Regulation of IRSp53-Dependent Filopodial Dynamics by Antagonism between 14-3-3 Binding and SH3-Mediated Localization

Abstract

Filopodia are dynamic structures found at the leading edges of most migrating cells. IRSp53 plays a role in filopodium dynamics by coupling actin elongation with membrane protrusion. IRSp53 is a Cdc42 effector protein that contains an N-terminal inverse-BAR (Bin-amphipysin-Rvs) domain (IRSp53/MIM homology domain [IMD]) and an internal SH3 domain that associates with actin regulatory proteins, including Eps8. We demonstrate that the SH3 domain functions to localize IRSp53 to lamellipodia and that IRSp53 mutated in its SH3 domain fails to induce filopodia. Through SH3 domain-swapping experiments, we show that the related IRTKS SH3 domain is not functional in lamellipodial localization. IRSp53 binds to 14-3-3 after phosphorylation in a region that lies between the CRIB and SH3 domains. This association inhibits binding of the IRSp53 SH3 domain to proteins such as WAVE2 and Eps8 and also prevents Cdc42-GTP interaction. The antagonism is achieved by phosphorylation of two related 14-3-3 binding sites at T340 and T360. In the absence of phosphorylation at these sites, filopodium lifetimes in cells expressing exogenous IRSp53 are extended. Our work does not conform to current views that the inverse-BAR domain or Cdc42 controls IRSp53 localization but provides an alternative model of how IRSp53 is recruited (and released) to carry out its functions at lamellipodia and filopodia.

The ability of a cell to rapidly respond to extracellular cues and direct cytoskeletal rearrangements is dependent on an array of signaling complexes that control actin assembly (58). The protrusive structures at the leading edges of motile cells are broadly defined as lamellipodia or filopodia (14). Lamellae are sheet-like protrusions composed of dendritic actin arrays that drive membrane expansion, with the “lamellipodium” representing a narrow region at the edge of the cell (in culture) characterized by rapid actin polymerization. This F-actin assembly is suggested to require Arp2/3 activity that nucleates new actin filaments from the sides of existing ones (58, 71) and capping proteins that limit the length of these new filaments and stabilize them (7). Arp2/3 activity in turn is regulated by the WASP/WAVE family of proteins, such as N-WASP and WAVE2 (68), whose regulation is a subject of intense interest (12, 29, 36, 41, 56, 76).

Filopodia contain parallel bundles of actin filaments containing fascin (22). These are dynamic structures that emanate from the periphery of the cell and are retracted, with occasional attachment (to the dish in culture). Thus, they have been thought to have a sensory or exploratory role during cell migration (28). This is the case for neuronal growth cones, where filopodia sense attractant or repulsive cues and dictate direction in axonal path finding (9, 17, 25, 35). Filopodia have been shown to be important in the context of dendritic-spine development (64, 77), epithelial-sheet closure (26, 60, 79), and cell invasion/metastasis (80, 83).

Lamellipodia have been well characterized since the pioneering work of Abercrombie et al. in the early 1970s (2, 3, 4). Filopodia require symmetry breaking at the leading edge (initiation), followed by elongation driven by a filopodial-tip protein complex (14, 28). A few proteins have been identified in this complex; Mena/Vasp serve to prevent capping at the barbed ends of bundled actin filaments (7, 53), and Dia2 promotes F-actin elongation (57, 85). Termination of filopodial elongation is not understood but nonetheless is likely to be tightly regulated. In the absence of F-actin elongation, retraction of the filopodium takes place by a rearward flow of F-actin and filament depolymerization (22).

IRSp53 is in a position to play a pivotal role in generating filopodia; this brain-enriched protein was discovered as a substrate of the insulin receptor (87). Subsequently, IRSp53 was identified as an effector for Rac1 (50) and Cdc42 (27, 38), where it participates in filopodium and lamellipodium production (38, 51, 54, 86), neurite extension (27), dendritic-spine morphogenesis (1, 15, 66, 67), cell motility and invasiveness (24). The N terminus of IRSp53 contains a conserved helical domain that is found in five different gene products and is referred to as the IRSp53/MIM homology domain (IMD) (51, 70). This domain has been postulated to bind to Rac1 (50, 70) in a nucleotide-independent manner (52), but no convincing effector-like region has been identified. A Cdc42-specific CRIB-like sequence that does not bind Rac1 (27, 38) allows coupling of this and perhaps related Rho GTPases. The structure of the IMD reveals a zeppelin-shaped dimer that could bind “bent” membranes; thus, its potential as an F-actin-bundling domain (51, 82) could be an in vitro artifact often attributed to proteins with basic patches (46). Although there are reports of F-actin binding at physiological ionic strength (ca. 100 mM KCl) (82, 19), this region when expressed in isolation does not decorate F-actin in vivo.

Two reports showed the IMD to be an “inverse-BAR” domain. BAR (Bin-amphipysin-Rvs) domains are found in proteins involved in endocytic trafficking, such as amphipysin and endophilin, and stabilize positively bent membranes, such as those on endocytic vesicles (31, 47). The IMD domains of both IRSp53 (70) and MIM-B (46) associate with lipids and can induce tubulations of PI(3,4,5)P₃ or PI(4,5)P₂-rich membranes, respectively. These tubulations are equivalent to membrane protrusions and are also referred to as negatively bent membranes. Ectopic expression of the IMD from IRSp53 (51, 70, 82, 86) or two other family members, MIM-B (11, 46) and IRTKS (52), can give rise to cells with many peripheral extensions. MIM-B is said to stimulate lamellipodia (11), while IRTKS generates “short actin clusters” at the cell periphery (52).

In IRSp53 is a CRIB-like motif that mediates binding to Cdc42 (27, 38), but the function of this interaction in unclear. Cdc42 could relieve IRSp53 autoinhibition as described for N-Wasp (38), but there is little evidence for this. It has been suggested that Cdc42 controls IRSp53 localization and actin remodeling (27, 38), but another study indicated that these events are Cdc42 independent (19). IRSp53 contains a central SH3 domain that may bind proline-rich proteins, such as Dia1 (23), Mena (38), WAVE2 (49, 50, 69), and Eps8 (19, 24). However, it seems unlikely that all of these represent bona fide partners, and side-by-side comparison is provided in this study. Mena is involved in filopodium production (37), Dia1 in stress fiber formation (81), and WAVE2 in lamellipodium extension (72). Thus, Mena is a better candidate as a partner for IRSp53-mediated filopodia than Dia1 or WAVE2.

There is good evidence for IRSp53 as a cellular partner for Eps8 (19). Eps8 is an adaptor protein containing an N-terminal PTB domain that can associate with receptor tyrosine kinases (65), and perhaps β integrins (13), and a C-terminal SH3 domain that can associate with Abi1 (30). Binding of the general adaptor Abi1 appears to positively regulate the actin-capping domain at the C terminus of Eps8 (18). It has been suggested that IRSp53 and Eps8 as a complex regulate cell motility, and perhaps Rac1 activation, via SOS (24); more recently, their roles in filopodium formation have been addressed (19). The involvement of IRSp53, but not MIM-B or IRTKS, in filopodium formation might be related to its role as a Cdc42 effector. We show here that, surprisingly, the CRIB motif is not essential for this activity, but rather, the ability of IRSp53 to associate via its SH3 domain is required, and that this domain is controlled by 14-3-3 binding.

