Migration of cells through the reorganization of the actin cytoskeleton is essential for morphogenesis of multicellular animals. In a cell culture system, the actin-related protein (Arp) 2/3 complex functions as a nucleation core for actin polymerization when activated by the members of the WASP (Wiskott-Aldrich syndrome protein) family. However, the regulation of cell motility in vivo remains poorly understood. Here we report that homologues of the mammalian Arp2/3 complex and N-WASP in Caenorhabditis elegans play an important role in hypodermal cell migration during morphogenesis, a process known as ventral enclosure. In the absence of one of any of the C. elegans Arp2/3 complex subunits (ARX-1, ARX-2, ARX-4, ARX-5, ARX-6 or ARX-7) or of N-WASP (WSP-1), hypodermal cell migration led by actin-rich filopodia formation is inhibited during ventral enclosure owing to the reduction of filamentous actin formation. However, there is no effect on differentiation of hypodermal cells and dorsal intercalation. Disruption of the function of ARX-1 and WSP-1 in hypodermal cells also resulted in hypodermal cell arrest during ventral enclosure, suggesting that their function is cell autonomous. WSP-1 protein activated Arp2/3-mediated actin polymerization in vitro. Consistent with these results, the Arp2/3 complex and WSP-1 colocalized at the leading edge of migrating hypodermal cells. The stable localization of WSP-1 was dependent on the presence of Arp2/3 complex, suggesting an interaction between the Arp2/3 complex and WSP-1 in vivo.
The actin-related protein (Arp) 2/3 complex polymerizes new actin filaments in response to various signals (Pollard et al., 2000; Pantaloni et al., 2001; Takenawa and Miki, 2001). It functions as a core for actin polymerization when activated by members of the Wiskott-Aldrich syndrome protein (WASP) family (Machesky et al., 1999; Rohatgi et al., 1999). Upon activation of the Arp2/3 complex, rapid actin polymerization results in the formation of branched filaments, which allow the cytoskeletal reorganization needed for formation of protrusive filopodia and lamellipodia (Blanchoin et al., 2000; Pantaloni et al., 2000; Suetsugu et al., 2001). All five members of the WASP family, including WASP (Derry et al., 1994), N-WASP (Miki et al., 1996) and WAVE1-3 (Miki et al., 1998b; Suetsugu et al., 1999), activate actin polymerization through the Arp2/3 complex. Although WASP and N-WASP function downstream of Cdc42 to induce the formation of filopodia (Symons et al., 1996; Miki et al., 1998a), WAVE functions downstream of Rac and induces formation of lamellipodia (Takenawa and Miki, 2001).
Cell migration occurs at several points during the development of multicellular organisms (Bard, 1992). Although proliferating stem cells are restricted to various regions of the body, differentiated cells with specific functions migrate to their proper location for function. However, the mechanism that drives cell migration in the development of multicellular organisms remains poorly understood. The Arp2/3 complex and WASP family proteins are the most likely candidate molecules for directly driving these processes. In Drosophila, the Arp2/3 complex, SCAR/WAVE and WASP proteins function in actin dynamics. Drosophila Arp2/3 complex and SCAR/WAVE regulate cell morphogenesis in blastomeres, CNS neurons, egg chamber and adult eyes (Hudson and Cooley, 2002; Zallen et al., 2002). Drosophila WASP function in Notch-mediated cell lineage determination (Ben-Yaacov et al., 2001). In spite of the genetic evidence presented for Drosophila, C. elegans represents an excellent model system in which to determine the function of genes in development, because, unlike Drosophila, cell lineage is tractable and observation of entire cells is possible. Here, we have investigated the function of the Arp2/3 complex and WASP family proteins in C. elegans in cell migration.
Materials and Methods
Wild-type (N2 Bristol) and C. elegans strain SU93 jcIs1[ajm-1::GFP], JR667wIs51[seam cell::GFP marker], sem-5 (n2019 and n2030), hmp-1 (zu278), hmp-2 (zu364), hmr-1 (zu389) and apr-1 (zu10) were obtained from the Caenorhabditis Genetics Center (University of Minnesota, St Paul, MN). The marker jcls1[ajm-1::GFP] is an integrated array hypodermal cell marker, and wls51[seam cell::GFP] is an integrated seam cell marker. Strains were cultured at 20°C on NGM plates with Escherichia coli OP50.
The cDNA clones corresponding to C. elegans Arp3 (arx-1), Arp2 (arx-2), p34Arc (arx-4), p21Arc (arx-5), p20Arc (arx-6), p16Arc (arx-7) and N-WASP (wsp-1) were kindly provided by Y. Kohara at the National Institute of Genetics, Mishima, Japan. These clones are yk170f5 for arx-1, yk543a12 for arx-2, yk277h3 for arx-4, yk508a3 for arx-5, yk393h10 for arx-6, yk311a10 for arx-7, yk184g1 (WSP isoform 1) and yk9h5 (WSP-1 isoform2) for wsp-1. Sequences were confirmed by Sequence kit (Amersham Pharmacia Biotech).
RNA-mediated interference (RNAI) and microinjection
For soaking RNAi, the cDNA clones for components of the Arp2/3 complex and N-WASP (described above), previously cloned into pBluescript (pBS) vector, were amplified by PCR using primers complementary to the T3 and T7 promoter sequences. The resulting PCR products were used as templates for in vitro transcription by T3 and T7 RNA polymerases (Stratagene). The resulting complementary single-stranded RNAs were combined to form double-stranded RNA (dsRNA) and extracted by phenol, phenol-chloroform and chloroform followed by precipitation with 0.3 M sodium acetate (pH 5.2) and isopropanol. The resulting dsRNA was used for RNAi and was administered by the soaking method described previously (Maeda et al., 2001). Approximately 10 μg of dsRNA was dissolved in 5 μl 1×M9 (Mg2+ free), 3 mM spermidine (Wako, Osaka, Japan), 0.05% gelatin and 0.12% DOTAP (Roche). L4 larvae were washed with M9 several times and placed on an NGM plate without OP50 for 1 hour. Five worms were selected and immersed in RNA solution for 24 hours and then allowed to recover by incubation at 20°C on NGM plates for 36 hours before observation.
