The C. elegans Hand gene controls embryogenesis and early gonadogenesis

An organ primordium consists of precursor cells that generate all the diverse cell types of the mature organ. For some organs, `organ selector'genes control all precursor cells within the organ primordium, regardless of cell type (Gaudet and Mango,2002; Gehring and Ikeo,1999). By contrast, precursor cells in other organ primordia are specified by the intersection of global patterning genes and may not rely on a single organ selector gene (Bradley et al.,2001; Lockwood and Bodmer,2002). However, the generality of these mechanisms remains unknown.

We have focused on developmental controls of the gonadal primordium in the nematode Caenorhabditis elegans. This organ primordium is unusually simple: it is composed of two somatic gonadal precursor cells (SGPs) and two primordial germ cells (PGCs). The SGPs generate all somatic tissues of the gonad proper (i.e. ovary or testis), as well as genital ducts (e.g. uterus,vas deferens), whereas the PGCs give rise to all germ cells, including gametes. The SGPs and PGCs arise from distinct embryonic blastomeres and assemble into the gonadal primordium midway through embryogenesis(Fig. 1A,B)(Sulston et al., 1983). Within the mature primordium, the SGPs (Z1 and Z4) reside at the distal poles and the PGCs (Z2 and Z3) are situated proximally(Fig. 1C)(Kimble and Hirsh, 1979). In addition to this proximal-distal polarity, the primordium displays left-right and dorsal-ventral polarity (Fig. 1C). Therefore, the four-celled gonadal primordium is patterned in three axes.

Somatic and germline precursors originate from distinct embryonic lineages during gonadogenesis in other organisms as well. In Drosophila, SGPs are specified by global patterning genes that subdivide the mesoderm into discreet regions (Boyle et al.,1997; Boyle and DiNardo,1995). Once specified, the clift (eyes absent– FlyBase) gene, which encodes a novel nuclear protein, maintains the SGP fate (Boyle et al., 1997). The mammalian gonadal mesoderm may similarly rely on genes that pattern the embryo as a whole (Capel,2000). In addition, several transcription factors affect development of the gonadal mesoderm in mice: Wt1, Sf1, Lim1(Lhx1 – Mouse Genome Informatics) and Emx2 control both gonadal and non-gonadal development(Kreidberg et al., 1993; Miyamoto et al., 1997),whereas Lhx9 appears specific for gonadogenesis(Birk et al., 2000). PGCs, on the other hand, are often formed outside the gonad and later migrate to the developing gonad (reviewed by Starz-Gaiano and Lehmann, 2001; Wylie,2000). Once there, the PGCs depend on the somatic gonad for survival and for cell fate decisions within the germ line(Kimble and White, 1981; McCarter et al., 1997; Starz-Gaiano and Lehmann,2001).

Several genes have been identified that control early gonadogenesis in C. elegans (Fig. 1C)(Hubbard and Greenstein,2000). For example, gon-2 and gon-4 control the onset and timing of gonadal cell divisions(Friedman et al., 2000; Sun and Lambie, 1997), and lin-17, sys-1, wrm-1, lit-1 and pop-1 govern the asymmetric division of the SGPs (Siegfried and Kimble, 2002; Sternberg and Horvitz, 1988). For germline development, pie-1 and nos-2 control PGC fate and influence their incorporation into the gonadal primordium (Seydoux et al.,1996; Subramaniam and Seydoux,1999; Tenenhaus et al.,2001). In this paper, we report that the C. elegans Hand bHLH transcription factor hnd-1 is important for early gonadogenesis as well as for embryogenesis. Specifically, hnd-1 influences the number and position of SGPs in the gonadal primordium, and affects body shape in the embryo. The hnd-1 gene is expressed broadly in the embryonic mesoderm and then more specifically in the SGPs. Our results suggest that hnd-1 governs maintenance of SGP fate and SGP survival. We also report the discovery of two genetic enhancers of hnd-1, named ehn-1 and ehn-3 (for enhancer of Hand), that have overlapping functions with hnd-1 in embryogenesis and gonadogenesis.

Animals were grown at 20°C unless otherwise noted. All strains were derivatives of Bristol strain N2 (Sulston and Horvitz, 1977). The following mutations are described by Hodgkin (Hodgkin, 1997) or cited references. LGI: gon-2(q388)(Sun and Lambie, 1997) and sys-1(q544) (Miskowski et al.,2001). LGII: hlh-1(cc450)(Chen et al., 1994); unc-104(e1265); rol-6(e187); and mnDf93(Sigurdson et al., 1984). LGIV: ced-2(e1752); ced-3(n717); gon-4(q519) (Friedman et al.,2000); unc-24(e138); unc-5(e53); dpy-13(e184); and nDf41. LGX: unc-9(e101). Dominant GFP balancers: mIn1[mIs14] for LGII(Edgley and Riddle, 2001); hT2[qIs48] for LGI; and nT1[qIs50] for LGIV. qIs48 and qIs50 are insertions of ccEx9747 onto hT2 and nT1, respectively. Molecular markers: qIs55[hnd-1(N)::GFP]; qIs69 [hnd-1::GFPlacZ]; qIs56 [lag-2::GFP](Siegfried and Kimble, 2002); leIs129 [pes-1::GFP](Molin et al., 2000); qIs61 [pes-1::GFP]; ayIs7 [hlh-8::GFP](Harfe et al., 1998); and qIs77 [unc-122::GFP](Miyabayashi et al., 1999). qIs56 and qIs61 were generated by microparticle bombardment(Praitis et al., 2001).

