A Wnt signaling pathway controls Hox gene expression and neuroblast migration in C. elegans

Conserved Hox genes are required to specify regional identity along the metazoan anterior/posterior (A/P) axis. Three of the four C. elegans Hox genes, like their Drosophila and vertebrate homologs, have been shown to be expressed in broad domains along the A/P axis, and to function where they are expressed to specify regional identity: lin-39, a Sex combs reduced homolog, patterns the mid-body; mab-5, an Antennapedia homolog, patterns the posterior body region; and egl-5, an Abdominal-B homolog, patterns the tail (Kenyon, 1986; Costa et al., 1988; Chisholm, 1991; Clark et al., 1993; Wang et al., 1993).

Limiting the expression of the Hox genes to their appropriate regions is essential for correct patterning; in many organisms, misexpression causes dramatic homeotic transformations of one body region into another. For example, in C. elegans, expression of lin-39 in the mid-body of hermaphrodites directs one particular ectodermal cell type, the Pn.p cells, to make a vulva in response to inductive signals; if the posterior Hox gene mab-5 is expressed in the mid-body in place of lin-39, then mid-body Pn.p cells adopt posterior-specific fates and make a mating structure normally produced by Pn.p cells in the posterior of males (Maloof and Kenyon, 1998). Restriction of mab-5 to the posterior is also important in the V cells, lateral ectodermal cells that extend in a row along each side of the worm. Normally mab-5 is only expressed in the descendants of the posteriormost V cells. In males this expression specifies the production of sensory structures known as rays instead of cuticular structures known as alae, which are made by anterior cells. Ectopic expression of mab-5 in the anterior causes the anterior V cells to adopt the posterior fate, making rays instead of alae (Salser and Kenyon, 1996).

In some cases Hox genes must be expressed in precise spatial and temporal patterns within their general domains of function to achieve correct development. For example, Drosophila Ubx is expressed in an intricate pattern in parasegment 5, and this expression is crucial for giving parasegment 5 its unique identity (Castelli-Gair and Akam, 1995). Hox expression patterns are very dynamic in vertebrates, and it is likely that in at least some cases this dynamic expression is required for proper development (e.g. Nelson et al., 1996). In C. elegans, spatio-temporal control of mab-5 and egl-5 expression is required for proper fate specification (Salser and Kenyon, 1992, 1996; Jiang and Sternberg, 1998). Precise control of mab-5 expression is crucial for achieving wild-type development of at least two cell types, the V cells and the migratory Q neuroblasts. In the V5 lineage, mab-5 is switched ON and OFF multiple times, and each change in expression regulates a different type of cell-fate decision (Salser and Kenyon, 1996). The two Q cells are neuroblasts born in identical A/P positions but on opposite sides of the animal: QL on the left and QR on the right. In wild-type animals, after a short posterior migration, QL switches on mab-5, which, in turn, causes descendants of QL to migrate posteriorly (Salser and Kenyon, 1992). In contrast, mab-5 remains off in QR and its descendants, and as a result the descendants of QR continue to migrate towards the anterior. Activation of a heat-shock-mab-5 fusion gene specifically within anteriorly migrating QR descendants with a focused laser microbeam causes the cells to reverse direction and migrate posteriorly (Salser and Kenyon, 1992; Harris et al., 1996). Thus, mab-5 acts as a switch to control the direction of Q descendant migration: in the absence of mab-5 activity, the descendants of both Q cells migrate towards the anterior, and conversely, ectopic expression of mab-5 in both cells causes the descendants of both cells to stay in the posterior.

A number of genes have been described that are specifically required for mab-5 to be switched on in QL (Harris et al., 1996). One of these, lin-17, has been found to encode a homolog of Drosophila frizzled (Sawa et al., 1996). Members of the frizzled family of seven-transmembrane proteins can act as receptors for Wnt ligands (Bhanot et al., 1996), suggesting that the Wnt pathway might be used to control mab-5 expression and, as a consequence, the direction of neuroblast migration, in the Q cell descendants (Harris et al., 1996).

The Wnt family of secreted signaling molecules provides positional information and establishes polarity in diverse tissues in many different organisms (reviewed by Cadigan and Nusse, 1997). Typically, Wnt signals activate a conserved signal transduction cascade that leads to translocation of β-catenin/Armadillo into the nucleus (Peifer et al., 1994). Once in the nucleus β-catenin/Armadillo acts in concert with Tcf/LEF-1 related transcription factors to activate expression of downstream genes such as engrailed and the Hox gene Ubx in Drosophila, and siamois in Xenopus (Brannon et al., 1997; van de Wetering et al., 1997; reviewed by Miller and Moon, 1996; Nusse, 1997). In the absence of Wnt signals, most of the β-catenin/Armadillo protein is found complexed with cadherins at adherens junctions. This is in part due to the Zeste-white 3 kinase, which, in the absence of Wnt signals, promotes rapid degradation of cytoplasmic β-catenin/Armadillo (Peifer et al., 1994). Activation of the Wnt pathway may cause nuclear translocation of Armadillo by inhibiting Zeste-white 3 kinase activity (Orsulic and Peifer, 1996; Pai et al., 1997). In C. elegans, a conserved Wnt pathway is used for establishment of embryonic polarity (Rocheleau et al., 1997; Thorpe et al., 1997), although in this pathway the role of the Tcf homolog appears to be reversed, since in this context it functions to repress signal transduction in the absence of Wnt signaling (Rocheleau et al., 1997; reviewed by Han, 1997). Wnt and frizzled homologs are also required for the determination of epidermal cell polarity in C. elegans, although in this case the downstream effectors are unknown (Herman et al., 1995; Sawa et al., 1996).

