The homeodomain protein CePHOX2/CEH-17 controls antero-posterior axonal growth in C. elegans

The development of the nervous system involves the navigation of axons towards their synaptic targets along complex and precise pathways. This navigation is directed by responses to repulsive and attractive cues (Tessier-Lavigne and Goodman, 1996; Chisholm and Tessier-Lavigne, 1999). In order to acquire their specific pattern of axonal projection, individual classes of neurons must differentially express receptors for these guidance cues and presumably other components of the growth cone navigational apparatus. Currently, little is known in any organism of the transcriptional control of axonal guidance. In particular, it is unclear to what extent the pattern of growth cone migration is controlled in a modular fashion, or conversely, is inextricably linked to other aspects of neuronal identity.

In vertebrates and Drosophila, the LIM class of homeodomain proteins has been implicated in binary axonal pathfinding choices. In Drosophila, the LIM-HD genes islet and lim3 combinatorially specify the choice of target muscle and the appropriate axonal pathway in a subclass of motor neurons (Thor et al., 1999). Similarly, both gain-of-function and loss-of-function experiments demonstrate that mouse Lhx3 and Lhx4 (the orthologues of Drosophila lim3) redundantly specify a ventral as opposed to a lateral exit point for spinal and some cranial motor neurons (Sharma et al., 1998). Since, for the moment, these motor neuron subclasses are distinguished exclusively by those very features that are controlled by Lhx3/4, i.e. axonal projection and position of the cell body, it remains possible that Lhx3/4 in fact controls a global fate switch. This is unambiguously the case for the C. elegans LIM-HD gene lim-4, which acts as a binary switch between two olfactory neuron identities, wherein axonal pathfinding is but one aspect (Sagasti et al., 1999). On the other hand, other LIM-HD transcription factors such as LIN-11, LIM-6 or TTX-3 (Hobert and Westphal, 2000, and references therein) disrupt neurite outgrowth patterns to various degrees, in ways which have been so far difficult to systematize. A possible case of discrete transcriptional regulation of axonal projection is that of the two Drosophila homeobox genes of the Iroquois complex, which specify a lateral, as opposed to medial, mode of axonal projection for the notum sensilla, and hence the somatotopy of mechanoreception (Grillenzoni et al., 1998).

In addition, in C. elegans, two cases of cell migration, thought to be mechanistically related to axonal pathfinding, have yielded insight into their transcriptional regulation. The different antero-posterior migrations of the descendants of the Q neuroblast (QR and QL) are determined at least in part by their differential expression of the Hox genes mab-5 and lin-39 (Wang et al., 1993; Harris et al., 1996). The anterior migration of the hermaphrodite-specific neuron HSN also depends cell-autonomously on a Hox gene, egl-5. In this case, a regulatory cascade has been outlined whereby egl-5 activates the zinc-finger gene ham-2, which cooperates with another, egl-5-independent, zinc-finger gene (egl-43) to promote anterior migration (Baum et al., 1999).

Here, we characterize CePhox2/ceh-17, a C. elegans member of the Q₅₀ paired-like class of homeobox genes, whose other members in the worm are unc-4 (Miller et al., 1992), unc-42 (Baran et al., 1999), ceh-8, ceh-10 (Svendsen and McGhee, 1995), and R08B4.2. CePhox2/ceh-17 is highly related to vertebrate Phox2a (Valarché et al., 1993) and Phox2b (Pattyn et al., 1997) and represents a probable species orthologue. In the mouse, Phox2a and its paralogue Phox2b are expressed in specific sets of neurons in which they coordinate multiple aspects of the neuronal differentiation program (Morin et al., 1997; Pattyn et al., 1997, 1999, 2000a,b; Guo et al., 1999). Like its vertebrate counterpart, CePhox2/ceh-17 is expressed exclusively in a subset of neurons; however, we show it to control much more discrete aspects of the neuronal phenotype. Specifically, both loss-of-function and gain-of-function experiments demonstrate its cell-autonomous involvement in longitudinal axonal guidance by a mechanism that does not depend on fasciculation with other axons.

