Protein 4.1R regulates cell migration and IQGAP1 recruitment to the leading edge

Protein 4.1R is the founding member of a large family of proteins, the band 4.1 superfamily, that contain a highly conserved region known as the ‘FERM domain’. This takes its name from the 4.1 (four point one) and ERM (ezrin, radixin, moesin) proteins, in which it was discovered (Chishti et al., 1998). FERM-containing proteins constitute a diverse group of eukaryotic proteins that bind membrane proteins and lipids. Some of its members (e.g. proteins 4.1, ERMs, talin, focal adhesion kinase) are involved in the organization of the actin cytoskeleton and its connection to the plasma membrane (Bennett and Baines, 2001). Protein 4.1R was originally identified as an 80 kDa component of human red blood cells. In these cells, protein 4.1R stabilizes the spectrin–actin network and mediates its attachment to the overlying lipid bilayer through interactions with integral membrane proteins (Bennett and Baines, 2001). Deficiency of 4.1R in erythrocytes leads to changes in their normal discoid morphology and the acquisition of an oval or elliptical shape with unstable membranes (Bennett and Baines, 2001). Although the expression pattern of 4.1R in mature mammalian red cells is relatively simple, multiple isoforms of 4.1R with variable N-terminal extensions and internal sequences are expressed in nonerythroid cells, mainly as a result of extensive alternative splicing of the 4.1R-encoding pre-mRNA (Conboy, 1999; Tang et al., 1990). The localization of protein 4.1R in nonerythroid cells is not restricted to the subjacent area beneath the plasma membrane, as 4.1R has also been identified at the nucleoskeleton (De Carcer et al., 1995), the centrosome (Krauss et al., 1997; Perez-Ferreiro et al., 2004), the endoplasmic reticulum (Luque et al., 1999) and microtubules (Perez-Ferreiro et al., 2001), thereby showing that 4.1R functions at multiple sites in the cell. The interaction of protein 4.1R with zonula occludens (ZO)-2 (Mattagajasingh et al., 2000) and with β-catenin and E-cadherin (Yang et al., 2009) suggested that 4.1R might act as a linker between tight and adherens junctions and the actin cytoskeleton. It has also been proposed that 4.1R might link splicing factors with the nucleoskeleton (Lallena et al., 1998). These data show that, in nonerythroid cells, 4.1R is a multifunctional protein that creates links between membranes and cytoskeletal and/or nucleoskeletal networks.

Cell migration is a key step in many physiological and pathological processes, such as wound repair, embryonic development, tissue regeneration, angiogenesis and metastasis. The failure of cells to migrate or their migration to inappropriate locations can result in abnormalities or life-threatening consequences (Ridley et al., 2003). One of the key regulators in the cell migration pathway is IQGAP1, an effector of the Rho-family small GTPases Rac1 and Cdc42 (Noritake et al., 2005). IQGAP1 acts as a scaffolding protein near the cell cortex at the leading edge of migrating cells, where it binds and cross-links actin filaments (Ho et al., 1999; Mateer et al., 2002) and associates with additional protein machinery (Fukata et al., 2002; Watanabe et al., 2009; Watanabe et al., 2004).

In this study, we have examined the role of protein 4.1R in cell migration by analyzing control and 4.1R-knockdown cells. Our results (i) show that protein 4.1R polarizes to the leading edges of cells migrating randomly or directionally, (ii) identify IQGAP1 as a partner of 4.1R, (iii) determine that the FERM domain of 4.1R is involved in binding IQGAP1, (iv) demonstrate that depletion of 4.1R impairs cell migration and, most importantly, (v) show that 4.1R is necessary for the recruitment of IQGAP1 to the leading edge, whereas IQGAP1 is not required for protein 4.1R localization. Collectively, these results situate 4.1R upstream of IQGAP1 in the cell migration pathway.