For feeding RNAi, the arx-1 cDNA (129-831) fragment was excised by SacI/SalI digestion of the yk221f3 construct and subsequently cloned into pPD129.36 (gift of A. Fire, Carnegie Institute, Washington, Baltimore). The construct was transformed into E. coli HT115(DE3) (obtained from CGC). After induction by IPTG, the bacterial cells were seeded on NGM plates containing 50 μg/ml ampicillin, 12.5 μg/ml tetracycline and 40 mM IPTG. L4 larvae were placed on the plates and allowed to lay eggs. As a control, L4 larvae were deposited on pPD129.36-transformed HT115(DE3).
For hypodermal cell-specific RNAi, cDNA fragments were cloned downstream of the tissue-specific promoter as described previously (Hoier et al., 2000). First, the sense and anti-sense strand of arx-1 cDNA (20-1278) and wsp-1 cDNA (537-1116) were amplified by PCR and cloned into pBS. The primer pairs used for the sense strand of arx-1 cDNA were: 5′senseBamHI: CGCGGATCCCGGCATGTGTGATCGACAATGG and 3′sensePstI: CCAATGCATTGGTTCTGCAGAGTAAGAGCTCCGAAAACTGG, and for anti-sense strand: 5′anti-sensePstI: AACTGCAGCGGCATGTGTGATCGACAATGG and 3′anti-senseBamHI: CGCGGATCCAGTAAGAGCTCCGAAAACTGG. The primer pairs for sense strand of wsp-1 cDNA were: WsSpe5′: GGACTAGTGATCCACTCCAACACGACAATTCG and WsPst3′: CCAATGCATTGGTTCTGCAGGCACCGCTTCCAATCGGAGCAGATG for anti-sense strand: WsPst5′: AACTGCAGGATCCACTCCAACACGACAATTCG and WsSpe3′: GGACTAGTGCACCGCTTCCAATCGGAGCAGATG. The resulting cDNA fragments cloned into pBS were subcloned by XbaI/KpnI digestion followed by insertion into the NheI/KpnI-digested plasmid pPD95.86(BamHI)-lin-26p(+), which has a lin-26 promoter region in pPD95.86 (gift of H. Qadota and K. Kaibuchi, Nagoya University, Japan). As an expression marker, the GFP fragment of pEGFP-C1 (Invitrogen) was digested by NheI/KpnI and subcloned into NheI/KpnI-digested pPD95.86(BamHI)-lin-26p(+). The sense and antisense constructs of arx-1 and wsp-1 cDNA, cloned downstream of lin-26 promoter, were injected along with the GFP marker in pPD95.86(BamHI)-lin-26p(+) into young adult worms. Progeny were observed by DIC and confocal microscopy (Leica). To obtain better identification of transgene-expressing embryos, embryos were stained with an anti-GFP antibody (A6455, Molecular Probes, Oregon).
For overexpression of the VCA region, the wsp-1 VCA region (1425-1794) was fused with VENUS (Nagai et al., 2002) (gift of Dr Miyawaki, RIKEN, Wako, Japan) at the WSP-1 VCA N-terminal, and subsequently cloned into pPD95.86(BamHI)-lin-26p(+). The construct was injected into young adult worms, and the progeny were observed by DIC and confocal microscopy (Leica).
Wild-type gravid hermaphrodites were cut transversely and the embryos placed on a 5% agar pad in M9 solution. RNAi-treated embryos, as described above, were collected from P0 hermaphrodites within 36 hours of the recovery period following RNAi immersion. The specimen was placed on a glass slide under a coverslip, sealed with Vaseline and its development was recorded by DIC microscopy. A four-dimensional series of images of GFP expression was recorded by a Multiphoton microscope as described previously (Simske and Hardin, 2001).
Antibodies and western blotting
Antibodies for ARX-1 and ARX-7 were obtained from rabbits immunized against synthetic peptides CGPSICRYNPVFGALT and CFGHGAIMRVFSGRQRL, respectively. Antisera were purified with activated CH-sepharose 4B beads (Pharmacia). Rabbit WSP-1 antiserum was raised against amino acids 475 to 598, corresponding to the VCA region of WSP-1. The antigen was first expressed as a GST fusion protein. The GST moiety was removed by thrombin digestion, and the resulting antigen was injected into rabbits. Antiserum was purified with CNBr-sepharose beads (Pharmacia) covalently coupled to the antigen. The specificity was examined by western blotting of total worm lysates. A monoclonal antibody against actin (MAB1501) was purchased from Chemicon (California), and its specificity for C. elegans actin was confirmed by western blotting. For western blotting, the gravid hermaphrodites were collected in M9 solution. Embryos were collected by treating adult hermaphrodites with bleach solution (2.5% NaHypochlorite, 0.5 M NaOH). In western blotting, total lysate or embryo extracts were run on a 12.5% polyacrylamide gel, blotted and probed. Approximately the same amount of embryo extract, judged from SDS-PAGE, was applied for the analysis of WSP-1 expression. Images were analyzed with NIH Image 1.62.
Gravid hermaphrodites of wild-type or RNAi-treated P0 hermaphrodites within 36 hours of recovery from RNA solution were placed on poly-L-lysine-coated slides (Sigma). A coverslip was overlaid, and pressure was applied to extrude the embryo. The slide was placed on dry ice for at least 10 minutes. Immediately after removal of the coverslip, the specimen was fixed with methanol for 5 minutes and washed three times for 10 minutes in phosphate-buffered saline (PBS) (125 mM NaCl; 16.6 mM Na2HPO4; 8.40 mM NaH2PO4) containing 0.1% Tween-20 (PBST). After blocking with PBST containing 20% fetal bovine serum (FBS) and 1% skim milk for 1 hour at room temperature under a coverslip, the specimen was incubated with a primary antibody diluted in PBST with FBS and skim milk overnight at 4°C. Antibodies against WSP-1, ARX-1 and ARX-7 were diluted 1:500, whereas the anti-actin antibody was diluted 1:1000. Anti-LIN-26 and MH27 antibodies were diluted 1:1500. Slides were washed three times for 10 minutes in PBST and then incubated with secondary antibodies for 1 hour at room temperature. Secondary antibodies were either AlexaFluor488-conjugated anti-rabbit IgG antibody (1:500, Molecular Probes) or AlexaFluor594-conjugated anti-mouse IgG antibody (1:500, Molecular Probes). Slides were washed three times in PBST and mounted using 25 mg/ml DABCO (Sigma) in 9:1 glycerol: PBS. Detection was performed by confocal laser scanning microscope (BioRad or Leica).
