Characterisation of IRTKS, a novel IRSp53/MIM family actin regulator with distinct filament bundling properties

The actin cytoskeleton is central to the movement, morphology and adhesion of eukaryotic cells (Pollard and Borisy, 2003). Actin filaments are used in a diverse array of cellular contexts and numerous actin binding proteins and regulators control the polymerisation and arrangement of filaments. In motile cells, polymerisation of actin at the leading edge drives protrusion of the membrane. Protrusive actin structures include lamellipodia, which are broad, flat sheet-like projections, and filopodia, which are thin, needle-like projections (Small et al., 2002). Lamellipodia consist of a dense network of short, branched actin filaments, and our understanding of how these structures are formed have seen a dramatic increase in recent years (Svitkina and Borisy, 1999). Central to lamellipodia formation is the Arp2/3 complex, which nucleates actin filaments while bound to the side of existing filaments, resulting in a branched filament network (Mullins et al., 1998). Signals are transduced from extracellular stimuli to the Arp2/3 complex by a pathway including the small GTPase Rac and Scar (WAVE) proteins (Bompard and Caron, 2004; Machesky and Insall, 1998). Filaments in lamellipodia are short due to the presence of capping protein, which binds to the barbed end of growing filaments shortly after nucleation, preventing further growth (Pollard and Borisy, 2003). In contrast to lamellipodia, filopodia consist of a combination of long, unbranched filaments arranged in tight, parallel bundles (Svitkina et al., 2003) and shorter unbundled filaments near the tips (Medalia et al., 2007). We currently do not have a clear understanding of the signals that control the formation of filopodia. Activation or overexpression of the GTPase Cdc42 induces filopodia in some cell types, but the key downstream effectors of Cdc42-induced filopodia are not known (Lamarche et al., 1996; Nobes and Hall, 1995). There is increasing evidence that Ena/VASP proteins are involved in filopodia formation, acting by preventing filament capping, resulting in the production of long parallel filaments (Bear et al., 2002; Lebrand et al., 2004; Mejillano et al., 2004). Diaphanous related formins (DRF) proteins are also key controllers of filopodia formation (Pellegrin and Mellor, 2005; Schirenbeck et al., 2005a; Schirenbeck et al., 2006; Schirenbeck et al., 2005b). In addition, numerous actin bundling and crosslinking proteins control the arrangement of actin filament protrusions (Revenu et al., 2004; Vignjevic et al., 2006). Bundling is a crucial determinant of the mechanical properties of actin protrusions (Gardel et al., 2004; Xu et al., 1998). Different actin structures contain different complements of bundling proteins, for instance fascin is found within filopodia and filamin A within lamellipodia (Flanagan et al., 2001; Kureishy et al., 2002). Like the proteins that control actin assembly, bundling proteins are subject to regulation and it is becoming apparent that in order to generate specific actin structures, actin assembly and bundling must be co-ordinately controlled (Revenu et al., 2004; Vignjevic et al., 2006).

