Kette regulates actin dynamics and genetically interacts with Wave and Wasp

The actin cytoskeleton plays a central role in cell motility, morphology,phagocytosis and cytokinesis. It is spatially and dynamically reorganized,providing force for the shape change and surface movement in most eukaryotic cells (Borisy and Svitkina,2000; Pantaloni et al.,2001). Rearrangement of actin is evoked rapidly by extracellular stimuli, and sets of actin-associated proteins are thought to act cooperatively in the polymerization, crosslinking and anchorage of actin filaments. Although the central dynamic nature of the actin system is known,actin organization and regulation are less well understood.

Actin dynamics crucially depend on the ability of the protein to switch from a monomeric (G-actin) to a filamentous form (F-actin). Polymerization of F-actin starts with the de novo nucleation of an actin trimer, a process that occurs relatively slowly and requires the action of the Arp2/3 complex(Machesky et al., 1994; May, 2001; Robinson et al., 2001; Welch et al., 1997). Subsequent elongation is fast and cells have to prevent spontaneous actin polymerization by expressing a variety of actin-binding proteins such as profilin (Schafer and Cooper,1995). The nucleation activity of the Arp2/3 complex in turn is regulated by a set of activators, such as the members of the Wasp(Wiskott-Aldrich syndrome protein) and Wave (Scar - FlyBase) families(Higgs and Pollard, 2001; Suetsugu et al., 2002b; Takenawa and Miki, 2001). Wasp proteins are usually self-inhibiting and require small G proteins of the Rho family for activation (Kim et al.,2000; Miki et al.,1998a). In its activated GTP-bound form, Cdc42 can bind to the Crib (Cdc42/Rac Interactive Binding) domain of Wasp, releasing the auto-inhibition and thereby leading to the activation of the Arp2/3 complex(Higgs and Pollard, 2001; Rohatgi et al., 1999). By contrast, Wave, which does not bind Cdc42, is trans-inhibited through its association with members of the Kette (Hem - FlyBase) family, Sra1(specifically Rac associated 1) and Abi (Abelson-interactor). Rac1 binding,presumably to Sra1, relieves the inhibitory function of this complex(Eden et al., 2002; Kobayashi et al., 1998; Miki et al., 2000).

Both, Cdc42 and Rac1 exert distinct functions in rearranging the F-actin cytoskeleton, GTP-bound Rac1 promoting lamellipodia and activated Cdc42 promoting filopodia formation (Hall,1998). This matches the finding that Wasp enhances the generation of filopodia, whereas Wave activation near the cell membrane results in an increase of lamellipodia (Miki et al.,1998a; Takenawa and Miki,2001).

Drosophila possesses only two genes, wasp and scar, that encode a Wasp and Wave homolog, respectively(Zallen et al., 2002; Ben-Yaacov et al., 2001). Both genes act largely independently of each other and have a strong maternal component. Only embryos lacking both the zygotic and the maternal gene functions display severe embryonic nervous system phenotypes. In the adult, wasp is required for formation of external mechanosensory organs,whereas scar/wave is - unlike wasp - required for normal formation of the compound eye. Recent structure function analyses demonstrated that Drosophila Wasp can perform at least part of its function independent of Cdc42 (Tal et al.,2002). Alternative modes of Wasp activation such as phosphorylation or SH3 (Src homology 3) domain binding have been described(Cory et al., 2002; Scott et al., 2002; Suetsugu et al., 2002a).

Obviously, regulation of the dynamic F-actin cytoskeleton is pivotal especially during cell movement and thus requires a close link to the plasma membrane and guidance receptors involved in the perception of extracellular signals. The latter generally induce conformational changes in the cytoplasmic domain of membrane-anchored receptors, which recruits other proteins to the ligand-receptor complex. In the case of receptor tyrosine kinases (Rtk),autophosphorylation of tyrosine residues leads to the recruitment of SH2 domain containing adapter proteins. One of these is the SH2 SH3 adapter protein Nck, which is able to link several Rtks as well as guidance receptors like the Netrin-receptor DCC to the actin cytoskeleton(Li et al., 2001; Li et al., 2002). Nck can recruit additional proteins to the cell membrane via its three SH3 domains. Among the highly conserved Nck-interacting proteins are Wasp(Rivero-Lezcano et al., 1995),the non-receptor tyrosine kinase Abl (Adler et al., 2000) and Nap1 (Nck-associated protein 1), which is a member of the evolutionary conserved Kette family(Baumgartner et al., 1995; Hummel et al., 2000; Kitamura et al., 1996; Soto et al., 2002).

