Kank attenuates actin remodeling by preventing interaction between IRSp53 and Rac1

The Rho family of small GTPases acts as molecular switches for a variety of extracellular signals (Hall, 1998). These signals are transduced through a rapid reorganization of the actin cytoskeleton that changes the cell shape, leading to cell adhesion and locomotion (Ridley et al., 1992). Rho GTPases are also implicated in many other cellular events and functions such as cell polarity, gene transcription, cell cycle progression in the G1 phase, microtubule dynamics, vesicular transport, and enzymatic processes (Kozma et al., 1996; Van Aelst and D'Souza-Schorey, 1997; Aspenström, 1999; Etienne-Manneville and Hall, 2002). The Rho family proteins such as Rac1 and cdc42 along with proteins like Wiskott-Aldrich syndrome protein (WASP), neural WASP, and Scar/WAVE participate in cell migration, neurite extension, and budding in yeast (Innocenti et al., 2004). These proteins can bind to the globular actin and Arp2/3 complex through their catalytic domain, which results in filament branching at the membrane (Takenawa and Miki, 2001). The regulatory actions of WASP and neural WASP proteins involve their binding to active cdc42 at their GTPase-binding domain (Higgs and Pollard, 2000; Takenawa and Miki, 2001). WAVE1 was found to be inactive in a complex with Nap1, Abi2, PIR121, and HSP300, and a GTP-loaded active Rac1 dissociated from this complex, relieving active WAVE1-HSP300 (Eden et al., 2002). In contrast, WAVE2 was found to bind to active Rac1 indirectly through insulin receptor substrate (IRS) p53 (Miki et al., 2000; Miki and Takenawa, 2002), a well-characterized adapter protein that connects actin remodeling proteins with the Rho family of small GTPases (Funato et al., 2004). IRSp53 contains several domains: a Rac1-binding domain in the N terminus, a half cdc42/Rac1 interactive binding (CRIB) motif, a proline-rich domain, and an Src homology 3 (SH3) domain. It also binds to cdc42 via the CRIB motif and stimulates the formation of filopodia through Mena (Govind et al., 2001; Krugmann et al., 2001). IRSp53 is involved in neuronal morphogenesis through a variety of proteins (Soltau et al., 2002, 2004; Choi et al., 2005; Hori et al., 2005).

During a comprehensive analysis of loss of heterozygosity in renal cell carcinoma patients, Kank, a kidney ankyrin repeat–containing protein, was identified as a growth suppressor in HEK293 cells and a disruptor of β-actin distribution in G-402 cells (Sarkar et al., 2002; Rodley et al., 2003). The protein contains a coiled-coil domain in the N-terminal region and an ankyrin repeat domain in the C-terminal region, both of which are likely to be involved in protein–protein interactions and thus may play a major role in cellular events. Interestingly, an orthologue of Kank in Caenorhabditis elegans, VAB-19, was reported to act in epidermal morphogenesis and to play a significant role in the regulation of the actin cytoskeleton (Ding et al., 2003).

In this study, we report that Kank disrupts the function of active Rac1 through IRSp53. The binding between IRSp53 and Kank inhibits the association of active Rac1 with IRSp53 rather than the association of active cdc42 with IRSp53. Kank inhibits the formation of lamellipodia and membrane ruffles induced by active Rac1 in NIH3T3 cells. In addition, the depletion of Kank induced lamellipodial membrane protrusions in NIH3T3 cells. Furthermore, Kank inhibits IRSp53-induced cell spreading and neurite outgrowth in N1E115 cells.

To identify proteins binding to Kank, we screened a human placental cDNA library using Kank (amino acids 77–977) as bait with the yeast two-hybrid system. Among >200 clones isolated, two encoded IRSp53, both lacking 49 amino acids from the N terminus. The specificity of interaction was confirmed by the growth of the yeast clones on plates lacking Leu/Trp/His and by β-galactosidase activity in the X-gal filter assay (unpublished data). To verify the result of the screening, reciprocal coimmunoprecipitation experiments were performed using antibodies against Kank and IRSp53 with mammalian cell lysates (Fig. 1 A). Immunoprecipitation of Kank resulted in the coprecipitation of IRSp53 as detected by Western blotting using the anti-IRSp53 antibody, whereas the result of the control experiment using a preimmune serum was negative. However, immunoprecipitation of IRSp53 resulted in coprecipitation of Kank as detected with the anti-Kank antibody, confirming binding in vivo between Kank and IRSp53. Immunocytochemical characterization of Kank in VMRC-RCW cells, a renal cell carcinoma cell line, also showed the colocalization of endogenous Kank and IRSp53 at the cell periphery (Fig. 1 B, arrowheads). To further authenticate our finding, we coimmunostained Kank with WAVE2, which is known to be localized to areas of ruffling, in VMRC-RCW cells. Kank and WAVE2 were colocalized at the cell periphery (Fig. 1 C, arrowheads). Furthermore, Kank and IRSp53 were both present in the membrane fraction (Fig. 1 D). In the spreading cells, IRSp53 was detected at the tips of both the lamellipodium and the filopodium (Nakagawa et al., 2003) where WAVE2 is also localized (Miki et al., 2000). Thus, we concluded that Kank is localized at the leading edge in the VMRC-RCW cells.