For immunostaining with labeled antibodies, anti-WSP-1 and anti-ARX-1 antibodies were labeled with Alexa Fluor 546 and Alexa Fluor 633 Protein Labeling Kit (Molecular Probes), respectively. Embryos were fixed and blocked as described above, then incubated with labeled antibodies overnight at 4°C. The dilutions were 1:100 for labeled anti-WSP-1 antibody and anti-ARX-1 antibody. The specimen were washed three times for 15 minutes in PBST then mounted and detected as described above.
Wild-type or RNAi-treated embryos were collected by cutting gravid hermaphrodites within 36 hours of recovery from RNA solution and were placed on a poly-L-Lysine-coated slide. The embryos were then treated as described previously (Williams-Masson et al., 1997). The fixed embryos were incubated with Rhodamine Phalloidin (Molecular Probes) for 1 hour at room temperature and washed three times with PBST. Embryos were mounted and viewed as described above.
Pyrene-actin assay was performed as described previously (Rohatgi et al., 1999). Briefly, Pyrene-labeled actin (9% final labeling) at 2 μM was added to 100 nM purified bovine N-WASP VCA or C. elegans WSP-1 VCA, 60 nM purified Arp2/3 complex. Bovine N-WASP VCA and C. elegans WSP-1 VCA were expressed as a GST fusion protein then purified with glutathione Sepharose 4B beads (Pharmacia). Arp2/3 complex was purified as described previously (Yamaguchi et al., 2002). The fluorescence was measured using a fluorescence spectrometer (Jasco).
Conservation of Arp2/3 complex subunits in C. elegans
The Arp2/3 complex is composed of seven subunits. The function and structure of the Arp2/3 complex is thought to be highly conserved in eukaryotes (Winter et al., 1999,Winter et al., 1999; Gournier et al., 2001; Robinson et al., 2001); however, functional analysis of the Arp2/3 complex in C. elegans has not yet been reported. Thus, we examined whether the Arp2/3 subunits are present in C. elegans by searching the complete C. elegans genome (Reboul et al., 2001; Fields et al., 1999). Candidate genes for each of the seven subunits were found, and they showed approximately 60% amino acid similarity with the homologous genes from human or Saccharomyces cerevisiae. These genes are Y71F9AL.16 for Arp3, K07C5.1 for Arp2, Y79H2A.6 for p41Arc, Y6D11A.1 for p34Arc, Y37D8A.1 for p21Arc, C35D10.1 for p20Arc and M01B12.3 for p16Arc. We confirmed that the cDNA sequences of six out of seven subunits (excluding Y79HA.6 for p41Arc) actually correspond to the predicted amino-acid sequences (data not shown). We named these putative subunits, as they correspond to the mammalian homologues, actin-related protein 2/3 complex (arx)-1 for Arp3, arx-2 for Arp2, arx-3 for p41Arc, arx-4 for p34Arc, arx-5 for p21Arc, arx-6 for p20Arc, and arx-7 for p16Arc.
Essential role of Arp2/3 complex in C. elegans morphogenesis
Next, we investigated Arp2/3 complex function in vivo by analysis of RNAi. We examined six of the seven subunits of Arp2/3 complex by the RNAi-by-soaking method (Maeda et al., 2001; Tabara et al., 1998), and the disruption of the expression of these genes by RNAi resulted in embryonic arrest during morphogenesis, with a similar terminal phenotype. This finding is consistent with the notion that the Arp2/3 complex functions as a seven-subunit complex (Winter et al., 1999,Winter et al., 1999; Gournier et al., 2001; Robinson et al., 2001). Hereafter, the suppression of a given gene by RNAi will be referred to by the gene followed by (RNAi), that is, arx-1(RNAi). RNAi was embryonic lethal for 100% of arx-1(RNAi) (n=211), 100% of arx-2(RNAi) (n=254), 83% of arx-4(RNAi) (n=234), 78% of arx-5(RNAi) (n=261), 88% of arx-6(RNAi) (n=285) and 87% of arx-7(RNAi) (n=296). To clarify at which point during development the Arp2/3 complex is essential, we observed RNAi-treated embryos under a DIC microscope until each embryo died.
Defects in ventral enclosure by depletion of Arp2/3 complex
During C. elegans morphogenesis, dynamic changes in hypodermal (external epithelial) cells occur in two distinguishable steps at approximately 300 minutes after the first cleavage. First, the two pairs of anterior contralateral hypodermal cells of the extreme ventral rows migrate toward the ventral midline by elongating filopodia until they meet and form junctions at the ventral midline. Second, eight pairs of posterior contralateral hypodermal cells are then drawn together by actin contractile forces to close and form junctions. Completion of these processes yields the basic worm shape and is called ventral enclosure (Priess and Hirsh, 1986; Raich et al., 1999; Simske and Hardin, 2001; Williams-Masson et al., 1997). At approximately 400 minutes, the embryo rotates 90° and begins elongation of the body along the anterior-posterior axis, with rearrangement of hypodermal cells and organogenesis of the pharynx and intestine (Fig. 1A). Twitching, which indicates muscle cell differentiation, is first observed at about 430 minutes.
Time-lapse observations of the lethal phenotype in embryos treated by RNAi revealed that ventral enclosure was not complete. We also observed the arrested embryos with AJM-1::GFP to visualize the shape of hypodermal cells. AJM-1 is a component of the zonula adherens at the cell-cell junction and can be used to visualize epithelial boundaries within the hypodermis during enclosure (Podbilewicz and White, 1994; Mohler et al., 1998; Koppen et al., 2001). Although alignment of hypodermal cells was observed on both the left and right sides (indicated by arrowheads in Fig. 1A-C), the hypodermal cells did not migrate, resulting in failed ventral enclosure (Fig. 1B,C). However, organogenesis, monitored with AJM-1 (Fig. 1F,G), and muscle cell differentiation, monitored by twitching, appeared to progress normally. The dying embryos exhibited distorted hypodermal cells, suggesting that the major defects caused by depletion of Arp2/3 complex by RNAi occur in cell migration during ventral enclosure (Fig. 1E-G). We show only the results of arx-2(RNAi) and arx-5(RNAi), but similar defects were observed with arx-1(RNAi),arx-4(RNAi), arx-6(RNAi) and arx-7(RNAi) (data not shown). Therefore, we conclude that the Arp2/3 complex plays an essential role in ventral enclosure in C. elegans.