Knowledge of how both intricate spatio-temporal and broad regional patterns of Hox gene expression are achieved is critical for understanding A/P patterning. Here we show that the egl-20 gene, known to be required for the specific activation of mab-5 in QL, encodes a Wnt signaling molecule. In addition, we show that two additional Wnt pathway genes control mab-5 expression in the V cells and Q cells. One of these, bar-1, encodes a β-catenin/armadillo-like protein (Eisenmann et al., 1998) that functions to transduce Wnt signals and that, like egl-20, is required for activating mab-5 in QL. The second, pry-1, is a negative regulator of the Wnt pathway. The Pry-1 phenotype can be suppressed by mutations in bar-1, but not by mutations in egl-20, suggesting that pry-1 normally functions downstream of the EGL-20/Wnt signal as a negative regulator of BAR-1/Armadillo activity. Together our findings show that a conserved Wnt pathway functions to specifically activate Hox gene expression in the QL cell, thereby controlling the migration of its descendants. Furthermore, mutations in pry-1 cause dramatic, widespread ectopic expression of other C. elegans Hox genes, indicating that pry-1 acts as a general repressor of Hox expression. Thus, our findings also suggest that inhibition of Wnt signaling plays an important role in restricting Hox gene expression to regional domains.

Strains were maintained using standard methods (Brenner, 1974; Wood, 1988). Unless otherwise noted, strains were maintained and analyzed at 25°C. The mutant alleles used are either described by this paper, described by Wood (1988), or referenced below.

Mutations used were: LGI: pry-1(mu38), pry-1(nc1) (S. Takagi, personal communication), mig-1(e1787) (Desai et al., 1988), lin-17(n671) (Ferguson and Horvitz, 1985), unc-101(m1), glp-4(bn2) (Strome et al., 1994), lev-10(x17). LGII: mig-14(mu71) (Harris et al., 1996). LGIII: mab-5(e2088). LGIV: egl-20(n585), egl-20(mu27), egl-20(mu25), egl-20(mu39) (Trent et al., 1983; Harris et al., 1996), him-8(e1489), lin-22(n372). LGV: him-5(e1490), muIs3[mab-5-lacZ + pRF4(rol-6d)] (Cowing and Kenyon, 1992; Salser and Kenyon, 1992), muIs13 [egl-5-lacZ + pRF4(rol-6d)] (Wang et al., 1993). LGX: bar-1(ga80) (Eisenmann et al., 1998); bar-1(mu63), bar-1(mu349), unc-6(n102), dpy-7(e88). The strain RW7000 (Williams et al., 1992) was used for STS mapping.

pry-1(mu38) was isolated in a screen for mutations affecting mab-5-LacZ expression. mab-5(e2088); muIs3 animals were mutagenized with 25-50 mM ethymethanesulfonate (EMS; Sigma) using standard procedures (Wood, 1988). F2 worms were picked clonally to individual wells of 12-well cluster plates filled with NG agar medium. Staged populations of F4s were collected as late L1s and then stained with X-gal, as described below.

pry-1(nc1) was identified in an EMS screen (S. Takagi, personal communication). bar-1(mu63) arose spontaneously in a pry-1(mu38); muIs3 background. It was noticed because unlike pry-1(mu38) single mutants, animals also carrying mu63 were active and healthy. bar-1(mu349) was identified in an EMS screen for mutations affecting the migration of the QL and QR descendants (Mary Sym, QueeLim Ch’ng and C. K., unpublished).

STS mapping (Williams et al., 1992) was used to map pry-1(mu38) to the right arm of chromosome I, closely linked to the markers Tcbn2 and hP4. 1/105 homozygous mu38 F2s acquired both STS markers and 1/105 acquired just hP4. Additional STS mapping was done by crossing an unc-101(m1) pry-1(mu38) double mutant strain to RW7000. 0/5 Pry non-Unc recombinants acquired the markers, whereas 7/11 Unc non-Pry worms acquired the markers, suggesting that pry-1 maps to the right of Tcbn2 and hP4. Three-factor mapping placed it between glp-4 and lev-10: from glp-4(bn2) lev-10(x17)/pry-1(mu38) heterozygotes, 5/9 Lev non-Glp recombinants segregated Pry progeny.

STS mapping (Williams et al., 1992) was used to map bar-1(mu63) to the middle of chromosome X, between the markers Stp103 and Stp156. Three-factor mapping placed it between unc-6 and dpy-7: from unc-6(n102) dpy-7(e88)/bar-1(mu63) heterozygotes, 4/8 Dpy non-Uncs segregated Bar progeny.

Males carrying an integrated GFP marker construct were crossed to pry-1(mu38); him-5(e1490) or pry-1(nc1); him-8(e1489) hermaphrodites. All glowing F1s (the cross progeny) were found to have the QR descendants in a wild-type position and to have normal alae/ray patterns.

pry-1(mu38)/+; GFP/+; him-5(e1490) males were crossed to nc1; him-8(e1489)hermaphrodites. Approximately one half of the glowing progeny had mis-positioned QR descendants (posterior of the vulva), and about one half of the glowing male progeny had no alae but instead had ectopic rays.

To show that mu63 and mu226 were alleles of bar-1, him-5(e1490); mu63 or him-5(e1490); mu226 males were crossed to unc-30(e191); bar-1(ga80) hermaphrodites. From each cross, all non-Unc hermaphrodite progeny had mutant QL descendants (anterior of ALM). To show that mu349 was an allele of bar-1, male GFP carrying worms were crossed to bar-1(mu226) hermaphrodites. The resulting GFP/+; bar-1(mu226)/0 males were crossed to mu349 hermaphrodites. 6/6 glowing hermaphrodite progeny from that cross were found to have QL descendants in a mutant (anterior of ALM) position.

pry-1(mu38) double mutants with mig-1(e1787), mig-14(mu71) and egl-20(n585) were all made essentially the same way. mig-1, mig-14 or egl-20 mutant males were crossed to pry-1 mutant hermaphrodites and non-Pry F1s were picked. The mig-1, mig-14 or egl-20 mutant alleles were rendered homozygous first, by picking F2s that had QL descendants in the anterior. Subsequently pry-1(mu38) was rendered homozygous by picking F3s with QR descendants in the posterior. The presence of mig-1, mig-14 and egl-20 in the strains was confirmed by observing their HSN migration defects (Desai et al., 1988; Harris et al., 1996).

