A novel actin bundling/filopodium-forming domain conserved in insulin receptor tyrosine kinase substrate p53 and missing in metastasis protein.

Insulin receptor tyrosine kinase substrate p53 (IRSp53) has been identified as an SH3 domain-containing adaptor that links Rac1 with a Wiskott-Aldrich syndrome family verprolin-homologous protein 2 (WAVE2) to induce lamellipodia or Cdc42 with Mena to induce filopodia. The recruitment of these SH3-binding partners by IRSp53 is thought to be crucial for F-actin rearrangements. Here, we show that the N-terminal predicted helical stretch of 250 amino acids of IRSp53 is an evolutionarily conserved F-actin bundling domain involved in filopodium formation. Five proteins including IRSp53 and missing in metastasis (MIM) protein share this unique domain and are highly conserved in vertebrates. We named the conserved domain IRSp53/MIM homology domain (IMD). The IMD has domain relatives in invertebrates but does not show obvious homology to any known actin interacting proteins. The IMD alone, derived from either IRSp53 or MIM, induced filopodia in HeLa cells and the formation of tightly packed parallel F-actin bundles in vitro. These results suggest that IRSp53 and MIM belong to a novel actin bundling protein family. Furthermore, we found that filopodium-inducing IMD activity in the full-length IRSp53 was regulated by active Cdc42 and Rac1. The SH3 domain was not necessary for IMD-induced filopodium formation. Our results indicate that IRSp53, when activated by small GTPases, participates in F-actin reorganization not only in an SH3-dependent manner but also in a manner dependent on the activity of the IMD.


INTRODUCTION
Insulin receptor tyrosine kinase substrate p53 (IRSp53) 1 , also known as brain-specific angiogenesis inhibitor 1 associated protein 2, is a multifunctional adaptor protein enriched in the central nervous system (1)(2)(3). The protein contains a unique N-terminal 250 amino acid stretch, a half-Cdc42/Rac interactive binding (CRIB) motif, a proline-rich domain, a Src homology 3 (SH3) domain, and a WW domain-binding motif (WWB). IRSp53 is directly regulated by Rho family small GTPases, Rac1 and Cdc42, and provides a molecular link between these GTPases and the actin cytoskeleton regulators Wiskott-Aldrich syndrome protein (WASP) family verprolin homologous-protein 2 (WAVE2) and mammalian enabled (Mena), which are involved in the formation of lamellipodia (4,5) and filopodia (6,7). Active Cdc42 binds to the half-CRIB motif (6,7),whereas Rac1 binds to the unique N-terminal domain (8). The association of Rac1 or Cdc42 is proposed to liberate the C-terminal SH3 domain masked intra-molecularly by its N-terminus, thereby allowing the SH3 domain to interact with its binding partners (4,7,9). Thus, the SH3 domain is thought to be essential for IRSp53-mediated actin reorganization. However, the N-terminal half of IRSp53 lacking the SH3 domain was reported to induce neurite outgrowth in a neuroblastoma cell line (6) and filopodia in B16 melanoma cells (10), suggesting that IRSp53 promotes actin reorganization independently of SH3-domain-mediated inter-molecular interactions.
Recently, a novel monomeric actin binding protein, missing in metastasis protein (MIM), containing a WASP homology 2 (WH2) domain in the C-terminus, was reported in human and mouse (11)(12)(13), and found to share the unique N-terminal domain with IRSp53 (13). We found that the N-terminal domains of MIM and IRSp53 also share other characteristic features; the predicted secondary structures are almost purely helical (see (9) for IRSp53) and the estimated isoelectric points are around nine. MIM induces actin cytoskeleton reorganization in cultured cells. This activity is not dependent on the C-terminal half (13), suggesting that the N-terminal half containing the IRSp53 homologous domain plays a key role in actin reorganization.
Here we show that IRSp53 and MIM belong to an evolutionarily related protein family sharing a well-conserved N-terminal helical domain (IMD, IRSp53/MIM homology domain) as a key constituent. We investigated the role of the IMD in actin reorganization. Our results indicate that the IMDs of IRSp53 and MIM induce filopodia in cultured cells and form tightly packed F-actin bundles in vitro. The filopodium-forming activity of the IMD in full length IRSp53 is regulated by small GTPases. Thus, upon association with active Rac1 or Cdc42, IRSp53 can induce actin cytoskeleton reorganization by dual mechanisms: The SH3-mediated recruitment of F-actin regulators and the action of the novel actin bundling domain in the N-terminus. Both mechanisms may work synergistically or additively in controlling cortical actin dynamics.
F-actin binding and bundling assays-F-actin was prepared from rabbit skeletal muscle as described (17). Glutathione S-transferase (GST)-fusion proteins of various fragments of IRSp53 and MIM (Fig. 1B) were expressed in BL21-Star (DE3) cells (Invitrogen Corp.), purified using glutathione-Sepharose (Amersham Biosciences), then buffer-exchanged into F-buffer (25 mM Hepes, pH 7.5, 100 mM KCl, 0.2 mM CaCl 2, 2 mM MgCl 2 , 2 mM EGTA, 0.2 mM ATP, 1 mM dithiothreitol) containing 0.1% C 12 E 8 (Nikko Chemicals, Tokyo, Japan). For binding assays, purified GST-fused fragments were clarified by centrifugation at 400,000 × g for 15 min to remove any aggregates, mixed with F-actin in the F-buffer, and incubated for 30 min on ice. The final concentration of the GST-fusions and F-actin were 1.2 µM and 5 µM (as for G-actin), respectively. The mixture was then centrifuged as above and equal aliquots of the supernatant and the pellet were analyzed by SDS-PAGE followed by Coomassie Blue staining. For quantitative analysis of F-actin binding and bundling, the IMDs were cleaved out from the GST-fusions expressed by pGEX-6P3 vectors using PreScission Protease (Amersham Biosciences) and further purified by cation-exchange chromatography (Resource S, Amersham Biosciences). To quantify F-actin binding, increasing amounts of F-actin was incubated with 2 µM IRSp53-IMD or MIM-IMD in the F-buffer for 3 h at room temperature. The samples were then centrifuged and analyzed as above. Protein bands were quantified by densitometry (Personal Densitometer SI, Amersham Biosciences). For quantitative bundling assay, increasing amounts of the IMDs were incubated with 1 µM F-actin in the F-buffer for 1 h at room temperature. The supernatant and the pellet were separated by low-speed centrifugation (10,000 × g for 30 min) and analyzed as above.