Based on its requirement for axonal growth, we have previously identified mutations in the Drosophila kette gene(Hummel et al., 2000). Loss of kette function primarily affects neurite growth and subsequently causes glial migration defects leading to a characteristic commissure phenotype in the embryonic CNS similar to the one observed in wasp^mat/zyg mutants(Zallen et al., 2002) and Drosophila Nck (dock) mutants(Desai et al., 1999; Garrity et al., 1996). Furthermore, kette genetically interacts with mutations in the Nck homolog dock as well as with the small GTPase Rac1, supporting the notion that Kette may provide a novel mechanism linking extracellular signals to the actin cytoskeleton (Hummel et al.,2000). Recently, the Kette homolog Nap1 was found in a large 500 kDa protein complex comprising PIR121/Sra1, Abi, HSPC300 and Wave that keeps Wave in an inactive state in vitro (Eden et al., 2002).

However, to date it is still unclear how Kette could regulate the organization of the actin cytoskeleton in vivo. As Kette has been predicted to be an integral membrane protein with six transmembrane domains(Baumgartner et al., 1995), it might serve as a receptor recruiting Nck/Dock or Rho-GTPases to the membrane. We present biochemical and genetic evidence revealing that Kette is found predominantly in the cytosol. Only a small amount of Kette is recruited to the plasma membrane. In vivo as well as in tissue culture models, Kette protein colocalizes with F-actin and co-sedimentation assays revealed a direct interaction with F-actin. Within the membrane, Kette accumulates at the insertion sites of large F-actin bundles, suggesting that targeted localization of Kette may be required for its function. Loss of Kette protein leads to a Scar/Wave-dependent accumulation of F-actin within the cell. Ectopic expression of wild-type or different truncated Kette proteins in tissue culture cells or during Drosophila development does not affect F-actin formation or viability. However, expression of membrane-tethered Kette efficiently induces ectopic bundles of F-actin in a process depending on Wasp but not on Scar/Wave. These data indicate that Kette fulfils a novel role in regulating F-actin organization by antagonizing Wave and activating Wasp-dependent actin polymerization.

All crosses were performed at 25°C unless otherwise indicated. The following strains were used: wasp¹/TM6 wasp³/TM6 (Ben-Yaacov et al., 2001); scar^k13811/CyO and scar^Δ37/CyO(Zallen et al., 2002); and kette^Δ2-6/TM6 and kette^J1-70/TM6 (Hummel et al., 2000). To determine the gain-of-function phenotype of Kette^Myr, we used the GAL4 system(Brand and Perrimon, 1993). Transgenic flies carrying the following constructs were generated following standard procedures: UAS-kette, UAS-kette^Myc;UAS-kette^1-374; UAS-kette^375-907;UAS-kette^908-1126 and UAS-kette^Myr. The Kette^Myr construct was made by fusing DNA encoding the first 88 amino acids from Drosophila Src1(Simon et al., 1985) to the first codon of Kette. Several independent transgenic lines were analyzed in each case.

To isolate proteins interacting with Kette protein, we inserted the kette as well as the wasp ORF in frame with the GAL4 DNA binding domain into the GBK-T7 vector (Clontech). With pGBK-Kette we screened a matchmaker Drosophila embryo cDNA library (1×10⁶clones tested) using the GAL4-based Two-hybrid System 3 from Clontech. pGAD-Abi contains the entire Abi-ORF.