To determine the binding domains of Kank and IRSp53, various mutants were constructed for IRSp53 (in GFP-tagged forms; Fig. 2 A) and Kank (in Flag-tagged forms; Fig. 2 C).2 HEK293T cells were transfected with the indicated constructs along with the wild-type Flag-Kank, and immunoprecipitation was performed using an anti-Flag antibody (Fig. 2 B). Positive GFP signals were observed for the wild-type GFP-IRSp53 (Fig. 2 A, construct a) and GFP-IRSp53(N) (N-terminal part containing the IMD/RCB domain and half CRIB domain; Fig. 2 A, construct c). In contrast, a mutant of IRSp53 having a deletion at the coiled-coil domain (amino acids 75–153; GFP-IRSp53^Δcoil; Fig. 2 A, construct b), a mutant containing the C-terminal part, including the SH3 domain (amino acids 362–521; construct d), and the GFP-vector (control; Fig. 2 A, construct v) did not show any signal for GFP in the immunoprecipitates, indicating that Kank binds to IRSp53 at its coiled-coil domain. The lanes for lysates showed the expression of the individual constructs. Similarly, Flag-tagged Kank and its mutants were expressed in HEK293T cells along with GFP-IRSp53(N) (Fig. 2 A, construct c) and were immunoprecipitated with polyclonal anti-GFP antibody (Fig. 2 D). Western blotting using an anti-Flag antibody showed strong signals for Flag-Kank (Fig. 2 C, construct e) and Flag-Kank^ΔAnk (Fig. 2 C, construct g) but no signal for Flag-Kank^Δcoil (Fig. 2 C, construct f). The result also indicates that Kank binds to IRSp53 at the coiled-coil domain. This result was further verified by coexpressing Flag-IRSp53(N) (Fig. 2 A, construct c) and a GFP-tagged coiled-coil region of Kank (Fig. 2 C, construct h) in HEK293T cells and by examining reciprocal immunoprecipitation followed by Western blotting (Fig. 2 E). The specificity of the binding was further examined by using a coiled-coil domain of kinectin (amino acids 1,116–1,356; GFP-kinectin^coil; Fig. 2 F). In this study, GFP-kinectin^coil did not bind to Flag-IRSp53(N). Thus, the results indicate that IRSp53 binds specifically to Kank through its coiled-coil domain.

As reported, active Rac1 binds to IRSp53 through its N-terminal 229 amino acid residues (Miki et al., 2000; Miki and Takenawa, 2002), which contain the Kank binding domain, so we tested whether Kank can compete against Rac1 in vitro (Fig. 3, A and B). Both His-Rac1^G12V and His-cdc42^G12V bound to GST-IRSp53(N), whereas GST alone did not, indicating that IRSp53 binds to Rac1 or cdc42 (Fig. 3, A and B, lanes 2 and 3). When this binding was challenged by maltose-binding protein (MBP; a control) or MBP-Kank^coil, MBP-Kank^coil inhibited the binding of GST-IRSp53(N) to His-Rac1^G12V in a dose-dependent manner more efficiently than the binding of GST-IRSp53(N) to His-cdc42^G12V (Fig. 3, A and B, lanes 6–8). MBP alone or MBP-kinectin^coil (containing a control coiled-coil domain) did not affect the binding (Fig. 3, A and B, lanes 4 and 5). In Fig. 2 B, we determined the binding domain of Kank in IRSp53 (amino acids 80–150), which also overlapped with the coiled-coil domain and was located within the RCB. Yamagishi et al. (2004) reported that the N-terminal half of IRSp53 contains a filopodium-inducing domain regulated by Rac1 and cdc42. IRSp53 contains a partial CRIB motif (Burbelo et al., 1995; Govind et al., 2001; Krugmann et al., 2001), a sequence of 16 amino acid residues (amino acids 267–283), which is located far from the Kank binding domain (amino acids 80–150). Our result showed that the binding of Kank to IRSp53 partially affected the binding between cdc42 and IRSp53 (Fig. 3, A and B). We then examined the effect of knockdown (KD) of Kank using endoribonuclease-prepared siRNA (esiRNA) on the binding between active Rac1 and IRSp53 in vivo (Fig. 3 C). For this, we established HEK293 cells stably expressing active Rac1 (Rac1^G12V). These cells were transfected with Kank esiRNA along with control esiRNA. After treatment with esiRNA for 30 h, the cells were subjected to immunoprecipitation using an anti-IRSp53 antibody that was raised against the C-terminal portion of IRSp53. The results of Western blot analysis revealed that Kank esiRNA efficiently decreased the level of Kank protein in the lysates but did not change the level of IRSp53 or Rac1 protein (Fig. 3 C, compare lanes 3 and 4 with lanes 1 or 2 in the lysate panel). The results of immunoprecipitation showed that depletion of Kank with Kank esiRNA resulted in less Kank (Fig. 3 C, compare lanes 3 and 4 with lane 2 in the top row) and, at the same time, increased the amount of active Rac1 (Rac1^G12V; Fig. 3 C, lanes 2–4 in the third row), both of which were immunoprecipitated with IRSp53. In other words, KD of Kank resulted in increased binding between IRSp53 and active Rac1. Normal IgG was used as a negative control, and GST-PAK1^CRIB was used for the pull down of active Rac1 (Fig. 3 C, lanes 1 and 5, respectively). From these results, Kank specifically inhibits the binding of Rac1 to IRSp53.

