ARHGAP22 Localizes at Endosomes and Regulates Actin Cytoskeleton

Rho small GTPases control cell morphology and motility through the rearrangement of actin cytoskeleton. We have previously shown that FilGAP, a Rac-specific GAP, binds to the actin-cross-linking protein Filamin A (FLNa) and suppresses Rac-dependent lamellae formation and cell spreading. ARHGAP22 is a member of FilGAP family, and implicated in the regulation of tumor cell motility. However, little is known concerning the cellular localization and mechanism of regulation at the molecular level. Whereas FilGAP binds to FLNa and localizes to lamellae, we found that ARHGAP22 did not bind to FLNa. Forced expression of ARHGAP22 induced enlarged vesicular structures containing the endocytic markers EEA1, Rab5, and Rab11. Moreover, endogenous ARHGAP22 is co-localized with EEA1- and Rab11-positive endosomes but not with trans-Golgi marker TNG46. When constitutively activated Rac Q61L mutant was expressed, ARHGAP22 is co-localized with Rac Q61L at membrane ruffles, suggesting that ARHGAP22 is translocated from endosomes to membrane ruffles to inactivate Rac. Forced expression of ARHGAP22 suppressed lamellae formation and cell spreading. Conversely, knockdown of endogenous ARHGAP22 stimulated cell spreading. Thus, our findings suggest that ARHGAP22 controls cell morphology by inactivating Rac but its localization is not mediated by its interaction with FLNa.


Introduction
Rho family small GTPases (Rho GTPases) regulate many fundamental cellular processes including cell adhesion, migration, vesicle trafficking, and differentiation [1,2,3]. Because Rho GTPases are involved in the control of actin cytoskeleton and cell migration, they are playing an important role in development, immune response, and cancer metastasis [4,5].
RhoA and Rac1 are well-studied members of Rho GTPases [6]. Rac1 stimulates actin polymerization by activating downstream effectors such as PAK protein kinases and WAVE adaptor proteins. Activation of Rac1 induces formation of lamellae in front of migrating cells and cell spreading on extracellular matrixes (ECMs). RhoA is involved in the generation of contractile force through phosphorylation and activation of myosin II. Thus, RhoA stimulates contraction at the rear of migrating cells and formation of focal adhesion. It is well established that RhoA and Rac1 antagonize each other and define the front-back of moving cells [7].
Rho GTPase functions as a molecular switch in cells. While GTP-bound active form stimulates downstream effectors, hydrolysis of GTP inactivates Rho GTPase. Therefore, they cycle between inactive GDP-bound state and 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. While GTPaseactivating proteins (GAPs) stimulate the intrinsic GTPase activity and inactivate them [8].
ARHGAP22 (also called RhoGAP2 and RhoGAP22) belongs to a family of RhoGAPs that includes FilGAP (ARHGAP24) and ARHGAP25 [9,10]. The domain structure of ARHGAP22 is similar to that of FilGAP. It contains pleckstrin-homology (PH) domain at its N-terminus, followed by GAP and coiled-coil (CC) domains. Recently, ARHGAP22 has been identified as a key mediator that suppresses Rac1 downstream of RhoA and involved in the amoeboid movement of melanoma cells in 3D environment [5,11,12,13]. Moreover, ARHGAP22 is phosphorylated downstream of Akt and the phosphorylation promotes binding to 14-3-3 protein [14,15]. We have shown that FilGAP binds to a widely expressed filamentous actin (F-actin) cross-linking protein Filamin A (FLNa) and FLNa binding targets FilGAP to the leading edge of the cell where it antagonizes Rac [9,10]. In FilGAP, an FLNabinding site resides to C-terminal to the CC domain [16]. Although ARHGAP22 contains FLNa-binding consensus sequence at its C-terminus [10], it is unclear whether ARHAGAP22 binds to FLNa. Moreover, localization of ARHGAP22 in mammalian cells is unknown.
In this study, we have studied the cellular distribution and function of ARHGAP22. We found that ARHGAP22 does not interact with FLNa. Moreover, we present the evidence that ARHGAP22 localizes at endosomes and is involved in downregulation of Rac.