All cloning was performed by standard methods(Sambrook et al., 1989). PCR products were sequenced. Primer sequences are available upon request. Transgenes were generated as simple arrays unless otherwise noted.

Using a probe from the coding region of C44C10.8, we isolated a hnd-1 cDNA from an embryonic C. elegans cDNA library (a gift from P. Okkema) and subcloned it to make pJK849. The hnd-1 5′end was cloned by RT-PCR using embryonic total RNA, a primer to the SL1 trans-spliced leader and internal hnd-1-specific primers. A full-length hnd-1 cDNA (pJK901) was assembled from the SL1 RT-PCR product and pJK849.

GFP coding sequences were amplified by PCR from pPD95.81 (a gift from A. Fire) and subcloned into a hnd-1 genomic fragment (pJK906). pJK850 includes 1568 bp of the hnd-1 sequence upstream of the 5′UTR and 182 bp downstream of the 3′UTR. pJK850 was injected with pRF4[Rol](Mello et al., 1991) into hnd-1 to generate qEx486; this array rescued hnd-1gonadal defects completely (n=136) and reduced lethality from 28% to 7% (n=190).

The first two exons and 1540 bp upstream of the hnd-1 5′UTR were PCR amplified and cloned into pPD95.81 (a gift from A. Fire). pJK848 was injected into unc-4(e120) with the co-injection marker pNC4-21[unc-4+] (Miller and Niemeyer, 1995) and N2 DNA to create qEx447 and,subsequently, qIs55. With the exception of SGPs, hnd-1(N)::GFP was detected in cells that also express hlh-1,a marker for body muscle (Krause et al.,1990).

The hnd-1 promoter (plus 11 N-terminal codons) was PCR amplified and cloned into pPD96.04 (a gift from A. Fire). pJK900 was injected with pRF4[Rol+] to create qEx492 and, subsequently, qIs69. pJK850 and pJK900, but not pJK848, express GFP in several head cells that we have not identified.

The hnd-1 cDNA from pJK901 was cloned into pPD49.78 (a gift from A. Fire) to generate pJK902, which was injected into qIs61 with the co-injection marker pRF4[Rol+] to make qEx493. Embryos were subjected to two 30-minute heat pulses at 33°C, with a one hour recovery interval. Resulting L1 larvae were scored for extra SGPs using pes-1::GFP.

A hnd-1::GFP fusion was generated by inserting GFP into the RsrII site of the full-length hnd-1 cDNA (pJK901). hnd-1::GFP was then cloned into pPD51.45(Krause et al., 1990) to generate pJK904, which was injected into hnd-1 with the co-injection marker pRF4[Rol+] to make qEx496; this array rescued hnd-1gonadal defects and marginally rescued lethality (20%, n=372).

A hnd-1::GFP fusion (see above) was cloned into pJK590(Blelloch et al., 1999) to generate pJK905, which was injected into hnd-1 with the co-injection marker pRF4[Rol+] to make qEx497; this array partially rescued hnd-1 gonadogenesis defects and did not rescue lethality (24%, n=192).

A plasmid carrying hnd-1 genomic DNA was amplified by PCR; it contained the same upstream and downstream sequences as in hnd-1(FL)::GFP. pJK906 was injected with pPD136.64[myo-3::YFP] (a gift from A. Fire) and pJK907 [pes-1::CFP]into hnd-1 to make qEx495; this array was used for mosaic analysis.

The pes-1 promoter from pUL#MJA1(Molin et al., 2000) was cloned into pPD136.64 (a gift from A. Fire).

Double-stranded hnd-1 RNA was generated, using pJK849 as template,and injected at 1 mg/ml. The hnd-1(q740) deletion was isolated essentially as described by Kraemer et al.(Kraemer et al., 1999), and backcrossed eight times. To test for maternal effects, hnd-1 females,generated by fog-1 RNAi (Jin et al., 2001), were crossed with N2 males [17% of the cross-progeny died as embryos or young larvae (n=313), and all adult progeny had normal gonads (n=260)]. To test for zygotic lethality, we scored progeny of unc-9 hnd-1/++ mothers [6% died as embryos or larvae(n=235)]. To investigate whether hnd-1(q740) was a null allele, RT-PCR was performed on mutant and wild-type worms, using primers to a region retained in the hnd-1 deletion. Template RNA was prepared from 20 gravid adults using TRI reagent (Molecular Research Center). A PCR product was obtained only from wild-type worms.

Progeny of hlh-1/+; hnd-1 mothers had 55% embryonic and larval lethality, compared with 25% defects for hlh-1/+(Chen et al., 1994) and 28%defects for hnd-1 (this work).

The following were evaluated using number of gonadal arms as a measure:

100% of sys-1/+; hnd-1/+ worms had two arms (n=89). sys-1/+; hnd-1 had 68% gonadal arms, compared with 70% for hnd-1 alone and <1% for sys-1/+. sys-1 dominantly enhances other Sys mutants (K. Siegfried, unpublished).

100% of gon-4/+; hnd-1/+ worms had two arms (n=112). gon-4; hnd-1 double mutants had 30% gonadal arms (n=46),compared with 43% for gon-4(Friedman et al., 2000) and 70% for hnd-1.