Involvement of Arp2/3 complex in hypodermal cell migration
Previous studies have helped to identify two classes of genes involved in ventral enclosure. The first class of genes, including C. elegans gex-2 (mammalian Sra-1) and gex-3 (mammalian HEM-2), are responsible for initiation of hypodermal cell migration. The second class of genes facilitates cell adhesion and includes genes such as hmp-1 (α-catenin) and hmp-2 (β-catenin) (Costa et al., 1998). Mutants of genes in this class are able to show some cell migration but either fail to form cell-cell junctions or the embryo eventually ruptures during elongation owing to weak adhesions between cells.
Analysis by DIC microscope (Fig. 1) could not completely identify the class to which the Arp2/3 complex belongs. It is possible that cells did not migrate or that cells did migrate but that the junction was not formed, resulting in dysfunctional cell-cell contacts. Therefore, we investigated whether there is a defect in cell migration during ventral enclosure in embryos depleted of Arp2/3 complex by studying the movement of epithelial cells by time-lapse observation of GFP-fused AJM-1 protein using a multiphoton microscope (Mohler et al., 1998; Mohler and White, 1998a; Mohler and White, 1998b).
The AJM-1 protein clearly indicated the boundary of migrating cells in the wild-type embryo (Simske and Hardin, 2001) (Fig. 2A). In arx-2(RNAi) embryos, the two pairs of anterior contralateral hypodermal cells did not migrate towards the ventral side (see cells numbered 1 and 2 in Fig. 2B). The posterior cells in arx-2(RNAi) embryos (including cells numbered 4 in Fig. 2B) also showed defects in migration. These arrested hypodermal cells failed to migrate, even during the period when the wild-type embryos began body elongation and organogenesis (Fig. 2B). We conclude that the abnormality in morphogenesis in arx-2(RNAi) embryos is caused by a defect in hypodermal cell migration not in the abnormal or incomplete formation of cell-cell junctions.
To examine the actin filament of arrested hypodermal cells, we stained RNAi-treated embryos with phalloidin. In wild-type cells, filamentous actin accumulated at the leading edge (Fig. 3A-D). By contrast, accumulation of filamentous actin was reduced in arx-2(RNAi) embryos arrested at ventral enclosure (Fig. 3E-H).
Differentiation of hypodermal cells in embryos depleted of Arp2/3 complex
To clarify whether or not arrested hypodermal cells had differentiated, we next examined the expression of hypodermal cell differentiation markers. First, we stained RNAi-treated embryos with an anti-LIN-26 antibody. LIN-26 is a zinc-finger transcription factor that is required for hypodermal cells to acquire their proper fates (Labouesse et al., 1996). It is expressed in the nuclei among the hypodermal cells, slightly before the migration of hypodermal cells during ventral enclosure. Although the ventral hypodermal cells from arx-1(RNAi) embryos were defective in migration, normal LIN-26 expression was observed for these cells (Fig. 4Ad,f). Next, we analyzed the expression of a seam cell marker by a SCM::GFP fusion protein. In wild-type embryos, SCM::GFP is expressed in the nuclei of 20 seam cells after the completion of ventral enclosure. We performed RNAi on embryos expressing SCM::GFP then stained the embryos with phalloidin. The arx-2(RNAi) embryo showed normal expression of SCM::GFP in the all of the 20 nuclei of the seam cells (Fig. 4Bd,f). Despite junction formation failure at the ventral side during ventral enclosure, seam cells are differentiated normally. Additionally, we confirmed the normal localization of AJM-1 by observing an AJM-1::GFP fusion protein by multiphoton microscopy (Fig. 2). Slightly before ventral enclosure, the normal localization of AJM-1 of embryos with depleted Arp2/3 complexes was observed by immunostaining with an MH27 antibody, which recognizes AJM-1 (data not shown; Fig. 4Ae). Together, these results revealed that the defect in cell motility in RNAi-treated embryos was not a result of a defect occurring during differentiation.
In concert with ventral enclosure, dorsal hypodermal cells are rearranged by a process known as dorsal intercalation (Podbilewicz and White, 1994; Simske and Hardin, 2001). The zinc finger protein, DIE-1, functions in dorsal intercalation but not in ventral hypodermal cell migration (Heid et al., 2001). Some mutants, including gex-2 (mammalian Sra-1), gex-3 (mammalian HEM-2) (Soto et al., 2002) and apr-1 (tumor suppressor APC related) (Hoier et al., 2000), which exhibit defects in ventral enclosure, also show defects in dorsal intercalation. However, the pattern of AJM-1 on the dorsal side was normal in arx-2(RNAi) embryos (Fig. 2E) as well as in embryos depleted of other Arp2/3 subunits by RNAi (data not shown). Therefore, the Arp2/3 complex does not appear to be involved in dorsal intercalation.
WASP family proteins in C. elegans
We then focused on the involvement of WASP family proteins in cell migration (Pollard et al., 2000; Pantaloni et al., 2001; Takenawa and Miki, 2001). In C. elegans, there are two WASP family proteins, WSP-1 (C07G1.4) and WVE-1 (R06C1.3), which correspond to mammalian N-WASP and WAVE, respectively. Since there are two verprolin homology (VPH) domains in C07G1.4, this gene is thought to correspond to N-WASP, which is the only WASP family protein with two VPH domains (Miki et al., 1996).
Analysis of transcripts corresponding to C07G1.4 revealed that there are two splicing variants of WSP-1 that would produce transcripts of 3.5 kb and 2.1 kb. Of 14 EST clones of C07G1.4, nine clones represented one variant (WSP-1 isoform 1) and five clones represented the other (WSP-1 isoform 2). In order to establish the existence of the splicing variants, we performed northern blot analysis on RNAs from embryos using probes against the 3′-terminal region, which is identical in both isoform 1 and 2. It revealed a major band at 2.1 kb and a minor band at 3.5 kb, suggesting that isoform 1 is the major variant (data not shown).
The open reading frame of WSP-1 isoform 1 predicts a protein of 598 amino acids (approximately 65 kDa) with high homology (approximately 50%) to mammalian N-WASP. Furthermore, the WSP-1 isoform 1 contains all of the domains found in mammalian N-WASP (Miki et al., 1996) (Fig. 5A,B), suggesting that the WSP-1 isoform 1 is the homologue of mammalian N-WASP.