The standard wild-type N2 Bristol strain, ceh-17(np1) LGI, vab-8(e1017) LGV, unc-104(e1265) LGII, lin-15(n765ts) X, Ex[flp-12::GFP] (a gift from C. Li) and muIs32 [mec-7::GFP] LGII (a gift from QueeLim Ch’ng and C. Kenyon) were grown at 20°C, unless otherwise indicated and maintained as described (Brenner, 1974). Transgenic arrays were generated in a wild-type or mutant background using standard techniques (Mello et al., 1991). Transgenic animals were identified by the dominant Roller phenotype conferred by coinjected plasmid pRF4 (that contains rol-6(su1006) DNA). Extrachromosomal arrays are designated as followed: npEx1 [pNP45 (pceh-17::GFP) + pRF4]; npEx6 [pNP46 (flceh-17::GFP) + pRF4]; npEx19 [pNP58 (ceh-17ΔHD::GFP) + pRF4]; npEx29 [pNP69 (pceh-17::GFP::3′) + pRF4]; npEx14 [pNP53 (pmec-3::ceh-17) + pRF4]; npEx18 [pNP64 (pmec-3::ceh-17ΔHD) + pRF4]; npEx10 [pNP51 (ceh-17lc) + pNP21 (pBunc-53::GFP) + pRF4]; npEx31 [pNP67 (ceh-17lcHDm) + pNP21 (pBunc-53::GFP) + pRF4]; npEx34 [p18 (odr-2::GFP; a gift from C. Bargmann) + pRF4].

A synthetic C-terminal peptide (YKEPGQISLEEIIDQI) of the CEH-17 protein was crosslinked to BSA and injected into rabbits (Neosystem). The antiserum was tested by ELISA.

To monitor ceh-17 expression, the following translational ceh-17::GFP fusions were constructed: pceh-17::GFP (pNP45) contains 1177 bp of genomic sequence upstream of the start codon of ceh-17, and the first six codons of ceh-17 (amplified with primers harboring restriction sites for NsiI and BglII at the 5′ and 3′ ends, respectively) fused to the GFP coding region of pPD95.75 (a gift from A. Fire). flceh-17::GFP (pNP46) was generated in an analogous manner, using the same 5′ primer, and a 3′ primer located 19 bp upstream of the ceh-17 stop codon, so that promoter, intronic and almost all of the coding sequences of ceh-17 were included upstream of GFP. ceh-17lc (pNP51) contains a PCR-amplified genomic sequence spanning the ceh-17 locus (positions 34954-38996 of cosmid D1007, GenBank#AF003151) cloned into pGEMT (Promega). ceh-17ΔHD::GFP (pNP58) is a derivative of pNP51 in which an AgeI-XmaI fragment, spanning the homeodomain and adjacent 5′ and 3′ sequences (−39 to +54 bp), was replaced by a GFP-containing XmaI fragment from pPD119.16 (a gift from A. Fire). pceh-17::GFP::3′ (pNP69) is a derivative of pNP58 in which the coding region of ceh-17 was removed, by replacing an AvrII-AgeI fragment by the corresponding AvrII-AgeI promoter-containing fragment of pNP45. Further details of the constructs are available upon request.

A UV/TMP-induced C. elegans deletion library of approximately 380,000 mutagenized chromosomes was made and screened using a PCR-based sib-selection procedure (Jansen et al., 1997). The ceh-17(np1) mutant was isolated, backcrossed 3 times and confirmed by PCR. The mutation is a deletion of 1353 bp whose breakpoints (positions 36642-37996 of cosmid D1007) were determined by sequencing. Rescue of the ALA axonal outgrowth phenotype of ceh-17 mutants was obtained with ceh-17lc (pNP51) and ceh-17lc+HDm (pNP67). The latter is a derivative of pNP51, in which the homeobox of ceh-17 was excised by AgeI and BamHI (39 bp upstream and 35 bp downstream of the homeobox, respectively) and replaced by the equivalent region of the mouse Phox2a gene amplified by PCR from plasmid pKS903 (Tiveron et al., 1996).

To visualize the ALA axons, we used an unc-53 promoter-GFP fusion construct, pBunc-53::GFP (pNP21) that is expressed in ALA and the motoneurons DA and AS (among others; N. P. et al., unpublished data). The extent of axonal outgrowth in wild type, ceh-17(np1), vab-8(e1017) and ceh-17(np1);vab-8(e1017) mutants and rescued transgenic lines was measured in L1 animals by scoring the position of the tip of the axon relative to the commissures of DA motoneurons. SIA axonal outgrowth was visualized in wild-type and ceh-17 strains carrying the npEx29[pceh-17::GFP::3′] extrachromosomal array, and scored by reference to the position of the anus and the presumptive gonad.

