The roles of two C. elegans HOX co-factor orthologs in cell migration and vulva development

In Caenorhabditis elegans, Hox genes specify patterns of cell migration, cell division, differentiation and morphogenesis along the anteroposterior body axis. In this study, we have characterized mutations in two genes: ceh-20, which encodes the C. elegans ortholog of the Hox co-factor Extradenticle (Exd/Pbx); and unc-62, which encodes the C. elegans ortholog of the Hox co-factor Homothorax(Hth/Meis/Prep). Like Hox mutations, mutations in these co-factors influence multiple cell-fate decisions. In this study, we have analyzed the effects of these mutations on two different processes, Q neuroblast migration and vulva development, both of which are known to be influenced by the C. elegans Hox gene lin-39.

The C. elegans Q neuroblasts, QL and QR, and their descendants migrate long distances along the anteroposterior body axis. QL and QR are born as bilaterally symmetric cells in the posterior body region of the animal(Fig. 1A)(Chalfie and Sulston, 1981; Sulston and Horvitz, 1977). The right Q cell (QR) and its descendants migrate towards the anterior,whereas the left Q cell (QL) and its descendants migrate towards the posterior. The stopping points of the Q descendants are not associated with any obvious landmarks, and the mechanisms by which their final positions are specified are not well understood.

Several genes are known to influence these stopping points, including lin-39, mig-13 and egl-20. lin-39 is a Hox gene, the C. elegans ortholog of the Drosophila homeobox genes Sex combs reduced/Deformed/proboscipedia (Clark et al., 1993; Wang et al.,1993). lin-39 is expressed in QR and its descendants and acts cell autonomously to promote the anterior migration of these cells(Clark et al., 1993; Wang et al., 1993). mig-13 encodes a novel transmembrane protein(Sym et al., 1999). This protein acts extrinsically of the Q descendants to promote their anterior migration in a dose-dependent manner, in which progressively higher protein levels, even if delivered throughout the body, promote progressively further anterior migration (Sym et al.,1999). In addition, increasing MIG-13 levels can cause cells that normally migrate posteriorly to stop migrating prematurely or reverse direction and migrate anteriorly. egl-20 encodes a Wnt protein(Maloof et al., 1999; Whangbo and Kenyon, 1999). Like mig-13, egl-20 is required to promote anterior migration of the QR descendants. Although egl-20 is expressed in the tail, ectopic or global egl-20 expression can restore full anterior migrations of QR descendants in animals with reduced egl-20 activity. Genetic and molecular experiments suggest that lin-39, mig-13 and egl-20act in partially independent pathways to regulate the anterior migrations of the QR descendants (Harris et al.,1996; Sym et al.,1999)

The egl-20 gene also influences the migrations of cells in the QL lineage, but in a different way. In QL, EGL-20 switches on expression of the Hox gene mab-5, the C. elegans ortholog of Drosophila Antennapedia, via a canonical Wnt signaling pathway. mab-5 acts cell-autonomously in the QL lineage, and is necessary and sufficient to promote the posterior migrations of QL descendants(Harris et al., 1996; Salser and Kenyon, 1992).

In addition to regulating neuroblast migration, the Hox gene lin-39 also regulates vulva development. The vulva is generated by a subset of the 12 ventral epidermal cells that are located in a row along the ventral margin of the animal (Fig. 1B) (for reviews, see Greenwald, 1997; Shemer and Podbilewicz, 2003; Wang and Sternberg, 2001). These cells, called P(1-12).p, are born during the first larval stage. Soon after their birth, all but the six central Pn.p cells, P(3-8).p, fuse with the epidermal syncytium, hyp7 (Sulston and Horvitz, 1977). P(3-8).p are prevented from undergoing cell fusion by LIN-39. CEH-20 also participates in cell fusion(Shemer and Podbilewicz,2002). Together with LIN-39, CEH-20 represses eff-1, a gene required for cell fusion (Mohler et al., 2002; Shemer and Podbilewicz, 2002). CEH-20 and LIN-39 also regulate the activity of egl-18 and elt-6, GATA factors that redundantly inhibit cell fusion (Koh et al.,2002). Furthermore, all of the Pn.p cells fused with hyp7 in most ceh-20(ay42) animals (Shemer and Podbilewicz, 2002).

During the L3 stage, three of the six vulval precursor cells (VPCs),P(5-7).p, are induced by the nearby gonadal anchor cell to generate the vulval cell lineages. The other VPCs, P(3,4, and 8).p, divide once, and their daughters then fuse with hyp7. Vulval cell-lineage specification is controlled by several interacting signaling pathways. The `synMuv' pathway plays an important role in determining whether a Pn.p cell generates vulval cell lineages or not. This pathway prevents vulval cell division in the absence of an inductive signal from the anchor cell (reviewed by Fay and Han, 2000). Animals defective in both class A and class B synMuv gene functions display a multivulval phenotype because most of the vulval precursor cells adopt vulval fates (Ferguson and Horvitz,1989). Many of the synMuv pathway components are members of the nucleosome remodeling and histone deacetylation (NuRD) complex or proteins that physically interact with histone deacetylases(Brehm et al., 1998; Lu and Horvitz, 1998; Solari and Ahringer, 2000). By antagonizing the Ras pathway, the NuRD complex helps to limit the number of cells adopting vulval fates.

Other signaling pathways involved in vulval lineage specification include:(1) the Ras, Wnt and Rb pathways, and the SEM-4 transcription factor, which upregulate lin-39 expression and allow the three Pn.p cells located closest to the anchor cell to overcome the silencing effects of the synMuv genes and generate vulval cell lineages(Chen and Han, 2001a; Eisenmann et al., 1998; Grant et al., 2000; Hoier et al., 2000; Maloof and Kenyon, 1998); (2)the Hox gene lin-39, which upregulates its own expression to permit vulval induction and to instruct cells to generate vulval cell types instead of different, posterior-specific, cell types(Clandinin et al., 1997; Clark et al., 1993; Maloof and Kenyon, 1998; Wang et al., 1993); and (3)the forkhead-family transcription factor LIN-31, which prevents a randomization of lineage patterns (Tan et al., 1998). Once the 22 vulval cells have been generated, they undergo a complex pattern of vulval morphogenesis which includes the migration of P5.p and P7.p descendants towards the descendants of P6.p during the final stages of postembryonic development.

In this study, we show that the C. elegans Hox co-factor orthologs ceh-20 and unc-62 are required for two processes that are also regulated by Hox gene lin-39: (1) anterior migration of the QR neuroblast and its descendants; and (2) vulva formation. Surprisingly, we find that ceh-20 and unc-62, but not lin-39, are required for MIG-13 to promote anterior migration. To our knowledge, ceh-20 and unc-62 are the only two genes that have been implicated to function with mig-13 in regulating anterior migration. We also find that ceh-20 and unc-62 are required for several steps in vulva formation, and mutations in these genes result in phenotypes that are starkly different from those previously shown for lin-39mutants. Thus, in both processes, these Hox co-factor orthologs may function,in part, independently of the Hox gene lin-39.

