FilGAP, a Rho–ROCK-regulated GAP for Rac, controls adherens junctions in MDCK cells

Cell–cell adhesion is essential for epithelial tissue integrity and function. Adherens junctions are responsible for epithelial cell–cell adhesions and are composed of a transmembrane protein E-cadherin that is connected to actin cytoskeleton through catenins (Green et al., 2010). Several lines of evidence indicate that the actin cytoskeleton and its regulators are crucially involved in the regulation of E-cadherin functions (Brieher and Yap, 2013; Niessen et al., 2011).

Rho family small GTPases (Rho GTPases) regulate many fundamental cellular processes including cell adhesion, migration, vesicle trafficking and differentiation (Heasman and Ridley, 2008; Jaffe and Hall, 2005; Parsons et al., 2010). The GTP-bound active form stimulates downstream effectors, whereas hydrolysis of GTP inactivates Rho GTPases. Therefore, these proteins cycle between an inactive GDP-bound state and an active GTP-bound state. Two classes of proteins mainly regulate this cycle. Guanine-nucleotide-exchange factors (GEFs) activate Rho GTPase by catalyzing the exchange of GDP for GTP, whereas GTPase-activating proteins (GAPs) stimulate the intrinsic GTPase activity and inactivate them (Bos et al., 2007). Because Rho GTPases are involved in the control of actin cytoskeleton, they play an important role in the regulation of E-cadherin-mediated cell–cell adhesion (McCormack et al., 2013).

Cell–cell adhesion is regulated spatiotemporally by Rho GTPases. Rac is activated as an early response to E-cadherin ligation (Ehrlich et al., 2002; Kovacs et al., 2002; Noren et al., 2001). Initial activation of Rac induces inactivation of RhoA through activation of p190RhoGAP (also known as ARHGAP35) and Rac activity is gradually replaced by RhoA, which is mainly responsible for adhesion strength and the maturation of adherens junctions (Guilluy et al., 2011; Noren et al., 2003; Yamada and Nelson, 2007). The requirement of RhoA signaling for cell–cell adhesion is mediated by actin-binding proteins and myosin (Niessen et al., 2011). It has been shown that a number of Rho GEFs and Rho GAPs are involved in the control of spatial regulation of Rho GTPases (Ngok et al., 2014; Niessen et al., 2011). However, it is largely unclear how RhoA and Rac are spatially segregated at cell–cell junctions.

FilGAP (also known as ARHGAP24) is a Rac-specific GAP that suppresses Rac-dependent lamellipodia formation and cell spreading (Nakamura, 2013; Ohta et al., 2006). FilGAP is phosphorylated by ROCK, and this phosphorylation stimulates its RacGAP activity (Ohta et al., 2006). Depletion of endogenous FilGAP induces a Rac-driven elongated mesenchymal morphology. By contrast, overexpression of FilGAP induces membrane blebbing and a rounded amoeboid morphology in a manner that requires Rho–ROCK-dependent phosphorylation of FilGAP (Saito et al., 2012). Thus, FilGAP mediates the antagonism of Rac by Rho that suppresses a mesenchymal morphology and promotes amoeboid migration (Guilluy et al., 2011; Nakamura, 2013; Saito et al., 2012).

Hepatocyte growth factor (HGF) stimulates disruption of cell–cell junctions and enhances cell migration and scattering in epithelial cell lines including MDCK cells (Kodama et al., 2000). The HGF receptor Met is often overexpressed or mutated in cancers (Graveel et al., 2013). However, it is not clear how Met signaling mediates diverse cellular responses such as motility, invasion and growth.

In this study, we demonstrate that FilGAP acts downstream of Rho–ROCK to specifically inactivate Rac at adherens junctions. We found that forced expression of FilGAP suppressed HGF-induced cell scattering. Hence, FilGAP might stabilize cell–cell adhesion through inactivation of Rac in MDCKII cells.