The pry-1(mu38); bar-1(mu349) double mutant was made by crossing pry-1(mu38)/+ males to bar-1(mu349) hermaphrodites. F2 pry-1(mu38) progeny were picked on the basis of their mutant QR descendant positions. Subsequently, bar-1(mu349) homozygotes were obtained by picking animals with mispositioned QL descendants. A complementation test was done to confirm the presence of pry-1(mu38).

To isolate total RNA, approximately 100 μl of mixed stage bar-1(mu63) or bar-1(mu349) worms were rinsed three times in distilled water (dH₂O) and then frozen in liquid N_2. For each tube, 1 ml Trizol reagent (Gibco BRL) was added to the worm pellets and then vortexed for 10 minutes. After a 5 minute 25°C incubation followed by a 5 minute spin at full speed in an Eppendorf microfuge, the bottom layer was removed to a fresh tube and then vortexed with 200 μl chloroform for 15 seconds. After another spin (as above), the upper, aqueous phase was removed to a fresh tube and mixed with 500 μl isopropanol. After 10 minutes at room temperature RNA was collected by a 10 minute spin (as above). The pellet was washed with 75% ethanol in DEPC-treated dH₂O, spun again for 5 minutes at 7,500 revs/minute, allowed to air dry for 20 minutes and resuspended in 100 μl DEPC-treated H₂O. First strand synthesis was done using SuperScript II (Gibco BRL) and a dT20 primer. PCR with the Pwo enzyme (Boehringer) was used to amplify the cDNA in three overlapping fragments: (1) 830 bp from primer SL1 (5′-GGTTTAATTACCCAAGTTTGAG-3′) to primer SP3C (5′-GGAGTCACATGAGTAATCGC-3′); (2) 1,134 bp from primer SP5B (5′-CAGGATAATGAGGTTGTGC-3′) to SP3B (5′-ATACAACTTTCAGCAGGAGAC-3′); and (3) 1,311 bp from SP5A (5′-CAAGGAGTTTATGTGTGGC-3′) to SP3A (5′-CCAATTCA-TGAACCCG-3′). Overhanging A residues were added using Taq polymerase (Perkin Elmer) and the fragments were cloned into pGEM-T (Promega) or pCR2.1 (Invitrogen). DNA was purified using a SNAP mini-prep kit (Invitrogen) and sequenced on an ABI sequencer by the UCSF Biomolecular Resource Center.

The egl-20 gene was previously localized to the gene cluster on Linkage Group IV between the cloned genes fem-3 and mec-3 (Trent et al., 1983; Harris et al., 1996). Using restriction-fragment-length polymorphisms in the region between fem-3 and mec-3, we mapped egl-20 to an interval covered by the cosmids from C28D4 to ZZH8. We generated transgenic lines of these candidate cosmids by germline transformation (Mello et al., 1991) and scored the lines for their ability to rescue both the QL migration and egg-laying defects caused by the egl-20(n585) mutation. A single cosmid located in this interval, W07H7, rescued the egl-20(n585) mutant phenotypes. Subsequently, the cosmid W07H7 was found to be identical by restriction digestion pattern to the cosmid W08D2, which had been sequenced by the C. elegans Genome Sequencing Project. One of the predicted open reading frames (ORFs) on the cosmid W08D2 (W08D2.1) was homologous to the Wnt family of signaling molecules. A 6.3 kb HindIII fragment of the cosmid W08D2 (pJW12), containing the entire predicted Wnt gene, was subcloned into the LITMUS38 vector (New England Biolabs). Transgenic lines carrying pJW12 in an egl-20(n585) background exhibited wild-type egg-laying and QL migration. We subsequently obtained the egl-20 cDNA by RT-PCR from mixed-stage N2 RNA. We first synthesized cDNA templates using an oligo(dT) primer, Q_T (Frohman, 1993). We then amplified the egl-20 cDNA by using egl-20 specific primers. We obtained a single 1.2 kb product, which was subcloned into the pCRII vector (Invitrogen) and sequenced.

We PCR-amplified two regions of the predicted Wnt gene from four egl-20 mutant alleles: n585, mu27, mu25 and mu39. Exons 1-4 were amplified with JW1: 5′-CTTAACCAGGCAAATCGGAA-3′ and JW5: 5′-CACACATAAGACAACACCTG-3′; exons 5-10 were amplified with JW3: 5′-CGTGTCGTTATGAAATACGC-3′ and JW4: 5′-TCTTGTTTTGCTAGGTCCCG-3′. These amplified regions included the entire coding region and all intron/exon boundaries of the predicted Wnt gene. Fragments were cloned and sequenced, as described above, from two independent PCR reactions. egl-20(mu27) was isolated in a non-complementation screen using egl-20(n585), as previously reported in Harris et al. (1996). Sequence of the Wnt gene from this allele revealed that the molecular lesion is identical to that of the egl-20(n585) allele. Thus this ‘allele’ was most likely a reisolation of the original egl-20(n585) mutation.

