C. elegans ankyrin repeat protein VAB-19 is a component of epidermal attachment structures and is essential for epidermal morphogenesis

During development, different tissues must coordinate their morphogenesis to form functional units. Caenorhabditis elegans offers unique experimental opportunities for the genetic analysis of morphogenesis(Chin-Sang and Chisholm, 2000; Simske and Hardin, 2001). The C. elegans embryo changes from being bean shaped into the long thin shape of the larva, a process known as elongation. Over a period of 4 hours,the embryo elongates from 50 μm to 250 μm along its long axis,simultaneously contracting in circumference. Elongation involves coordinated changes in the shapes of cells within the epidermis. Cell ablation experiments suggest that these shape changes occur cell autonomously and require circumferential actin microfilament bundles (CFBs) in epidermal cells(Priess and Hirsh, 1986). Analysis of elongation-defective mutants has identified epidermal proteins that regulate actin-based contractions, leading to the model that actomyosin-based contraction of lateral epidermal cells provides the driving force for epidermal cell shape change(Piekny et al., 2000; Wissmann et al., 1997; Wissmann et al., 1999). Loss of function in the SMA-1/β_H-spectrin and the SPC-1/α-spectrin results in mild disorganization of epidermal actin bundles and slower or blocked elongation(McKeown et al., 1998; Norman and Moerman, 2002),showing that epidermal actin organization is important for elongation.

During early epidermal elongation, body muscle cells migrate to positions underlying dorsal and ventral epidermis. At the same time, attachment structures begin to form in the epidermis adjacent to the muscle cells. Attachment structures were first observed using electron microscopy, and were termed fibrous organelles (Francis and Waterston, 1991; Hresko et al., 1994). Attachment structures are confined to the regions of the epidermal cells overlying muscle and some other cells, suggesting that muscle cells regulate attachment structure localization within the epidermal cytoskeleton (Hresko et al.,1994). Epidermal attachment structures form part of a molecular link between the contractile apparatus of the muscle and the cuticular exoskeleton.

Epidermal elongation is not dependent solely on cell autonomous processes in the epidermis but also requires nonautonomous function of muscle cells. Mutations that severely disrupt muscle function also block epidermal elongation soon after muscle contractions normally start, leading to the Pat(paralysed, arrested at twofold) phenotype(Barstead and Waterston, 1989; Barstead and Waterston, 1991; Gettner et al., 1995; Williams and Waterston, 1994). The requirement for muscle function in elongation remains unexplained;defective muscles might hinder epidermal cell shape changes, or muscle contraction might provide additional stretching forces on the epidermis allowing elongation beyond the twofold stage.

An alternative explanation comes from the observation that muscle cells influence the organization of attachment structures in the epidermis. Loss of embryonic attachment structure components results in elongation defects. The spectraplakin VAB-10A is the C. elegans ortholog of Plectin, a conserved component of hemidesmosomes; VAB-10A is localized to attachment structures and is essential for elongation(Bosher et al., 2003). The cytoplasmic intermediate filament IFB-1A is a component of attachment structures (W.-M. Woo and A.D.C., unpublished) and is also essential for elongation beyond the twofold stage(Karabinos et al., 2001). The transmembrane protein Myotactin localizes to attachment structures and is essential for elongation; in the absence of Myotactin, attachment structures do not remain localized to muscle-adjacent regions of the epidermis,suggesting that Myotactin receives a muscle-dependent signal that anchors attachment structures (Hresko et al.,1999).

Here, we report that vab-19 is essential for epidermal elongation and that it encodes the C. elegans ortholog of a tumor suppressor locus, Kank (Sarkar et al.,2002). A functional VAB-19 fusion with green fluorescent protein(VAB-19::GFP) localizes to epidermal attachment structures. The function of vab-19 is not required for the assembly of attachment structures but is required for their localization to muscle-adjacent regions of epidermis. VAB-19 is also required for normal epidermal actin organization, and vab-19 mutant phenotypes are specifically suppressed by loss of function in the β_H spectrin SMA-1. VAB-19 might antagonize the function of β_H spectrin in the apical actin cytoskeleton,permitting attachment structures to develop beyond a certain stage and allowing epidermal cells to change shape.