To study the role of FilGAP in the maintenance of adherens junctions, MDCKII cells were transfected with FilGAP siRNA for 48 h. siRNAs targeting FilGAP effectively reduced the expression of endogenous FilGAP in MDCKII cells (Fig. 1B). Depletion of endogenous FilGAP by siRNA alone did not induce cell–cell detachment (Fig. 1A,C). However, knockdown of FilGAP stimulated HGF-induced cell scattering (Fig. 1A,C). When control cells were treated by HGF, they started to detach after 6 h of stimulation, and the cells gradually dispersed. By contrast, FilGAP-depleted cells started to detach after 2 h of stimulation with HGF. Live cell-imaging analysis showed that although control cells migrated at a constant speed after detachment, the migration speed of FilGAP-depleted cells increased during scattering (Fig. 1D; supplementary material Movies 1,2). Similar results were obtained using another FilGAP siRNA (data not shown). Forced expression of FilGAP in MDCKII cells suppressed cell scattering induced by HGF (Fig. 1E). MDCKII cells transfected with FilGAP adhered to each other and did not separate even at 18 h after addition of HGF. Interestingly, the total amount of FilGAP protein expressed in MDCKII cells gradually decreased after the addition of HGF (Fig. 1F,G). The results suggest that FilGAP suppresses cell dissociation and migratory activity induced by HGF.

FilGAP is a Rac-specific GAP and we first examined whether GAP activity of FilGAP is necessary for the suppression of HGF-induced scattering. MDCKII cells were transduced with lentiviral vectors expressing wild-type and a GAP-domain-deficient mutant FilGAP (ΔGAP FilGAP). Although wild-type FilGAP suppressed cell scattering after the stimulation of the cells with HGF for 18 h, ΔGAP FilGAP failed to suppress cell scattering induced by HGF (Fig. 2A,B). Therefore, Rac GAP activity of FilGAP might be necessary for the suppression of cell scattering induced by HGF. In accordance with this result, depletion of endogenous FilGAP by siRNA increased the amount of active Rac1 (GTP-Rac1) both under steady-state conditions and during HGF-induced cell scattering (Fig. 2C,D).

We next studied whether the effect of FilGAP on HGF-induced scattering is caused by the modulation of E-cadherin-dependent cell–cell adhesion. Endogenous FilGAP is localized at cell–cell junctions in MDCKII cells (Fig. 2E). The staining of FilGAP at cell–cell junctions is weak but specific because knockdown of endogenous FilGAP with siRNA eliminated the staining of FilGAP at cell–cell junctions (supplementary material Fig. S1A). FilGAP colocalized with E-cadherin at contacts but not with the tight junction marker ZO-1 (Fig. 2F–H). Therefore, FilGAP might localize at adherens junctions. These results suggest that FilGAP might stabilize adherens junctions through targeting E-cadherin at cell–cell adhesion.

E-cadherin mediates cell–cell adhesion by Ca²⁺-dependent homophilic interactions. Using the Ca²⁺ switch model in MDCKII cells, we examined E-cadherin-mediated cell–cell adhesion. When cultured with a normal Ca²⁺ level, MDCKII cells formed E-cadherin-containing adhesions. By contrast, when cells were treated with EGTA, a Ca²⁺ chelator, these adhesions detached and dispersed (Fig. 2I,J). After restoration of Ca²⁺, cell–cell adhesion was reconstructed in control cells, but FilGAP-depleted cells were unable to form cell–cell adhesions (Fig. 2I,J). These results suggest that FilGAP is required for E-cadherin-mediated cell–cell junction formation.

We next examined whether FilGAP affects the localization of E-cadherin at adherens junctions to modulate cell–cell adhesion. MDCKII cells were transfected with FilGAP siRNA for 2 days. Then the cells were fixed and stained for filamentous actin (F-actin) and E-cadherin. In control MDCKII cells, E-cadherin localized at adherens junctions together with F-actin (Fig. 3A). Depletion of FilGAP by siRNA reduced the accumulation of E-cadherin at adherens junctions (Fig. 3A,C–E). Similar results were obtained using another FilGAP siRNA (data not shown). Conversely, forced expression of FilGAP induced the accumulation of E-cadherin at cell–cell junctions (Fig. 3B–E).

The reduction of E-cadherin accumulation at adherens junctions is more evident in MDCKII cells plated on Transwell membranes, which induce cell polarization (Fig. 3F). Depletion of FilGAP in cells plated on the membrane reduced E-cadherin intensity (Fig. 3G) and increased the cell area, probably owing to the reduced tension between the cells (Fig. 3H). To see whether the reduction in E-cadherin intensity is caused by the loss of E-cadherin expressed on the cell surface, surface-exposed E-cadherin was biotinylated and examined. As shown in Fig. 3I,J, depletion of FilGAP significantly reduced the surface expression of E-cadherin. Moreover, total amount of E-cadherin protein was also reduced by the depletion of FilGAP (Fig. 3K,L). Thus, knockdown of FilGAP might reduce both surface expression and total amount of E-cadherin.