We have focused on the regulation of Cdc42 effectors that bind 14-3-3, including IRSp53 and PAK4, which are found as 14-3-3 targets in various proteomic projects (32, 44). In this study, we characterize the binding of 14-3-3 to IRSp53 and uncover how this activity regulates IRSp53 function. The phosphorylation-dependent 14-3-3 binding is GSK3β dependent, and 14-3-3 blocks the accessibility of both the CRIB and SH3 domains of IRSp53, thus indicating its primary function in controlling IRSp53 partners. This regulation of the SH3 domain by 14-3-3 is critical in the proper localization and termination of IRSp53 function to promote filopodium dynamics.

MATERIALS AND METHODS

Reagents and materials.

The anti-Flag monoclonal antibody (MAb) M2-Sepharose and protein A-Sepharose were from Sigma. Glutathione-Sepharose was from GenScript. Anti-Flag (rabbit; used for immunoblotting [IB] at 1:1,000 and for immunofluorescence assay [IF] at 1:100) and rabbit antiphosphothreonine were purchased from Sigma. Antihemagglutinin (anti-HA) (SC-805; used for IB at 1:1,000) and anti-14-3-3 (SC-629; used for IB at 1:2,500) were purchased from Santa Cruz. Rabbit anti-green fluorescent protein (anti-GFP) (used for IB at 1:1,000 and for IF at 1:100) was from Invitrogen (A11122). An IRSp53 MAb raised against residues 1 to 364 has been previously described (27) and was used at a concentration of 1:50 for IF. Anti-EPS8 was from Transduction Labs. Horseradish peroxidase-conjugated rabbit anti-mouse IgG and goat anti-rabbit IgG were from Dako (used at 1:4,000). The secondary antibodies used in immunofluorescence experiments were Alexa Fluor 488- or Alexa Fluor 546-conjugated anti-mouse/rabbit IgG (Molecular Probes). F-actin was visualized with rhodamine-conjugated phalloidin (Sigma; used at 1:1,000).

Mammalian and bacterial expression constructs.

pXJ/HA-IRSp53 was subcloned into the BamHI-NotI sites in pXJ/Flag and pXJ/GFP (45). C-terminal truncations of IRSp53 were generated by PCR using the BamHI N-terminal primer and NotI C-terminal primers (restriction sites are in italics) for the 1-to-440 construct (5′-GATGCGGCCGCACTGCCATCGCTGTCCAAG-3′) and for the 1-to-375 construct (5′-GATGCGGCCGC-TTAGCCATTGCGCTCCAGGCC-3′). IRSp53(T340,360A) and IRSp53(I402P) were generated by PCR-mediated mutagenesis using the QuikChange Site-Directed Mutagensis Kit (Stratagene) according to the manufacturer's instructions. pXJ/GST-IRS(337-370) and GST-IRS(337-370/TA) were generated by using 5′-GATGGATCCTACTCCACCACACTCCCC-3′ and 5′-GACGGTACCTTAGCCGGCTGCCATGGAGCT-3′ to amplify an IRSp53(337-370) BamHI-Asp718 fragment from IRSp53 or IRSp53(T340,360A) and then subcloned into the BamHI-Asp718 site in pXJ/GST. pXJ/HA-Eps8 and GFP-Eps8 were generated by amplifying murine Eps8 from an EST clone (I.M.A.G.E., 4240899) with an N-terminal BamHI primer (5′-CGCGGATCCGATAGGAATTATGACGCAGTC-3′) and a C-terminal NotI primer (5′-ATAAGAATGCGGCCGCTCAGTGGCTGCTCCCTTC-3′) and subcloning the fragment into the BamHI-NotI site in either pXJ/HA or pXJ/GFP. Internal deletions of residues 266 to 309 or 298 to 309 in Eps8 were generated by PCR using the N-terminal BamHI primer with 5′-CCGCTCGAGCCGGGCTGGCCATCATCTC-3′ (1 to 266) or 5′-CCGCTCGAGGGACCTGGAGAGGGGGTT-3′ (1 to 298) to amplify BamHI-XhoI fragments and 5′-CCGCTCGAGAGAAAGCTCAGAAAACGCTTC-3′ with the C-terminal NotI primer to amplify an XhoI-NotI fragment encompassing amino acids 309 to 821. The fragments were then digested, ligated together, and subcloned into the BamHI-NotI sites in pXJ/HA and GFP. Glutathione S-transferase (GST)-Eps8(200-520)/pGEX-4T-1 was generated by using 5-GAGGATCCAGGATGATTGCCAAAGCAG-3′ and 5′-CATGCGGCCGCTACTTGGTGATTAGGAGTTGA-3′ to amplify an Eps8(200-520) BamHI-NotI fragment from Eps8, Eps8(Δ266-309), or Eps8(Δ298-309) and then subcloned into the BamHI-NotI site in pGEX-4T-1 (GE Life Sciences). pXJ/GFP-VASP was generated by amplifying a VASP BamHI-XhoI fragment from an EST clone (human; Image, 5210128) using the primers 5′-CGCGGATCCATGAGCGAGACGGTCATCTGT-3′ and 5′-CCGCTCGAGTCAGGGAGAACCCCGCTTCCT-3′ and then subcloned into the BamHI-XhoI sites in pXJ/GFP. The Cdc42(G12V) constructs have been described previously (46).

Tissue culture.

COS-7 cells were maintained in either high-glucose (4.5 g/liter) or low-glucose (1 g/liter) Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS). HeLa cells were maintained in minimal essential medium (MEM) containing 10% fetal bovine serum, 10 mM sodium pyruvate (Invitrogen), 2 mM l-glutamine (Invitrogen), 100 μM nonessential amino acids (Invitrogen), and 0.15% sodium bicarbonate. All cells were maintained at 37°C and 5% CO₂.

Immunoprecipitation and Western blot analysis.

COS-7 cells were transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Briefly, cells were grown to 90% confluence in 60-mm dishes and transfected with DNA-liposome at a ratio of 4 μg to 15 μl Lipofectamine 2000 (preincubated in 300 μl of serum-free DMEM for 20 min at room temperature) in 3 ml of 10% FBS-DMEM for 4 h prior to the medium being replaced with 5 ml of fresh growth medium. The following day, the cells were treated if indicated and then harvested in 500 μl of ice-cold lysis buffer (25 mM HEPES, pH 7.2, 100 mM NaCl, 5 mM MgCl₂, 0.5% Triton X-100, and 4% glycerol). The lysates were incubated on ice for 10 min and then spun at 14,000 × g for 12 min at 4°C. For input samples, 45 μl of the soluble cell lysate (SCL) was mixed with 15 μl of 4× SDS sample loading buffer and heated for 10 min at 100°C. For affinity precipitations, 400 μl of the SCL was rolled with either a 15-μl bead volume of anti-Flag M2-Sepharose (immunoprecipitation [IP]) or a 30-μl bead volume of glutathione-Sepharose (pulldown [PD]) at 4°C for 3 h. The samples were then spun at 14,000 × g for 1 min, and the supernatant was removed gently by aspiration and washed with 1 ml of lysis buffer. This was repeated three times, and then the beads were boiled in 45 μl 1.5× SDS sample loading buffer for 12 min each time. For Western blot analysis, 15 μl of the input sample (2.5% of the total) and 20 μl of the IP/PD (33% of the total) were resolved on SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membranes prior to being immunoblotted with the indicated antibodies.