For detection of β-galactosidase, a PAP pen (Research Products International) was used to draw hydrophobic lines on poly(lysine)-coated slides, creating a single well (12 wells were created on each slide for mutant screens). 1.5 μl of 25% glutaraldehyde (Sigma) was added to each well, followed by 15 μl of worms that had been rinsed several times in dH₂O + 0.01% Triton X-100 (Sigma). After 3 minutes of fixation an 18 mm₂ coverslip was placed over the worms, excess liquid was removed by aspiration, and the slides were frozen on dry ice. After 5 or more minutes, coverslips were prised off with a razor blade and the frozen slides were dipped in room temperature acetone (Sigma) for 1 minute. Slides were air dried and then stained overnight using a variation of the β-galactosidase stain solution described by Fire et al. (1990), with 10 mM MgCl₂.

Anti-MAB-5 staining was done as described by Salser and Kenyon (1996).

Anti-LIN-39 staining and staining using the monoclonal antibody MH27 was done as described by Maloof and Kenyon (1998).

In all cases the nuclear stain DAPI was also included to assist in cellular identification.

In the wild type, the QL cell migrates a short distance toward the posterior and then switches on expression of the Hox gene mab-5. QL’s bilateral homolog QR migrates a short distance anteriorly and does not switch on mab-5. How is mab-5 switched on in the migrating QL neuroblast? Previously we found that four genes, mig-14, egl-20, lin-17 and mig-1 are required for this process (Harris et al., 1996). In animals carrying mutations in these genes, QL migrates normally towards the posterior but does not turn on mab-5. As a consequence, the descendants of QL migrate toward the anterior. To look for additional genes that might participate in this process, we decided to screen for the opposite phenotype in hopes of finding genes required to keep mab-5 off in QR. To do this, we mutagenized a strain carrying a mab-5-lacZ reporter construct and then looked for changes in lacZ expression. One mutation, mu38, caused ectopic expression of the mab-5 reporter in QR and its descendants. We subsequently used anti-MAB-5 antibodies to confirm that the endogenous mab-5 gene was expressed ectopically (Fig. 1). mu38 defines a new gene, pry-1 (polyray); a second, slightly weaker allele, nc1, has recently been isolated in an independent screen (S. Takagi, personal communication). This allele exhibits the same spectrum of phenotypes observed in the pry-1(mu38) allele.

As expected from the ectopic mab-5 expression, pry-1(mu38) causes the descendants of QR to stay in the posterior, just like the wild-type QL descendants (Table 1). To confirm that the posterior migration of the QR descendants was due to mab-5 activity, we made a pry-1(mu38); mab-5(e2088lf) double mutant and found that in this double mutant the Q cells migrated into the anterior (Table 1). Both alleles of pry-1 were found to be recessive (see Materials and methods), suggesting that the mutations reduce gene activity and that wild-type pry-1 activity functions to keep mab-5 expression off in QR. In contrast to the large number of mutations that affect mab-5 expression in QL, few mutations affecting QR have been described. Only one mutation has previously been described that, without affecting QR’s anterior migration, causes ectopic activation of mab-5 in QR: this is a gain-of-function promoter mutation in the mab-5 gene itself (Hedgecock et al., 1987; Salser and Kenyon, 1992). Thus pry-1 defines a new class of genes involved in regulating mab-5 expression in the Q cells: genes that are required in wild-type animals to repress mab-5 in QR.

In mig-14, egl-20, lin-17 and mig-1 mutants, mab-5 expression is not initiated in QL and as a result the descendants of QL migrate into the anterior (Harris et al., 1996). Since pry-1(mu38) causes ectopic mab-5 expression in QR, we wondered if it would also allow mab-5 to be expressed in QL in mig-14, egl-20 or mig-1 mutant animals. Therefore we constructed strains that were mutant for pry-1 and either mig-14, egl-20 or mig-1 (we were unable to construct lin-17 pry-1 double mutants because the double mutant was too unhealthy). We found that in all three double mutants, mab-5 expression was activated in both QL and QR, and subsequently descendants of both both QL and QR adopted a posterior-specific pattern of migration similar to that of cells in the wild-type QL lineage (Table 1). This demonstrates that pry-1(mu38) is epistatic to mutations in all three of these genes, and suggests that pry-1 acts downstream of mig-14, egl-20 and mig-1 to negatively regulate mab-5 expression. In other words, these findings imply that in wild-type animals, mig-14, egl-20 and mig-1 activate mab-5 expression in QL by relieving pry-1-mediated repression of mab-5.

In addition to causing ectopic mab-5 expression, the pry-1(mu38) mutation also causes worms to be unhealthy: pry-1(mu38) animals are scrawny, often herniated, and uncoordinated. The mu63 mutation was discovered when healthy, active animals spontaneously appeared in a population of pry-1(mu38) animals, suggesting that pry-1(mu38) was being suppressed by a second mutation. To test whether or not suppression was due to an extragenic mutation, we outcrossed the suppressed strain to wild type. Approximately 1/4 of the F2 progeny from the outcross showed a pry-1 phenotype, demonstrating that the suppressor was not linked to pry-1 and therefore was extragenic.

We mapped mu63 to a small region of the X chromosome between the cloned genes unc-6 and dpy-7. Interestingly, the gene bar-1, which encodes a β-catenin/Armadillo-like protein, mapped to the same interval (Eisenmann et al., 1998). Mutations in bar-1 previously had been identified on the basis of a defect in vulval development suggestive of Hox gene mis-regulation. For this reason we hypothesized that mu63 might be an allele of bar-1. Indeed, we found that mu63 failed to complement bar-1(ga80), supporting the idea that mu63 is an allele of bar-1. To confirm this, we sequenced bar-1 cDNA from mu63 and found a missense mutation in the first arm repeat, which changed the conserved Leu-130 residue to Phe. Subsequently, a number of additional alleles of bar-1 were identified in a screen for mutations affecting Q cell migration (Mary Sym, Queelim Ch’ng and C. K., unpublished). We found that one of these new alleles, bar-1(mu349), causes Gln-147 (also in the first arm repeat) to be changed to a stop codon. mu349 is therefore likely to be a strong loss-of-function or null allele.