C. elegans was grown on NGM plates as described by Brenner(Brenner, 1974). Bristol N2 was used as the wild type. vab-19(e1036) was isolated by J. Lewis in screens for mutants with epidermal morphogenesis defects and displays a cold-sensitive phenotype. vab-19(ju406) was isolated in a semiclonal screen for epidermal morphogenesis mutants (M.D., W.-M. Woo and A.D.C.,unpublished). EMS-mutagenized hermaphrodites were allowed to self-fertilize and two F₁ progeny were put on a single plate as a semiclone. The F₂ progeny of each semiclone were examined for the segregation of dead larvae resembling e1036. Siblings then were recovered to produce progeny. 8000 mutagenized genomes were screened.

We made four-dimensional movies as previously described(Chin-Sang et al., 1999). We recorded movies from ten vab-19(e1036cs) embryos (progeny of parents shifted to 15°C as L4s), eight vab-19(ju406) and five vab-19(ju406)/nDf2 embryos. All vab-19 mutant embryos arrested at the twofold stage.

The vab-19(e1036) mutation was genetically mapped to the lin-42 sup-9 interval on the left arm of chromosome II (map data are available from Wormbase). The chromosomal deficiencies nDf2 and nDf3 fail to complement vab-19. Cosmids were obtained from the Sanger Institute (Hinxton, UK). DNA preparation and subcloning followed standard procedures (Sambrook and Russell,2001).

In rescue experiments, we assayed rescue of the vab-19 lethal phenotypes and of the Vab phenotypes of e1036 animals raised at a semipermissive temperature. Rescue of lethal phenotypes was scored qualitatively (growth or no growth of the transgenic strain at 15°C). Rescue of Vab phenotypes was scored in Rol animals; over 1000 animals were scored for each array tested. At 20°C, 20% of vab-19(e1036)animals are Vab; rescue was defined as a statistically significant reduction in the proportion of Vab in Rol animals.

Initial transformation experiments showed that cosmid F42G2 contained partial vab-19 rescuing activity; the vab-19 exons contained in F42G2 encode the N-terminal 484 amino acids of VAB-19. The minimal genomic fragment that fully rescues vab-19 phenotypes is contained in clone pCZ440, made by ligating a 6.6 kb MluI-XbaI fragment from cosmid F42G2 with a 6.5 kb MluI-SalI fragment of cosmid K02G6. vab-19 corresponds to Wormbase gene model T22D2.1. cDNAs for vab-19 had been isolated in a genome-wide expressed sequence tag project (Maeda et al., 2001). The longest cDNA, yk481c2, was sequenced completely and found to lack the first seven bases of the predicted coding sequence. In the process of allele sequencing, we also identified one error in the published genomic sequence,which changes one predicted intron to an exon. The predicted vab-19gene structure between residues 470 and 504 was confirmed by sequencing yk481c2 and another cDNA, yk656c2. The 6.5 kb MluI-SalI genomic DNA fragment in pCZ440 was replaced with the corresponding cDNA fragment from yk481c2 to make a vab-19 minigene (pCZ437). To make the VAB-19::GFP fusion construct pCZ441, GFP coding sequence from pPD113.37 (Fire lab vector kit) was inserted in frame at the BamHI site of pCZ437.

To identify lesions in vab-19 alleles, vab-19 genomic DNA(including all exons and exon-intron junctions) was amplified from vab-19 mutant and wild-type animals. DNA sequences were determined using ³³P-labeled primers and the fmol sequencing kit(Promega). All lesions were confirmed on both strands and from DNAs prepared in independent polymerase chain reactions (PCRs). Sequences of primers used are available on request.

To determine the effect of the ju406 mutation on vab-19splicing, reverse-transcription PCR (RT-PCR) was performed on total RNA isolated from vab-19(ju406) animals (primer sequences available on request). We sequenced 13 RT-PCR products and found two kinds of mutant transcript containing frame shifts and premature stop codons in exons 3 or 4.

Germ-line transformation followed standard procedures(Mello et al., 1991) using 5-10 ng μl^–1 of vab-19 DNA and 50 ngμl^–1 pRF4 coinjection marker. In all transgenic analysis,at least five independent lines were scored. Integrants of functional VAB-19::GFP arrays were obtained by TMP-UV mutagenesis. Five independent integrants of VAB-19::GFP (juIs167, juIs168, juIs169, juIs170 and juIs171) were identified and backcrossed multiple times before data collection. The VAB-19::GFP transgenes displayed identical expression patterns and rescue of vab-19 phenotypes. juIs167 was used in Figs 3, 4, 5, and juIs169 was used in Fig. 6.