Standard methods for culturing and genetic analysis were used(Brenner, 1974; Sulston and Hodgkin, 1988). Strains were maintained at 20°C. Strains used in this study are:

N2 wild-type var. Bristol,

tax-4::gfp,

mig-13(mu225) X; tax-4::gfp,

ceh-20(mu290) III; tax-4::gfp,

unc-62(mu232) V; tax-4::gfp,

mab-5(e2088) III; tax-4::gfp,

lin-39(n1760) III; tax-4::gfp,

ceh-20(mu290) III,

ceh-20(mu290) III; muIs32[mec-7::gfp] II,

unc-62(mu232) V,

unc-62(mu232) V; muIs32[mec-7::gfp] II,

unc-62(ku234) V,

unc-62(s472) V unc-46(e177);yDp1(IV;V;f),

mig-13(mu225) X, lin-39(n1760) III,

ceh-20(mu290) lin-39(n1760) III,

unc-62(mu232) V; lin-39(n1760) III,

mab-5(e2088) III,

ceh-20(mu290) mab-5(e2088) III,

unc-62(mu232) V; mab-5(e2088) III,

mig-13(mu31) X,

mig-13(mu225) X,

egl-20(n585) IV,

mig-13(mu31) X; egl-20(n585) IV,

egl-20(mu320) IV,

ceh-20(mu290) III; egl-20(mu320) IV,

ceh-20(mu290) III; egl-20(n585) IV,

ceh-20(mu290) III; mig-13(mu31) X,

unc-62(mu232) V; egl-20(mu320) IV,

unc-62(mu232) V; egl-20(n585) IV,

unc-62(mu232) V; mig-13(mu31) X,

lin-39(n1760) III; egl-20(mu320) IV,

lin-39(mu26) III; mig-13(mu31) X,

muEx89 [Punc-119::mig-13GFP],

mig-13(mu225) X; muEx89 [Punc-119::mig-13GFP],

lin-39(n1760) III; muEx89 [Punc-119::mig-13GFP],

ceh-20(mu290) III; muEx89 [Punc-119::mig-13GFP],

unc-62(mu232) V; muEx89 [Punc-119::mig-13GFP],

ceh-20(mu290) III;muEx261 [ceh-20::gfp + odr-1::rfp],

rrf-3(pk1426); muIs35 [mec-7::gfp],

nDf16/qC1[dyp-19(e1259) glp-1(q339)] III,

unc-62(s472) unc-46(e177) V; yDp1(IV;V;f),

wIs52[scm::gfp::lac-Z; unc-119(+)](Koh and Rothman, 2001),

ceh-20(mu290) III; wIs52[scm::gfp::lac-Z;unc-119(+)],

unc-62(mu232) V; wIs52[scm::gfp::lac-Z; unc-119(+)],

jcIs1[jam-1::gfp; rol-6(d)](Mohler et al., 1998),

ceh-20(mu290) III; jcIs1[jam-1::gfp; rol-6(d)],

unc-62(mu232) V; jcIs1[jam-1::gfp; rol-6(d)],

ceh-20(mu290) III; ayIs9[egl-17::gfp],

mu232; ayIs9[egl-17::gfp],

muIs3[mab-5::lacZ, rol-6(d)],

ceh-20(mu290) III; muIs3[mab-5::lacZ, rol-6(d)],

unc-62(mu232); muIs3[mab-5::lacZ, rol-6(d)],

ceh-20(mu290) III; unc-62(mu232) V

The positions of the Q.pa daughters were determined directly using Nomarski optics at the end of L1. At this stage, the hypodermal V cells have divided once and the P nuclei have all descended into the cord. The positions of the Q.pa daughters were scored relative to the V cell daughters. The Q.ap cell positions were determined using an integrated tax-4::GFP fusion(Sym et al., 1999). Cell positions of Q descendants were analyzed only in animals without additional neurons in the body. In principle, this could introduce a selection bias;however, we addressed this by extensive lineage analysis and other methods(see Fig. S1 in the supplementary material). If the cell(s) of interest migrated near a Vn.a cell that had divided, the `Vn.a position' was defined as the distance bounded by a dorsoventral line through the center of the Vn.aa nucleolus and a dorsoventral line through the center of the Vn.ap nucleolus. Statistical analyses on the QR.pax distributions were performed using the Mann-Whitney test on AVM and SDQR separately.

The mu290 mutation was identified in a Nomarski screen for mutants with misplaced Q.ap cells in mutagenized tax-4::gfp animals. Mutants were generated with ethyl methanesulfonate (EMS). The QR.pax cell positioning defect was assayed in mapping experiments. mu290 was mapped using STS mapping (Williams et al.,1992) to a 0.7 map unit region tightly linked to the physical marker stP120 at the center of LG III. Chromosomal deficiencies were then used to refine the map position. The QR.pax positioning defect was generated when mu290 was placed in trans to the deficiency nDf16. Transformation rescue of the QR.pax cell positioning phenotype was obtained with pools of cosmids in this region. Ultimately, the rescuing activity of the QR.pax cell positioning defect, as well as the multivulval and egg-laying defects, were obtained with a single cosmid, F31E3. This cosmid also rescued the extra ventral protrusion phenotype of mutant hermaphrodites. A candidate open reading frame for mu290 was F31E3.1, which encodes a C. elegans ortholog of Exd/Pbx(Burglin and Ruvkun, 1992). A 3.6 kb PCR fragment that encompasses this gene was injected at 1 ng/μl with pPD93.97 (myo-3::gfp) at 100 ng/μl into mu290 animals. Four out of four lines generated with this PCR fragment gave 48-75% rescue of the QR.pax defect and over 85% rescue of the vulval defect. To identify the molecular lesion in ceh-20(mu290), we isolated and sequenced cDNA clones generated by RT-PCR (Frohman,1993) from wild-type and ceh-20(mu290) DNA. The oligo(dT) primer was used to obtain the 3′ end and the spliced leader sequence SL1 primer to obtain the 5′ cDNA end. Only a single isoform was isolated from several clones. A single G to A transition was identified in the cDNA product derived from ceh-20(mu290). This mutation changes a conserved arginine (R) in the 3rd helix of the homeodomain to a histidine. This R has been implicated as part of a nuclear localization signal(NLS) in the third helix of the homeodomain that is less conserved than the classical NLS RRKRR found at the N-terminal arm of the homeodomain. The R that is mutated in ceh-20(mu290) is the second R in this less conserved and weaker NLS, KRIRYKKNI (Abu-Shaar et al., 1999; Saleh et al.,2000a). ceh-20(mu290)/qC1[dyp-19(e1259) glp-1(q339) III; him-5(e1490) was used for crosses since it was very difficult to mate into ceh-20(mu290). Animals were outcrossed at least three times before phenotypes were studied.