In isoform 2 of WSP-1, the N-terminal region of the WSP-1 isoform 1, including the WASP homology 1 domain and IQ motif, was replaced with amino acids with no homology to any known proteins (Fig. 5A,B). We have not, so far, determined the methionine at which the translation begins. Translation may start at either of the methionines indicated by asterisks (*1 and *2) in Fig. 5A, with a 54 kDa and a 51 kDa protein being translated from the 1*-start and 2*-start sites, respectively. If neither of these methionines is the translation start, WSP-1 isoform 2 would be more than a 70 kDa protein containing more than 660 amino acids. In all of these cases, the WSP-1 isoform 2 contains all the domains present in mammalian N-WASP except the WH1 domain and IQ motif, which suggests that isoform 2 may also form a functional interaction with Cdc42, the Arp2/3 complex, as well as SH3-domain-containing proteins such as Ash/Grb2/SEM-5.
To determine whether these WSP-1 variants are translated into proteins, we raised a specific antibody against WSP-1 using the C-terminal VCA region as the antigen. Western blot analysis of mixed stage worm cell lysates using this antibody showed only the presence of a 65 kDa protein, indicating that the majority of WSP-1 appears to be the 65 kDa isoform 1 (Fig. 5C).
WSP-1 activates Arp2/3 complex in vitro
Because the C-terminal region (VCA domain) of WASP family proteins has been shown to function as the Arp2/3-complex-activating domain in yeast and mammals, we examined the ability of the VCA domain of WSP-1 to activate Arp2/3-complex-mediated actin polymerization in vitro. In this assay, bovine Arp2/3 complex was used and the activity of WSP-1 was compared to bovine N-WASP VCA, which was included as a positive control. The VCA region of WSP-1 increased Arp2/3-complex-mediated actin polymerization (Fig. 5D). Although actin polymerization mediated by WSP-1 VCA was somewhat slower than that mediated by bovine N-WASP VCA, this is probably due to differences in the specificities of bovine N-WASP and C. elegans WSP-1 VCA for the bovine Arp2/3 complex.
Function of WSP-1 in ventral enclosure
We investigated the in vivo role of WSP-1 in ventral enclosure by using RNAi. We synthesized dsRNA from cDNA encoding WSP-1 isoform 1 (yk184g1); however both isoforms of WSP-1 mRNA are thought to be degraded owing to the identical C-terminal region. RNAi-by-soaking with the dsRNA yielded several phenotypes: 15% were embryonic lethal, 20% showed larval arrest and 20% were sterile (n=134). Delivery of dsRNA by injection showed similar results (data not shown). In order to clarify the relationship between the Arp2/3 complex and WSP-1, we further examined the embryonic-lethal phenotype. Although the arrested embryos possessed no obvious worm shape, they did exhibit twitching, indicating muscle cell differentiation. The time-lapse series by DIC microscopy revealed incomplete ventral enclosure (Fig. 1D). AJM-1 protein localization of the terminal RNAi embryos revealed a normal pharynx and intestine; however, the distorted hypodermal cells did not cover the entire body but instead accumulated around the periphery of the embryo (Fig. 1H). We then investigated the process under a multiphoton microscope and failed to observe hypodermal migration in arrested embryos by wsp-1(RNAi) (Fig. 2C). We also examined filamentous actin formation in wsp-1(RNAi)-treated embryos and observed a reduction in filamentous actin formation at the edge of the hypodermal cells (Fig. 3I-L). The edge of arrested hypodermal cells (indicated with arrows) does not show filamentous actin accumulation. These phenotypes were similar to those of arx-1(RNAi), arx-2(RNAi), arx-4(RNAi), arx-5(RNAi), arx-6(RNAi) and arx-7(RNAi) embryos. The similarity in phenotypes suggests that the Arp2/3 complex and WSP-1 function in the same signaling pathway in C. elegans. In addition, abnormality of the dorsal side was not observed, which is again similar to that seen in embryos depleted of Arp2/3 components (data not shown). Obvious defects exhibited by the other 85% viable embryos were not related to ventral enclosure (data not shown). We also treated embryos with RNAi against both arx-2 and wsp-1, but this treatment did not result in the observation of a different phenotype (n=43). Finally, distribution of the differentiation markers, LIN-26, SCM::GFP and AJM-1 were not affected by wsp-1(RNAi) (data not shown).
Function of Arp2/3 complex and WSP-1 is cell autonomous
Previous investigations into the role of vab-1 and vab-2 suggested that the signaling pathway originating from the neuroblast guides hypodermal cell migration during ventral enclosure (George et al., 1998; Chin-Sang et al., 1999). By contrast, Drosophila WASP (Wsp) functions in the Notch signaling pathway, which suggests that WASP also functions in cell-cell signaling as well as in cell motility (Ben-Yaccov et al., 2001). Therefore, we investigated whether the Arp2/3 complex and WSP-1 function cell autonomously in hypodermal cell migration. First, we performed hypodermal cell-specific RNAi for arx-1 and wsp-1 by expressing sense and anti-sense RNA under the expression of the lin-26 promoter. During embryogenesis, LIN-26 is expressed in hypodermal cells, in the deirid sheath cells and transiently in the neuroblast ALM/BDU located mid-posterior of the embryo (Labouesse et al., 1996). A plasmid, containing GFP under the expression of the lin-26 promoter, was injected together with the RNAi constructs to monitor lin-26-promoter-driven expression. Although injection of the GFP marker did not result in a change of phenotype, we observed 100% embryonic lethality in embryos expressing lin-26::arx-1 (n=22) and 63% in embryos expressing lin-26::wsp-1 (n=11), respectively. Immunostaining with anti-ARX-1 and anti-WSP-1 antibodies revealed that ARX-1 and WSP-1 are reduced in hypodermal cells, whereas other tissues expressed ARX-1 or WSP-1 at normal levels (data not shown). Although lethal embryos exhibited the structure of a pharynx and muscle formation, they did not develop the normal worm shape (Fig. 6A,B; data not shown). These phenotypes are similar to those observed for soaking RNAi against the Arp2/3 complex subunits and wsp-1.
It has been previously been reported that overexpression of the VCA region of N-WASP has a dominant-negative effect on the Arp2/3 complex (Machesky and Insall, 1998). In order to investigate this in vivo in C. elegans, we expressed the VCA region of WSP-1 under the control of the lin-26 promoter. Overexpression of WSP-1 VCA region resulted in embryonic lethality, with a defect in ventral enclosure. The lethal embryos showed the structure of a pharynx and were twitching, indicating muscle formation, but did not develop into a normal worm shape (data not shown). Together, these results suggest that the function of the Arp2/3 complex and WSP-1 in ventral enclosure is cell autonomous.