The N-terminal helical domain is evolutionarily conserved in IRSp53 family proteins and MIM
family proteins-IRSp53 and MIM share the N-terminal stretch of 250 amino acids (22% identical/18% similar) while the remaining parts of the molecules show only marginal similarity.
To explore whether this similarity is based on real homology, we searched the GenBank database for proteins having similar sequences. First we found three more genes encoding homologous N-terminal sequences in the human genome: Insulin receptor tyrosine kinase substrate (IRTKS), a hypothetical gene FLJ22582 and ABBA-1 ( Fig. 1A; see Supplemental Table 1 for detail).
IRTKS and FLJ22582 are IRSp53-related proteins containing an SH3 domain in the C-terminal half. However, both of them lack the half-CRIB motif found in IRSp53 and FLJ22582 further lacks the WWB (PPPXY) (Fig. 1B). ABBA-1 is a MIM-related protein which possesses a WH2 domain in the C-terminus. Further database searches have shown that each of these five proteins has a putative ortholog in chicken and zebra fish, indicating that they are well conserved through vertebrate evolution . As pointed out previously (9,13), related proteins are also found in invertebrates (C. elegans M04F3.5 protein and D. melanogaster CG32082 protein). In an amino acid sequence alignment of the N-terminal region of these proteins ( Fig.   1A), clusters of basic amino acids, proline, glycine and clusters of hydrophobic amino acids are well conserved. There is a signature sequence of ALxEE [RK][RG]RFCx(0,1)F[IL] in the C-terminal half of the stretch. As expected from the number of basic amino acid clusters in this domain, the estimated isoelectric points are highly basic, ranging from 8.5 for human MIM to 9.2 for human FLJ22582.  M04F3.5 and CG32082, respectively. Helix-breaking amino acid residues at the four breaking sites of the IRSp53-related proteins are also conserved in MIM/ABBA family proteins (asterisks in Fig. 1A).
Thus, all of these proteins appear to have a common segmentation pattern of helices, helix I to V ( Fig. 1A). Although human IRSp53 lacks helix V, it is predicted to be present in the chicken ortholog.
A phylogenetic tree ( Fig. 1C) based on the alignment of the IMDs shows that the vertebrate IRSp53/MIM family is divided into two major groups: The IRSp53 sub-family and the MIM/ABBA sub-family. The putative invertebrate homologs are positioned between them. The tree of the IMDs exactly reflects the hierarchy of domain composition of these proteins. The IRSp53 sub-family members contain an SH3 domain and the MIM/ABBA sub-family proteins contain a WH2 domain. The vertebrate SH3-containing sub-family is further divided into three groups according to the presence or absence of the WWB and the half-CRIB motif. These data suggest that the IRSp53/MIM family originated from a common ancestor and diverged through evolution. This hypothesis is supported by the fact that IRTKS and FLJ22582 but not M04F3.5 or CG32082 share highly homologous C-termini with the MIM/ABBA sub-family members (Supplemental Fig. 6). Our analyses suggest the presence of an evolutionarily conserved IRSp53/MIM family and that the IMDs are the key components for the functional roles of proteins belonging to this family.

