The VAB-1 Eph receptor tyrosine kinase and SAX-3/Robo neuronal receptors function together during C. elegans embryonic morphogenesis

Cell movements are crucial for the development of all embryos. If cells are not correctly positioned in the developing embryo, tissues or organs may not be formed or they may be formed in the wrong place or at the wrong time. Cells can navigate to their correct position by sensing their environment for positional cues. In C. elegans, the VAB-1 Eph receptor tyrosine kinase and its ephrin ligand EFN-1 function primarily as guidance cues on neuroblasts for proper neuronal and epidermal cell movements(Chin-Sang et al., 1999; George et al., 1998). The vertebrate Eph receptor tyrosine kinases (RTKs) and their ligands, the ephrins, have diverse roles during development. These include axon guidance,angiogenesis and the regulation of cancer (reviewed by Adams and Klein, 2000; Dodelet and Pasquale, 2000; Drescher, 2000; Flanagan and Vanderhaeghen,1998; Kullander and Klein,2002). Understanding how the vertebrate Eph receptors signal during development is under intense research, but the large number of receptors and ephrin ligands, as well as the promiscuity of their binding partners, can complicate the picture. By contrast, C. elegans has only one Eph RTK and four ephrin ligands.

Null mutations in either the vab-1 Eph RTK or efn-1ephrin result in defective cell movements, and, as a result, the embryos usually die or have severe morphogenesis defects(Chin-Sang et al., 1999; George et al., 1998). However,the fact that the penetrance of the lethality is not complete suggests that there is genetic redundancy in C. elegans Eph RTK signaling, that is,other signaling pathways may work together or in parallel with the VAB-1 Eph RTK. Indeed a divergent ephrin, EFN-4, and a leukocyte common antigen-related(LAR) receptor protein tyrosine phosphatase (RPTP), PTP-3, work redundantly,or in parallel, with the VAB-1 Eph RTK(Chin-Sang et al., 2002; Harrington et al., 2002).

We report that the VAB-1 Eph RTK shows a dosage-dependent genetic interaction with the conserved axon guidance receptor, SAX-3/Robo. The Roundabout (Robo) family of receptors are evolutionary conserved and have been implicated in mediating axon guidance events, specifically axon repulsion upon binding its ligand Slit (Brose et al.,1999; Challa et al.,2001; Kidd et al.,1998; Lee et al.,2001; Seeger et al.,1993; Zallen et al.,1998) (for a review, see Wong et al., 2002). Like the vertebrate and Drosophilacounterparts, the C. elegans Robo homolog SAX-3 is required for multiple axon guidance events, including midline crossing, ventral guidance and nerve ring positioning (Zallen et al.,1999; Zallen et al.,1998). However, the role of SAX-3 in early embryonic cell movements has not been described. Four-dimensional video microscopy revealed that sax-3 embryos display defects in neuroblast gastrulation cleft closure, as well as ventral movements of the epidermal cells, during ventral closure, similar to those observed in the vab-1 (Eph RTK) and efn-1 (Ephrin) mutants. sax-3 mutants also display epidermal phenotypes that are not observed in vab-1 mutants, and we provide evidence for a role of SAX-3 in the morphogenesis of epidermal cells independently of VAB-1. We report that double-mutant combinations in sax-3 and vab-1 display a completely penetrant embryonic lethal phenotype. Furthermore, double-mutant combinations between vab-1 and slt-1, the ligand for SAX-3, show significantly enhanced cell migration defects in comparison with vab-1 single mutants, revealing a role for SLT-1 in embryogenesis. We also show that the VAB-1 tyrosine kinase binds to the intracellular portion of SAX-3,specifically at the juxtamembrane and CC1 (conserved cytoplasmic region 1)region. The SAX-3 receptor and VAB-1 Eph RTK are co-expressed on a subset of neuroblasts, which is consistent with these two receptors forming a complex. Our results suggest that SAX-3/Robo and VAB-1 can act together to regulate early neuroblast movements and axon guidance.

C. elegans strains were cultured as described by(Brenner, 1974) at 20°C unless noted. Mutants used in this study were as follows.

Chromosome II: vab-1 (e2, e699, dx31)(George et al., 1998); ptp-3 (op147) (Harrington et al.,2002); mec-7::gfp (muIs32) a GFP marker for mechanosensory neurons (gift from Dr C. Kenyon's lab).

Chromosome IV: efn-4(bx80)(Chin-Sang et al., 2002); ajm-1::gfp (jcIs1) (Mohler et al., 1998).

Chromosome X: sax-3(ky123), the ky123 allele deletes the signal sequence and the first exon and is likely to be a null allele(Zallen et al., 1998); slt-1(eh15), the eh15 allele is a duplication of the slt-1 locus with an out-of-frame deletion in both copies(Hao et al., 2001).