Immunofluorescence and live-cell imaging.

HeLa cells were grown on acid-washed coverslips either placed into 35-mm dishes or glued to the bottoms of stamped 35-mm dishes to a confluence of 50% and transfected with Optifect (Invitrogen) according to the manufacturer's instructions. Briefly, cells were transfected with DNA-Optifect at a ratio of 1 μg to 5 μl (preincubated in 100 μl of serum-free MEM at room temperature for 30 min) for 4 h in 1 ml 10% FBS-MEM prior to being replaced in 2 ml fresh growth medium overnight. The following day, the cells were either used for live-cell imaging (see below) or fixed in 3% paraformaldehyde for 10 min. The cells were then permeabilized in 0.1% PBST (phosphate-buffered saline [PBS] plus 0.1% Triton X-100) for 10 min and blocked in 2% bovine serum albumin (BSA) in PBS for 1 h at room temperature. The coverslips were then incubated with primary antibodies diluted in blocking buffer for 2 h at room temperature prior to being washed three times with 0.1% PBST. Secondary antibodies were added in the presence of rhodamine-phalloidin in blocking buffer for 1 h at room temperature. After being washed three times in 0.1% PBST and once in double-distilled H₂O (ddH₂O), the coverslips were air dried and mounted on slides using AquaPerm (Shandon Immunon) and sealed with nail polish. Samples were viewed on a Zeiss 510 meta-laser scanning confocal microscope with a C-Apochromat 63×/1.2 Wcorr objective using excitation wavelengths of 488 nm (Alexa Fluor 488 and GFP) or 546 nm (Alexa Fluor 543 and rhodamine).

For live-cell-imaging experiments, cells were placed in fresh growth medium and equilibrated for 30 min on an Olympus Fluoview laser scanning confocal microscope with a 63×/1.4 objective equipped with a humidified sealed chamber maintained at 37°C and 5% CO₂ prior to visualization. GFP fusion proteins were observed using an excitation wavelength of 488 nm, and cells were observed in parallel using differential interference contrast (DIC). To visualize filopodium dynamics, samples were collected under a 4× zoom every 10 s for a period of 20 min.

Filopodium quantifications.

To determine the number of filopodia, still images were obtained from the 20-min live-cell recording sessions at the 5-, 10-, and 15-min time points. Over a region of 50 μm in each image, the filopodia that were present were counted. The quantifications from these three images were then averaged, and this number was used as the final “average” number of filopodia on a given cell. This was repeated on 5 different cells in a least 3 separate experiments, and the final results were tabulated and subjected to an analysis of variance (ANOVA). To quantify the filopodial life span, the 120 still images of each recording session were carefully analyzed for the emergence and retraction of filopodia. The number of frames from the point of emergence to its complete loss was determined and multiplied by 10 s to achieve the filopodial life span. Twenty-five filopodia from 5 different cells from 3 separate experiments were recorded, and then the 125 filopodia were assigned to one of three categories: (i) short (50 s or less), (ii) average (60 to 180 s), or (ii) long (greater than 180 s).

RESULTS

IRSp53 associates with 14-3-3.

Proteomic studies of 14-3-3 binding proteins (including our own) have revealed that IRSp53 is enriched in pulldowns using various isoforms of 14-3-3 (32, 44). Since 14-3-3 serves as a transducer of serine-threonine phosphorylation signals (5), we decided to investigate how this Cdc42 target might be regulated by 14-3-3 binding. The amount of HA-tagged 14-3-3ζ bound to Flag-IRSp53 was augmented by treatment of COS-7 cells with the phosphatase inhibitor calyculin A (Fig. 1a), showing the interaction was likely a conventional one (i.e., phosphorylation dependent). Both endogenous 14-3-3 and ectopically expressed 14-3-3ζ bound Flag-IRSp53 (Fig. 1a). Using relevant antibodies, we found that endogenous 14-3-3 was recovered with IRSp53 (Fig. 1b), which appears as a doublet, likely due to alternative C-terminal splicing (87).

FIG. 1.

IRSp53 associates with 14-3-3 in a GSK3β-dependent manner. (a) Flag-tagged IRSp53 was transfected alone or together with HA-tagged 14-3-3ζ in COS-7 cells. The cells were left untreated or treated with calyculin A prior to lysis. IRSp53 was immunoprecipitated, and associated 14-3-3 was detected by anti-14-3-3. IRSp53 associates with 14-3-3 in a calyculin A-sensitive manner. (b) Endogenous 14-3-3 was immunoprecipitated from untreated or calyculin A-treated COS-7 cells, and associated IRSp53 was determined by an anti-IRSp53 MAb. Anti-p75NTR was used as a negative control. (c) Flag-IRSp53-transfected cells were left untreated or treated with the indicated inhibitors for 1 h prior to lysis. IRSp53 was immunoprecipitated, and associated 14-3-3 was detected by anti-14-3-3 immunoblotting. LiCl abrogates binding of 14-3-3 to IRSp53. (d) Schematic of IRSp53 C-terminal deletion truncation constructs. (e) HA-tagged IRSp53 truncations were cotransfected with Flag-14-3-3ζ in COS-7 cells and then subjected to anti-Flag immunoprecipitations after calyculin A treatment. Associated IRSp53 was detected with anti-HA immunoblotting. The N-terminal 375 residues of IRSp53 are required for binding to 14-3-3.

We noted that the concentration of glucose in the medium affected the degree of binding between 14-3-3ζ and IRSp53. This suggested that phosphorylation was responsive to kinases that respond to extracellular glucose levels, including protein kinase A (PKA) (20), GSK3β (33), phosphatidylinositol 3-kinase (PI3-K) (34), and mTOR (88). To determine which of these might be involved in driving 14-3-3 binding to IRSp53, transfected cells (in high-glucose medium) were tested with kinase inhibitors prior to immunoprecipitation. While inhibition of PKA, PI3-K, and mTOR had no effect on 14-3-3 binding to IRSp53, lithium chloride (LiCl), a potent and specific inhibitor of GSK3β, significantly reduced binding (Fig. 1c). Although the association of 14-3-3 with IRSp53 is GSK3β dependent, we were unable to find direct phosphorylation of IRSp53 by GSK3β in vitro (data not shown) or evidence for the required priming sites (Fig. 2).

FIG. 2.

Mapping the phosphorylation sites on IRSp53 responsible for 14-3-3 binding. (a) Alignment of amino acids in the region found to bind 14-3-3 comparing human, mouse, and zebrafish IRSp53 proteins. Sequences flanking the two threonine residues involved in 14-3-3 binding (asterisks) are conserved, including the TLP sequence (underlined). Serine residues tested by serine-to-alanine mutagenesis are indicated with an A. (b) IRSp53 residues 360 to 365 are essential for 14-3-3 binding. HA-IRSp53 constructs were cotransfected with Flag-14-3-3ζ in COS-7 cells and treated with calyculin A for 10 min before being harvested. The amount of IRSp53 (anti-HA) associated with the anti-Flag immunoprecipitates is shown. (c) Scansite analysis of all putative 14-3-3 binding sequences in IRSp53 are shown with their probability scores. Phosphopeptides with sequences as shown contained a C-terminal Gly-bA-bA-bA linker (bA, beta-alanine) and were synthesized in situ on cellulose (Jerini Biotools), blocked with 1% BSA, and overlaid for 1 h with 10 μg/ml biotin-labeled 14-3-3ζ (43) in 1% BSA. Bound 14-3-3 detection was with strepavidin-horseradish peroxidase (HRP) (1:5,000; 30 min). Control indicates the peptide control corresponding to Raf1 pS259. Those IRSp53 sites that have been identified as phosphorylated by proteomic studies in vivo (78) are indicated. pS, phosphoserine; pT, phosphothreonine. (d and e) IRSp53 constructs were tested for 14-3-3 binding as for panel b. The S364/S365 residues that might act as GSK3β priming sites are dispensable for 14-3-3 binding. Single or double mutation of T340 and T360 led to similar loss of 14-3-3 binding.

