Cdc42 (Bashour et al., 1997; Kuroda et al., 1996), Arf6 (Hu et al., 2009) and Rap1 (Awasthi et al., 2010; Jeong et al., 2007) and their downstream effectors (supplementary material Table S3) (Noritake et al., 2004; Pelikan-Conchaudron et al., 2011; Sakurai-Yageta et al., 2008). A recurring theme from these studies, and from ours, is that modulation of GTPase activity by IQGAP1 is due to the recruitment of other proteins. Therefore, it is tempting to speculate that IQGAP1 could act as a ‘small GTPase scaffolding platform’, recruiting both small GTPase regulators and effectors to the leading edge, to orchestrate cell behaviour precisely. Given that different extracellular stimuli can promote both Rac1 activation and inhibition, via IQGAP1, it is likely that the mechanisms by which IQGAP1 coordinates GTPase activity will be exquisitely dependent on both micro environmental and cellular context. The next challenge will be to determine if these molecules are recruited to IQGAP1 simultaneously or at discrete subcellular locations and/or time, and to identify how the convergence of intracellular signals downstream of other receptors, such as growth factor receptors or syndecans, and/or post-translational modifications dictate the differential recruitment of these GTPase regulators and effectors to IQGAP1.

RacGAP1, as part of the centralspindlin complex, has been described previously to play a regulatory role during cytokinesis (Hirose et al., 2001; Minoshima et al., 2003; Oceguera-Yanez et al., 2005; Zhao and Fang, 2005). Interestingly, FLNa, IQGAP1 and Rac1 have also been reported to be important for cell division (Canman et al., 2008; Eng et al., 1998; Mondal et al., 2012; Padmanabhan et al., 2011). In addition, cytokinesis regulators and spindle components such as septins (isoforms 2 and 11) (Estey et al., 2010), calmodulin (Yu et al., 2005), kinesin-like proteins (KIF11 and KIF23) (Lee et al., 1995) and cohesin (subunits SA-1 and SA-2) (Hauf et al., 2001) were identified by mass spectrometry in the FLNa–GFP, IQGAP1–GFP and Rac1–GFP pull-downs (supplementary material Table S1). The possibility that the pathway described here could play a role in cell division will be the subject of future investigations.

Our data demonstrated that RacGAP1 plays an important role in regulating Rac1 activity at the cell leading edge during interphase. Recent work has also identified an interphase role for RacGAP1 at cell–cell junctions (Ratheesh et al., 2012). These reports of cytokinesis- and centralspindlin-independent RacGAP1 functions suggest that RacGAP1 has a broader range of functions than previously anticipated. In this context, IQGAP1 is known to localise at cell–cell junctions (Krishnan et al., 2012; Lehtonen et al., 2005; Noritake et al., 2004) and cell–cell junction proteins were identified as IQGAP1-associated proteins (e.g. β-catenin) (supplementary material Table S3). Therefore, IQGAP1 may also be involved in recruiting RacGAP1 at cell–cell junctions.

Our data demonstrated that FLNa and IQGAP1 specifically co-purified and localised with activated β1 integrin. Both IQGAP1 and FLNa have been previously reported to be recruited to β1 integrin (Calderwood et al., 2001; Kiema et al., 2006; Nakajima et al., 2005; Takahashi and Suzuki, 2006; Travis et al., 2004); however, neither FLNa or IQGAP1 localised to classical focal adhesions, which are sites of integrin activation (Askari et al., 2010), but instead to lamellipodia, at early adhesion structures (Fig. 2C). It has been reported that, in lamellipodia, unligated integrin can get activated by actin polymerisation and therefore be ‘primed’ to probe for adhesion sites (Galbraith et al., 2007). Therefore, it is possible that FLNa and IQGAP1 specifically target the sub-set of unligated primed integrins to restrict the formation of uncontrolled protrusions. In addition, the recruitment of FLNa to active integrin is surprising. The current model of the integrin–FLNa association would predict that FLNa binds inactive integrins. FLNa has been reported to directly bind to the β1 integrin tail (Calderwood et al., 2001; Ehrlicher et al., 2011; Travis et al., 2004), in a region overlapping with the talin binding site (Kiema et al., 2006). FLNa binding to integrin is thought to compete with talin, a key β1 integrin activator (Tadokoro et al., 2003), and therefore to inactivate integrin. However, our data demonstrated that FLNa is recruited specifically to activated integrins along with talin. Therefore, in this context we predict that the recruitment of FLNa to integrin is unlikely to be direct and one of the next challenges leading from this study will be to determine how FLNa and IQGAP1 are recruited to activated integrins.