The rabbit anti-Kette antibody (97/82), directed against a peptide derived from the middle region of Kette (652-666), KHFDDIRKPGDESYR, was made by Eurogentec (Belgium). In addition, polyclonal antibodies were generated against parts of Kette fused with a His₆-tag (Qiagen). pQE plasmids express amino acid regions 1-374, 375-906 and 907-1126 of Kette, respectively. Using these expression constructs, His₆-Kette fusion proteins were expressed in E. coli and purified with Ni-NTA resin (Qiagen) under denaturing conditions. Rabbits were immunized with purified proteins by Davids Biotechnologie (Germany).

Drosophila S2 cells were propagated in 1× Schneider's Drosophila media (Gibco) supplemented with 10% FBS, 50 units/ml penicillin and 50 μg/ml streptomycin in 75 cm³ T-flasks(Sarstedt) at 25°C. For transfections, 5×10⁵ cells were plated on glass cover slips (pretreated with fibronectin) in 24-well plates,cultured for 24 hours and transfected with Fugene 6 (Roche) at a transfection reagent:DNA ratio of 3:1. Cells were incubated 24 hours after transfection,fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, stained with Drosophila anti-Kette antibody 97/82 (1:2000 dilution) followed by Alexa-fluorophore-conjugated goat anti-rabbit IgG antibody (Molecular Probes, 1:1000 dilution). For filamentous actin and nucleic acid staining,cells were prepared as described above and incubated in 1-2 units Alexa-fluorophore-conjugated phalloidin (Molecular Probes) and in 300 nM DAPI(Molecular Probes), respectively. Tyrosine phosphorylated proteins were detected using the PT-66 anti-phosphotyrosine antibody (Sigma). The samples were mounted in a 25% (w/v) Mowiol (Sigma) solution containing DABCO (Sigma),and visualized on a Leica LSM.

570 bp of the kette ORF and 750 bp of the wasp and scar/wave ORF were inserted into the vector pLITMUS-28i (BioLabs)flanked by T7 promoters. After linearization, dsRNA was generated using a HiScribe RNAi Transcription Kit (BioLabs). dsRNA products were resuspended in water and annealed by incubation at 65°C for 30 minutes followed by slow cooling to room temperature and stored at -20°C. For RNAi treatment,5×10⁵Drosophila S2 cells in 0.5 ml serum-free medium were plated on fibronectin coated cover slips in 24-well plates(Sarstedt). dsRNA (5 μg) were added directly to the media followed by vigorous agitation. The cells were incubated for 30 minutes at 25°C followed by addition of 1 ml medium containing 15% FCS. dsRNA treated cells were incubated for additional 2-3 days and analyzed by immunofluorescence and western blot analysis.

Actin was purified from rabbit skeletal muscle following the methods of Spudich and Watt (Spudich and Watt,1971) and stored as G-actin at -172°C(Spudich and Watt, 1971). Actin was polymerized in a buffer of 100 mM KCl, 2 mM MgCl₂, 20 mM HEPES, pH 7.4, 0.5 mM β-mercaptoethanol and 2 mM NaN₃. Prior to use, MBP, MBPKette and actin were centrifuged at 200,000 gfor 30 minutes in a tabletop ultracentrifuge (Beckmann Instruments) to pellet any aggregated protein. For binding assays, increasing amounts of full-length or truncated Kette proteins were incubated with 3 μM F-actin for 30 minutes at room temperature in polymerization buffer. The mixtures were centrifuged at 200,000 g for 30 minutes at 4°C. Equal amounts of supernatant and pellet were separated by SDS-PAGE followed by Coomassie blue staining.

Fractionation of Drosophila embryonic and S2 cell extracts were performed as described (Zhang and Hsieh,2000). The pellets were solubilized in SDS sample buffer and supernatants were precipitated in trichloroacetic acid, followed by solubilization in SDS sample buffer. Western blot analysis was carried out as described previously (Bogdan et al.,2001).

Kette was predicted to be an integral membrane protein(Baumgartner et al., 1995). To test this, we generated polyclonal antisera against the Kette protein(Fig. 1A). Western blot analyses were performed on protein extracts of wild-type embryos and embryos overexpressing the full-length Kette protein under the control of the GAL4 system. All antisera recognized a 120 kDa protein, which is in good agreement with the predicted molecular mass of 125 kDa for Kette(Fig. 1B). The specificity of the antisera was further supported by the detection of a recombinant Myc-tagged Kette protein recognized in vivo by both anti-Kette and anti-Myc antibodies (Fig. 1B) and RNAi experiments (see Fig. 4A). No evidence for post-translational proteolytic cleavage was obtained.