To determine whether the ability of bar-1 mutations to suppress pry-1 was specific to the mu63 allele, we constructed a pry-1(mu38); bar-1(mu349) double mutant strain. We found that this putative null allele of bar-1 was able to suppress pry-1(mu38), confirming that wild-type bar-1 is necessary for the manifestation of the Pry-1 mutant phenotype.

The BAR-1-like protein Armadillo is known to act as a transcriptional co-activator in Drosophila (van de Wetering et al., 1997), suggesting two possible ways that bar-1 mutations might act to suppress pry-1(mu38). First, wild-type bar-1 could function as an activator to promote mab-5 expression. In this case, mutations in bar-1 would suppress pry-1(mu38) by preventing ectopic mab-5 expression. Alternatively, bar-1 might act as a co-activator with mab-5 to initiate expression of downstream genes. In this case, mutations in bar-1 would not affect mab-5 expression. To distinguish between these possibilities, we examined expression of mab-5 in the Q cells of pry-1(mu38); bar-1(mu63) and pry-1(mu38); bar-1(mu349) strains. We found that mab-5 was no longer ectopically expressed in either strain (Table 1 and data not shown). Therefore, we infer that bar-1 mutations suppress the Pry-1 phenotype by preventing ectopic mab-5 expression.

Since bar-1 is required for mab-5 expression in pry-1(mu38) mutants, we wondered if it was also required for mab-5 expression in QL in otherwise wild-type animals. We found that in bar-1(mu63) mutants, mab-5 was expressed only weakly in QL, and that in the stronger bar-1(mu349) allele, mab-5 was not expressed in QL at all (Table 1 and data not shown). In addition, mutations in bar-1 caused the descendants of QL to migrate into the anterior (Table 1). Thus, bar-1 functions like mig-14, egl-20, lin-17 and mig-1, to activate mab-5 expression in QL. However, unlike these genes, bar-1 acts downstream of pry-1, since bar-1 is still required for mab-5 expression in the absence of wild-type pry-1 function.

To further investigate how mab-5 is switched on in QL, we cloned the upstream activator egl-20 by positional mapping and transformation rescue (see Materials and methods). A single cosmid, W07H7, was found to rescue the egl-20(n585) egg-laying and QL descendant migration defects. This cosmid was located in a sequenced region of the genome, and was found to contain an open reading frame (ORF) predicted to encode a Wnt family member. We found that a 6.3 kb fragment containing this ORF (as the only complete ORF) also rescued egl-20(n585). To confirm that egl-20 encoded this Wnt family member, we sequenced the entire coding region and all intron/exon boundaries of the predicted Wnt gene from three mutant alleles of egl-20: egl-20(n585), egl-20(mu25) and egl-20(mu39). We identified mutations in highly conserved cysteine residues of the putative Wnt protein in all three alleles, indicating that the egl-20 gene does encode this Wnt protein. In addition, we have recently found egl-20 mutants with similar phenotypes that contain early stop and splice junction mutations (data not shown), suggesting that these mutations might eliminate gene activity.

Comparison of the predicted EGL-20 protein sequence with other members of the Wnt family (Fig. 2) shows that EGL-20 shares the greatest identity with mouse Wnt-7B (47% amino acid identity) and mouse Wnt-7A (35% amino acid identity). As in other Wnt proteins, there is a hydrophobic region at the N terminus that could serve as a signal sequence (von Heijne, 1986). In addition, the predicted EGL-20 protein has three potential sites for N-linked glycosylation and contains 22 conserved cysteine residues.

Previous work from our laboratory showed that egl-20 and the lin-17/frizzled genes are required to activate mab-5 expression in QL (Harris et al., 1996). As described above, in this study we have found that the β-catenin/armadillo-like gene bar-1 is also required to turn on mab-5 in QL and that egl-20 encodes a Wnt molecule. Together these findings indicate that a Wnt signaling pathway, with components conserved from the Wnt signal to an Armadillo-related transactivator, functions to control the direction of Q cell migration by regulating mab-5 expression. These results also strongly suggest that bar-1 functions to transduce the Wnt signal, possibly in a manner analogous to its Drosophila homolog armadillo.

Effects of the egl-20 mutations are limited primarily to the Q cells and the cells in their immediate vicinity (Harris et al., 1996). In contrast, we found that pry-1 affects patterning throughout the animal. For example, pry-1 (polyray) is named for the most dramatic transformation caused by pry-1 mutations, the formation of ectopic male sensory rays all along the A/P body axis. In wild-type males, mab-5 expression is restricted to the descendants of V5 and V6, and specifies that they generate sensory rays. The descendants of V1-V4 do not express mab-5 and, as a result, produce ridged cuticular structures known as alae (Fig. 3A,B). Strikingly, we found that in pry-1 mutants none of the V cells made alae; instead, all the V cells generated ray-like structures and cell types (Fig. 3C-E; Table 2). These ectopic rays were dependent on wild-type mab-5 activity; in pry-1(mu38); mab-5(e2088) double mutants, almost all of the V cells gave rise to cells that made alae (Table 2).

pry-1(mu38) also affects a second V cell patterning decision: the formation of the postdeirid. The postdeirid is a sensory structure normally made by the descendants of V5.pa; however, ectopic mab-5 expression in V5.pa prevents postdeirid formation (Salser and Kenyon, 1996). We examined pry-1 mutants and found that they failed to make a postdeirid (Table 2). This defect also depends on mab-5 activity, since it is corrected by mab-5(−) mutations (Table 2).