ARHGAP22 suppresses lamellae formation
Previous study has shown by using RNA interference that ARHGAP22 is involved in regulating the switch between mesenchymal and amoeboid modes of cell migration in 3D environment [11]. Depletion of endogenous ARHGAP22 by RNAi increased GTP-bound Rac and increased the number of mesenchymal melanoma cells [11]. However, it is unclear where ARHGAP22 localizes in cells and how ARHGAP22 regulates actin cytoskeleton. Many growth factors such as EGF induce lamellae through activation of Rac [9]. Therefore, we investigated if ARHGAP22 could function as a RacGAP and suppress lamellae formation induced by EGF.
A7 melanoma cells transfected with ARHGAP22 were stimulated with EGF (50 nM) for 30 min and lamellae formation was analyzed by F-actin staining. More than 90% of control cells produced lamellae ( Figure 1A and C), but less than ,30% of cells produced lamellae when cells were transfected with ARHGAP22 ( Figure 1B and C). On the other hand, neither ARHGAP22 lacking its GAP domain (amino acids 163-365; DGAP) nor GAP deficient ARHGAP22 (R211A) mutants [14] did not suppress EGF-induced lamellae formation ( Figure 1B and C). The GAPdeficient mutants were still capable of inhibiting lamellae formation. The reason is unclear but overexpression of ARH-GAP22 could inhibit lamellae formation through RacGAPindependent fashion. We found that ARHGAP22 is mainly localized at punctate structures inside the cells and does not appear to co-localize with actin filaments ( Figure 1B and see below). Nonetheless, ARHGAP22 suppresses lamellae formation induced by EGF and its GAP activity is mainly responsible for lamellae suppression.
To confirm if ARHGAP22 functions as a RacGAP, a GSTfusion protein of ARHGAP22 encompassing amino acids 168-365 was prepared and its effect on the intrinsic GTPase-activity of Cdc42, Rac1, and RhoA was determined. The GAP domain of ARHGAP22 stimulated GTPase activity of both Cdc42 and Rac1 but not of RhoA ( Figure S1A). Moreover, forced expression of ARHGAP22 effectively reduced the level of GTP-Rac1 but not GTP-Cdc42 and GTP-RhoA as determined by pull-down assay ( Figure S1B). These results demonstrate that ARHGAP22 inactivates Rac1 and suppresses lamellae formation.