To examine whether Kette is an integral membrane protein, we first analyzed its distribution in Drosophila cell extracts prepared from embryos or Drosophila S2-tissue culture cells by differential ultracentrifugation and flotation assays(Fig. 1C,D). In both cases the results were identical. After ultracentrifugation at 200,000 gKette was found predominantly in the cytosolic supernatant, whereas smaller amounts could be detected in the pellet containing membranes and large protein complexes. The particulate fraction of Kette was efficiently extractable by sodium carbonate treatment (0.1 M, pH 11.5), but it was retained in the 1%NP-40 insoluble cytoskeletal fraction (Fig. 1C). To further investigate how much Kette protein really binds to membranes, a flotation assay was performed. We fractionated post-nuclear supernatants from Drosophila embryo lysates by isopycnic centrifugation on a discontinuous sucrose density gradient. Unlike the integral membrane protein FasIII (Patel et al., 1987), some of the Kette protein could float out of the most dense sucrose during centrifugation to equilibrium, while most of the Kette protein remains in high dense sucrose (Fig. 1D). Taken together, these data show that Kette is not an integral membrane protein but small amounts are associated with membranes.

The recovery of Kette protein in the 1% NP-40 insoluble fraction containing F-actin and other cytoskeletal proteins suggest an association between Kette and F-actin. To test this directly, we determined a possible interaction of Kette with F-actin in a co-sedimentation assay. Soluble, maltose-binding protein (MBP) fusion proteins containing full-length Kette or its N-terminal third (Kette^1-374) were generated. Both fusion proteins, and unaltered MBP as a control, were subjected to ultracentrifugation either alone or in a mixture with F-actin. Only the fusion proteins pelleted with F-actin(Fig. 2A,B), indicating that actin binding is mediated by the N terminus of Kette, which, however, lacks known F-actin binding sequence motifs. Furthermore, we identified three different actin isoforms (Act42A, Act5C and Act57B) as putative interaction partners of Kette in a yeast two-hybrid screen using full-length Kette as bait(Fig. 2C). Based on sequence analyses of these functionally redundant actin isoforms(Wagner et al., 2002), Kette is able to bind to the C-terminal 110 amino acids of G-actin.

Unfortunately, none of the antisera generated detects endogenous Kette protein in whole-mount preparations of wild-type embryos. As many epitopes are sensitive to methanol treatment required to prepare Drosophilawhole-mount embryos, we studied the subcellular localization of the endogenous Kette protein in freshly dissociated cells from Drosophila embryos as well as in Drosophila S2 tissue culture cells(Fig. 3). In neuronal cells,which can be identified by expression of the HRP epitope(Jan and Jan, 1982), Kette is localized in the cytosol as well as in a punctuated pattern along the emerging neurites (Fig. 3A). In muscle cells, Kette is expressed in long fibers that resemble F-actin fibers. Indeed double staining experiments using phalloidin showed prominent co-localization with F-Actin (Fig. 3B). Both cell types are affected by the kette mutation(Hummel et al., 2000).

As shown by western blot analyses, Drosophila Schneider S2 tissue cells express high levels of Kette (Fig. 4A). When S2 cells are grown on fibronectin coated substrates,they adhere to the culture dish and adopt a characteristic pancake shape(Fig. 3C,D). The majority of the Kette protein is found in the cytosol around the nucleus; however, small amounts of Kette can be detected at the leading edge of lamellipodia-like extensions (Fig. 3C,arrowheads). Increased levels of Kette expression were noted on spoke-like structures extending from the nucleus to the membrane(Fig. 3C, arrows). Highest Kette concentration is generally found close to the cell membrane overlapping with tyrosine-phosphorylated proteins, which mark focal contact sites(Fig. 3D)(Parsons et al., 2000). F-actin is similarly distributed as Kette and the spokes showing high Kette expression correspond to thick F-actin bundles. At the cell margin, however,Kette appears to be localized closer to the membrane than the cortical F-actin network. In line with the above mentioned fractionation results this again indicates that some Kette protein is associated with the membrane.

