The Cdc42 Effector IRSp53 Generates Filopodia by Coupling Membrane Protrusion with Actin Dynamics*

The Cdc42 effector IRSp53 is a strong inducer of filopodia formation and consists of an Src homology domain 3 (SH3), a potential WW-binding motif, a partial-Cdc42/Rac interacting binding region motif, and an Inverse-Bin-Amphiphysins-Rvs (I-BAR) domain.We show that IRSp53 interacts directly with neuronal Wiskott-Aldrich syndrome protein (N-WASP) via its SH3 domain and furthermore that N-WASP is required for filopodia formation as IRSp53 failed to induce filopodia formation in N-WASP knock-out (KO) fibroblasts. IRSp53-induced filopodia formation can be reconstituted in N-WASP KO fibroblasts by full-length N-WASP, by N-WASPΔWA (a mutant unable to activate the Arp2/3 complex), and by N-WASPH208D (a mutant unable to bind Cdc42). IRSp53 failed to induce filopodia in mammalian enabled (Mena)/VASP KO cells, and N-WASP failed to induce filopodia when IRSp53 was knocked down with RNA interference. The IRSp53 I-BAR domain alone induces dynamic membrane protrusions that lack actin and are smaller than normal filopodia (“partial-filopodia”) in both wild-type N-WASP and N-WASP KO cells. We propose that IRSp53 generates filopodia by coupling membrane protrusion through its I-BAR domain with actin dynamics through SH3 domain binding partners, including N-WASP and Mena.

Filopodia and lamellipodia are ubiquitous and dynamic actinbased structures at the leading edge of cells that play important roles in processes such as cell invasion, cell migration, phagocytosis, and axonal guidance. Thus, understanding how the formation of filopodia and lamellipodia is regulated will give us insight into the fundamental aspects of cell biology and what goes wrong in disease states such as cancer. Work over the last few years has revealed many of the important players involved in cell signaling events that regulate actin dynamics associated with filopodia and lamellipodia formation (1). Rho GTPases (e.g. Cdc42, Rac1, and RhoA) are intimately involved in communication between cell surface receptors and proteins that control actin dynamics (2). Cdc42 is a major regulator of filopodia formation in mammalian cells. The isolation and identification of Cdc42 effectors have opened up the possibility of defining the molecular mechanisms responsible for the regulation and formation of filopodia. To date the Cdc42 effectors N-WASP, 2 IRSp53, PAK, and MRCK have been implicated in filopodia formation. In this study we focus on the role of Cdc42 effector IRSp53 and the mechanism by which it induces filopodia formation.
The WASP/N-WASP (Wiskott-Aldrich syndrome protein and neuronal-Wiskott-Aldrich syndrome protein) and WAVE1-3 (Wasp family verproline-homologue) family protein complexes are major downstream targets for Rho GTPase (3). WASP and WAVE proteins are activators of actin nucleation in vitro and adaptor proteins composed of a number of distinct domains as follows: WH1 (WASP Homology 1) and WH2 (WASP Homology 2), a basic stretch binding phosphatidylinositol 4,5-biphosphate, and the WA domain (W (verprolin and cofilin) and Acidic region). The latter domain is involved in binding to the Arp2/3 complex (Actin-related proteins 2 and 3). Rac1 interacts directly with one of the proteins present in the WAVE complex, p140 Sra-1, although the function of this interaction is unclear (4,5). WAVE1 and -2 complex proteins include Abi-1, p125 Nap-1, p140 Sra-1, and HSPC300 (6), respectively. Phosphatidylinositol 4,5-biphosphate and Toca-1 binding to N-WASP unfolds the protein to make the WA domain available to interact with the Arp2/3 complex (7,8). How the activities of "WASP" family proteins are coordinated both spatially and temporally is under intense study (3).
The adaptor protein IRSp53 was identified in a yeast twohybrid screen using the WAVE1 polyproline sequence as bait (9). Subsequent analysis suggested IRSp53 interacted with WAVE2 and might be involved in linking Rac1 to WAVE proteins (9,10). IRSp53 was also identified in a yeast two-hybrid screen using Cdc42 and shown to be an effector for Cdc42 (11,12). IRSp53 consists of an I-BAR 3 domain (Inverse-Bin-Am-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1-S4 and Movies 1-7. 1 To whom correspondence should be addressed: Institute of Medical Biology, 8A, Biomedical Grove, Immunos, Singapore 138665. Tel.: 65-6407-0165; Fax: 65-6464-2048; E-mail: sohail.ahmed@imb.a-star. edu.sg. 2 The abbreviations used are: N-WASP, neuronal Wiskott-Aldrich syndrome protein; BAR, Bin-Amphiphysin-Rsv domain; CRIB, Cdc42/Rac interacting binding region; I-BAR, inverse-Bin-Amphiphysin-Rsv domain; IMD, IRSp53-MIM homology domain; Mena, mammalian enabled; SH3, Src homology domain 3; GFP, green fluorescent protein; KO, knock-out; WT, wild type; RNAi, RNA interference; oligo, oligonucleotides; ROI, region of interest; FRET, fluorescence resonance energy transfer; GST, glutathione S-transferase; FCS, fluorescence correlation spectroscopy; DIC, differential interference contrast; mRFP, monomer red fluorescent protein; KD, knockdown. 3 For the I-BAR domain, the N-terminal 250 amino acid residues of IRSp53 have weak protein sequence similarity to the BAR domain (14,41). Structural analysis of the N terminus of IRSp53 shows that it dimerizes, forms a cigar shape, and has a stronger relationship to the BAR domains (41). Functionally, the BAR domain of IRSp53 deforms membranes with a curvature opposite that seen by other BAR domains (17). Here we designate this domain of IRSp53 the term Inverse-BAR or I-BAR.
phiphysins-R; also referred to as the IMD 4 [IRSp53 and Missing in metastasis homology Domain] see Refs. 13,14), a partial-CRIB 5 motif interrupted by an SH3-binding site, an SH3 domain (11,12), a potential WW domain binding site, and a PDZ domain binding site in some isoforms. The I-BAR domain has been suggested to play a role in F-actin bundling (14,15). It has also been suggested that the N-terminal region of IRSp53 incorporating the I-BAR domain is able to bind Rac1 directly (10). Recent work has suggested that the IRSp53 I-BAR domain is linked with membrane deformation/curvature of lipids in vitro but not F-actin bundling (16,17). Suetsugu et al. (18) have shown that that the IRSp53 I-BAR domain can produce membrane protrusion in cells. The partial-CRIB has high affinity for Cdc42, binds Rac1 extremely weakly, but does not bind RhoA (11,12). To date, the SH3 domain of IRSp53 has been shown to bind a number of proteins, including dentatorubral pallidoluysian, WAVE1 and -2, mDia1 (mouse Diaphanous1), Mena, Espin, Eps8, Pro/Shank, and bovine angiogenesis inhibitor (9, 12, 19 -24). IRSp53 interacts with the MALS (mammalian Lin7 homologue) protein through its PDZ domain (25). IRSp53 has recently been shown to bind the Rac1 exchange factor Tiam1 at a site near the partial-CRIB domain (26). IRSp53 has five splice variants, and two of them contain a functional actin monomer binding WH2 domain at the C terminus (27)(28)(29).
We identify N-WASP as an essential mediator of IRSp53induced filopodia formation. The SH3 domain of IRSp53 can bind N-WASP directly, and IRSp53 fails to induce filopodia in N-WASP KO fibroblasts (30) but does induce lamellipodia formation and membrane ruffling in these cells. In N-WASP reconstitution experiments of N-WASP KO fibroblasts, IRSp53 regains its ability to induce filopodia formation. Intriguingly, the WA domain or the Cdc42-binding site of N-WASP is not required to allow IRSp53 to induce filopodia formation, but the WA domain may play a role in filopodia turnover. IRSp53 also fails to induce filopodia in Mena/VASP KO cells, and N-WASP failed to induce filopodia when IRSp53 was knocked down with RNAi. The I-BAR domain alone induces dynamic membrane protrusions that lack actin and are smaller than normal filopodia ("partial-filopodia") in both N-WASP WT and N-WASP KO cells. We propose that IRSp53 generates filopodia by coupling membrane protrusion through its I-BAR domain with actin dynamics through proteins, including N-WASP and Mena via its SH3 domain.