Whole mount immunofluorescence staining was performed following the protocol of Finney and Ruvkun (Finney and Ruvkun, 1990). To analyse VAB-19::GFP in embryos, Clorox-treated animals were fixed in 1% paraformaldehyde for 3.5 hours on ice and incubated with anti-GFP antibodies (Chemicon) at 1:100 dilution and appropriate secondary antibodies. MH4 and MH46 (Developmental Hybridoma Studies Bank,University of Iowa) were used at 1:300 dilution. Images were collected using a Leica TCS-NT Confocal microscope. VAB-19::GFP expression was observed using a Zeiss Axioskop equipped with a HQ-FITC filter (Chroma). Phalloidin staining was performed as described by Costa et al.(Costa et al., 1997). Embryos were dislodged from NGM plates. After rinsing in M9 buffer, the eggshells were digested for 10 minutes using chitinase (Sigma). Embryos were fixed for 20 minutes in 4% paraformaldehyde, washed in PBS and stained for 1 hour with 0.3μM phalloidin labeled with Alexa Fluor 488 (Molecular Probes).

L1 animals were processed for electron microscopy as previously described(Hallam et al., 2002). Serial thin sections were taken from three wild-type L1s and two vab-19(e1036) L1s; the vab-19 mutants were progeny of parents that had been shifted to 15°C at the L4 stage. The wild-type sections are longitudinal sections from the head region.

The vab-19 gene is defined by two mutant alleles, e1036cs and ju406. At permissive temperatures (22.5°C), about 90% of e1036 animals survive to adulthood and display variable defects in epidermal morphology (the Vab phenotype). At restrictive temperatures(15°C), all e1036 mutants arrest at the twofold stage of embryonic elongation (Fig. 1H). The phenotypes of e1036/Df animals and e1036/e1036 animals raised at 15°C were indistinguishable, suggesting that e1036causes complete loss of function at 15°C. In all studies of mutant phenotypes described below, ju406 behaved indistinguishably from e1036 at 15°C.

To determine more precisely the role of vab-19 in embryonic morphogenesis, we analysed vab-19 mutant embryos using time-lapse Nomarski microscopy and found that vab-19 mutant embryos display defects in epidermal elongation. Wild-type embryos begin to elongate at∼350 minutes after first cleavage and elongate three- to fourfold by the time of hatching (at 800 minutes). At the 1.75-fold stage (∼430 minutes),body wall muscle contractions begin and the embryos start to twitch weakly. By the twofold stage (∼450 minutes), the embryos roll vigorously and continue their elongation until threefold (∼520 minutes, Fig. 1C). vab-19mutant embryos displayed normal elongation prior to the twofold stage(Fig. 1E,F). Within 10-20 minutes of reaching the twofold stage, vab-19 mutant embryos stopped elongating (Fig. 1G). vab-19 embryos also displayed detachment of body muscles from the epidermis (Fig. 1H, arrowhead; Fig. 1I,J). Consistent with a gradual failure of muscle attachments, vab-19 mutant embryos twitched normally but never rolled, and were essentially paralysed after the twofold stage. Development continued, as indicated by cuticle formation and development of a pharynx, and most vab-19 embryos hatched a few hours later than the wild type, although they did not grow after hatching. In general, the vab-19 phenotype resembles the `weak Pat' phenotype described by Williams and Waterston(Williams and Waterston,1994). Expression of the adherens junction marker AJM-1(Koppen et al., 2001) in vab-19 mutants was similar to the wild type (compare Fig. 1M,N with 1K,L), showing that epidermal cells are specified correctly and have normal cell-cell contacts.

Although our analysis did not reveal defects earlier than elongation in vab-19 mutants, we could not rule out the possibility that lack of VAB-19 causes defects in early development and that these defects indirectly affect elongation. To address this question, we performed temperature shift experiments using the cold-sensitive allele e1036. When vab-19(e1036) embryos grown at 22.5°C were shifted to 15°C before elongation, almost 100% of these embryos arrested at the twofold stage. When vab-19(e1036) embryos grown at 15°C were shifted to 22.5°C before elongation, about 80% of these animals survived to adulthood and appeared wild type in morphology (Fig. 1O). Thus, these experiments identify a vab-19temperature-sensitive period at the twofold stage of elongation, suggesting that VAB-19 function is directly involved in epidermal elongation.

