Radixin regulates cell migration and cell–cell adhesion through Rac1

The ezrin/radixin/moesin (ERM) family of proteins act as reversible linkers between cell surface receptors and the actin cytoskeleton, and participate in signal transduction pathways regulating cell migration and cell polarity (Fehon et al., 2010). They contribute to the formation of dynamic actin filament-containing membrane structures such as filopodia, uropods, lamellipodia and membrane ruffles, the cleavage furrow in dividing cells, and cell–cell adhesions (Bretscher et al., 2002; Gautreau et al., 2002; Niggli and Rossy, 2008).

Ezrin, radixin and moesin are closely related ∼80-kDa proteins that consist of three domains: an N-terminal FERM (four point one-ezrin–radixin–moesin) domain linked to a C-terminal F-actin-binding domain through an α-helical domain (Fehon et al., 2010; Neisch and Fehon, 2011). The FERM domain interacts either directly with membrane receptors such as CD44, CD43, and intercellular adhesion proteins (ICAMs) or indirectly through scaffold proteins such as EBP50 (ezrin binding protein 50). It also binds to phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P₂] (Bretscher et al., 2002; Gautreau et al., 2000; Gautreau et al., 2002; Matsui et al., 1998; Nakamura et al., 1995; Oshiro et al., 1998). ERM proteins can exist in an inactive ‘closed’ conformation, where the C-terminal domain binds to the N-terminal FERM domain. In this conformation they are found in the cytosolic fraction of cells and the binding sites for plasma membrane proteins and F-actin are masked. ERM proteins are activated through conformational changes induced by binding to PtdIns(4,5)P₂ and subsequent phosphorylation of a conserved C-terminal threonine residue [T567 in ezrin, T564 in radixin and T558 in moesin] (Brown et al., 2003; Gautreau et al., 2000; Nakamura et al., 1995; Roch et al., 2010; Suda et al., 2011). Several kinases, including members of the protein kinase C, ROCK and MRCK families have been reported to regulate ERM function through phosphorylation of this threonine (Baeyens et al., 2010; Belkina et al., 2009; Hebert et al., 2008; Matsui et al., 1998; Nakamura et al., 2000; Ng et al., 2001; Parisiadou et al., 2009; Yamane et al., 2011).

ERM proteins can act both upstream and downstream of Rho GTPases, which are key regulators of cell polarity and migration (Hall, 2009; Hughes and Fehon, 2007). Rho GTPases, cycle between an active GTP-bound and an inactive GDP-bound conformation. They are activated by guanine nucleotide exchange factors (GEFs) and inactivated by GTPase activating proteins (GAPs). Some Rho GTPases are also regulated by binding to GDP dissociation inhibitors (GDIs) or by phosphorylation (Hirao et al., 1996; Ivetic and Ridley, 2004; Mackay et al., 1997). RhoGDI can associate with ERM proteins, thereby reducing RhoGDI interaction with GTPases and consequently promoting their activation (Dovas and Couchman, 2005). Ezrin has been reported to interact with PLEKHG6 (pleckstrin homology domain containing family G with RhoGEF domain member 6), a GEF for RhoG and Rac1, to induce morphological changes in the apical region of epithelial cells (D’Angelo et al., 2007). In vitro studies have also reported a direct interaction between the RhoGEF Dbl and radixin (Prag et al., 2007; Takahashi et al., 1998). In this report it is suggested that this interaction would compete with RhoGDI (Takahashi et al., 1998). On the other hand, RhoA and Rac stimulate ERM activity by increasing the levels of PtdIns(4,5)P₂ and/or by stimulating C-terminal phosphorylation, leading to induction of microvilli enriched in activated ERM proteins (Auvinen et al., 2007; Shaw et al., 1997; Yonemura et al., 2002). ERM proteins also bind to the cytoplasmic domain of the transmembrane receptor CD44 and this interaction is enhanced by RhoA activity (del Pozo et al., 1999; Lee et al., 2004; Tsuda et al., 2004). Finally, Rac1 activation in T-lymphocytes leads to inactivation of ERM proteins via dephosphorylation of the C-terminal threonine (Cernuda-Morollon et al., 2010).