Truncation analysis was performed to assess which regions of IRSp53 were needed for 14-3-3 association; initially, only C-terminal truncations were assessed, because the N-terminal IMD is required for its dimerization (51). Flag-14-3-3ζ was coexpressed with the HA-tagged IRSp53 constructs depicted in Fig. 1d, and IRSp53 levels were assessed by Western blotting (Fig. 1e). Removal of the SH3 domain of IRSp53(375-440) diminished but did not abolish 14-3-3 binding. Since there is no predicted or actual 14-3-3 binding site within the SH3 domain (see below), this represents a conformational effect and the loss of binding to the kinase(s) responsible for modifying IRSp53. The region between the SH3 domain and the CRIB motif of IRSp53 (residues 305 to 375) was found to be essential for 14-3-3 interaction.

IRSp53 modification of threonine 340 and 360 drives 14-3-3 binding.

Sequence alignment showed significant numbers of conserved residues in the uncharacterized IRSp53(305-375) (Fig. 2a) that are not present in other family members, such as IRTKS. Based on deletion analysis, sequences between the CRIB region and the SH3 domain were likely to contain critical 14-3-3 binding sequences. Starting with IRSp53(1-375), we generated constructs with sequential 5-amino-acid deletions (Fig. 2b). Residues 361 to 365 were absolutely required for 14-3-3 binding, consistent with these residues being involved directly in 14-3-3 interaction. The 14-3-3 binding pocket accommodates approximately 5 residues flanking the phospho-residue.

To identify the potential 14-3-3 binding sites across the entire protein, we used Scansite (55) to identify putative motifs (including noncanonical ones [16]) and then used an overlay assay (which will be expanded upon elsewhere) to directly test these sequences using synthetic peptides. As can be seen, Raf1 pS495 (a validated site) was detected by this protocol (Fig. 2c). Only pT340 and pT360 significantly bound 14-3-3, although we detected a faint signal with pS366 (but this site was excluded as being required for binding [Fig. 2b]). The C-terminal region of IRSp53 including the SH3 domain (375 to 524) showed no association with 14-3-3.

To substantiate the exclusive involvement of pT340 and pT360, a number of additional mutations were made. Neither substitution of S335/S337 (see Fig. S1a in the supplemental material) nor that of S364/S365 affected the extent of 14-3-3 binding (Fig. 2d). The latter set of mutations demonstrated that GSK3β (which requires a priming phosphorylation at the +3/4 position) is not involved in direct modification of IRSp53 T360 (see below). In summary, critical 14-3-3 binding determinants include IRSp53 residues 360 to 365 (Fig. 2b) and residues 330 to 342 (see Fig. S1a, lane 4, in the supplemental material), consistent with the peptide array data showing that T340 and T360 are the dominant 14-3-3 binding sites. T340 and T360 include the putative modified motif pTLP; both residues have been identified as phosphorylated in vivo (78). We assessed the role of modification of these threonines by alanine substitution (Fig. 2e). Substantial loss of binding with mutation of a single site indicated that T340 and T360 function as a pair. The ability of 14-3-3 to associate with IRSp53(337-370) (see Fig. S1b in the supplemental material) indicates that the linking sequence probably spans the two halves of 14-3-3. In this context, substitution of either T340 or T360 was sufficient to prevent 14-3-3 binding. To confirm phosphothreonine modification of IRSp53, we probed the protein recovered in 14-3-3 pulldown using a phosphothreonine antibody (see Fig. S1c in the supplemental material); the immunosignal was correlated with the amount of IRSp53 present in the 14-3-3 complex (compare lanes 2 and 3).

The IRSp53/14-3-3 complex cannot bind Cdc42 or WAVE2.

Because the 14-3-3 binding region of IRSp53 lies between the Cdc42-binding CRIB region (38) and the SH3 domain, it seemed possible that 14-3-3ζ could compete for Cdc42-GTP or WAVE2 binding. Immobilized GST-Cdc42V12 was able retain HA-IRSp53, as expected; however, 14-3-3ζ was not detected in the complex (Fig. 3a), in contrast to the Cdc42 effector PAK4, which also binds 14-3-3 (Y. Baskaran, personal communication) and serves as a control. Note that under identical conditions GST-Cdc42V12 was competent to form a trimeric complex with IRSp53 and WAVE2 (Fig. 4d). When Flag-14-3-3ζ was used to precipitate HA-IRSp53, no GST-Cdc42V12 could be recovered in the complex (Fig. 3b). Then we tested if Flag-14-3-3/HA-IRSp53 complex could include the SH3 target WAVE2 (Fig. 3c). As a control, we recovered an equivalent amount of Flag-IRSp53 and confirmed the presence of HA-WAVE2. The 14-3-3 immunoprecipitates were completely devoid of WAVE2, showing that the trimeric complex is unfavorable.

FIG. 3.

The binding of 14-3-3 to IRSp53 prevents association of Cdc42-GTP and WAVE2. (a) GST-Cdc42V12 was cotransfected with HA-IRSp53 or HA-PAK4 in the absence or presence of Flag-14-3-3ζ. After calyculin A treatment, Cdc42V12 was precipitated, and associated IRSp53, PAK4, and 14-3-3 were detected by their respective antibodies. (b) Flag-14-3-3ζ was cotransfected with GST-Cdc42V12, HA-IRSp53, or both prior to being immunoprecipitated subsequent to calyculin A treatment. Associated Cdc42V12 and IRSp53 were detected by anti-GST or anti-HA, respectively. (c) HA-WAVE2 was cotransfected with either Flag-IRSp53 or both HA-IRSp53 and Flag-14-3-3ζ and treated with calyculin A. Recovery of WAVE2 by IRSp53, but not 14-3-3-complexed IRSp53, was determined after Flag immunoprecipitation with anti-HA immunoblotting.

FIG. 4.

Binding of 14-3-3 to IRSp53 blocks association with WAVE2 or EPS8. (a) HA-tagged PAK1, WAVE2, N-WASP, or PAK4 was transfected with Flag-IRSp53 and tested by coimmuniprecipitation in COS7 lysates. Although PAK1 and N-WASP contain multiple proline-rich regions, only WAVE2 was detected as an IRSp53 partner. (b) Flag-IRSp53 can bind a variety of proteins found at lamellipodia. IRSp53 was tested by cotransfection with tagged versions of VASP or WAVE2 or by mixing IRSp53-containing lysates with those containing Eps8 (lanes marked with asterisks) (see Materials and Methods). (c) IRSp53 bound to Flag-14-3-3ζ does not contain WAVE2 or Eps8. All cells were subjected to 10 min of calyculin A treatment prior to being harvested. 14-3-3ζ was recovered on anti-Flag Sepharose, and immunoprecipitates (IP) were tested with anti-GFP. (d) GST-Cdc42V12 was cotransfected with GFP-tagged VASP or WAVE2 in the absence or presence of HA-IRSp53 and treated with calyculin A prior to being harvested. GFP-Eps8-containing cell lysate was expressed separately and mixed with lysates as indicated. Proteins bound to glutathione-Sepharose were assessed by Western blotting. VASP, WAVE2, and Eps8 were able to bind to the Cdc42V12/IRSp53 complex. The asterisks indicate lanes in which Eps8 was expressed separately and mixed after cell lysis.