While we discovered that RacGAP1 was recruited to a complex containing FLNa and IQGAP1 upon integrin activation, it has been reported previously that FLNa can recruit an alternative suppressor of Rac1 activity, FilGAP, and this association is modulated by mechanical strain (Ehrlicher et al., 2011; Nieves et al., 2010; Ohta et al., 2006). Although we were unable to detect the association of FilGAP with GFP–FLNa in our mass spectrometric analyses, the fact that multiple mechanisms exist to limit Rac1 activity upon integrin activation, or during integrin-mediated mechanosensation, highlights the importance of constraining Rac1 activity during these processes. Given the fundamental role that Rac1 plays in regulating actin polymerisation and cellular behaviour it is perhaps not surprising that Rac1 activity must be tightly coordinated both spatially and temporally.

Our experiments demonstrated that suppression of FLNa, IQGAP1 or RacGAP1 expression disrupted cell migration. Silencing any of these proteins not only increased the speed of migration but also triggered a loss in the ability to migrate in a directionally-persistent manner. Others have reported previously either no migratory defect (Baldassarre et al., 2009) or a decreased speed of cell migration (Leung et al., 2010) following FLNa knockdown for cells on FN. These discrepancies could be explained by the different mode of migration that cells utilise on fibrillar matrices compared to 2D FN (Petrie et al., 2009). Migration defects observed on CDM could be explained by high Rac1 activity and the concomitant increase in protrusion dynamics measured in these cells (Bass et al., 2007; Pankov et al., 2005). This is consistent with our proposed model that FLNa, IQGAP1 and RacGAP1 are recruited to newly formed protrusions in order to restrict their expansion by locally suppressing Rac1 activity. Indeed, during a protrusion event, a cell explores its environment by making new transient cell–ECM adhesions. Integrin activation at these adhesion sites results in Rac1 activation and promotion of the initial protrusion. In the absence of an off switch to break the cycle of integrin and Rac1 activation, protrusion events would become dysregulated and lack control. The recruitment of the FLNa–IQGAP1 complex to sites of integrin activation may be one mechanism whereby cell protrusion events can be spatially restricted.

In conclusion, we report a signalling pathway linking β1 integrin to Rac1 activity that controls cell migration. We envision these components as the nodes of a much broader regulatory network, integrating multiple matrix-associated and soluble signals from the microenvironment to control cell behaviour spatially. The next challenge will be to identify other components of this network, their inter-connection, and how they transduce information to regulate cell migration. It is likely that systems-level proteomic analysis of the components of this network, using IQGAP1 and FLNa binding partners as a starting point, will help to unravel the structure, the dynamics, and the regulatory mechanisms of this network.

Wild-type and β1-integrin–GFP-expressing MEFs were maintained in Dulbecco's modified Eagle's medium (DMEM; Sigma-Aldrich) supplemented with 10% (vol/vol) foetal calf serum (FCS; Biosera) and 20 U/ml mouse interferon γ (Sigma-Aldrich) at 33°C, 5% (vol/vol) CO₂. Human foreskin fibroblasts (HFFs), mouse 3T3 cells, human 293T embryonic kidney cells and human U2OS osteosarcoma cells were cultured in DMEM supplemented with 10% (vol/vol) FCS at 37°C, 5% (vol/vol) CO₂.