Insulin receptor substrate of 53 kDa (IRSp53) was originally identified as a phosphorylation substrate for the insulin receptor, however, subsequently research has focused on its role in regulating the actin cytoskeleton (Yeh et al., 1996). Several groups have observed that overexpression of IRSp53 results in the formation of filopodia-like protrusions, and IRSp53 localises to the tips of both filopodia and lamellipodia (Govind et al., 2001; Krugmann et al., 2001; Nakagawa et al., 2003; Yamagishi et al., 2004). IRSp53 is a modular protein containing several protein interaction domains, including an SH3 domain, which binds to several important actin regulators including Scar2/WAVE2 and the Ena/VASP protein, Mena (Krugmann et al., 2001; Miki et al., 2000). IRSp53 also possesses distinct binding sites for the two Rho family GTPases, Rac and Cdc42, leading to the proposal that IRSp53 functions as an adaptor, transducing signals from these GTPases to cytoskeletal regulating proteins (Krugmann et al., 2001; Miki et al., 2000). However, recent work has suggested that IRSp53 also possesses effector functions (Yamagishi et al., 2004). The N-terminal 250 amino acids of IRSp53 consist of a domain conserved in five mammalian proteins including the previously identified actin regulator missing in metastasis-B (MIM-B), hence the name IRSp53/MIM homology domain (IMD). The crystal structure of the IRSp53 IMD reveals it to be a zeppelin-shaped dimer with one F-actin-binding site per monomer (Millard et al., 2005). In isolation the IMD can bundle actin filaments and induce filopodia-like protrusions, however, mutation of the actin binding sites abrogates protrusion formation by both the isolated IMD and by full-length IRSp53 (Millard et al., 2005). The structure of the IMD is closely related to that of the BAR domain, a function of which is to tubulate membranes (Gallop and McMahon, 2005). Intriguingly, recent data has indicated that the IRSp53 IMD also has the capacity to tubulate membranes and it has been suggested that this activity contributes to protrusion formation (Suetsugu et al., 2006). Like IRSp53, MIM-B can bundle filaments via its IMD, however, outside the IMD, MIM-B has little sequence conservation with IRSp53, suggesting a functional divergence (Yamagishi et al., 2004). Unlike IRSp53, MIM-B does not possess an SH3 domain, instead it contains an actin monomer binding WASP homology 2 (WH2) domain and a region that binds receptor protein tyrosine phosphatase δ (Gonzalez-Quevedo et al., 2005; Mattila et al., 2003; Woodings et al., 2003). Overexpression of MIM-B results in the formation of structures resembling actin microspikes and lamellipodia, but not long, filopodia-like protrusions as observed on overexpression of IRSp53 (Mattila et al., 2003; Woodings et al., 2003). Yamagishi et al. reported that five mammalian proteins contain IMDs, however, of these, only IRSp53 and MIM have been studied, the remaining three being predicted proteins, based on cDNA sequences (Yamagishi et al., 2004). Of these three proteins, two are closely related to IRSp53 and one to MIM-B. In this study we have characterised insulin receptor tyrosine kinase substrate (IRTKS), one of the two IRSp53-related IMD proteins. We find that IRTKS has properties clearly distinct from those of IRSp53.

In a search for IRSp53-related proteins we identified a human cDNA sequence (accession number: NM_018842) containing an open reading frame, translation of which would result in a 511 amino acid protein with 39% identity and 59% similarity to human IRSp53 (also known as BAI1-associated protein isoform 2 or IRSp53-S, NM_017450). This sequence has previously been identified by Yamagishi et al. who referred to it as IRTKS in sequence alignments (Yamagishi et al., 2004). Homologous sequences have also been identified for several other species, including mouse (NP_080109, 87% identity), chicken (XP_414749, 70% identity) and zebrafish (AAH68330, 52% identity). The human IRTKS gene is located on chromosome 7 at q21.3–q22.1, whereas IRSp53 is on chromosome 17 at q25.

Alignment of human IRTKS with IRSp53 demonstrates that the region corresponding to the IMD of IRSp53 is well conserved, as are the SH3 domain and WW domain interaction motif (Fig. 1). Outside of these regions conservation is low, with IRTKS notably lacking the partial CRIB domain which mediates binding of Cdc42 to IRSp53. The C terminus of IRTKS has no similarity to that of the well-characterised short (S) splice variant of IRSp53, however, it does have homology to the C-termini of two other IRSp53 splice variants, the L (long) and M (medium) forms (Fig. 1B) (Miyahara et al., 2003; Yamagishi et al., 2004). Note that unless otherwise stated, IRSp53 refers to the S splice variant in this manuscript.

We used peptides based on the sequence of human and mouse IRTKS (see Materials and Methods) to raise a polyclonal antibody, which recognised a band of 60 kDa in extracts from the mouse myoblast cell line C2C12 (see Fig. S1A in supplementary material). Immunoprecipitations from C2C12 extract were performed using this antibody and the precipitated material subjected to SDS-PAGE. A band of 60 kDa was excised from the gel and analysed by mass spectrometry after trypsin digestion. Three of the five peptide sequences derived were found to match the murine form of the novel protein (Fig. S1B in supplementary material). The tissue distribution of IRTKS was studied by immunoblotting murine tissue extracts. Western blots were inconclusive as the IRTKS antibody produced a high background signal in some tissues (not shown), equally, attempts at northern blotting failed. We could, however, detect IRTKS by immunoprecipitating IRTKS from the tissue extracts using the IRTKS antibody covalently linked to protein G beads, and then blotting the immunoprecipitated material. Clear expression of IRTKS was observed in bladder, liver, testes, heart and lung, and trace amounts were found in spleen, brain and skeletal muscle (Fig. S1C in supplementary material). None was detected in intestine or kidney. This shows that IRTKS has widespread tissue distribution in mouse. IRSp53 was originally identified as a substrate for the insulin receptor (Yeh et al., 1996), so we tested whether IRTKS shared this characteristic. COS7 cells were co-transfected with Myc-IRTKS and the insulin receptor β-subunit. The transfected cells were stimulated with insulin and Myc-IRTKS immunoprecipitated from the extract, and probed for the presence of phosphotyrosine by immunoblotting. We observed tyrosine phosphorylation of IRTKS, which was dependent on the presence of the insulin receptor and stimulation with insulin, indicating that IRTKS is an insulin receptor substrate (Fig. S1D in supplementary material).

