The Slit Receptor EVA-1 Coactivates a SAX-3/Robo–Mediated Guidance Signal in C. elegans

See allHide authors and affiliations

Science  28 Sep 2007:
Vol. 317, Issue 5846, pp. 1934-1938
DOI: 10.1126/science.1144874

This article has a correction. Please see:


The SAX-3/roundabout (Robo) receptor has Shiga-like toxin 1 (SLT-1)/Slit–dependent and –independent functions in guiding cell and axon migrations. We identified enhancer of ventral-axon guidance defects of unc-40 mutants (EVA-1) as a Caenorhabditis elegans transmembrane receptor for SLT-1. EVA-1 has two predicted galactose-binding ectodomains, acts cell-autonomously for SLT-1/Slit–dependent axon migration functions of SAX-3/Robo, binds to SLT-1 and SAX-3, colocalizes with SAX-3 on cells, and provides cell specificity to the activation of SAX-3 signaling by SLT-1. Double mutants of eva-1 or slt-1 with sax-3 mutations suggest that SAX-3 can (when slt-1 or eva-1 function is reduced) inhibit a parallel-acting guidance mechanism, which involves UNC-40/deleted in colorectal cancer.

The UNC-6/netrin guidance cue and its neuronal receptors, UNC-5 and UNC-40/deleted in colorectal cancer (DCC), are used in different combinations to guide growing axons toward (by attraction) or away (by repulsion) from the ventral nerve cord (VNC) of Caenorhabditis elegans (1). The incomplete penetrance of pioneer-axon guidance defects observed in unc-6/netrin and unc-40 single- and double-null mutants (Table 1) suggests that other mechanisms act in parallel with netrin signaling to guide axons toward the VNC. One such mechanism involves the Shiga-like toxin 1 (SLT-1)/Slit guidance cue, a large secreted protein with several predicted N- and O-glycosylation sites (2), and its receptor SAX-3, a homolog of the transmembrane (TM) roundabout (Robo) receptor (36). Both Drosophila and vertebrate Slit bind to Robo receptors (3, 7). C. elegans SLT-1/Slit is expressed predominantly by dorsal body-wall muscles and repels SAX-3/Robo–expressing AVM and PVM pioneer axons toward the VNC (2), concomitant with UNC-40–mediated attraction of these same axons toward the VNC by ventral sources of UNC-6 (1).

Table 1.

Misguided AVM and PVM pioneer axons in mutant lines. % defective indicates the percentage of misguided AVM pioneer axons. SD was calculated for N experiments; n, total number of animals that were scored; ND, not determined.

View this table:

In C. elegans, slt-1 and sax-3 mutations affect the guidance of several of the same pioneer axons (8). For example, the pioneer axon of the lateral AVM sensory neuron in the anterior body extends toward and then along the VNC in wild-type (WT) animals (Fig. 1, A and B), but in slt-1 and sax-3 mutants, the AVM axon frequently grows directly toward the head (Fig. 1C). Cell-specific rescue experiments have demonstrated that sax-3(+)dependent guidance of AVM axons is cell-autonomous (6, 8). Although SAX-3/Robo is the only previously known receptor for SLT-1, slt-1 mutants of C. elegans do not exhibit the nerve-ring and epithelial defects of sax-3/robo mutants, suggesting that SAX-3/Robo has both Slit-dependent and -independent functions in development (2).

Fig. 1.

AVM axon guidance defects in eva-1 mutants. (Anterior is left in all panels.) (A) Representation of a lateral view of touch-receptor neurons in a WT animal. (B and C) Lateral view of AVM neurons labeled by a mec-7p::GFP transgene array (10) showing cell-body (asterisk) and axon (arrow) trajectories in a WT animal and an eva-1 mutant, respectively. (B) In a WT animal, the AVM axon pioneers a path (region of arrowhead) from its lateral cell body(asterisk) to the VNC (arrow) before turning to grow along the VNC toward the head. (C) In an eva-1(ev751) mutant animal, as in sax-3(ky123) and slt-1(eh15) mutant animals, the AVM axon frequently (Table 1) grows longitudinally toward the head (arrow) without first extending to the VNC. Scale bar, 10 μm.