The expression of ERM proteins has been investigated in several cancer types, including prostate cancer. For example, ezrin expression has been correlated with adverse prognosis, showing high levels of expression in both high grade prostate intraepithelial neoplasia and prostate adenocarcinoma (Pang et al., 2004; Valdman et al., 2005). Androgen-induced cell invasion requires hormonal regulation of ezrin phosphorylation in the prostate cancer cell line, LNCaP (Chuan et al., 2006). Recently, a study analysing radixin and moesin expression in normal prostate, high grade prostate intraepithelial neoplasia and prostatic adenocarcinoma tissues showed a slight decrease in radixin levels in adenocarcinomas (Bartholow et al., 2011). On the other hand, radixin has been implicated in promoting breast cancer metastasis in an experimental metastasis model (Valastyan et al., 2009).

Although the functions of ezrin and moesin have been extensively studied in cell migration, a role for radixin has not been described. Here we show that radixin depletion severely impairs migration of prostate cancer PC3 cells. This inhibition is accompanied by a change in cell shape and increased spread area concomitant with reorganization of the actin cytoskeleton, focal contacts and cell–cell contacts. We demonstrate that Rac1, which is activated by radixin depletion, mediates these changes. Our results indicate that radixin levels regulate the transition from a non-migratory epithelial morphology with cell–cell adhesions to a migratory morphology by altering Rac1 activity.

To elucidate the role of radixin in cell migration, we depleted ERM protein expression in prostate carcinoma-derived PC3 cells by transient transfection with siRNA (sequences in supplementary material Table S1). Radixin, ezrin and moesin were each efficiently knocked down by 4 different siRNAs (Fig. 1A) Radixin depletion did not affect the levels of ezrin or moesin (Fig. 1B).

Cell migration was analysed by tracking cells in movies generated by time-lapse microscopy (Fig. 1C; supplementary material Movies 1 and 2). PC3 cells normally grow and migrate as single cells with a mesenchymal morphology, and migrate with a mean speed of 1.34±0.04 µm/minute. In contrast, radixin-depleted PC3 cells had a migration speed of approximately 38% of that of control siRNA-transfected cells, observed with 4 independent siRNAs targeting radixin (Fig. 1D; supplementary material Movies 1 and 2). The inhibition of migration following radixin suppression appeared to be independent of the substrate as similar results were obtained on uncoated plastic and on collagen-I or fibronectin-coated surfaces (data not shown).

Radixin-depleted cells were flatter with a more circular shape than control cells; they also showed persistent membrane ruffling around their whole periphery, rather than the small and localised membrane ruffles in discrete regions observed in control cells (Fig. 1E; supplementary material Movies 3 and 4). In addition, we observed a fourfold increase in the cell spread area of radixin-depleted cells compared to control cells, either quantified from fixed cells stained for F-actin (Fig. 1F) or from analysis of cells in time-lapse movies (data not shown). In contrast, depletion of moesin or ezrin did not affect cell area (Fig. 1F). Although most of the control cells had a migratory phenotype with a defined leading edge, we observed some cells in the whole population with a similar flat shape to radixin-depleted cells. We therefore quantified the number of cells that had a circular, flat shape in both control and radixin knockdown cells. We observed that almost 90% of the radixin knockdown cells were circular and flat but just a 13% of control cells showed a similar phenotype (Table 1).

Since actin cytoskeleton and cell adhesion dynamics are required for cell migration, we investigated whether radixin knockdown was accompanied by changes to the actin cytoskeleton and cell adhesions. We observed that radixin-depleted PC3 cells had more peripheral actin filament bundles than control cells, mainly localised close to the plasma membrane. This actin reorganization was accompanied by a re-localisation of the focal contact marker paxillin (Schaller, 2001; Schaller, 2004) from a uniform distribution around the cell periphery to larger focal contacts that mainly localised at the tips of actin filament bundles (Fig. 2A). Active, C-terminally phosphorylated ERM proteins (p-ERMs) often localise to membrane ruffles and lamellipodia (Niggli and Rossy, 2008). In control PC3 cells, p-ERMs were primarily localised to filopodia and other cell protrusions. In contrast, p-ERMs had a more punctate distribution on the dorsal surface in radixin-depleted cells, similar to the dorsal distribution of F-actin (Fig. 2B), a characteristic of microvilli of epithelial cells (Fehon et al., 2010; Lan et al., 2006). This suggests that radixin-depleted PC3 cells have a more epithelial phenotype. Quantification indicated that more than 80% of the radixin-depleted cells were positive for punctate F-actin and p-ERM in the top 20% of the cell height whereas only 12% of the control cells had similar F-actin/p-ERM distribution at that height (Fig. 2C). Radixin therefore clearly regulates p-ERM distribution, although it did not alter total levels of p-ERM, as determined by western blotting (supplementary material Fig. S1).