We mapped vab-19 between lin-42 and sup-9 on the left arm of chromosome II, and tested genomic DNA clones from this region for their ability to rescue vab-19(e1036) phenotypes in transgenic lines(Fig. 2A). vab-19phenotypes were partly rescued by cosmid F42G2. Further rescue analysis showed that vab-19 corresponds to the gene T22D2.1 (see Materials and methods).

The predicted VAB-19 protein is a member of a conserved protein family containing four ankyrin repeats in their C-termini (VAB-19 residues N863-T999; Fig. 2B). The Drosophila genome contains a single VAB-19 ortholog, CG10249. The human genome contains two orthologs, KIAA1518 and KIAA0172/Kank, and the mouse and rat genomes each contain two VAB-19-related genes. The cellular functions of these VAB-19 family members have not been determined. The KIAA0172 gene, also known as Kank, is downregulated in renal cell carcinomas, suggesting that it acts as a tumor suppressor(Sarkar et al., 2002). Within the ankyrin repeats, C. elegans VAB-19 is 59% identical to human Kank and 58% identical to Drosophila VAB-19(Fig. 2B). Alignments of VAB-19 family members also revealed two short regions of conservation in the N-terminus (Fig. 2C). These motifs appear to be unique to the VAB-19 family and are termed motif A (VAB-19 residues P64-R103) and motif B (VAB-19 residues S229-G285). Expression of the human Kank cDNA under the control of the vab-19 promoter partly rescued vab-19 mutant phenotypes(Fig. 2A), suggesting that VAB-19 and Kank have similar cellular roles.

We determined the molecular lesions of vab-19 alleles. ju406 alters the exon 3 splice acceptor (tccag|AAA to tccaa|AAA); vab-19(ju406) animals express aberrantly spliced transcripts that encode truncated proteins (see Materials and methods). The cold-sensitive allele e1036 results in a premature stop codon(R732opal) before the ankyrin repeat domain. e1036 behaves genetically as a null at 15°C and as a hypomorph at higher temperatures. At the restrictive temperature, e1036 appears to make the mRNA unstable and subject to nonsense-mediated decay by the SMG system(Mango, 2001), because the lethal and Vab phenotypes of e1036 were almost completely suppressed by a mutation in smg-3 (Table 1); smg-3 did not suppress ju406 (data not shown). vab-19(e1036) might retain partial function owing to read-through of the stop codon at the permissive temperature or in a smg background; alternatively, truncated proteins made in e1036 might themselves be cold sensitive.

To determine how vab-19 might function in vivo, we generated reporter constructs containing translational fusions of the full-length vab-19 coding sequence to GFP. These VAB-19::GFP constructs completely rescued vab-19 null mutant phenotypes(Fig. 2), suggesting that they reflect the endogenous VAB-19 expression pattern. VAB-19::GFP was first detected in embryos at the onset of elongation(Fig. 3A). At this stage,VAB-19::GFP localization was diffuse in dorsal and ventral, but not in lateral, epidermal cells. During early elongation, VAB-19::GFP started to accumulate in muscle-adjacent regions of dorsal and ventral epidermal cells(Fig. 3B, arrow). During intermediate stages of elongation, VAB-19::GFP became localized to the parts of the epidermis adjacent to body wall muscles(Fig. 3C). Finally, during the late elongation, larval and adult stages, VAB-19::GFP localized to circumferential bands in epidermis adjacent to muscles(Fig. 3D-F) and to mechanosensory neuron processes (not shown). Within these regions, the pattern of circumferential bands is interrupted by gaps where other neuronal processes intervene between muscles and the epidermis (e.g. Fig. 3F). This pattern of localization of VAB-19::GFP resembles that of epidermal attachment structures. Furthermore, transgenes containing the VAB-19 cDNA under the control of the ajm-1 promoter, which is expressed in the epidermis and other epithelial tissues but not in muscles(Köppen et al., 2001),also fully rescued vab-19 phenotypes (not shown), consistent with VAB-19 functioning in epidermal cells.

To determine whether VAB-19 is a component of epidermal attachment structures, we asked whether VAB-19::GFP localized with known components of attachment structures. Attachment structures contain several intermediate filament (IF) proteins, some of which are recognized by monoclonal antibody MH4 (Francis and Waterston,1991). During early elongation, MH4 staining accumulated in a single patch in dorsal epidermal cells(Fig. 4B), whereas VAB-19::GFP remained diffuse (Fig. 4A). During the intermediate elongation stage, VAB-19::GFP and MH4 staining displayed more extensive colocalization(Fig. 4D-F). During later elongation and afterwards, VAB-19::GFP and MH4 staining fully co-localized to circumferential bands (Fig. 4G-L). Because MH4-positive IFs become localized before VAB-19::GFP, VAB-19 probably is not required for the initial recruitment of IFs to attachment structures.