Rearrangement: mIn1[mIs14 dpy-10 (e128)] II(Edgley and Riddle, 2001); mIn1 mIs14 (a.k.a. mIn1GFP throughout this paper) is a dominant green fluorescent protein (GFP) balancer for chromosome II, including the region of vab-1 and ptp-3.

Mutations not referenced here can be found elsewhere(Riddle et al., 1997). All C. elegans strains were obtained from the C. elegansGenetics Stock Center, care of T. Stiernagle (University of Minnesota).

vab-1(weak); sax-3 double mutants were completely inviable and were maintained as balanced strains of the genotype vab-1(e2 or e699)/mIn GFP; sax-3(ky123). These strains segregate only viable GFP animals, as non-GFP animals (vab-1; sax-3) are dead. Balanced strains with the vab-1(dx31) null could not be maintained because of the vab-1 dosage effect (see below). efn-4; sax-3 double homozygotes were maintained as homozygous lines balanced by an extrachromosomal array (juEx350) carrying wild-type copies of vab-1 and efn-4(Chin-Sang et al., 2002). ptp-3; sax-3 double mutants were analyzed from balanced strains of genotype ptp-3/mIn1GFP; sax-3. The penetrance of lethality and morphogenetic defects were quantified as described previously(George et al., 1998).

vab-1 males were crossed to mIn1GFP/+; sax-3(ky123) animals to obtain vab-1/mIn1GFP; +/sax-3 GFP cross progeny. GFP-positive sax-3 animals were picked based on the Sax-3 notched-head phenotype (vab-1/mIn1GFP; sax-3). We were able to isolate vab-1(dx31)/mIn1GFP; sax-3(ky123) from vab-1/mIn1GFP(+);+/sax-3 mothers because sax-3 animals exhibited maternal rescue. In the subsequent generations, only mIn1GFP; sax-3 animals were observed, which is consistent with sax-3 animals requiring two copies of the vab-1 gene to survive. Of 35 putative vab-1/mIn GFP;sax-3(ky123) GFP notched-head animals, four were of the genotype vab-1/mIn1GFP;+/sax-3(ky123) double heterozygous and therefore displayed non-allelic non-complementation for the notched-head phenotype. Note that vab-1 does not show dominant phenotypes and is completely recessive (i.e. +/vab-1(dx31), 100% of embryos viable, n>500). To confirm that even in the presence of one copy of the wild-type vab-1 gene the sax-3 animals are inviable, we crossed vab-1(dx31)/mIn1GFP males to sax-3(ky123) animals and scored for the male cross progeny. Of >100 male cross progeny scored,only GFP animals (+/mIn1GFP; 0/sax-3) were observed, suggesting that the non-GFP vab-1(dx31)/+; 0/sax-3 males were dead. Other crosses using GFP-marked vab-1 males confirmed this result (see below).

Double heterozygous +/vab-1; +/sax-3 animals were constructed by crossing GFP-marked vab-1 chromosome (vab-1 (dx31) GFP (mec-7:gfp muIs32)) males to sax-3 females (animals feminized by feeding RNAi with fem-1) and +/vab-1(dx31) GFP; + or 0 /sax-3(ky123)cross progeny was identified as GFP-positive animals. All cross progeny (GFP positive) that survived developed as hermaphrodites, suggesting that the males were dead. 57% of the GFP embryos were dead (assume that 50% should be XO male embryos), suggesting that about 7% of the XX double heterozygous animals were dead (t-test compared with control cross, P<0.001). A control cross of mec-7::GFP males crossed to sax-3 females produced male and hermaphrodite survivors, and 27% of the GFP cross-progeny embryos were dead. To score the doubly heterozygous +/vab-1; +/sax-3for amphid neuron defects, we crossed vab-1(dx31) males to mIn1GFP; sax-3(ky123) hermaphrodites to isolate vab-1/mIn1GFP;+/sax-3. These animals were stained with the fluorescent dye DiI(Molecular Probes/Invitrogen) and scored for axon guidance defects (lack of ventral guidance or anterior positioning, 25%, n=87), as described previously (Zallen et al.,1999).

Yeast were grown on standard complete and selective media, as appropriate(Sherman, 1991). Yeast transformation was performed using a lithium acetate method(Schiestl and Gietz, 1989). For deletion analysis, pGBKT7 and pGADT7 (Clontech) were used as bait and prey cloning vectors, respectively, and β-galactosidase activity was measured qualitatively by X-GAL overlay assays(Serebriiskii and Golemis,2000), or was quantified by liquid β-galactosidase(Ausubel et al., 1989). At least three independent liquid β-galactosidase experiments were performed, and the mean and standard error of the mean (s.e.m.) were calculated. The VAB-1 kinase domain (669aa-985aa) was cloned into the pGBKT7 GAL-4 DNA-binding domain vector. SAX-3 deletion constructs were made in pGADT7 by cloning cDNA (PCR derived) encoding the various SAX-3 regions, and their sequence was verified. Amino acid sequences correspond to the SAX-3B isoform(Wormbase Release WS130). Primer sequences and details of plasmid constructs are available upon request.