To detect CEH-17 and UNC-17, larvae and adults were washed, fixed and permeabilized according to Finney and Ruvkun (1990). To detect FMRFamide reactivity worms were treated according to Li and Chalfie (1990). Fixed worms were incubated overnight in primary antibody solution, i.e. a 1:200 dilution of rabbit anti-CEH-17, 1:10 dilution of mouse anti-UNC-17 (a gift of J. Duerr and J. Rand), or 1:100 dilution of rabbit anti-FMRFamide (a gift of C. Li). For double staining of transgenic worms that expressed GFP, animals were incubated with mouse or rabbit anti-GFP antibodies (Roche and Clontech, respectively) at a 1:100 dilution. The worms were then washed three times during the course of a day, and incubated overnight with a secondary antibody solution (1:200 dilution of Cy3-conjugated donkey anti-rabbit or anti-mouse, and 1:100 dilution of FITC-conjugated rat anti-mouse or donkey anti-rabbit antiserum (Jackson)). The worms were washed, mounted on 3% agar pads and observed using a Zeiss confocal microscope.

Neurons expressing the UNC-17 protein were identified in an unc-104 background, in which the synaptic vesicles are localized to the cell body (Hall and Hedgecock, 1991; Alfonso et al., 1993). UNC-17 staining was compared between an unc-104(e1265);npEx29[pceh-17::GFP::3′′] strain and a ceh-17(np1);unc-104(e1265); npEx29[pceh-17::GFP::3′′] strain.

To ectopically express ceh-17 in mechanosensory neurons, we cloned an amplicon encompassing the entire genomic ceh-17 coding sequence and containing NheI and NcoI restriction sites at the 5′ and 3′ ends, respectively, into pPD57.56 (a gift of A. Fire) to produce pmec-3::ceh-17 (pNP53). As a negative control, the homeodomain of ceh-17 was replaced by GFP. Briefly, a PCR fragment of the pNP58 construct (in which the ceh-17 homeobox is deleted and replaced by GFP, see above) was subcloned in pGEMT; a SpeI-SacII blunted fragment was then introduced into the NheI-EcoRV digested pPD57.56, resulting in the pmec-3::ceh-17ΔHD (pNP64) construct. These plasmids were independently coinjected with pRF4 into muIs32, a mec-7::GFP integrated line and transgenic Roller lines established.

To ectopically express ceh-17 in glr-1⁺ neurons, we cloned the same ceh-17 genomic fragment as above into the NheI-NcoI sites of the PV6 plasmid, containing the glr-1 promoter (a gift from S. Clark) to produce pglr-1::ceh-17 (pNP77). In parallel, we cloned the GFP coding sequence amplified by PCR from the pPD95.75 plasmid (a gift from A. Fire) into the SalI-SacI sites of PV6, to produce pglr-1::GFP (pNP78). These plasmids were coinjected with a plasmid containing the wild-type lin-15 gene (a gift from S. Clark) into the lin-15(n765ts) mutant and transgenic non-multivulva lines were established at 25°C.

Egg-laying, defecation and response to light body touch were assayed as previously described (Trent et al., 1983; Thomas, 1990; Chalfie and Sulston, 1981).

Inspection of the C. elegans genome revealed the presence of a 714-bp-long open reading frame predicted to encode a Q₅₀ paired-type homeodomain protein, initially named D1007.1 and now CEH-17 (T. Bürglin, personal communication), with a homeodomain highly similar to that of the murine proteins Phox2a and Phox2b (Valarché et al., 1993; Pattyn et al., 1997). Using RT-PCR, a cDNA was synthesized and sequenced, confirming that the ceh-17 gene is expressed, trans-spliced with SL1 and contains three exons spliced according to the predicted pattern (Fig. 1A). The homeodomain, encoded by the third exon, is 88% identical to that of the Phox2 proteins at the amino acid level (Fig. 1B) while the remainder of the protein sequence does not show any significant match to murine Phox2a/b or any other known proteins. The next murine member of the PRX superfamily of homeodomain proteins (Bürglin, 1994) most related to CEH-17 is Arx (80%) and the next C. elegans member most related to Phox2a/b is RO8B4.2 (77%) (Fig. 1B). These data and a phylogenetic analysis of the PRX superfamily (Galliot et al., 1999) strongly suggest orthology between ceh-17 and vertebrate Phox2a/b.