We identified a TM protein, enhancer of ventral-axon guidance defects of unc-40 mutants (EVA-1), that is required to guide the AVM pioneer axon to the VNC (Fig. 1, A and B) by acting as a receptor for SLT-1. EVA-1 acts cell-autonomously, and ectopic expression of EVA-1 in SAX-3–expressing cells confers SLT-1 sensitivity to their migration. Thus, EVA-1 is predicted to be a receptor for SLT-1 that acts in conjunction with SAX-3 (as a likely co-receptor) to provide cell specificity for the activation of SAX-3 signaling by SLT-1. We also discovered a previously unknown in vivo function for SAX-3/Robo, which is to inhibit a signaling mechanism that normally functions in parallel to SLT-1 to guide pioneer axons along the dorsal/ ventral axis. We show that this parallel-acting mechanism involves UNC-40/DCC.

AVM and PVM neurons are left, right (L/R) lineal analogs with lateral cell bodies (9) that extend pioneer axons along the basal surface of the epidermis toward the VNC, which they then follow toward the head (Fig. 1A). Using a mec-7p::gfp reporter to mark touch-receptor axons (10), we identified several AVM and PVM axon guidance mutants after standard ethyl methanesulfonate mutagenesis (11), including eva-1(ev751). A predicted in-frame deletion mutant, eva-1(tm0974), was provided by S. Mitani (Tokyo Women's University) (Fig. 2 and fig. S1). In eva-1 mutants, as in slt-1 and sax-3 mutants, the initial pioneer phase (to the VNC) of AVM axon extension fails frequently, and axons instead grow along the lateral epidermis toward the head (Fig. 1C). Similar defects in PVM also occur in eva-1 mutants, but at much reduced penetrance (Table 1). Both eva-1 mutants are recessive for AVM axon guidance defects, and neither has any obvious maternal inheritance or any obvious sax-3 mutant–like body morphology defects (6, 8).

Fig. 2.

Molecular characterization of eva-1. (A to C) DNA and cDNA sequencing methods are summarized in the SOM. (A) A subclone representing F23A7.3a in the overlapping region of cosmids F14B11 and F32A7 rescues eva-1(ev751) (11). Rescue results are shown to the right, with ++ indicating nearly full rescue). (B) eva-1 mutations are also rescued by various other constructs (SOM text). Exons are indicated by orange boxes, 5′ and 3′ untranslated regions are shown with thick lines, and mec-7 promoter regions are indicated by blue lines. eva-1(ev751) has a missense mutation in the second conserved Cys (codon 186) in the fourth exon. A construct carrying the ev751 mutation only slightly rescued significantly at 20°C (Table 1) but not at 25°C, indicating that the ev751 mutation makes this protein thermolabile. (C) (Top) Nematode EVA-1 has a signal sequence for secretion (ss), two predicted Gal-binding lectin domains (green), a predicted TM domain (yellow), and predicted cytoplasmic Ser/Arg–rich domains (SR rich, orange circles). This domain organization is found in CBP04604 (Caenorhabditis briggsae), HuC21 orf63, and MmC21orf63 [proteins identified from the Down's syndrome project (12)]. The numbers show the percentage identity of EVA-1 with predicted Gal-binding and intracellular domains of related proteins (see fig. S1 for direct sequence comparisons).

The AVM axon guidance defect in eva-1 (ev751) is slightly temperature-sensitive, but deficiency heterozygotes indicate that it is null at 25°C (Table 1). eva-1(tm0974) causes a strong loss of function at 20° and 25°C. To eliminate EVA-1 function, eva-1(ev751) was used at 25°C in all experiments unless otherwise specified. eva-1 was cloned by functional rescue with injected DNAs (Fig. 2 and table S1). Partial cDNA clones were obtained from Y. Kohara (National Institute of Genetics, Mishima, Japan). The 5′ end of the eva-1 coding region was isolated by reverse transcription polymerase chain reaction using SL1- and exon-specific primers. The entire eva-1 cDNA was sequenced and was found to encode a protein of 461 amino acid residues (fig. S1A) predicted to have a hydrophobic signal sequence for secretion (residues 1 to 23), a single TM domain, and a predicted ectodomain containing N- and C-terminal predicted lectinlike galactose (Gal)–binding domains (Fig. 2 and fig. S1B). The predicted cytodomain of EVA-1 is only 70 residues and contains a Ser/Arg–rich region (residues 400 to 412) and a single phosphotyrosine binding domain consensus binding sequence (Asn-Pro-His-Tyr). EVA-1 shares domain organization with a protein encoded within the candidate Down's syndrome region of human chromosome 21 (12) and also with related mammalian and nematode proteins (Fig. 2C).