The IMDs of IRSp53 and MIM induce filopodia in HeLa cells-
To explore the functional roles of the IMD, we first examined the morphological effects of ectopic expression of the IRSp53-IMD and the MIM-IMD in HeLa cells. Cells expressing the GFP-tagged IMD of IRSp53 formed numerous long filopodia that were F-actin rich as demonstrated by Rhodamine-phalloidin staining ( Fig. 2, a, a' and b, b'). The MIM-IMD also induced filopodia but these were reduced in length (Fig. 2,c,c' and d,d'). In addition, MIM-IMD promoted the formation of microvillus-like protrusions on the apical cell surface. IRSp53-IMD and MIM-IMD localized to and occasionally were concentrated in these protrusions (arrows in Fig. 2,c' and d'). Both IMDs appeared not to be associated with stress fibers. There were no obvious signs of enhanced lamellipodial activity or disruption of stress fibers in these IMD-expressing cells. GFP used as a negative control did not induce any morphological changes (Fig. 2, e and f). Truncated fragments of IMD, IRSp53-N-IMD (aa 1-161) and IRSp53-C-IMD (aa 105-250) could not stimulate filopodium formation (data not shown). These data indicate that both IMDs are capable of inducing filopodia in cells. Since IRSp53 and MIM represent the most divergent members of the vertebrate IRSp53/MIM protein family (Fig. 1C), the filopodium-inducing activity of the IMD is likely to be conserved in all family members.
IMD does not act upstream of Rac1 or Cdc42 for filopodium formation-Actin cytoskeletal reorganization is often a hallmark of Rho family GTPases. Previous reports have shown that Rac1 binds to the N-terminus of IRSp53 (8) and Cdc42 binds to the aa 202-305 fragment containing the half-CRIB motif (6). Therefore, we examined whether Cdc42 or Rac1 activation was involved in IMD-induced filopodium formation. The formation of numerous filopodia induced by IRSp53-IMD was not perturbed by the co-expression of dominant negative Cdc42 or Rac1 (Fig. 3). There was no quantitative difference in the ratio of filopodium forming cells among HeLa cells transfected with IRSp53-IMD alone, those transfected with IRSp53-IMD and Cdc42N17, and those transfected with IRSp53 and Rac1N17 ( Fig. 3,d), suggesting that the IMD itself is not regulated by these small GTPases. This result also excludes the possibility that the domain functions upstream of these small GTPases.