ARHGAP22 does not interact with FLNa
As shown in Figure 2A, FilGAP, a close relative of ARHGAP22, localizes at lamellae with FLNa in A7 cells stimulated by EGF. On the other hand, ARHGAP22 localizes at punctate structures at the cytoplasm and does not co-localize with FLNa at lamellae (Figure 2A). We therefore determined if ARHGAP22 binds to FLNa. HEK cells were transfected with HA-tagged full length ARHGAP22 or HA-FilGAP and the proteins were immunoprecipitated from the cell lysates. Although FLNa was co-precipitated with HA-FilGAP, ARHGAP22 failed to precipitate FLNa ( Figure 2B). We showed that the C-terminal Repeat 23-24 of FLNa mediates a stable complex with FilGAP [9]. A recombinant GST-Repeats 23-24 construct exhibited strong FilGAP binding activity but the repeats did not bind to ARHGAP22 ( Figure 2C).
The C-terminal region of ARHGAP22 contains CC domain similar to FilGAP and has FLNa-binding motif [10,16]. However, ARHGAP22 does not bind to FLNa in vitro and in vivo ( Figure 2B and C). The high-affinity binding of FilGAP to FLNa is dependent on the dimerization of FilGAP [16]. Therefore, we investigated if ARHGAP22 can dimerize in vivo. To investigate whether ARHGAP22 can dimerize in cells similar to FilGAP, we transfected HEK cells with HA-ARHGAP22 and FLAG-ARH-GAP22. HA-ARHGAP22 was immunoprecipitated from cell extracts using anti-HA agarose. The immunoprecipitate was separated by SDS-PAGE and immunoblotted for the presence of FLAG-and HA-ARHGAP22 ( Figure 2D). Although HA-ARH-GAP22 was detectable in the immunoprecipitate as expected, little HA-ARHGAP22 or HA-FilGAP was immunoprecipitated from cell extracts using anti-HA agarose, and bound proteins were identified by immunoblot using anti-HA and anti-FLNa antibodies. (C) HEK cells were transfected with HA-ARHGAP22 or HA-FilGAP. Cell extracts were prepared and then incubated with GST-FLNa-Repeat 23-24 or GST alone, and precipitated with glutathione-Sepharose beads. Bound proteins were analyzed by immunoblot using anti-HA antibody. Asterisks indicate non-specific bands. (D) HEK cells were transfected with HA-FilGAP (or HA-ARHGAP22) in the presence or absence of FLAG-FilGAP (or FLAG-ARHGAP22). Then, HA-FilGAP (or HA-ARHGAP22) was immunoprecipitated from cell extracts using anti-HA agarose, and bound proteins were identified by immunoblot using anti-HA and anti-FLAG antibodies. Asterisk indicates a nonspecific band. (E) HEK cells were transfected with HA-ARHGAP22 or HA-FilGAP. The cells were lysed after treatment with indicated concentrations of DSP, boiled with 1% SDS in the absence or presence of 2-mercaptoethanol (2ME) and analyzed by immunoblot using anti-ARHGAP22 or anti-FilGAP antibody. doi:10.1371/journal.pone.0100271.g002 FLAG-ARHGAP22 was included in the HA-ARHGAP22 immunoprecipitate. On the other hand, FLAG-FilGAP was readily detectable in the HA-FilGAP immunoprecipitate, which was precipitated from co-transfected HEK cells. These results suggest that FilGAP can oligomerize in vivo, but ARHGAP22 does not.
We further analyzed if ARHGAP22 could dimerize in vivo by using chemical cross-linker DSP (Dithiobis [Succinimidyl propionate]) ( Figure 2E). HEK cells transfected with HA-FilGAP were treated with DSP and cell lysates were subjected to SDS-PAGE followed by Western blot for FilGAP. Molecular mass of ,240 kDa was detected in the presence of DSP. This suggests that FilGAP is present as a multimer in cells. On the other hand, when HA-ARHGAP22 was transfected in HEK cells and treated with DSP, only single band ,80 kDa was detected and formation of multimer was not detected. These results strongly suggest that ARHGAP22 may present as a monomer in vivo.

Identification of domains mediating subcellular localization of ARHGAP22
We delineated respective domains of ARHGAP22 mediating its punctate subcellular localization. We have generated HA-tagged ARHGAP22 constructs ( Figure 3A and B), and expressed them in A7 cells. We found that the C-terminal CC domain of ARHGAP22 alone localizes at subcellular punctate structures ( Figure 3C). On the other hand, the N-terminal PH domain or GAP domain localizes diffusely in the cell. Moreover, all the ARHGAP22 constructs lacking CC domain failed to localize at punctate structures ( Figure 3C). These findings indicate that the CC domain of ARHGAP22 is essential for targeting of ARHGAP22 to punctate structures. Interestingly, we found that ARHGAP22 mutant lacking CC domain (DCC) is predominantly localized in nucleus ( Figure 3C).