FilGAP is phosphorylated by ROCK, and phosphorylation stimulates its Rac GAP activity (Ohta et al., 2006). Moreover, E-cadherin-dependent cell–cell adhesion is dependent on Rho–ROCK signaling (Smutny et al., 2010). Therefore, we investigated whether FilGAP is located downstream of ROCK to modulate the formation of adherens junctions. Forced expression of full-length Myc-ROCK induced accumulation of E-cadherin at cell–cell junction (Fig. 4A,B). Depletion of endogenous FilGAP by siRNA reduced the accumulation of E-cadherin at adherens junctions (Fig. 4A,B). Similar results were obtained using another FilGAP siRNA (data not shown). We further examined whether ROCK-mediated phosphorylation of FilGAP is required for accumulation of E-cadherin at adherens junctions. As shown in Fig. 4A and supplementary material Fig. S1B,C, depletion of FilGAP was rescued by overexpression of siRNA-resistant wild-type FilGAP but not a non-phosphorylatable FilGAP ST/A mutant. This demonstrates that ROCK stimulates accumulation of E-cadherin at cell–cell junction that might require ROCK-dependent phosphorylation of FilGAP.

Forced expression of wild-type FilGAP induced the accumulation of E-cadherin at adherens junctions. This accumulation was blocked by the treatment of the MDCKII cells with the ROCK-specific inhibitor Y27632 (Fig. 4C). However, overexpression of the phospho-mimic FilGAP ST/D mutant induced the localization of E-cadherin at adherens junctions in the presence of ROCK inhibitor. Thus, FilGAP might play a role in the regulation of localization of E-cadherin at adherens junctions downstream of ROCK.

We further examined whether phosphorylation of FilGAP by ROCK is involved in E-cadherin-dependent adhesion. MDCKII cells were transduced with lentiviral vectors expressing non-phosphorylatable (ST/A) or phospho-mimic (ST/D) HA–FilGAP mutants. We found that phospho-mimic FilGAP ST/D-expressing cells did not scatter after stimulation with HGF. By contrast, non-phosphorylatable FilGAP ST/D expressing cells dispersed even better than control cells after HGF treatment (Fig. 4D,E). Taken together, these data suggest that FilGAP might regulate E-cadherin localization downstream of ROCK.

FilGAP is a Rac-specific GAP and overexpression of FilGAP induced downregulation of Rac in MDCKII cells. Therefore, we expected that depletion of Rac1 would stimulate cell–cell adhesion induced by E-cadherin. MDCKII cells were transfected with Rac1 siRNA, and Rac1 was effectively depleted after 48 h (Fig. 5B). In Rac1-depleted cells, E-cadherin accumulated at cell–cell junctions (Fig. 5A,C). Knockdown of endogenous FilGAP reduced the accumulation of E-cadherin at adherens junctions but depletion of Rac1 rescued the loss of E-cadherin at adherens junctions induced by FilGAP siRNA (Fig. 5A,C). We also found that depletion of Rac1 suppressed HGF-induced cell scattering (Fig. 5D,E; supplementary material Movie 3). Similar results were obtained using another Rac1 siRNA (data not shown). Moreover, depletion of Rac1 blocked HGF-induced cell scattering stimulated by FilGAP siRNA (Fig. 5D,E; supplementary material Movie 4). These results suggest that FilGAP might induce accumulation of E-cadherin and block HGF-induced cell scattering through downregulation of Rac1.

We further examined whether FilGAP stabilizes adherens junctions through Rac1-mediated reorganization of the F-actin network. Forced expression of the constitutively activated Rac1 Q61L mutant induced a ruffle-like junctional F-actin network (Fig. 5F). A similar F-actin network was induced by depletion of FilGAP (Fig. 5F). The ruffle-like immature F-actin network at cell–cell junctions that was induced by knockdown of FilGAP was transformed into stress fibers by overexpression of the dominant-negative inhibitor Rac1 T17N mutant (Fig. 5F). Thus, downregulation of Rac1 by FilGAP might stabilize adherens junctions through re-organization of F-actin network.