In eva-1(ev751), a point mutation in exon 4 changes the second conserved Cys of the second lectinlike domain to a Tyr (Cys186 → Tyr186) [supporting online material (SOM)] (fig. S1B). The eva-1(tm0974) allele is deleted for base pairs (bp) 5477 to 5994 in the genomic sequence of cosmid F32A7. Splicing around this would delete residues 111 to 223, predicting an in-frame protein of the size observed on Western blots (fig. S2).

A construct EVA-1(delC), deleted for the C-terminal 69 residues of the EVA-1 cytodomain, appears to rescue both eva-1 alleles (Fig. 2B and table S1). This may be because the cytodomain is not required or because of complementation via multimerization of EVA-1 (delC) with either EVA-1(ev751) or EVA-1 (tm0974), both of which are predicted to have an abnormal ectodomain but a normal TM and cytodomain. Consistent with this idea, preliminary data indicate that EVA-1 can combine as a homomultimer (fig. S3).

AVM pioneer-axon guidance is affected by mutations in unc-6, unc-40, slt-1, and sax-3 genes. To examine whether EVA-1 acts in the same pathway as UNC-6 and UNC-40, we analyzed eva-1(ev751); unc-6(ev400) and unc-40 (e1430)eva-1(ev751) double-null mutants. Their defects were either roughly equal to the additive effects of each gene (eva-1; unc-6) or had moderate synergistic effects (unc-40 eva-1) (Table 1). These results suggest that EVA-1, like SLT-1 (2), acts in parallel with UNC-6 and UNC-40 and somewhat redundantly with UNC-40 for AVM pioneer-axon guidance.

All or nearly all AVM pioneer-axon guidance is lost in the eva-1(ev751); unc-6(ev400) or unc-40(e1430)eva-1(ev751) double mutants, as was previously shown for unc-40(e1430); slt-1(eh15) (2). In contrast, PVM axon guidance is barely sensitive to mutations in eva-1 or slt-1, but relevant double mutants suggest that SLT-1 and EVA-1 have guidance functions that are almost totally redundant with UNC-6 and UNC-40 in PVM (Table 1).

We found that AVM pioneer-axon guidance defects in eva-1(ev751); sax-3(ky123) and eva-1(ev751); slt-1(eh15) double-null and eva-1(ev751); sax-3(ky123)slt-1(eh15) triple-null mutants were not significantly more penetrant than the corresponding single mutants (Table 1), demonstrating that EVA-1, SLT-1, and SAX-3 all act in the same pathway for AVM axon guidance.

To examine the cell-specific expression pattern of EVA-1, we generated green fluorescent protein (GFP) transcriptional and rescuing-translational (cDNA and genomic) reporters (Fig. 2B), driven by 3.6 (first two constructs only) or 4.1 kb of eva-1 upstream sequence and 1248 bp of endogenous 3′ genomic sequence. The translational GFP fusion protein rescued AVM (table S1) and PVM [13 out of 550 (13/550) mutant defects rescued to 0/550, respectively] and localized at or near the surface of several cell types (fig. S4F). This finding is in accordance with the predicted signal sequence and TM domain of EVA-1 (which suggests that EVA-1 is a TM protein) and is further supported by the ability of live (nonpermeabilized) EVA-1–expressing mammalian cells to bind SLT-1 (see below).

The EVA-1::GFP fusion protein is widely expressed in the developing nervous system during the period of embryogenesis (the comma stage), when most axon growth is occurring. The same result was found previously for SAX-3 (2). In first-larval-stage animals, strong expression is observed in ventral and dorsal nerve cords and in the PVM neurons but surprisingly not in the AVM neurons, which are more frequently misguided in eva-1 mutants. GFP is also detected in other neurons in the head and tail and in muscle and the hypodermis, uterus, and vulva (fig. S4). Most GFP-expressing cells have no obvious defects in eva-1 mutants but may possess signaling mechanisms that function redundantly with EVA-1, as we know to be true for PVM (Table 1).