At 20°C, the progeny of gon-2; hnd-1 worms had 78% gonadal arms (n=89), compared with 70% for hnd-1 and 100% for gon-2 (Sun and Lambie,1997). The progeny of gon-2; hnd-1 worms shifted to 25°C as L4s resembled gon-2 alone (most had no visible gonad).

Embryos were fixed essentially as described by Miller and Shakes(Miller and Shakes, 1995), and then stained with α-PGL-1, a component of P granules(Kawasaki et al., 1998),α-HLH-1 (Krause et al.,1990) or α-UNC-54(Miller et al., 1983) for body muscle. Secondary antibodies were used at 1:400 (Jackson Labs, West Grove,PA). DAPI staining was performed as described by Kadyk and Kimble(Kadyk and Kimble, 1998).

ehn-1(q638) was identified in a EMS mutagenesis screen for gonadogenesis mutants (L.D.M., K. Siegfried, F.-H. Markussen, unpublished). ehn-1(q690) and ehn-3(q689) were obtained in an ehn-1 non-complementation screen of 2251 haploid genomes: EMS mutagenized males were crossed to ehn-1(q638) rol-6 hermaphrodites,and F2 progeny screened for gonadal defects. By three-factor mapping, we positioned ehn-1 between unc-104 and rol-6 on linkage group II, and ehn-3 between dpy-13 and unc-5 on linkage group IV. ehn-1 is almost maternally and zygotically sufficient for gonadogenesis. From an ehn-1(q638) rol-6mother, 1% of heterozygous cross progeny had defects (n=221); from an ehn-1(q638) rol-6/++ mother, none of the ehn-1 rol-6homozygous progeny had defects (n=199). ehn-3 has minor dominance: <1% of ehn-3 unc-5/++ had gonadogenesis defects(n=222).

The C. elegans genome encodes a single Hand bHLH protein(Ledent and Vervoort, 2001; Ruvkun and Hobert, 1998). We initially investigated hnd-1 (C44C10.8) because of its sequence similarity to vertebrate proteins implicated in Notch signaling(Saga et al., 1997). The hnd-1 gene is predicted to have four exons encoding a 226 amino acid protein with one bHLH domain and no other motif(Fig. 2A,B). We confirmed these four exons in a single cDNA and identified the 5′ end using RT-PCR with the trans-spliced leader SL1 (Fig. 2A; Materials and Methods). During the course of this work, we isolated hnd-1(q740), a deletion mutant that removes 674 bp from the hnd-1 5′ flanking region, as well as its first two exons and introns (Fig. 2A). Consistent with hnd-1(q740) being a null allele, we did not detect hnd-1 mRNA in hnd-1 mutants (Materials and Methods). Furthermore, hnd-1(RNAi) and the hnd-1 deletion caused similar defects (Table 1),suggesting that both represent strong and perhaps complete loss-of-gene function.

To explore hnd-1 function, we examined both hnd-1(RNAi)animals and the hnd-1-deletion mutant. Similar defects were observed,and penetrance, recessivity and maternal effects were analyzed in the mutant(Materials and Methods). The hnd-1 phenotype includes a partially penetrant lethality (28%, n=909), as well as partially penetrant gonadal defects in surviving adults (52%, n=656). Aside from gonadal defects and minor body wall abnormalities, hnd-1 survivors appeared normal. The hnd-1 mutant was recessive: the gonadal defects had no maternal effect, but both maternal and zygotic hnd-1 activities were important for viability.

hnd-1 mutants can die as embryos or young larvae with variable body shape defects, typically in the posterior(Fig. 3A,B). Most hnd-1 embryos contained pharynx and gut(Fig. 3C,D), as well as muscle,as evidenced by twitching. Because hnd-1::GFP is expressed in mesodermal precursors (see below), we compared body wall muscles in wild-type and hnd-1 embryos using an α-myosin antibody(Miller et al., 1983). Both wild-type and hnd-1 late-stage embryos have four quadrants of body muscle (Fig. 3E,F)(Miller et al., 1983),although muscle fibers were sometimes disorganized in mutants(Fig. 3F).

hlh-1 has striking similarities to hnd-1. The hlh-1 gene encodes a MYOD-like bHLH protein, and hlh-1mutants have severe defects in embryo morphogenesis but generate body muscle normally (Chen et al., 1994). We examined hnd-1; hlh-1 double mutants to determine whether these two bHLH proteins might have overlapping functions, but double mutants made body muscle (not shown). Furthermore, we found no significant genetic interaction between hlh-1 and hnd-1 (Materials and Methods). Therefore, hnd-1 and hlh-1 appear to function independently.

Wild-type hermaphrodites possess two gonadal arms. By contrast, adult hermaphrodites depleted for hnd-1 displayed a range of gonadal shapes: two gonadal arms (Fig. 4A); a single gonadal arm (Fig. 4B); no apparent gonad (Fig. 4C); or abnormal gonads (Fig. 4D). `Abnormal gonads' include a variety of shapes, most typically an amorphous mass (Fig. 4D). One-armed and two-armed gonads were frequent and were usually fertile, whereas absent and abnormal gonads were less common and were always sterile(Table 1). A similar, but less penetrant effect was seen in males (not shown). We conclude that hnd-1 is important, but not essential, for gonadogenesis.