The above data showed that Kette can be detected at the membrane and it is able to bind to F-actin. First evidence for a regulatory function of Kette in F-actin organization stems from our phenotypic analysis of kettemutants (Hummel et al., 2000). Loss of kette function affects the organization of the F-actin cytoskeleton and is characterized by an excess of disorganized F-actin bundles. To better analyze the function of Kette in F-actin dynamics we reduced kette expression in Drosophila S2 cells by RNA interference (see Materials and Methods). As judged by western blot and immunofluorescence analyses, treatment of S2 cells with a 570 bp dsRNA fragment generated from the kette-coding region resulted in a marked reduction of Kette protein expression (compare Fig. 4A with Fig. 5A,B). No interference was seen with sense and antisense RNAs (data not shown). With increasing time after dsRNA treatment, cells with reduced levels of Kette show dramatic alterations in cell morphology and concomitant changes of the F-actin cytoskeleton. Within 3 days, cells appear to collapse and accumulate large amounts of F-actin. Two days after RNAi treatment, intense granular F-actin structures are observed which are similar to those in kette-null mutants (Hummel et al., 2000)(Fig. 5B, arrowhead).

These data led to the question how Kette influences actin dynamics. F-actin formation is controlled by the action of the Arp2/3 complex, whose activity is in turn regulated by members of the Wasp/Wave family. Drosophila has a single Wasp as well as a single Wave homolog. Recently, the Kette homolog Nap1 was found in a large 500 kDa protein complex comprising PIR121/Sra1, Abi and HSPC300 that keep Wave inactive in vitro(Eden et al., 2002). In agreement with this work disruption of kette function leads to an excess formation of cytoplasmic F-actin. In contrast to kettedepletion, treatment of S2 cells with scar dsRNA leads to a reduced F-actin formation in the cytosol. Only some cortical F-actin is detected(Fig. 5C). Interestingly,disruption of Scar/Wave function resulted in a fairly uniform distribution of Kette in the cytosol and spoke-like F-actin structures cannot be recognized any more (Fig. 5C). This suggests an intact actin cytoskeleton is required to position Kette in the cell and may point to the functional relevance of the F-actin binding properties of Kette. Treatment of S2 cells with both scar/wave and kette dsRNA suppresses the excess formation of cytoplasmic F-actin induced by kette dsRNA alone. In all cell shape changes appeared slightly more severe compared with those evoked via scar dsRNA alone(Fig. 5C,D).

Further support for a Kette mediated repression of Scar/Wave activity stems from genetic analyses. Wild-type embryos are characterized by clearly separated anterior and posterior commissures, which is due to the migrating midline glia (Fig. 6A). Embryos that lack zygotic kette function display a characteristic CNS phenotype and commissures appeared fused(Fig. 6B). By contrast, loss of zygotic Scar/Wave expression does not affect embryonic nervous system development (Zallen et al.,2002) (Fig. 6D). To test whether the kette mutant phenotype might due to an upregulation of scar/wave activity we removed one copy of the scar/wavelocus in a kette background, which indeed suppressed the homozygous kette phenotype and distinct commissures are now recognizable(Fig. 6C). Taken together,these data suggest that in vivo Kette represses scar/waveactivity.

To further understand how Kette influences the dynamics of the F-actin cytoskeleton, we first conducted overexpression studies. In contrast to the kette loss of function phenotype, overexpression of full-length protein as well as the expression of several deletion constructs did not lead to any phenotype in wild-type flies or S2 cells(Fig. 4B, Fig. 5E; data not shown). As the overexpressed Kette protein was found in the cytosol, it appeared likely that membrane-recruitment of Kette is crucial for its function. In addition,biochemical as well as genetic data suggest that Kette - like the human homolog Nap1, which binds to the membrane-associated SH2SH3 adapter NCK/Dock -acts at the membrane (Hummel et al.,2000; Kitamura et al.,1996).