14-3-3 binding to IRSp53 blocks SH3 interactions.

Given the finding that 14-3-3 binding prevents the association of WAVE2 with IRSp53, it was of interest to investigate how phosphorylation of IRSp53 and 14-3-3 binding affected its ability to bind another partner protein(s). Since the SH3 domain of IRSp53 might interact with a host of proline-rich proteins, the specificity of partner interaction was first tested. As has been shown previously (49), IRSp53 can bind WAVE2 (Fig. 4a). However, other Cdc42 effectors with a range of proline-rich regions (cf. PAK1, N-WASP, and PAK4) were unable to associate, illustrating some degree of IRSp53 SH3 selectivity. Recent work has suggested that IRSp53 can bind N-WASP (42), but it is detectable only via Förster resonance energy transfer (FRET) techniques, suggesting that the interaction is transient or weak. In addition to WAVE2, other IRSp53 partners that show association by coimmunoprecipitation are VASP and Eps8 (Fig. 4b). Because IRSp53 coexpression with Eps8 causes both to become Triton X-100 insoluble (reference 19 and data not shown), COS-7 lysates containing Eps8 were mixed with those containing IRSp53 to test for coimmunoprecipitation. We concluded that Eps8 binds to IRSp53 much more efficiently than other proline-rich proteins (19), since VASP and WAVE2 did not bind with these in vitro incubations (Fig. 4b and data not shown).

These partners were also tested for their abilities to associate with IRSp53 in the presence of 14-3-3. As shown in Fig. 3c, no WAVE2 could be recovered via IRSp53 (Fig. 4c). VASP associates directly with 14-3-3 (59), but there was no increase in levels with additional IRSp53. Finally, a minor amount of Eps8 (right lane) likely reflected the “reverse” association of Eps8-SH3 with the IRSp53 PXXDY motif (residues 467 to 471) (19). We concluded that one function of 14-3-3 binding is to block access to the flanking SH3 domain. Whether 14-3-3 binding can dislodge a bound SH3-binding partner following phosphorylation of the critical threonines has yet to be determined. Further, the possibility that other phosphorylation sites contribute to target dissociation prior to 14-3-3 binding cannot be ruled out. What is clear from our results, however, is that once IRSp53 is complexed to 14-3-3, its SH3 domain is no longer capable of interacting with its partners.

Cdc42 can be found in a ternary complex with IRSp53 and Mena (38), and thus, Cdc42 binding to IRSp53 does not block the SH3 domain (nor is there augmentation of SH3 binding [data not shown]). The indirect association of Cdc42V12 with other targets—VASP, WAVE2, and Eps8—was analyzed in the absence or presence of IRSp53 (Fig. 4d). Low levels of VASP could coprecipitate with Cdc42V12 and significantly more with the addition of IRSp53. Neither WAVE2 nor Eps8 associated with Cdc42V12 under these conditions unless IRSp53 was coexpressed. Thus, while the association of 14-3-3 blocks accessibility of the SH3 domain of IRSp53, Cdc42 is able to form a ternary complex with IRSp53 and its SH3 domain-binding partners. Interestingly, our data also suggest that VASP is efficiently coupled to Cdc42 via an alternate effector. This will be investigated further.

The SH3 domain of IRSp53 is required for lamellipodial localization.

Since IRSp53 CRIB, and the SH3 domain are characterized protein-protein interaction sites, we were interested to investigate how they contribute to localization of the protein. It has been suggested that Cdc42 is critical for IRSp53 localization to filopodia (38), but as we show below, Cdc42 does not appear to play such a role in the absence of SH3 function. Single point mutations in the CRIB motif are sufficient to abolish Cdc42 binding to IRSp53 (38). A mutant with loss 14-3-3 binding to IRSp53 (T340,360A) can be used, but gain of function is impossible, since phosphomimetic substitution is ineffective in 14-3-3 binding. The IRSp53(I402P) SH3 mutant cannot bind WAVE2 (15), VASP, or Eps8 (Fig. 5a).

FIG. 5.

The SH3 domain of IRSp53 is required for lamellipodial localization. (a) Flag-IRSp53 or IRSp53(I402P) was cotransfected with VASP, WAVE2, or Eps8 and tested for interaction as in Fig. 4b. In all cases, IRS(I402P) failed to interact with target proteins. (b) The localization of GFP-tagged IRSp53, IRSp53(I267N), or IRSp53(I402P) in HeLa cells was determined by anti-GFP indirect immunofluorescence and confocal microscopy. The boxed regions are enlarged above to illustrate the lamellipodial region. Both IRSp53 and IRSp53(I267N) were enriched in the lamellipodium, but not IRSp53(I402P). Scale bar = 20 μm. (c) Enlarged regions (×4) show typical filopodia in cells like those in panel b. F-actin was visualized with rhodamine-phalloidin. The arrows point to IRSp53 and IRSp53(I267N) localization at filopodium tips, while the arrowheads point to filopodia with no discernible IRSp53(I402P) staining. Scale bar = 5 μm. (d) Localization of HA-IRSp53(1-440) and IRSp53(1-375) in HeLa cells with respect to a MAb to Eps8, which marks lamellipodia. Note that the clear colocalization is dependent on the presence of the SH3 domain. (e) GFP-IRSp53(440) or IRSp53(375) was introduced into HeLa cells that were imaged by confocal microscopy for 60 frames at 10 s/frame. Typical filopodial morphology and dynamics are illustrated. Cells expressing IRSp53(1-440) (top) exhibit numerous and rigid filopodia versus cells expressing IRSp53(1-375) (bottom).

The localization of IRSp53 in HeLa cells is shown in Fig. 5b. Both endogenous and exogenously expressed IRSp53 are located in a small discrete band at the cell edge (Fig. 5b, top row, left), which is a well-organized lamellipodium in HeLa and MRCK cells (73), and in other sites that are largely dorsal membrane ruffles (data not shown). This was also seen in NIH 3T3 cells (data not shown). At these moderate levels of expression, no perturbation of cell morphology or F-actin organization was observed. The Cdc42-defective IRSp53(I267N) localized in a manner indistinguishable from that of the wild type (Fig. 5b, middle). Strikingly, without a functional SH3 domain, IRSp53 failed to localize to the lamellipodium (right). Weak IRSp53(I402P) staining at the lamellipodium likely resulted from dimerization with endogenous IRSp53.