Mouse anti-human β1 integrin monoclonal antibody, JB1A was provided by John Wilkins (University of Manitoba, Winnipeg, MB, Canada). Mouse anti-human β1 integrin monoclonal 12G10 and rat anti-human β1 integrin monoclonal 9EG7 (which recognise receptor with high affinity for ligand, termed ‘active’) and mAb13 (which recognises receptor with low affinity for ligand, termed ‘inactive’) were produced as reported previously (Mould et al., 1995; Mould et al., 1996). Function-blocking mouse anti-human β1 integrin monoclonal 4B4 was obtained from Beckman Coulter and function-neutral mouse anti-human β1 integrin monoclonal K20 was obtained from Santa Cruz Biotechnology. Other mouse monoclonal antibodies used were directed against human paxillin (clone 349; BD Biosciences), human transferrin receptor (H68.4; Invitrogen; and OKT9; European Collection of Cell Cultures), mouse and human actin (AC-40; Sigma-Aldrich), mouse IQGAP1 (K0100-3; MBL international), mouse and human Rac1 (BD Biosciences), human RacGAP1 (1G6; Sigma-Aldrich) and GFP (Roche). Polyclonal antibodies used were goat anti-human CD98hc (C20; Santa Cruz Biotechnology), rabbit anti-human IQGAP1 (H-109; Santa Cruz Biotechnology), mouse anti-human IQGAP1 (D-3; Santa Cruz Biotechnology), rabbit anti-human RacGAP1 (PA5-22265; Thermo Scientific), rabbit anti-human FLNa (H-300; Santa Cruz Biotechnology), rabbit anti-mouse FLNa (A301-133A; Bethyl Laboratories) and goat anti-human talin (C-20; Santa Cruz Biotechnology). Secondary Alexa-Fluor-647-conjugated anti-mouse and Alexa-Fluor-488-conjugated anti-rabbit antibodies were used for immunofluorescence, and Alexa-Fluor-680-conjugated or Alexa-Fluor-800-conjugated antibodies were used for western blotting; all were obtained from Invitrogen. Rabbit IgG and bovine plasma FN were obtained from Sigma-Aldrich. A codon optimised RacGAP1–FLAG was synthesised by Genscript, and cloned into the pcDNA3 vector (EcoRI-BamHI). The following plasmids were obtained as gifts: GST-PAK-1 CRIB domain (from Kozo Kaibuchi, Nagoya University School of Medicine, Nagoya, Japan), FLNa–RFP and FLNa–GFP (from Fumihiko Nakamura, Brigham and Women's Hospital, Boston, MA), IQGAP1–eGFP (from Robert Grosse, Institute of Pharmacology, University of Marburg, Marburg, Germany), and Lifeact-mEGFP (from Roland Wedlich-Söldner, Max Planck Institute of Biochemistry, Martinsried, Germany).

Poly(dimethyl siloxane) (PDMS; Sylgard 184; Dow Corning) stamps with patterns (2 µm in diameter) were cast from photoresist patterned silicon wafers as described previously (Tan et al., 2004). To spatially constrain β1 integrin conformation, PDMS was inked with activation-state-specific β1 integrin antibodies (12G10, active; 4B4, inactive) at 50 µg/ml in PBS for 1 hour at room temperature. The stamps were then submerged in sterile water and dried using a stream of N₂. The stamps were placed feature-face down on PDMS-coated glass coverslips that were treated with ultraviolet ozone (Jelight Company) to transfer the patterned protein to the substrates. Finally, Pluronic F-127 (Sigma-Aldrich) was adsorbed to prevent unwarranted protein adsorption to the PDMS substrates.