The mu232 mutation was identified in a screen for mutants with misplaced Q.pax cells in mutagenized mec-7::gfp animals, as described by Ch'ng et al. (Ch'ng et al.,2003). In mu232/+ animals, the QR.pax cells were in wild-type positions, indicating that mu232 is recessive and thus likely to be a reduction or loss-of-function allele. We mapped mu232to chromosome VL near stP3 and unc-62. mu232 failed to complement s472, a null unc-62 allele, for the QR.pax cell-positioning defect, suggesting that mu232 is an allele of unc-62. mu232/s472 animals have a stronger QR.pax phenotype than mu232 animals, consistent with mu232 being a hypomorph. unc-62 exons were amplified from wild-type and mu232 using primers around each predicted exon. Sequencing these genomic fragments revealed that mu232 harbored a T→A transition that changed the predicted initial Met(ATG) codon of exon 1b to a Lys(AAG) codon. We also sequenced unc-62(ku234) (Van Auken et al., 2002) and discovered a G→A transition in the same codon. mu232 failed to complement ku234 (gift from Meera Sundaram) for the QR.pax positioning and Egl phenotypes. Because mu232 and ku234 exhibited the lowest percentage of embryonic and larval lethality among the known unc-62 homozygous alleles, we examined larval phenotypes using mu232 after noting that mu232 and ku234 exhibit very similar migration and vulval phenotypes. mu232 was outcrossed more than three times before phenotypes were studied.

After verifying that the extrachromosomal array,muEx89(P_unc-119::mig-13GFP) (Q. Ch'ng, PhD thesis, University of California, 2001), could create an anterior-promoting environment mimicking that created by hs-mig-13 [see Fig. 5 for rescued positions of QR.pax and BDU in mig-13(-)], we crossed muEx89 into ceh-20(mu290), unc-62(mu232), and lin-39(n1760). In animals that were green, we believe that mig-13::GFP was ectopically expressed because ALM (data not shown) and CAN (see Fig. S4 in the supplementary material) were redirected to more anterior positions.

A translational ceh-20::GFP fusion was constructed by inserting the GFP coding region of pPD119.45 into an AgeI site engineered just 5′ to the last stop codon of ceh-20. Transgenic arrays were generated using standard techniques (Mello et al., 1991). The resulting plasmid (pLY11) was injected at 50 ng/μl with odr-1::rfp (generously provided by Cori Bargmann's laboratory) at 100 ng/μl into ceh-20(mu290) hermaphrodites. Several transgenic lines rescuing the QR.pax defect and the muv defect were established, and all showed similar expression patterns. The line used in this study was muEx261.

LIN-39 antibody staining was performed as described by Maloof and Kenyon(Maloof and Kenyon, 1998). Staged populations of animals were stained with both the LIN-39 antibody and the monoclonal antibody MH27 (Francis and Waterston, 1991). Individual slides with animals of different genotypes were stained simultaneously under identical conditions prior to comparison, and the differences in staining intensity were repeatable. For mab-5 expression, staged populations of muIs3[mab-5::lacZ,rol-6(d)], ceh-20(mu290); muIs3[mab-5::lacZ, rol-6(d)],and unc-62(mu232); muIs3[mab-5::lacZ, rol-6(d)] animals were grown at 20°C. Larvae were fixed on slides and stained as described by Maloof et al. (Maloof et al.,1999).

Analysis of Q lineages and Vn.a lineages were performed from hatching to late L1/early L2. Of the 27 wild-type Q cells whose lineages were analyzed,Q.pp and Q.aa were still present by late L1/early L2 in only four (14.8%) and three animals (11.1%), respectively, and none of the Q cell descendants divided inappropriately. Vulval lineages were observed starting with the Pn.px cell stage until the L4 molt by Nomarski optics.

rrf-3(pk1426) is more sensitive than wild-type animals to RNAi(Simmer et al., 2002). rrf-3(pk1426); muIs35[mec-7::gfp] animals in L4 were grown on bacteria containing vector only or on bacteria expressing ceh-20dsRNA, unc-62 dsRNA or lin-39 dsRNA. Their progeny were analyzed for QR.pax positions in late L1, vulval induction in L4, and number of ventral protrusions as young adults. Animals fed with the control plasmid did not display the QR.pax positioning defect, vulval induction defects or additional ventral protrusions.

Pn.p cell fusion status was scored by first crossing jam-1::gfp(Mohler et al., 1998) into ceh-20(mu290) and unc-62(mu232). The early L1 fusion event was scored in L2 when the adherens junctions of unfused cells were visible as a circle. The later fusion event was scored in L3.

Vulval induction was scored at the L4 stage under Nomarski optics(Sternberg and Horvitz, 1986). Nuclei that were not hypodermal, neuronal or muscle were counted. The number of vulval nuclei was used to extrapolate the number of induced Pn.p cells. A Pn.p cell in which both daughter cells divide one more time, and both granddaughters divide to generate seven or eight great granddaughters and no hypodermal tissue was scored as 1.0 cell induction. A Pn.p, which generates more than two but fewer than seven or eight cells was scored as 0.5 cell induction. In wild-type, P5.p, P6.p and P7.p each undergo a 1.0 cell induction whereas the other Pn.p cells are not induced, resulting in a total cell induction of 3.0.

In the wild-type, the Q.ap cells migrate either towards the head (QR lineage) or tail (QL lineage). We isolated the mutation mu290 (see Materials and methods) in a screen for mutants with improperly positioned Q.ap cells using a tax-4::GFP fusion, which is expressed in the Q.ap cells. Subsequent genetic and molecular analyses indicated that mu290is an allele of the Hox co-factor ceh-20 (see Materials and methods). The molecular lesion in ceh-20(mu290) is a G→A transition that alters a conserved arginine to histidine in a nuclear localization signal consensus site in the third helix of the homeodomain(Abu-Shaar et al., 1999; Passner et al., 1999; Piper et al., 1999; Saleh et al., 2000a).

In ∼90% of ceh-20(mu290) animals, QR.ap was located between the birthplace of QR and its wild-type location in the head(Fig. 2). In addition, the QR.pax cells were in the central body region rather than in the anterior(Fig. 3), suggesting that cells in the QR lineage stopped migrating prematurely. In ∼33% of ceh-20(mu290) animals, QL.ap was not located in its normal position in the tail (Fig. 2). In fact,QL.ap could be found anterior to the birthplace of QL in ∼10% of ceh-20 mutants, suggesting that this cell had migrated anteriorly instead of posteriorly.

Unexpectedly, we also found additional neuronal cells in the mid-body region (∼25% on both left and right sides). By observing the sequence of Q cell divisions using Nomarski optics, we found that these cells resulted from inappropriate cell survival and division in the Q cell lineages. In wild-type animals, Q.pp and Q.aa die by apoptosis, and most of these cells disappear by the end of L1. Of the 27 wild-type Q cells lineaged to the end of L1, only four Q.pp and three Q.aa cells remained visible, and none of the Q cell descendants divided inappropriately. However, in nine of the 20 ceh-20(mu290) animals we analyzed, Q.pp remained visible. Furthermore, Q.pp divided in two ceh-20 mutants and Q.ap divided in one animal.

To determine whether these phenotypes result from reduced ceh-20activity, we first analyzed ceh-20(mu290)/+ heterozygotes. In ceh-20(mu290)/+ animals, the QR.pax cells were mostly in wild-type positions (see Fig. S1 in the supplementary material) and there were no extra body neurons visible. We also found that the QR.pax phenotype was not enhanced in ceh-20(mu290)/deficiency heterozygotes. Finally, as RNA interference reduces gene activity, we fed ceh-20 dsRNA to an RNAi-sensitive strain, rrf-3(Simmer et al., 2002). The progeny of these animals showed similar defects to those in ceh-20(mu290) animals, including QR.pax cells in the mid-body region and extra body neurons (see Fig. S2 in the supplementary material). Taken together, these analyses indicate that ceh-20(mu290) is likely to reduce gene function. Therefore, we conclude that in wild-type animals, ceh-20 is required for the proper generation and migration of cells in the Q lineage. ceh-20 may also have an earlier function in Q development because, in stronger alleles of ceh-20 (obtained from M. Stern), the Q cell failed to complete its lineage (3/3 animals; data not shown).