We used a GST `pull-down' assay to confirm the SAX-3/VAB-1 interaction. A cDNA encoding the vab-1 intracellular region (581aa-1117aa) was cloned in frame to Glutathione-S-Transferase (pGEX4T-2, Amersham), expressed in E. coli (BL21 Tuner) and purified on glutathione agarose beads(GST-Bind, Novagen), according to the manufacturer's protocol. A cDNA encoding the SAX-3 juxtamembrane to CC1 region (900aa-1030aa) was fused in-frame to Maltose-Binding Protein (pMALp2x, New England Biolabs) and expressed in E. coli. Soluble extract containing MBP-SAX-3 (Load) was incubated overnight at 4°C with either GST (3 mg/ml) or GST-VAB-1 (0.25 mg/ml) bound to 100μl beads. Unbound fractions were collected, protein bound to beads was washed three times [25 mM HEPES (pH 7.5) 250 mM NaCl, 5% glycerol, 0.05%Triton-X-100] and a proportional loading of each sample was analyzed by SDS-PAGE followed by western blotting. MBP-SAX-3 was detected by using anti-MBP (New England Biolabs) and secondary HRP-anti-rabbit antibodies(Upstate) followed by ECL (Pharmacia Biotech).

Fixation and staining of embryos was performed as described previously(Chin-Sang et al., 1999; Finney and Ruvkun, 1990). Chicken polyclonal antibodies against GFP (Chemicon) were used at 1:200 dilutions. Rabbit anti-VAB-1 antibodies (antigen: VAB-1-6XHIS intracellular 581aa-1117aa) were used at 1:100 dilutions. MH27 monoclonal antibody(Francis and Waterston, 1991)(anti-AJM-1) was used at a concentration of 1:500. Rhodamine-conjugated goat anti-mouse (Chemicon), FITC-conjugated goat anti-chicken, and Texas Red-conjugated goat anti-rabbit secondary antibodies (Jackson ImmunoResearch Laboratories) were used at 1:500 dilutions.

To create a rescuing SAX-3::GFP reporter (quEx89), we used a PCR fusion-based approach (Hobert,2002) to generate a PCR (Roche Expand Long PCR) product consisting of 1.2 kb of the sax-3 promoter, the sax-3 genomic region(exons 1-5) followed by the rest of the sax-3 cDNA fused in frame to GFP and unc-54 3′UTR sequences derived from pPD95.75 (Dr A. Fire's laboratory). A final 10.4 kb PCR product (10 ng/μl) was co-injected with pRF4 (30 ng/μl) marker and transgenics were obtained as described(Mello et al., 1991). The SAX-3 minigene (quEx99) consisted of 1.5 kb of the sax-3promoter fused to the whole 3.8 kb sax-3 cDNA and followed by 0.7 kb of the sax-3 3′UTR. The final mini-gene PCR product was co-injected with odr-1::RFP transformation marker (gift from Dr C. Bargmann's Laboratory). For ajm-1::SAX-3 (quEx100) and F25B3.3::SAX-3 (quEx102), we used the same strategy as for the mini-gene, where sax-3 promoter sequence was replaced with 1.6 kb sequence of ajm-1 promoter and 1.5 kb sequence of F25B3.3promoter, respectively. The vab-1 rescuing mini-gene (pCZ47) was as described previously (George et al.,1998).

4D video microscopy was carried out essentially as described previously(Chin-Sang et al., 2002). Recordings were made at room temperature (22°C) using a Zeiss Axioplan 2 microscope with a 63× Plan-neofluar objective lens, Axiocam and Axiovison software. Fifteen to twenty (0.5 to 1 μm) z sections were taken at 1-2 minute intervals throughout development. We recorded 20 sax-3(ky123) embryos of which 10 did not complete embryogenesis. For time-lapsed imaging of GFP-expressing embryos, we replaced the standard UV(100 W Mercury short arc burner) lamp with a 250 W Halogen fiber-optic cold light source (Zeiss KL2500 LCD). Images were captured every 5-8 minutes.

The incomplete penetrance of the embryonic lethality of vab-1suggested that other genes function with vab-1 during embryogenesis. We identified sax-3 in a candidate gene screen for genes that are synthetic lethal with vab-1 (S.G. and I.D.C.-S., unpublished). We observed 48% embryonic lethality for sax-3(ky123), n=1210, and 54%embryonic lethality for vab-1(dx31), n=1065(Fig. 1). We made double-mutant combinations with a putative null allele of sax-3(ky123)(Zallen et al., 1998) and various alleles of vab-1 [e2, e699, dx31; weak, intermediate and strong (null), respectively] (George et al., 1998). All vab-1 alleles showed 100% embryonic lethality with sax-3(ky123), and the weak vab-1 alleles showed a synergistic interaction with sax-3, as the lethality was more than additive (Fig. 1).