Kank affected the interaction between IRSp53 and Rac1 and partially affected that between IRSp53 and cdc42 (Fig. 3). This led us to investigate the effect of Kank expression on the formation of lamellipodia or filopodia mediated by Rac1 or cdc42 (Figs. 4 and 5). Initially, IRSp53 was reported to be a downstream target of Rac1 that links Rac1 activity to WAVE/Scar and the Arp2/3 complex (Miki et al., 2000; Etienne-Manneville and Hall, 2002). However, others reported that IRSp53 preferentially binds to active cdc42 (Govind et al., 2001; Krugmann et al., 2001). Overexpression of Kank severely impaired the development of lamellipodia and cell spreading induced by a constitutively active form of Rac1, Rac1^G12V (Fig. 4 A, lane 4). Overexpression of Kank^S167A, which lacks the ability to bind to 14-3-3 and has inhibitory effects on active RhoA and cell migration (Kakinuma et al., 2008), also impaired the formation of lamellipodia induced by Rac1^G12V (Fig. 4 A, lane 5). However, Kank^Δcoil, which cannot interact with IRSp53, had no effect on the Rac1^G12V-induced development of lamellipodia (Fig. 4 A, lane 6). To confirm that this effect of Kank depends on the interaction with IRSp53, we performed coexpression of Kank with IRSp53 (Fig. 4 B). Coexpression of GFP-Rac1^G12V and Kank with IRSp53 resulted in the disappearance of Kank's inhibitory effect on the Rac1^G12V-dependent formation of lamellipodia (Fig. 4 B, lane 4). However, coexpression of GFP-Rac1^G12V and Kank with IRSp53^R11E/Q23E or IRSp53^K143E, which could not bind to Rac1 and could bind to Kank (Fig. S1, available; Suetsugu et al., 2006b), resulted in the retention of the inhibitory effect on the Rac1^G12V-dependent development of lamellipodia (Fig. 4 B, lanes 5 and 6). This result implies that Kank inhibits Rac1–IRSp53 signals for the formation of lamellipodia. Moreover, to confirm that this Rac1^G12V-dependent formation depends on IRSp53, we performed KD of IRSp53 along with GFP-Rac1^G12V expression (Fig. 4 C). We first examined the KD of mouse IRSp53 (mIRS) in NIH3T3 cells using various candidate plasmids (Fig. S2). These plasmids contained an H1 promoter–based expression system for siRNAs in mammalian cells (Steffen et al., 2004). Based on the results, we used #513 siRNA for the KD of mIRS. Coexpression of GFP-Rac1^G12V and mIRS-KD reduced the Rac1^G12V-dependent formation of lamellipodia (Fig. 4 C, lane 3), whereas on the coexpression of human IRSp53 with GFP-Rac1^G12V and mIRS-KD, GFP-Rac1^G12V retained its function (Fig. 4 C, lane 4). However, the coexpression of human IRSp53^R11E/Q23E or IRSp53^K143E with GFP-Rac1^G12V and mIRS-KD had little effect on mIRS-KD's function in the Rac1^G12V-dependent formation of lamellipodia (Fig. 4 C, lanes 5 and 6). Therefore, the interaction between IRSp53 and active Rac1 is important for lamellipodia to form. These results suggest that IRSp53 is one of the targets of Rac1^G12V in the Rac1^G12V-dependent development of lamellipodia. However, overexpression of Kank and Kank^S167A had little effect on the formation of filopodia induced by a constitutively active form of cdc42, cdc42^G12V (Fig. 5, lanes 3 and 4), which may imply that Kank was not involved in the cdc42^G12V-mediated formation of microspikes.

Based on the finding that Kank bound to IRSp53 and inhibited active Rac1-mediated lamellipodial development, as well as reports of IRSp53 as a downstream target of Rac1 (Miki et al., 2000; Miki and Takenawa, 2002), we examined the effect of KD of Kank on the formation of lamellipodia (Fig. 6).6 NIH3T3 cells were transfected with a Kank KD vector that targeted mouse Kank transcripts, a control KD vector, and an mIRS-KD as indicated in Fig. 6. Because we used plasmids containing a neo-gfp fusion gene in this assay, transfected cells can be stained with an anti-GFP antibody against GFP or can be detected with GFP fluorophore. The cells were cultured for 48 h after transfection and fixed and stained with anti-GFP antibody (Fig. 6 B; green) or with phalloidin (Fig. 6 B; red). Transfection of a KD vector (Kank-KD) effectively suppressed the expression of Kank protein by ∼80% with respect to control-KD (Fig. 6 A, lanes 2 and 3), and mIRS-KD suppressed the expression of mIRS protein by ∼100% with respect to control-KD (Fig. 6 A, lanes 3 and 4). A careful observation of cell morphology revealed that silencing of Kank resulted in the formation of lamellipodia (Fig. 6 B, lane 2). However, the silencing of both Kank and IRSp53 simultaneously had little effect (Fig. 6 B, lane 3). These findings support the idea that Kank inhibits lamellipodia from forming by interrupting the interaction between active Rac1 and IRSp53.