We also screened for Q migration mutants using mec-7::gfp, which is expressed in QR.paa and QL.paa(Ch'ng et al., 2003). In this screen, we isolated mu232, a mutant with QR.pax cells located posterior to their normal stopping points(Fig. 3). Subsequent genetic and molecular analyses (Materials and methods) indicated that mu232is a reduction-of-function allele of the C. elegans homothorax(hth) ortholog unc-62, a member of the Meis family of Hox co-factors.

In Drosophila, hth and exd mutants have similar phenotypes. Likewise, we found that unc-62 mutants share the cell positioning defects exhibited by ceh-20 mutants. In unc-62mutants, 70% of QR.ap cells and 100% of QR.pax cells stopped posterior to their wild-type positions, in distributions similar to those in ceh-20 mutants (Figs 2and 3). Likewise, in ∼70%of unc-62 mutants, QL.ap was located anterior to the birthplace of QL(Fig. 2). Finally, QL.pax cells were in wild-type positions in unc-62 mutants, as in ceh-20mutants (data not shown). The one Q phenotype that ceh-20(mu290)animals exhibited that unc-62(mu232) animals did not was the Q lineage defect.

Mutations in ceh-20 and unc-62 also affected the anteriorwards migration of BDU, a neuron whose cell body migrates a short distance anteriorly during embryogenesis. In ∼70% of animals with a mutation in ceh-20(mu290) or unc-62(mu232), this migration was incomplete (see below and Fig. 5).

In theory, the shortened anterior migration of cells in mutants could simply be due to a requirement for CEH-20 and UNC-62 activity in the basic mechanisms of cell motility. However, we found that a mutation in ceh-20 or unc-62 could also change the direction of cell migration. First, direct observation of cell movements by Nomarski optics revealed that QR.a or QR.p migrated towards the posterior rather than the anterior direction in two out of nine ceh-20(mu290) animals. In addition, these ceh-20 or unc-62 mutations could redirect the anterior-bound cells of the QR lineage towards the posterior when combined with a mutation in a Wnt homolog egl-20 (see below and Fig. 4). Although we cannot exclude the possibility that ceh-20 and unc-62 are required for some aspect of cell motility, these findings indicate that CEH-20 and UNC-62 activities are required for the guidance or positioning of cells along the AP axis.

If CEH-20 and UNC-62 function as Hox co-factors like their Drosophila and vertebrate homologs, then ceh-20 and unc-62 mutants should share phenotypes with Hox mutants. Close examination of the final positions of the QR descendants in lin-39(n1760null), ceh-20(mu290) and unc-62(mu232) mutants revealed that the ceh-20 and unc-62 mutants had a more severe cell positioning defect than did lin-39 null animals(Fig. 3). In addition, the QR.pax cell positioning defect was more severe in the ceh-20(mu290)lin-39(n1760) double mutant than in either single mutant and, similarly,the lin-39(n1760); unc-62(mu232) double mutant phenotype was more severe than that of either single mutant(Fig. 3). Thus, CEH-20 and UNC-62 could function as LIN-39 co-factors, but they must also have a function that is independent of LIN-39 in promoting anterior migration.

QR descendants may migrate further anteriorly in lin-39 mutants than they do in ceh-20 or unc-62 mutants because cells in lin-39 mutants, but not in ceh-20 or unc-62mutants, retain the ability to respond to MIG-13. Consistent with this hypothesis, the distribution of QR.pax cells in the mig-13(mu31);lin-39(n1760) double mutant was shifted further posteriorly than in either single mutant (data not shown) (Sym et al., 1999). This distribution nearly overlapped with the distribution seen in ceh-20 or unc-62 single mutants,raising the possibility that ceh-20 and unc-62 may have a function in the mig-13 pathway. A mig-13(mu31null) mutation can cause cells in the QR lineage to reverse direction and migrate toward the posterior in an egl-20(-) background(Fig. 4A)(Harris et al., 1996; Sym et al., 1999). Similarly,in ceh-20 or unc-62 double mutants with a null or haploinsufficient allele of egl-20, QR.pax cells could be posterior to the birthplace of QR (Fig. 4B,C). However, combining null mutations in lin-39 and egl-20 did not cause cells in the QR lineage to migrate posteriorly(Fig. 4D). In addition,removing mig-13 activity from ceh-20(mu290) or unc-62(mu232) did not result in the posterior migration of QR descendants (Fig. 4B,C). Finally, in ceh-20(mu290); unc-62(mu232) double mutants, there were QR.pax cells posterior to the birthplace of QR (see Fig. S3 in the supplementary material). Collectively, these results suggest that ceh-20 and unc-62, but not lin-39, may act in the same pathway as mig-13 to promote anterior migration.

To test directly whether ceh-20 and unc-62 were required for cells to respond to the anterior-promoting activity of mig-13, we asked whether expression of mig-13::gfp in our ceh-20(mu290)and unc-62(mu232) mutants using the pan-neural unc-119promoter (see Materials and methods) could rescue the shortened anterior migrations of cells in the Q lineage. In mig-13(null) and lin-39(null) animals, pan-neuronal mig-13 expression could rescue the QR.pax positioning defect; however, in ceh-20(mu290) or unc-62(mu232) animals it could not(Fig. 5). In addition,P_unc-119::mig-13 failed to fully rescue a different anterior cell migration, that of BDU, in ceh-20 and unc-62 mutants (Fig. 5). In these mutant backgrounds, pan-neuronal expression of mig-13 could not fully restore anterior migration of BDU or cells in the QR lineage but could shorten the posterior migration of another cell, CAN(see Fig. S4 in the supplementary material). Thus, we believe that mig-13 was appropriately misexpressed in these mutants. Together,these results suggest that ceh-20 and unc-62, but not lin-39, are required for QR descendants and BDU to respond to the anterior-promoting activity of mig-13.

Another possible explanation for the truncated anterior migration of QR descendants in ceh-20 and unc-62 mutants was that cells in the QR lineage inappropriately express the Hox gene mab-5. If so,then removing mab-5 activity in our ceh-20 or unc-62 mutants should allow QR descendants to migrate further anteriorly. However, the cell distributions of QR.p descendants in the mab-5(-) double mutant with either ceh-20(mu290) or unc-62(mu232) were nearly indistinguishable from those of ceh-20(mu290) or unc-62(mu232) single mutants, respectively,indicating that ectopic expression of mab-5 was unlikely to underlie the migration shortening in these mutants(Fig. 3). In concordance with this finding, mab-5::lacZ was not ectopically expressed in QR or its descendants in these mutants (data not shown).