Cell Culture and Transfection
N1E115 neuroblastoma cells were grown as described in Ref. 11. The N-WASP flox/flox control fibroblast (WT) and N-WASP del/del (KO) were derived from N-WASP flox/flox mice prepared on E14 cells and immortalized with retroviruses, resulting in the expression of a temperature-sensitive simian virus 40 large T-antigen. N-WASP del/del KO cell lines were then selected after transient expression of Cre, with the resulting lineage been comparable with their respective parental precursor lines (see Lommel et al. (30) for details of these cells). Mena WT (GMH3.0) and KO (MDV7) cells were cultured as described previously (31). Cells were transfected as described in Ref. 11.

Microinjection and Live Cell Imaging of N-WASP/Mena WT and KO Cells
Cells were plated out at approximately ϳ10 5 cells per glassbottom dish and grown overnight at 32°C in Dulbecco's modified Eagle's medium low with supplements. cDNA of required constructs were prepared at 50 ng/l in double distilled H 2 O and centrifuged at 16,000 ϫ g at 4°C for 30 min. 6 l of DNA mix was loaded into a microinjection needle, and cells were injected at a constant pressure of 20 p.s.i. for 100-ms duration. Microinjection was performed on a custom microinjection setup and Olympus microscope (IMT-10). Between 100 and 150 cells were injected per dish, and cells were left to express protein for 1-6 h before they were imaged or fixed and stained. For DIC/fluorescence time-lapse analysis, cells were incubated on a heated stage at 37°C and imaged with a monochromator on a Zeiss Axiovert 200 microscope enclosed in an incubator with CoolSNAP CCD camera. Generally, images were taken over a period of 10 min at 10-s intervals. The supplemental movies were compiled using the Metamorph software.

Knockdown of IRSp53 Protein by RNAi
The IRSp53 oligos and the negative control oligo were obtained from Invitrogen. The oligos were transfected using HiPerfect (Qiagen), according to the manufacturer's protocol. The oligos were transfected at a final concentration of 5 nM. Cells were harvested 27 and 48 h after transfection and lysed for Western blots to determine the level of IRSp53 knockdown. Oligo sequences were as follows: 5Ј AUG GUA AGC AGC AGA GUU CUU GGC C 3Ј, 5Ј AUU GCU AUU GGC CAU CUG UUG CAU C 3Ј, and 5Ј AUU CAU GAC AGG UAC AUU CUC CUG A 3Ј.

In Vitro Transcription/Translation and Binding Assay
N-WASP and GFP were in vitro transcribed and translated using the TNT T7-coupled reticulocyte lysate systems (Promega, L4610) with pXJ40-N-WASP-HA as the template for N-WASP and pXJ40-GFP as the template for GFP following the manufacturer's protocol. 4 IMD is an N-terminal 250-residue domain found in IRSp53 and Missing in Metastasis. 5 CRIB is a consensus amino acid sequence for a Cdc42 and Rac interacting binding site. Partial-CRIB lacks some of the consensus amino acid sequence.

IRSp53 and N-WASP Constructs
For mRFP-IRSp53, IRSp53 was subcloned from HA-IRSp53 in pXJ40 into mRFP-pXJ40 vector between the BamHI and NotI site. For mRFP-N-WASP, N-WASP was subcloned from HA-N-WASP in pXJ40 into mRFP-pXJ40 vector between the HindIII and NotI site. The 4K mutants were generated using a site-directed mutagenesis kit (Stratagene) as per the manufacturer's protocol. The mutants are as follows: (i) IRSp53-4K, (ii) GFP-I-BAR-4K, and (iii) GST-I-BAR-4K.