To test this possibility directly, we generated a membrane-tethered form of Kette by fusing a myristyolation signal derived from the DrosophilaSrc1 protein to the N terminus of Kette (Kette^Myr). Overexpression of comparable amounts of Kette^Myr in S2 cells also led to a dramatic rearrangement of the cytoskeleton and F-actin accumulates close to the membrane at sites of Kette expression(Fig. 4B, Fig. 5F). Thus, membrane tethered Kette protein is able to induce local actin polymerization.

In order to assess whether the membrane-tethered Kette protein performs similar to the wild-type protein we established UAS-kette^Myr transgenic flies and conducted genetic rescue experiments. Expression of Kette^Myr in the rhomboidpattern in all CNS midline cells of mutant kette embryos suppresses the mutant phenotype in such that the segmental commissures can be clearly identified again which is not possible in mutant kette embryos(Fig. 6F). This indicates that the Kette^Myr protein performs wild-type functions.

Elevated expression of the activated, membrane-tethered, Kette protein is able to induce a dominant phenotype in a dose-dependent manner. Flies carrying a scabrous-GAL4 driver (Klaes et al., 1994) and three copies of the UAS-Kette^Myrtransgene are characterized by pronounced alterations in bristle morphology(Fig. 7). Whereas in wild-type flies macrochaete and microchaete form as thin and relatively straight cuticular structures, expression of Kette^Myr leads to thicker and often forked bristles (Fig. 7A,B,E,G). In addition to the bristle phenotypes we observed defects in the epidermal hairs. In the wild-type, each epithelial cell will develop one fine hair, whereas Kette^Myr expression results in multiple hairs emerging from one cell (Fig. 7B,C, arrows). When we expressed only two copies of the Kette^Myr transgene, bristle development and epidermal hair formation was affected more moderately(Fig. 7C,F).

The bristle and hair phenotypes induced by Kette^Myr expression,reflect alterations in the organization of the apical F-actin cytoskeleton during pupal development when apical F-actin bundles prefigure bristles(Fig. 7H). In flies overexpressing Kette^Myr, a rearrangement of the F-actin cytoskeleton is apparent. In contrast to wild type, large and irregular F-actin bundles emerge. In addition, cell morphology is slightly altered,cells exhibit more irregular shapes and the cortical F-actin network appears fuzzy (Fig. 7I).

Treatment of S2 cells expressing the membrane-bound Kette^Myrprotein with wave dsRNA did not suppress the Kette^Myrinduced phenotype, suggesting the Kette may induce F-actin formation independent of Wave (data not shown). Another important regulator of F-actin formation is Wasp, which has a single Drosophila homolog. Removal of one copy of the wasp gene significantly restores the bristle phenotype evoked by overexpression of Kette^Myr(Fig. 7D). If Kette can act via activating Wasp, similar phenotypes might be expected following disruption of either gene. This is indeed the case, and loss of zygotic and maternal Wasp function results in a kette-like embryonic CNS phenotype(Tal et al., 2002; Zallen et al., 2002). To test whether a reduction in the gene dose of wasp may also affect the mutant kette phenotype, we generated a kette^C3-20wasp³ double mutant. However, the kette waspphenotype appeared identical to the kette mutant phenotype(Fig. 6E), which is in agreement with the notion that Kette and Wasp act positively in the same pathway.

In summary, Kette appears to repress Wave and, after membrane association,is able to activate Wasp. In the cytosol Kette is found in a complex with Wave but to date no interaction between Kette and Wasp has been reported. To elucidate further how Kette may regulate Wasp activity, we determined possible protein-protein interactions between Kette and Wasp in a yeast two hybrid assay (Fig. 8). Although no direct interaction was found between Kette and Wasp, it could be demonstrated between Kette and Abi as well as between Abi and Wasp, which thus is able to link Kette and Wasp.