The transmembrane protein Myotactin maintains localization of attachment structures to muscle-adjacent parts of the epidermis(Francis and Waterston, 1991; Hresko et al., 1999). Myotactin itself initially localizes to muscle-adjacent regions of the epidermis and later becomes organized into attachment structures. VAB-19::GFP colocalized with Myotactin before the twofold stage(Fig. 4M-O). Both Myotactin and VAB-19::GFP localized to circumferential bands(Fig. 4P-R). These bands are interrupted by gaps where neuronal processes intervene between muscle and epidermis (Fig. 4J,P,Q,arrows). In contrast to Myotactin or VAB-19::GFP, MH4 staining is present or enhanced at such positions (Fig. 4K, arrow). These differences suggest that VAB-19 does not colocalize with all epidermal IFs but only those associated with muscle adjacent attachment structures.

VAB-19::GFP was also expressed in pharyngeal marginal cells, which contain prominent tonofilament-like attachment structures(Albertson and Thomson, 1976). Pharyngeal attachment structures resemble epidermal attachment structures in both ultrastructure and molecular composition, because they contain MH4-positive IFs, VAB-10A and Myotactin(Francis and Waterston, 1991). In contrast to Myotactin, which is localized to the outer basal surface of marginal cells (Fig. 4T),VAB-19::GFP was localized throughout the apical-basal axis, although it was concentrated at the apical and basal surfaces(Fig. 4S, arrow; compare with Fig. 4U). Thus, in pharyngeal marginal cells, VAB-19 might be a component of the ends of IF-containing attachment structures.

To examine how VAB-19 becomes localized to epidermal attachment structures,we fused GFP to different portions of VAB-19 and expressed these truncated proteins under the control of the vab-19 promoter. Truncated proteins lacking the ankyrin repeats were localized normally(Fig. 3G). By contrast, a truncated protein lacking the N-terminal domain was diffusely localized in dorsal and ventral epidermal cells (Fig. 3H). Thus, the VAB-19 N-terminal domain, containing the conserved motifs A and B, might be both necessary and sufficient for localization of VAB-19. GFP-tagged fusion proteins containing only motif A or only motif B were diffusely expressed (not shown), suggesting that both motifs are required together for normal VAB-19 localization.

Because, during later stages of embryogenesis, VAB-19 localizes in or close to attachment structures, we examined the effect of vab-19 mutations on the localization of other attachment structure components. One of the first proteins to become localized to attachment structures is VAB-10A, a C. elegans spectraplakin orthologous to Plectin and Drosophila Shot(Bosher et al., 2003). During early and intermediate (twofold) stages of epidermal elongation, the localization of intermediate filaments (MH4) and VAB-10A appeared normal in vab-19 mutants (Fig. 5). However, during later elongation, IFs or VAB-10A became delocalized, expanding into regions not adjacent to body wall muscle, yet remaining in regular circumferential bands (compare Fig. 5C and 5F for MH4 staining). Thus, VAB-19 is a component of attachment structures but is not essential for their formation.

Because of the similarities between VAB-19 and Myotactin distribution, we also analysed the localization of Myotactin in vab-19 mutants. Myotactin staining appeared normal in 1.5-fold and twofold vab-19embryos (Fig. 5). However, in vab-19 mutants, Myotactin never became localized into circumferential stripes and instead remained localized to longitudinal bands in muscle-adjacent regions of the epidermis (compare Fig. 5I and 5M). Thus, VAB-19 is required to recruit Myotactin to attachment structures but not for its localization to muscle-adjacent regions.

To learn whether VAB-19 is required for aspects of attachment structure morphology not discernible by immunostaining, we used electron microscopy to compare attachment structures in wild-type and vab-19 mutant L1s. Attachment structures were present in vab-19 mutants and were correctly distributed with respect to actin bundles, as judged by the registration of attachment structures and cuticle annuli (which form directly above actin CFBs) in longitudinal sections(Fig. 5O and inset). The epidermis appears thicker in vab-19 mutants and attachment structures are extended, possibly because of the reduced length of the animal. Thus,VAB-19 is not essential for the correct distribution of attachment structures with respect to actin CFBs.