We analyzed the expression pattern of CEH-17 using an antibody raised against the C-terminal 16 amino acids (see Materials and Methods). The specificity of the antibody was confirmed by the absence of reactivity on a ceh-17 deletion mutant and restoration of the signal in rescued transgenics (see below). CEH-17 expression was strong at the 3-fold to L1 stage and decreased thereafter to reach barely detectable levels in adults (not shown). At the L1 stage, CEH-17 was detected exclusively in the nuclei (as expected for a transcription factor) of five neurons of the ring ganglia (Fig. 2A). One of the five CEH-17⁺ neurons was dorsal and just posterior to the nerve ring, while the four others were ventral and close to the posterior bulb of the pharynx. The projections of these five neurons were characterized by using a GFP fusion construct, pceh-17::GFP (Fig. 1D), which includes 1.2 kb of upstream sequence and the first six codons of ceh-17. The dorsal neuron could be seen projecting two processes towards the ring (Fig. 2B), which then turned back and coursed along the lateral cords, all the way to the tail of the animal, identifying this neuron as ALA (White et al., 1986). In adults, these axons terminated beyond the anus, in the lumbar ganglion (data not shown), supporting the notion that ALA is the source of the lateral cord-derived input to the PVC interneuron, as tentatively proposed by White (1986). The four ventral neurons projected axons to the nerve ring and then back to the dorsal or ventral sublateral cords (Fig. 2C), consistent with their belonging either to the SIA or SMB class (White et al., 1986). To distinguish between these two classes, we performed a double anti-CEH-17 and anti-GFP immunostaining on two transgenic strains that are known to express GFP in the SMB neurons (containing odr-2::GFP and flp-12::GFP constructs; C. Bargman and C. Li, respectively, personal communication). No colocalisation of GFP and CEH-17 was observed. Moreover, the four ventral CEH-17 positive neurons were located just posterior to the SMBs (Fig. 2D,E), as was described for the SIAs (White et al., 1986). Collectively, these data identify the four ventral ceh-17 expressing neurons as the SIA neurons (SIADR, SIADL, SIAVR and SIAVL).

In addition to ALA and the four SIAs, a few other neurons also expressed the pceh-17::GFP construct. To verify that these extra neurons, which were never stained by the CEH-17 antiserum, represented an ectopic expression, as often displayed by transgenic arrays in C. elegans, we designed additional constructs. First, we fused GFP to the C terminus of ceh-17, thereby adding all intronic sequences (construct flceh-17::GFP, see Fig. 1D); this resulted in the same expression pattern, albeit with a nuclear localization of GFP, probably owing to the targeting property of the homeodomain. We then replaced the homeodomain with GFP and added 0.6 kb of sequences downstream of ceh-17 (construct ceh-17ΔHD::GFP); this effectively restricted GFP expression to the five neurons seen with the antibody (i.e. ALA and the four SIAs, data not shown). The fluorescence had a granular appearance and was restricted to the cell body, possibly due to trapping of the chimeric protein in the endoplasmic reticulum. To circumvent this problem, we replaced all coding and intronic sequences with GFP (construct pceh-17::GFP::3′), which resulted in the same expression pattern (i.e. ALA and the four SIAs), but with axonal staining. These data (summarized in Fig. 1D) demonstrate that intronic sequences are dispensable for correct expression of ceh-17 and that essential cell-specific repressor elements are located 3′ of the coding sequence.