IRSp53 has been shown to bind to Rac via a site within the N-terminal 230 amino acids and also to Cdc42 via a partial CRIB motif located between residues 268 and 280 (Govind et al., 2001; Krugmann et al., 2001; Miki et al., 2000). The Rac binding site is within the IMD, which is well conserved between IRSp53 and IRTKS, whereas the Cdc42 binding site is found within a region of poor conservation between the two proteins. Binding to the two GTPases was studied by GST pulldown assay using Myc-tagged IRSp53 and IRTKS expressed in COS7 cells and GST-GTPases expressed in E. coli. Full-length IRSp53 bound to the constitutively active L61 mutant of Cdc42 but not the dominant negative N17 mutant, whereas binding to both Rac mutants could be detected (Fig. 2). The Rac-binding site on IRSp53 was located within the N-terminal 241 amino acids, and the Cdc42-binding site was within amino acids 242-521, consistent with published data. Full-length IRTKS bound similarly to both L61 and N17 Rac mutants, but no binding to either Cdc42 mutant was detected (Fig. 2). As observed for IRSp53, binding of IRTKS to Rac was mediated by the N-terminal 240 amino acids. The GTPase binding characteristics of endogenous IRTKS were also studied using extract from C2C12 cells. Binding to both Rac mutants, but not to either Cdc42 mutant was detected, consistent with the results obtained with the exogenously expressed Myc-IRTKS.

We compared the effects on the actin cytoskeleton of expression of IRTKS and IRSp53 in COS7 cells. Expression of IRSp53 results in the formation of numerous long, often wavy filopodia-like extensions, to which IRSp53 is localised (Fig. 3A,B) (Govind et al., 2001; Krugmann et al., 2001; Millard et al., 2005; Yamagishi et al., 2004). Overexpression of IRTKS resulted in a clearly distinct actin phenotype. In cells expressing relatively low levels of Myc-IRTKS numerous small actin microspikes were observed at the cell periphery (Fig. 3C,E). There was also a notable clearing of F-actin from the cytoplasm, which could be the result of reorganisation of the filaments or depolymerisation, but it was not possible to determine the cause. At higher expression levels, the actin microspikes induced by IRTKS appeared to coalesce into clusters of brightly stained protrusions (Fig. 3D,F). In cells expressing the highest IRTKS levels, the majority of cellular F-actin was observed within these actin clusters. In confluent monolayers of IRTKS-expressing cells, the actin clusters were notably abundant along cell junctions (Fig. 3G,H). In all cases the Myc-IRTKS was localised to these actin structures, but not necessarily concentrated there. Dense clusters of F-actin such as those observed in Myc-IRTKS-expressing cells were never observed on overexpression of IRSp53. Expression of a construct lacking the IMD of IRTKS failed to induce formation of the actin clusters (data not shown), indicating that the IMD is required for formation of the clusters. Although it would also be desirable to study the effects of knockdown of IRTKS expression on the actin cytoskeleton, we have, thus far, been unable to achieve more than about 50% knockdown, which does not result in a detectable phenotype (data not shown).

Despite a high level of sequence similarity, overexpression of IRSp53 and IRTKS resulted in very different effects on the actin cytoskeleton. We reasoned that this difference in behaviour may be caused by a region of sequence divergence between the proteins. A notable difference between the form of IRSp53 studied here and IRTKS is the presence of an extension at the C terminus of IRTKS (Fig. 1). As previously noted, this C-terminal region has sequence similarity with the WH2 domain, a common actin monomer-binding motif. This C-terminal region is similar to, but distinct from, that of the L and M forms of IRSp53 (Fig. 1), which we address later in this manuscript. An IRTKS construct lacking the C-terminal 24 amino acids was expressed in COS7 cells. Expression of this construct did not result in the formation of actin clusters, as was observed for full-length IRTKS. Instead, this construct induced the formation of long filopodia, reminiscent of those induced by IRSp53 (Fig. 3I,J). This suggested that this C-terminal (Ct) extension of IRTKS is important for the formation of the actin clusters. We prepared full-length IRSp53 fused to the Ct extension of IRTKS (amino acids 483-511). Expression of this construct (IRSp53+Ct) resulted in the formation of actin clusters similar to those observed on expression of full-length IRTKS (Fig. 3K,L). These data suggest that the unique Ct extension of IRTKS is important in controlling the effect of IRTKS on the actin cytoskeleton and appears to convert a filopodia-inducing activity into an actin cluster-forming activity.