In the course of making double-mutant combinations with vab-1 and sax-3/Robo, we identified a gene-dosage dependence of vab-1in that +/vab-1(dx31); sax-3 heterozygotes were completely inviable(see Materials and methods). We also confirmed the gene dosage requirement for vab-1 by crossing vab-1(dx31) males to sax-3animals. Because sax-3 is on the X chromosome, half of the cross progeny should be +/vab-1(dx31); 0/sax-3 (males) and the other half +/vab-1(dx31); +/sax-3 (hermaphrodites). From several independent crosses, all cross progeny developed as hermaphrodites, suggesting that heterozygous +/vab-1; 0/sax-3 males die as embryos(Fig. 1). Therefore, removing just one copy of the vab-1 gene reveals an essential role for sax-3 during embryogenesis. The double heterozygous +/vab-1;+ or 0/sax-3 embryos (hermaphrodites and males) from vab-1 crossed to sax-3 displayed greater than 50% lethality. Furthermore, +/vab-1(dx31); +/sax-3(ky123) showed weak penetrance notched-head phenotypes, demonstrating non-allelic non-complementation (see Materials and methods).

If VAB-1 functions with the SAX-3 receptor during embryogenesis, we reasoned that sax-3 mutations may also show a synthetic lethal phenotype with other genes known to function with vab-1. Two proteins have been shown to function redundantly with VAB-1 Eph RTK signaling during embryonic morphogenesis: PTP-3, a LAR (Leukocyte Common Antigen Related)Receptor Tyrosine Phosphatase (Harrington et al., 2002) and EFN-4 a divergent Ephrin(Chin-Sang et al., 2002). We made double mutants between sax-3(ky123) and ptp-3(op147) or efn-4(bx80). As predicted, double-mutant combinations caused 100%embryonic lethality. Unlike vab-1, we did not see any gene dosage effects or non-allelic non-complementation with either ptp-3 or efn-4 mutants. The embryonic lethality appears to be caused by defects in neuroblasts and epidermal movements similar to those seen in the vab-1 ptp-3 and vab-1; efn-4 double mutants, respectively(Chin-Sang et al., 2002; Harrington et al., 2002)(Fig. 2G). Both ptp-3and efn-4 single mutations have very modest embryonic lethality (10%and 14%, respectively), therefore sax-3 mutations act synergistically with ptp-3 and efn-4.

Because it is possible that VAB-1 and SAX-3 may function together during axon guidance, we asked whether these two genes display a gene-dosage interaction, as seen in the embryonic lethality phenotype. We show that the double-heterozygous +/vab-1; +/sax-3 animals display defects in axon guidance in head neurons. Twenty-five percent of the double heterozygous animals displayed an anterior positioning and/or defects in ventral guidance of the amphid neurons similar to the phenotype of sax-3 mutant animals (Fig. 3). Thus, vab-1 and sax-3 show non-allelic non-complementation in axon guidance.

We used four-dimensional (4D) video microscopy to record the embryonic development in sax-3 mutant embryos. sax-3 mutant embryos displayed defects in two phases of embryonic cell movements: closure of the ventral gastrulation cleft by short-range neuroblast movements; and enclosure of the embryo by epidermal cell shape changes, which are similar to those seen in vab-1 and efn-1 mutants(Chin-Sang et al., 1999; George et al., 1998). For this reason, they were classified using the same criteria as those used to classify vab-1 and efn-1 embryonic phenotypes (see Fig. 2).

In 30% of the sax-3 embryos examined, the gastrulation cleft remained open at the time of epidermal ventral enclosure and the epidermal cells failed to contact each other at the ventral midline. As a result, the embryos ruptured at the ventral side (Fig. 2C, Class I, strongest phenotype). In 20% of the embryos there were no apparent defects in gastrulation cleft closure, and although the epidermal cells met at the ventral side, they failed to form stable contacts(Class II). These animals also ruptured at the ventral midline. Another striking phenotype displayed by vab-1 and efn-1 mutants was an epidermal notched head, which was most likely caused by improper epidermal ventral enclosure in the anterior region. sax-3(ky123) mutant animals also show a `notched' head phenotype (17% strong notches), similar to that seen in vab-1 and efn-1 mutants, which is consistent with sax-3 exhibiting similar embryonic defects as Eph RTK signaling defective animals (Fig. 5A).