If CEH-20 and UNC-62 do not repress mab-5 expression in QR descendants, then do they regulate migration of these cells by controlling lin-39 expression? By staining wild-type animals with an antibody against LIN-39, we show that QR could express lin-39 in its nucleus before it delaminates from the epithelium, but expression at this early time is usually not high (see Fig. S5A in the supplementary material, Table 1). After delamination but before division, more QR cells expressed lin-39 strongly (see Fig. S5B in the supplementary material). After QR divides, lin-39expression persisted and continued throughout the lineage. In unc-62(mu232) animals, lin-39 expression in QR and its daughters resembled that of wild-type(Table 1). In ceh-20(mu290) animals, however, lin-39 expression was often stronger than that of wild-type in QR before delamination, after delamination and after division (see Fig. S5D-F in the supplementary material, Table 1). Neighboring V and P cells in ceh-20(mu290) also had increased LIN-39 staining. These differences were validated in many repeated staining comparing wild-type and ceh-20(mu290) animals under identical staining conditions. Because lin-39 overexpression does not inhibit anterior cell migration in the QR lineage (Hunter and Kenyon,1995), the altered lin-39 expression we observed in ceh-20 mutants was unlikely to contribute to the shortened migrations of these cells.

On the left side, we also observed stronger lin-39 expression in QL and/or its daughters in ceh-20(mu290) and unc-62(mu232)mutants (see Fig. S5G-L in the supplementary material; Table 1). In addition, we found that mab-5 expression was reduced. A mab-5::lacZfusion was expressed in the QL daughters in 72% (n=98) of wild-type animals but only in 23% (n=39) of ceh-20(mu290) animals and 30% (n=59) of unc-62(mu232) animals. Although this altered expression might have contributed to the anterior migration of the QL.ap cell,it did not affect the position of the QL.pax cells: not only were these cells positioned normally in ceh-20(mu290) and unc-62(mu232)animals, but addition of either mutation to mab-5(-) or egl-20(-) shortened the anterior migration of QR.p descendants (see Fig. S6 in the supplementary material).

In addition to having Q lineage defects, all ceh-20(mu290) and unc-62(mu232) hermaphrodites were egg-laying defective (Egl). Furthermore, all ceh-20(mu290) hermaphrodites formed `bags of worms'as their progeny hatched internally. Instead of having a single, centrally located vulval protrusion, as in wild-type, 94% (n=100) of ceh-20(mu290) mutants have multiple vulval protrusions (Muv phenotype) on the ventral surface. We tested whether this phenotype was a loss-of-function phenotype, using RNA interference and found that over 20%(n=50) of the progeny of animals fed bacteria expressing ceh-20 dsRNA had more than one ventral protrusion. This finding,together with our finding that mu290/+ animals have no ventral protrusions (100%, n=132), suggest that mu290 reduces the level of an activity of ceh-20 required for normal vulva formation.

Extra ventral protrusions can be caused by ectopic induction of Pn.p cells that do not normally contribute to vulva formation, or by defective morphogenesis (Chen and Han,2001b; Sternberg and Han,1998). Examination of vulva development using Nomarski optics revealed that in over 95% of ceh-20(mu290) hermaphrodites, Pn.p cells that do not normally undergo vulva development, that is, P3.p, P4.p and/or P8.p, are often ectopically induced to form a vulval invagination(Fig. 6B; Table 2). In addition, anterior and posterior Pn.p cells that normally fuse with the hypodermal syncytium,P(1,2,9-11).p, could also be induced so that their descendants formed a small invagination (Fig. 6C). In the six ceh-20(mu290) hermaphrodites lineaged, P1.p divided in two animals, P1.px divided in one animal, P2.p divided in four animals, P2.px divided in one animal, P3.px divided in three animals and P8.px divided in one animal. In one animal, not only were the great-granddaughters of P4.p inappropriately generated, but one even divided despite the fact that even the great-granddaughters of P(5-7).p do not normally divide. In contrast to these ectopic inductions, vulval precursor cells that are normally induced,P(5-7).p, were sometimes not induced in ceh-20(mu290) hermaphrodites(Table 2). When they were induced, P(5 or 7).p descendants could form invaginations separate from the one formed by the P6.p descendants (Fig. 6B), suggesting that P5.p or P7.p descendants sometimes failed to migrate toward P6.p descendants during morphogenesis (48%, n=50). Curiously, P7.p descendants formed a separate invagination or failed to fully migrate towards P6.p descendants in ∼40% of unc-62(mu232)(Fig. 6F) and ∼50% of unc-62(ku234) hermaphrodites (n=25 per strain). This defect probably contributes not only to their Egl phenotype but also to the 8% of animals of each strain which had an extra ventral protrusion (n=100 per strain). In ceh-20(RNAi) animals, we also observed ectopic induction of P3.p and P4.p (Fig. 6D, Table 2) as well as separate invaginations formed by P5.p or P7.p (10%, n=50). Thus, both ectopic induction and defective morphogenesis contribute to the multiple ventral protrusions in hermaphrodites with reduced CEH-20 activity,and UNC-62 activity is also required for proper morphogenesis.

ceh-20(mu290) mutants also displayed vulval lineage defects. In wild-type animals, P6.p adopts the primary fate, the fate in which all granddaughters divide transversely to generate eight descendants (reviewed by Greenwald, 1997). P5.p and P7.p normally adopt the secondary fate to generate seven descendants each. Their granddaughters divide longitudinally (L), transversely (T) or not at all(N), depending on their positions along the anteroposterior axis (from anterior to posterior, LLTN for P5.p and NTLL for P7.p). In ceh-20(mu290) animals, the induced Pn.p cells rarely generated the eight or seven descendants characteristic of primary or secondary fates,respectively. Out of 23 animals, there was only one animal in which P6.p had divided to generate eight cells. All of the other animals had fewer than eight P6.p descendants, and all of the animals had fewer than the expected seven descendants of P5.p and P7.p in the L4 stage. We confirmed that vulval cell divisions are prematurely truncated in ceh-20(mu290) animals by lineage analysis. In two of the six ceh-20(mu290) hermaphrodites lineaged, at least one of the P(5-7).p daughters failed to divide. Furthermore, at least one of the P(5-7).p granddaughters that normally divide failed to divide in two animals. In addition to the premature truncation of vulval lineages, cells could also divide along an unusual axis in ceh-20(mu290) animals. In one of the six ceh-20(mu290)hermaphrodites lineaged, all six P.6p granddaughters divided longitudinally instead of transversely. In this same animal, P5.ppa divided obliquely instead of transversely, P7.ppa divided obliquely instead of longitudinally, and both P5.ppp and P7.paa divided obliquely instead of remaining undivided. Taken together, these observations suggest that ceh-20 is also required for proper execution of VPC fates.

Because the extreme anterior and posterior Pn.p cells, (P1,2,9-11).p, often divided in ceh-20(mu290) mutants, we hypothesized that these cells failed to fuse to the hypodermal syncytium in mutant animals. To characterize this defect better, we used the jam-1::GFP marker, which labels cell adhesion junctions of unfused, but not fused, Pn.p cells(Mohler et al., 1998). In wild-type hermaphrodites, only P(3-8).p remain unfused after the early fusion event during L1. P3.p fuses by L3 in ∼50% of wild-type animals. In ceh-20(mu290) hermaphrodites, as in wild-type, the six central Pn.p cells remained unfused, but P1.p, P2.p and P(9-11).p sometimes also failed to fuse during the early fusion event (Table 3A). Some anterior and posterior Pn.p cells, as well as P3.p,still remained unfused in the L3 stage(Table 3B), suggesting that CEH-20 activity is required to generate the proper number of VPCs. Like ceh-20(mu290) animals, unc-62(mu232) hermaphrodites also exhibited the Pn.p cell fusion defect, although with lower penetrance(Table 3).