We also examined colocalization of tagged EVA-1 and SAX-3 proteins in mammalian cells using Alexa-conjugated secondary antibodies (see SOM for procedures). The cells were visualized with goat antibody to mouse (anti-mouse) [Alexa 488 (green in Fig. 3B)] and anti-rabbit [Alexa 546 (red in Fig. 3A)] fluorescent secondary antibodies (1:250) (Molecular Probes, Invitrogen, Carlsbad, CA). Both proteins localize on or near cell membrane surfaces and in an uneven distribution that includes large regions of overlap (Fig. 3). This colocalization is consistent with the protein interaction results described below.

Fig. 3.

EVA-1 and SAX-3 colocalization in mammalian cells. Alexa tags used for 293T cell colocalization studies were observed and photographed with a Leica SP2 confocal microscope in a single z-axis plane (11). (A to D) Confocal localization patterns of EVA-1 [secondary antibody is Alexa 488–tagged (green)] and SAX-3 [secondary antibody is Alexa 546–tagged (red)] in 293T cells are shown in (A) and (B). The overlap between Alexa 546 and Alexa 488 in (C) is yellow. (D) Differential interference contrast view of cells shown in (A) to (C). Scale bar in (A), 8 μm.

Cell-specific rescue and RNA interference experiments indicate that eva-1(+) functions in AVM for pioneer-axon guidance; however, these results are not as definitive as those of mosaic analysis (SOM text). For mosaic analysis, we made transgenic animals carrying an extrachromosomal array with both sur-5p:: SUR-5::mCherry (to mark the nuclei of all cells carrying the transgene) and the rescuing-translational reporter in a strain carrying a genomic reporter for examining AVM axon morphology (mec-7p::GFP) (10). We then scored for loss of the sur-5p::SUR-5:: mCherry; eva-1p::EVA-1::GFP array in cells that readily mark appropriate lineages (fig. S5). Four out of 10 losses in a progenitor to AVM had axon guidance defects, whereas 14/14 losses in other lineages had no AVM pioneeraxon guidance defects. Two out of three mosaic animals that lost the array in QR but not in its sister cell V5R had AVM axon guidance defects, whereas 6/6 losses in V5R had normal AVM axons. These results indicate that the absence of eva-1(+) in the descendants of QR, which include only three neurons (AVM, SDQR, and AQR) but not descendants of V5 (or other cells in these animals; fig. S5) causes AVM pioneer-axon guidance defects.

SAX-3/Robo is known to have SLT-1–dependent and –independent functions (2, 8). The phenotypes and penetrance of the eva-1 null closely mimic the slt-1 null and also represent a fraction of sax-3–null mutant phenotypes and penetrance. This includes VNC midline axon crossing, nerve-ring axon guidance, CAN cell migration (2), viability, and body morphology and locomotion defects (8) (table S2). Thus, the combined data suggest that EVA-1 is required in all aspects of SLT-1–dependent but not in SLT-1–independent functions of SAX-3/Robo. Our interpretation is that there are two classes of cells that express and use SAX-3. One coexpresses EVA-1, allowing the SAX-3 receptor to respond to SLT-1, whereas the other expresses SAX-3 but not EVA-1 and consequently does not respond to SLT-1.

If this is correct, then adding EVA-1 to SAX-3–dependent SLT-1–independent cells could make them responsive to SLT-1. We found that the expression of functional EVA-1 (delC) in the touch-receptor neurons causes ALM cell-body displacements toward the head (failure to fully migrate posteriorly), like those observed in sax-3 but not in slt-1 mutants (Table 2). We found that these EVA-1–induced displacements were suppressed by mutations in slt-1, indicating that ectopic EVA-1 made the migration of these cells dependent on SLT-1 function. Thus, EVA-1 provides the specificity needed for SLT-1 to mediate the SAX-3 signaling required to guide migrating cells and, by inference, migrating growth cones.

Table 2.