Lamellipodia and filopodia are formed at the leading edge in migrating cells in response to growth factors, chemokines, and extracellular matrix molecules (Ridley et al., 2003). Insulin acts as a growth factor, and insulin-stimulated KB cells, a strain derived from human epidermoid carcinoma, show membrane ruffling and Rac1 activation (Nishiyama et al., 1994). Overexpression of Kank inhibited active Rac1-dependent lamellipodial development (Fig. 5), and KD of Kank enhanced the formation of lamellipodia (Fig. 6). We then investigated whether Kank can inhibit insulin-induced membrane ruffling (Fig. 7). Whereas serum-starved NIH3T3 cells without treatment showed little membrane ruffling (Fig. 7, lane 1), insulin-stimulated cells exhibited extensive membrane ruffling (Fig. 7, lane 2). When Kank was overexpressed, the number of cells with membrane ruffling was significantly decreased (Fig. 7, lane 3). However, the effect of overexpression of Kank was not observed when IRSp53 was coexpressed (Fig. 7, lane 5). This result suggests that IRSp53 abrogates the inhibitory effect of Kank on insulin-induced membrane ruffling and, thus, that Kank inhibits insulin-induced membrane ruffling through IRSp53.

Integrins are major adhesion receptors in vertebrates and play major roles in cell adhesion and in a variety of intracellular signaling pathways (Hynes, 2002). The Rho family of small GTPases such as Rac1 and cdc42 are involved in integrin-mediated cell spreading (Clark et al., 1998; Price et al., 1998; Katoh and Negishi, 2003; Choi et al., 2005). As Kank interrupted the interaction between active Rac1 and IRSp53 (Fig. 3), we hypothesized that it might inhibit integrin-mediated cell spreading. To investigate the role of Kank downstream of integrin signaling, NIH3T3 cells were transfected, seeded on fibronectin-coated glass substrata, fixed, and stained with phalloidin (Fig. 8 A).8 The cells adhered rapidly to the substrata and completed spreading in 30 min, showing a normal shape (Fig. 8 A, lane 1). However, the expression of the wild-type Kank radically changed cell shape and markedly inhibited cell spreading (Fig. 8 A, lane 2). Notably, this change was not caused by apoptosis because nuclear staining of transfected cells with DAPI did not show any sign of apoptosis (unpublished data) and the cells were viable, as they could spread later (unpublished data). Kank^Δcoil, which cannot bind to IRSp53, did not show any significant changes in cell shape or cell spreading (Fig. 8 A, lane 3). Kank^S167A, which cannot bind to 14-3-3, partially inhibited cell spreading (Fig. 8 A, lane 4). We obtained similar results using COS-7 and HeLa cells (unpublished data). In contrast, a control protein with the coiled-coil domain, kinectin^coil, did not affect the spreading of NIH3T3 cells (Fig. 8 A, lane 5). From these results, Kank inhibits cell spreading specifically through its coiled-coil domain.

To see how Kank inhibits cell spreading and whether this phenomenon depends on IRSp53, Kank was coexpressed with IRSp53 in NIH3T3 cells (Fig. 8 B). When Kank was coexpressed with wild-type IRSp53, cell spreading was enhanced compared with that when only Kank was expressed (Fig. 8 B, compare lane 2 with lane 3). Integrin activates Rac1 and cdc42 (Price et al., 1998; Hynes, 2002). We then examined whether the inhibitory effect of Kank on cell spreading depends on IRSp53–Rac1 signaling. Kank was coexpressed with IRSp53^R11E/Q23E or IRSp53^K143E in NIH3T3 cells when the cells were spreading (Fig. 8 B, lanes 4 and 5). Surprisingly, the rates of spreading differed little between wild-type IRSp53, IRSp53^R11E/Q23E, and IRSp53^K143E when Kank was present (Fig. 8 B, lanes 3–5). Next, we investigated whether IRSp53 was involved in the spreading of NIH3T3 cells (Fig. 8 C). KD of mIRS resulted in the inhibition of cell spreading (Fig. 8 C, compare lane 1 with lane 3). In contrast, KD of mIRS with overexpression of the wild type or the mutants of human IRSp53 did not affect the spreading of NIH3T3 cells (Fig. 8 C, lanes 4–6). Furthermore, there was almost no change in cell spreading with or without KD of IRSp53 when Kank was overexpressed (Fig. S3, available).

We then investigated the effect of Kank and IRSp53 expression on cell morphology when NIH3T3 cells were spreading upon stimulation with fibronectin (Fig. S4, available). Cell shapes were classified according to those in the process of matrix-based cell spreading (Applewhite et al., 2007). Overexpression of Kank decreased the number of cells with the filopodial phenotype, whereas overexpression of IRSp53 increased the number of smooth-edged cells. Notably, coexpression of Kank and IRSp53 decreased the number of cells with the filopodial phenotype, whereas there was no increase in the number of the cells with the smooth-edged phenotype. These results suggest that Kank partially inhibits the spreading of NIH3T3 cells through IRSp53, although this phenomenon was not part of the Rac1–IRSp53 signaling pathway. The reason why IRSp53 partially recovers cell spreading inhibited by Kank might be that Kank inhibits the formation of both filopodial microspikes and lamellipodial sheets, whereas IRSp53 contributes only to lamellipodial sheets when the cells are spreading.