The latter construct (pceh-17::GFP::3′) was used to examine embryonic expression of ceh-17. At the comma stage, expression was detected in two dorsal neurons in the head, presumably corresponding to ALA and its sister RMED, and in six cells on the ventral side of the head, just anterior to the excretory pore (Fig. 2F,G). These cells probably correspond to the four SIAs, in which the GFP signal was strongest and revealed sprouting axons, and the two SIBVs, sisters of the SIADs (Sulston et al., 1983). Further supporting this notion was the detection of two other weakly positive cells on the ventral side, one in the mid-body region and one in the tail, at positions fitting those of motor neurons DB5 and DA8, sisters of the SIAVs (Fig. 2F,G). The earliest time at which we could detect ceh-17::GFP expression was at the end of gastrulation (Fig. 2H,I) in four pairs of cells, which could correspond to the four SIAs and their sisters or to the four SIA mothers and their own sisters. Since this is the time when the mothers of SIAs are undergoing their last division, we cannot ascertain whether ceh-17 expression precedes or follows the last cell division.

To explore the function of ceh-17, a ceh-17 deletion mutant was obtained and phenotypically characterized. This allele, np1, was isolated by PCR screen of a UV/TMP mutagen-induced deletion library (Jansen et al., 1997; B. Barstead and G. Moulder, personal communication). The deletion in ceh-17(np1) encompasses 1353 bp including 549 bp upstream of the initiator ATG and extending to the fifth codon of the homeodomain (Fig. 1A,C). Such a deletion should be a molecular null. Confirming this, no antibody staining was detected in ceh-17 worms. ceh-17 homozygotes were fully viable, fertile and exhibited no flagrant abnormalities of development. Similarly, their behavior was wild type, as judged by locomotion, mechanosensation, rate of defecation and egg-laying (data not shown), thermotaxis and thermoavoidance behavior (N. Wittenburg and R. Baumeister, personal communication). Since ALA is thought to synapse onto PVC, we tested ceh-17 mutants for their posterior body touch response in which this interneuron is involved (Driscoll and Kaplan, 1997). The response of ceh-17 mutants to posterior touch was indistinguishable from that of wild-type worms.

We then examined the presence and morphology of the five ceh-17 expressing neurons in the mutants. To visualize ALA, we expressed in ceh-17 mutant animals a pBunc-53::GFP reporter construct that strongly marks the ALA neuron as well as DA motor neuron commissures, thus providing convenient landmarks along the anteroposterior axis (N. P., unpublished observations). The two ALA axons normally run along the lateral cord all the way to the tail (Fig. 3A), stopping beyond the anus in the adult. In the mutant, ALA was present and its axons initially followed their normal trajectory, but they stopped prematurely (Fig. 3B). This stalling phenotype was highly penetrant (Fig. 3D), with 100% of ALA axons shorter than in wild type and 76% stopping close to the gonad primordium, between the DA4 and DA6 motor neuronal commissures. In a few cases (5%), the axons not only stopped before the gonad but then backtracked along the lateral cord (Fig. 3C). Other axons expressing pBunc-53::GFP (the DA and AS motoneurons) or other punc-53::GFP constructs (N. P., unpublished data) such as ALN and PLN which, like the SIAs project on the sublateral cords, had a normal axonal morphology (data not shown). The phenotype in ALA was largely rescued by reintroduction of the wild-type ceh-17 gene or of a chimeric construct in which the ceh-17 homeobox had been replaced by that of Phox2a (Fig. 3D), showing that ceh-17 is necessary for correct longitudinal outgrowth of ALA and further supporting the notion that Phox2a/b are the vertebrate orthologues of ceh-17.

We then examined the morphology of the SIA axons using the pceh-17::GFP::3′ construct as reporter. In a wild-type background, the axons of the SIAs stopped at various positions ranging from the gonad primordium to the anus. In a mutant ceh-17 background, a dramatic shortening was observed, with more than 60% of SIA axons stopping next to the gonad

Another intracellular protein reported to affect the posterior growth of ALA axons is VAB-8 (Wightman et al., 1996; Wolf et al., 1998). To test for possible interactions between ceh-17 and vab-8 we scored the ALA axonal length in a vab-8(e1017) mutant (corresponding to a strong allele) and in vab-8(e1017);ceh-17(np1) double mutants. The shortening of the ALA axons observed in vab-8 mutants was less dramatic than that observed in ceh-17 mutants, with a small proportion of axons reaching nearly wild-type positions and a wider scatter of the stop points (Fig. 3D). In ceh-17;vab-8 double mutants, axons were further shortened compared to either single mutant, with a distribution centered on the DA3 commissure (Fig. 3D). Given that ceh-17 is a null allele, this additive effect suggests that vab-8 and ceh-17 function in pathways that are at least partly parallel.