Indirect immunostaining of fixed samples showed that IRSp53 was often present in filopodial structures or retraction fibers. Filopodium-enriched IRSp53 was confirmed by live-cell imaging (see Videos S1 to S3 in the supplemental material). From such analysis, the spots in Fig. 5c likely represent IRSp53 at the tips of filopodia (arrows). The IRSp53(I267N) Cdc42-binding-defective protein was also localized to filopodium tips (middle, arrows), indicating that Cdc42 binding is not necessary to bring IRSp53 to these sites, as previously suggested (38). In contrast, IRSp53(I402P) was never enriched at these sites and was largely lacking in actin-rich protrusions (right, arrowheads). This is even more evident in the live-cell images (see Video S3 in the supplemental material), where filopodia could clearly be seen in the DIC image while no IRSp53(I402P) could be detected in the shafts or tips of these filopodia. Together, these imaging studies demonstrate that the SH3 domain of IRSp53 is critical for its enrichment in the lamellipodia and filopodia, rather than the CRIB motif.

The critical involvement of the SH3 domain was again illustrated by comparing the lamellipodial enrichment of IRSp53(1-440), which lacks the actin binding WH2 domain, or IRSp53(1-375), which also lacks the SH3 domain. The IRSp53(1-440) signals were indistinguishable from those of full-length IRSp53 at the lamellipodium (Fig. 5d, top), while IRSp53(1-375) exhibited even staining across cells. Endogenous Eps8 was used as a marker for the lamellipodium (see Fig. 7). Typical live-cell imaging of the dynamic nature of the IRSp53-enriched structures (Fig. 5e) showed signal in both lamellipodia and filopodia. Cells expressing IRSp53(1-440) displayed typical dynamic but “rigid” filopodia found in HeLa cells. With IRSp53(1-375) expression, we observed fainter GFP signals associated with the dynamic protrusions. IRSp53(1-440) was again seen in the lamellipodial region between filopodia (see Fig. 8b). Interestingly, filopodia in cells expressing IRSp53(1-375) lacked normal rigidity, suggesting a dominant inhibitory effect. Together, the data suggest that IRSp53 promotes, but is not necessary for, filopodium production.

FIG. 7.

Eps8 enhances the lamellipodial localization of IRSp53. (a) Schematic of the domain structure of Eps8. Residues conserved between human and zebrafish are boxed in black, while the IRSp53-binding region is marked above. (b) Flag-tagged IRSp53 or Eps8 was transfected into HeLa cells, and the colocalization of endogenous Eps8 or IRSp53 was visualized with confocal microscopy. Endogenous Eps8 (left) and endogenous IRSp53 (right) colocalized strongly at the plasma membrane with either Flag-tagged IRSp53 or Eps8, respectively. The scale bars represent 10 μm. (c) Flag-tagged IRSp53 was expressed in HeLa cells either alone or with GFP-Eps8, and protein localization was determined by anti-Flag indirect immunofluorescence. IRSp53 localizes consistently to the lamellipodium, but this localization is significantly enhanced by the coexpression of Eps8. (d) The fluorescence intensity of IRSp53 signal was analyzed across the lamellipodium as indicated by the boxed regions in panel c using ImageJ software, and the quantification of the lamellipodium/cytoplasm ratio is shown. Eps8 expression significantly increased the localization of IRSp53 to the lamellipodium (P < 0.0001; Student's t test).

FIG. 8.

14-3-3 binding-defective IRSp53 exhibits more robust lamellipodial and filopodial staining and influences filopodial dynamics. (a) Expression of GFP-tagged IRSp53 or IRSp53(T340,360A) in HeLa cells was visualized by anti-GFP immunofluorescence. F-actin was visualized with rhodamine-phalloidin. (Right) Enlargements of the areas indicated by the white boxes on the left. IRSp53(T340,360A) localized more strongly to the lamellipodial region, and in some cases filopodium staining was apparent. Scale bars = 20 μm (left) and 5 μm (right). (b) Quantification of the number of filopodia across a 50-μm region of HeLa lamellipodium by live imaging. The symbols indicate significance between filopodial numbers seen for GFP-IRSp53 and GFP (*) or between GFP and IRSp53(I402P) (#) (P < 0.02). The IRSp53(I402P) SH3 mutant did not stimulate filopodia. Other mutants did not promote filopodia to a greater extent than the wild type. (c) Quantification of the life spans of filopodia by live imaging of HeLa cells expressing the indicated GFP-tagged constructs. IRSp53(T340,360A)-expressing cells exhibited increased filopodial life span compared to IRSp53 or IRSp53(I267N). (d) Model of the findings uncovered in this study. IRSp53 is recruited to the lamellipodium via its SH3 domain by partners such as Eps8. Activated Cdc42 can associate with the CRIB motif of IRSp53 and might help to stabilize IRSp53 to sites of filopodial activity. Once recruited, IRSp53 can coordinate membrane tubulation via its IMD; the function of the WH2 domain is unclear. Phosphorylation of T340 and T360 promotes the binding of 14-3-3, which then blocks access to the SH3 domain of IRSp53 by other partners and binding of Cdc42-GTP to the CRIB region. Dephosphorylation of IRSp53 is likely inhibited by 14-3-3 binding but then allows subsequent recruitment to the lamellipodium by its SH3-binding partners.

SH3 domain specificity underlies the lamellipodial localization of IRSp53.

The importance of SH3 binding for IRSp53 localization was unexpected given the previous focus on Cdc42 and lipid interactions (63). To evaluate this issue in more detail, we compared the localization of IRSp53 with that of a related family member, IRTKS (52). IRTKS contains a domain architecture similar to that of IRSp53 but lacks the central CRIB motif and 14-3-3 binding region. It has been noted previously that IRTKS overexpression produces a distinct actin phenotype in COS7 cells (52), which was attributed to the C-terminal WH2 domain. Compared with IRSp53, IRTKS is only 47% identical in the IMD and 62% in the SH3 domain (Fig. 6c). Interestingly GFP-IRTKS showed quite different disposition in HeLa cells monitored by confocal live-cell microscopy. GFP-IRSp53 was enriched in lamellipodia and filopodia (Fig. 6a; see Video S5 in the supplemental material). In contrast, the GFP-IRTKS protein was concentrated at lamellae and within membrane ruffles (Fig. 6a; see Video S6 in the supplemental material). This confirmed the previous observation that these two proteins have different cellular dispositions in other cell types (52).

FIG. 6.

The SH3 domain of IRSp53 is required for lamellipodial localization. (a) Confocal images of live HeLa cells expressing either GFP-IRSp53 or GFP-IRTKS. IRSp53 localizes to lamellipodia, while IRTKS localization is more restricted to lamellae and membrane ruffles. (b) Flag-tagged IRSp53 or IRTKS were cotransfected with the indicated constructs prior to immunoprecipitation. GFP-mDia2 and Eps8 were expressed separately, and the lysates were mixed prior to anti-Flag immunoprecipitation (asterisks are above the indicated lanes). Bound proteins as indicated were detected by Western blotting with anti-GFP. Only mDia2 associated to any appreciable extent with IRTKS. (c) Schematic representation of IRSp53, IRTKS, and two chimeras generated. (d) Cell lysates containing Flag-tagged constructs as indicated were mixed with HA-Eps8-containing lysates. Eps8 bound to IRSp53 and IRTKS/IRSp53, but not to IRTKS or IRSp53/IRTKS. (asterisks are above the lanes where Eps8 was not cotransfected). (e) Live confocal images of GFP-tagged IRSp53, IRTKS, and IRSp53/IRTKS localization in HeLa cells. IRSp53 and IRTKS/IRSp53 exhibited clear lamellipodial localization. Both IRTKS and IRSp53/IRTKS were enriched in dorsal membrane ruffles but not in the lamellipodium. Scale bar = 5 μm.