The gonadal morphologies in hnd-1 mutants suggested a defect early in gonadogenesis. Therefore, we examined SGPs in hnd-1 gonadal primordia, using either nuclear pes-1::GFP(Molin et al., 2000)(Fig. 5A-C) or cytoplasmic lag-2::GFP, which reveals cellular processes(Blelloch et al., 1999)(Fig. 5G,J). Whereas all wild-type gonadal primordia had two SGPs(Table 2A), hnd-1primordia could have two (Fig. 5A,B), one (Fig. 5C), zero (Fig. 5D)or even three SGPs (Table 2A). Therefore, hnd-1 is important for determining SGP number.

SGP position was also affected in hnd-1 mutants. In wild type, the two SGPs reside at the distal poles of the primordium, flanking the PGCs and extending cytoplasmic processes to meet mid-ventrally. In most hnd-1mutants, SGPs occupied similar polar positions(Table 2B), and extended ventral processes (Fig. 5E-G). However, in some hnd-1 mutants, one or both SGPs were not at the pole, but instead were found more centrally in the primordium(Table 2B). When an SGP was misplaced dorsally, it extended cytoplasmic processes along the dorsal surface of the primordium (Fig. 5H-J). Finally, in some hnd-1 mutants, SGPs were observed ectopically(Table 2B). These ectopic SGPs could be in animals with either two or three total SGPs; as predicted, ectopic gonadal arms have been observed in rare hnd-1 mutants. Therefore, hnd-1 can affect the position of the SGPs within the primordium and within the animal.

We used hnd-1 mutants born with aberrant gonadal primordia to investigate how organization of that primordium affected gonadogenesis. Specifically, we used pes-1::GFP to score SGPs in hnd-1 L1 larvae, permitted the animals to develop and then examined them again as L4s. Our results (Table 3) led to three conclusions. First, most hnd-1 primordia with a wild-type appearance (two SGPs placed at the poles) generated wild-type appearing adults with two gonadal arms (93%, n=70). Therefore, hnd-1 appears to play little or no role in gonadogenesis after formation of the gonadal primordium. Second, most primordia containing one SGP generated adult gonads with only a single arm (98%, n=53); none made two arms. Finally,primordia with wild-type SGP number but aberrant SGP position often generated defective gonads (26%, n=23). Therefore, SGP position within the primordium may be important for gonadogenesis.

The gonadal primordium in hnd-1 mutants sometimes lacked PGCs. In primordia with two SGPs, 7% lacked one or both PGCs (n=59), and in primordia with one SGP, 40% were missing at least one PGC (n=47). To investigate whether the two PGCs were made in hnd-1 mutants, we stained ∼100-cell embryos with a germline-specific antibody, α-PGL-1(Kawasaki et al., 1998): all had two PGL-1-staining cells (n=237). Therefore, hnd-1 does not affect the generation of PGCs, but instead affects their maintenance within the gonadal primordium. A simple interpretation is that hnd-1acts in SGPs, which in turn are essential for germline survival and positioning.

To investigate hnd-1 expression, we constructed three GFP reporters (Fig. 2C). hnd-1(FL)::GFP inserts GFP coding sequences into the third exon of the full-length HND-1 protein; this reporter rescued hnd-1 mutants(Materials and Methods). hnd-1(N)::GFP inserts GFP more N-terminally and replaces the hnd-1 3′UTR with the unc-543′UTR. hnd-1::GFPlacZ replaces most of the hnd-1coding region with GFP and β-galactosidase coding sequences, and the unc-54 3′UTR. All three hnd-1 reporters expressed GFP in largely the same cells, but hnd-1(N)::GFP and hnd-1::GFPlacZ expressed GFP at a higher level and expression persisted longer. The rescue by hnd-1(FL)::GFP suggests that its expression is relevant to hnd-1 function.

The hnd-1 reporters expressed GFP in the MS, C and D embryonic lineages (Fig. 6A). Expression was first observed in four MS great-granddaughters, four C great-granddaughters and two D daughters(Fig. 6B). These MS descendants give rise to the SGPs and other mesodermal cells(Fig. 6A)(Sulston et al., 1983); the C-and D-expressing cells all generate body wall muscle(Sulston et al., 1983). Expression continued through one cell division(Fig. 6C) and then became difficult to detect using hnd-1(FL)::GFP. hnd-1(N)::GFP remained detectable in some cells within these MS and C lineages(Fig. 6D), but disappeared from most body muscle cells by the comma stage of embryogenesis(Fig. 6E). Then, the hnd-1 reporters were expressed in the SGPs (Z1 and Z4) as they approached the PGCs to form the gonadal primordium(Fig. 6E,F). Shortly after the primordium was assembled, hnd-1(N)::GFP expression was reduced or disappeared (Fig. 6G). GFP was not detected in the SGPs at hatching or post-embryonically (not shown). Therefore, hnd-1 appears to be expressed during embryogenesis in mesodermal precursor cells that generate predominantly body wall muscle, and then in SGPs.

The most common gonadal defect in hnd-1 mutants is a missing SGP(Table 2A). To investigate whether both SGPs are made in hnd-1-mutant embryos, we used the hnd-1(N)::GFP reporter, which is an early SGP marker. We found both SGPs present and in their normal position near the PGCs(Fig. 6H; n=29). Therefore, hnd-1 is not necessary for SGP specification or SGP migration to the PGCs.

After formation of the gonadal primordium, SGPs remained associated with PGCs in hnd-1 embryos; none were seen detaching. Instead, SGP nuclei sometimes became smaller and hnd-1 reporter expression faded prematurely (38%, n=13). The simplest hypothesis is that hnd-1 is required for maintenance of SGP fate and possibly SGP survival.