In this study, we show that FilGAP mediates Rho–ROCK-dependent inactivation of Rac to stabilize adherens junctions. Rho GTPases are required for the formation and maintenance of adherens junctions. Rac is activated as an early response to E-cadherin-mediated cell–cell adhesion. Rac activity is then replaced by Rho and Rho, which is required for expansion and stabilization of adherens junctions (Noren et al., 2003; Yamada and Nelson, 2007). Therefore, the balance in the activity of Rac and Rho at adherens junctions should be spatiotemporally determined by Rho GTPase regulators (Ngok et al., 2014; Niessen et al., 2011). Our present study suggests that FilGAP is involved in the maintenance of the ‘Rho zone’ at adherens junctions.

The integrity of adherens junctions is regulated by Rho–ROCK signaling (Shewan et al., 2005). Myosin IIA is a major downstream effector of the Rho–ROCK pathway (Smutny et al., 2010). Phosphorylation of myosin IIA by ROCK stabilized actin cytoskeleton at adherens junctions. Our present study suggests that ROCK is not only involved in the activation of myosin II but is also required for the suppression of Rac through activation of FilGAP. First, depletion of endogenous FilGAP blocked the accumulation of E-cadherin at adherens junctions induced by overexpression of ROCK (Fig. 5A). Moreover, the accumulation of E-cadherin at adherens junctions was recovered upon the expression of siRNA-resistant wild-type FilGAP but not by the expression of the non-phosphorylatable FilGAP ST/A mutant. Second, forced expression of the phospho-mimic FilGAP ST/D mutant induced accumulation of E-cadherin at adherens junctions in the presence of the ROCK-specific inhibitor Y27632. Taken together, our study suggests that stabilization of E-cadherin at adherens junctions might require activation of FilGAP downstream of ROCK. A previous study has demonstrated that the centralspindlin complex associates with α-catenin and supports Rho activation through recruitment of the Rho GEF ECT2 and suppression of Rac through modulating its Rac GAP activity (Ratheesh et al., 2012). Once this pathway is activated, FilGAP could function downstream of Rho–ROCK to maintain Rho activation. Although FilGAP is colocalized with E-cadherin at adherens junctions, we could not demonstrate the direct association of FilGAP with E-cadherin and/or α-catenin. Further work is necessary to determine how FilGAP is targeted to adherens junctions.

It is unclear how FilGAP stabilizes E-cadherin at adherens junctions. However, our study suggests that inactivation of Rac by FilGAP might play a role in targeting of E-cadherin at adherens junctions. First, the Rac GAP activity of FilGAP is required for stabilization of cell–cell adhesions (Fig. 2A). Second, ROCK-mediated activation of FilGAP seems to be necessary for E-cadherin accumulation at adherens junctions (Fig. 4). Finally, knockdown of endogenous FilGAP reduced the accumulation of E-cadherin at adherens junctions but depletion of endogenous Rac1 rescued the localization of E-cadherin at adherens junctions in the presence of FilGAP siRNA (Fig. 5). A previous study has shown that depletion of TEM4, a Rho GEF, impairs cell–cell junction formation in MDCK cells and reduces the amount of activated RhoA (Ngok et al., 2013). Thus, the balance in the activity of Rac and Rho as a result of activation of Rho or inactivation of Rac might be necessary for the maintenance of adherens junctions.

It remains to be determined how inactivation of Rac1 induces accumulation of E-cadherin at adherens junctions. Our results showed that the F-actin network at adherens junctions is affected by Rac1 activity. Both overexpression of dominant-negative Rac1 (Rac1 T17N) and knockdown of endogenous Rac1 induced the formation of a dense cortical F-actin network at adherens junctions, which was accompanied by the accumulation of E-cadherin at adherens junctions (Fig. 5). Therefore, inactivation of Rac1 by FilGAP might stabilize the F-actin network and maintains E-cadherin localization at adherens junctions. Rac and Rho might antagonize each other at adherens junctions, and inactivation of Rac1 at adherens junctions might induce activation of Rho, which would in turn stabilize the F-actin network (Ratheesh et al., 2012).

We showed that depletion of FilGAP reduced the total amount as well as cell surface expression of E-cadherin (Fig. 3I–L). The mechanism of regulation is unknown; however, FilGAP could stabilize E-cadherin at the cell surface through regulation of endocytosis of E-cadherin. Previous studies have shown that Rac could inhibit or activate E-cadherin endocytosis (Akhtar and Hotchin, 2001; Izumi et al., 2004; Lozano et al., 2003). A similar mechanism might be responsible for the reduced E-cadherin localization at adherens junctions upon knockdown of FilGAP.