Localization of C. elegans ARX-1 and ARX-7 at hypodermal cells during ventral enclosure
We attempted to raise antibodies against ARX-1, ARX-2, ARX-5 and ARX-7. However, only antibodies against ARX-1 and ARX-7 were isolated successfully. The specificities of these antibodies were examined by western blotting mixed-staged C. elegans lysates. Bands at 48 kDa and 16 kDa were detected with anti-ARX-1 and anti-ARX-7 antibodies, respectively (Fig. 7A). Next, we double-stained actin and either ARX-1 or ARX-7 with their respective antibodies. During early embryogenesis, ARX-1 and ARX-7 were localized in the cytoplasm of all of the cells (Fig. 7B,E). A lateral view showing expression of these proteins in hypodermal cells during ventral enclosure in C. elegans embryos is shown (Fig. 7C,F). ARX-1 and ARX-7 were localized among a broad range of cells including the hypodermal cells. The similarity in localization of ARX-1 and ARX-7 supports the hypothesis that they form a complex. A reduction in the fluorescent intensities in embryos treated for both arx-1(RNAi) and arx-7(RNAi) confirmed the specificity of the staining against ARX-1 and ARX-7, respectively. In arx-1(RNAi) and arx-7(RNAi) embryos, we did not observe any obvious changes in actin localization by using an anti-actin antibody, although there was defect in actin filament formation (Fig. 3, Fig. 7Be,Cf,Ee,Ff,Fig. 7Be,Cf,Ee,Ff). To determine whether WSP-1 influences localization of ARX-1 and ARX-7, we stained wsp-1(RNAi) embryos with anti-ARX-1 and anti-ARX-7 antibodies. However, no obvious change in the localization of the Arp2/3 complex was observed in wsp-1(RNAi) embryos (n>200) (Fig. 7Ci,Fi).
WSP-1 localizes at the periphery of cells
We next examined the localization of WSP-1 (Fig. 8). During early embryogenesis, WSP-1 colocalized with actin at the blastomere boundary (Fig. 8Aa-c). During ventral enclosure, WSP-1 colocalized with actin at the periphery of cells, including in hypodermal cells migrating to the ventral side (Fig. 8Aj-m), suggesting that WSP-1 functions in cell migration at cell boundaries. In wsp-1(RNAi) embryos, WSP-1 protein levels were reduced and localization at boundaries was diminished (Fig. 8Ad,n). Reduction of the protein level of WSP-1 was observed in almost all the RNAi-treated embryos, including those that would undergo normal embryogenesis. Anti-actin staining of wsp-1(RNAi) embryos did not significantly alter the actin skeleton compared to wild-type embryos (Fig. 8Ae,o).
Determination of WSP-1 localization by Arp2/3 complex
To investigate whether WSP-1 and the Arp2/3 complex interact in vivo, we first double-stained embryos expressing ajm-1::GFP with anti-WSP-1 antibodies conjugated to Alexa Fluor 546 and anti-ARX-1 antibodies conjugated to Alexa Fluor 633. Colocalization of WSP-1 and ARX-1 was observed at the leading edge of migrating hypodermal cells (Fig. 9). We further examined the localization of WSP-1 in embryos with disrupted Arp2/3 complexes. In 60% of arx-5(RNAi)-treated embryos (n=55), localization of WSP-1 at the periphery of cells was disrupted. In these embryos, WSP-1 was partially shifted to cell nuclei in blastomeres and migrating epithelial cells (data not shown). Changes in localization of WSP-1 were not only observed in arx-5(RNAi) but also in arx-1(RNAi) and arx-2(RNAi) embryos (Fig. 8Ag,i,r,t,u; data not shown). These findings suggest that the Arp2/3 complex is required for WSP-1 localization at the periphery of cells, including at the leading edge of migrating cells during ventral enclosure. WSP-1 and the Arp2/3 complex appear to associate and function at cell boundaries during ventral enclosure in C. elegans, and localization of WSP-1 is dependent on the Arp2/3 complex.
We examined the effect of molecules that regulate N-WASP on WSP-1 localization. The SH3 domain of Ash/Grb2/SEM-5 was the first N-WASP-binding partner identified (Miki et al., 1996). SEM-5, the C. elegans homologue of mammalian Ash/Grb2, has the conserved two SH3 domains. WSP-1 was localized at boundaries in sem-5 (n2019 and n2030) (Clark et al., 1992), mutants of sem-5 in which one or both SH3 domains are deleted (data not shown). Cdc42 is the upstream regulator of N-WASP in mammals (Miki et al., 1998a). In the C. elegans embryos depleted of the Cdc42 homologue (cdc-42(RNAi)), WSP-1 was localized at boundaries during early cell division in common with that in wildtype (data not shown). In other mutants with ventral enclosure defects, including apr-1 (APC-related), hmp-1 (α-catenin), hmp-2 (β-catenin) and hmr-1 (cadherin), WSP-1 localization at cell boundaries was also maintained (data not shown).
Expression of WSP-1 is increased in arx-1(RNAi)
Next, we examined the amount of WSP-1 protein in arx-1(RNAi) embryos by western blotting (Fig. 8B). To harvest the embryos required for western blotting we deleted arx-1 using RNAi by feeding. We confirmed that arx-1(RNAi) by feeding results in embryonic lethality (85%, n=294) with similar phenotypes to those observed by RNAi by soaking. The amount of ARX-1 was reduced in arx-1(RNAi) embryos whereas the amount of actin remained the same in control and arx-1(RNAi) embryos. In arx-1(RNAi) embryos, the anti-WSP-1 antibody recognized two bands, at approximately 65 kDa and 72 kDa, that correspond with isoform1 and 2, respectively. The intensity of the band of isoform 1 was four-fold stronger than that of the band detected in the lysate of the control embryo, as determined by densitometry, suggesting that expression of WSP-1 is increased by the reduction of functional ARX-1.