Mass Spectrometry Analysis
Proteins associated with the SH3 domain of IRSp53 were isolated by affinity purification from lysates of adult rat brain with the GST fusion protein of the SH3 domain of IRSp53 immobilized on Sepharose beads. The protein complex was eluted and resolved by 10% SDS-PAGE and detected by colloidal Coomassie Blue (Pierce). Protein bands detected by colloidal Coomassie Blue were excised and subjected to in-gel reduction, S-alkylation, and trypsin hydrolysis. Liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis of the peptides was performed on a Finnigan LCQ Deca ion trap mass spectrometer (Thermo Finnigan) fitted with a nanospray source (MDS Proteomics). Chromatographic separation was conducted using a Famos autosampler and an Ultimate gradient system (LC Packings) over Zorbax SB-C18 reverse phase resin (Agilent) packed into 75 m inner diameter PicoFrit columns (New Objective). Protein identifications were made using the search engines Mascot (Matrix Sciences) and Sonar (Pro-teoMetrics). 44 peptides were obtained for N-WASP, with 60% coverage of the protein.

Yeast Two-hybrid
Yeast two-hybrid analysis was done by mating AH109 and Y187 strains carrying the appropriate plasmids as described in the Clontech manuals.

FRET Measurement
FRET was measured by acceptor photobleaching method (32) by making necessary settings in a Zeiss LSM 510 confocal microscope with a C-Apochromat 63 ϫ 1-2-water objective. The fusion proteins of GFP/mRFP were excited using 488 and 561 nm laser line as excitation source, by selecting 405/488/561 dichroic mirror and 490,565 secondary dichroic mirrors for GFP and mRFP emission, respectively. The emission was monitored by selecting GFP (BP 505-550) and Red (LP 575) emission filters to record the fluorescence intensity. ROI was selected and photobleached using 70% of 561 nm laser power by selecting 50 iterations. The increase in GFP fluorescence intensity followed by mRFP bleaching was measured as FRET. FRET efficiency (%FE) was calculated using the change in background subtracted fluorescence intensity as 100 ϫ ((post-bleach intensity) Ϫ (pre-bleach intensity)/(post-bleach intensity)).
To verify that the change in intensity was not due to artifacts, we obtained the Pearson product moment correlation coefficient r, a dimensionless index that ranges from Ϫ1.0 to 1.0 inclusive and reflects the extent of a linear relationship between the two fluorescence intensity data of GFP and mRFP while bleaching. In our case we expect Ϫ1.0 as the perfect fitting of the linear relation because during acceptor bleaching donor intensity increases while acceptor intensity decreases. However we selected the range of Ϫ0.7 to Ϫ1.0 as the best range of index (see supplemental Fig. S1 for details). The GFP/mRFP pair has been used previously to measure FRET (33).

Statistical Analysis of Filopodia, Lamellipodia/Membrane Ruffling, and Neurite-like Processes
Morphological phenotypes were quantitated using the following definitions.
Filopodia-Protrusions that contain actin are dynamic, with a width of ϳ0.6 -1.2 m and an average length between 6 and 15 m. The number of filopodia per cell was determined.
Lamellipodia/Membrane Ruffling-Lamellipodia/membrane Ruffling are areas of cell flattening/wavy membrane thickening. Each cell was divided into eight sectors, and each sector was assessed for the presence of lamellipodia or membrane ruffling. Each sector contributes a maximum of 12.5% morphological activity. The eight sector values for each cell were then combined to give % lamellipodia/membrane ruffle per cell.
Neurite-like Processes-In CHO-1 cells, N-WASP (WT) fibroblasts, and Mena WT fibroblasts, we observed that IRSp53 induced the formation of cell extensions that resemble neurites. Neurite-like processes are defined as cell extensions greater than two cell body lengths. This definition in no way attributes any functional value to these structures and is purely a reflection of morphological similarity to neurites. Filopodia lifetimes were determined by following individual filopodia from appearance from the cell membrane to disappearance. Filopodia assembly is defined as the time taken from appearance to maximum length. Filopodia disassembly is defined as the time taken from maximum length to disappearance.
For each experiment ϳ6 -12 cells were evaluated for filopodia, neurites/neurite-like structures, and lamellipodia/ membrane ruffles. At least three independent experiments were carried out for any one set of conditions giving an n value of ϳ36. Values presented in bar charts and tables represent mean Ϯ S.D.

F-actin Bundling Assays
F-actin bundling was monitored by a sedimentation assay and by visualization of F-actin bundles with fluorescence microscopy as described (14,15). Fluorescence correlation spectroscopy (FCS) was used to measure GFP-IRSp53 concentrations in vivo. For FCS analyses, the fluorescence intensity fluctuations arise from single molecules that are diffusing in and out of a defined confocal volume. Details of the instrumentation and method can be found in Ref. 34.

RESULTS
Filopodia are a ubiquitous but diverse group of cell structures making them difficult to define (35). However, in mammalian cells filopodia are more homogeneous. In studies examining the ability of the IRSp53 or its I-BAR domain to induce filopodia in mammalian cells, filopodia have not been defined (14,15,36). In addition, in most studies on filopodia measurement of filopodial dynamics have not been done, and this makes interpretation and comparison of the data difficult. Thus we felt it important to define mammalian filopodia at the outset and determine which domains of IRSp53 were essential for this activity using dynamic measurements.
Definition of Filopodia-To define filopodia more accurately, we measured their formation in a range of mammalian cells. N1E115, HeLa, COS7, and B16F1 cells were transfected with GFP-actin, and by using sequential time-lapse wide field fluorescence/DIC imaging, we scored the protrusions for length and lifetime. Mammalian cell line filopodia are actin-based structures with lengths between 8 and 15 m and lifetimes between 79 and 142 s ( Table 1). In addition, we noticed the following features of filopodia: (i) they have a width between 0.6 and 1.2 m (see Table 1 legend); (ii) are unbranched; and (iii) rarely protrude together. In contrast to filopodia, retraction fibers are nondynamic, tapered, sometimes branched, and can protrude as clusters. Thus for structures to be designated bona fide filopodia in mammalian cells, they must be dynamic actin-based structures with characteristics similar to those presented in Table 1.