We then tested whether other phenotypic traits of ALA and SIAs were affected in the mutants. Among the very few markers available for these neurons, a FRMFamide-like reactivity is normally detected in ALA (Schinkmann and Li, 1992) and the vesicular acetylcholine transporter (VAChT) encoded by unc-17 (Alfonso et al., 1993) is expressed in SIAs (J. Duerr, personal communication). Expression of these markers was unaffected in ceh-17 mutants (see Materials and Methods and data not shown).

To investigate further the role of ceh-17 in longitudinal axonal outgrowth, we tested the effect of its ectopic expression in anteriorly projecting neurons. The gene was placed under the control of the promoter of a mechanosensory neuron-specific gene, mec-3, which directs expression in the six touch receptors (or microtubule cells, MCs): 2 ALMs and PLMs, AVM and PVM (Way, 1989). Mechanosensory neuron outgrowth was visualized with an integrated mec-7::GFP transgene (Hamelin et al., 1992). The main process (which is dendritic in nature) of the ALMs and PLMs normally courses anteriorly, dorsal and ventral to the lateral hypodermal ridge, respectively, whereas that of AVM and PVM first joins the ventral cord where it turns anteriorly. These processes stop along the anteroposterior axis at highly stereotyped positions, which do not correspond to any obvious landmarks (Fig. 4A,B,E,F). In wild-type worms and in the ceh-17 mutant the axonal morphology of the MCs was normal, as expected. When we ectopically expressed ceh-17 in the MCs, the processes of the AVM, PLM and ALM neurons started their normal trajectory, but they all failed to arrest their outgrowth at the usual position, and continued instead, often for a substantial distance (Fig. 4C,D,G,H and Table 1). For example, the PLM axon reached a position anterior to the AVM cell body in more than 20% of cases, and was occasionally more than twice its normal length, reaching as far as the nerve ring (Fig. 4G,I and Table 1). In the case of the AVM axon, most often its exaggerated anteriorward progress was stopped only when it reached the nose, whereupon it frequently turned back and grew posteriorly (Fig. 4C,D). Another phenotype is the growth of a posteriorly directed axon from the ALM cell body, which could be seen in 24% of the cases as opposed to 2% in the wild type (Fig. 4G,H and Table 1). These worms did not display any blatant mechanoreception deficit (data not shown), arguing that ectopic ceh-17 expression left untouched many aspects of the mechanoreceptor phenotype. Deletion of the homeodomain abolished the effects of the transgene (Fig. 4I and Table 1), strongly arguing that ectopic CEH-17 acts in mechanoreceptors through its transcriptional properties on specific promoter sequences.

To ensure that the overgrowth of mechanoreceptor neurites was not an idiosyncratic response of MCs to ectopic CEH-17, we performed a second gain-of-function experiment. We placed ceh-17 under the control of the glr-1 promoter, which directs expression in several head and tail neuronal classes (Maricq et al., 1995; Hart et al., 1995). In 37% of pglr-1::ceh-17 animals (n=54), an unidentified glr-1⁺ head neuron was seen projecting ectopically beyond the nerve ring towards the head, sometimes as far as the nose. In addition, in 6% of cases, the glr-1⁺ tail neuron PVQ had acquired a posterior projection never seen in wild type.

Altogether these gain-of-function data show that ectopic ceh-17 is sufficient to cell-autonomously induce excessive longitudinal axonal growth, both anteriorly and posteriorly, in many neurons.

We report here on ceh-17, a C. elegans homeobox gene of the Q₅₀ paired-like family (Galliot et al., 1999) and the probable orthologue of vertebrate Phox2a (Valarché et al., 1993) and Phox2b (Pattyn et al., 1997). Like its vertebrate counterparts, ceh-17 is expressed exclusively in the nervous system. We identified the five ceh-17 positive neurons as ALA, SIADR, SIADL, SIAVR and SIAVL. Detailed analysis of a null mutant shows that ceh-17 is dispensable for their birth, cell body position, initial pattern of axonal projection and neurotransmitter phenotype, as judged by the expression of VAChT and of a FRMFamide-like reactivity. It is, however, required for the second leg of the posterior projection of their axons (from the mid-body to the anus). Conversely, ectopic expression of ceh-17 in mechanosensory neurons leads to a striking longitudinal axonal overgrowth, without obvious perturbation of other aspects of their phenotype. Thus, although we cannot exclude other roles, ceh-17 offers the first example of a transcription factor which controls, in a rather discrete fashion, the pattern of longitudinal axonal outgrowth. We discuss below possible mechanisms.