The expression of proteins of interest was reduced using 100 nM ON-TARGETplus modified siRNA (Dharmacon, Thermo Fisher) and Lipofectamine 2000 (Invitrogen) according to manufacturer's instructions. Two rounds of transfection were necessary to achieve efficient knockdown. The ON-TARGETplus non-targeting siRNA (Dharmacon) was used as a negative control. The siRNA oligos used to knock down IQGAP1 were siIQGAP1 #1 (mouse IQGAP1, J-040589-12), siIQGAP1 #2 (human IQGAP1, J-004694-06) and siIQGAP1 #3 (mouse IQGAP1, J-040589-10). The siRNA oligos used to knock down FLNa in mouse cell lines were siFLNa #1 (J-058520-20) and #3 (J-058520-18). The FLNa smart pool (L-012579-00, Dharmacon) was used to knock down FLNa in human cell lines. The siRNA oligos used to knock down RacGAP1 in human cell lines were siRacGAP1 #1 (J-008650-06) and #2 (J-008650-08). Levels of knockdown were verified by western blotting for every experiment.

Cells were plated on FN for 2 hours, washed with cold PBS and lysed in CSK buffer [0.5% (wt/vol) Triton X-100, 10 mM Pipes, pH 6.8, 150 mM NaCl, 150 mM sucrose, 3 mM MgCl₂, 10 µg/ml leupeptin, 10 µg/ml aprotinin, 0.5 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 2 mM Na₃VO₄]. Lysates were clarified (centrifugation at 12,000 g, 4°C for 1 minute) and incubated with antibody-coated protein G magnetic beads (1 µg of antibody per 10 µl of beads) (Dynabeads; Invitrogen) for 1 hour at 4°C. Complexes bound to the beads were isolated using magnets, washed three times with ice-cold lysis buffer and eluted in Laemmli reducing sample buffer.

Active Rac1 was affinity purified from lysates using an effector pull-down approach with GST-PAK beads. Cells were serum starved and maintained in suspension for 1 hour before plating on 10 µg/ml FN-coated dishes. At various times, cells were lysed in ice-cold lysis buffer [20 mM Hepes, pH 7.5, 1% (wt/vol) Igepal, 0.5% (wt/vol) sodium deoxycholate, 140 mM NaCl, 4 mM EDTA, 4 mM EGTA, 10% (wt/vol) glycerol), and lysates were clarified by centrifugation at 12,000 g, 4°C, for 1 min. Lysates were then incubated with GST-PAK beads for 1 hour at 4°C. Beads were washed three times with ice-cold lysis buffer, and active GTPase was eluted off beads by addition of Laemmli reducing sample buffer. Samples were then resolved by SDS-PAGE and analysed by western blotting.

GFP-tagged proteins were pulled down from 293T cell lysates using GFP-Trap agarose beads (Chromotek) according to the manufacturer's instructions. 293T cells transiently expressing bait GFP-tagged protein were lysed in 0.5% (wt/vol) NP-40, 10 mM Tris-HCl, pH 7.2, 250 mM NaCl, 0.5 mM EDTA, 10% (wt/vol) glycerol, 10 µg/ml leupeptin, 10 µg/ml aprotinin, 0.5 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 2 mM Na3VO4. Lysates were then sonicated (3 pulses of 10 seconds; VibraCell VCX 500; Sonics & Materials) and centrifuged at 12,000 g for 5 min. Clarified lysates were incubated with GFP-Trap agarose beads for 1 hour at 4°C. Complexes bound to the beads were isolated by centrifugation, washed three times with ice-cold lysis buffer and eluted in Laemmli reducing sample buffer. GFP tags of Rac1–GFP, RCC2–GFP, FLNa–GFP and IQGAP1–GFP were all N-terminal.

Protein extracts were separated under denaturing conditions by SDS-PAGE (4–12% Bis-Tris gels; Invitrogen) and transferred to nitrocellulose membrane. Membranes were blocked overnight at 4°C with blocking buffer (Sigma-Aldrich) and then incubated with the appropriate primary antibody diluted in blocking buffer (Sigma-Aldrich) for 2 hours. Membranes were washed with PBS and then incubated with the appropriate fluorophore-conjugated secondary antibody diluted 1∶5000 in blocking buffer for 30 minutes. Membranes were washed in the dark and then scanned using an Odyssey infrared imaging system (LI-COR Biosciences). Band intensity was determined by digital densitometric analysis using Odyssey software (version 2.1; LI-COR Biosciences).