The dependence of the postdeirid and ray phenotypes on wild-type mab-5 activity strongly suggested that pry-1(mu38) was causing ectopic mab-5 expression in the V cells. To determine if this was the case, we examined mab-5 expression in mutant pry-1 animals. We found that mab-5 was expressed ectopically in anterior V cell lineages (Fig. 4A-D). In addition, we observed that pry-1(mu38) causes mab-5 expression to begin too early in the V5 lineage (see below), starting in the V5.p cell instead of V5.pp. It is likely that this early mab-5 expression causes the loss of the postdeirid. Thus, in wild-type worms, pry-1 is required to restrict mab-5 expression spatially to the posterior V cells and temporally to a later portion of the V5 lineage so that the postdeirid and alae/ray fates can be specified correctly.

In addition to observing ectopic mab-5 expression in the V cell lineages, we also saw ectopic mab-5 expression in the anterior ventral cord neurons (Fig. 4E,F). This indicates that wild-type pry-1 is important for limiting mab-5 expression in other cell types.

As mentioned above, mutations in bar-1 could suppress the ectopic mab-5 expression in QR caused by pry-1(mu38). We found that bar-1(mu63) and bar-1(mu349) also suppressed the ectopic expression of mab-5 in the V cells and ventral cord caused by pry-1(mu38) (Fig. 4G and data not shown). Consistent with this observation, the postdeirid and ray phenotypes of pry-1 were fully suppressed in pry-1; bar-1 double mutants (Table 2).

Since the ectopic expression of mab-5 seen in pry-1 mutants required bar-1 activity, we wondered whether bar-1 played a role in wild-type V cell patterning. Surprisingly, when we examined expression of mab-5 in bar-1 single mutants, we found that mab-5 expression was normal in the V cells. In addition, the V rays were made almost as frequently as in wild type: 64 out of 76 bar-1(ga80); him-5(e1490) males examined had all six V rays, as do wild type; 12 animals had five rays, one animal had four rays, and one had no rays. In comparison, 67 of 76 him-5(e1490) animals had all 6 rays, and nine had only 5 rays. This suggests that even though bar-1 is downstream of pry-1 in the V cells, either bar-1 is not a major component of the machinery used to activate mab-5 expression in wild-type V cells, or it is redundant with other activators of mab-5. Together these results suggest that although the downstream components of the Wnt signaling pathway are functional in the V cells to repress mab-5 expression, Wnt signaling may not play a major role during wild-type development to activate mab-5 expression in the V cell lineages.

We decided to investigate how bar-1 interacts with another gene required to regulate mab-5 expression in the V cells: lin-22. lin-22 encodes a hairy/enhancer of split homolog that is required to specify the fates of the anterior V cells V1-V4 (Fixsen, 1985; Wrischnik and Kenyon, 1997). In lin-22(−) mutants, these cells adopt V5-like fates: each generally makes an ectopic postdeirid and two rays, and expresses mab-5 late in development, in a pattern similar to that of wild-type V5 descendants. Since bar-1(mu63) can suppress the ectopic mab-5 expression and the ectopic rays caused by pry-1(mu38), we wondered if it would also be able to suppress the lin-22 mutant phenotype. We constructed lin-22(n372); bar-1(mu63) double mutants, and found that its ray and postdeirid phenotypes were indistinguishable from those of lin-22(n372) alone. This result demonstrates that bar-1 mutations cannot suppress lin-22 mutations, and, therefore that bar-1 activity is not required for the lin-22(−) phenotype.

Because pry-1 is important for limiting mab-5 expression in a broad, general way – affecting the Q cells, V cells and ventral cord neurons – we wondered if it might also be involved in limiting expression of other Hox genes. To determine if this was the case, we examined the effect of pry-1(mu38) on the expression of the Abd-B homolog egl-5 and the Scr homolog lin-39. egl-5 is normally expressed in the tail of the worm where it specifies tail cell fates (Chisholm, 1991; Wang et al., 1993). We used an egl-5-lacZ fusion (Wang et al., 1993) to monitor egl-5 expression, and found that pry-1(mu38) caused it to be expressed ectopically in anterior cells (Fig. 5A,B). Similarly, expression of lin-39, which normally specifies cell fates in the mid-body of the worm (Clark et al., 1993; Wang et al., 1993; Maloof and Kenyon, 1998), was expanded anteriorly (Fig. 5C,D). These results show that pry-1 functions as a general repressor that restricts Hox gene expression to the correct domains.

To investigate whether or not bar-1 was required in pry-1(mu38) animals for ectopic expression of Hox genes other than mab-5, we examined lin-39 expression in the pry-1(mu38); bar-1(mu349) double mutant strain. We found that bar-1(mu349) was able to suppress the ectopic expression of lin-39 normally caused by pry-1(mu38) (Fig. 5E). This result indicates that bar-1 activity is required for expression of multiple Hox genes in pry-1 mutants.

During our studies we observed that young pry-1(mu38) worms show relatively normal patterns of mab-5 and egl-5 expression (Fig. 6A,B); ectopic expression does not begin until later in development. In addition, the spatial limits of lin-39 expression are similar in wild type and in pry-1(mu38) mutants at hatching (n=20); apparently spatial misexpression of this gene also does not begin until later. However, while the spatial limits of lin-39 expression are correct at hatching, expression levels in anterior ventral cord neurons (within the normal domain of expression) are higher in pry-1(mu38) than in wild type (data not shown).

To further investigate the timing of ectopic mab-5 expression in the V cells, we examined expression at 4 hour intervals, starting at hatching (Fig. 6C). We found that ectopic expression was first evident in the V5.p cell. Subsequently, ectopic mab-5 expression appeared in most of the Vn.ppp cells. Thus it seems that in these cells, wild-type pry-1 is not required to establish the initial anterior limits of expression, but rather is required to prevent ectopic expression of the Hox genes after the initial patterns are set. Thus, pry-1 may perform a function analogous to the Polycomb-group genes of Drosophila.