The filopodium-inducing IMD activity of wild-type IRSp53 is regulated by Cdc42 and
Rac1-The common mechanism of effecter activation by Rho family GTPases appears to be dependent on the disruption of intra-molecular autoinhibitory interactions. Cdc42-induced conformational changes have also been demonstrated for the molecule containing a half or semi-CRIB motif, Par6 (19). First we found that GFP-tagged IRSp53-WT or GFP-tagged IRSp53-∆SH3, when expressed in moderate levels, could not induce filopodia ( Fig. 4A, a, b, and c). As reported earlier (7,10), cells expressing very high amounts of IRSp53 often formed dendritic extensions accompanied with sever retraction of the cell body. As noted in Fig. 2 legend, these cells were omitted from our analyses. Next, we examined whether the IMD function was regulated by Cdc42 and Rac1 in IRSp53-WT and in IRSp53-∆SH3 containing the half-CRIB motif. Co-expression of the active Cdc42 with these IRSp53 constructs led to massive formation of wavy filopodia (IRSp53+Cdc42 phenotype as shown in Fig. 4A, d and e) that was clearly distinguishable from straight filopodia induced in cells co-expressing GFP and active Cdc42 (Cdc42 phenotype as shown in Fig. 4A, f). A similar level of filopodium induction, mixed with Rac1-dependent enhanced lamellipodia activity (Fig. 4A, i) was induced by the co-expression of active Rac1 (Fig. 4A, g and h). These results suggest that the SH3 domain is not necessary for IMD-dependent filopodium formation. Our results also suggest that the filopodium-inducing IMD activity in wild-type IRSp53 is regulated by Cdc42 and Rac1. The central region of IRSp53 containing the half-CRIB motif appears to be essential for this regulation, as previously suggested for the regulation of the SH3 domain (7,9).
To further confirm that the IMD-induced filopodium formation is independent of SH3-binding molecules, we used two non-functional SH3 mutants, IRSp53-W/R and IRSp53-FP/AA (7). Both mutants could induce filopodia when expressed with active Cdc42 ( In vitro F-actin bundling activity of IMD-The filopodium-promoting activity of the IMDs of IRSp53 and MIM in cultured cells led us to examine whether these IMDs have F-actin binding and bundling activity. We examined F-actin binding/bundling activity of the GST-fused IMD and other fragments, and also tag-free purified IMDs in vitro. As shown in Fig. 5A, GST-fused IRSp53-IMD, IRSp53-∆SH3 and MIM-IMD, but not GST were co-sedimented with F-actin in a high-speed assay (total binding). To exclude the possible contribution of GST-tag or contaminating bacterial proteins to F-actin binding and bundling, the activities of purified tag-free IMDs (Fig. 5B, left panel) were examined. In the high-speed assays, the IMDs of IRSp53 and MIM bound to F-actin in a concentration-dependent and saturable manner (Fig. 5B, right panel). The apparent half-maximum concentrations of F-actin for IMD binding were almost the same, 0.5 µM, irrespective of the variation between the maximum extents of these IMDs, suggesting that both IMDs have roughly the same affinity to F-actin. Low levels of the maximum extent of bound IMDs, about 30% for IRSp53-IMD and 20% for MIM-IMD, can be explained by improper protein folding of the bacterial-made IMDs or their denaturation during the purification process.
The GST-fusions capable of F-actin binding induced thick F-actin bundles (Fig. 5C).
Although GFP-tagged IRSp53-∆SH3 required activation by Rac1 or Cdc42 for filopodium formation, GST-fused IRSp53-∆SH3 alone could induce F-actin bundling. It is possible that the bacterial-made protein may not be folded properly to form the self-inhibitory conformation.
IRSp53-∆SH3 showed stronger bundling than IRSp53-IMD or MIM-IMD, however, different levels of bundling activity among these proteins may simply reflect differences in stability of these fusion proteins. To quantify the bundling activity, the tag-free IMDs were examined in low-speed sedimentation assay (Fig. 5 D). The bundling activity was concentration-dependent and most of F-actin could be incorporated into bundles in high concentrations of the IMDs.
The IMD-induced F-actin bundles could be seen under a phase contrast microscope and their thickness was measured at 0.1 to 0.2 µm by electron microscope observation of negatively stained materials (Fig. 6A). Observation of thin sections of the bundles revealed tight packing of parallel actin filaments in the bundles (Fig. 6B). The bundle as a whole was not a paracrystal in which actin filaments were packed into a hexagonal array with a constant spacing of 11.5 nm, as previously described (20). However, actin filaments in the bundles tended to be arranged in a line and partly packed into a hexagonal pattern (Fig. 6B, inset). The center-to-center distance between neighboring actin filaments aligned in a line was nearly constant and was measured at 11.2 nm in transverse sections (Fig. 6C). These observations indicate that IRSp53 acts as a typical parallel-actin-bundle-forming molecule such as fimbrin and fascin and suggest that the IRSp53/MIM family is a novel actin bundling protein family.
Self-association of IMD-Actin bundling activity requires at least two independent F-actin binding sites or a combination of one binding site plus a self-association site in an actin bundling domain. The ability of IMDs to form dimers or oligomers was examined by chemical cross-linking using a zero-length cross-linker EDC. The apparent molecular weight of the tag-free IMDs in SDS-PAGE was progressively shifted from 30 kDa, which matched to the calculated molecular weights of the IRS-IMD (28,972) and MIM-IMD (28,640) including a short linker sequence, to the dimer one of 60 kDa (Fig. 7A). Both of the tag-free IMDs were effectively cross-linked into dimers but not into trimers or tetramers. This result suggests that the purified IMDs are present as dimers. Next we examined the IMD's self-association in cultured cells. The IMD of IRSp53 or MIM could associate with each other and with the full-length molecule but not with the C-terminal half lacking the IMD (∆IMD) in co-transfected 293T cells (Fig. 7B). Consistent with the IMD-dependent self-association, ∆IMD was not co-immunoprecipitated in any combinations whereas the full length IRSp53 and MIM associated with themselves. These results indicate that the IMD is a self-associating domain and suggest that IRSp53 and MIM can be present as dimeric forms in mammalian cells.