ARHGAP22 localizes at endosomes
To investigate if the punctate structures correspond to any particular organelles in cells, we compared the localization of ARHGAP22 in cells with various organelle markers including: early endosome (anti-EEA1 and anti-Rab5 antibodies), recycling endosome (anti-Rab11 antibody; RE), lysosome (anti-LAMP-1 antibody), Golgi apparatus (anti-GM130 antibody), and trans-Golgi network (anti-TGN46 antibody). We found that punctate structures induced by HA-ARHGAP22 contain endocytic markers EEA1, Rab11, and Rab5 in A7 cells ( Figure 4A). Forced expression of HA-ARHGAP22 in mouse myoblast C2C12 cells induced enlarged vesicular structures that also contained Rab11 and Rab5 ( Figure 4B). We found that the CC domain of ARHGAP22 alone is sufficient to localize at Rab11-positive structures ( Figure 5A). The ARHGAP22 mutant lacking PH domain (DPH) showed perinuclear localization ( Figure 3C). However, the perinuclear structures still contain Rab11 but do not contain Golgi apparatus marker GM130 ( Figure 5B). Therefore, the CC domain of ARHGAP22 may mediate targeting to vesicle structures that contain endosome markers.
To determine the subcellular localization of endogenous ARHGAP22 in mammalian cells, we have prepared rabbit polyclonal antibody directed against amino acid residues 469-485 (RGHRRASSGDRLKDSGS) of human ARHGAP22. The antibody recognized ARHGAP22 but not other family members FilGAP (ARHGAP24) and ARHGAP25 ( Figure S2A). The anti-ARHGAP22 antibody also recognized HA-ARHGAP22 protein, which was overexpressed in A7 cells ( Figure S2B). By using this antibody, we have analyzed endogenous ARHGAP22 expression in various cell types. Among cell lines tested, we found that mouse C2C12 myoblasts express endogenous ARHGAP22 protein [14].
We therefore determined localization of endogenous ARHGAP22 in C2C12 cells. We found that endogenous ARHGAP22 is localized at punctate structures, which are partly overlapped with Rab11-positive endosomes in C2C12 cells ( Figure 6A). Moreover, co-localization of endogenous ARHGAP22 and Rab11 was diminished when the primary antibody was pre-absorbed with the cell lysates expressing HA-ARHGAP22 ( Figure 6A). The punctate staining also disappeared after depletion of endogenous ARHGAP22 by siRNA treatment ( Figure 6B). The endogenous ARHGAP22 is also partly co-localized with EEA1-positive endosomes but not with trans-Golgi marker TNG46 ( Figure 6C and D). Thus, at least in part, endogenous ARHGAP22 seems to localize at Rab11-and EEA1-positive endosomes in C2C12 cells.

ARHGAP22 suppresses integrin-mediated cell spreading
A7 cells that were plated on collagen-coated dishes adhered within 20 min and then spread circumferentially. Transfection of full-length ARHGAP22 abolished spreading, but ARHGAP22D-GAP and R211A mutants enhanced initial cell spreading ( Figure 7A and B), which is consistent with the finding that activation of Rac and Cdc42 by integrin mediates cell spreading [17].
The CC domain of ARHGAP22 mediates targeting of ARHGAP22 to punctate structures ( Figure 3 and 5A). Therefore, we have examined if targeting of ARHGAP22 to punctate structures has any role in the control of cell spreading. Forced expression of mutant ARHGAP22 lacking CC domain (DCC) failed to suppress cell spreading on collagen. Thus, localization of ARHGAP22 at punctate structures is critical for suppression of cell spreading.
To explore the role of endogenous ARHGAP22 in cell spreading, we have transfected C2C12 mouse myoblast cells with small interference RNAs (siRNAs) targeting ARHGAP22 and spreading on fibronectin was analyzed by F-actin staining. Twoindependent siRNAs targeting ARHGAP22 (KD#1 and KD#3) reduced the expression of endogenous ARHGAP22 in C2C12 cells ( Figure 8A), and depletion of ARHGAP22 by these siRNAs promoted much more rapid spreading ( Figure 8B and C). The spread area that was occupied by ARHGAP22 RNAi-silenced cells is much bigger than that of control cells 10 min after spreading ( Figure 8D).
We introduced 5 silent mutations into the siRNA-targeting sequence of ARHGAP22 (KD r ) and examined if the spreading on fibronectin that was induced by ARHGAP22 siRNA could be prevented by KD r . After two days treatment with ARHGAP22 siRNA, control HA-ARHGAP22 protein was significantly depleted, whereas KD r protein was abundant ( Figure 8E). C2C12 cells expressing KD r did not spread on fibronectin in the presence of ARHGAP22 siRNA ( Figure 8F).