Ectopic expression of IRSp53 in mouse neuroblastoma N1E115 cells induces neurite outgrowth (Govind et al., 2001; Miki and Takenawa, 2002), with Rac1 and cdc42 acting as positive regulators and Rho inhibiting the process (Kozma et al., 1995, 1997; Luo, 2000). To assess whether Kank is also involved in IRSp53-mediated neurite outgrowth, N1E115 cells were transfected with GFP-tagged IRSp53 to reproduce the morphological changes described previously (Govind et al., 2001). Overexpression of GFP-Kank inhibited neurite outgrowth under serum starvation (Fig. 9 A). The cells transfected with GFP-IRSp53 showed vigorous neuritic development in the presence or absence of serum (Fig. 9 B, second row). There were no round or adherent neurites when the cells were not transfected in the presence of serum. When Kank was coexpressed with GFP-IRSp53, the cells showed a phenotype similar to that seen with GFP alone (Fig. 9 B, bottom, Serum [+]). The quantitative analysis showed that Kank significantly reduced the extent of the IRSp53-induced outgrowth of neurites both in the presence (from 40 to 14%) and in the absence (from 85 to 25%) of serum (Fig. 9 B, graph). These findings demonstrate that Kank is involved in the IRSp53-mediated formation of neurites.

In this study, we demonstrated that Kank plays a major role in the negative regulation of lamellipodial development in fibroblasts. The regulation is likely to be mediated by interaction with IRSp53 in ruffled areas of the cell. This binding seems to be constitutive because it was not affected by stimulation with a growth factor like PDGF or EGF (unpublished data). IRSp53 is a multidomain scaffolding protein (Govind et al., 2001; Krugmann et al., 2001; Soltau et al., 2002, 2004; Funato et al., 2004; Choi et al., 2005; Connolly et al., 2005; Hori et al., 2005) and a downstream target of Rac1 and links WAVE2 to Rac1 (Miki et al., 2000; Miki and Takenawa, 2002), which plays an important role in the reorganization of the actin cytoskeleton.

Kank contains two important domains, the coiled-coil and ankyrin repeat domains, both of which have been described in several structural proteins as well as in proteins involved in an array of cellular functions (Sedgwick and Smerdon, 1999; Burkhard et al., 2001). We showed in this study that the coiled-coil domain of Kank is necessary for binding with IRSp53 and the negative regulation of cell spreading and lamellipodial development induced by constitutively active Rac1 (Fig. 4).

Although active Rac1- or active cdc42-dependent actin polymerization has been described for the regulation of WASP and WAVE proteins (Bompard and Caron, 2004) and the regulation of Rac1 activity through three major classes of regulators, GTPase-activating proteins, guanine nucleotide exchange factors, and GDP dissociation inhibitors, has been identified, albeit with little understanding of the signaling pathways that modulate their activity (Burridge and Wennerberg, 2004), Kank does not possess any of these classical regulatory functions in Rac1 activation. This raises the possibility that Kank functions via other proteins also.

IRSp53 makes a complex with Eps8 and activates Rac1, which is important to cancer cell motility and invasion (Funato et al., 2004). Active cdc42 binds to the complex of IRSp53 and Eps8 and controls its cellular distribution, thereby contributing to the generation of actin bundles and promoting filopodial protrusions (Disanza et al., 2006). Eps8 was shown to form a complex with Abi-1, Sos-1, E3B1, and PI3-K, which activate Rac1 and transduce signals from Ras to Rac1 (Biesova et al., 1997; Scita et al., 1999, 2001; Innocenti et al., 2003). Tiam1, another Rac1 activator, also binds to IRSp53, leading to increased activation of Rac1 and enhanced binding of IRSp53 to both active Rac1 and WAVE2. Tiam1 promotes the localization of IRSp53 to Rac1-induced lamellipodia rather than cdc42-induced filopodia (Connolly et al., 2005). Similarly, Kank specifically inhibits the binding between IRSp53 and Rac1^G12V rather than that between IRSp53 and cdc42^G12V (Fig. 3, A and B), which is consistent with the finding that the coexpression of Kank with active Rac1 or active cdc42 in NIH3T3 cells showed marked differences: the formation of lamellipodia induced by active Rac1 was inhibited (Fig. 4), whereas microspikes or filopodia generated by active cdc42 showed some but not great alteration (Fig. 5). Because the IMD/RCB domain of IRSp53 overlaps with the Kank binding domain, Kank might disrupt the association of active Rac1 to IRSp53. In contrast, the cdc42-binding domain of IRSp53 is located far from the Kank binding domain. This may be why Kank weakly inhibits cdc42-IRSp53 association. KD of Kank by RNAi in NIH3T3 cells revealed that depletion of Kank resulted in an increase in lamellipodia, whereas KD of both Kank and IRSp53 resulted in normal or unaltered cell shapes (Fig. 6 B). Moreover, Kank inhibited insulin-induced membrane ruffling, and coexpression of Kank with IRSp53 abrogated this inhibitory effect (Fig. 7). This suggests that Kank inhibits the IRSp53-dependent formation of lamellipodia. However, Kank inhibited integrin-dependent cell spreading through IRSp53, but this was not caused by Rac1 (Fig. 8, B and C). Kank also inhibited IRSp53-dependent neurite outgrowth in N1E115 (Fig. 9). Kank binds to the IMD/RCB domain of IRSp53, and it has been shown that this domain binds not only to active Rac1 but also to actin and lipids (Miki et al., 2000; Yamagishi et al., 2004; Millard et al., 2005; Suetsugu et al., 2006b). The IMD/RCB domain was also shown to have actin-bundling activity (Millard et al., 2005), although this activity may only be seen under nonphysiological conditions such as at a low ionic strength (Lee et al., 2007; Mattila et al., 2007). Consistent with this, we also found that actin was present in a complex immunoprecipitated with an anti-IRSp53 antibody in HEK293 cells (Fig. S5, available). Expression of Kank is likely to inhibit the association of IRSp53 with actin, which implies that Kank negatively regulates the association of IRSp53 with actin. This might be why Kank inhibits the spreading of NIH3T3 cells and the outgrowth of neurites in N1E115 cells through the binding of IRSp53.