The L isoform of IRSp53 has a C-terminal region similar to that of IRTKS, so we tested the effect of expression of this isoform. We found that expression IRSp53-L resulted in a phenotype indistinguishable from that of IRSp53-S (Fig. 4). In addition to (or perhaps because of) the different changes that expression of IRSp53 and IRTKS effect on the actin cytoskeleton, the cell shape also is affected differently by these two proteins. Cells expressing IRSp53 S or L isoforms (Fig. 4) typically have long thin filopodial extensions and a diminished cell surface area, as measured by drawing around the periphery of the cells (Fig. 4). The area of IRSp53 (S or L forms)-expressing cells was roughly half of the control (GFP alone), IRTKS or IRSp53+Ct-expressing cells. This could be because the long filopodial extensions make up half of the area in IRSp53-expressing cells, but we were not able to quantify these long protrusions. An alternative explanation is that IRSp53 expression reduces the surface area by membrane tubulation or cell retraction.

To further characterise the actin clusters observed with IRTKS expression, we also examined the location of VASP, vinculin, Arp2/3 complex and cortactin in IRTKS-expressing cells. VASP is typically found in filopodia, focal adhesions and lamellipodia, whereas cortactin localises in lamellipodia, and vinculin is found in focal complexes and focal adhesions, so these proteins served as markers of various cytoskeletal structures. Fig. 5A-D shows that VASP localises abundantly in IRTKS-induced actin clusters. Cortactin (Fig. 5E-H) and Arp2/3 complex (not shown), markers for dynamic actin networks, were also highly enriched in these structures. Vinculin was present in patches coinciding with IRTKS-induced actin clusters, but did not generally co-localise to the extent of the other proteins, indicating that perhaps the clusters contained some focal-complex-like structures, but were not major sites of focal adhesion formation (Fig. 5I-L). We also observed similar co-localisation for IRSp53+Ct expression for VASP (Fig. S2A-D in supplementary material) and vinculin (Fig. S2E-H in supplementary material), cortactin (not shown) and Arp2/3 complex (not shown), indicating that the structures formed by IRSp53+Ct and IRTKS expression were of similar composition and appearance.

Since it was previously shown that the protrusion-generating ability of IRSp53 depends on four basic residues near the distal poles of the IMD (Millard et al., 2005), we mutated the four corresponding lysine residues in IRTKS IMD: K141E, K142E, R145E, K146E and found that this protein could no longer induce actin clusters (Fig. S3 in supplementary material).

The IMD of IRSp53 and MIM-B have previously been demonstrated to bundle actin filaments in vitro (Millard et al., 2005; Yamagishi et al., 2004). To test whether the corresponding region of IRTKS bundles actin, amino acids 1-249 (IMD) were expressed as a GST-fusion in E. coli, and purified to homogeneity after cleavage from the GST. Bundling was then assessed using a low speed co-sedimentation assay in which F-actin was incubated with varying concentration of IRTKS IMD and then centrifugation at 10,000 g. The presence of IRTKS IMD resulted in a concentration-dependent increase in actin pelleting, indicating filament bundling by IRTKS IMD (Fig. S4A in supplementary material). This demonstrates the amino acids 1-249 of IRTKS constitutes a functional IMD. In addition, we wished to study the effect of the IRTKS Ct extension on the in vitro bundling characteristics of IRTKS, but we were unable to generate purified full length IRTKS recombinantly. We thus tested the ability of a shortened version of IRTKS, termed IMD-Ct, composed of the IMD of IRTKS (amino acids 1-253) fused to the Ct extension (amino acids 483-511). We did not find any difference in the bundling ability of IMD-Ct compared with IMD alone (Fig. S4B in supplementary material).