HGF stimulates disruption of cell–cell junctions and enhances cell migration and scattering in a number of epithelial cell lines including MDCK cells. Our study suggests that FilGAP might suppress cell scattering induced by HGF. First, depletion of endogenous FilGAP in MDCKII cells stimulated the detachment of cells after the addition of HGF. Moreover, the speed of cell migration increased during HGF-induced scattering. Second, forced expression of FilGAP suppressed detachment of cells after the addition of HGF. Third, the total amount of FilGAP protein decreased after the addition of HGF. This might facilitate the disassembly of adherens junctions stabilized by FilGAP. ARHGAP12 has been identified as a Rac GAP that is associated at adherens junctions and is also transcriptionally downregulated by HGF (Gentile et al., 2008; Matsuda et al., 2008). Moreover, depletion of ARHGAP12 stimulates HGF scattering (Gentile et al., 2008). Therefore, the biological function of ARHGAP12 seems to be similar to that of FilGAP. However, it is unclear how the activity of ARHGAP12 is regulated. Further study is required to understand the relations between FilGAP and ARHGAP12.

It remains to be determined how FilGAP suppresses cell scattering, but downregulation of Rac seems to be involved not only in the stabilization of adherens junctions but also the suppression of HGF-mediated scattering. We showed that depletion of Rac1 blocked HGF-induced cell scattering (Fig. 5). Knockdown of FilGAP failed to stimulate HGF-induced cell scattering in the absence of endogenous Rac1 (Fig. 5). In agreement with this result, previous studies have shown that activation of Rac1 and its effector PAK1 are involved in HGF-induced cell scattering (Bright et al., 2009; Ridley et al., 1995). Moreover, depletion of Numb, an endocytic adaptor protein, stimulates cell scattering induced by HGF, which is inhibited by the Rac1 inhibitor NSC23766 (Lau and McGlade, 2011). Numb depletion in epithelial cells causes increased Rac1 activation, which correlates with destabilized cell–cell junctions and potentiation of HGF-induced cell dispersal. Our study also showed that depletion of FilGAP induced activation of Rac1 (Fig. 2).

Malignant tumor cells invade through tissue extracellular matrix by using two interconvertible modes of cell motility referred to as mesenchymal and amoeboid migration (Friedl and Alexander, 2011; Sanz-Moreno and Marshall, 2010). Our previous study showed that FilGAP contributes to the regulation of mesenchymal to amoeboid transition (Saito et al., 2012). Forced expression of FilGAP induced membrane blebbing and a rounded amoeboid morphology that requires Rho–ROCK-dependent phosphorylation of FilGAP (Ohta et al., 2006). Recent studies have shown that hyperactivation of Met induced membrane blebbing, which leads to amoeboid cell motility and invasion (Laser-Azogui et al., 2014). Moreover, transformation of MDCK cells with moloney sarcoma virus cells induced membrane blebbing, which is dependent on Rho–ROCK signaling and requires autocrine c-Met activation (Jia et al., 2006). FilGAP could be involved in the HGF and Met signaling in those invasive cancer cells.

We have shown previously that FilGAP functions downstream of Arf6, and that Arf6 stimulates FilGAP-dependent suppression of Rac (Kawaguchi et al., 2014). Arf6 has been shown to suppress Rac by inhibiting Tiam-1 Rac GEF at adherens junctions (Palacios et al., 2002). It is therefore possible Arf6 might utilize FilGAP to modulate the integrity of adherens junctions.

In summary, our present study demonstrates that FilGAP might function to maintain adherens junction integrity through inactivation of Rac downstream of Rho–ROCK signaling. FilGAP might be crucially involved as a Rho GTPase regulator to determine the balance of Rac and Rho activities at adherens junctions.

The HA-tagged FilGAP (wild-type, ST/A, ST/D and ΔGAP) constructs in pCMV5 vector were as described previously (Ohta et al., 2006). EGFP-C1-Rac1(T17N) was generated by PCR. pcDNA3-EGFP-Rac1 Q61L was purchased from Addgene (plasmid ID 12981, Cambridge, MA). CSII-EF-MCS–HA-FilGAP [wild type (WT), ST/A ST/D and ΔGAP] vectors were generated by PCR. The FilGAP siRNA KD#2-resistant construct (KD#2r) was generated by introducing point mutations at nucleotide positions 1980, 1986, 1992 and 1995 of the FilGAP coding sequence using the QuikChange site-directed mutagenesis kit. siRNA oligonucleotide duplexes targeting FilGAP and Rac1 were purchased from Invitrogen (Carlsbad, CA). The targeting sequences were as follows: FilGAP, KD#1 5′-AAGAUAGAGUAUGAGUCCAGGAUAA-3′ (nt 1975–1999) and KD#2 5′-CAGUGAUGAUUAGCAAACAUGAUUG-3′ (nt 956–980); Rac1, KD#1 5′-UUUACCUACAGCUCCGUCUCCCACC-3′ (nt 24–48) and KD#2 5′-AGCAAAUUAAGAACACAUCUGUUUG-3′ (nt 220–244).