We also examined the lethality of double vab-1(e2); sax-3(ky123)mutants. These double-mutant embryos died with the most severe gastrulation cleft defects, in addition to defects in epidermal ventral enclosure(Fig. 2E, Class I). We did not see any new phenotypes in the double mutants that were not observed in the single mutants (other than the frequency of the defects)(Fig. 2E). SLT-1 is predicted to be the only C. elegans ligand identified so far for SAX-3(Hao et al., 2001), and a null mutation in slt-1 does not display embryonic lethality(Fig. 1). When we examined the vab-1(dx31); slt-1(eh15) double mutant, we found 60% embryonic lethality and these embryos arrested during ventral enclosure with a larger gastrulation cleft than vab-1 single mutants. Often in these animals,the epidermis did not initiate migration toward the ventral side and remained on the dorsal side (Fig. 2F). More cells failed to ingress during gastrulation cleft closure and the posterior pharynx was often abnormal (data not shown). This is a new phenotype, which was not observed in vab-1, sax-3, or vab-1;sax-3 mutants, suggesting that SLT-1 has a SAX-3-independent role. Therefore the vab-1; slt-1 double mutation does not lead to synthetic lethality; however, it can drastically enhance the vab-1 phenotype and unmasks an embryonic role for slt-1 in embryonic cell migration(see below). Together, these findings suggest that sax-3 and vab-1 have common and partially redundant roles during these morphogenetic cell movements.

Previously (Chin-Sang et al.,1999; George et al.,1998) suggested that the epidermal cell migration defects seen in ephrin signaling mutants arise from defects in the organization of the neuronal substrate that they migrate over, and not directly from a lack of ephrin signaling in epidermal cells. Because the DIC time-lapse showed that sax-3 mutants have similar defects in epidermal cell migration during ventral enclosure, we questioned whether there is the same deficient mechanism that is proposed in ephrin signaling. We analyzed the movement of epithelial cells in greater detail by 4D time-lapse microscopy of an AJM-1::GFP reporter construct and MH27 (anti-AJM-1) antibodies, which localize to adherens junctions of epidermal cells. Wild-type embryos carrying AJM-1::GFP (jcIs1) illustrate two epidermal morphogenetic steps that occur at approximately the same time on opposite sides of the embryo: (1) ventral migration of the contralateral pairs of epidermal cells to the ventral midline where they contact one another to wrap the embryo in the epidermal monolayer(Fig. 4A); and (2)intercalation of the two dorsal-most rows of epidermal cells that later fuse to each other and form the dorsal syncytium(Fig. 4C)(Podbilewicz and White, 1994; Williams-Masson et al., 1998). In sax-3 embryos, AJM-1::GFP was expressed and localized to the junctions of the epidermal cells, as in wild-type embryos. We analysed 45 embryos and we observed three kinds of epidermal defects in sax-3(ky123) embryos. The most dramatic defect was the disruption of the migration of the ventral pocket cells due to the lack of underlying substrate cells, caused by the failure of gastrulation cleft closure (13 embryos). In nine of those 13 embryos, the leading cells met at the ventral midline, and, in the other four embryos, the leading ventral epidermal cells failed to migrate or migrated abnormally. These 13 embryos ruptured at the ventral midline and could be classified as Class I, as described for DIC time-lapse above (Fig. 4D). The second defect observed was a change in the shape and position of the dorsal epidermal cells that led to cell intercalation and fusion defects (four embryos; Fig. 4F). In addition,three of these embryos also failed to undergo ventral enclosure. The last defects observed were abnormalities in the lateral epidermis (12 embryos). In C. elegans embryos, 10 epidermal cells (H0, H1, H2, V1-V6 and T) form a lateral row of seam cells on each side of the embryo(Fig. 4B). In sax-3mutants, one or more seam cells were displaced. For example, V1, V3 and V5 were often shifted ventrally or dorsally, resulting in inappropriate cell contacts between adjacent cells (Fig. 4E). The remaining 16 embryos enclosed and elongated without epidermal defects. In summary, the sax-3 gene is necessary for at least three different processes during epithelial morphogenesis of the embryo:for ventral epidermal migration, dorsal epidermal cell migration and fusion,and for the alignment of the lateral seam cells. In addition, sax-3is required for the neuroblast movements during gastrulation cleft closure.