Regulation of growth cone and cell motility involves coordinated control of cytoskeleton dynamics (Baum and Perrimon,2001; Mitchison and Cramer,1996). We present biochemical and genetic evidence that the Drosophila Nap1 homolog, Kette, is able to modulate the activity of both Wasp and Scar/Wave and thus contribute to the regulation of F-actin dynamics. Wasp and Wave together with Cdc42 and Rac1 control different aspects of cortical actin dynamics (Hall,1998; Takenawa and Miki,2001). Cdc42 and Wasp are required for filopodia formation,whereas dominant-negative Wave disrupts the Rac1-dependent formation of branched network of F-actin bundles required to form lamellipodia(Fukuoka et al., 2001; Hall, 1998; Miki et al., 1998a). Wave localizes to membrane ruffles induced by activated Rac1 and Wasp accumulates in microspikes containing bundled F-actin(Miki et al., 1998b; Nakagawa et al., 2001).

Recent work from the Kirschner laboratory showed that a large 500 kDa complex comprising Nap1/Kette, PIR121/Sra1, Abi, HSPC300 and Wave silences the otherwise constitutive activity of Wave in stimulating actin polymerization(Eden et al., 2002). The 500 kDa complex is stabilized by direct protein-protein interactions that have been demonstrated between Nap1/Kette and Abi(Tsuboi et al., 2002; Yamamoto et al., 2001), and Nap1/Kette and PIR121/Sra1 (also called Gex2)(Soto et al., 2002) (see Fig. 9 for a model). The exact binding partners of Wave are presently unknown.

Two modes of Wave activation have been demonstrated in vitro. On the one hand, activated Rac1 is able to bind to PIR121/Sra1(Kobayashi et al., 1998) and can relieve Wave inhibition by dissociating the Nap1/Kette, PIR121/Sra1 and Abi sub-complex (Eden et al.,2002). On the other hand, PIR121/Sra1 is able to bind to the first SH3 domain of SH2SH3 adapter Nck (Kitamura et al., 1996; Kitamura et al.,1997), which is sufficient to activate the Wave complex(Eden et al., 2002). In vivo these two mechanisms may work at the same time to fully activate Wave.

In agreement with the work of Eden et al.(Eden et al., 2002) disruption of Kette function leads to an excess formation of cytoplasmic F-actin. However, expression of even very high levels of wild-type Kette protein do not evoke any mutant phenotype. Thus, in wild type cells the inactive Wave complexes are already formed and Kette overexpression does not result in an additional sequestering of Wave into the silencing complexes and/or an incorporation of additional Kette into these complexes. In addition, as Kette requires Sra1 function to bind to membrane-associated adapters such as Nck,excess cytosolic Kette will not be able to reorganize subcellular Wave distribution.

Further support for the notion that Kette mediates repression of Scar/Wave activity stems from genetic analyses. Embryos lacking zygotic kettefunction display a characteristic CNS phenotype, whereas loss of zygotic Scar/Wave expression does not affect embryonic nervous system development(Zallen et al., 2002). The kette mutant phenotype, which is due to defects in neurite outgrowth(Hummel et al., 2000), could significantly be suppressed by reducing the dose of Scar/Wave expression. This demonstrates that in wild-type embryos, Kette acts as a negative regulator of Scar/Wave. Similar results were obtained when we reduced Kette and or Scar/Wave expression in Drosophila S2 cells. These experiments also revealed that Scar/Wave is required for the normal subcellular distribution of Kette, which may, however, be an indirect effect caused the disruption of the F-actin actin cytoskeleton.

Kette protein localizes to the plasma membrane where it accumulates in focal adhesion contacts. A prime candidate that may mediate recruitment of Kette to the membrane is the SH2 SH3 adapter protein Nck which, besides binding of the Sra1/Kette/Abi complex, also recruits numerous other proteins to focal contact sites (Li et al.,2001). Among these is Wasp, which binds to the third SH3 domain of Nck (Quilliam et al., 1996; Rivero-Lezcano et al., 1995)(Fig. 9).