MEFs were transfected using Lipofectamine PLUS reagent (Invitrogen) and U2OS cells were transfected using Lipofectamine 2000 reagent. Cells were seeded in six-well plates and allowed to reach 80% confluence prior to transfection. For Lipofectamine Plus-mediated transfections, the transfection mixture consisted of 2 µg plasmid pre-incubated for 10 minutes at room temperature with 5 µl PLUS reagent diluted in 250 µl OptiMEM and 20 µl of Lipofectamine pre-incubated for 10 minutes at room temperature with 250 µl OptiMEM. For Lipofectamine 2000-mediated transfections, the transfection mixture consisted of 4 µg plasmid diluted in 250 µl OptiMEM and 7 µl Lipofectamine 2000 pre-incubated with 250 µl OptiMEM for 5 minutes at room temperature. In each case, the transfection mixtures were then incubated for 20 minutes at room temperature. The medium was removed from cells and replaced with 500 µl of OptiMEM and the appropriate transfection mixture. Following 4 hours of incubation, the transfection mixture was removed, fresh medium was added, and transfected cells were used the following day.

293T cells were transfected using 100×polyethyleneimine reagent (Sigma-Aldrich). Cells were seeded in a 15 cm dish and allowed to reach 80% confluence prior to transfection. The transfection mixture consisted of 12 µg plasmid diluted in 500 µl OptiMEM and 54 µl 1×polyethyleneimine (diluted with 150 mM NaCl) pre-incubated with 446 µl of OptiMEM for 5 minutes at room temperature. Transfection mixtures were incubated for 20 minutes at room temperature. Medium was then removed from cells and replaced with 10 ml growth medium and the transfection mixture. Following 10 hours of incubation, the transfection mixture was removed, fresh medium was added, and transfected cells were used the following day.

Cells were fixed in 4% (wt/vol) paraformaldehyde for 20 minutes, washed with PBS and permeabilised with PBS containing 0.5% (vol/vol) Triton X-100 for 3 minutes. Cells were then washed with PBS, incubated with primary antibodies for 45 minutes at room temperature, washed with PBS and incubated for a further 45 minutes with the appropriate secondary antibodies at room temperature. After washing, coverslips were mounted using Prolong Gold mounting medium (Invitrogen), dried overnight at room temperature in the dark and stored at 4°C prior to analysis.

Unless stated otherwise, all images were acquired on a DeltaVision RT restoration microscope (Applied Precision) using a 60×/NA 1.42 Plan Apo objective and the Sedat Quad Set filter set (Chroma 86000v2). The images were collected using a Coolsnap HQ camera (Photometrics) and analysed with ImageJ software. Image co-localisation analysis was performed using the Intensity Correlation Analysis ImageJ plug-in to create PDM (product of the differences from the mean) images, where PDM  =  (red intensity−mean red intensity)×(green intensity−mean green intensity).

Proximity ligation assays were performed according to the manufacturer's instructions (Duolink) using the Duolink In Situ PLA Probes Anti-Rabbit PLUS and Anti-Mouse MINUS, and the Duolink In Situ Detection Reagents Red (Sigma). Cells were counterstained using FITC–phalloidin (Invitrogen), imaged using a DeltaVision RT restoration microscope, and analysed using ImageJ. The number of positive PLA spots was quantified, for each cell, using the particle analysis plugin of ImageJ.

For optimal image resolution, normal culture medium was replaced by Ham's F-12 (Gibco) containing 25 mM HEPES in the absence (Rac1 activity measurement by FLIM) or presence (membrane protrusion analyses) of 2% (vol/vol) FCS, prior to image acquisition. For analyses of membrane protrusion, U2OS cells transiently expressing Lifeact-mEGFP were plated on FN on four-compartment glass-bottom dishes (Greiner Bio-One) for 4 hours and then imaged at 37°C on an Axiovert 200M live cell imaging system (Carl Zeiss) using a 63×/NA 1.40 Plan Apochromat objective, the 62HE filter set and a 470 nm Colibri light-emitting diode fluorescent light source (Carl Zeiss). Images were collected with a Coolsnap HQ camera using μManager (version 1.4) (Edelstein et al., 2010) at 3 minute intervals for 7 hours. The outline of each cell was segmented, divided into 400 points connected by edges, and these 400 points were tracked over the recording period using the ImageJ plug-in, QuimP11 (Tyson et al., 2010). The instant velocity of each of the 400 points was calculated between each frame using QuimP11 and velocity maps of the cell outline were generated using the QuimP11 MATLAB script in MATLAB (MathWorks). The average membrane velocity was calculated for each cell as an average of the absolute values of the instant velocity of the 400 points over the recording period.