ARHGAP22 co-localizes with constitutively activated Rac at the plasma membrane
To determine if ARHGAP22 could function as a GAP for Rac in cells, we co-expressed ARHGAP22 and constitutively activated mutant Rac (Q61L) in A7 cells. When constitutively activated Rac Q61L mutant was expressed, ARHGAP22 concentrated in sites of membrane ruffles and co-localized with Rac Q61L mutant ( Figure 9). Thus, ARHGAP22 could bind to and inactivate Rac at the cell surface although it localizes to the punctate structures in the absence of activated Rac (Figure 4). Targeting of ARHGAP22 to activated Rac at the plasma membrane requires its GAP domain. The GAP deficient ARHGAP22 R211A mutant colocalizes with constitutively activated Rac at the plasma membrane whereas ARHGAP22 mutant lacking its GAP domain (DGAP) failed to translocate to the plasma membrane and colocalize with activated Rac Q61L. Thus, GAP domain seems to be a predominant site for interaction with Rac. Forced expression of another constitutively activated Rac G12V mutant induced membrane ruffling and ARHGAP22 was translocated to the ruffles ( Figure S1C). On the other hand, translocation of ARHGAP22 to the plasma membrane did not occur when activated mutants of Cdc42 G12V or RhoA G14V were transfected with HA-ARHGAP22 ( Figure S1C).

ARHGAP22 did not affect transferrin receptor-mediated endocytosis
Forced expression of ARHGAP22 induced enlarged vesicles that contain endocytic markers Rab11 and Rab5. Therefore, we examined if ARHGAP22 affects receptor-mediated endocytosis using transferrin. After incubation of A7 cells with Alexa Fluor 568-transferrin for 30 min, Alexa Fluor 568-transferrin was endocytosed and partially co-localized with recycling endosome marker Rab11 ( Figure 10A). Forced expression of ARHGAP22 induced enlarged vesicles but did not affect incorporation of Alexa Fluor 568-transferrin ( Figure 10A and B). The internalized transferrin did not accumulate at the enlarged vesicles ( Figure 10A).