We also investigated the onset of mab-5 expression in QL and QR. For these experiments we determined developmental age by observing QL’s position in fixed, stained animals (the migration of QL over V5 is not affected by pry-1 mutations or by changes in mab-5 expression). ‘Young’ animals were those in which QL had not yet migrated over V5 (approximately 1.5-2.5 hours old); ‘older’ animals were those in which QL had migrated at least 50% over V5 (approximately 2.5-3.5 hours old). In young wild-type animals no mab-5 expression was seen in QL (or QR) (n=10; also S. Salser and C. K., unpublished observations). In older wild-type animals expression was seen in QL in 50% of the animals (n=10). Similarly, we found that in young pry-1(mu38) animals expression of mab-5 was off in both QL and QR (n=13). In older pry-1(mu38) animals mab-5 was expressed in QL 64% of the time and in QR 38% of the time (n=47). Thus, expression of mab-5 in the Q cells begins in pry-1 mutants at approximately the same stage that expression begins in wild-type animals, suggesting that there is temporal control of mab-5 expression operating independently of pry-1.

In this study, we have investigated the regulation of Hox gene expression in C. elegans. Our results demonstrate that a Wnt signaling pathway regulates Hox gene expression in a number of cell types. We have found that the egl-20 gene, which activates expression of the Hox gene mab-5 in the QL neuroblast, encodes a Wnt signaling molecule and that egl-20 is part of a conserved Wnt pathway that involves the lin-17/frizzled receptor homolog and the bar-1/β-catenin/armadillo-like gene. We also find that the pry-1 gene functions in this pathway downstream of egl-20/Wnt to inhibit the ability of bar-1 to transduce the Wnt signal. Our studies of these genes show that proper regulation of Wnt signaling is critical: activation of this pathway is required to promote Hox gene expression in selected cells in order to achieve proper patterning, and inhibition of this pathway is required to prevent widespread ectopic activation of Hox genes and the resulting homeotic transformations.

Prior to this study, it was known that lin-17, which activates mab-5 expression in QL (Harris et al., 1996), encodes a frizzled homolog (Sawa et al., 1996). In Drosophila, frizzled family members are known to participate in two different signaling pathways: the first one is known to be well conserved and includes an upstream Wnt signal and the downstream β-catenin/Armadillo and dTcf/Pangolin proteins. The second is a less well-defined pathway required for tissue polarity (reviewed by Adler, 1992). Our findings that egl-20 encodes a Wnt signal and that the β-catenin/Armadillo-like protein BAR-1 is required to turn on mab-5 in QL argues that the C. elegans mig-14, egl-20, lin-17, mig-1, pry-1 and bar-1 genes all participate in a conserved pathway analogous to the first described above. The mig-14 and mig-1 sequences have not yet been reported. However, another gene, mig-5, has been reported to be a homolog of the Wnt pathway member disheveled, and to have a QL migration phenotype (Guo, 1995). It is not known whether this mutation affects the same aspect of Q migration, but it seems possible that it too functions in the Wnt pathway to activate expression of mab-5 in QL.

Our epistasis results imply that pry-1 functions downstream of egl-20 and mig-1 but upstream of bar-1. Together with the expression data, this suggests that in wild-type worms PRY-1 inhibits BAR-1 activity, thereby preventing Hox gene expression. Our genetic analysis is consistent with PRY-1 acting either downstream of or in parallel to BAR-1. PRY-1 protein could directly repress BAR-1 activity, or it could act at a more downstream point in the pathway; for example, by binding to the targets of BAR-1 and rendering them insensitive to BAR-1 activity (see below). Although the molecular identity of pry-1 is not known, it could encode a C. elegans homolog of a Wnt pathway regulator such as zeste-white 3 (Peifer et al., 1994; Siegfried et al., 1992), Axin (Zeng et al., 1997), or APC (Munemitsu et al., 1995; Polakis, 1997; Rubinfeld et al., 1996), or it could encode a novel protein. The fact that egl-20, lin-17 and mig-1 mutants share some phenotypes that bar-1 mutants do not have (for example a defect in the migration of the HSN neuron (Harris et al., 1996; data not shown) raises the possibility that they may activate a second type of signaling pathway in other cells.

One of the most interesting aspects of Q cell migration is the fact that in spite of being born in bilaterally symmetric A/P positions, QL and QR and their descendants migrate in opposite directions. One critical difference between these cells is the expression of mab-5 in QL but not in QR. This difference is necessary and sufficient for controlling the behavior of the Q descendants. Why is it that mab-5 is switched on in QL but not in QR? In principle, this regulation could be exerted at the level of the Wnt signaling pathway, or, alternatively, QR might be intrinsically different from QL and simply be immune to Wnt signaling. The observation that pry-1 is required to keep mab-5 off in QR demonstrates that QR is sensitive to activation of the Wnt pathway and thus has all of the components downstream of pry-1 necessary for activation of mab-5. Therefore, in wild-type animals, some step in the Wnt signaling pathway upstream of pry-1 must be blocked in QR.