We also examined vab-1(dx31) mutants for AJM::GFP phenotypes found in sax-3 mutants. As in sax-3 mutants, the most dramatic defect was the presence of an open gastrulation cleft, which prevented the epidermal cells from migrating and enclosing the embryo (Class 1, 6 out of 20 embryos examined, Fig. 4G). Furthermore, four embryos had no apparent ventral enclosure defects, but eventually ruptured later at the ventral midline. Two embryos had similar seam cell displacements to those observed in sax-3 embryos. For example,V5 was shifted ventrally, and V4 made direct contact with V6(Fig. 4H). However, unlike sax-3 mutants, vab-1 mutants displayed no defects in the dorsal epidermal cells. By contrast, the vab-1; slt-1 double mutant did exhibit more severe epidermal defects than those observed in vab-1 alone. Like in sax-3 mutants, the epidermal structure was highly disorganized on both ventral (8 out of 22 embryos) and dorsal sides(2 out of 22 embryos; Fig. 4J,L) and more seam cells had abnormal shapes or were displaced than in vab-1 mutants alone (four embryos)(Fig. 4K,L). Together, these findings indicated that the VAB-1 Eph RTK is required for epidermal morphogenesis but only for a specific subset of cells. Furthermore vab-1;slt-1 double mutants exhibited phenotypes not exhibited in either of the single mutants, suggesting functional redundancy during epidermal morphogenesis.

Because the dorsal epidermal defects observed in sax-3 mutant embryos cannot be easily reconciled by neuroblast signaling (as is proposed for the ventral epidermal defects observed in vab-1 and efn-1), we questioned whether SAX-3 could function cell autonomously in epidermal cells. We expressed SAX-3 specifically in either epidermal or neuronal cells using the ajm-1 and F25B3.3 promoters,respectively. SAX-3 tissue-specific expression was scored for its ability to rescue the embryonic lethality of sax-3 mutants. The ajm-1gene is activated only after the epidermal cells are specified and therefore the ajm-1 promoter is suitable to quantify the activity of SAX-3 in epidermal cells (Mohler et al.,1998). F25B3.3 is a RAS1 guanine nucleotide-exchange factor that is expressed early in neuroblasts and throughout the adult nervous system (Altun-Gultekin et al.,2001). Both ajm-1::SAX-3 and F25B3.3::SAX-3 constructs were able to partially rescue the lethality in sax-3(ky123) mutant embryos (Fig. 5A), reducing the embryonic lethality from 48% to 23% for ajm-1::SAX-3; sax-3 transgenic animals, and to 17% lethality for F25B3.3::SAX-3; sax-3 transgenic animals, suggesting that the sax-3 gene functions both in epidermal and neuroblast cells. Surprisingly, the epidermal `notched-head' phenotype was rescued only by the neuronal F25B3.3::SAX-3 construct and not the epidermal ajm-1::SAX-3 construct (Fig. 5A).

SAX-3 expression has been reported in the nervous system, particularly during the stage of axonal outgrowth in the embryo(Zallen et al., 1998). However, we re-examined the SAX-3 expression pattern (specifically in the embryo), and we were able to extend the SAX-3 expression pattern by creating transgenic animals that carried the entire SAX-3 receptor fused to GFP expressed from its own promoter (see Material and methods). The SAX-3::GFP construct was able to partially rescue the sax-3 embryonic lethality and the notched-head phenotype, suggesting that SAX-3::GFP has functional activity (Fig. 5A). SAX-3::GFP is expressed in early neuroblasts during gastrulation cleft closure and in the underlying neurons during ventral enclosure, similar to VAB-1 and EFN-1. However, unlike VAB-1 or EFN-1, SAX-3::GFP was also observed in epidermal cells (Fig. 5B)(Chin-Sang et al., 1999; George et al., 1998). To examine SAX-3 and VAB-1 co-localization, we used anti-VAB-1 antibodies and anti-GFP antibodies on transgenic embryos expressing VAB-1 (pCZ47)(George et al., 1998) and the SAX-3::GFP construct. VAB-1 and SAX-3 co-localized on neuroblasts, consistent with SAX-3 and VAB-1 functioning together during this stage of development. However, the co-localization is not exact and VAB-1 and SAX-3 were also expressed in separate cells. In general, we found that co-localization of VAB-1 and SAX-3 occurred more frequently in the early neuroblasts, but in later embryos (post-ventral enclosure), we found that SAX-3 and VAB-1 were expressed in separate groups of neighboring cells, reminiscent of the reciprocal expression patterns of ephrin ligands and their receptors(Fig. 5C).

Non-allelic non-complementation or gene-dosage sensitivity between two genes can be interpreted genetically as two proteins acting in a complex. Because VAB-1 exhibits gene-dose sensitivity with SAX-3 in embryogenesis and in axon guidance, we tested whether these two receptors can physically interact. We used the VAB-1 tyrosine kinase region fused to the GAL-4-binding domain, and the full-length intracellular region of SAX-3 fused to the GAL-4-activation domain, to show that these two proteins can interact in a yeast two-hybrid assay (Fig. 6A,B).