Since Cdc42-binding is not critical to IRSp53 localization (Fig. 5b), we hypothesized that the difference might lie in the specificities of their SH3 domains. We then tested the IRTKS SH3 domain for target binding (including the formin Dia2, which has recently emerged as an important player in filopodium dynamics [85]). While IRSp53 bound Eps8, WAVE2, and VASP (Fig. 6b), none of these proteins significantly interacted with IRTKS. However, like Dia1 (23), Dia2 bound to both IRSp53 and IRTKS, showing that the IRTKS SH3 domain is accessible and functional under our conditions. The different biochemical properties of the SH3 domains led us to test if the proteins acquired their discrete subcellular localizations via their SH3 selectivity. For this purpose, chimeric constructs were created by splicing constructs encoding IRSp53 and IRTKS at the SH3 domain boundary (as shown in Fig. 6c). This retained the 14-3-3 regulatory sequences in the case of the IRSp53-IRTKS splice. The various chimeras depicted in Fig. 6c were tested for Eps8 binding by coimmunoprecipitation (Fig. 6d); these results were entirely in line with an expected switch in behavior upon SH3 domain substitution and showed that other sequences do not overtly alter the SH3 domain properties. Comparison of the dispositions of GFP-tagged chimeras with those of wild-type proteins by confocal live-cell imaging (see Videos S7 and S8 in the supplemental material) indicated that the SH3 domains dictate protein localization. Notably, the IRSp53/IRTKS chimera, which contains the IMD and CRIB regions from IRSp53, nonetheless was preferentially enriched in membrane ruffles versus lamellipodia (Fig. 6e).

Eps8 enhances IRSp53 localization to the lamellipodium.

Our results (Fig. 5) and previous reports (19) indicated that Eps8 is a physiological partner of IRSp53 requiring SH3 interactions. Eps8 localizes to the lamellipodial region of the plasma membrane in HeLa cells (which are not usually polarized), which have been well characterized for typical lamellipodial markers (8, 73). The lamellipodial profiles of these proteins in HeLa cells (Fig. 7b) are essentially identical for all nonmitotic cells (data not shown). Tagged Eps8 and endogenous IRSp53 costained in a pixel-to-pixel manner at the lamellipodium, though IRSp53 showed higher cytosolic levels, perhaps indicative of an “inactive” 14-3-3-bound pool. If SH3 domain interactions promote lamellipodial localization of IRSp53, we would expect increased levels of an associated protein, such as Eps8, to enhance the IRSp53 signal in this region. To assess this model, we took cells (n = 10 per sample) expressing as nearly as possible the same amount of IRSp53 and analyzed a lamellipodial region from each in the absence or presence of overexpressed Eps8 (Fig. 7c; see Fig. S3 in the supplemental material). Inspection showed clearly that Eps8 expression enhanced the ratio of lamella to lamellipodial IRSp53 signal. The typical fluorescence intensity signal (pixels averaged parallel to the lamellipodium in the boxed region) is shown in Fig. 7d, and the average lamellipodium/cytoplasmic-protein ratio for the sets of 10 cells was plotted (Fig. 7d).

Analysis of a 14-3-3 binding-defective IRSp53 mutant.

Given the essential role of the SH3 domain of IRSp53 in its localization to the lamellipodium, we next wished to explore the role of 14-3-3, since its binding to IRSp53 regulates access to this domain. GFP-IRSp53 that is mutated in the two 14-3-3 binding sites [IRSp53(T340,360A)] (Fig. 2d) was monitored by indirect immunofluorescence in fixed cells or by live-cell imaging (see Video S4 in the supplemental material), in all cases with analysis of low-expressing cells. IRSp53(T340,360A) exhibited more robust lamellipodial enrichment (Fig. 8a, bottom row). We did not attempt to assess filopodial-tip localization (Fig. 8a, right), since these structures are unstable to fixation. IRSp53(T340,360A) could promote more elongated and branched protrusions from the cell periphery (see Fig. S2a in the supplemental material). These types of structures are reported with wild-type IRSp53 (52) or by IMD expression alone (70), likely by IMD-mediated membrane tubulation rather than actin reorganization (70). At a low level of expression, wild-type IRSp53 protein did not promote these aberrant branched filopodia, suggesting that IRSp53(T340,360A) lacks negative regulatory cues.

The branched protrusions that are reported in cells expressing IRSp53 fragments including the IMD (70) were assessed by live-cell microscopy (see Fig. S2c in the supplemental material); most protrusions were not dynamic and more closely resembled retraction fibers. The black arrowheads designate two points of attachment to the substratum. While the protrusion slowly moved, these attachments remained fixed.

Filopodium dynamics are modulated by 14-3-3 binding to IRSp53.

The ability of IRSp53 to promote filopodia is established (63), if not the underlying protein interactions. To address how 14-3-3 might modulate this function, we investigated the number of filopodia in HeLa cells expressing various GFP-tagged versions of IRSp53 by live-cell imaging. In order to quantify the numbers of filopodia produced by these constructs, cells expressing low levels of GFP proteins were recorded for a period of 20 min at 10-s intervals. Analysis at the 5-, 10-, and 15-min time points was done, but not of the initial condition to avoid bias. The number of new filopodia per 50 μm of plasma membrane was then assessed. This analysis was conducted for 5 individual cells over 3 experiments. Cells expressing wild-type GFP-IRSp53 had 2-fold more new filopodia versus GFP (Fig. 8b). Cdc42-binding-defective IRSp53(I267N) and 14-3-3 binding-defective IRSp53(T340,360A) were both able to increase filopodium density, suggesting these mutants drive production of filopodia equally. Expression of IRSp53 with an inactive SH3 domain, IRSp53(I402P), did not induce additional filopodia, however.

Although IRSp53(T340,360A) did not enhance the effect of IRSp53 on filopodium numbers, their lifetimes were extended. IRSp53-positive filopodia were imaged by time lapse microscopy (Fig. 8c). The times taken for filopodia to emerge and fully retract were binned into three groups: <1 min, 1 to 3 min, and >3 min. It has previously been suggested that average filopodial lifetimes are in the range of 1 to 3 min (54). The lifetimes of most (73%) of the HeLa filopodia in IRSp53-expressing cells were in the 1- to 3-min range (Fig. 8c). IRSp53(I267N) was similar in this respect; in contrast, IRSp53(T340,360A)-expressing cells exhibited increased filopodial lifetimes, with >50% stable for longer than 3 min. Thus, binding of 14-3-3 to IRSp53 promotes retraction of filopodia (Fig. 8c). These observations fit with the importance of SH3 interactions for IRSp53 function and are in agreement with the antagonism between 14-3-3 and the flanking domains.

In summary, we have uncovered a mechanism by which IRSp53 phosphorylation in turn regulates both Cdc42-GTP binding to the CRIB and the abilities of certain proline-rich proteins to bind the SH3 domain (Fig. 8d). This in turn affects the dynamic properties of IRSp53 and therefore its site of action in the cell.

DISCUSSION

There has been considerable progress in uncovering the underlying biochemistry of filopodium assembly. Components of the presumptive “filopodial-tip complex” include myosin X, Mena/Vasp, mDia2, and IRSp53 (22), which in various ways contribute to the initiation and elongation of filopodia. In some cases, small interfering RNA (siRNA) experiments clearly demonstrate a requirement for these proteins. In this context, IRSp53 appears to have a nonredundant role of linking this tip complex to membranes (63) to coordinate membrane tubulation with actin elongation. We showed here that the closely related IRTKS cannot play the same role because of differences in SH3 binding (Fig. 6).