As a first step to studying the role of protein 4.1R in cell migration, we chose to examine the distribution of protein 4.1R in subconfluent, randomly migrating human epithelial ECV304 and PC3 cells using antibodies to different conserved regions of the 4.1R molecule (Fig. 1A). In addition to previously reported nuclear and cytoplasmic distributions (De Carcer et al., 1995), analysis of 4.1R staining with all the available antibodies revealed 4.1R to be localized to the leading edge of the cell and colocalized with lamellipodial F-actin (Fig. 1B,C). The polarized distribution of 4.1R was not due to increased cytoplasmic volume at the leading edge (supplementary material Fig. S1A–F) and was also observed in confluent ECV304 cells induced to migrate directionally in wound-healing assays (Fig. 1D). No 4.1R staining was detected at stress fibres or cell cortex areas enriched in F-actin other than at the leading edge. Ectopic expression of three representative types of 4.1R isoforms, 4.1R¹³⁵, 4.1R⁸⁰ and 4.1R⁶⁰, generated by the use of three alternative translation-initiation sites and, hence, varying in their N-terminal regions (Fig. 1A), suggested that only those containing a complete FERM domain (4.1R¹³⁵ and 4.1R⁸⁰) localize to plasma membrane extensions (Fig. 1E). In agreement with this result, ectopic expression of the 4.1R FERM domain alone was sufficient for lamellipodial localization (Fig. 1E, bottom panels). Western blot analysis revealed that ECV304, PC3 and COS-7 cells, the last used for ectopic expression experiments, expressed similar levels of endogenous 4.1R¹³⁵, 4.1R⁸⁰ and 4.1R⁶⁰ groups of isoforms (Fig. 1F), consistent with previous studies performed in nonerythroid cells (Anderson et al., 1988; Luque et al., 1998). It should be mentioned that the three groups of 4.1R isoforms are generally referred to as 4.1R¹³⁵, 4.1R⁸⁰ and 4.1R⁶⁰, even though their sizes range from ~95 to 160 kDa, ~65 to 80 kDa and ~45 to 60 kDa, respectively, because of alternative expression of internal coding exons of the gene encoding 4.1R (Conboy et al., 1988; Tang et al., 1988). We chose ECV304 cells for our functional study of protein 4.1R in cell migration because they display an extended leading edge, migrate quickly, and have a very suitable size and morphology for cellular studies.

The leading edge compartmentalizes different protein complexes that orchestrate oriented cell motility (Ridley et al., 2003). Because protein 4.1R localized to this region, we analyzed whether 4.1R was involved in cell migration by using a loss-of-function strategy with specific small interfering RNAs (siRNAs) used individually (siRNA1, siRNA2) or as a pool (siRNA_pool). The siRNA1 was less effective than the siRNA2 or the siRNA pool in reducing 4.1R¹³⁵ and 4.1R⁸⁰ levels (Fig. 2A,B). By contrast, none of the 4.1R siRNAs knocked down the 4.1R⁶⁰ isoform, which was previously shown to localize in the cell nucleus (Luque and Correas, 2000), with significant efficiency (Fig. 2A,B). The 4.1R⁶⁰ isoform is better observed in a long exposure blot (Fig. 2A). Immunofluorescence analysis of 4.1R-knockdown cells showed a pronounced reduction of 4.1R immunostaining both at the leading edge and in the cytoplasm, but not in the nucleus (Fig. 2C), consistent with the nuclear localization of the remaining 4.1R⁶⁰. Morphological analysis of 4.1R-deficient cells showed that their area was increased (supplementary material Fig. S1G), hence facilitating their identification within the entire cell population. We then analyzed whether 4.1R-deficient cells had altered cell migration. Persistence is the intrinsic propensity of cells to continue migrating in the same direction for a sustained period without turning, even in the absence of exogenous stimuli (Huttenlocher, 2005). Quantitative time-lapse video microscopy analysis revealed that 4.1R-knockdown cells completely lost directionality and displayed much shorter net translocation (the shortest linear distance from the starting point to the end point of the time-lapse recording) compared with control cells (Fig. 2D; supplementary material Movie 1). 4.1R-knockdown cells displayed a significantly lower index of directionality (ID), which is a measure of the directional persistence of the cells (Fig. 2E), as well as an impaired movement capacity, as reflected by their lower mean velocity (Fig. 2F).