DISCUSSION
In this study we show that the N-terminal helical domain, the IMD, identified in IRSp53 and MIM, induces filopodium formation in vivo and F-actin bundling in vitro and suggests that these domains are conserved in an evolutionarily related protein family, the IRSp53/MIM family.
We propose that the IRSp53/MIM family is a novel F-actin bundling protein family which  (10) have shown that IRSp53 is able to self-localize in filopodia using its N-terminus, levels of accumulation appear not to be high. Considering that actin bundling proteins require a relatively high molar ratio to actin in order to function, this level of specificity may not be sufficient to support dynamic behavior of the cell periphery in non-transfected cells. Both WAVE2 and Mena are shown to localize at the filopodial tip (33-35), again suggesting the functional redundancy of these protein complexes with increased specificity of localization.
Here we show that the activity of the IMD is tightly regulated by Rac1 and Cdc42, in a similar manner to that of the SH3 domain (4,7). Our results suggest that the central region of IRSp53, including the half-CRIB motif, is essential for the autoinhibition of the IMD. The N-terminus aa 1-178 of IRSp53 has been shown to interact with the region around the half-CRIB motif and inhibit binding of the SH3 domain to Mena (7). The autoinhibitory mechanisms of the IMD and the SH3 domain may work together within the same molecule. Conversely, F-actin association of the IMD and the SH3-ligand binding are likely to activate or stabilize each other.
We propose that IRSp53 is a direct effecter of Cdc42 and Rac1, acting in concert with various partner proteins recruited by the SH3 domain. Further analyses are required to evaluate the activities of various IRSp53/partner protein complexes and their specific roles in the regulation of cortical actin dynamics. Although MIM has been shown to interact with protein tyrosine phosphatase delta (13), its regulation remains unknown. Our present study reveals that the IMDs are highly conserved both structurally and functionally. So far we have not found any apparent sequence homology of this domain with known F-actin interacting proteins. Future work including crystallographic studies will be needed to ascertain precise molecular mechanisms for F-actin bundling by the IMDs as well as to clarify their regulation, especially by small GTPases in IRSp53.         NT_010498.13: join (19441131..19441197, 19436376..19436431, 19436201..19436273, 19435377..19435461, 19435192..19435283, 19435036..19435110, 19434720..19434728, 19433658..19433815, 19433198..19433305, 19429712..19429934, 19420872..19420946, 19420326..19420502, 19420004..19420169, 19419038..19419855). All the exons except one keep AG/exon/GT boundaries and the exon 13 uses GC instead of GT. Northern blot analysis of human ABBA-1 showed a single 6 kb band in various tissues with the highest expression level in the brain.

Cloning of Zebrafish IRSp53-Zebrafish
IRSp53 was PCR-amplified from Zebrafish adult brain cDNAs using a pair of primers predicted from the zebrafish cDNAs and the genomic sequences (An UniGene cluster Dr.16661, WGS 126238745 and a genomic contig BX323027

Cloning of Zebrafish ABBA-1 and -2-Zebrafish
ABBA-1 and -2 were similarly cloned from the brain cDNAs using two pairs of PCR primers predicted from the zebrafish database. We found two N-terminal and two C-terminal sequences that are more closely related to ABBA-1 than MIM. A forward primer (F1) 5'TTCTTCCAGCCGATTTTTGTTCATTCATTGGGAGCAG was predicted from ESTs AL914911 and AL914912 and the other one (F2) 5'TTGGGCTGGTAAAATGGATGCGGGAATGG was predicted from genomic sequences BX530017 and BX548167. Similar to human ABBA-1, the last exons of zebrafish putative ABBAs encode one third of the whole proteins, so that we could obtain two different relatively long 3'-sequences from AI959191 and overlapping WGSs and from BQ419430 and WGSs. The contains 710 aa and shows 66% identity/9% similarity to human ABBA-1 and 51% identity/12% similarity to human MIM. ABBA-2 contains 737 aa that shows 73% identity/10% similarity to human ABBA-1 and 56% identity/11% similarity to human MIM.

Database searches for putative IRSp53/MIM family proteins in vertebrates-We found five
candidates proteins in the human genome, IRSp53, IRTKS, FLJ22582, MIM and ABBA-1. To reveal the conservation of these proteins in vertebrates, we intensively searched Gallus gallus (chicken) sequences for the bird, Xenopus laevis (african clawed frog) and Silurana tropicalis (western clawed frog) sequences for the amphibia, and Danio reio (zebra fish) sequences for the fish. Basically overlapping EST clones were used to make putative full-length mRNA sequences.
Mismatches and possible sequence errors were corrected by overlapping ESTs and genomic sequences, mainly whole genome shotgun (WGS) sequences. We have found that the exon compositions of these genes are almost completely conserved and more than five WGSs could be obtained for most of the exons. GenBank accession numbers of representative clones used are sited below.