ALM cell migration defects in mec-7p::EVA-1(delC) animals. ALML/R % anterior indicates the percentage of animals in which the final position of ALM is anterior to the WT position. SD was calculated for N experiments; n, total number of animals that were scored.

View this table:

The simplest interpretation of our genetic results and the known domain structure of EVA-1 suggest that EVA-1 acts as a receptor for SLT-1 required for SAX-3 signaling. To further examine this possibility, we did two kinds of binding experiments (Fig. 4 legend and fig. S6). First, we found that roughly 100-fold concentrated alkaline phosphatase (AP)–tagged SLT-1 (SLT-1::AP) could bind to 293T cells expressing either SAX-3 (as expected) or EVA-1, but not to UNC-40–expressing cells (see Fig. 4A and fig. S6 for additional evidence). Similar results were obtained with the use of luminescence-based mammalian interactome (fig. S6). Second, we found that immunoprecipitation of tagged SAX-3, but not of the unrelated transforming growth factor–β type II receptor, specifically coimmunoprecipitated EVA-1 (Fig. 4B). These experiments demonstrate that SLT-1 binds to SAX-3, as expected, and also binds to EVA-1. Because EVA-1 functions in the same pathway as SLT-1, the simplest interpretation of these results is that EVA-1 binds SAX-3 as a coreceptor for SLT-1. This idea is supported by the ability of EVA-1 to provide cell specificity to the action of SLT-1 through SAX-3.

Fig. 4.

Binding of EVA-1, SLT-1, and SAX-3 to each other. (A) 293T cells were transfected with either 3xFLAG::SAX-3–, 3xFLAG::EVA-1–, UNC-40::3xFLAG– (control at far right), SLT-1::AP– or AP-expressing plasmids. Nonpermeabilized transfected cells were incubated with SLT-1::AP or AP media (containing equal activities of AP). The FLAG-tagged proteins were expressed roughly equally, as determined by immunoblotting (fig. S7). SLT-1::AP or AP bound to the cell surface was detected by an AP color reaction (7). AP activity is visible on cells transfected with 3xFLAG::SAX-3 (SAX-3) or 3xFLAG::EVA-1 (EVA-1) and incubated with SLT-1::AP, but not on UNC-40::3xFLAG transfected cells incubated with SLT-1::AP (control at right) or on receptor transfected cells incubated with AP alone (bottom). (B) The cell lysates of 293T cells transiently cotransfected with 3xFLAG::SAX-3 and/or 3xFLAG::TGFbetaII receptor (TGFβIIR) and/or HA::EVA-1 (top) were (i) lysed and immunoblotted (IB) with anti-FLAG (total cell lysate) or (ii) immunoprecipitated (IP) with anti-FLAG (Genhunter, Nashville, Tennessee) or anti-HA, subjected to SDS–polyacrylamide gel electrophoresis, and immunoblotted with anti-FLAG or anti-HA, as shown at the left of each blot.

By scoring large numbers of animals grown at 25°C, we identified a previously unappreciated genetic interaction between sax-3 and slt-1 and a similar interaction between sax-3 and eva-1. The AVM guidance defects are roughly twice as penetrant in slt-1(eh15) (48%) and eva-1(ev751) (53%) putative null alleles than in the sax-3(ky123) putative null (22%), suggesting that although SLT-1 and EVA-1 can function with SAX-3, they may also have an equally potent SAX-3–independent function in guiding AVM pioneer axons. However, sax-3(ky123)slt-1(eh15) and eva-1(ev751); sax-3(ky123) double mutants plus eva-1(ev751); sax-3(ky123)slt-1(eh15) triple mutants each show similar penetrance to the sax-3(ky123) single mutant (21, 19, 20, and 22%, respectively) for AVM guidance defects, suggesting that EVA-1 does not have a SAX-3–independent function (Table 1). In fact, sax-3 mutations appear to be epistatic to the AVM axon guidance defects caused by eva-1 and slt-1 mutations, demonstrating that when SAX-3 is nonfunctional, ventrally oriented AVM pioneer-axon guidance is facilitated. This facilitation could occur because SAX-3, when not bound to SLT-1 or EVA-1, normally inhibits a guidance mechanism that acts in parallel to SLT-1 signaling.