As a second approach to analyzing cell motility, we used wound-healing assays to determine the effects of protein 4.1R silencing on directional cell migration by using time-lapse video recording. Whereas control cells nearly closed the wound after 7 hours (Fig. 2G), it was still not closed in 4.1R-knockdown cells after that time. This delay in wound closure was quantified by measuring the surface of the wound covered by control and 4.1R-knockdown cells at different times after wounding (Fig. 2H). Together, these results indicate that protein 4.1R participates in cell migration.

IQGAP1, a key regulator of cell migration, accumulates at the leading edge of migrating cells, where it binds and cross-links actin filaments (Ho et al., 1999; Mateer et al., 2002). The best-known function of protein 4.1R derives from the studies performed in human erythrocytes. In these cells, protein 4.1R localizes beneath the plasma membrane, binds actin filaments, and establishes links between the subcortical actin skeleton and the plasma membrane (Bennett and Baines, 2001). Thus, in nonerythroid cells, 4.1R might regulate scaffolding proteins that coordinate actin dynamics at the leading edge, such as IQGAP1. As shown in Fig. 3A, these two cytoskeletal regulators, 4.1R and IQGAP1, colocalized at the leading edges of polarized cells migrating in wound-healing assays. IQGAP1 does not localize to the plasma membrane immediately after wounding of the cell monolayer, but, instead, accumulates at the leading edge once cells at the border of the scratch polarize and begin to migrate to close the wound (Watanabe et al., 2004). Interestingly, protein 4.1R dynamics paralleled IQGAP1 translocation to the cell border (Fig. 3B).

The parallel dynamics of 4.1R and IQGAP1 in translocating to the leading edge in wound-healing assays and their colocalization in this region prompted us to analyze their possible physical association by using coimmunoprecipitation assays. The anti-4.1R antibody immunoprecipitated endogenous 4.1R and IQGAP1 (Fig. 3C); reciprocally, the anti-IQGAP1 antibody immunoprecipitated IQGAP1 and 4.1R (Fig. 3D). No association of N-WASP or paxillin with 4.1R was detected in the 4.1R immunoprecipitates (Fig. 3E). To determine whether the 4.1R–IQGAP1 interaction was mediated by actin, we performed coimmunoprecipitation assays in the presence of cytochalasin D, which disrupts actin filaments. Under these conditions, 4.1R also immunoprecipitated IQGAP1 (Fig. 3C) and IQGAP1 immunoprecipitated 4.1R (Fig. 3D), which suggests that actin filaments do not mediate the interaction of 4.1R and IQGAP1. To map the 4.1R region involved in the association, we performed pull-down experiments using GST fusions of 4.1R fragments that encompass the entire 4.1R⁸⁰ molecule (referred to as FERM, CORE and Cter) (Fig. 4A). Using equal amounts of recombinant GST–4.1R proteins (Fig. 4B), endogenous IQGAP1 bound to the FERM and the CORE, but not to the Cter fragment (Fig. 4C,D). The same results were obtained using in-vitro-translated IQGAP1 instead of cell lysates (Fig. 4E,F). Because the FERM and CORE contain a common sequence, corresponding to the C-terminal region of the FERM domain (Fig. 4A), our analysis suggested that this region, C-FERM, is responsible for the association of protein 4.1R with IQGAP1. The interaction of the C-FERM fragment of 4.1R with IQGAP1 was confirmed by pull-down analysis (Fig. 4E,F).