Discussion
In this study, we showed that ARHGAP22 localizes at punctate vesicular structures that contain endocytic markers, and regulates actin cytoskeleton to suppress lamella formation and cell spreading on ECMs. Although ARHGAP22 has similar domain structure to that of FilGAP, FilGAP localizes along actin filaments and concentrates at lamellae whereas ARHGAP22 primarily localizes at punctate structures that contain endosome markers [9,18,19]. Localization of FilGAP at lamellae is mediated by its binding to F-actin crosslinking protein FLNa [9]. FilGAP binds to FLNa through its CC domain. Localization of ARHGAP22 at punctate structures is also dependent on its CC domain. However, we could not detect the binding of ARHGAP22 to FLNa. ARHGAP22 contains consensus sequence for FLNa binding at its CC domain [10]. It has been shown that high affinity binding of FilGAP to FLNa is dependent on dimerization of FilGAP [16]. FLNa is also present as dimer in cells and binding affinity of FilGAP/FLNa interaction is high due to their dimerization [16]. Our study showed that ARHGAP22 seems to present as a monomer in cells. Thus, dimerization of ARHGAP22 may be required for its interaction with FLNa.
It is unclear how ARHGAP22 is targeted to vesicular structures. ARHGAP22 has been shown to interact with several factors. For example, short isoform of ARHGAP22 lacking PH domain (p68RacGAP) binds to transcriptional factor Vezf1 and association induces translocation into nucleus [20]. In neuron, ARHGAP22 (RhoGAP2) is localized to excitatory synapses, and its C-terminal tail interacts with the TIR domain of IL1RAPL1 [21]. Both Vezf1 and IL1RAPL1 are reported to interact with the C-terminus of ARHGAP22. However, they do not seem to be responsible for targeting ARHGAP22 into punctate structures. ARHGAP22 also interacts with b isoform of 14-3-3 [14]. It has been shown that 14-3-3f ortholog (PAR-5) plays a role in Rab11-positive recycling endosome positioning and apicobasal cell polarity [22]. However,  PH domain of ARHGAP22 has been shown to interact with 14-3-3 [15] and our study demonstrates that CC domain of ARHGAP22 is responsible and sufficient for targeting to Rab11positive vesicular structures. It is likely, therefore, that the binding of CC domain of ARHGAP22 to punctate structures is mediated by as yet unidentified factors.
Although wild-type ARHGAP22 is mainly localized at vesicular structures, mutant ARHGAP22 lacking CC domain is exclusively localized in nucleus. ARHGAP22 isoform lacking PH domain (p68RacGAP) is also localized in the nucleus when co-expressed with transcription factor Vezf1 [20]. We showed in this study that ARHGAP22 mutant lacking PH domain is localized at perinuclear structures but not in the nucleus. It remains to be determined if ARHGAP22 could function in the nucleus.
Recent studies have shown that membrane trafficking of Rho GTPases plays critical roles in the regulation of cell migration [27]. Rac is endocytosed by Rab5 and activated by RacGEF Tiam1 at the early endosome and recycled back to the plasma membrane through Arf6-dependent pathway [28]. Cdc42 is also activated by aPIX at the early endosome and transported to leading edge by Arf6 to regulate cell polarization [29]. It is therefore possible that ARHGAP22 could function as a RacGAP at endosomes. We showed that forced expression of ARHGAP22 induced colocalization with activated Rac at the plasma membrane. However, little activated Rac was co-localized with ARHGAP22 at endosomes. It is likely that ARHGAP22 may be reserved at endosomes and translocated to the plasma membrane to downregulate activated Rac.
Although forced expression of ARHGAP22 induced enlarged vesicles in A7 cells, ARHGAP22 did not change the distribution of endocytosed transferrin. The internalized transferrin did not accumulate at the enlarged vesicles induced by ARHGAP22. This is in contrast to Rab5 Q79L-and Rab22-induced vesicles that were accessible to internalized transferrin [23,26]. Therefore, ARHGAP22 does not seem to have dramatic effects on membrane trafficking. However, regulation of endocytosis and trafficking is receptor-dependent and also relies on experimental conditions. Some studies have reported unaffected recycling of transferrin receptor by Rab5 Q79L [24], while others reported defects in recycling [25]. Rab5 Q79L has been shown to affect EGF receptor recycling [23,30]. Rab22 has been shown to induce enlargement of early endocytic compartments and recycling of transferrin from Rab22-positive enlarged compartments was blocked [26]. It is well documented that Rab-regulated endocytosis and trafficking of  [31,32]. Further study is required to determine if ARHGAP22 could regulate endocytic pathways of specific membrane receptors and adhesion molecules that are involved in cell migration. Our present study showed that ARHGAP22 suppresses lamellae formation and cell spreading on ECMs such as fibronectin and collagen. Therefore, ARHGAP22 could regulate actin cytoskele-ton in a similar way conducted by FilGAP. Moreover, both FilGAP and ARHGAP22 have been shown to mediate the antagonism between RhoA and Rac1 and AMT (amoeboid-to- The surface area of spreading cells (n = 100) 10 min after plating was calculated and shown as box and whisker plots. **, p,0.01. Statistical significance was determined by Welch's t-test. (E) HEK cells were treated with control or ARHGAP22 siRNA for 24 h followed by transfection with HA-tagged ARHGAP22 constructs. The cells were cultured for another 24 h. ARHGAP22 and tubulin were analyzed by immunoblot using anti-HA and anti-tubulin antibodies, respectively. (F) C2C12 cells were treated with control or ARHGAP22 siRNA KD#1 for 24 h followed by a transfection with rescue constructs (KD r ). The cells were cultured for another 24 h and serum-starved. The cells were fixed and stained with anti-HA antibody for HA-KD r (green) and phalloidin (red). Merged fluorescent images are shown. The cells were also stained with hoechst 33258 for nuclei (blue). Scale bar, 20 mm. doi:10.1371/journal.pone.0100271.g008  mesenchymal transition) in 3D environment [11,13,33]. We previously showed that FilGAP specifically inactivates Rac through Rho/ROCK-mediated phosphorylation of FilGAP [9,33]. It has been shown that depletion of ARHGAP22 inhibited Rho-mediated inactivation of Rac1 and activation of ARHGAP22 is dependent on ROCK/myosin activity [11]. However, it is unclear how ROCK/myosin activity regulates ARHGAP22 at the molecular level. Moreover, our present study demonstrated that cellular localization of ARHGAP22 is completely different from that of FilGAP. ARHGAP22 localizes at endosomes whereas FilGAP is targeted to lamellae. Further study is required to understand how ARHGAP22 is regulated to inactivate Rac downstream of RhoA.