Myosin X has been suggested to initiate the formation of filopodia by localizing to the barbed ends of actin filaments via its motor domain and merging these filaments together via oligomerization (75). These actin microspikes appear to elongate into filopodia through the actions of Mena/Vasp, which bind myosin X (74), and more importantly, the actin nucleator Dia2 (85). We suggest that IRSp53 may be recruited to the lamellipodium and potentially the filopodium by its SH3-binding partners, such as Eps8, Mena/Vasp, and/or Dia2 (Fig. 8d). The binding of 14-3-3 to two conserved phosphothreonines can block access to the SH3 domain. Concentration of IRSp53 at the filopodia is likely permissive for further elongation. It seems possible that Mena (38), Vasp, and the formin Dia2 (this paper) compete for SH3 binding, suggesting multiple mechanisms for IRSp53 recruitment. Loss of IRSp53 following 14-3-3 binding is consistent with the reported loss of IRSp53 localization at filopodium tips during retraction (54).

IRSp53 SH3 mutants are reported to be effective in driving filopodium formation in combination with activated Cdc42 (82), suggesting some redundancy in protein localization cues. In the absence of elevated Cdc42-GTP, we found that IRSp53 is mislocalized without SH3 function (Fig. 5) and is unable to accumulate at lamellipodia. Many reports indicate that IRSp53 induces protrusions (19, 27, 38, 46, 51, 52, 67, 70) via its IMD, which generates few dynamic filopodia but rather predominantly nondynamic membrane tubules that often lack filopodium-based markers, such as myosin X (42). Full-length IRSp53 promotes bona fide (dynamic) filopodia, which are in any case present in most cultured cells when viewed by live-cell microscopy. Thus, the IMD alone is sufficient to tubulate membranes (62) when highly expressed but fails to concentrate at the cell edge, where filopodia are generated.

Previous studies of IRSp53 SH3 function based on a F428A/P429A mutation, which has reduced binding to Mena (38), is, however, wild type with respect to Vasp and WAVE2 binding (our unpublished observations). In contrast, the SH3 mutation used in this study (I402P) is null and phenocopies SH3 deletion. It has been suggested that the SH3 domain is required for the synergistic formation of filopodia by IRSp53 and Mena (38). The notion that the SH3 domain is autoinhibited by associating with a proline-rich region adjacent to the CRIB motif and relieved by Cdc42-GTP binding (38) is not supported by any experiments we have performed. Truncated constructs containing the SH3 domain (but lacking the CRIB and proline-rich region) do not bind any better to targets such as Dia2 (data not shown). Similarly, a point mutation in the CRIB motif (I267N) that blocks Cdc42 binding has no effect on IRSp53 SH3 binding.

Nakagawa et al. (54) have suggested that the SH3 domain of IRSp53 is not required for membrane or lamellipodial localization. This inconsistency with our data might be due to the different structures under analysis in their study and ours. In their paper, they analyzed the localization of IRSp53 to membrane ruffles in actively migrating cells. In our study, we looked at the steady-state localization in HeLa cells that have well-formed and biochemically characterized lamellipodia (73). Our work emphasizes the importance of the SH3 domain for IRSp53 localization and function. Loss of SH3 function blocks the proper localization of IRSp53 to the lamellipodium and thus its targeting to filopodia (clearly demonstrated in Video S3 in the supplemental material). This, then, would block its biological activity in promoting filopodia [Fig. 8b; cells expressing IRSp53(I402P) do not have any more filopodia than GFP-expressing cells]. Surprisingly, the related IRTKS, with a highly conserved IMD but without the SH3 regulatory 14-3-3 binding sites, does not localize to lamellipodia. While it is suggested that Cdc42 binding positively regulates the SH3 domain (38, 67), we provide evidence (Fig. 4) that 14-3-3 association blocks binding, which might be an alternate means of IRSp53 association with filopodia. The lamellipodial recruitment of IRSp53 is enhanced by increased Eps8 levels (Fig. 7), consistent with the importance of the SH3 domain in IRSp53 for targeting to this region (Fig. 5). A body of work indicates that Eps8 binds to IRSp53 with high affinity (due to the dimeric nature of both proteins) and that a loss of Eps8 leads to a concomitant loss of IRSp53-induced filopodia (19). While Eps8 localizes very selectively to the lamellipodial region, it is not clear if it is required for the localization of IRSp53 to filopodia (although the Eps8 protein is also found in the tip complex). Thus, it is conceivable that IRSp53 may also be targeted to filopodia by distinct binding partners. Two potential candidates are Mena/Vasp family members and Dia2. Mena (38, 84) and Vasp (this paper) both associate with IRSp53 and play roles in filopodium production (6, 7, 39, 40, 48). Dia2 has recently emerged as an important component of filopodia (57, 85) and robustly binds the SH3 domain of IRSp53 (Fig. 6).

Since 14-3-3 binding to IRSp53 is sensitive to LiCl, a GSK3β-dependent kinase is clearly involved, although we have ruled out direct phosphorylation, which would require an upstream priming site(s). Connection with Cdc42 pathways has been noted; Cdc42V12 can inactivate GSK3β via PKCζ phosphorylation (21). GSK3β activity is inhibited during wound healing (10), when Rac1 and Cdc42 are activated, which could promote IRSp53-mediated filopodia at the leading edge. Since IRSp53 activity is required for neurite outgrowth and growth cone motility (27), GSK3β-stimulated growth cone collapse may in part arise from IRSp53 inhibition. It is also worth noting that Eps8 has been shown to recruit Dishevelled (Dsh) to the plasma membrane and filopodia (61). Since Dsh is a known inhibitor of GSK3β, Eps8 may promote IRSp53-mediated filopodia in two distinct manners: first, by recruiting IRSp53 to the membrane, and second, by inhibiting GSK3β activity at the sites of filopodia to maintain IRSp53 in an active non-14-3-3-bound state. Inhibition of a number of kinases (PKA, PKC, PAK, ROK, PI3K, mTOR, and ERK) (Fig. 1c and data not shown) did not affect 14-3-3 binding. Thus, this key modulator of IRSp53 has yet to be identified, and because the kinase is sensitive to glucose, we are currently focusing on the potential role of AMPK.

In summary, our model for IRSp53 function in filopodium dynamics is shown in Fig. 8d. IRSp53 is dynamically associated with the lamellipodial region by association with its SH3-binding partners, such as Eps8. Binding to Cdc42-GTP may help to stabilize the open (non-14-3-3-bound) state of IRSp53, perhaps by preventing kinase access. Thus, IRSp53 would more likely be recruited to emerging filopodia under the influence of either Eps8 or filopodial-tip complex proteins, such as Mena/Vasp or Dia2. This would conveniently couple actin elongation with membrane tubulation. It has not escaped our notice that phosphorylation of the conserved IRSp53 Y338, followed by SH2 binding, would also antagonize 14-3-3. Termination of IRSp53 function is suggested to occur following Cdc42 dissociation, kinase phosphorylation of T340 and T360, and subsequent 14-3-3 binding, which competes for SH3 partners, thus allowing filopodial retraction.