IQGAP1 is a scaffold protein required for the recruitment of Dia1, CLIP-170 and APC to the leading edge (Brandt et al., 2007; Fukata et al., 2002; Watanabe et al., 2004). We investigated whether IQGAP1 also regulates the recruitment of protein 4.1R. Two IQGAP1 siRNAs (siRNA1 and siRNA2) were used individually or as a pool to silence IQGAP1 expression to less than 5% of its normal level (Fig. 5A,B). Under these conditions, protein 4.1R levels were unaffected (Fig. 5A,B). Immunofluorescence of cells directionally migrating in wound-healing assays revealed that, unlike the accumulation of proteins 4.1R, IQGAP1 and actin at the wound front observed in control cells (Fig. 5C), IQGAP1 and F-actin accumulation was notably impaired in 4.1R-knockdown cells (Fig. 5C–E; supplementary material Fig. S2). By contrast, the recruitment of 4.1R to the wound front was not affected in IQGAP1-knockdown cells (Fig. 5C,F). The reduced recruitment of IQGAP1 at the wound front of cells depleted in protein 4.1R was not due to a reduction in IQGAP1 expression (Fig. 5G). Moreover, biochemical fractionation assays revealed that IQGAP1 association with actin-enriched insoluble fractions was not significantly modified upon 4.1R depletion, suggesting that 4.1R does not alter the intrinsic capacity of IQGAP1 to bind actin polymers (Fig. 5H–K).

We wanted to know whether the recruitment of proteins 4.1R and IQGAP1 in randomly migrating cells was regulated in a similar manner as in cells directionally migrating into a wound. Immunofluorescence assays revealed that, unlike the control cells, most of the cell population treated with 4.1R siRNA, which includes cells with 4.1R depleted to different extents, lacked typical extended lamellipodia (Fig. 6A,C). The analysis of cells efficiently silenced for 4.1R (55.2% of all cells) and, hence, completely lacking 4.1R at the plasma membrane indicated that they had reduced levels of actin at the cell periphery (Fig. 6B,C) and presented no accumulation of IQGAP1 at any specific part of the plasma membrane (Fig. 6C, middle panels). APC and Dia1 recruitment to the plasma membrane was also affected in these cells (supplementary material Fig. S3). It is of particular note that, whereas the recruitment of IQGAP1 to the plasma membrane was dependent on 4.1R expression, 4.1R accumulated normally in IQGAP1-knockdown cells (Fig. 6C, bottom panels). The analysis of the cells with only partially depleted 4.1R (24.8% of all cells) showed a reduction in the accumulation of both 4.1R and IQGAP1 at the plasma membrane, compared with control cells. Both proteins, 4.1R and IQGAP1, accumulated with actin at small plasma membrane protrusions resembling peripheral ruffles, which did not become typical sustained lamellipodial extensions (supplementary material Fig. S4 and Movie 1).

Given that protein 4.1R regulated IQGAP1 recruitment to the plasma membrane, whereas IQGAP1 did not regulate that of protein 4.1R, we wanted to compare the effects of silencing 4.1R, IQGAP1 and both proteins 4.1R and IQGAP1 on the most important parameters altered in 4.1R-knockdown cells, that is, the loss of cell directionality and velocity of migration. We found similar defects in directionality and mean velocity of migration in cells singly or doubly knocked down for 4.1R and IQGAP1 (Fig. 6E–G). These defects in random migration were observed for 4.1R-siRNA-treated cells either containing or lacking ruffles (Fig. 6D; supplementary material Fig. S4 and Movies 1 and 2).

The actin cytoskeleton plays a pivotal role in cell migration by promoting the formation of protrusions at the leading edge and providing force, together with molecular motors, to move the cells. Actin nucleators and actin-binding proteins distribute at the leading edge to polymerize new actin filaments or to serve as scaffolds to recruit other proteins necessary for cell migration (Ridley et al., 2003). Although 4.1R was identified as an actin-binding protein almost three decades ago, its role in cell migration has not been investigated before. Our results yield new insights into the function of protein 4.1R in nonerythroid cells by showing that 4.1R plays a key role in cell migration and in the recruitment of IQGAP1 to the leading edge.