We wished to study the effect of the IRTKS Ct extension on the in vitro bundling characteristics of IRTKS. The IMD-Ct recombinant protein was incubated with actin filaments fluorescently labelled with Cy3 and the resulting mixture was then visualised using a fluorescence microscope. We and others have previously used this assay to observe bundles formed by the IMD of IRSp53 (Millard et al., 2005; Yamagishi et al., 2004). The IMD of IRTKS resulted in the formation of long bundles, frequently greater than 100 μm in length, which were similar to those we have previously observed using the IMD of IRSp53 (Fig. 6A) (Millard et al., 2005). By contrast, the bundles induced by the IMD-Ct were noticeably shorter (usually <20 μm) and were frequently found in small clusters (Fig. 6B). The same was observed if the Ct extension of IRTKS was fused to the IMD of IRSp53 (data not shown). Without added bundling protein, the individual filaments were too small and disordered to resolve (Fig. 6C). This demonstrates that the Ct extension of IRTKS modifies the actin bundling characteristics of the IMD. Notably, the effect of the Ct extension on bundling in vitro mirrors that observed in vivo, in that it switches from induction of long straight bundles (such as those found in long filopodia) to short, clustered bundles. We were unable to produce recombinant IMD fused to the C-terminal region of IRSp53, so we cannot rule out the possibility that addition of this short extension might also alter bundling characteristics of the IMD. We did not observe any qualitative differences between IRTKS IMD and IRTKS + Ct IMD in standard low-speed pelleting assays to measure actin bundling (Fig. 6D,E), nor was this data noticeably different from that of IRSp53 IMD (Millard et al., 2005) or MIM IMD (Bompard et al., 2005). It would be an interesting next step to characterise the IRTKS bundles by electron microscopy as has been done for IRSp53 IMD (Yamagishi et al., 2004).

As previously noted by Yamagishi et al. alignment of the Ct extension of IRTKS with other proteins shows that there is clear similarity between the Ct extension of IRTKS and the WH2 domain located at the C terminus of MIM, as well as the WH2 domain of WASP interacting protein (WIP) (Fig. 7A). WH2 domains are actin monomer-binding motifs found within many actin regulators (Paavilainen et al., 2004; Paunola et al., 2002). Mutational analysis of the WH2 domain of thymosin-β4 identified four residues that are critical for actin monomer binding and are conserved in all WH2 domains (Paunola et al., 2002; Van Troys et al., 1996). The Ct extension of IRTKS lacks one of these residues, an invariant isoleucine, suggesting it may not function as a conventional WH2 domain (Fig. 7A). We prepared a construct consisting of the Ct extension of IRTKS (aa 483-511) fused to the C terminus of GST (GST-Ct) to test for actin monomer binding. As a control, a construct consisting of the WH2 domain of Scar/WAVE1 (residues 496-528) fused to the C terminus of GST (GST-ScarWH2) was used. In vitro, WH2 domains sequester monomers and this results in an increase in the apparent critical concentration, which is the lowest actin concentration at which polymerisation can occur. The critical concentration can be measured by recording the fluorescence of a range of concentrations of pyrene actin at equilibrium (Carlier et al., 1993). The addition of GST-ScarWH2 to actin resulted in a clear increase in the apparent critical concentration from 0.17 μM ± 0.01 to 0.65 μM ± 0.1 (± s.d.; based on data from three experiments; Fig. 7B). By contrast, the GST-Ct had no significant effect on the apparent critical concentration (0.18 μM ± 0.06), suggesting no sequestration of monomers. This indicates that the Ct extension of IRTKS is not a monomer sequestering WH2 motif. To address whether there might be direct actin monomer binding between this WH2 motif and G-actin, we attempted to synthesise peptides corresponding to the WH2 to allow `in-solution' actin binding studies to fluorescent peptides. However, the peptides proved difficult to synthesise, precluding this analysis. Instead, we used a native gel shift assay with GST-Ct and GST-ScarWH2 to determine their ability to shift G-actin from its normal migration on the gel (Costa et al., 2004). Whereas 10 μM GST-ScarWH2 shifted 2.5 μM G-actin quantitatively on this assay, 10 μM GST-Ct caused only a slight smearing of the actin band and shifted very little to the higher mobility position (data not shown). We performed this experiment three times with similar results and we can conclude that GST-Ct likely has a very weak affinity for actin monomers, which may explain why it does not effectively sequester them. Further experiments to measure the affinities of the WH2 motifs in the context of the full-length proteins will need to be performed to completely resolve this issue.

Since WH2 proteins have also been shown to be actin-filament-binding proteins in some cases (Irobi et al., 2004), we next tested the ability of the GST-Ct and GST-ScarWH2 to bind actin filaments using a co-sedimentation assay. Owing to the monomer-sequestering activity of GST-ScarWH2, the presence of this construct resulted in a reduction in the proportion of actin that was polymerised. To achieve consistent levels of F-actin we therefore stabilised filaments with phalloidin. This did not appreciably affect binding of either construct to filaments. We found that GST-Ct co-sedimented with actin filaments in a concentration-dependent and saturable manner, whereas little GST-ScarWH2 co-sedimented with filaments (Fig. 7C). A K_d of approximately 1 μM was obtained for the binding of GST-Ct to filaments. This suggests that the Ct extension of IRTKS may act as an F-actin binding motif.