In ceh-17 mutants, the axons of ALA and SIAs stop prematurely in the mid-body region of the animal. Misexpression of ceh-17 in mechanosensory neurons causes the opposite phenotype: excessive longitudinal growth of axons, which overshoot their normal stopping points. Collectively, these data seem to rule out three mechanisms of ceh-17 action. First, although we do not know whether ALA or SIAs pioneer their tract, the gain-of-function phenotype shows that ceh-17 action is independent of the capacity of axons to fasciculate, since the overgrowth of ALMs, AVM and PLM occurs in a region where they do not contact any axons. Thus ceh-17 stands in contrast to other genes whose loss of function disrupts longitudinal growth by interfering with fasciculation, such as unc-34, unc-71 and unc-76 (McIntire et al., 1992; Bloom and Horvitz, 1997). Second, a requirement of ceh-17 for the basic mechanics of axonal elongation, which could underlie the ALA and SIA axonal shortening, is not easy to reconcile with the excessive axonal outgrowth of MCs ectopically expressing ceh-17, and is strongly argued against by the occasional reversal of ALA growth cone migration at the gonad primordium in ceh-17(np1) mutants. Third, unlike the other main genes studied for their involvement in longitudinal neuronal or growth cone migrations, mab-5 (Harris et al., 1996), vab-8 (Wightman et al., 1996; Wolf et al., 1998) and mig-13 (Sym et al., 1999), ceh-17 does not provide the cells with directional information since it is not capable of rerouting posteriorly the anterior-bound axons of the mechanoreceptors.

How then can we explain the action of ceh-17? One possibility is that ceh-17 is indeed involved in the response to an antero-posterior guidance system, while the sign of the response, attractive or repulsive, would depend on a cell-specific or even axon-specific state of the navigational apparatus. The unc-5/unc-40 system provides a precedent for a combinatorial intracellular switch between repulsive and attractive responses to the same guidance cue, UNC-6 (see Chisholm and Tessier-Lavigne, 1999, for a review). Further arguing that ceh-17 enhances longitudinal growth without specifying an anterior or posterior direction, is its capacity to increase from 2% to 24% the occurrence of a posterior projection in ALM (see Table 1) and to induce an ectopic posterior projection in PVQ (data not shown). It should be stressed that the axonal pathfinding defects in both gain-and loss-of-function mutants do not involve dorsoventral misplacement of the axons (except after ALMs and AVM axons have bumped into the nose of the animal), and that both PVM and AVM correctly pathfind ventrally to the cord before turning anteriorly. This is in keeping with the general observation that dorsoventral and longitudinal guidance are largely regulated by distinct pathways and further suggests that ceh-17 specifically controls the latter.

A second hypothesis, compatible with both gain-and loss-of-function phenotypes and suggested by the strong bias of the mutant axons towards stalling at the gonad, is a role of ceh-17 in blinding the growth cone to a stop signal. Thus, in the absence of ceh-17, an otherwise cryptic signal from the mid-body region would induce the ALA and SIA growth cones to stop. Conversely, in the presence of ceh-17, the mechanoreceptors would ignore the stop signal(s) that normally locate the tip of their axons at precise positions on the anteroposterior axis. The source of the signal could be the same for ALA and SIAs on the one hand and the PLMs on the other, which normally stop just anterior to the presumptive vulva. It is noteworthy that the mid-body region is the final destination of several migrating cells, such as CAN, HSN and the sex myoblasts, and that several signalling centers have been identified there: the gonad primordium is the likely source of chemoattraction and repulsion for the sex myoblasts (reviewed in Antebi et al., 1997; Burdine et al., 1998) and the CAN neuron has been implicated in stopping the anterior migration of HSN (Forrester and Garriga, 1997). Also, the mid-body region contains a cryptic stalling point for the excretory canals that is unmasked by the unc-53 and unc-73 mutations (Hedgecock et al., 1987). The source of the putative stop signal, however, is clearly different in the case of ALMs and AVM, which normally peter out at the level of the pharyngeal procorpus. It is conceivable that, whatever the signal and its source, there is a generic stalling process at play inside every neuron, possibly related to growth cone collapse, that ceh-17 impinges on.