Radixin depletion similarly induced cell spreading, increased peripheral actin filament bundles and accumulation of p-ERM on the dorsal surface in MDA-MB-231 breast cancer cells, indicating that the phenotype is not PC3-cell specific (supplementary material Fig. S2).

From time-lapse movies, it was apparent that radixin-depleted cells formed more stable cell–cell contacts than control cells (Fig. 3A; supplementary material Movies 5 and 6). This was quantified by determining the proportion of cells that maintained contact with each other during the duration of each movie: over 90% of radixin-depleted cells maintained contact with their neighbour(s) over 16 h (Fig. 3B). In order to investigate in more detail the basis for this radixin-regulated change in cell–cell adhesion we analysed the distribution of adherens junctions, which are known to be important for establishing and maintaining cell–cell contacts (Millan et al., 2010; Shapiro and Weis, 2009). PC3 cells express N-cadherin but not E-cadherin (supplementary material Fig. S3A), and radixin knockdown did not induce E-cadherin expression (supplementary material Fig. S3B). In control PC3 cells N-cadherin was predominantly distributed throughout the cytoplasm with some localisation at the plasma membrane. After radixin knockdown, N-cadherin accumulated at cell–cell contact regions (Fig. 3C). Similarly, β-catenin, which binds directly to cadherin cytoplasmic tails (Shapiro and Weis, 2009; Sung et al., 2008), was localised to sites of cell–cell contact in radixin-depleted cells, whereas in control cells it was homogenously distributed throughout the cytoplasm (Fig. 3C).

Although cell spreading in radixin-depleted cells would be expected to increase cell–cell contact, very few control PC3 cells that interacted with each other showed N-cadherin and β-catenin accumulation or colocalisation in the regions of contact. Quantification showed that 79% of radixin-depleted cells had cell–cell contacts that were positive for N-cadherin, whereas only 3.5% of the control cells showing a similar flat phenotype had N-cadherin in the cell–cell contact regions (Table 1). The interaction of control cells did not seem to depend on whether they had an elongated or spread morphology, whereas for radixin knockdown cells there was a significant correlation between the presence of N-cadherin in cell–cell contacts and a spread phenotype. This indicates that the spread phenotype observed in some control cells is not likely to be related to that induced by radixin knockdown.

Interestingly, when cells were grown to very high confluence, the radixin-depleted cells showed a significant increase in cell height (Fig. 3D), indicating that cell–cell contacts are likely to be more mature. In these high-density cultures, the cadherin-associated protein p120-catenin (Carnahan et al., 2010) localised to cell–cell contacts of radixin-depleted cells but not those of control cells (supplementary material Movies 7 and 8).

Rho GTPases are known to be important both for cell migration and for formation of adherens junctions in epithelial cells (Parri and Chiarugi, 2010; Popoff and Geny, 2009; van Duijn et al., 2010). Since radixin knockdown inhibited migration and enhanced adherens junction formation, we investigated whether radixin affected Rho GTPase activity. Radixin depletion led to an increase in Rac1 activity, but no change in Cdc42 or RhoA activity was observed (Fig. 4A). Total levels of Rac1, Cdc42 and RhoA were not changed. Consistent with a role for Rac1 in mediating the phenotypic changes induced by radixin suppression, expression of constitutively active Rac1 (GFP–Rac1–L61) but not dominant negative Rac1 (GFP–Rac1–N17) stimulated cell spreading and increased N-cadherin levels at sites of cell–cell contact (Fig. 4B; supplementary material Fig. S4). GFP–Rac1–L61 colocalised with N-cadherin at cell–cell contacts (Fig. 4B).