The identification and characterisation of a new conserved actin bundling and membrane interacting domain, the IMD, has considerably progressed our understanding of how the IRSp53 and MIM proteins function (Millard et al., 2005; Yamagishi et al., 2004). In this study, we describe the characterisation of IRTKS, an IRSp53-related protein that possesses an IMD and a WASP homology WH2 motif. We propose that IRTKS has a wide tissue distribution and different biochemical properties to IRSp53. Because IRSp53 has several splice variants and isoforms, it is important to carefully define probes used for detection of related proteins. To avoid confusion, we have specifically determined that our antibody does not react with IRSp53 (not shown) and that it does immunoprecipitate a protein matching the sequence of IRTKS, using Fourier transform ion cyclotron resonance (FTICR) mass spectrometry. Therefore, we present evidence that although IRSp53 has been shown to be highly enriched in brain tissues (Govind et al., 2001; Okamura-Oho et al., 2001), IRTKS is low or nonexistent in brain but is present in bladder, liver, testes, heart and lung tissue. Like IRSp53, IRTKS can be tyrosine phosphorylated in an insulin-signalling-dependent manner.

The IMD regions of IRTKS and IRSp53 appear to have largely the same properties biochemically and when expressed in cells. We found that the N-terminal 249 amino acids of IRTKS bundle actin filaments similarly to the IRSp53 IMD and similarly to MIM IMD (Bompard et al., 2005). The IMD of IRTKS also binds to Rac, in both the GTP and GDP forms. Sequence conservation among IRSp53, IRTKS and MIM reveals a cluster of basic residues that are important for the actin bundling (Millard et al., 2005), which also appears to be important for the interaction with Rac (Bompard, 2005) and probably with phosphoinositides (Suetsugu et al., 2006).

Expression of the IMD of MIM, IRSp53 (Yamagishi et al., 2004) or IRTKS (this study) produces a similar effect of filopodia-like structures protruding from the periphery of expressing cells. However, expression of full-length IRSp53, IRTKS and MIM cause clearly distinct changes in actin cytoskeletal organisation, suggesting that the divergent regions of these proteins modulate their behaviour. Whereas IRSp53 induces long filopodia-like protrusions, IRTKS expression results in clusters of short actin clusters around the cell periphery. These clusters co-localise with endogenous cortactin, Arp2/3 complex and VASP, and to a lesser extent contain some vinculin. We find that a short extension at the C terminus of IRTKS is important in producing this effect. Removal of the C-terminal extension, causes IRTKS to induce long filopodia-like protrusions, reminiscent of those induced by IRSp53, rather than actin clusters. Conversely, fusing the C terminus of IRTKS on to IRSp53 causes actin clusters rather than filopodia to be induced by IRSp53. Notably, expression of IRSp53-L, a splice variant of IRSp53 with a C-terminus homologous to that of IRTKS, does not result in the formation of actin clusters, and it would be interesting to identify precisely which residues are responsible for the divergent behaviours of these two proteins. We find that, when fused to the IMD, this extension also affects IMD-induced bundling in vitro, causing the formation of shorter, often clustered bundles. We attempted to test whether the shorter clusters might be formed as a result of severing or depolymerisation activity of IRTKS, using pyrene-labelled actin fluorescence depolymerisation assays, but we could not obtain reliable data, perhaps due to solubility problems with the IMD proteins at low salt concentrations (Millard et al., 2005). When IMD-Ct is expressed in cells, similar structures to those observed with IRTKS expression are also formed. This suggests that the C-terminal extension is important for the actin organising function of IRTKS. The Ct extension of IRTKS has sequence similarity to the WH2 motif, an actin monomer binding sequence (Paunola et al., 2002). However, we find that the Ct extension binds to F-actin but does not sequester monomers and only very weakly shifts the actin mobility in native gel assays (data not shown). This is not without precedent, since one of four WH2 domains found within tetrathymosinβ has been shown to bind F-actin, but only relatively poorly to G-actin (Van Troys et al., 2004). Furthermore, espins are actin-bundling proteins that contain a WH2 motif, which although it binds to G-actin also aids in bundle formation (Loomis et al., 2006).