The best-known function of protein 4.1R derives from the original studies performed in human red blood cells, in which 4.1R was viewed as a bridging protein between the actin cytoskeleton and the plasma membrane (Bennett and Baines, 2001). The FERM domain of protein 4.1R, which is also present in many proteins including the ERMs, is responsible for 4.1R binding to the cytoplasmic domain of transmembrane proteins (Anderson and Marchesi, 1985; Nunomura et al., 1997; Pasternack et al., 1985). Recent studies have reported that the FERM domain of 4.1R also binds proteins other than transmembrane proteins (Lallena et al., 1998; Perez-Ferreiro et al., 2001). Our study shows that IQGAP1 association with 4.1R also takes place through its FERM domain. FERM domains consist of three lobes, the N, the α and the C lobes, which adopt a cloverleaf-like structure (Han et al., 2000). p55, a palmitoylated peripheral membrane protein belonging to a membrane-associated guanylate kinase homologue family of signalling and cytoskeletal proteins (Nunomura et al., 2000), and tubulin bind to the C lobe (Perez-Ferreiro et al., 2001). Our mapping analysis shows that the C lobe is also responsible for IQGAP1 association.

IQGAP1 modulates actin function by binding directly to F-actin (Mateer et al., 2004) and indirectly through Cdc42 and Rac1 (Watanabe et al., 2004). In addition, IQGAP1 captures growing microtubules by binding CLIP-170, leading to a polarized microtubule array (Fukata et al., 2002). IQGAP1 also binds to Dia1, N-WASP and adenomatous polyposis coli (APC), a multifunctional protein that links IQGAP1 to microtubules (Brandt and Grosse, 2007). APC and Dia1 directly mediate actin filament assembly (Okada et al., 2010), whereas IQGAP1 stimulates actin polymerization through the N-WASP–Arp2/3 complex (Le Clainche et al., 2007). It has been demonstrated that IQGAP1 and APC or CLIP-170 recruitment to the leading edge are interdependent (Watanabe et al., 2004). By contrast, recruitment of IQGAP1 and Dia1 are not mutually regulated, because absence of IQGAP1 abrogates localization of Dia1, whereas IQGAP1 distributes normally to the leading edge in the absence of Dia1 (Brandt et al., 2007). Importantly, our data indicate that 4.1R is required for IQGAP1 localization and efficient actin accumulation to the leading edge of cells migrating into a wound or to the plasma membrane of randomly migrating cells, whereas IQGAP1 is required for neither localization nor accumulation of 4.1R. Because IQGAP1 interacts with F-actin (Mateer et al., 2004), the reduced accumulation of actin in 4.1R-depleted cells might also contribute to the diminished accumulation of IQGAP1. These findings place 4.1R upstream of APC, CLIP-170 and Dia1, all of which require IQGAP1 for their recruitment to the leading edge. The possibility that the effect on IQGAP1 localization observed in 4.1R-depleted cells was due to a decrease in the expression levels of IQGAP1 in 4.1R-siRNA-transfected cells was ruled out. Consistent with the requirement for IQGAP1 distribution at the leading edge for recruitment of APC and Dia1 (Brandt et al., 2007; Watanabe et al., 2004), we observed deficient targeting of APC and Dia1 in 4.1R-knockdown cells. Our results indicate that protein 4.1R controls cell migration by mediating polarized accumulation of actin, formation of both peripheral ruffles and extended lamellipodia, and recruitment of IQGAP1. Because 4.1R acts upstream of IQGAP1, different phenotypes could be expected from cells silenced for 4.1R or for IQGAP1. Indeed, although both 4.1R- and IQGAP1-knockdown cells showed similar defects in directionality and movement capacity in random migration assays, 4.1R-knockdown cells lacked lamellipodia, whereas IQGAP1-knockdown cells did not.