Materials and Methods
Plasmids cDNAs encoding full length FilGAP (NM_001025616) and ARHGAP22 (BC126444) were described previously [9,13]. cDNAs encoding ARHGAP22 (wild-type, PH, GAP, CC, DPH, DGAP, DCC, R211A) constructs were inserted into pCMV5-HA or pCMV5-FLAG vector. They were generated as follows; ARH-GAP22 was digested with EcoRI and SphI to produce PH domain. The GAP domain of ARHGAP22 was generated by PCR. ARHGAP22 was digested with EcoRI and SalI to produce CC domain. ARHGAP22 cDNA lacking PH domain was generated by PCR. ARHGAP22 cDNA lacking GAP domain was generated by digesting full-length ARHGAP22 with XbaI and BamHI. ARH-GAP22 lacking CC domain was generated by digesting with SmaI and self-ligation. Mutation of R211A of ARHGAP22 construct was generated by introducing point mutations at nucleotide positions 631 and 632 of ARHGAP22 coding sequence using QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). pcDNA3-EGFP-Rac1 Q61L was purchased from Addgene (plasmid ID 12981, Cambridge, MA).

Immunostaining
Cells were cultured on coverslips (poly-L-lysine coated) transfected with relevant plasmids. After 24 h, cells were washed once with PBS and fixed with 3.7% formaldehyde in PBS for 10 min. The fixed cells were then permeabilized with 0.5% Triton X-100 in PBS for 10 min, then incubated with blocking buffer (10% blocking one [Nakarai Tesque, Kyoto, Japan] in PBS) for 30 min, and immunostained with primary antibodies in blocking buffer for 1 h. Cells were then washed and incubated with Alexa Fluor dyelabeled secondary antibodies (Invitrogen) in blocking buffer for 1 h. For visualization of F-actin and nuclei, cells were stained with Alexa Fluor 568 conjugated-phalloidin and Hoechst 33258, respectively, in PBS for 1 h. After wash with PBS, cells were observed under an Olympus IX81 fluorescence microscope with a 10x, 20x, or 40x objective (Olympus, Tokyo, Japan). Images were acquired by a charge-coupled device camera (ORCA-ER; Hamamatsu photonics, hamamatsu, Japan) and analyzed by MetaMorph software (Molecular Devices, Sunnyvale, CA). Quantification of Pearson's Colocalization Coefficient (PCC) was calculated by Colocalization finder for ImageJ (NIH).

EGF stimulation
A7 cells were cultured on coverslips (poly-L-lysine coated) transfected with relevant plasmids for 5 h and serum-starved. The cells were fixed after the treatment with 50 nM EGF for 30 min.