We asked whether attachment structure components were required for VAB-19 localization by introducing VAB-19::GFP into animals lacking Myotactin or VAB-10A. In Myotactin (let-805) mutants, VAB-19::GFP was normally localized to muscle-adjacent regions of the epidermis but never became distributed into circumferential stripes. Within 15-20 minutes of muscle contraction starting (∼430 minutes), VAB-19::GFP became dispersed into puncta within epidermal cells (Fig. 6B). Thus, VAB-19 requires Myotactin for its subcellular localization in epidermal cells. By contrast, in vab-10A(ju281ts) mutants, VAB-19::GFP remained confined to the regions of epidermal cells overlying muscle(Fig. 6D) but never became localized to attachment structures. Thus, neither Myotactin nor VAB-10A is required for VAB-19 to become localized adjacent to muscles but both are required for VAB-19 to be recruited to attachment structures.

The driving force for epidermal elongation is provided by contraction of circumferential actin bundles (CFBs)(Priess and Hirsh, 1986). Because vab-19 mutants have defects in elongation, we asked whether VAB-19 is required for normal actin distribution. During elongation, actin CFBs are circumferentially oriented in the epidermis(Fig. 7A-D). Actin distribution in vab-19 mutants appeared normal until the 1.5-fold stage(Fig. 7E). As the mutant embryos developed to the twofold stage, the epidermal actin cytoskeleton became disorganized. For example, actin bundles were absent or reduced from apical epidermis (Fig. 7F,arrow). Actin CFBs were often misoriented or splayed(Fig. 7G,H). Although some of these actin bundles might represent misplaced actin from other parts of the epidermis, the apical CFBs constitute the most prominent actin structures in the epidermis at this stage. The disorganization of actin bundles was usually more obvious where body muscles had detached from epidermal cells. This result suggests that muscle detachment might result in the disorganization of actin bundles, which then blocks epidermal elongation.

To determine whether VAB-19 displayed interactions with other components of the apical actin cytoskeleton we tested for genetic interactions between vab-19 and other genes required for normal elongation. The SMA-1β _H spectrin is required for organization of the actin cytoskeleton within the epidermis (McKeown et al., 1998). Mutants in sma-1 are viable yet elongate more slowly than the wild type. We found that loss of sma-1 function strongly suppressed the elongation arrest and lethal phenotypes of vab-19 mutants (Table 1). Viable vab-19 sma-1 double mutants were indistinguishable from sma-1 mutants in both elongation rate and terminal phenotype (data not shown). Multiple alleles of sma-1suppressed both alleles of vab-19. By contrast, sma-1mutations did not suppress the phenotypes of other attachment structure or epidermal elongation mutants including vab-10A/Plectin, vab-10B/MACF, pat-3/β-integrin, let-805/Myotactin and mup-4 (data not shown). The gene-specific suppression of vab-19 by sma-1 suggests that VAB-19 and SMA-1 have antagonistic functions. The expression level of SMA-1 protein appeared normal in vab-19 mutants (data not shown),suggesting that these proteins antagonize at the level of function but not expression. Epidermal actin distribution in vab-19 sma-1 double mutants resembled the mildly disorganized pattern of sma-1 mutants rather than that of vab-19 mutants(Fig. 7I,J), suggesting that the antagonism of VAB-19 and SMA-1 occurs at the level of their effects on actin distribution.

The vab-19 gene encodes an ankyrin repeat-domain-containing protein essential for C. elegans epidermal morphogenesis. The vab-19 mutant phenotype resembles that caused by loss of function in epidermal components of attachment structures. VAB-19 is produced in the epidermis and localizes to attachment structures. VAB-19 is not essential for the formation of attachment structures but is necessary for their later localization to muscle-adjacent regions of the epidermis. VAB-19 is also necessary for the normal organization of epidermal actin. We hypothesize that VAB-19 participates in cross-talk between IF-containing attachment structures and the actin cytoskeleton.