IRSp53-SH3 and I-BAR Domains Are Required for Filopodia Formation-IRSp53 induces filopodia in a Cdc42-dependent manner as a Cdc42-binding mutant
IRSp531267N is unable to induce these structures. Induction of filopodia is also significantly reduced in an IRSp53 mutant deleted of its C-terminal domain comprising the SH3 domain (11). The C terminus of IRSp53 contains at least three protein binding domains, the SH3 domain, a potential WW domain binding motif, and a PDZ domain binding motif that could be responsible for filopodia formation (Fig. 1).
To determine whether the SH3 domain was responsible for the filopodia formation, we compared the phenotypes of wild-type protein with that of IRSp53-FP/AA (an SH3 mutant defective in binding to ligands) using GFP-actin and timelapse microscopy. Filopodia were scored on the definition outlined above. Filopodia formation was absent in the IRSp53-FP/AA mutant (Fig. 1).
The I-BAR domain of IRSp53 has been reported to bundle F-actin, and this activity is reduced by mutation of four lysine residues (142, 143, 146, and 147) that form a potential actin-binding site (14). Using time-lapse analysis of GFP-

TABLE 1 Filopodia characteristics of mammalian cells
Cells were transfected with GFP-actin and left for 18 -24 h to allow for expression. GFP-actin-expressing cells were picked randomly, and morphological activity was then followed by sequential time-lapse DIC and fluorescence imaging. Filopodia were followed for 10 min at 6 frames/min. Filopodia scored positive for GFP-actin. Filopodia width was measured using GFP-actin-transfected cells. Measurements were taken at the base of the filopodia. The average width for the cells examined ranged between 0.6 and 1.2 M and was affected by growth conditions. We attempted to measure filopodial characteristics of CHO-1 cells but failed to detect any endogenous filopodia formation and so were unable to make these measurements. All measurements are presented as an average Ϯ S.D., n ϭ 15 per experiment, from three experiments.

Cell line (endogenous filopodia)
Length ؎S.D. Lifetime ؎S.D. Complexes-Because the SH3 domain is essential for filopodia formation, we wanted to identify novel binding partners and thus carried out an IRSp53 GST SH3 domain affinity purification of interacting proteins from brain lysates followed by mass spectrometry analysis of the peptides to identify interactors. Interestingly, WAVE1 and the WAVE1 complex proteins, Abi-1/2b, p125 Nap-1, and p140 Sra-1, were present on the GST-SH3 column (Fig. 2). Actin, tubulin, dynamin, and mDia1 were also present on the column. In similar experiments using T-cell lysates instead of brain, we detected Mena, mDia2, and WAVE2 as found previously (9, 12 and data not shown). In addition to the WAVE1/2 complex, we found N-WASP and CR-16 binding to the affinity column (Fig. 2). From the proteins that bound to the IRSp53 SH3 domain column, N-WASP was unique in that it had been linked to filopodia formation previously (37,38) but not with IRSp53. Thus we decided to investigate the relationship between IRSp53 and N-WASP in filopodia formation further.
IRSp53 Interacts with N-WASP Directly-The presence of N-WASP and CR-16 on SH3 domain affinity columns suggested that IRSp53 might induce filopodia formation by direct interaction with N-WASP. To investigate this, we used in vitro transcription/translation to produce N-WASP labeled with 35 S-labeled methionine and GST-IRSp53-SH3 in pulldown experiments. GST-Cdc42 binding was used as positive control (Fig. 3A, lane 3) and GST-IRSp53-SH3/ 35 S-GFP as a negative control (Fig. 3A, lane 5). We show that the SH3 domain of IRSp53 interacted with N-WASP specifically (Fig. 3A, lane 4), and the FP/AA mutant failed to interact (Fig. 3A, lane 6).
The IRSp53 SH3 domain interaction with N-WASP was also analyzed by using the yeast two-hybrid system. The SH3 domain was cloned into the bait vector (Clontech, system 3) and mated with a strain carrying the N-WASP cDNA cloned in the prey vector or with a strain carrying an empty prey vector (pACT2) as a control. IRSp53-N-WASP were found to interact as diploids grew on quadruple dropout plates and possessed significant ␤-galactosidase activity (Fig. 3B, panels i and ii). The IRSp53 FP/AA fails mutant to interact with N-WASP in similar yeast two-hybrid experiments (Fig. 3B, panels iii and iv).
To determine whether the IRSp53-N-WASP interaction occurred in vivo we used a FRET approach to measure proteinprotein interactions (Fig. 3C). mRFP-IRSp53 and GFP-N-WASP were cotransfected into N1E115 and CHO-1 cells and allowed to express for 36 h (to allow proper folding of the fluorescent moieties). We used the acceptor photobleaching method to determine FRET. Briefly, FRET was measured in the following way. ROI were chosen, and mRFP-IRSp53 (acceptor) was bleached, and changes in GFP-N-WASP (donor) and acceptor fluorescence were measured. If acceptor bleaching induces an increase in donor fluorescence, FRET is occurring and can be quantitated as a percentage efficiency, %FE (see under "Materials and Methods"). The %FE is a measure of the distance between mRFP-IRSp53 and GFP-N-WASP. For FRET to occur the distance between donor and acceptor has to be 10 nm or less. We also reasoned that if FRET is present then there should be a negative correlation between rates of change of acceptor/donor fluorescence. This we define with a correlation coefficient (CC; see under "Materials and Methods" and supplemental Fig. S1 for details). We used four controls in our FRET experiments; cytoplasmic GFP and mRFP, mRFP-IRSp53 and GFP, mRFP and GFP-N-WASP, and a tandem GFP-mRFP.
Positive FRET values reached a maximum of 28% with CC ϭ Ϫ0.99, whereas background %FE varied between 0.96 and 2.12% with CC ϭ 0.17 to Ϫ0.34. Any FRET efficiency above 3% with CC ϭ Ϫ1.0 to Ϫ0.7 is defined as positive FRET and indicates that protein-protein interaction was occurring.
In both CHO-1 cells and N1E115 cells, we observed positive FRET between IRSp53 and N-WASP in filopodia-like structures, neurites, neurite-like processes, and the cell body.  we compared the effect of IRSp53 cDNA microinjection on the morphology of the N-WASP WT and KO cell lines. cDNA for GFP-actin was included in the microinjection to identify expressing cells and to facilitate the imaging of actin dynamics (Fig. 4). When cells were injected with GFP-actin cDNA alone, and scored for filopodia, neurite-like processes, and membrane ruffling, there was no difference between N-WASP WT and N-WASP KO cells (Fig. 4A, panels a and b). For the basal morphological activity of N-WASP WT and KO cells, see supplemental Fig. S2A. IRSp53 induced filopodia formation in the N-WASP WT cells and caused the formation of neurite-like processes (Fig. 4A, panels d, e, and f; see supplemental Movie 1). In N-WASP KO cells no filopodia or neurite-like processes were seen. However, membrane ruffling (and lamellipodia) formation was strongly stimulated when IRSp53 was expressed in N-WASP KO cells (Fig. 4A, panels c and g; see supplemental Movie 2). N-WASP KO cells expressing IRSp53 appeared to be thicker than WT cells possibly because of dorsal ruffling (Fig.  4A, panel c) Expression levels of IRSp53 in both N-WASP WT and KO cells were similar (Fig. 4A, panel h).
IRSp53 induced membrane ruffling and lamellipodia formation in N-WASP KO cells. To rule out the possibility that these morphological activities were masking filopodia formation, we used Rac1N17. 6 The presence of Rac1N17 with IRSp53 in N-WASP WT cells increased the observed filopodia formation (Fig. 5A, compare panels a and c). However, Rac1N17 did not affect the ability of IRSp53 to induce filopodia formation in N-WASP KO cells (Fig. 5A, compare panels b and d), although membrane ruffling/lamellipodia formation were reduced significantly (Fig. 5A, panel f).  JULY 18, 2008 • VOLUME 283 • NUMBER 29