It is not yet clear how the filament-binding activity of the Ct extension contributes to bundling or organisation of actin. One possibility is that interaction with a second filament binding site outside the IMD alters the orientation of bundled filaments to produce non-parallel, poorly ordered bundles. This could explain why IRTKS frequently induces clustered structures rather than distinct, parallel bundles. Another possibility is that filament binding by the Ct extension causes severing or destabilisation of filaments. Binding of ADF and/or cofilins to F-actin causes a small, destabilising twist in the filament, which leads to depolymerisation (Paavilainen et al., 2004). Filament severing would give rise to shortened bundles (Maciver et al., 1991).

The regulation of IRSp53 and IRTKS by GTPases remains largely unexplored. Interestingly, IRSp53 binds to Cdc42, a known activator of filopodia formation, whereas IRTKS does not. However, the interaction with Cdc42 is not required for filopodia induction, as the IMD alone will produce filopodia-like protrusions when expressed in cells. All IMD proteins have numerous protein interaction motifs and as such act as scaffolds as well as effectors. Interactions with upstream regulators could recruit IMD proteins to particular cellular locations, for instance to filopodia by Cdc42 or to lamellipodia by Rac. The lack of a Cdc42-binding site in IRTKS may suggest that it plays a role in regulating lamellipodia, rather than filopodia. The relative importance of the bundling activity of the IMD with the activities of recruited binding partners or other regions of the proteins has not yet been determined, and will no doubt be the subject of future studies.

Monoclonal antibodies 9E10 (anti-Myc) and PY72 (anti-phosphotyrosine) and protein G beads were obtained from Cancer Research U.K. Anti-p34-Arc of the Arp2/3 complex and anti-cortactin were obtained from Upstate Biotechnology, VASP273 was a kind gift from Juergen Wehland (Braunschweig, Germany). Alexa Fluor 488-conjugated anti-mouse secondary antibody was obtained from Molecular Probes. Unless otherwise stated all chemicals and reagents were purchased from Sigma-Aldrich UK.

COS7 and C2C12 cells were grown in DMEM containing 10% FBS (Gibco-BRL), supplemented with penicillin and streptomycin at 37°C, 5% CO₂. Cells were transfected using GeneJuice (Novagen) according to the manufacturer's instructions and fixed or lysed 24 hours later. Cells were fixed using 4% formaldehyde, blocked with 50 mM NH₄Cl and permeabilised with 0.1% Triton X-100 as previously described (Millard et al., 2005). After incubation in appropriate antibodies, coverslips were mounted in Mowiol and antifade (p-phenylenediamine) as previously described (Millard et al., 2005).

cDNA encoding IRTKS (accession no. BC013888) was obtained from MRC gene service (Cambridge, UK). The coding sequence was amplified by PCR and cloned into pRK5Myc (Lamarche et al., 1996). IRSp53pRK5Myc and GTPase constructs in pGEX2T were gifts from Dr Alan Hall (Krugmann et al., 2001). Truncated forms of IRSp53 and IRTKS were generated by PCR amplification and products cloned into pRK5Myc for mammalian expression or pGEX4T2 (Amersham Pharmacia) for bacterial expression. Insulin receptor β subunit expression construct was a gift from Jeremy Tavare (University of Bristol, UK). Enzymes were purchased from New England Biolabs.

Peptides CLIEISSTHKKLNESLD and CIEYVETVTSRQSEIQK (peptide 2 is identical in mouse and human, whereas peptide 1 has one S to T change at position 15 in mouse) corresponding to residues 75-90 and 161-176, respectively, of human IRTKS, were synthesised by Alta Biosciences (Birmingham UK). Peptides were coupled to maleimide-activated keyhole limpet haemocyanin (Pierce) according to the manufacturer's instructions and conjugates were used to immunise rabbits (Eurogentec). The same peptides were coupled to Affigel 10 (Bio-Rad) and used to affinity purify antisera.

Four 10 cm dishes of confluent undifferentiated C2C12 cells were lysed in 150 mM NaCl, 20 mM, Tris-HCl pH 7.5, 1% N-octyl-β-D-glucopyranoside supplemented with the protease inhibitors chymostatin, leupeptin, antipain, pepstatin A (all 1 μg/ml) and 1 mM phenylmethylsulfonyl fluoride. Lysates were incubated with 20 μg of the affinity purified IRTKS antibody bound to protein G beads. Following several washes with lysis buffer samples were boiled in sample buffer and analysed by SDS-PAGE. Following staining with Coomassie Blue, a 60 kDa band was excised. Trypsin digestion of the sample and mass spectrometry using a MicroMass Q-TOF Global was carried out by University of Birmingham functional genomics and proteomics laboratories.