Mouse anti-HA (12CA5), and anti-α-tubulin monoclonal antibodies were purchased from Sigma. Rabbit anti-Myc antibody was purchased from MBL (Nagoya, Japan). Mouse anti-E-cadherin antibody was purchased from BD biosciences (Bedford, MA). Mouse anti-Rac1 monoclonal antibody was purchased from Millipore (Billerica, MA). Rabbit anti-FilGAP polyclonal antibody was prepared as described previously (Ohta et al., 2006). Mouse anti-ZO-1 antibody and secondary antibodies conjugated to Alexa Fluor 488 or 568, and Alexa-Fluor-568–phalloidin were purchased from Invitrogen. HGF was from Millipore.

MDCKII cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Sigma-Aldrich, St Louis, MO) supplemented with 10% (v/v) fetal bovine serum (FBS) and 50 U/ml penicillin-streptomycin at 37°C under 5% CO₂. Cells were transfected with plasmid DNA for 24 h or FilGAP siRNA for 48–72 h using Lipofectamine 2000 and Lipofectamine RNAiMax (Invitrogen) according to the manufacturer's instructions. To disrupt E-cadherin-mediated cell–cell contacts, serum-starved MDCKII cells were incubated with medium containing 4 mM EGTA at 37°C for 30 min. Then, the cells were incubated with DMEM containing a normal Ca²⁺ concentration to induce cell–cell contacts for indicated times at 37°C.

For immunofluorescence staining, cells were washed twice with PBS and fixed with 3.7% (v/v) formaldehyde for 10 min. Fixed cells were permeabilized with 0.5% Triton X-100 in PBS for 10 min and stained with Alexa-Fluor-568–phalloidin for F-actin or immunostained with primary antibodies. Cells were then washed and incubated with Alexa-Fluor-dye-labeled secondary antibodies (Invitrogen). After washing with PBS, cells were observed under an Olympus IX81 fluorescence microscope (Olympus, Tokyo, Japan). Images were acquired by a charge-coupled device (CCD) camera (ORCA-ER; Hamamatsu Photonics, Hamamatsu, Japan) or an EMCCD camera (iXon3; Andor, South Windsor, CT) and MetaMorph software (Molecular Devices, Sunnyvale, CA). Confocal images were obtained using Olympus DSU (Disk scanning unit)-IX81with a 100× objective. Time-lapse DIC images were recorded at 12 min intervals for 24 h.

MDCK cells treated with control or FilGAP siRNA were cultured for 48 h. After incubation with or without 20 ng/ml HGF for 4 h, cells were directly lysed in RIPA buffer [20 mM Tris-HCl pH 7.5, 500 mM NaCl, 10 mM MgCl₂, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS and protease inhibitor cocktails (Sigma)] at 4°C. The cell lysates were pre-cleared and samples of supernatants fluids were withdrawn for the determination of total Rac protein, and the remaining supernatants were incubated with 20 μg of GST-PAK-CRIB protein in the presence of glutathione–Sepharose beads (GE Healthcare). The beads were washed and the level of GTP-bound Rac1 was determined by immunoblotting using anti-Rac1 antibody.

Cell surface biotinylation was performed as described previously (Fujita et al., 2002). Briefly, MDCK cells grown on filters (cell culture inserts, Thermo Scientific) were incubated with 0.5 mg/ml sulfo-NHS-SS-biotin (Thermo Scientific) for 1 h at 0°C, which was applied to both apical and basal sides of the filter, followed by washing by blocking reagent (50 mM NH₄Cl, 1 mM MgCl₂, 0.1 mM CaCl₂). Cells were lysed in RIPA buffer and cell extracts were incubated with streptavidin beads (Sigma) to collect bound biotinylated proteins. Samples were then analyzed by SDS-PAGE and immunoblotting.

The statistical significance was accessed by two-tailed unpaired Student's t-test. Differences were considered to be statistically significant at P<0.05. Error bars (s.e.m.) and P values were determined from results of at least three experiments.