Deletion analysis of the SAX-3 protein determined that the juxtamembrane and CC1 domains of SAX-3/Robo are necessary and sufficient for binding the VAB-1 kinase (Fig. 6B). Neither the juxtamembrane nor the CC1 domain alone showed significant binding to VAB-1, suggesting that both these regions are necessary for the interaction. We further confirmed the interaction between SAX-3 and VAB-1 by Glutathione S-Transferase (GST) pull-down experiments (see Materials and methods). A fusion protein consisting of Maltose Binding Protein (MBP) fused to the SAX-3 juxtamembrane and the CC1 region (MBP-SAX-3 juxta CC1) specifically co-precipitated with GST-VAB-1 (Fig. 6C).

We show that the SAX-3/Robo receptor functions with the VAB-1 Eph RTK in neuronal and epidermal morphogenesis. Time-lapsed video analysis revealed that sax-3 mutants have similar cell movement defects to those of vab-1 (Eph RTK) and efn-1 (Ephrin) mutants. Our genetic,expression pattern and in vitro interaction analyses suggest that SAX-3 and VAB-1 may function together to receive signals on neuroblasts and neurons to regulate cell movements and axon guidance. In addition, SAX-3/Robo has roles in epidermal morphogenesis that function independently of VAB-1.

We report that SAX-3 is necessary for cell migration in two phases of embryonic development, closure of the gastrulation cleft by short-range movement of neuroblasts, and epidermal cell movement during ventral enclosure;defects in cell migration during these two phases are also observed in ephrin-signaling mutants.

In C. elegans, cells are generated from a stereotypical cell lineage during embryogenesis. Most of the epidermal cells are descended from the AB founder cell, with the exception of the posterior dorsal epidermal cells, which are derived from the C founder cell(Podbilewicz and White, 1994; Sulston et al., 1983). Our results indicate that sax-3 mutants have epidermal defects only from cells derived from the AB lineage and not the epidermal C lineage. Previous results showed that SAX-3 has roles in cell positioning and axon guidance. The CAN, and HSN neurons are mispositioned in sax-3 mutant animals(Zallen et al., 1999). The ALM neuron is also mispositioned in sax-3 and slt-1 mutants(Hao et al., 2001). Similarly, vab-1 animals show defects in CAN and PLM cell positioning (A. Mohamed and I. D. Chin-Sang, unpublished). One simple hypothesis for the epidermal defects observed in sax-3 embryos is the failure of AB epidermal precursors to reach their final position because of incomplete migration, and slow or erratic movements. As a result, the mispositioned epidermal cells observed in sax-3 animals could reflect an indirect consequence of the failure of neuroectoblast (AB lineage) movement earlier in development. This is supported by the observation that expression of F25B3.3::SAX-3 in neuroblasts was able to partially rescue the sax-3 defects. However, the ability of ajm-1::SAX-3 to rescue sax-3 embryonic lethality clearly indicates that SAX-3 could function autonomously in epidermal cells.

The lethality in both sax-3 and vab-1 embryos was caused by a defect in cell movements and not as a result of defective cell differentiation. For example, the structure of the pharynx and intestine were formed correctly in Class I terminal-stage vab-1 and sax-3embryos, and the GABAergic motoneurons and mechanosensory neurons were specified and expressed using the unc-25::GFP and mec-4::GFP markers, respectively (data not shown). The body muscles were also differentiated judging by the presence of muscle twitching in terminal stage mutants.

The Robo receptor family is characterized by an extracellular domain consisting of two to five immunoglobulin domains and two to three fibronectin type III repeats, and a cytoplasmic domain containing various combinations of four short Conserved Cytoplasmic motifs called CC0, CC1, CC2 and CC3 (Bashaw et al., 2000; Kidd et al., 1998). These motifs are thought to serve as binding sites for various intracellular signaling molecules, including Enabled (Ena) and the Ableson (Abl) tyrosine kinase (Bashaw et al., 2000). We have shown an interaction between the VAB-1 tyrosine kinase domain and the juxtamembrane and CC1 region of SAX-3/Robo. SAX-3 does not have a CC0 consensus, which is located in the juxtamembrane region of Robo receptors;however, there are seven tyrosine residues in the SAX-3 juxtamembrane region that VAB-1 could potentially phosphorylate. Interestingly, the human and mouse Robo3/RIG-1 lacks the CC1 domain and differs from the other Robo homologs because it is required for midline crossing, which is opposite to the role of Robo receptors. It appears to accomplish this by repressing Robo1/2 signaling(Sabatier et al., 2004). Thus,the CC1 region may be important in regulating a repulsive signaling downstream of Robo. It will be of interest to determine whether VAB-1 phosphorylates SAX-3 to alter its function, or whether the SAX-3/VAB-1 interaction affects the UNC-40/SAX-3 interaction or the ability of Abl to regulate SAX-3/Robo function.