The Hox gene lin-39 is required during the L1/L2 stage for the P(3-8).p cells to remain unfused, and, later, to generate vulval cell divisions (Clandinin et al.,1997; Clark et al.,1993; Shemer and Podbilewicz,2002; Wang et al.,1993). In lin-39(null) animals, P(3-8).p cells all fuse,causing the animals to be vulvaless. To ask whether the ceh-20 Muv phenotype might result from altered lin-39 activity, we examined vulva development in ceh-20(mu290) lin-39(n1760null) double mutants. Surprisingly, though the number of protrusions was reduced, we found that 44%(n=100) of these animals had one or more ectopic ventral protrusions. It is probable that these protrusions arose, in part, from ectopic induction of P.3p and P4.p, as well as P2.p in these double mutants(Table 2). When P5.p or P7.p were induced, their descendants sometimes failed to migrate towards the descendants of P6.p, resulting in a separate invagination(Fig. 6E). Thus, the Muv phenotype of ceh-20 mutants is at least partially independent of LIN-39.

A row of six lateral hypodermal cells (V1-V6) runs along each side of newly hatched wild-type animals. During the first larval stage, each V cell divides asymmetrically along the AP axis, generating an anterior daughter (Vn.a) that fuses with hyp7 and a posterior daughter (Vn.p) that adopts the seam cell fate and continues to divide (Sulston and Horvitz, 1977).

We found that the nuclei of the anterior daughters of some V cells were sometimes replaced by two small nuclei in ceh-20(mu290) animals, and more often in unc-62(mu232) animals (see Fig. S7 in the supplementary material, Table 4). We determined that the two small nuclei were indeed daughters of Vn.a cells by lineage analysis (data not shown). These cells did not express a seam-specific GFP marker (Koh and Rothman,2001) and thus were not transformed into Vn.p-like seam cells(data not shown). Instead, these cells appeared to fuse with hyp7: first, the adherens-junction marker jam-1::GFP failed to label these cells at the L2 stage (data not shown); and second, we determined the lineage of a dividing Vn.a cell in an unc-62(mu232) animal harboring jam-1::GFP and found that there was only a short period of time when Vn.ax cells were outlined in green (see Fig. S8 in the supplementary material). Because the cell outlines of Vn.ax can be observed, albeit transiently, Vn.a division occurs before fusion. Whether ceh-20 and unc-62 actively inhibit cell division or simply promote fusion of Vn.a cells so that cell division cannot occur is currently unknown.

In addition to these defects, we found that in unc-62(mu232)animals, V5.a was often completely missing and replaced by neuronal cells (see Fig. S9 in the supplementary material). Lineage analysis of V5.a on the right(n=2) and left (n=1) sides in unc-62(mu232) animals showed that V5.a could generate a Q-like lineage pattern, producing three migratory cells. Using mec-7::gfp which is expressed in six differentiated cells including Q.paa, we confirmed that in unc-62(mu232) animals, V5.a descendants can express a fate marker normally expressed in Q descendants (data not shown).

We constructed a ceh-20::gfp translational fusion gene(pLY11) to investigate where and when ceh-20 is expressed. Transgenic ceh-20(mu290) animals bearing pLY11 were rescued for the QR.pax positioning phenotype, the Muv/Egl, and the Vn.a division phenotype,suggesting that ceh-20::gfp was expressed in cells that require its function for these processes. In all cells expressing ceh-20, the expression was stronger in the nucleus than in the cytoplasm. We detected ceh-20::gfp expression in QR and QL(Fig. 7A,B) and their descendants throughout their migrations to the end of L1. V cells and their daughters also expressed ceh-20::gfp(Fig. 7A-C). Expression persisted in the descendants of the V cells through the adult stage (data not shown). At hatching, all P cells expressed ceh-20::gfp(Fig. 7A,B). Before the anterior and posterior Pn.p cells fuse, they also expressed ceh-20. In L3 hermaphrodites, expression was maintained in P(3-8).p(Fig. 7D). We identified ceh-20::gfp expression in several other cell types. These included M, BDU, ALM, HSN, body wall muscle cells, I4, all L1 ventral cord neurons and a few unidentified neurons in the head behind the posterior bulb of the pharynx. ceh-20::gfp expression and nuclear localization did not change in the unc-62(mu232) background (data not shown).

The C. elegans orthologs of Hox co-factors Exd/Pbx(ceh-20) and Hth/Meis/Prep (unc-62), have been shown previously to play an important role in the post-embryonic development of the mesoderm and the embryo, respectively (Liu and Fire, 2000; Van Auken et al., 2002). In this study, we provide evidence that ceh-20 and unc-62 are also required for the post-embryonic development of the ectoderm, including the Q, V and P cell lineages. Our analyses of ceh-20 and unc-62 mutants indicate that they play a crucial role in ensuring that these cells and their descendants undergo their invariant patterns of cell division, migration, fusion and morphogenesis. Some, but not all, of the ceh-20 and unc-62mutant ectodermal phenotypes resemble those of Hox gene lin-39mutants. Thus, CEH-20 and UNC-62 could conceivably function as lin-39co-factors in some of the cell-fate decisions they influence, but probably not in others.

Our interest in these ceh-20 and unc-62 mutants was initially stimulated by the fact that their cell migration phenotypes are quite similar to those produced by loss-of-function mutations of the transmembrane protein MIG-13 (Sym et al.,1999). Mutations in all of these genes prevented the anteriorwards migration of BDU and severely shortened the anteriorwards migration of the QR descendants. Moreover, we found that our ceh-20 and unc-62mutations, in combination with other Q migration mutants, produced phenotypes similar to those we had seen previously with mig-13. For example,combining a unc-62 or ceh-20 mutation with a lin-39null allele abolished nearly all anteriorwards migration of the Q descendants. Combining an unc-62 or ceh-20 mutation with an egl-20/Wnt null mutation could cause cells in the QR lineage to reverse direction and migrate posteriorly rather than anteriorly. Finally, we found that global overexpression of MIG-13 was unable to stimulate anteriorwards migration in ceh-20 or unc-62 mutant backgrounds. Together, these findings suggest that, in Q cell migration, ceh-20 and unc-62 may function in the same regulatory step as MIG-13. The possibility that ceh-20 and unc-62 are simply required for mig-13 expression is ruled out by the finding that these ceh-20 and unc-62 mutations do not affect the pattern of MIG-13::GFP (data not shown). Because CEH-20 and UNC-62 are known to regulate transcription, one possibility is that they regulate the expression of another, as yet unidentified, molecule that functions in the same process as MIG-13 to promote anteriorwards migration. Alternatively, CEH-20 and UNC-62 may act in a pathway that is independent of and parallel to other pathways involving LIN-39, MIG-13, and EGL-20. Our data would also be consistent with such a model in which CEH-20 and UNC-62 act in a pathway that is dominant over these other pathways, increasing the difficulty with which mig-13overexpression rescues the shortened anterior migration in ceh-20 or unc-62 mutant animals. Determining which cells require ceh-20 and unc-62 for proper Q descendant migration and identifying ceh-20 and unc-62 effector(s) may facilitate the understanding of mechanisms that regulate cell migration along the anteroposterior axis in C. elegans.