For analysis of Rac1 activity by FLIM, U2OS and 3T3 cells transiently expressing the Raichu-Rac1 probe (Raichu-1011X) (Itoh et al., 2002), were plated on FN on four-compartment glass-bottom dishes (Greiner Bio-One) for 30 minutes and then imaged at 37°C. Images were collected using a frequency domain FLIM system (Marianis system) equipped with Lambert instruments FLIM module (3i) and a 100× objective at 15 minute intervals for 1 hour. Image analysis was performed using Slidebook (3i) and the polar plot method to calculate nearest single lifetime of CFP. Förster resonance energy transfer (FRET) efficiency was calculated using the following formula: FRET efficiency (%)  =  100*(1–t_da/t_d) where ‘t_d’ represents the lifetime of the donor (CFP; experimentally derived as 2.4594587 nseconds), and ‘t_da’ represents the lifetime of the donor in the presence of the acceptor (CFP/YFP probe).

To generate CDM, HFFs were seeded at a density of 50,000 cells/ml in a 24-well plate. When confluent, cells were cultured for a further 10 days, with medium changed every 48 hours to complete medium supplemented with 50 µg/ml ascorbic acid (Sigma-Aldrich) to ensure collagen cross-linking. Mature matrices were then denuded of cells using lysis buffer (PBS containing 20 mM NH₄OH and 0.5% (vol/vol) Triton X-100). Following PBS washes, matrices were incubated with 10 µg/ml DNase I (Roche) at 37°C for 30 minutes. Matrices were then stored in PBS containing 1% (vol/vol) penicillin/streptomycin and 1% (vol/vol) Fungizone at 4°C prior to use.

For cell migration analyses, MEFs and U2OS cells were seeded at a density of 5,000 cells/ml on CDM and allowed to spread for 4 hours. Cells were then filmed using an AS MDW live-cell imaging system (Leica) with a 5×/NA 0.15 HC Plan Fluotar air objective (magnification, 1.5×). Point visiting was used to allow multiple positions to be imaged within the same time course, and cells were maintained at 37°C and 5% (vol/vol) CO₂. Images were collected using a Coolsnap HQ camera, and eight movies (one picture every 10 minutes) were generated for each condition. To assess cell migration, the speed and directionality of 80 cells was measured (10 cells per movie) using the Manual Tracking plug-in of ImageJ. Cell tracking was performed over a 16 hour time period. Results were computed and analysed by the ImageJ plug-in Chemotaxis Tool.

Protein samples were separated by SDS-PAGE and allowed to migrate 10 mm into a 4–12% polyacrylamide gel. Following staining with InstantBlue (Expedeon), gel lanes were sliced into five 2-mm bands and subjected to in-gel digestion with trypsin as described previously (Shevchenko et al., 1996) with modifications (Humphries et al., 2009).

Samples were analysed by liquid chromatography tandem mass spectrometry using a nanoACQUITY UltraPerformance liquid chromatography system (Waters) coupled online to an LTQ Velos (Thermo Fisher Scientific). Peptides were concentrated and desalted on a Symmetry C18 pre-column (20 mm×180 µm, 5 µm particle size; Waters) and separated on a bridged ethyl hybrid C18 analytical column (250 mm×75 µm, 1.7 µm particle size; Waters) using a 45-minute linear gradient from 1% to 25% (vol/vol) acetonitrile in 0.1% (vol/vol) formic acid at a flow rate of 200 nl/minute. Peptides were selected for fragmentation automatically by data-dependent analysis.