We next investigated whether ectopic hnd-1 expression could transform other cells to the SGP fate. To this end, we used a heat-inducible promoter to express hnd-1 during embryogenesis but found no ectopic SGPs, as assayed by pes-1::GFP (Materials and Methods). We also expressed a full-length HND-1::GFP fusion protein under control of either of two mesodermal promoters (see below), but again did not observe ectopic SGPs. These results are consistent with the proposed role for hnd-1 in controlling SGP maintenance or survival.

The missing SGPs in hnd-1 mutants might be explained by transformation to a different cell type, or by cell death. To explore the first idea, we reasoned that the most likely transformation would be to a different mesodermal cell type. We tested this using hlh-8::GFP to mark the M mesoblast (Harfe et al.,1998) and unc-122::GFP to mark coelomocytes(Miyabayashi et al., 1999). All wild-type L1s had a single hlh-8::GFP-expressing M mesoblast, as expected (n=61). Similarly, most hnd-1 mutants had a single M cell, but a few had two M cells (5%, n=63) or no M cell (2%, n=63). Those with an additional M cell had two SGPs, suggesting that extra M cells were not transformed SGPs. Likewise, occasional extra coelomocytes were seen, but overall hnd-1 mutants had marginally fewer coelomocytes than wild type (5.5 versus 5.9 on average per animal, n>30). Importantly, the extra coelomocytes could be in worms with two gonadal arms. Therefore, hnd-1 appears to have a low-penetrance effect on M cells and coelomocytes, but this is unlikely to account for the missing SGPs.

To determine whether SGPs are lost as a result of programmed cell death in hnd-1 mutants, we examined hnd-1; ced-3 double mutants using the pes-1::GFP marker. The ced-3 gene is required for all programmed cell deaths (Ellis and Horvitz,1986). In hnd-1 single mutants, 42% were missing at least one SGP (Table 2A), and, in ced-3; hnd-1 double mutants, 45% lacked at least one SGP(n=69). Therefore, SGP loss does not appear to rely on ced-3-dependent programmed cell death.

Next, we investigated whether SGPs died in hnd-1 mutants. In C. elegans, cell corpses resulting from either programmed or necrotic cell death are engulfed by their neighbors(Chung et al., 2000; Ellis et al., 1991). The engulfment of cell corpses relies on several genes, including ced-2(Ellis et al., 1991). In ced-2 single mutants, no cell corpses were evident near the gonad(Fig. 7A,B; n=54);however, in ced-2; hnd-1 double mutants, cell corpses were found near the gonad (Fig. 7C,D; 28%, n=50). Importantly, the presence and site of corpses correlated with SGP absence. We observed no cell corpses near gonads with two SGPs(n=53), mostly anterior or right cell corpses near gonads missing Z1(4/5; Fig. 7C), and only posterior or left cell corpses in those missing Z4 (8/8). Indeed, in one cell corpse, hnd-1::GFP was faintly expressed, indicating that it had been specified originally as an SGP (data not shown). Therefore, SGPs appear to die in hnd-1 mutants.

Both SGPs and PGCs are affected in hnd-1 mutants. To learn where hnd-1 functions, we used a combination of mosaic analysis and transgenic experiments driving the hnd-1 coding region with tissue-specific promoters.

In C. elegans, mosaic animals can be made by loss of extra-chromosomal arrays that carry transgenes and that are transmitted with varying fidelity at each cell division(Herman, 1984). For this study, we created an extra-chromosomal array that carries a rescuing hnd-1 genomic fragment and two fluorescent markers(myo-3::YFP to mark body muscle and pes-1::CFP to mark SGPs). We then identified `germline mosaics', animals that retained the array in somatic tissues but failed to transmit it to their progeny; such animals have lost the array in divisions generating the germline blastomere P4 (see Fig. 6). All six germline mosaics had a wild-type gonadal primordium, which suggests that hnd-1activity acts in somatic tissues rather than in the germ line.

To further explore where hnd-1 acts, we used either of two promoters: hlh-1, which is expressed in body muscle and not in SGPs(Krause et al., 1990); or lag-2, which is first expressed in the AB and MS lineages(Moskowitz and Rothman, 1996),and then in SGPs (Miskowski et al.,2001). Each promoter was fused to a full-length, rescuing hnd-1::GFP cDNA and expressed in hnd-1 mutants. Expression of HND-1::GFP by the hlh-1 promoter rescued the hnd-1gonadogenesis defects, from 52% to 5% defective (n=42). By contrast,HND-1::GFP driven from the lag-2 promoter, which is expressed in the two SGPs (Fig. 5G), did not appreciably rescue hnd-1 gonadogenesis defects (36% defective, n=108). The latter experiment has the caveat that this promoter is switched on after SGPs assemble into the gonadal primordium and it may not be expressed in dying SGPs. From the hlh-1::hnd-1GFP result, we suggest that HND-1 acts in early mesodermal lineages.

The hnd-1 SGP defects are the earliest observed to date among any genes controlling C. elegans gonadogenesis. To investigate whether hnd-1 might function with other early gonadogenesis genes, we explored genetic interactions between hnd-1 and two mutant classes. The first type, represented by gon-2 and gon-4(Friedman et al., 2000; Sun and Lambie, 1997),controls the onset of cell divisions in the gonad but not in other tissues(Fig. 1C). Gonadal divisions are delayed in gon-2 or gon-4 single mutants(Friedman et al., 2000; Sun and Lambie, 1997), but not in hnd-1 mutants (n=5). Moreover, hnd-1; gon-2 and hnd-1; gon-4 double mutants have additive phenotypes (Materials and Methods). Therefore, hnd-1 does not affect the onset or timing of gonadal divisions and acts independently of gon-2 and gon-4.