JOURNAL OF BIOLOGICAL CHEMISTRY 20459
Cdc42/Rac1N17 Fails to Induce Filopodia Formation in N-WASP KO Cells-Cdc42 has been reported to induce filopodia formation in N-WASP KO cells (30) using Cdc42L61/ Rac1N17/C3 toxin microinjections. In our hands, cells retracted and died when injected with Cdc42L61/Rac1N17/C3 toxin (possibly through collapse of significant numbers of focal adhesions/complexes; data not shown). So we re-examined the ability of Cdc42 to induce filopodia formation in N-WASP KO cells. Cdc42V12 7 alone was found to induce membrane ruffling/lamellipodia formation as the main phenotype, and few filopodia were seen in N-WASP WT cells and none in KO cells (Fig. 5B, compare panels a and b).
When Cdc42V12 was microinjected with Rac1N17, 6 membrane ruffling/lamellipodia formation was strongly inhibited in both cell types (Fig. 5B, compare panels a and c,  and b and d) and filopodia were induced in N-WASP WT cells but not in N-WASP KO cells. N-WASP was able to reconstitute Cdc42V12/Rac1N17-induced filopodia formation in KO cells showing that the N-WASP KO cells were competent to generate filopodia (Fig. 5B, compare panels e and f; statistics are presented in panels g-j). Expression of N-WASP alone induced membrane ruffling in KO cells (Fig.  5B, panel j). Taken together, these results suggest that N-WASP is essential for IRSp53-and Cdc42-induced filopodia formation.
Mena/VASP KO Cells and IRSp53 Knockdown (KD)-Mena has been linked to IRSp53-mediated filopodia formation (12). To examine this further, we used Mena/VASP KO cells. IRSp53 was unable to induce filopodia formation in Mena/VASP KO 7 Cdc42 Q61L and G12V are mutant forms of the protein that have low intrinsic GTPase activity such that the protein remains in the GTP-bound form. The intrinsic GTPase activity Q61L is slightly lower than G12V; however, both are in the "on" state and have similar phenotypes and can be used interchangeably. Thus both N-WASP and Mena are important for IRSp53-mediated filopodia formation but not for membrane ruffling/lamellipodia formation (Fig. 6, A and B). We also determined the effect of IRSp53 KD on the N-WASP-driven filopodia formation. As shown in Fig. 6, D and E, IRSp53 KD significantly inhibited N-WASP driven filopodia formation.  with IRSp53 cDNA. Low levels of N-WASP cDNA were tolerated by cells and allowed us to carry out reconstitution experiments with IRSp53. N-WASP KO cells microinjected with N-WASP cDNA responded to IRSp53 by filopodia formation (Fig. 7A, compare panels a and b; see supplemental Movies 3 and 4), and neurite-like processes were also formed (data not shown). IRSp53-FP/AA was not competent to generate filopodia in N-WASP reconstitution experiments (Fig. 7A, panels d and e) suggesting that IRSp53 interaction with N-WASP is essential for filopodia formation in these reconstitution experiments (controls determining the morphological activity of IRSp53FP/AA alone in N-WASP WT and KO cells are shown in supplemental Fig. S2B).

Reconstitution of N-WASP in N-WASP KO Cells-In
Finally, we were able to demonstrate positive FRET between mRFP-IRSp53 and GFP-N-WASP in KO cells (data not shown) similar to that seen in N1E115 and CHO-1 cells (Fig. 3C).
N-WASP⌬WA, a mutant unable to interact with the Arp2/3 complex, was used next in reconstitution experiments. The IRSp53-N-WASP⌬WA combination was able to reconstitute filopodia formation in KO cells, but neurite-like processes were not observed. Reconstitution with the N-WASP⌬WA mutant induced filopodia on the dorsal surface of the KO cells as well as on the periphery (Fig. 7A, panel c; see supplemental Movie 5).
Using the N-WASPH208D mutant in the reconstitution system, we examined the role of Cdc42-N-WASP interaction. The   N-WASPH208D mutant was able to reconstitute filopodia formation driven by IRSp53 (Fig. 7A, panels d and e).