In order to determine whether Rac1 was required for the phenotype of radixin-depleted cells, cells were sequentially transfected with siRNAs to Rac1 then radixin, or radixin then Rac1 (Fig. 4C). Rac1 suppression both prevented the increase in cell spread area induced by radixin depletion and reverted the radixin knockdown phenotype to that of control cells. The changes induced by radixin depletion are therefore dependent on Rac1.

To determine how radixin affected Rac1 activity, we first tested whether there was an interaction between radixin and regulators of Rac1, including Dbl and RhoGDI, which have previously been reported to interact with radixin in vitro (Takahashi et al., 1998), and Tiam1, which is known to promote adherens junction formation via Rac1 (Buongiorno et al., 2008). We could not detect interaction between endogenous proteins in any case by immunoprecipitation (data not shown). The GEF activity of Vav GEFs is known to be activated by phosphorylation of a conserved tyrosine (Y174) in their acidic regions, adjacent to the PH-DH GEF domains (Lopez-Lago et al., 2000). Interestingly, we observed an increase in Y174 tyrosine phosphorylation of Vav GEFs only in radixin-depleted cells but not ezrin- or moesin-depleted cells (Fig. 5A). In order to determine whether Vav GEFs were required for the phenotype of radixin-depleted cells, cells were transfected with siRNAs to radixin or Vav GEFs alone or together (Fig. 5B). Vav depletion prevented the increase in cell-spread area induced by radixin depletion, although Vav depletion alone did not alter cell morphology.

The increase in Vav phosphorylation upon radixin depletion suggests that Vav GEFs could be responsible for the increase in Rac1 activity. Depletion of Vav GEFs significantly reduced the increase in Rac1 activity induced by radixin depletion (Fig. 6). No change in Rac1 activity was observed upon Vav GEF depletion in the absence of radixin knockdown (Fig. 6). Together, these results indicate that the Vav GEF activation following radixin depletion is required for Rac1-mediated cell spreading.

Compared to ezrin and moesin, little is known of radixin function. Here we show for the first time that radixin regulates cell shape, migration and cell–cell adhesion, and that it acts by altering Rac1 activity. The inhibition of migration in radixin-depleted cells is accompanied by an increase in cell area and cadherin-mediated contacts. Our results indicate that the different effects of ezrin, moesin or radixin depletion are not simply explained by tissue-selective expression of the 3 isoforms. Only radixin depletion but not ezrin or moesin depletion has a profound effect on cell morphology in PC3 and MDA-MB-231 cells, even though ezrin and moesin are both expressed.

Radixin depletion induces changes in PC3 cell shape and adhesion similar to acquisition of a more epithelial morphology, despite the fact that they do not re-express E-cadherin. However, mesenchymal cells (including fibroblasts and astrocytes) frequently form N-cadherin-mediated adherens junctions (Li et al., 2001; Zhu et al., 2010), and thus the junctions formed in radixin-depleted PC3 cells are likely to resemble these. The changes induced by radixin depletion are dependent on Rac1 since Rac1 knockdown inhibits the radixin knockdown-induced increase of cell area, and expression of constitutively active Rac1 induces cell spreading and N-cadherin accumulation at cell–cell contacts. The role of Rac1 in inducing and maintaining adherens junctions is well established (Hage et al., 2009; Ray et al., 2007), and thus increased Rac1 activity could explain the high level of adherens junctions in radixin-depleted cells. Several RacGEFs have been implicated in adherens junction assembly. For example, RacGEF Tiam1 stimulates adherens junction assembly in MDCK epithelial cells (Malliri et al., 2004). Vav GEFs have also been implicated in adherens junction signalling to Rac1 (Fukuyama et al., 2006). Both Tiam1 and Vav have been implicated downstream of phosphoinositide 3-kinases in regulating adherens junctions (Rivard, 2009). The increased Rac1 activity in radixin-depleted PC3 cells might be due to Vav GEFs rather than Tiam1, since tyrosine phosphorylation of Vav GEFs was increased following radixin knockdown and they were required for the increase in cell size and Rac1 activity observed upon radixin depletion.