Analysis of VAB-19 homologs in Drosophila and vertebrates has shown that the VAB-19 family is defined by a highly conserved C-terminal domain containing four ankyrin (ANK) repeats, and a less well conserved N-terminal domain containing two novel motifs. ANK repeats are found in many proteins with diverse functions; according to the protein families database Pfam (Bateman et al., 2002),over 112 C. elegans proteins are predicted to contain ANK repeats. The ANK repeat domains of VAB-19 family members are 58-60% identical to each other; outside the VAB-19 family, the closest ANK repeats (37% identity) are from zebrafish Diversin (Schwarz-Romond et al., 2002), which shares no other similarities with VAB-19. ANK repeats function as protein-protein interaction interfaces(Sedgwick and Smerdon, 1999). The high degree of conservation in the ANK repeats suggests that VAB-19 family members bind one or more conserved proteins via this domain.

The ANK-repeat domain of VAB-19 is neither necessary nor sufficient for the normal subcellular localization of the protein, as assayed using GFP fusion proteins. By contrast, two conserved motifs at the N-terminus are necessary and sufficient for VAB-19::GFP subcellular localization to attachment structures. The N-terminus of VAB-19 shows no similarity to IF-associated proteins, suggesting that VAB-19 is localized indirectly to IF-containing structures. Three non-IF proteins are found in embryonic attachment structures– the spectraplakin VAB-10A and the transmembrane receptors Myotactin and MUP-4. Our molecular epistasis experiments suggest that the final localization of VAB-19 requires both VAB-10A and Myotactin. VAB-10A is required for VAB-19 localization to attachment structures but not to muscle-adjacent regions; by contrast, Myotactin is required for both aspects of VAB-19 localization. These proteins might direct VAB-19 localization via the VAB-19 N-terminal motifs.

Using molecular epistasis, we have explored the functional relationships between VAB-19 and other components of attachment structures. In vab-19 mutants, the localization of known attachment structure components (IFs/MH4 and VAB-10A/Plectin) appears normal until late in embryogenesis. At this point, attachment structures redistribute into non-muscle-adjacent regions in the epidermis, although they remain in regular circumferential bands (Fig. 5F), suggesting that trans-epidermal attachments initially assemble normally but, in the absence of VAB-19, they lose their normal association with muscle cells. Consistent with this interpretation, our ultrastructural analysis has shown that attachment structures are present and correctly placed with respect to cuticular structures in vab-19mutants.

Like VAB-19, Myotactin anchors attachment structures to muscle-adjacent regions (Hresko et al., 1999). Myotactin and VAB-19 are mutually required for their correct final localization. Myotactin's early localization to muscle-adjacent regions does not require VAB-19, but the assembly of Myotactin into circumferential bands is dependent on VAB-19 (Fig. 5M). Myotactin is a large transmembrane protein with an extracellular domain consisting of ∼32 fibronectin type-III repeats,suggesting that it could bind a ligand on muscle cells or muscle-associated extracellular matrix and thereby anchor attachment structures in muscle-adjacent regions. Unlike VAB-10A or intermediate filaments, the final localization of VAB-19 to attachment structures is completely dependent on Myotactin. VAB-19 might function in the Myotactin-dependent stabilization of attachment structures in muscle-adjacent regions; in the absence of Myotactin,VAB-19 would lose its association with attachment structures, and attachment structures lose their association with muscle.

The refinement of attachment structure localization into circumferential bands must involve interaction with CFBs. VAB-19 is not required for the segregation of attachment structures from actin CFBs, because attachment structures retain their striped appearance in vab-19 mutants. However, VAB-19 could play a role in cross-talk between attachment structures and the cortical actin cytoskeleton. At present, the evidence for a specific interaction between VAB-19 and the actin cytoskeleton comes from the suppression of vab-19 by sma-1. By analogy to other spectrins, SMA-1/β_H-spectrin probably cross-links apical actin and might stabilize integral membrane proteins within the apical epidermis(Dubreuil and Grushko, 1998). Is VAB-19 localized to the apical ends of attachment structures, consistent with a role in regulating the cortical actin cytoskeleton or SMA-1? Although epidermal attachment structures are too short to resolve apical versus basal localization, we can partly address this question by examining pharyngeal marginal cells, which contain longer attachment structures. In these cells,VAB-19::GFP appeared to be concentrated at the apical and basal ends of attachment structures. Although we do not know the role of vab-19 in pharyngeal marginal cells, the similarities between pharyngeal and epidermal attachment structures suggest that, by analogy, VAB-19 might be present at the apical surface of the epidermis, consistent with a role in regulation of the cortical actin cytoskeleton. We infer from our genetic analysis that VAB-19 might locally inhibit or antagonize SMA-1 function, perhaps to allow attachment structures to move during the cell shape changes of epidermal elongation. Thus, in sma-1 mutants, apical membrane proteins might be less highly cross-linked to cortical actin, partly bypassing the function of vab-19. Further tests of these models will require identification of proteins that interact with VAB-19.