Kank regulates RhoA activity through its 14-3-3 binding motif inside the coiled-coil domain (Kakinuma et al., 2008). Kank inhibited Rac1–IRSp53 signaling, which is mediated by the coiled-coil domain (Fig. 4). However, the formation of lamellipodia induced by active Rac1 was inhibited by both Kank and Kank^S167A, which cannot bind to 14-3-3 and was ineffective against active RhoA (Fig. 4 A, lanes 4 and 5). From this finding, the effect of Kank on active Rac1-induced lamellipodial development can be separated from the effect of the binding of 14-3-3 to Kank. In contrast, overexpression of Kank^S167A was relatively ineffective compared with the effect of wild-type Kank on integrin-dependent cell spreading (Fig. 8 A, compare lane 2 with lane 4). A deletion mutant, Kank^Δcoil, was not completely effective against integrin-dependent cell spreading (Fig. 8 A, lane 3); Kank may affect the integrin-dependent cell spreading mediated by both IRSp53 and 14-3-3 or more likely through the Kank coiled-coil domain.

Although there are orthologues of Kank in different species, in which the coiled-coil domain along with the ankyrin repeat domain are conserved, only VAB-19 of C. elegans has been studied and was shown to be a component of the epidermal attachment structure and to interact with actin (Ding et al., 2003). This further strengthens our finding that Kank functions through the coiled-coil domain in the regulation of actin remodeling. This function of Kank in the reorganization of the actin cytoskeleton may be conserved in other orthologues as well. Further studies are needed to explore other potential roles of Kank or the Kank–IRSp53 complex in response to external stimuli, which may contribute to our understanding of the cell signaling mechanisms in renal cell carcinoma.

Cytomegalovirus-driven Flag-tagged constructs (Agilent Technologies), GFP-tagged constructs (BD), the vector pcDNA-3.1 (Invitrogen), the hEF1-α–driven expression vector pEF1/myc-His (Invitrogen), and a bacterial expression vector for GST, pGEX-5x-1(GE Healthcare), were generated by recombinant PCR. Full-length human IRSp53 cDNA was obtained by RT-PCR. Kank mutants and human IRSp53 mutants were made by PCR-mediated recombination. The coiled-coil domain of kinectin (kinectin^coil) was obtained by RT-PCR and used as a control (Vignal et al., 2001). All of the constructs were verified by sequencing. Polyclonal and monoclonal antibodies against Kank were described previously (Sarkar et al., 2002; Roy et al., 2005). The polyclonal antibody against mouse Kank recognizes a peptide spanning amino acids 861–975 (Kakinuma et al., 2008). The polyclonal antibody against IRSp53 was a gift from M. Yamada (National Research Institute for Child Health and Development, Tokyo, Japan; Okamura-Oho et al., 1999) or was generated in a rabbit using the C-terminal part of the protein (163 residues long, amino acids 358–521) fused to GST as an antigen and was prepared by affinity purification. A monoclonal antibody against autofluorescent proteins (MP Biomedicals), a rabbit antiserum against GFP (Invitrogen), a polyclonal antibody against GFP (Clontech Laboratories, Inc.), a monoclonal antibody against cdc42 protein, a polyclonal antibody against EGF receptor (Santa Cruz Biotechnology, Inc.), a monoclonal antibody against Rac1 (Transduction Laboratories), a goat polyclonal antibody against WAVE2 (Santa Cruz Biotechnology, Inc.), and monoclonal antibodies against Flag, GFP, β-actin, and β-tubulin (Sigma-Aldrich) were purchased.