IRSp53 Couples Membrane Protrusion and Actin Dynamics
Function of the IRSp53-N-WASP Interaction-Next, we compared filopodia induced by IRSp53/N-WASP with IRSp53/ NWASP⌬WA in reconstitution experiments for length and lifetime (a control determining the morphological activity of N-WASP⌬WA alone in N-WASP KO cells is shown in supplemental Fig. S2C). The filopodia induced in the two situations differed significantly only in lifetime. The effect of the WA deletion was to increase the lifetime of the filopodia from 60 to 160 s (Fig. 7B), which was caused by an increase in time taken for disassembly.
IRSp53 Interacts with Mena and Eps8 in Filopodia-Unlike Cdc42, IRSp53, or N-WASP (Fig. 1A), overexpression of Mena or Eps8 fails to induce filopodia (data not shown). Both GFP-Mena and GFP-Eps8 were located primarily in distinct vesicular structures, and filopodial localization was not observed (data not shown). When GFP-Mena and GFP-Eps8 were transfected with mRFP-IRSp53, their distribution was changed, and they could be observed in filopodia. To determine whether IRSp53 interacted directly with Mena and Eps8 and thereby recruit these proteins to filopodia, we used the acceptor photobleaching FRET approach as described above. Fig. 8 shows examples of individual filopodia where positive FRET signals between GFP-Mena or GFP-Eps8 and mRFP-IRSp53 were found. In these experiments Eps8 was found throughout the filopodia, whereas Mena had a preference for the tip complex (Fig. 8).
We also examined the role of the IRSp53-mDia2 interaction in filopodia formation. Expression of mRFP-IRSp53 with either myc-mDia2 or YFP-mDia2 led to inhibition of the IRSp53-mediated filopodia formation (data not shown). In most cells there were no filopodia. IRSp53 and mDia2 colocalized in the perinuclear area.
Role of the IRSp53 I-BAR Domain in Filopodia Formation-Recent studies have suggested that the I-BAR domain has F-actin bundling activity, and the role of IRSp53 in filopodia is to bundle F-actin (14,15,36). We compared the F-actin bundling activity of IRSp53 with that of Fascin using low speed sedimentation assays as first described for IRSp53 in Ref. 15. Fascin was able to bundle F-actin effectively at 60 nM. With the IRSp53 I-BAR domain we obtained variable results, but at least 5-10 M was required to see any F-actin bundling (supplemental Fig.  S3). We then assayed the cellular concentration at which GFP-IRSp53 was able to generate filopodia formation using FCS (see under "Materials and Methods"). Cellular concentrations of the IRSp53 ranged from 29.7 to 453 nM (supplemental Fig. S4), with 29.7 nM being sufficient to generate filopodia.
Next we overexpressed the GFP-I-BAR domain with mRFPactin (Flag-tagged-I-BAR and GFP-actin were also used in , and cells were incubated for up to 6 h for expression. GFP-actin-positive cells were imaged using DIC time-lapse microscopy as described under "Materials and Methods." Filopodia generated by either treatment were followed during formation and disassembly, and the time taken for both processes was monitored. The time-lapse series shown gives an example of one such measurement. Filopodia were measured for length (panel c), lifetime (panel d), and duration (panel e). Blue bars represent assembly, and green bars represent disassembly (panel e). Bar ϭ 5 m. Lifetime, determined by following individual filopodia from appearance from the cell membrane to disappearance. Duration, filopodia assembly is defined as the time taken from appearance to maximum length. Filopodia, disassembly is defined as the time taken from maximum length to disappearance. All measurements are presented as an average Ϯ S.D., n ϭ 7 per experiment, from three experiments. some experiments) in cells and measured morphological activity by using time-lapse microscopy. At high levels of I-BAR expression static protrusions which contain actin were seen (these are similar to the protrusions observed by Yamagishi et al. (15) and Millard et al. (14)). Furthermore, the morphology of these static I-BAR domain-induced protrusions is distinct from filopodia; they were clustered and nonuniform (Fig. 9A,  panel a). At lower levels of I-BAR expression, we observed membrane protrusions, but these were not associated with actin ( Fig. 9, panel b). Some of these membrane protrusions lacking actin were dynamic, like filopodia, but were shorter and thinner than filopodia (Fig. 9, panel c). For comparison, the characteristics of IRSp53-induced filopodia is presented in Fig.  9A, panel d. We term the I-BAR domain-induced dynamic membrane protrusions lacking actin partial-filopodia. As expected, the structures induced by the I-BAR domain were independent of the presence on N-WASP. Interestingly, the I-BAR-4K mutant induced partial-filopodia, but not the static actin-containing protrusions, again independent of the presence of N-WASP (Fig. 9C). Thus the four lysine residues, 142, 143, 146, and 147, of the I-BAR may play a role in linking membrane protrusion with actin association, but their role in filopodia formation is unclear.
We attempted colocalization of IRSp53 and BAR with actin using Metamorph software for the calculation of correlation coefficients for colocalization (CC* ; Fig. 9D). The actin images of the IRSp53 induced filopodia gave uniform signal along their length. In contrast, the static I-BAR-induced structures that contained actin were nonuniform, and actin appeared as aggregates (Fig. 9D, panel a). IRSp53 colocalized with actin in filopodia with CC* of 0.86 Ϯ 0.064 (Fig. 9D, panel c). Dynamic structures induced by I-BAR (seen with GFP-labeling of I-BAR) were devoid of actin (Fig. 9D, panel b; see supplemental Movies 6 and 7), and this was reflected in the low CC* value of 0.24 Ϯ 0.2. Taken together, these data show that I-BAR can induce dynamic membrane protrusions, but they do not contain actin. The characteristics of Cdc42, IRSp53, and N-WASP induced filopodia are presented in Table 1 for comparison with the I-BAR-induced protrusions (Fig. 9). We conclude that the I-BAR domain does not induce bona fide filopodia formation.
IRSp53 Binds Actin-To further characterize the differences between IRSp53 and its I-BAR domain, we attempted to measure FRET between GFP-actin and mRFP-IRSp53 as well as between mRFP-actin and GFP-I-BAR and GFP-I-BAR-4K. There was positive FRET between GFP-actin and mRFP-IRSp53 in filopodia-like structures, ribs, neurite-like processes, and areas of ruffling (Fig. 10). However, both I-BAR and I-BAR-4K failed to show positive FRET with actin. These data suggest that the I-BAR domain does not interact with actin in the same way as IRSp53 (Fig. 10).