The genetic interaction of kette and dock which encodes the Drosophila Nck homolog has recently been shown(Hummel et al., 2000). We have demonstrated that membrane recruitment of Kette is sufficient to activate actin polymerization in the cell cortex mediated by Wasp. How is this brought about? One explanation might be that recruitment of Kette to the membrane disintegrates the inhibitory Wave complex - independent of the Nck/Sra1 association. This would then lead to an excess of Wave activity and subsequently to an excess of actin polymerization. However, the genetic data clearly show that membrane bound Kette functions independent of Scar/Wave but depends on Wasp.

Wasp usually adopts an auto-inhibited conformation and is activated after Cdc42, Nck binding or phosphorylation (Kim et al., 2000; Miki et al.,1998a). A structure-function analysis of the DrosophilaWasp demonstrated that the Cdc42-binding domain is not necessary for function,suggesting that alternative pathways, such as phosphorylation can activate Wasp (Tal et al., 2002). Kette might be a part of such an alternative pathway, as we could demonstrate a genetic interaction between kette and wasp in the regulation of actin dynamics. Regulation of Wasp by Kette is not mediated by direct protein-protein interaction, but probably involves Abi that is able to link Kette and Wasp. The Abl interactor (Abi) protein localizes to sites of actin polymerization at the tips of lamellipodia and filopodia and has been implicated in the cytoskeletal reorganization in response to growth factor stimulation (Stradal et al.,2001). As a positive regulator of the non-receptor tyrosine-kinase Abelson (Abl) Abi may bring Abl into position to phosphorylate and thus activate Wasp. Abl is known to phosphorylate many proteins regulating focal adhesion and F-actin dynamics and overexpression of activated Abl induces F-Actin formation in Cdc42-independent manner(Woodring et al., 2002). Some tyrosine kinases activate by phosphorylation of Wasp(Cory et al., 2002; Scott et al., 2002; Suetsugu et al., 2002a);however, direct phosphorylation of Wasp by Abl remains to be demonstrated.

Further support of the model (Fig. 9) that in vivo Kette activates Wasp but suppresses Wave are the phenotypic analyses of Drosophila kette, wasp and scar/wavemutants. Mutations in kette have been isolated due to defects in commissure formation in the embryonic CNS(Hummel et al., 1999). If Kette acts via activating Wasp, similar phenotypes are expected following disruption of either gene. This is indeed the case and loss of zygotic and maternal Wasp function results in a kette-like embryonic CNS phenotype (Hummel et al.,2000; Zallen et al.,2002).

In agreement with the proposed function of Kette in regulating both, Wasp and Wave, is its subcellular distribution. Whereas the majority of Kette is present in the cytoplasm to keep Wave in its inactive state(Eden et al., 2002) some is present leading edge of lamellipodia-like structures. However, highest amounts of Kette are present at the insertion points of large F-actin bundles where N-Wasp is also present (Nakagawa et al.,2001). Kette might be recruited to these focal adhesion sites via Sra1/Nck (Goicoechea et al.,2002) and via Wasp could enhance the formation of F-actin bundles.

How Wasp activity results in straight F-actin bundles, whereas Wave stimulates the formation of a meshed F-actin network is presently unclear(Takenawa and Miki, 2001). In the cytosol, Kette may act as a scaffold protein that keeps Wave close to F-actin and recruits additional factors to F-actin such as Profilin, which not only binds to Kette but also enhances actin nucleation(Tsuboi et al., 2002; Witke et al., 1998; Yang et al., 2000). Thus,Kette could promote the formation of a meshed F-actin network characteristic for lamellipodia. At the membrane other proteins may interact with Kette and in this respect it is interesting to note that the F-actin crosslinking protein Filamin, which plays an important role in filopodia formation, also binds to Kette (S.B., unpublished). This suggests that Kette, in addition to regulating Wasp and Wave, may also contribute to the decision whether filopodia or lamellipodia are formed.

We thank E. Schejter for sending wave and scar mutants;M. Bähler for help during the co-sedimentation experiments and reagents;M. Fröhlich and K. Krukkert for excellent technical assistance; M. Bähler, G. Edenfeld, V. Gerke and A. Püschel for many helpful discussions and comments on the manuscript; and members of the Klämbt laboratory for help throughout the project. This work was funded through a grant of the DFG to C.K. and S.B.