Because IQGAP1 is not required for 4.1R localization, the question remains as to how 4.1R is targeted to leading edges. The FERM domain of 4.1R is able to interact with phosphatidylinositol-4,5-biphosphate [PtdIns(4,5)P₂] (An et al., 2006) and local enrichment of PtdIns(4,5)P₂ is required at the leading edge for lamellipodia formation (Golub and Caroni, 2005; Ling et al., 2006). One possibility might be that local accumulation of PtdIns(4,5)P₂ at the leading edge recruits 4.1R. It has been reported that engagement of CD44 promotes its redistribution to the leading edge and the subsequent activation of Rac1 (Murai et al., 2004). As CD44 is known to interact with protein 4.1R (Nunomura et al., 1997), a second possibility might be that 4.1R is recruited though interactions with transmembrane receptors, such as CD44. These two possibilities are not mutually exclusive as it is known that PtdIns(4,5)P₂ modulates the interaction of 4.1R with specific transmembrane proteins (An et al., 2005; An et al., 2006).

In conclusion, our work establishes a new role for 4.1R in nonerythroid cells as an important component of the cell migration machinery, acting as an anchoring protein that recruits IQGAP1 and establishes a bridge between IQGAP1, the actin cytoskeleton and the plasma membrane at the leading edge (Fig. 7).

ECV304, PC3 and COS-7 cells were grown in DMEM containing 10% foetal bovine serum at 37°C in a 95% air:5% CO₂ atmosphere. siRNA transfection in ECV304 cells was performed using Lipofectamine 2000, according to the manufacturer's instructions (Invitrogen). Cells were processed 72 hours after transfection or at the times indicated.

Anti-4.1R (10b) antibody was an affinity-purified polyclonal antibody generated as described previously (Correas et al., 1986). Rabbit polyclonal anti-4.1R (937) antibody was raised against a fusion protein (GST–Cter) consisting of GST fused to the C-terminal region of protein 4.1R (Perez-Ferreiro et al., 2001). Mouse monoclonal antibody (mAb) M21 to 4.1R was raised against a synthetic peptide (MESVPEPRPSEWDKC) whose sequence is encoded by exon 17. Specificity of antibodies 937 and M21 is shown in supplementary material Fig. S5. The mouse mAbs to IQGAP1, to paxillin and to Dia1 were from BD Transduction Laboratories and the mouse mAb to actin (AC-40) was from Sigma. Anti-c-Myc mouse mAb 9E10 was from the American Type Culture Collection. The rabbit polyclonal antibodies to N-WASP and APC were from Santa Cruz Biotechnology. The rabbit mAb to GAPDH (14C10) was from Cell Signaling Technology. Horseradish-peroxidase-labelled secondary antibodies were from Southern Biotechnology Associates. Secondary antibodies conjugated to Alexa-Fluor-488, -594 or -647 were from Molecular Probes. TRITC-phalloidin and phalloidin-647 were obtained from Sigma.

21-nucleotide siRNA duplexes targeting mRNA sequences to human 4.1R (siRNA1: 5′-GGUGGUCGUCCACCAGGAG-3′; siRNA2: 5′-CCCCUUCAGCCUAGCCUCU-3′) and IQGAP1 (siRNA1: 5′-UGCCAUGGAUGAGAUUGGA-3′; siRNA2: 5′-CAAUAGGGAUGGUAGGAUU-3′) with dTdT overhangs at their 3′ end were purchased from Sigma. Sequences were submitted to BLAST search to ensure targeting specificity. Cells treated with only Lipofectamine 2000 or transfected with a scrambled (control) siRNA were used as controls in all the experiments.