HEK293 cells stably expressing Rac1^G12V were constructed with the pcDNA4/TO expression vector (Invitrogen). NIH3T3, HeLa, N1E115 (a gift from Y. Kanaho, Tsukuba University, Ibaraki, Japan), HEK293 (a gift from M. Noguchi, Hokkaido University, Hokkaido, Japan), and HEK293T cells were cultured in DME supplemented with 10% FBS. VMRC-RCW cells were cultured in MEM with 10% FBS. Transfection was performed using Lipofectamine 2000 (Invitrogen) and Effectine (QIAGEN) according to the manufacturers' instructions. After 30 h, cells were harvested, washed with ice-cold PBS, and lysed for 15 min on ice in buffer A (50 mM Tris-HCl, pH 7.5, 140 mM NaCl, 10% glycerol, 1% Nonidet P-40, 100 mM NaF, 200 mM Na₃VO₄, 1 mM PMSF, 10 µg/ml leupeptin, 10 µg/ml aprotinin, and 10 µg/ml chymotrypsin). The cell lysates were clarified by centrifugation, and the proteins were immunoprecipitated, washed in the lysis buffer, and boiled in an SDS-PAGE loading buffer. Proteins were separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes (Immobilon; Millipore), and detected with respective antibodies. Signals were enhanced with secondary antibody against AP (Sigma-Aldrich) or HRP (GE Healthcare) and developed with nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (Sigma-Aldrich) or ECL (GE Healthcare). The intensity of the bands was quantified by ImageJ software (National Institutes of Health).

The cells on coverslips were fixed with 3.7% formaldehyde-PBS for 30 min and permeabilized with 0.25% Triton X-100–PBS for 10 min. The cells were subsequently incubated with the primary and secondary antibodies. The secondary antibodies used were Alexa Fluor 594–conjugated goat anti–mouse and anti–rabbit IgG (Invitrogen), Alexa Fluor 488–conjugated goat anti–mouse and anti–rabbit IgG (Invitrogen), RITC-conjugated donkey anti–goat IgG (Santa Cruz Biotechnology, Inc.), and Cy5-conjugated goat anti–rabbit IgG (Millipore). Filamentous actin (F-actin) was stained with TRITC-conjugated phalloidin (Sigma-Aldrich). Images of cells after they were mounted were acquired at room temperature in 80% (wt/vol) glycerol in PBS using AxioVision 3.1 software (Carl Zeiss, Inc.) and a fluorescence microscope (Axioskop 2 Plus; Carl Zeiss, Inc.) with a Plan-NEOFLUAR 40× NA 0.75 objective (Carl Zeiss, Inc.) and a color camera (AxioCam HRc; Carl Zeiss, Inc.). Confocal laser microscopy was performed using a laser-scanning confocal microscope (Axiovert 200M; Carl Zeiss, Inc.) with LSM510 META software (Carl Zeiss, Inc.). Optical thickness was set at 1.0 µm in Z height with a Plan-Apochromat 63× NA 1.4 oil immersion objective (Carl Zeiss, Inc.). All images were prepared with Photoshop 7.0 (Adobe).

NIH3T3 cells were transfected as indicated and incubated for 24 h (for overexpression) or 48 h (for KD). The cells were then trypsinized and seeded on 10 µg/ml fibronectin-coated coverslips. After incubation overnight, they were subjected to immunostaining as described in Immunocytochemistry. The cells with lamellipodia were those with broad ruffles, whereas the cells with filopodia were those with >15 microspikes in phase-contrast images.

NIH3T3 cells were transfected as indicated. After 4 h, the cells were trypsinized and seeded onto fibronectin-coated coverslips. When they had adhered to coverslips completely, the cells were serum starved by culturing them in 0.5% FBS-DME for 2 d. Then, the cells were treated with 10 µg/ml insulin for 6 h. The cells with membrane ruffling were counted from phase-contrast images and the images for F-actin (phalloidin) staining.

Integrin stimulation was performed as described previously (Katoh and Negishi, 2003). In brief, cells were transfected as indicated, incubated with trypsin, and suspended in a serum-free medium containing 1 mg/ml of trypsin inhibitor (from soybeans; Sigma-Aldrich). After 1 h, the cells were transferred onto fibronectin-coated coverslips and cultured for 30 min. The number of spreading cells was counted for the cells that were completely adhered and examined by phase-contrast microscopy. The classification of spreading cell morphology was performed as described previously (Applewhite et al., 2007). In brief, the cells were transfected as indicated and after 24 h were trypsinized and seeded on fibronectin-coated coverslips. The cells were incubated for 30–40 min to allow them to attach and spread. They were immunostained as described in Immunocytochemistry. Morphological typing was performed based on F-actin (phalloidin) staining and phase-contrast images.

The N1E115 cells were seeded on 10 µg/ml laminin-coated coverslips and cultured overnight. The next day, they were transfected with the respective expression vectors in the presence of serum. After 16 h, the cells were serum starved for another 12 h, immunostained as described in Immunocytochemistry, and analyzed as described previously (Miki and Takenawa, 2002; Yamazaki et al., 2002). In brief, cells with neurites were defined as the cells possessing at least one neurite two times longer than the diameter of the cell body from phase-contrast images.