Spindle orientation in wsp-1(RNAi) embryos
WSP-1 localization at cell boundaries suggests that WSP-1 functions in the formation of cell polarity. WSP-1 has a conserved Cdc42-binding motif similar to that of mammalian N-WASP and WASP. Thus, WSP-1 may function downstream of Cdc42, like PAR proteins, in the formation of cell polarity. In wild-type embryos, the one-cell embryo divides asymmetrically, producing the larger AB blastomere and the smaller P blastomere. In the second division, the spindle of the AB blastomere is oriented longitudinally, whereas it is oriented horizontally in the P blastomere (Fig. 10A). However, cdc42(RNAi) embryos show abnormal orientation of the spindles (Kay and Hunter, 2001; Gotta et al., 2001), such that spindles in both AB and P blastomeres are oriented horizontally (Fig. 10C). On the other hand, the wsp-1(RNAi) embryos did not show abnormal spindle orientation and divided asymmetrically, but later showed defects in ventral enclosure (n=7, Fig. 10B). Viable embryos did not show abnormal spindle positioning (n=43, data not shown). We also confirmed that the proper asymmetric cell division occurred even in the presence of reduced WSP-1 protein levels (Fig. 8Ad,f). Together, the results of these RNAi experiments suggest that WSP-1 does not contribute to spindle orientation during early development of C. elegans.
Ventral enclosure marks the beginning of morphogenesis in C. elegans, with a change in embryo shape from ovoid to vermiform. This process is initiated by hypodermal cell migration and is dependent on actin polymerization. The involvement of actin filament dynamics was revealed by treatment of embryos with cytochalasin D, which inhibits actin filament formation. Cytochalasin D abolished cell migration during ventral enclosure and caused retraction of epithelial cells (Williams-Masson et al., 1997).
In the present study, we demonstrated that the Arp2/3 complex is essential for cell migration during ventral enclosure. WSP-1 functions with the Arp2/3 complex to drive cell migration, most probably by inducing actin polymerization, and these actions are likely to be cell autonomous in hypodermal cells. In addition, hypodermal cell differentiation was shown to be normal before and after ventral enclosure, indicating that WSP-1 and the Arp2/3 complex are not involved in cell differentiation. Furthermore, the Arp2/3 complex and WSP-1 seemed to form a complex at the leading edge of hypodermal cells and the results of in vitro actin polymerization assays indicate that the WSP-1-Arp2/3 pathway is likely to be conserved in C. elegans and mammals. On the basis of these results, we conclude that the Arp2/3 complex and WSP-1 are necessary for the actin-mediated generation of forces needed for cell migration during ventral enclosure and that these factors function cell autonomously.
During ventral enclosure, two remarkable events occur. One is hypodermal cell migration, and the other is formation of cell adhesions. In this study, we provide evidence that the C. elegans Arp2/3 complex and WSP-1 function in cell migration. Various strategies have been used to identify the genes involved in ventral enclosure, and cells with mutations in many of these genes show some cell migration but fail to form cell-cell junctions, or the embryo ruptures during elongation owing to weak adhesions between cells. These mutants include apr-1 (tumor suppressor APC related), hmp-1 (α-catenin), hmp-2 (β-catenin) and hmr-1 (cadherin), all of which encode cell adhesion molecules involved in junction formation during ventral enclosure. Another gene, mab-2 (semaphorin) (Roy et al., 2000), is needed for proper cell-cell contact, and disruption of this gene leads to embryo rupture. To date, there is little information regarding molecules involved in the initial process of hypodermal cell migration during ventral enclosure, with gex-2 (mammalian Sra-1) or gex-3 (mammalian HEM-2) being the lone identified genes directly involved in cell migration and dorsal intercalation during ventral enclosure (Soto et al., 2002). Mutations in these genes lead to defects in cell migration and dorsal intercalation. With the similarities in phenotypes and the possible involvement of Rho family small GTPases, which govern the signaling cascade in reorganization of actin cytoskeletons, the relationship between WSP-1, GEX-2 and GEX-3 in C. elegans ventral enclosure requires further study. Another class of mutants defective in ventral enclosure are vab-1 (ephrin receptor protein-tyrosine kinase) (George et al., 1998) and vab-2 (ephrin) (Chin-Sang et al., 1999), which function in neuroblasts. Mutation of these genes results in improperly positioned neuronal cells, which presumably disrupt a signaling pathway required for hypodermal cell migration, or the leakage of neural cells, which blocks hypodermal cell migration. Although it might be because neuronal cells are resistant to RNAi, depletion of Arp2/3 complex or wsp-1 did not cause any defects in neural cells, which would inhibit the hypodermal cell migration.
In fact, ARX-1 and ARX-2 have high homology to actin. The cDNA sequence of arx-2 that we used as a template for dsRNA synthesis in RNAi experiments has a region at the 3′-terminal with 90% homology to act-2, one of the five actin genes in C. elegans. Although this high homology may result in the disruption of actin, the labeling of actin was not reduced significantly in arx-2(RNAi) embryos (Fig. 8Ah,s). Thus, the phenotype of arx-2(RNAi) embryos is thought to be due solely to the disruption of ARX-2.
When RNAi against wsp-1 was driven under the expression of the lin-26 promoter, lin26::wsp-1(RNAi), embryonic lethality was as high as 63% (n=11); soaking or injection of RNAi against wsp-1 only yielded 15% embryonic lethality. A likely explanation is inefficient RNAi by soaking or injection. Although immunofluorescence confirmed that the levels of WSP-1 were reduced after soaking RNAi, it is possible that a small amount of WSP-1 remained beyond the limits of immunofluorescent detection. In addition, it should also be taken into account that only a small number of embryos expressing lin26::wsp-1(RNAi) could be detected owing to the difficulty in detecting the GFP expression marker. In this regard, it is important to consider that greater detectable GFP expression under lin-26 promoter could reflect a greater amount of dsRNA expression, and this is the cause of the higher embryonic lethality seen for RNAi under the expression of the lin-26 promoter.
A second explanation is the existence of alternative signaling pathways that direct the Arp2/3 complex during ventral enclosure. In lin-26::wsp-1(RNAi), which resulted in a high frequency of embryonic lethality, a sudden decrease of WSP-1 may not be compensated for by other proteins. WAVE (WVE-1), another WASP family member, is a candidate for regulating the Arp2/3 complex, but wve-1(RNAi) did not cause a detectable altered phenotype. Following double RNAi against wsp-1 and wve-1, there was no increase in embryonic lethality observed (data not shown). However, we cannot conclude that WVE-1 is not involved in ventral enclosure, because the efficiency of RNAi is also dependent on the target gene.