On the basis of sequence and the functional equivalence of their homeodomains, ceh-17 and mouse Phox2a/b are close relatives, probably species orthologues; however, no conserved developmental role readily appears from our current study. One theme in the function of Phox2a/b is that, in all neuronal types examined, they have early and pleiotropic roles whose abrogation results in an early differentiation block and the cell death of precursors (Morin et al., 1997; Pattyn et al., 1997, 1999, 2000a,b). This contrasts with the discrete phenotypic alterations observed in the five CEH-17 positive neurons in ceh-17 mutants. For example, Phox2a/b are required for the specification of neurotransmitter phenotype, both noradrenergic (Morin et al., 1997; Pattyn et al., 1999, 2000a) and cholinergic (Pattyn et al., 2000b; V. Dubreuil, M.-R. Hirsch, J.-F. Brunet and C. Goridis, unpublished data), while ceh-17 is dispensable for expression of the two neurotransmitter markers, VAChT and FMRFamide, in SIAs and ALA, respectively. Another striking theme of vertebrate Phox2 function is that most neurons that depend on it are connected, and form the reflex circuitry of the autonomic (or visceral) nervous system (Tiveron et al., 1996; Pattyn et al., 1997). Strikingly, a hodological correlate has also been proposed in mouse for a close relative of the Phox2 genes, DRG11 (Saito et al., 1995) and it was recently proposed that the C. elegans Q₅₀ paired-like gene unc-42 controls the development of connected neurons, several of which form the circuitry for nose-touch avoidance (Baran et al., 1999). In contrast, however, no interconnection has been described between SIAs and ALA (which are, in fact, remarkable for the paucity of their synapses; White et al., 1986). As to a higher level conservation of the role of Phox2, i.e. a functional relationship of the ALA and SIA neurons to the visceral nervous system of vertebrates, this cannot be assessed yet, since no function so far has been assigned to ALA and SIAs.

It is nevertheless noteworthy that the ALA axons lie in close apposition to the excretory canal and to the axons of the CAN neuron, both thought to be involved in osmoregulation (Nelson and Riddle, 1984; Forrester and Garriga, 1997), one of the roles of Phox2-dependent neurons in the mouse (Pattyn et al., 1999) (and A. Pattyn and J.-F. Brunet, unpublished observations). We could not, however, detect the phenotypes associated with abrogation of excretory canal or CAN function in ceh-17 mutants (data not shown).

Such an apparent lack of structure/function conversation between ceh-17 and Phox2 genes could be an artifact of our still patchy knowledge of their developmental roles. Alternatively, the structure/function relationship of these proteins is evolutionary labile. A third possibility is suggested by the conservation in CEH-17 and Phox2, often seen in pairs of orthologues, of residues in the homeodomain that are not involved in secondary structure or DNA binding and are otherwise highly variable throughout homeodomain proteins (Gehring et al., 1994) (e.g. positions 29,30): functional conservation could lie at the still largely unexplored level of protein interactors of the homeodomain itself. Such interactors could be revealed in the worm by second-site suppressor screens of ceh-17 ALA loss-of-function and MC gain-of-function phenotypes. The same screens could also reveal repressed and activated CEH-17 transcriptional targets, respectively, and new genes involved in longitudinal axonal navigation.

We thank G. Moulder for invaluable advice on target selected mutagenesis and R. Terranova for help with construction of a mutant library; C. Bargman, C. Li, C. Kenyon, C. QueeLim, A. Fire, J. Rand, J. Duerr, M. Weill, S. Clark, A. Coulson and the Caenorhabditis Genetics Center (funded by the NIH National Center for Research Resources) for the generous gift of mutant, transgenic lines, reporter constructs, cosmids and antibodies; Jim Thomas for advice on defecation tests; and C. Goridis for stimulating discussions and incisive comments on the manuscript. This work was supported by institutional grants from CNRS and Université de la Méditerranée and by specific grants from the European Community Biotech program (BIO4-CT98-0112) and the Association pour la Recherche sur le Cancer.