The second class of early gonadogenesis genes, represented by sys-1 (Miskowski et al.,2001), is required for SGPs to produce daughter cells with different developmental potential (Fig. 1C). In wild type, each SGP generates one distal tip cell (DTC),whereas in sys-1 mutants they make no DTCs(Miskowski et al., 2001). Most hnd-1 SGPs that were properly positioned generated DTCs (96%, n=193; Table 3), and no genetic interactions were found with sys-1 (Materials and Methods). Therefore, hnd-1 does not appear to affect SGP asymmetric divisions, but instead ensures that two SGPs are present and properly positioned in the gonadal primordium.

To identify additional genes controlling SGP development, we screened for EMS-induced mutants with a hnd-1-like gonadogenesis phenotype and discovered loss-of-function mutations of ehn-1 and ehn-3[for enhancer of Hand (Materials and Methods)]. The ehn-1 and ehn-3 mutants had low-penetrance gonadal defects(Table 1). For ehn-1,gonadal defects could be rescued either maternally or zygotically, but ehn-3 exhibited no maternal effect (Materials and Methods). Furthermore, ehn-1 had low-penetrance lethality, but lethality was negligible in ehn-3 mutants (Table 4).

The gonadal primordia of ehn-1 and ehn-3 mutants had absent or misplaced SGPs, as described above for hnd-1 mutants(Table 2A,B). However, hnd-1, ehn-1 and ehn-3 displayed subtle differences in their spectrum of defects. For example, ectopic SGPs were seen in ehn-1 but not ehn-3 mutants, and only hnd-1 mutants generated extra SGPs (Table 2A). Despite these minor differences, the primary defects were similar among the three mutants,which suggests that they may function in a common pathway to control early gonadal development.

To investigate the functional relationships between the hnd-1 and EHN genes, we first investigated double and triple mutants(Table 4). Although hnd-1,ehn-1 and ehn-3 single mutants all had relatively low-penetrance gonadal defects, the double and triple mutants showed increased penetrance(Table 4). For example, 80-90%of ehn-1 and ehn-3 single-mutant adults had two gonadal arms, but almost none of the ehn-1; ehn-3 double mutants had two gonadal arms (2-4%; Table 4). Similarly, only 5% of the double mutants had two SGPs at hatching(Table 2A). By contrast, larval lethality did not increase in the ehn-1; ehn-3 double mutant. Therefore, ehn-1 and ehn-3 may be partially redundant for SGP development.

The ehn-1; hnd-1 and ehn-3; hnd-1 double mutants were also more defective than any of the single mutants, but each double mutant was unique. For the gonadogenesis defects, ehn-3 enhanced hnd-1more strongly than did ehn-1. Thus, some ehn-1; hnd-1 double mutants made two gonadal arms and only about one-third had no apparent gonad. By contrast, no ehn-3; hnd-1 double mutants had two gonadal arms, and most had no visible gonad (Table 4). Intriguingly, this situation was reversed for lethality: ehn-1 enhanced hnd-1 more strongly than did ehn-3for both embryonic and larval lethality. One simple explanation is that the three genes are all partially redundant, but that each has acquired an individual role in the repertoire of activities normally carried out by hnd-1/EHN genes (see Discussion).

The ehn-1; ehn-3; hnd-1 triple mutant appears additive for the ehn-1; hnd-1 and ehn-3; hnd-1 defects. Thus, the penetrance of the triple mutant with respect to lethality is similar to that of the ehn-1; hnd-1 double mutant, and the penetrance of the triple mutant with respect to gonadal defects is similar to that of the ehn-3;hnd-1 mutant (Table 4). The fact that the triple mutant is not fully penetrant may suggest the existence of one or more additional genes involved in the process, or it may indicate that the ehn-1 or ehn-3 mutant is not a null.

To begin addressing relationships between the ehn genes and hnd-1 at a molecular level, we examined expression of hnd-1(N)::GFP to mark hnd-1 transcription and SGP formation. The ehn-1 and ehn-3 single mutants both expressed hnd-1(N)::GFP in two SGPs (ehn-1, n=35; ehn-3,n=31). Therefore, ehn-1 and ehn-3 do not control hnd-1 transcription in SGPs, which is consistent with the idea that they function in parallel to hnd-1. Furthermore, ehn-3;hnd-1 double mutant embyros made two SGPs (n=8), but few possessed SGPs at hatching (4%; Table 2A). Therefore, like hnd-1, the ehn-1 and ehn-3 genes do not affect SGP specification but instead influence SGP fate or survival.

In this paper, we investigate the controls governing development of the C. elegans gonadal primordium. Our primary focus is the hnd-1 gene, which encodes the single Hand transcription factor in the C. elegans genome (Ledent and Vervoort, 2001; Ruvkun and Hobert, 1998). Animals lacking hnd-1 activity have partially penetrant defects in gonadogenesis and embryogenesis (this work). In addition, we have found two genetic enhancers of hnd-1 that have overlapping functions. We discuss the roles played by these genes in C. elegans organogenesis and compare our findings to similar studies in other organisms.