Both ezrin and moesin are implicated in regulating cell–cell adhesion, but radixin has not previously been linked to this process. For example, in the mouse intestinal epithelium, ezrin deficiency results in fusion of villi, accompanied by the appearance of elongated cell–cell junctions (Saotome et al., 2004). In Caenorhabditis elegans, the single ERM protein erm-1 is required for lumen formation and positioning of adherens junctions (Gobel et al., 2004). In zebrafish, moesin and the endothelium-specific VE-cadherin both regulate endothelial lumen formation during development of blood vessels (Zhu et al., 2010).

Radixin localises to bile canicular membranes in mouse liver, and radixin-deficient mice accumulate high levels of bilirubin in their blood, possibly because they are unable to secrete it into the bile (Kikuchi et al., 2002). However, this phenotype is dependent on the genetic background of the mice (Fukumoto et al., 2007), and the mice grow normally and are fertile. The underlying defect is therefore unclear, although it could relate to a decrease in microvilli in the apical bile canalicular membranes (Kikuchi et al., 2002). Radixin-null mice also become deaf due to progressive degeneration of cochlear stereocilia, which are actin filament-based apical projections on hair cells (Kitajiri et al., 2004). Cadherin-23 mutations lead to deafness in humans (Muller, 2008), and thus it would be interesting to know whether radixin affects cadherin-23 localisation in the cochlea, given that we observe that radixin regulates N-cadherin localisation in PC3 cells.

Increased ezrin expression is a well-known prognostic marker for tumour aggressiveness in a variety of cancers, and influences tumour survival and progression (Hunter, 2004). Moesin expression is also upregulated in some cancers, correlating with progression (Cui et al., 2009). So far little is known about radixin in cancer, although interestingly it is a target of the microRNA miR-31, which inhibits breast cancer metastasis in animal models (Valastyan et al., 2009). Depletion of the miR-31 targets radixin, RhoA and integrin α5 together phenocopies the effects of miR-31 expression in suppressing metastasis (Valastyan et al., 2010). The relative contribution of each of these three genes is not known, however. Our data are consistent with a role for radixin in promoting cancer cell migration, which could contribute to cancer progression.

PC3 and DU145 prostate cancer cells were grown in RPMI containing 25 mM Hepes and 2 mM glutamine supplemented with 10% FCS, 100 µg/ml streptomycin, and 100 units/ml penicillin (5% CO₂, 37°C). MDA-MB-231 breast cancer cells were grown in DMEM containing 25 mM Hepes and 2 mM glutamine supplemented with 10% FCS, 100 µg/ml streptomycin, and 100 units/ml penicillin (5% CO₂, 37°C). Antibodies were obtained from the following sources: radixin from Sigma-Aldrich (cat. no. R3653); ezrin, moesin, phospho-ERM, MLC and phospho-paxillin from Cell Signalling (cat. no. 3145, 3142, 2466, 3672 and 2541, respectively); N-cadherin, E-cadherin, p120-catenin, β-catenin from BD Transduction Laboratories (cat. no. 610921, 610404, 610134 and 610154, respectively); Rac1 clone 23A8 from Millipore (cat. no. 05-389); Cdc42 from Millipore (cat. no. 07-1466); RhoA from Cytoskeleton (cat. no. ARH03-A); Vav and phospho-Vav from Santa Cruz (cat. no. sc-132 and sc-16408-R, respectively); GAPDH from Sigma-Aldrich (cat. no. G-8795); actin filaments were visualized with Alexa-Fluor-546- or Alexa-Fluor-633-labelled phalloidin from Invitrogen (cat. no. A22283 and A22284, respectively). Alexa Fluor 488 anti-rabbit, Alexa Fluor 546 anti-mouse, and Alexa Fluor 633 anti-mouse secondary antibodies were from Invitrogen (cat. no. A11008, A11103 and A21126, respectively). HRP-anti-rabbit and HRP-anti-mouse from GE-Healthcare (cat. no. NA934 and NA931, respectively). Except when specifically stated, all the chemical reagents were obtained from Sigma-Aldrich.