DISCUSSION
Definition of Filopodia-Filopodia are F-actin-based morphological structures at the periphery of cells. In particular, neurons have prominent filopodia in their growth cones, and these structures are thought to help axons find their targets (39). Filopodia are constructed from parallel bundles of F-actin that lie perpendicular to the cell periphery. The length of endogenous filopodia in the mammalian cells examined in this study varied between 8 and 15 m. These filopodia are highly dynamic with lifetimes between 79 and 142 s. Mammalian cells form a number of structures that resemble filopodia. For example, retraction fibers are F-actin-based structures present at the cell periphery and have similar overall dimensions to filopodia. However, retraction fibers are static structures that arise from the cell membrane withdrawing from the leading edge. Thus, it is essential to use time-lapse analysis with GFP-actin to distinguish filopodia from other structures at the cell periphery.
Cdc42 Pathways to Filopodia Formation-N-WASP was the first Cdc42-interacting protein implicated in filopodia formation (37). Expression of dominant negative N-WASP constructs or microinjection of N-WASP antibodies into bradykinin-treated cells inhibited filopodia formation. Subsequently,   (38) were unable to discriminate between retraction fibers and filopodia, and thus it is difficult to comment on the basal filopodial activity observed in N-WASP KO cells in this study. However, Snapper et al. (38) did find substantial reduction in "filopodia" numbers in N-WASP KO cells, which led them to conclude that their data "support earlier studies (37) implicating a crucial role for N-WASP in the induction of filopodia by Cdc42." In contrast, Lommel et al. (30) found the ability of Cdc42 to generate filopodia was unaffected by N-WASP KO. The data presented in our study taken together with previous work (37,38) suggest an important role for N-WASP in filopodia formation. The use by Lommel et al. (30) of the mixture of Cdc42L61/RacN17/C3 toxin to investi-gate filopodia formation represents one major difference in the studies. C3 toxin may induce the dissociation of a RhoA-mDia2 complex allowing mDia2 to work with Cdc42 (40) in the absence of N-WASP. Another potential discrepancy between the studies is the definition of filopodia. We believe it is crucial that bona fide filopodia have the characteristics described in Table 1.
IRSp53 SH3 Domain Binding Partners-The SH3 domain of IRSp53 is clearly important in the morphological activities of this protein. The IRSp53 SH3 domain mutant IRSp53-FP/AA does not bind to its ligands or induce filopodia in both N1E115 and N-WASP WT fibroblasts. This extends the initial observations made by Krugmann et al. (12) that the IRSp53 SH3 domain is important for filopodia formation in Swiss 3T3 cells possibly through interaction with Mena. We searched for imaged for 10 min with frames taken every 10 s. Static protrusions i and ii did not turn over during the 10 min and are scored as having a lifetime of more than 10 min. Dynamic protrusions iii and iv are those that are appearing/disappearing over the 10 min. Lifetime is determined as the time taken from appearance to disappearance. F-actin was followed with mRFP signal. All measurements are presented as an average Ϯ S.D., n ϭ 7 per experiment, from three experiments. C, N-WASP WT and KO cells were microinjected with GFP-I-BAR or GFP-I-BAR-4K with mRFP-actin. Cells were left for 1-6 h for cDNA expression before live cell imaging was carried out. All measurements are presented as an average Ϯ S.D., n ϭ 7 per experiment, from three experiments. D, cells were transfected with either mRFP-actin/GFP-I-BAR or GFP-actin/mRFP-IRSp53. GFP and RFP channels were followed sequentially as described under "Materials and Methods." Images were inverted using Adobe Photoshop to allow visualization of signals.  IRSp53 SH3 domain binding partners and identified N-WASP as a novel interactor. Yeast two-hybrid and pulldown experiments in vitro demonstrated that IRSp53 can interact directly with N-WASP. Furthermore, FRET experiments show that IRSp53 and N-WASP interact in vivo and that, using the IRSp53-FP/AA mutant, the IRSp53 SH3 domain is essential for this interaction.
IRSp53 Phenotypes in N-WASP and Mena KO Cells-To investigate the role of the IRSp53-N-WASP and IRSp53-Mena interactions, we employed the use of KO fibroblasts and RNAi mediated KD. Overexpression of IRSp53 in N-WASP and Mena WT fibroblasts induced strong filopodia formation. However, in N-WASP KO and Mena/VASP KO cells, overexpression of IRSp53 induced membrane ruffling but not filopodia. Furthermore, N-WASP-induced filopodia formation was significantly reduced by IRSp53 KD. We conclude that both N-WASP and Mena are essential for IRSp53-mediated filopodia formation.
We also noticed that IRSp53 induced significant stimulation of lamellipodia/membrane ruffling in N-WASP and Mena KO. We suspect that IRSp53-SH3-WAVE2 and IRSp53-Tiam1 (Tiam1 is a Rac1 exchange factor) interactions are responsible for this phenotype. In support of this, both the FP/AA mutation and Rac1N17 reduced the ability of IRSp53 to stimulate lamellipodia/membrane ruffling. Thus the IRSp53 phenotype and function may be determined by competition between targetinteracting proteins.
Because the IRSp53-N-WASP interaction has not been described before we investigated the role of N-WASP further, IRSp53 in combination with Rac1N17 revealed even stronger filopodia formation in N-WASP WT cells but still none in N-WASP KO cells. If N-WASP cDNA is included with IRSp53 in the injection/transfection of N-WASP KO cells, reconstitution of filopodia formation is observed. The IRSp53 SH3 domain is essential for these effects as the FP/AA mutation did not generate filopodia in N-WASP WT cells and N-WASP could not reconstitute filopodia formation with this mutant in N-WASP KO cells.