We propose a model in which SAX-3 has two cooperative roles. The first is to function together with VAB-1 in AB neuroectoblast movement during late gastrulation and ventral enclosure to provide a substrate or a signal for proper epidermal cell movements specifically on the ventral side of the embryo. A second role for SAX-3 may be to function independently of VAB-1 in the epidermis for epidermal movements. For example, on the dorsal side,because sax-3 animals display dorsal epidermal defects that are not shared by the vab-1 null (Fig. 7). Similarly, sax-3 animals display axon guidance defects not seen in vab-1 mutants, suggesting that SAX-3 has VAB-1-independent roles during axon development(Zallen et al., 1999). This is also consistent with their expression patterns, where SAX-3 and VAB-1 are co-expressed early in development but later are expressed in separate cells. George et al. showed by genetic mosaic analysis that loss of vab-1 in AB lineages caused strong morphogenetic defects, including the `notched-head'epidermal defect (George et al.,1998). Similarly, the sax-3 `notched-head' defects could arise as a result of a requirement for SAX-3 in the AB lineages, as F25B3.3::SAX-3 expression in the AB neuroectoblast specifically rescues this defect; by contrast, the ajm-1::SAX-3 expression in epidermal cells did not rescue the epidermal notched-head defect. Furthermore,the double vab-1; sax-3 mutants do not exhibit any new phenotype over those observed in single mutants, suggesting that SAX-3 and VAB-1 have common and partially redundant functions during morphogenetic movements.

In vitro binding of the VAB-1 kinase domain, and the juxtamembrane and CC1 region of SAX-3, support the formation of a co-receptor complex between VAB-1 and SAX-3. Non-allelic, non-complementation and a gene-dosage requirement for VAB-1 in the absence of the SAX-3/Robo receptor also provide genetic evidence to support the formation of a complex between VAB-1 and SAX-3. Proteins that physically interact with Robo receptors appear to be sensitive to dosage or show non-allelic non-complementation. For example, in Drosophila,null mutations in the slit ligand fail to complement null alleles of Robo (Kidd et al.,1999), and Robo mutants dominantly enhance Vilse, which encodes a Rac/Cdc42 GAP(Lundstrom et al., 2004) and binds Robo specifically at the CC2 region. In C. elegans,removing a single copy of UNC-34/enabled, an interacting effector of SAX-3/Robo signaling, significantly enhances the nerve ring phenotypes of a weak allele of sax-3 (Yu et al.,2002). Robo receptors have been reported to have the potential to form homodimers, as well as heterodimers with the Netrin receptor UNC-40/DCC(Hivert et al., 2002; Stein and Tessier-Lavigne,2001; Yu et al.,2002). What is the significance of a SAX-3/VAB-1 receptor complex?The Robo receptors do not have any apparent catalytic activity and thus the association of VAB-1 Eph RTK and SAX-3/Robo may provide SAX-3/Robo with a different function than when it signals independently. In C. elegansthere is only one Robo receptor, whereas in Drosophila there are three and in vertebrates there are at least four Robo homologs. A combinatorial code of Robo receptors has been proposed where various homo- and heterodimeric combinations of Robo, Robo2 and Robo3 can determine axon trajectory in the Drosophila CNS(Rajagopalan et al., 2000; Simpson et al., 2000). Because C. elegans possesses only one Robo receptor, this combinatorial code does not exist per se, therefore the interaction with other receptors, such as VAB-1 Eph RTK, may provide the signaling diversity required during development. The sax-3 and vab-1 genetic interactions are reminiscent of the unc-5 and unc-40 genes in hermaphrodite distal tip cell (DTC) migration. These two genes encode Netrin receptors that can function either independently or together for repulsion in the hermaphrodite gonad distal tip cells (Hong et al., 1999; Merz et al.,2001).

The vertebrate Robo receptors and Eph RTKs have been shown to mediate axonal repulsion, and in some cases to work in the same neurons. It has even been suggested that the downstream signaling between these two neuronal receptors converge on phosphatidylinositol-3-kinase (PI3K) signaling(Wong et al., 2004). The vertebrate Eph RTKs have also been shown to function during angiogenesis, and,curiously, Magic Roundabout (Robo4) is expressed specifically in the vasculature and is essential for angiogenesis(Bedell et al., 2005; Suchting et al., 2005). It will be interesting to know whether in other organisms the Eph RTKs and Robo receptors work together in a similar fashion to that which we propose for the early morphogenetic movements in C. elegans.

Some nematode strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center of Research Resources (NCRR). We thank T. Papanicolaou for help with microinjections. We thank members of the Chin-Sang Laboratory for reviewing and commenting on the manuscript. Special thanks to Drs D. Lum, J. Gaudet, A. Colavita and A. Chisholm for advice on the manuscript. This work is supported by a Canadian Institutes of Health Research (CIHR) operating grant and infrastructure awards from the Canadian Foundation for Innovation and Ontario Innovation Trust. I.D.C.-S. is a Research Scientist of the Canadian Cancer Society through an award from the National Cancer Institute of Canada.