Spreading assay
A7 cells were cultured on 6 cm dish transfected with relevant plasmids for 5 h and serum-starved. After 20 h, the cells were trypsinized and suspended in MEM. The cells were plated on collagen-coated coverslips (50 mg/cm 2 ) and fixed 20 min after plating. C2C12 cells were cultured on 6 cm dish treated with ARHGAP22 siRNAs for 48 h and serum-starved. The cells were trypsinized and suspended in DMEM. The cells were plated on fibronectin-coated coverslips (10 mg/ml) and fixed at 10, 20, 30, and 40 min after plating.
Transferrin uptake A7 cells were cultured on coverslips (poly-L-lysine coated) transfected with HA-ARHGAP22. After 24 h, the cells were incubated in serum-free growth medium for 1 h. The cells were then incubated with 20 mg/ml Alexa Fluor 568-conjugated transferrin (Invitrogen) at 37uC for 30 min and fixed immediately after incubation.

Antibodies
Mouse anti-FLAG (M2) and anti-a-tubulin monoclonal antibodies and rabbit anti-HA and anti-FLAG polyclonal antibodies were purchased from Sigma. Mouse anti-EEA1, anti-GM130, and anti-Rab11 monoclonal antibodies were purchased from BD Biosciences (Bedford, MA). Rabbit anti-TGN46 polyclonal antibody was purchased from Abcam (Cambridge, UK). Mouse anti-LAMP-1 monoclonal and Rabbit anti-Rab5 polyclonal antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse anti-HA (12CA5) and anti-FLNa monoclonal antibodies were purchased from Roche Applied Science (Indianapolis, IN) and Millipore (Billerica, MA), respectively. Secondary antibodies conjugated to Alexa Fluor 488 or 568, Alexa Fluor 568-phalloidin (Invitrogen), hoechst 33258 (Dojido laboratories, Kumamoto, Japan) were also purchased from commercial sources. Rabbit anti-FilGAP polyclonal antibody was prepared as described previously [9]. Rabbit anti-ARHGAP22 polyclonal antibody was directed against amino acid residues 469-485 (RGHRRASSGDRLKDSGS) of human ARHGAP22. The peptide was coupled through cysteine at the NH 2 -terminal residue to keyhole limpet hemocyanin (KLH) and was used to raise the antiserum. The antiserum specific to ARHGAP22 was affinitypurified with the immobilized peptide. siRNA siRNA oligonucleotide duplexes targeting human ARHGAP22 (BC126444) were purchased from Invitrogen. The targeting sequences were as follows: ARHGAP22, KD#1 59-GAUACAU-CUGCAAGUUUCUGGAUGA-39 (nt 902-926) and KD#3 59-GGAAAUAAAGCUGCGGAACUCUGAA-39 (nt 2004-2028). For siRNA rescue assay, 5 silent mutations were introduced to the siRNA targeting sequence (nucleotides 902-926). The final mutant was changed into GG 903 TACATA 909 TG-CAAA 915 TTC 918 CTGGAC 924 GA by PCR. The cells were treated with siRNA for 24 h followed by a transfection with rescue constructs. The cells were cultured for another 24 h and proceed for Western blot or spreading assay.

Statistical analysis
The statistical significance was accessed by two-tailed unpaired Student's t-test or Welch's t-test. Differences were considered to be statistically significant at p value of ,0.01. Error bars (s.e.m.) and p values were determined from results of at least three experiments. (A) Specificity of anti-ARHGAP22 antibody was shown by immunoblotting from HEK293 cells transfected with a control plasmid (pCMV5-HA) or pCMV5-HA plasmids encoding human FilGAP, ARHGAP22, or ARHGAP25. The HA-epitope and tubulin (loading control) were also detected by immunoblotting using anti-HA and anti-tubulin antibodies, respectively. Arrowheads and asterisks indicate HA-tagged proteins and non-specific bands, respectively. (B) A7 cells were transfected with HA-ARHGAP22. After 24 h, the cells were fixed and stained with anti-ARHGAP22 (green) and anti-Rab11 (red) antibodies. Merged fluorescent image is shown. The cells were also stained with hoechst 33258 for nuclei (blue). Scale bar, 20 mm. (TIF)