To extend this analysis we carried out the reconstitution with the mutant, N-WASP⌬WA, which is unable to bind the Arp2/3 complex. N-WASP⌬WA was fully competent to reconstitute filopodia formation. However, the filopodia observed in the N-WASP⌬WA reconstitution experiments had a longer lifetime because of a slower disassembly phase. These results clearly show that the role of N-WASP in IRSp53-mediated filopodia formation is not to activate actin nucleation via the Arp2/3 complex. The N-WASPH208D mutant was able to reconstitute filopodia formation, and thus it is unlikely that the Cdc42-N-WASP interaction is important for IRSp53-mediated filopodia formation. So what could be the role of IRSp53-N-WASP interaction? Three possible roles for IRSp53-N-WASP interaction in filopodia formation could be determined as follows. (i) N-WASP induces a conformational change in IRSp53 allowing activation of the membrane deformation activity of protein. (ii) IRSp53 could sequester N-WASP, reducing actin branching via the Arp2/3, and thereby promoting actin filament elongation. (iii) N-WASP is localized to vesicles and is implicated in promoting their movement. The IRSp53-N-WASP interaction may promote vesicle recycling, which could FIGURE 11. Model of IRSp53 mediated filopodia formation. A, IRSp53 I-BAR domain generates three distinct membrane protrusions (type 1-3) but not filopodia. Only full-length IRSp53 generates filopodia with uniform actin and with similar characteristics (length, width, and lifetime) to filopodia generated by Cdc42 and N-WASP. The use of GFP-actin and time-lapse microscopy is essential to determine whether bona fide filopodia are being induced. B, step 1: Cdc42-GTP recruits IRSp53 to the plasma membrane. We have shown previously that a mutant IRSp53 that is unable to bind Cdc42 fails to localize with actin and does not generate filopodia (11). Disanza et al. (36) have shown biochemically that Cdc42 can recruit IRSp53 to the plasma membrane.
Step 2: IRSp53 normally exists in a closed conformation where the SH3 domain is masked. Krugmann et al. (12) have shown that C-terminal of IRSp53 binds N-terminal. IRSp53 needs to adopt an open conformation, but the mechanism for this is unclear. One possibility is that N-WASP could promote an open IRSp53 conformation. Further work is necessary to address this issue.
Step 3: recruitment by the IRSp53 SH3 domain of N-WASP, Mena and Eps8 allows coupling of I-BAR mediated membrane protrusion with actin dynamics and subsequent generation of filopodia formation. be important for delivering proteins to the membrane for filopodia formation. Further work will be necessary to understand the contribution played N-WASP in IRSp53-and Cdc42-mediated filopodia formation.
BAR Domain Function-The role of BAR domain proteins in membrane trafficking has been recognized for some time (41). In these proteins the BAR domain induces membrane curvature, and this is coupled to actin dynamics via proteins such as dynamin to facilitate membrane vesicle formation and endocytosis. Members of the BAR domain family also include proteins Toca-1, CIP4, and FBP17 that have a similar overall structure to IRSp53, namely a BAR domain linked to a Cdc42-binding site and an SH3 domain. In the case of Toca-1 it is clear that its SH3 domain binds N-WASP, and Toca-1 can regulate N-WASP actin polymerization activity in vitro. Thus the domain structure of these proteins (Toca-1, IRSp53, CIP4, and FBP17) could allow the coupling of the membrane curvature with actin dynamics under the control of Cdc42. Interestingly, structural and functional studies of the IRSp53 I-BAR domain suggest it would induce membrane curvature opposite that of proteins such as amphiphysin leading to membrane protrusion rather than membrane invagination (16,42) and hence the designation I (Inverse)-BAR domain. Here, we show that the IRSp53 I-BAR domain induces dynamic membrane protrusions that lack actin in live mammalian cells (see Fig. 11A).
Function of the I-BAR Domain in Filopodia Formation-Yamagishi et al. (15) and Millard et al. (14) have suggested that the I-BAR domain of IRSp53 is sufficient for filopodia formation. Our experiments fail to detect bona fide filopodia formation with the I-BAR domain. At high levels of I-BAR, we detect static actinbased protrusions that do not possess typical filopodial morphology. At lower expression levels, the I-BAR domain induces membrane protrusions, some of which are dynamic, but which do not contain actin. These membrane protrusions are smaller and thinner than endogenous or IRSp53-induced filopodia. We term these I-BAR domain induced structures as partial-filopodia. The 4K mutation in the I-BAR domain eliminates the induction of static actin-based structures but does not affect the membrane protrusions lacking actin. Thus, the 4K mutation allows us to dissociate the I-BAR domain-induced membrane protrusions from changes in actin. The I-BAR domain-induced structures were similar in both N-WASP WT and N-WASP KO cells suggesting that N-WASP does not play a role in the induction of these structures. Further work is necessary to clarify the physiological link between actin and membrane protrusion for the I-BAR domain.
Mechanism of IRSp53-mediated Filopodia Formation-IRSp53 KD has recently been reported to block Cdc42-induced filopodia formation (36) validating the importance of IRSp53 in this pathway. We have shown here that N-WASP-driven filopodia formation requires IRSp53. A number of studies (14,15,36) suggest that the apparent F-actin bundling activity of IRSp53 (possibly with Eps8) is sufficient to induce filopodia formation. In parallel assays we found the IRSp53 I-BAR domain has F-actin bundling activity ϳ83-fold weaker than Fascin. In vitro assays require at least 5 M I-BAR domain to see any F-actin bundling compared with 60 nM for Fascin. We estimate by FCS that cellular IRSp53 concentrations of 30 nM are sufficient to generate filopodia. Furthermore, our FRET analy-sis suggests that the I-BAR domain does not bind actin in the static actin-based protrusions, whereas full-length IRSp53 does bind actin in filopodia-like structures. In line with our findings, two recent reports (16,17) fail to see F-actin bundling by the IRSp53 I-BAR domain (named IMD in these studies) at physiological salt concentrations. Thus the I-BAR domain-mediated F-actin bundling is unlikely to be the main mechanism used by IRSp53 to generate filopodia.
The I-BAR domain was unable to induce bona fide filopodia as determined by time-lapse microscopy with GFP-actin or mRFP-actin. However, the I-BAR domain and the I-BAR-4K mutant did generate dynamic membrane protrusions that lacked actin and were of a smaller size than filopodia. Recent studies using lipid vesicles in vitro demonstrate that the I-BAR domain of IRSp53 can generate membrane deformation in the form of buds independent of F-actin (18). Taken together, these results suggest that the prime function of the I-BAR domain is to induce membrane protrusion.
In conclusion, our data suggest a mechanism for filopodia formation; Cdc42 helps to localize IRSp53 through the partial-CRIB domain (see Fig. 11). Subsequently, IRSp53 mediates the coupling of I-BAR-mediated membrane protrusion with actin dynamics via its SH3 domain binding partners N-WASP and Mena. Other SH3 domain binding proteins such Eps8 (36) are also likely to be involved in filopodia formation. Work is currently underway to define the exact molecular roles played by N-WASP, Mena, and Eps8 8 in actin dynamics (F-actin microfilament assembly and disassembly) during IRSp53-mediated filopodia formation.