Wild-type C. elegans has four gonadal precursors, two SGPs and two PGCs, in stereotyped positions within the primordium. hnd-1 mutants affect the presence and position of these precursors in the primordium and within the animal. Thus, hnd-1 mutants can possess fewer than normal,as well as mispositioned, SGPs or PGCs. The hnd-1 gene probably acts cell autonomously in the SGPs or their precursors to control early gonadogenesis. However, hnd-1 does not affect SGP specification,because the correct number of SGPs is generated in all hnd-1-mutant embryos. Nor does it cause SGPs to be transformed into either of two mesodermal types (coelomocytes and the M mesoblast), although it remains possible that they are transformed into muscle cells. Instead, we suggest that hnd-1 is required for SGP survival.

What happens to SGPs in hnd-1 mutants? We used two classes of cell death mutants: ced-3, which eliminates all programmed cell death(Ellis and Horvitz, 1986), and ced-2, which is defective in cell corpse engulfment(Ellis et al., 1991). At first glance, our results appear contradictory: we observed extra cell corpses in hnd-1; ced-2 double mutants, but saw no increase in the number of SGPs in hnd-1; ced-3 double mutants. One simple explanation is that SGPs die by a ced-3-independent pathway. Alternatively, if the SGPs no longer expressed markers of their fate (e.g. pes-1), they would not have been identified in our analysis of hnd-1; ced-3 mutants. Therefore, it remains possible that hnd-1 mutant SGPs fail to maintain their fate and die via programmed cell death. In either case, the correlation between missing SGPs and extra cell corpses strongly supports the idea that hnd-1 is required for SGP survival.

Why might SGPs die in hnd-1 mutants? One simple explanation is that inhibition of apoptosis is part of the normal developmental program, as has been suggested for the winged-helix transcription factor Fork head in Drosophila salivary gland development(Myat and Andrew, 2000). Alternatively, cells may be programmed to die when they receive ambiguous developmental cues. This idea is supported by the extensive apoptosis seen in many developmental mutants [e.g. Pax6/eyeless mutants(Halder et al., 1998)]. Because hnd-1 SGPs initially show evidence of their fate (they express SGP markers and migrate to the PGCs), we favor the idea that hnd-1 is required for maintenance of cell fate and that in its absence the SGPs die. Similarly, Pax6/eyeless mutants generate eye primordia that express early markers of their fate (e.g. ey-eye enhancer lacZ) and later undergo programmed cell death(Halder et al., 1998).

In addition to gonadal defects, hnd-1 mutants can die as embryos or young larvae with body morphogenesis defects. Elongation of the embryo is driven largely by cell shape changes in the hypodermis(Priess and Hirsh, 1986). However, mutants affecting muscle development also disrupt the process (e.g. Bejsovec and Anderson, 1988; Chen et al., 1994; Waterston, 1989). Of particular interest to this work is the hlh-1 gene, which encodes the C. elegans myoD homolog (Krause et al., 1990); its loss disrupts development of body wall muscles and causes a characteristic morphogenesis defect(Chen et al., 1994). We explored the possibility that hnd-1 may similarly be involved in body muscle development. However, hnd-1; hlh-1 double mutants still make body muscle, suggesting that these bHLH proteins control different aspects of muscle development. Although speculative at the current time, we suggest that hnd-1 may play a role in muscle fate that parallels its role in controlling SGP fate.

The hnd-1 deletion has incompletely penetrant gonadal and embryonic defects. Yet, the mouse and zebrafish Hand mutants are completely penetrant (Firulli et al.,1998; Riley et al.,1998; Srivastava et al.,1997; Yelon et al.,2000). Why might a hnd-1-null mutant exhibit partially penetrant defects? One simple explanation is genetic redundancy. We have identified two genes, ehn-1 and ehn-3, that enhance the hnd-1 phenotype. All three single mutants have partially penetrant gonadal defects, and mutations in two of the three, hnd-1 and ehn-1, also affect viability. Each of the double mutants has a more severe gonadogenesis defect, which suggests at least two pathways control SGP survival. Redundancy frequently results from gene duplication(Ohno, 1970). However, only one Hand homolog exists in the C. elegans genome, and neither ehn-1 nor ehn-3 maps to a region containing any predicted bHLH protein. Therefore, the hnd-1 and EHN genes redundantly control SGP development, but they are unlikely to represent paralogous pathways.

Intriguingly, ehn-1 enhances hnd-1 lethality more strongly than it enhances the hnd-1 gonadal defect, whereas ehn-3 enhances the hnd-1 gonadal defect but not its lethality. The identity of hnd-1 as a putative bHLH transcription factor provides a molecular framework for considering the enhancement of hnd-1 by ehn-1 and ehn-3. One idea is that ehn-1 and ehn-3 might encode, or control the activity of,transcription factors that cooperate with hnd-1 in the regulation of partially overlapping sets of target genes. Regardless of the molecular mechanism, the hnd-1/ehn genes clearly have overlapping, but non-equivalent, functions in embryonic development and gonadogenesis.