HeLa cells suspended in hypotonic buffer (10 mM Tris-HCl, pH 7.5, 1 mM EGTA, 1 mM MgCl₂, 50 mM NaF, 1 mM Na₃VO₄, and a protease inhibitor cocktail; Sigma-Aldrich) were homogenized by 10 passages through a 22-gauge needle. The samples were then centrifuged at 4,300 g for 10 min, and the supernatants were again centrifuged at 50,000 g for 1 h to separate the cytosolic and membrane fractions. The membrane fraction was washed three times with the hypotonic buffer and dissolved in buffer A with 1% SDS. The protein in the cytosolic and membrane fractions was quantified using a DC Protein Assay kit (Bio-Rad Laboratories). Each fraction was mixed with SDS sample buffer, and aliquots of the samples containing 30 µg of protein were analyzed by Western blotting.

An H1 promoter–based mammalian expression vector, pSUPER.neo, or pSUPER.neo+gfp, which coexpresses siRNA and a neo-gfp fusion gene (OligoEngine), was used for the expression of siRNA in mouse NIH3T3 fibroblasts. A 19-nucleotide sequence corresponding to nucleotides 609–627 (5′-GAGGCGAAAGCCATCTGTG-3′) of the mouse cDNA clone A530060J16 (available from GenBank/EMBL/DDBJ under accession no. AK041011.1) was used as the Kank RNAi (Kank-KD) vector. A control vector was constructed using a 19-nucleotide sequence (5′-ACTAGACGAAGCGGTACTG-3′) with no significant homology to any mammalian gene sequences and used as a nonsilencing control (control #2 in Suetsugu et al., 2006a). For KD of mIRS, we searched for candidate KD sequences by Sfold (http://sfold.wadsworth.org/), tested their efficiency using NIH3T3 cells (Fig. S2), and selected a 19-nucleotide sequence (#513) corresponding to nucleotides 513–531 (5′-GTAAGAACCCTCAGAAGTA-3′) of mIRS (available from GenBank/EMBL/DDBJ under accession no. AF390179). For transfection, after cells were grown to 80–90% confluency, they were detached with trypsin/EDTA, washed with PBS, suspended at ∼10⁶ cells/ml in PBS, and subjected to electroporation. Electroporation was performed with 18 µg of plasmid DNA using a Gene Pulser II System (Bio-Rad Laboratories) at 270 mV and 975-µF capacitance. Under these conditions, we estimated the transfection efficiency to be around 50–70% by comparing the level of expression of GFP in a parallel experiment. After electroporation, the cells were kept on ice for 5 min and transferred to plates with 10 ml of a complete medium. After 36 h, the cells were lysed in a lysis buffer for checking the expression of Kank protein. Kank esiRNA was prepared as described previously (Yang et al., 2002; Kittler et al., 2004). In brief, sense and antisense RNA was transcribed using a T7 promoter system (for both strands) supplied in the MEGAscript kit (Applied Biosystems). After annealing the sense and antisense strands, RNA was treated with ShortCut RNase III (New England Biolabs, Inc.) and purified with Sepharose Q (GE Healthcare). Kank esiRNA and control esiRNA containing the Xenopus laevis elongation factor gene (from the MEGAscript kit) were transfected with Lipofectamine 2000 in HEK293 cells stably expressing Rac1^G12V.

1 µg His-Rac1^G12V or His-cdc42^G12V, which was produced in bacteria and purified with an Ni column, was added with 2 µg of control GST or GST-IRSp53(N) and 10 µg MBP or 1, 3, or 10 µg MBP-Kank^coil as competitor in binding buffer A. After these mixtures were rotated for 1 h at 4°C, glutathione–Sepharose 4B (GE Healthcare) was added to recover the protein complexes, and rotation was continued for 1 h at 4°C. The complexes were washed four times with the binding buffer, boiled for 5 min in an SDS-PAGE loading buffer, and subjected to 5–20% SDS-PAGE, Coomassie blue staining, and Western blot analysis using anti-Rac1 or anti-cdc42 antibodies to detect the respective proteins. The intensity of the bands was quantified by CS Analyzer (ATTO Corp.).

Fig. S1 shows IRSp53 mutants that bind to Kank^coil but not to Rac1^G12V. Fig. S2 shows the candidate sequences for IRSp53 KD. Fig. S3 shows that KD of IRSp53 does not affect fibronectin-stimulated cell spreading when Kank is overexpressed. Fig. S4 shows that Kank inhibits the formation of the cells with filopodial- and IRSp53-induced smooth-edged phenotypes when cells are spreading. Fig. S5 shows that overexpression of Kank inhibits the association of IRSp53 with actin.

We thank Professor T. Akiyama (the University of Tokyo, Tokyo, Japan) for advice and discussions, Dr. M. Yamada for the anti-IRSp53 serum, Professor M. Noguchi for HEK293 cells, Professor Y. Kanaho for N1E115 cells, and Professor R. Matsuoka and Dr. Y. Furutani (Tokyo Women's Medical University, Tokyo, Japan) for laser-scanning confocal microscopy.

This work was supported by the Special Coordination Fund for Promoting Science and Technology by the Ministry of Education, Culture, Sports, Science, and Technology of Japan and by a fund for promoting collaboration with small and medium enterprises from the National Institute of Advanced Industrial Science and Technology.