Of possible relevance, it has been found for Xenopus neurons that netrin–UNC-40/DCC–mediated axon guidance in cultured cells can be silenced by a mechanism involving the stimulation of Robo by Slit2 and a consequent interaction between the cytodomains of Robo and DCC (13). These results suggest that the parallel axon guidance pathway that is inhibited by SAX-3 in C. elegans could be the UNC-6/netrin–UNC-40/DCC pathway. This possibility is even more enticing because C. elegans SAX-3 is also known to bind UNC-40, but the biological function of this interaction is unknown (14).

There is no obvious reason for silencing to exist in axons such as AVM pioneer axons, which would otherwise be pushed by one mechanism (SLT-1 signaling) and simultaneously pulled to the same place (the VNC) by a second mechanism (UNC-6/netrin signaling). However, it is possible that silencing is required to dampen a guidance-related signaling mechanism that would otherwise be oversaturated by a high concentration of the guidance cue. For example, as an AVM growth cone approaches the VNC, it may encounter such a high oversaturating concentration of UNC-6 that it can no longer sense the UNC-6 gradient. The reduced amount of SLT-1 near the VNC could allow SAX-3 to dampen UNC-6 signaling by binding and inhibiting an UNC-6 signaling component such as UNC-40, thereby restoring the ability of the growth cone to correctly interpret the guidance information provided near the high end of the UNC-6 gradient. If this were true, then at least some of the AVM pioneer-axon guidance defects of a sax-3 mutant might happen because UNC-6 signaling near the VNC would not be dampened.

The above considerations raise the counterintuitive possibility that reducing UNC-40 function would suppress the AVM guidance defects of a sax-3 mutant or a sax-3 slt-1 double mutant. To examine this prediction, we scored AVM axon guidance defects in unc-40/balancer; sax-3 slt-1 animals and found a significantly lower penetrance (P < 0.005) of AVM defects (10%) than in sax-3 single (22%) or sax-3 slt-1 double mutants (24%) (Table 1). This counterintuitive result is consistent with the model that SAX-3, in the absence of SLT-1 or EVA-1 function, can dampen or silence signaling through the UNC-40 receptor, which normally mediates the attraction of UNC-40–expressing axons toward ventral sources of UNC-6 (Fig. 5) (SOM text).

Fig. 5.

A dual-function model of how SAX-3 operates in AVM ventral guidance. In AVM, UNC-6 signaling and SLT-1/EVA-1 signaling function in parallel. Green indicates functions that are active (even inhibitory functions), whereas red indicates functions that are inactive. (A) In WT animals, SLT-1 activation and EVA-1 expression prevent SAX-3 from inhibiting a ventral-guidance signaling mechanism involving UNC-40/DCC. Therefore, both SAX-3 and parallel signaling are fully functional for ventral guidance. (B) In the absence of SLT-1 or EVA-1 function, SAX-3 no longer works as a pioneer-axon guidance receptor but is freed to inhibit ventral-guidance signaling involving UNC-40/DCC. This accounts for the higher penetrance of AVM axon guidance defects in predicted slt-1 and eva-1 null mutants as compared with sax-3 null mutants (Table 1).

EVA-1 is a previously unknown receptor for SAX-3/Robo in C. elegans with an apparent counterpart in mammals. EVA-1 acts to allow SAX-3/Robo to elicit a guidance response to SLT-1/Slit in cells expressing both receptors. EVA-1, like SLT-1, can also regulate attractive signaling of UNC-6/netrin by dampening the signaling through UNC-40/DCC in the AVM in a manner dependent on the SAX-3/Robo cytodomain (SOM). This dampening is proposed to be a mechanism for reducing the set-point sensitivity of the AVM growth cone as it moves up the UNC-6/netrin gradient, preventing the UNC-6 signals from becoming oversaturating. EVA-1 has putative Gal-binding lectinlike domains that could bind to complex carbohydrates on SAX-3/Robo, SLT-1/Slit, or even heparan sulfate proteoglycans, which have also been shown to interact with Slit and may help to localize this secreted cue in Drosophila and mammals (1517).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S8

Tables S1 and S2


References and Notes

View Abstract

Navigate This Article