All siRNAs were from Dharmacon (Thermo Scientific); esiRNA (a mixture of siRNAs that all target the same mRNA sequence) for vav1, vav2 and vav3 were from Sigma-Aldrich; their sequences are specified in supplementary material Table S1. Cells were reverse transfected with 50 nM siRNA or esiRNA using Lipofectamine 2000 (Invitrogen) in complete medium without antibiotics. Medium was changed to complete medium 14 hours after transfection and samples were collected. For experiments involving radixin knockdown, radixin 2 siRNA (as specified in supplementary material Table S1) was used unless otherwise indicated.

For DNA transfection cells were transfected using indicated plasmids. 1 µg of DNA in 100 µl of serum-free medium and 3 µl of Lipofectamine 2000 (Invitrogen) in 100 µl of serum-free medium were mixed within 5 minutes and the mixture added into cells in medium without antibiotics for 8 hours before replacement with normal growth medium.

Cell lysates for western blot analysis were prepared at the specified times after treatment using lysis buffer (50 mM Tris-HCl pH 8, 0.5 mM EDTA, 150 mM NaCl, 1% Triton X-100 plus protease and phosphatase inhibitors cocktails). Lysates were mixed with 2× Final Sample Buffer (FSB; Invitrogen) and resolved on 4–12% polyacrylamide gels (Invitrogen), transferred to nitrocellulose membranes and blotted with the indicated antibodies.

PC3 cells grown on 13-mm diameter coverslips were fixed with 4% paraformaldehyde for 20 min. Cells were permeabilised with 0.1% Triton X-100, blocked with blocking buffer (1% BSA, 2% FCS in PBS) and incubated with the indicated primary antibodies diluted in blocking buffer for 1 hour. After three washes with PBS, cells were incubated with fluorochrome-conjugated secondary antibody for 45 minutes. After five washes with PBS, coverslips were mounted in slides using fluorescent mounting medium (Dako).

Images of PC3 cells were acquired at 37°C and 5% CO₂ using a fully motorized, multi-field time-lapse microscope (Eclipse TE 2000-E; Nikon) with a charge-coupled device camera (ORCA; Hamamatsu Photonics). Images were acquired at different intervals using a 20× or 40× objective. Images were acquired using MetaMorph software (MDS Analytical Technologies), and cells were tracked using ImageJ software (National Institute of Health). Time-lapse videos were acquired for 16 to 24 hours with images taken at 10-min intervals. Analysis of cell speed was performed using the Chemotaxis plug in from Integrated BioDiagnostics GmbH (IBIDI).

Images were acquired at room temperature using a confocal microscope (LSM 510 META; Carl Zeiss, Inc.) with three single photomultiplier tube detectors mounted on a inverted microscope (AxioObserver Z1; Carl Zeiss, Inc.) using a 40× 1.3 NA oil immersion objective (Carl Zeiss, Inc.). Images were acquired using Zen software (Carl Zeiss, Inc.) and processed using Zen and Photoshop (Adobe) software. Image analysis for cell height and morphology was performed using Zen and Image J software.

Cells treated with siRNA against radixin or with siRNA control were scrapped and processed for the Rac/Cdc42 activation assay or the RhoA activation assay according to the manufacturer's instructions (Upstate/Millipore and Cytoskeleton Inc, respectively) using the following lysis buffer: 50 mM Tris-HCl, pH 7.5; 1mM EDTA; 500 mM NaCl, 10 mM MgCl₂, 1% (v/v) Triton X-100; 0.5% sodium deoxycholate; 0.1% SDS, 10% (v/v) glycerol; 0.5% (v/v) 2-mercaptoethanol; protease inhibitor cocktail complete (from Roche) and phosphatase inhibitor cocktail II and IV (from Calbiochem). Total input and bound samples were subjected to SDS-PAGE and immunoblot analysis to determine the total and GTP bound Rac1, Cdc42 or RhoA levels.

We are grateful to the members of the Ridley laboratory for their insights during the development of the manuscript and to the Microscopy Facility at St George's University of London for their help during the acquisition of some of the images included in the manuscript.