Tissues were isolated, washed in ice-cold PBS and then dounce homogenised in 25 mM HEPES pH 7.6, 150 mM NaCl, 0.5% Triton X-100, 1 mM EDTA and protease inhibitors (as above). Following centrifugation at 53300 g for 1 hour at 4°C supernatants were frozen at –80°C. Protein concentrations were determined using the Bio-Rad protein assay. IRTKS antibodies were covalently coupled to protein G beads using dimethyl pimelimidate as previously described (Harlow and Lane, 1988). An equal volume of beads was then incubated with 1.5 mg of each tissue extract for 2 hours at 4°C. Beads were washed with homogenisation buffer, boiled in sample buffer and analysed by immunoblotting.

COS7 cells were transfected with Myc-IRTKS and the insulin receptor β-chain. Cells were stimulated 24 hours later with 100 nM insulin, then, 10 minutes later were lysed on ice in 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 1 mM Na₂VO₄ and protease inhibitors (as above). Lysates were immunoprecipitated with 9E10 bound to protein G beads, and bead-bound material was then analysed by immunoblotting.

GST-fused GTPases were expressed in E. coli and purified on glutathione-agarose beads. Levels of bead-bound protein were estimated by SDS-PAGE analysis. Myc-IRSp53-p60 constructs were transfected into COS7 cells, which, 24 hours later, were lysed in 20 mM Tris-HCl pH 7.5, 0.5% NP-40, 50 mM NaCl, 5% glycerol, protease inhibitors (as above). An equal volume of cell lysate was added to each set of GST-GTPase beads. 20 μg of GST-GTPase was used per pulldown. Beads were rotated for 1 hour at 4°C and they were then washed with lysis buffer, boiled in Laemmli buffer and subjected to SDS-PAGE and western blotting.

Actin was prepared from rabbit muscle as previously described (Spudich and Watt, 1971). Actin was pyrene labelled as previously described (Mullins and Machesky, 2000). F and G actin binding assays were performed in F buffer: 100 mM KCl, 2 mM MgCl₂, 0.2 mM EGTA, 2 mM Tris pH 8.0, 0.2 mM ATP, 0.5 mM DTT, 0.2 mM.

GST-Ct and GST-ScarWH2 were expressed in E. coli, purified on gluathione-agarose and eluted with 20 mM glutathione, 50 mM Tris-HCl pH 8.0. Co-sedimentation assays were performed as previously described (Millard et al., 2005). 2.5 μM GST-Ct or GST-ScarWH2 were mixed with varying concentrations of F-actin (0-12 μM) stabilised with a twofold molar excess of phalloidin. Equal fractions of pellet and supernatant were analysed by SDS-PAGE and levels quantified using NIH image. Proportions pelleting were calculated and corrected for pelleting in the absense of actin.

A range of actin concentrations (0.025-1.3 μM) containing 5% pyrene-labelled actin were mixed with 2.5 μM GST-Ct or GST-ScarWH2 or with F-buffer and incubated at 4°C for 16 hours. Fluorescence was then recorded using a PTI C60 spectrophotometer (Photon Technology International) at excitation and emission wavelengths of 370 nm and 407 nm, respectively. Graphs were plotted and linear regression performed using Microsoft Excel.

Rabbit muscle actin at 2.5 μM was mixed with fusion proteins (GST, GST-Scar/WAVE1WH2 or GST-IRTKSWH2 at a range of 5-10 μM) in F-buffer. The mixture was centrifuged at 100,000 g for 30 minutes and the supernatant fraction was loaded on a 7.5% native gel as previously described (Costa et al., 2004; Palmgren et al., 2001). The gel was stained with Coomassie Blue dye and a shift upwards in the mobility of the G-actin was considered to indicate binding.

IRTKS IMD and IRTKS IMD-Ct were produced and in E. coli and purified as described previously for IRSp53 IMD (Millard et al., 2005). Low speed co-sedimentation assays were performed as previously described, except that centrifugation was performed at 10 000 g instead of 16 000 g (Millard et al., 2005). Bundling of fluorescent filaments was performed as previously described (Millard et al., 2005).

We thank Guillaume Bompard for careful reading of the manuscript and helpful suggestions throughout this study. We thank Tim Dafforn for help with the linear regression analysis. We thank Anthony Jones and Helen Cooper for the FTICR analysis of IRTKS. This work was supported by an MRC Senior fellowship to L.M.M. (G117/399). J.D. is supported by an AICR grant to L.M.M. (04-009).