Tandem mass spectra were extracted using extract_msn (Thermo Fisher Scientific) executed in Mascot Daemon (version 2.2.2; Matrix Science). Peak list files were searched against the IPI Human database (version 3.70, March 2010) modified to contain ten additional contaminant and reagent sequences of non-human origin. Searches were submitted to an in-house Mascot server (version 2.2.03; Matrix Science) (Perkins et al., 1999). Carbamidomethylation of cysteine was set as a fixed modification and oxidation of methionine was allowed as a variable modification. Only tryptic peptides were considered, with up to one missed cleavage permitted. Monoisotopic precursor mass values were used, and only doubly and triply charged precursor ions were considered. Mass tolerances for precursor and fragment ions were 0.4 Da and 0.5 Da, respectively. Data were further analysed using the search engine X! Tandem (version 2007.01.01.1) (Craig and Beavis, 2003) implemented from within Scaffold (version 3.00.03; Proteome Software). Data were validated in Scaffold using a threshold of identification of at least 90% probability at the peptide level, at least 99% probability at the protein level and assignment of at least two unique, validated peptides. These acceptance criteria resulted in an estimated protein false discovery rate of 0.1% for all datasets.

Three biological replicates were performed for each GFP pull-down. Relative protein abundance was calculated using the unweighted spectral count of a given protein normalised to the total number of spectra observed in the entire sample and to the molecular weight of that protein (normalised spectral count). Proteins in the FLNa–GFP and IQGAP1–GFP datasets enriched more than twofold over controls (GFP and RCC2) were considered to be specifically recruited. For the Rac1–GFP dataset, since RCC2 has been previously reported to interact with Rac1 (Mollinari et al., 2003), GFP was used as the negative control. Data were converted using PRIDE Converter (version 2.5.5) (Barsnes et al., 2009) and deposited in the PRIDE database (http://www.ebi.ac.uk/pride) (Vizcaíno et al., 2010) under accession numbers 26836–26850. Details of all identified proteins are provided in supplementary material Tables S1–S4.

Gene Ontology analysis was performed using DAVID (version 6.7) (Huang et al., 2009). Proteins annotated with the Gene Ontology term ‘GTPase regulator activity’ (GO:0030695) were selected as GTPase regulators. Proteins were hierarchically clustered on the basis of uncentred Pearson correlation using Cluster 3.0 (C Clustering Library, version 1.50) (de Hoon et al., 2004) and visualised using Java TreeView (version 1.1.6r2) (Saldanha, 2004). PPI network analysis was performed using Cytoscape (version 2.8.1) (Smoot et al., 2011). Proteins were mapped onto a merged human interactome consisting of PPIs reported in the Protein Interaction Network Analysis platform Homo sapiens network (June 2011) (Wu et al., 2009) and literature-curated databases of integrin adhesion-associated proteins (Zaidel-Bar et al., 2007), FLNa and IQGAP1 binders and Rac1 regulators (this study).

Statistical analysis was performed when appropriate, and P values are indicated by an asterisk in the figure legends. Z-tests were performed when the sample size was higher than 30, in all other cases the Student's t-test was used (unpaired, two-tailed, unequal variance).

We thank Fumihiko Nakamura (Brigham and Women's Hospital, Boston, MA), Kozo Kaibuchi (Nagoya University School of Medicine, Nagoya, Japan), Roland Wedlich-Söldner (Max Planck Institute of Biochemistry, Martinsried, Germany) and Robert Grosse (Institute of Pharmacology, University of Marburg, Marburg, Germany) for providing constructs and John Wilkins (University of Manitoba, Winnipeg, MB, Canada) for the JB1A antibody. We thank Alexandre Carisey (University of Manchester, Manchester, UK) for help with the microscopy analyses. We thank Stacey Warwood and David Knight (University of Manchester) for help with the mass spectrometric analyses and Julian Selley (University of Manchester) for bioinformatic support.