Cells were fixed with 10% formalin (37% formaldehyde solution; Sigma), permeabilized, blocked, stained with the indicated antibodies followed by the appropriate secondary antibodies conjugated with Alexa-Fluor-488 (excitation at 488 nm and emission collected at 505–530 nm), Alexa-Fluor-594 (excitation at 543 nm and emission at 585–615 nm) or Alexa-Fluor-647 (excitation at 633 nm and emission from 650 nm) or with TRITC-phalloidin (excitation at 543 nm and emission at 570–615 nm), and processed as described (De Carcer et al., 1995). Specifically, for triple immunofluorescence stainings, TRITC-phalloidin and secondary antibodies conjugated with Alexa-Fluor-488 and Alexa-Fluor-647 were used; for double immunofluorescence stainings, either secondary antibodies conjugated with Alexa-Fluor-488 and Alexa-Fluor-594 or TRITC-phalloidin and secondary antibody conjugated with Alexa-Fluor-488 were used. In supplementary material Fig. S1, phalloidin-647 (excitation at 650 nm and emission at 668 nm) and a secondary antibody conjugated with Alexa-Fluor-594 were used. Controls to assess labelling specificity included incubations with control primary antibodies or omission of the primary antibodies. Fluorescence was examined using confocal laser-scanning microscope LSM 510 in conjunction with inverted microscope Axiovert 200 M (Zeiss). LSM images were converted to TIFF format and quantified with ImageJ.

60 hours after transfection, control and 4.1R-knockdown ECV304 cells were transferred to a microscope stage heated to 37°C. Image series were acquired using a 10×/0.45 Plan-Apochromat Ph1 objective lens on an inverted Axiovert200 microscope (Zeiss). Images were processed and analyzed with MetaMorph software. To determine cell trajectories, the position of the centroid of the cells in phase-contrast image series was tracked. Cells were imaged at 15-minute intervals for 8 hours. At least 60 cells per experiment were tracked using the track object function of MetaMorph. To determine the closure of the wound, the surface covered by the cells was quantified at 15-minute intervals for 12 hours. Three independent experiments were performed.

Protein samples were subjected to immunoprecipitation and immunoblotting as described previously (Perez-Ferreiro et al., 2006), except that the lysis buffer contained 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, 10 mM β-glycerol-phosphate, 10 mM sodium orthovanadate and 5 mM NaF and was supplemented with a cocktail of protease inhibitors. The extract was centrifuged and the supernatant used as input in the pull-down assays. The expression and purification of GST–4.1R fusions and the procedure for the pull-down assays have been described elsewhere (Perez-Ferreiro et al., 2006). Cytochalasin D was purchased from Sigma and used in cell cultures at 5 μM for 1 hour.

72 hours after transfection, 5×10⁶ control and 4.1R-knockdown ECV304 cells were lysed in 200 μl of buffer containing 5 mM sodium phosphate pH 7.0, 150 mM NaCl and 0.5% Triton X-100, mixed well and incubated on ice for 1 hour. After centrifugation at 20,000 g for 15 minutes at 4°C, the supernatant was separated and the pellet was resuspended in 200 μl of the lysis buffer. Samples were subjected to western blot analysis with anti-IQGAP1, anti-actin and anti-GAPDH antibodies.

Data are expressed as mean ± s.e.m. A paired Student's t-test was used to establish the statistical significance of differences between the means. Data were processed from at least three independent experiments. *P<0.03; **P<0.005; ***P<0.001.

We thank F. Martín-Belmonte, C. M. Pérez-Ferreiro and E. P. Lospitao for critical discussions, and L. Fernández and A. Gosálbez for their technical expertise (CBMSO, Madrid). We also thank M. García-Gallo and M. Llorente for their help in the generation of antibody M21 (Protein Tools Unit; CNB, Madrid). We are very grateful to O. M. Antón for her generous help with the design of the scheme in Fig. 7. The expert technical advice of the personnel of the Optical and Confocal Microscopy Facility is gratefully acknowledged. A.R.-S. is a recipient of a fellowship from the Ministerio de Ciencia e Innovación (MICINN), Spain. This work was supported by grants BFU2008-02460 to I.C., BFU2009-07886 and CONSOLIDER COAT CSD2009-00016 to M.A.A., SAF-2008-01936 to J.M., all from MICINN, and S-GEN-0166/2006 from the Comunidad de Madrid to I.C.