Our genetic studies of ceh-20 and unc-62 demonstrate that they play multiple roles in regulating vulva development. First, in ceh-20 and unc-62 single mutants, cells anterior and posterior to the VPCs sometimes fail to fuse with the syncytial hypodermis. Although ectopic lin-39 expression could explain this phenotype, it does not, because, as in wild type, lin-39 is expressed in only the six central Pn.p cells in ceh-20 and unc-62 mutants even when some anterior and posterior Pn.p cells are unfused (data not shown). Thus, UNC-62 and CEH-20 appear to regulate cell fusion independently of LIN-39. Second, ceh-20 is required to prevent cells outside of the vulval equivalence group from generating vulval cell lineages and also for ensuring that the central VPCs do generate vulval cell lineages. Consistent with this, ∼26% of P6.p cells in L3 ceh-20(mu290) hermaphrodites(versus 0% in wild-type) fail to express egl-17::gfp, an early marker for the primary fate (data not shown). Although we did not see induction and/or lineage defects in unc-62(mu232) mutants, it is probable that unc-62 is also involved in these processes since non-vulval VPC descendants and vulval-like descendants of Pn.p cells outside of the vulval equivalence group have been observed when a weak unc-62 allele has been placed over a unc-62(null) or a deficiency encompassing the unc-62 gene (Van Auken et al.,2002). In addition, cell divisions of P(5-7).p were often prematurely terminated or misoriented in ceh-20(mu290)hermaphrodites, suggesting that CEH-20 is required for the proper execution of the primary and secondary vulval fates. Finally, in ceh-20 and unc-62 mutants, the descendants of P5.p and P7.p sometimes failed to migrate towards the descendants of P6.p(Fig. 6)(Van Auken et al., 2002),indicating that proper morphogenesis of the VPC descendants requires CEH-20 and UNC-62 activity. Thus, both ceh-20 and unc-62participate in multiple steps in vulva formation.

The vulval phenotypes exhibited by ceh-20(mu290) animals are reminiscent of those in NuRD complex mutants such as lin-40 MTA(metastasis associated factor) and hda-1 HDAC (histone deacetylase). Like ceh-20(mu290) animals, lin-40 or hda-1 mutants also have disrupted axes of transverse divisions, and their P5.p and P7.p descendants may form a separate invagination(Chen and Han, 2001b; Dufourcq et al., 2002). Animals doubly mutant for lin-40 and a class B synMuv gene also exhibit underinduction of P(5-7).p and overinduction of P(3,4,8).p(Chen and Han, 2001b). In addition, mutating lin-40 increases the frequency of unfused P3.p cells, and, in an egl-27 MTA mutant background, a lin-40mutation causes posterior P(9-11).p cells to also remain unfused(Chen and Han, 2001a).

We note that whereas the VPCs remained unfused in our ceh-20(mu290) animals, these cells have been reported to fuse with hyp7 in ceh-20(ay42) mutants(Shemer and Podbilewicz,2002). One possible explanation for this discrepancy is that,although both mutations alter residues in the conserved homeodomain, they may disrupt different functions of ceh-20: mu290 changes a residue that is predicted to form hydrogen bonds with the guanine of the TGAT core in the major groove of the DNA, whereas ay42 changes a residue that is predicted to contact the Hox protein (E. Chen, PhD thesis, Yale University, 1996) (Passner et al.,1999; Piper et al.,1999).

How might ceh-20 and unc-62 interface with known regulators of vulva development? It has been shown that CEH-20, together with LIN-39, participates in Pn.p fusion by regulating the expression of fusion effector eff-1 and GATA factors egl-18 and elt-6(Koh et al., 2002; Shemer and Podbilewicz, 2002). Our findings suggest that CEH-20 and UNC-62 may have additional functions during vulva formation. The C. elegans NuRD complex components repress transcription during vulval fate specification by regulating chromatin structure (Ahringer, 2000). A recent study demonstrated that in mammalian cells, the ceh-20 homolog PBX1 can interact with histone deacetylase HDAC(Saleh et al., 2000b). By analogy to their mammalian counterparts, CEH-20 might interact with HDA-1(HDAC). Thus, in general terms, our findings raise the possibility that CEH-20 and UNC-62 are required either for the expression of specific NuRD complex proteins, or for the proper function of the NuRD complex during vulva development (see L. Yang, PhD thesis, University of California, San Francisco,2003). Future work analyzing the relationships between ceh-20,unc-62, and NuRD complex components may provide a more comprehensive view on how vulva development is controlled in C. elegans.

Mutations in ceh-20 and unc-62 affect processes that Hox gene lin-39 also affects, such as cell migration and vulva development. Because ceh-20 and unc-62 have been shown to act as Hox co-factors in the C. elegans mesoderm and embryo(Liu and Fire, 2000; Van Auken et al., 2002), we considered the possibility that the ceh-20(mu290) and unc-62(mu232) phenotypes could be caused by altered target-site specificity of LIN-39. However, based on the many phenotypic differences we observe between ceh-20(mu290) or unc-62(mu232) and lin-39(n1760) animals, it is probable that ceh-20 and unc-62 have functions that are at least partly independent of lin-39 in the ectoderm.

Some of the Ceh-20 and Unc-62 phenotypes we observed were initially suggestive of a LIN-39-co-factor role for CEH-20 and UNC-62. For example, in ceh-20(mu290), unc-62(mu232) and lin-39(null) mutants,anterior migrations of QR descendants are shortened. Consistent with this, ceh-20 is expressed in the nucleus of QR and its descendants where lin-39 acts autonomously (Clark et al., 1993; Wang et al.,1993). In vulva development, our observation that P(5-7).p cell divisions sometimes terminated prematurely in ceh-20(mu290) animals is reminiscent of that seen in lin-39 mutants under conditions in which VPC fusion to the hypodermis was prevented and in animals with reduced activity of a LIN-39 target, egl-18(Koh et al., 2002; Shemer and Podbilewicz,2002).

Other phenotypes, however, suggest that ceh-20 and unc-62may function somewhat independently in processes that lin-39 also regulates. For example, QR descendants migrate further anteriorly in lin-39(null) mutants than in ceh-20(mu290) or unc-62(mu232) animals. In addition, unlike ceh-20(mu290) or unc-62(mu232) animals, lin-39(null) mutants do not share the shortened BDU migration phenotype with mig-13 mutants or exhibit QR descendants migrating in the wrong direction in the absence of egl-20(Harris et al., 1996; Sym et al., 1999). Furthermore, ectopic mig-13 expression can rescue the anterior migration defect of QR descendants in lin-39(null) animals but not in ceh-20(mu290) or unc-62(mu232) animals. For this reason, the hypothesis that ceh-20 and unc-62 act in a mig-13-dependent process rather than a strictly lin-39-dependent process seems reasonable. Although it is theoretically possible that another C. elegans homolog of exd, ceh-40, is also involved in this process, we have observed no migration phenotypes upon RNA interference of ceh-40 in the rrf-3 background, and no enhancement of migration phenotypes in the ceh-20; rrf-3 background (data not shown).