The hnd-1 gene encodes the single Hand transcription factor in the C. elegans genome (Ledent and Vervoort, 2001). Higher vertebrates contain two Hand genes(eHand/Hand1 and dHand/Hand2), whereas a single family member has been identified in zebrafish(Yelon et al., 2000),ascidians (Dehal et al., 2002)and flies (Moore et al.,2000). Vertebrate dHand is expressed in lateral plate mesoderm and is important for development of mesodermal organs, including heart and limbs (Firulli et al.,1998; Riley et al.,1998; Srivastava et al.,1997; Yelon et al.,2000). The Drosophila Hand gene is expressed in the dorsal vessel (heart) and visceral mesoderm, but its function is not known(Moore et al., 2000). The C. elegans Hand gene, hnd-1, is first expressed broadly in mesodermal precursors that generate striated muscles, and then is restricted to the somatic gonadal precursors; its function appears to affect both muscle and gonadal development. Therefore, all Hand genes explored to date are expressed in mesodermal cells and, where studied, are important for mesoderm development.

The defects in hnd-1 have intriguing similarity to the defects in the zebrafish Hand gene called hands off (han). Thus, han mutants generate the normal number of precardiac cells, but these cells cannot differentiate and a midline heart tube fails to form(Yelon et al., 2000). Similarly, SGPs are specified correctly in hnd-1 mutants, but they often fail to maintain their fate and can subsequently die. The fate of the cardiac precursors in zebrafish han mutants is not known(Yelon et al., 2000). We suggest that the zebrafish and nematode Hand genes may play parallel roles in controlling cardiac and gonadal precursor cells, respectively. Interestingly,zebrafish han may also be important for gonadogenesis: hanmutants have defects in migration of germ cells to the gonad as well as abnormalities in pax2.1 expression in the putative gonadal mesoderm(Weidinger et al., 2002). Therefore, zebrafish han mutants, like hnd-1 mutants, might have defects in development of the gonadal mesoderm. Our identification of hnd-1 as a regulator of somatic gonadal development in C. elegans raises the possibility that Hand genes are ancient regulators of gonadogenesis.

How does the control of SGP development by hnd-1/ehngenes compare to the genetic regulation of early gonadogenesis in other animals? Although genes have been identified that govern formation of the early gonad in both Drosophila and vertebrates, the genetic control of early gonadogenesis remains relatively uncharted territory. Perhaps most analogous to C. elegans hnd-1/ehn genes is Drosophila clift, which encodes a novel nuclear protein required for both SGP development in the gonad and for photoreceptor survival in the eye(Boyle et al., 1997). The clift effect on SGPs is remarkably similar to that of the hnd-1/ehn genes: SGPs are generated in cliftmutants, but they do not coalesce into a gonadal primordium and are lost over time (Boyle et al., 1997). Furthermore, ectopic clift expression, like ectopic hnd-1expression, did not increase SGP number(Boyle et al., 1997). Therefore, like hnd-1, clift may not be sufficient to direct SGP development on its own. In mice, several transcription factors have been implicated in development of the genital ridge, a mesodermal swelling destined to generate the somatic gonad (Birk et al.,2000; Capel, 2000). In Sf1 and Wt1 knockout mice, the genital ridge forms initially but it does not develop further; instead, the ridge regresses because of programmed cell death(Kreidberg et al., 1993; Luo et al., 1994). Therefore,although these genes all encode different transcription factors, the similarities in mutant phenotype suggest parallels in the genetic control of somatic gonadal precursors in flies, mammals and worms.

How does C. elegans gonadogenesis compare with the development of other organs? Some organs rely on `selector' genes, which regulate (directly or indirectly) all the genes needed to generate a particular organ. One simple example of an organ selector gene is C. elegans pha-4, which encodes a forkhead transcription factor that appears to regulate most, and perhaps all, pharyngeal genes (Gaudet and Mango,2002; Horner et al.,1998). Another example is that of the Drosophila Pax6/eyeless gene, which encodes a paired homeodomain transcription factor that is crucial for eye development(Quiring et al., 1994). Dramatically, Pax6/eyeless induces extra eyes when expressed ectopically (Halder et al.,1995). Our data suggest that hnd-1 does not fit into the organ selector model. Although loss of hnd-1 function can cause the complete loss of gonadal development, global expression of hnd-1 did not induce ectopic gonadal development. Therefore, gonad `identity' in C. elegans might rely on the coordinate regulation of several genes. Based on our analysis of hnd-1/ehn double mutants, the ehn genes represent likely candidates for additional regulators of the gonadal fate. Similarly, during Drosophila salivary gland development, several genes, including the winged-helix transcription factor encoded by fork head (fkh), are independently regulated and required for the development of specific salivary gland cell types (reviewed by Bradley et al., 2001). Interestingly, one aspect of fkh function is to inhibit apoptosis in the salivary gland primordia (Myat and Andrew, 2000). Therefore, like hnd-1, fkh acts after the salivary gland primordia are specified and is required for the survival of specific salivary gland cell types. We suggest that the role of hnd-1is to maintain the somatic gonadal fate and thereby prevent the death of the somatic gonadal precursors. It remains to be seen whether the maintenance of cell fate and cell survival are intimately linked during the development of other organs.

The authors thank Mike Krause, Ian Hope, Piali Sengupta and Susan Strome for strains, antibodies and reporter constructs. Additional strains were provided by the Caenorhabditis Genetics Center. We are grateful to Peggy Kroll-Conner for her contributions to the deletion screen and for meticulous strain maintenance. We thank Kellee Siegfried and Christopher Tilmann for critical reading of the manuscript, and two anonymous reviewers for helpful comments. L.D.M. was supported by an NIH postdoctoral fellowship and by the Howard Hughes Medical Institute. J.K. is an investigator with the Howard Hughes Medical Institute.