Depletion of the Arp2/3 complex or WSP-1 did not result in arrest during early embryogenesis. During ventral enclosure, filamentous actin formation is greatly reduced by the depletion of Arp2/3 complex or WSP-1; however, actin accumulation and changes in cell shape were still observed in the ventral hypodermal cells. Immunostaining and western blotting indicated that the expression of actin did not change during RNAi against Arp2/3 complex subunits or wsp-1. Cultured cells injected with an antibody inhibiting the function of the Arp2/3 complex were reported to have a normal cell shape, but were unable to form lamellipodia or filopodia (Bailly et al., 2001). Furthermore, fibroblasts from N-WASP-deficient mice have normal actin cytoskeletons; however, they were defective in some types of filopodia formation (Lommel et al., 2001; Snapper et al., 2001). Together, these findings suggest that there may be other factors that regulate actin cytoskeleton reorganization independently from the Arp2/3 complex.
Although the Arp2/3 complex has been established as the nucleation core in actin polymerization, Arp2/3-complex-independent mechanisms of actin polymerization have recently been identified. Formin, a conserved Rho-GTPase effector among eukaryotes, is predicted to function in cytokinesis, polarized growth and stress fiber formation by activating actin assembly independently from the Arp2/3 complex (Evangelista et al., 2002; Pruyne et al., 2002; Sagot et al., 2002). In addition to formin, Ena/VASP regulates cell motility by binding to actin directly to promote barbed end elongation with binding to actin directly (Bear et al., 2002). Although the Arp2/3-complex-independent pathways also function in C. elegans from cytokinesis to morphogenesis, they may not nucleate actin as efficiently as Arp2/3-dependent pathways in dynamic reorganization of the actin cytoskeleton. Thus, defects due to disruption of Arp2/3 complex and WSP-1 might first appear at ventral enclosure with dynamic cell migration.
It is also possible that the Arp2/3 complex, WSP-1 or WVE-1 functions in other processes. Since RNAi does not represent disruption of the gene and is rather a means to inhibit gene expression, the method does not always yield complete disruption of gene expression. The nervous system, which requires cell motility during development, is resistant to RNAi (Tavernarakis et al., 2000). Furthermore, in our analysis, RNAi was performed on L4 and adult animals by soaking, feeding or microinjection of plasmid-based dsRNA under the expression of the lin-26 promoter. Taking these considerations into account, we may not have detected all of the functions of the Arp2/3 complex, WSP-1 or WVE-1. In Drosophila, disruption of the Arp2/3 complex or WAVE/SCAR causes abnormalities during oogenesis. These mutants have defects in the ring canal, an actin-dependent structure that functions as a bridge for the flow of cytoplasm from the nurse cell to the oocyte (Hudson and Cooley, 2002; Zallen et al., 2002). Although C. elegans does not have an obvious structure corresponding to the ring canal, it is possible that Arp2/3, WSP-1 or WVE-1 in C. elegans may be involved in oogenesis, since maternal proteins may ameliorate the effect of RNAi in oogenesis in our procedure. In Drosophila, a mutant of WASP has been isolated and characterized. In Drosophila, Wsp functions in a Notch-mediated signaling pathway involved in cell lineage decision, in particular in the central nervous system and mesoderm (Ben-Yaacov et al., 2001). However, we did not detect a phenotype related to cell linage decision.
By western blotting and immunostaining, we found that the expression of WSP-1 protein is affected by the expression of ARX-1 and, furthermore, stable localization of WSP-1 requires binding to the Arp2/3 complex from early cell division to morphogenesis. WSP-1 expression is enhanced when Arp2/3 complex is decreased. These results indicate that there is a feedback system inducing WSP-1 expression when the Arp2/3 complex is absent. Immunostaining results suggest an interaction between the Arp2/3 complex and WSP-1 in vivo. In mammals, activated WASP and N-WASP bind to the Arp2/3 complex, and this might indicate that WSP-1 is activated during early cell division as well as during ventral enclosure. However, it is unclear why WSP-1 localizes in cell nuclei in embryos depleted of the Arp2/3 complex, because no nuclear N-WASP-binding molecule has been reported. WSP-1 may be recruited to nuclei in concert with its increased expression. WSP-1 may function in the rearrangement of cortical actin filaments during early cell divisions, although we did not detect a phenotype associated with this period. The localization of the Arp2/3 complex in wsp-1(RNAi) embryos was not affected. We assume there is additional machinery that influences the localization of the Arp2/3 complex. The dependency of WSP-1 on Arp2/3 for localization, and not vice versa, raises several questions about the relationship between Arp2/3 complex and WSP-1 localization. One is that WSP-1 and the Arp2/3 complex are both contained within a huge protein complex, and the Arp2/3 complex might determine the localization of this putative complex. In this regard, it is notable that the N-WASP regulator in mammals does not affect WSP-1 localization. Although the most characterized pathway in mammals is activation of WASP by Cdc42, Cdc42 does not influence WSP-1 localization in early cell divisions of C. elegans. However, during ventral enclosure, cdc42(RNAi) strongly influenced cells to arrest during early cell divisions, and this prevented us from examining Cdc42 function at a later stage. It has been reported that C. elegans Cdc42 (CDC-42) localizes to the periphery of hypodermal cells during morphogenesis, and this may influence WSP-1 localization (Chen et al., 1996). Similarly, in Drosophila, Cdc42 is not required for regulation of Wsp during cell lineage determination (Tal et al., 2002). Another candidate that may activate WSP-1 is Ash/Grb2/SEM-5. Although sem-5 mutants have been isolated in C. elegans, no abnormality in ventral enclosure was observed. WSP-1 localization is also normal in sem-5 mutants (n2019 and n2030), suggesting that SEM-5 is not a regulator of WSP-1 in ventral enclosure. Therefore, the regulation of the Arp2/3-WASP pathway in C. elegans remains unclear.
We thank the following people for providing materials used in this work: Y. Kohara for cDNA clones for the Arp2/3 complex and WSP-1, M. C. Hersko and R. H. Waterston for the mAbMH27 and M. Labouesse for the anti-LIN-26 antibody. We are also grateful to H. Qadota and K. Kaibuchi for providing pPD95.86(BamHI)-lin-26p(+), A. Fire for pPD129.36 vector and A. Miyawaki for VENUS. Some of the strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the National Center for Research Resources of the National Institutes of Health.
- Accepted January 8, 2003.
- © The Company of Biologists Limited 2003