In vulva development, we observed that many ceh-20(mu290) and ceh-20(RNAi) hermaphrodites, and some unc-62(mu232) animals,are multivulval, whereas lin-39(null) mutants are vulvaless. Removing lin-39 activity causes all Pn.p cells, including P(3-8).p, to fuse in late L1 (Clark et al., 1993; Wang et al., 1993). By contrast, in ceh-20(mu290) and unc-62(mu232) mutants,P(3-8).p remain unfused as in wild type, and Pn.p cells that normally fuse sometimes fail to fuse. In lin-39(null) animals rescued for the early fusion defect, P(5-7).p often fail to complete their lineages, but induction of P(3,4,8).p has not been observed(Maloof and Kenyon, 1998). In ceh-20(mu290) and ceh-20(RNAi) animals, however, P(3,4,8).p could also be induced, and, for ceh-20(mu290), this ectopic induction could occur even in a lin-39(null) background(Table 2). In addition, ceh-20 and unc-62 have a role in vulval morphogenesis(Fig. 6), whereas lin-39 has been shown to have no role in promoting cell migration during vulva formation (Shemer and Podbilewicz, 2002).

Although it has previously been shown that CEH-20/LIN-39 heterodimers bind enhancers of GATA factors that regulate P(3-8).p fusion(Koh et al., 2002) and that CEH-20 and LIN-39 together repress the fusion effector eff-1(Shemer and Podbilewicz,2002), we hypothesize that ceh-20 and unc-62have some functions that are independent of lin-39 for vulva formation based on the stark differences in their phenotypes. We speculate that there are partners for CEH-20 and UNC-62 besides LIN-39 in the Pn.p lineages. As vulval phenotypes have not been shown for other C. elegans Hox gene loss-of-function mutants, it is possible that CEH-20 pairs with a non-Hox homeodomain protein. There is precedence for this,because a Drosophila ortholog of CEH-20, EXD acts as a co-factor for the non-Hox homeodomain protein Engrailed(Peifer and Wieschaus, 1990; van Dijk and Murre, 1994).

Do ceh-20 and unc-62 function in similar ways as their homologs? First, previous studies in Drosophila and vertebrates have demonstrated that the phenotypes produced by mutations in the homologs of ceh-20 and unc-62 resemble one another. For example, in Drosophila, exd and hth mutants both exhibit posterior transformations of embryonic segments(Peifer and Wieschaus, 1990; Rieckhof et al., 1997). In zebrafish, mutations in either lazarus or meis (the exd and hth homologs, respectively) cause defects in hindbrain segmentation (Choe et al.,2002; Popperl et al.,2000; Waskiewicz et al.,2001; Waskiewicz et al.,2002). Here, we show that this property of these genes is conserved: mutations in C. elegans ceh-20 and unc-62 both disrupt, in similar ways, the migration of Q descendants, Pn.p fusion, vulval morphogenesis and the patterning of V cell descendants.

Second, as mentioned above, in many situations, mutations in Hox co-factor genes produce phenotypes that mimic those of Hox mutants. In fact, Drosophila exd and hth were identified based on their sharing embryonic segmentation defects with Hox mutants(Peifer and Wieschaus, 1990; Rieckhof et al., 1997). In C. elegans, although there are some similarities between the Q migration defects of unc-62, ceh-20 and Hox gene mutants, the details of their phenotypes are very different. The same holds true for vulva development. Thus, ceh-20 and unc-62 may have functions independent of lin-39 in anteriorwards migration and vulva development. In Drosophila, Hox-independent functions of exdand hth have been demonstrated for antennal formation(Casares and Mann, 1998; Yao et al., 1999). However, in contrast to the C. elegans situation in which lin-39 is required and expressed in the cells with defects in migration, fusion or division, Drosophila Hox genes are neither required nor expressed in the eye-antennal disc from which the antenna forms(Casares and Mann, 1998; Clark et al., 1993; Wang et al., 1993; Yao et al., 1999).

Third, Hox gene expression is regulated, in part, by Hox co-factors. In Drosophila, exd and hth are required to maintain Sex-combs reduced expression in the salivary gland(Henderson and Andrew, 2000). In zebrafish, Hox genes interact with their co-factors to cross-regulate the expression of other Hox genes (Maconochie et al., 1997; Popperl et al.,1995; Studer et al.,1998). As a result, reducing the function or altering the intracellular localization of zebrafish lazarus or meis/prepgenes resulted in decreased expression of various Hox genes.(Choe et al., 2002; Popperl et al., 2000; Waskiewicz et al., 2001). In C. elegans, although ceh-20 does not activate mab-5or lin-39 expression in the mesoderm, it is required for the autoregulatory expression of ceh-13, the C. elegans ortholog of Drosophila Hox gene labial(Liu and Fire, 2000; Streit et al., 2002). We found that lin-39 expression was upregulated, instead of downregulated, in the Q cells of some animals with reduced ceh-20 or unc-62function. Although this was somewhat unexpected, it is possible that ceh-20 and unc-62 downregulate Hox gene expression in conjunction with non-HOX proteins downstream of signaling pathways. Although the wingless and the TGFβ pathways have not been shown to inhibit Hox gene expression, they do influence Hox gene expression in C. elegans(Stoyanov et al., 2003; Streit et al., 2002). Because lin-39 upregulation was not restricted to the Q cells in ceh-20 or unc-62 mutants, we raise the possibility that ceh-20 and unc-62 could function with repressors of lin-39 expression such as SAM domain proteins(Zhang et al., 2003).

Taken together, these comparisons suggest that C. elegans ceh-20and unc-62 share many properties with their homologs, but the degree to which the Hox-independent and Hox-dependent roles of these genes and their homologs dominate may differ in different organisms and even between different tissues within the same organism.

We thank Joy Alcedo, Scott Alper, Javier Apfeld, Queelim Ch'ng, Douglas Crawford, Natasha Libina, Yung Lie, Jennifer Whangbo, Lisa Williams and other members of the Kenyon laboratory for technical advice, supportive environment,stimulating discussions and critical comments on the work. We especially thank Natasha Libina for her expertise in preparing figures for publication. We thank Yung Lie for rrf-3(pk1426); muIs35, Queelim Ch'ng for P_unc-119::mig-13 and Andy Dillin for the control RNAi vector. We also thank the Sanger Center for cosmids and the CGC for numerous strains. In addition, we are indebted to Michael Stern for providing the egl-17 promoter, unpublished data and ceh-20 alleles. We are also grateful to William Wood and Barbara Robertson for communicating unpublished results and for additional unc-62 alleles. We thank Meera Sundaram for unc-62(ku234). Finally, we thank Cori Bargmann for her excellent advice and her laboratory for reagents. L.Y. was supported by the Medical Scientist Training Program (NIH-NIGMS GM07618). M.S. was supported by a postdoctoral fellowship from the Jane Coffin Childs Fund for Medical Research. This work was supported by NIH grant GM37053 to C.K.