pry-1 is required in many tissues to keep Hox genes from being expressed outside of their normal domains of function along the A/P body axis. Interestingly, the initial expression patterns of both mab-5 and egl-5 in the epidermis are largely normal; dramatic ectopic expression is not seen until many hours after hatching. For this reason, pry-1 (polyray) mutants phenotypically resemble the Drosophila Polycomb-group (PcG) mutants, which are characterized by exhibiting normal patterns of Hox gene expression early in development, but later exhibiting widespread Hox gene misexpression. The cloning of some of these genes has suggested that they repress Hox gene expression by affecting chromatin dynamics (reviewed by Pirrotta, 1997). Why might pry-1 and PcG mutants have analogous phenotypes? The simplest interpretation of our findings is that the molecular mechanism of pry-1 activity is different from that of the cloned PcG genes. This is because pry-1 appears to be a member of a conserved Wnt signaling pathway, since pry-1 mutations can be suppressed by mutations in the β-catenin/armadillo-like gene bar-1. Since Armadillo is thought to be regulated by phosphorylation, which in turn affects protein stability, it would not be surprising if the C. elegans PRY-1 protein regulated BAR-1 in a similar fashion. Nevertheless, it is possible that there is a connection between Wnt signaling and the PcG genes. To explore this possibility it would be interesting to determine if mutations in the Polycomb-group genes can be suppressed by armadillo mutations in Drosophila.

The Drosophila Wnt homolog wingless (wg) is known to control transcription of Ubx in the developing midgut (van de Wetering et al., 1997). However, in contrast to the C. elegans pathway described here, the Wnt pathway in Drosophila is not known to play a large, organism-wide role in regulating Hox expression. There are several possible explanations for the apparent difference. One possibility is that the Drosophila Wnt pathway may be more important for regulating Hox genes than has been realized. The Hox misexpression phenotype may have been missed due to other more dramatic consequences of mutations in the Wnt pathway. If so, it is possible that some of the uncloned Polycomb-group members could in fact be regulators of Wnt signal transduction. Another possibility is that pry-1, like some known Polycomb- group members, affects chromatin stability. In this scenario one would imagine that PRY-1 activity somehow renders chromatin resistant to transcriptional activation by BAR-1/Armadillo. Finally, the differences in the importance of Wnt signaling for Hox gene expression may simply reflect evolutionary differences between worms and flies. For this reason it will be interesting to learn whether the Wnt pathway is a relatively minor or major regulator of Hox gene expression in other species.

We found that the bar-1(mu63) mutation was not able to suppress the ectopic rays caused by lin-22(n372). This finding indicates that lin-22(+) activity is still required to repress ray production in anterior V cells even in the absence of bar-1. One possibility is that wild-type bar-1 activity promotes the formation of ectopic rays, at least in part, by inhibiting or bypassing lin-22-mediated ray repression. The pattern of ectopic mab-5 expression differs in lin-22 and pry-1 mutants. In lin-22 mutants, the V cells express mab-5 in an ON-OFF pattern similar to that of wild-type V5 (Wrischnik and Kenyon, 1997), whereas in pry-1 mutants, mab-5 is expressed at an earlier time than normal and simply remains on. Our model, which would be analogous to the role of the Polycomb system in Drosophila, is that lin-22(+) activity can only commence repression of mab-5 if Hox gene expression already off. In wild-type animals, pry-1 activity keeps the Hox genes off by preventing bar-1 from activating their expression. In pry-1 mutants, this early repression fails, so lin-22 cannot act. However, in bar-1(−) animals, mab-5 remains off during early larval development so that lin-22 is also able to commence repression of mab-5.

bar-1 and pry-1 function quite differently in different cell types. Not only do the Hox genes under their control vary from tissue to tissue, but the precise roles that bar-1 and pry-1 play during wild-type development vary as well. pry-1 represses the Abd-B homolog egl-5 in epidermal cells, the Scr homolog lin-39 in ventral cord neurons, and the Antp homolog mab-5 in the Q cells, the V cell descendants, and the ventral cord neurons. In the Q cells, bar-1 and pry-1 function as part of an ON/OFF switch that keeps mab-5 expression off in QR and initiates mab-5 expression in QL. In the V cells, however, their primary role seems to be to prevent mab-5 expression in the anterior body region; pry-1 mutations cause ectopic mab-5 expression, but loss of bar-1 has little or no effect on mab-5 expression or V cell development. In contrast, bar-1 has been found to be required in the Pn.p cells to maintain lin-39 activation after expression has been initiated by other factors (Eisenmann et al., 1998).

Cells also change their responsiveness to bar-1 and pry-1 during development. For example, although pry-1 mutations cause ectopic expression of mab-5 in QR, this ectopic expression does not appear until the time that wild-type mab-5 expression begins in QL, as if the Q cells are not competent to respond to activation of the Wnt pathway until that time. Similarly ectopic expression in the V5 lineage does not begin until late in the first larval stage. It seems that there are at least three phases to the regulation of mab-5 in the V5 lineage: first mab-5 is kept off by a pry-1-independent mechanism; later, at the time when expression begins in the pry-1 mutant, mab-5 is kept off by pry-1; and yet later pry-1 repression is relieved (directly or indirectly), allowing mab-5 expression to begin in wild type. Taken together these observations suggest that other spatially and temporally regulated factors interact combinatorially with bar-1 and pry-1 to modulate and specify their actions. Identifying these factors and understanding their interactions with bar-1 and pry-1 should provide interesting insights into the control of the Wnt pathway during development.

We thank Shin Takagi, David Eisenmann and Stuart Kim for sharing results and reagents before publication. Some strains were obtained from the Caenorhabditis Genetics Center, which is supported by the National Center for Research Resources of the National Institutes of Health. We thank Jon Alexander, Stacey Harmer and members of the Kenyon Laboratory for careful reading of the manuscript and thoughtful discussions. J. N. M. was supported by a HHMI predoctoral fellowship. J. W. was supported by a NSF predoctoral fellowship. G. D. J. was supported by an A.C.S. Post-doctoral grant. C. K. is the Herbert Boyer Professor of Biochemistry and Biophysics. This work was supported by NIH grant GM37053 to C. K.