Why are epidermal attachment structure components required for epidermal elongation? One model suggests that the function of attachment structures in elongation is indirect and reflects their role in muscle contraction. In mutants such as vab-19, muscles detach from the epidermis soon after muscle contraction begins, presumably because of a failure in the mechanical strength of epidermal muscle attachments. Muscle function is, for unknown reasons, required for epidermal elongation and so muscle detachment leads to elongation arrest. Alternatively, attachment structures might directly influence the organization of epidermal actin bundles, which are themselves required for elongation. Epidermal attachments and circumferential actin bundles are regularly interspersed in the muscle-adjacent parts of epidermal cells; during later stages, attachment structures are in register with the cuticular ridges known as annuli, whereas actin bundles coincide with the furrows separating annuli (Costa et al.,1997). The complementary distribution of actin CFBs and attachment structures suggests that they are mutually required for each other's organization. This is exemplified by VAB-10A/Plectin (an attachment-structure component) and VAB-10B/MACF, an actin-associated protein, which are mutually required for their organization into complementary circumferential bands(Bosher et al., 2003). The complementary pattern of actin and attachment structures arises gradually:actin and attachment structures might partly co-localize at early stages and only later segregate into alternating bands. Thus, the elongation defects of mutants defective in muscle might arise because muscles induce the development of epidermal attachment structures, which themselves influence the organization of epidermal actin.

There are VAB-19 orthologs in the Drosophila, mouse, rat and human genomes. At present, however, little is known of the function of these VAB-19 family members. Based on the conservation of sequence in both the N-terminal motifs and the C-terminal ankyrin repeats, VAB-19 family members probably interact with conserved protein partners and function in regulating the cytoskeleton.

The Drosophila genome contains a single vab-19 ortholog,whose function has not yet been determined. However, unlike C. elegans,Drosophila lacks cytoplasmic IFs(Fyrberg and Goldstein, 1990; Goldstein and Gunawardena,2000). Instead, Drosophila muscle attachment structures contain microtubule (MT) bundles (Mogensen and Tucker, 1988). The Drosophila spectraplakin Shot(previously known as Kakapo) is a component of attachment structures and contains a MT-binding domain but not an IF-binding domain(Gregory and Brown, 1998; Röper et al., 2002). It will be important to determine whether the Drosophila VAB-19 is a component of MT-based muscle-attachment structures. The existence of a conserved VAB-19 family member in a species lacking cytoplasmic IFs further supports the hypothesis that VAB-19 does not interact directly with IFs. Nevertheless, VAB-19 and VAB-10-like spectraplakins could represent an evolutionarily conserved core of cell-matrix attachment structures.

The only vertebrate VAB-19 homolog for which functional data exist is the human gene Kank, which might function as a tumor suppressor for renal cell carcinoma, the most common malignancy of the adult kidney(Sarkar et al., 2002). Expression of Kank in HEK293 cells suppressed cell growth and morphological changes, and expression of Kank in human kidney tumor(G-402) cells resulted in disorganization of actin(Sarkar et al., 2002). In light of the possible interaction of VAB-19 with actin in C. elegans,the VAB-19 family might play a conserved role in the actin cytoskeleton. The ability of human Kank partly to rescue vab-19 mutant phenotypes supports the contention that the VAB-19 family has a conserved cellular function. It will clearly be informative to determine the subcellular localization and function of VAB-19 family members in vertebrate cells.

We thank J. Lewis for isolating e1036, S. George and R. Harrington for initial characterization of e1036, Y. Kohara for vab-19cDNAs, the Kazusa DNA Research Institute for the KIAA0172 cDNA clone, the C. elegans genome consortium for the sequences and cosmid DNAs, A. Fire for GFP vectors, and V. Praitis for her phalloidin staining protocol and SMA-1 antisera. We thank X. Huang, W.-M. Woo and N. Brown for discussions and comments on the manuscript. Some strains used here were obtained from the Caenorhabditis Genetics Center, which is supported by the NIH. YJ is an Assistant Investigator of the HHMI. This work was supported by a grant to ADC from the NIH (GM54657). The GenBank Accession Number for the vab-19 cDNA sequence is AY273823