mab-3 is a direct tra-1 target gene regulating diverse aspects of C. elegans male sexual development and behavior

Sexual dimorphism is controlled by cascades of regulatory proteins that respond to a primary cue, either chromosomal or environmental, and coordinate the sexual differentiation of diverse tissues. The terminal global regulator in the C. elegans sex-determination cascade is encoded by the tra-1 gene (Hodgkin and Brenner, 1977; Hodgkin, 1987). tra-1 activity directs female somatic development, and also plays a major role in germline sex determination (Hodgkin, 1987; Schedl et al., 1989). The crucial role of tra-1 is illustrated by the effects of gain-of-function and loss-of-function tra-1 mutations, which can cause full sex reversal, irrespective of the activities of upstream regulators (Hodgkin 1980, 1987). Thus tra-1 can regulate, directly or indirectly, all downstream genes necessary for somatic sexual development. Identifying the genes tra-1 regulates and determining how they mediate sexual differentiation and behavior are crucial next steps in understanding how tra-1 controls sexual dimorphism.

tra-1 encodes a DNA-binding protein, TRA-1A, which contains zinc fingers closely related to those of ci and odd paired of Drosophila and the Gli genes of vertebrates (Zarkower and Hodgkin, 1992). Thus it is likely that TRA-1A controls sexual fate by activating transcription of female-specific genes and/or by repressing transcription of male-specific genes required for sexual differentiation. TRA-1A also has been proposed to regulate gene expression post-transcriptionally (Graves et al., 1999). tra-1 controls a wide array of sexually dimorphic features including cell lineages, cell deaths, migrations, morphogenesis and behavior (reviewed in Hodgkin 1988). Several genes have been identified that are required for sexual development, and any of these might be direct targets of TRA-1A regulation. So far, however, only one somatic gene, egl-1, has been reported to be directly regulated by TRA-1A (Conradt and Horvitz, 1999). TRA-1A represses egl-1 transcription in a pair of cells, the HSN neurons, preventing their death in hermaphrodites (Conradt and Horvitz, 1998, 1999). In the vast majority of somatic cells, however, the regulatory targets of TRA-1A and their functions are unknown.

One gene that appears to act downstream of tra-1, based on genetic epistasis analysis, is mab-3 (Shen and Hodgkin, 1988), which has at least two functions in male development. mab-3 represses vitellogenin (yolk protein) gene transcription in the male intestine, thereby acting as a direct regulator of sexual differentiation (Yi and Zarkower, 1999). In the nervous system, mab-3 promotes differentiation of a class of male sense organs of the peripheral nervous system called V rays. In mab-3 mutant males, the six bilateral pairs of V ray neuroblasts differentiate primarily as hypodermal cells rather than undergoing V ray differentiation to produce sensory neurons and support cells (Shen and Hodgkin, 1988).

mab-3 encodes a protein with two copies of a nonclassical ‘zinc finger’ DNA-binding motif called a DM domain (Raymond et al., 1998). The DM domain was first identified in the doublesex (dsx) gene, a downstream sexual regulator in Drosophila (Erdman and Burtis, 1993). As expected from its unusual sequence, the DM domain is structurally distinct from other zinc fingers, and it binds in the DNA minor groove (Zhu et al., 2000). In addition to the DM domain, mab-3 and dsx share several other characteristics, suggesting that they may be derived from an ancestral sex-determining gene. Both genes act downstream of the global regulators in their respective sex-determination cascades, controlling a subset of sexually dimorphic features (Baker and Ridge, 1980; Shen and Hodgkin, 1988). Both genes are direct transcriptional regulators of yolk protein genes (Coschigano and Wensink, 1993; Yi and Zarkower, 1999) and are required for differentiation of sex-specific sense organs (Baker and Ridge, 1980; Shen and Hodgkin, 1988). The male-specific isoform of DSX can substitute for MAB-3 in the C. elegans male peripheral nervous system, indicating that the two proteins are functionally very similar (Raymond et al., 1998). dsx appears to be more widely required than mab-3, regulating all external sexually dimorphic features (Baker and Ridge, 1980) and playing a role in the central nervous system that is essential for mating behavior (Villella and Hall, 1996). Based on their extensive similarities, mab-3 and dsx may represent the first example of evolutionary conservation between distantly related sex-determination pathways (Raymond et al., 1998). Recent work suggests that vertebrates use similar genes to control sexual development. Expression of the DM domain gene Dmrt1 in vertebrates with XX/XY, ZZ/ZW and environmental sex-determining mechanisms is consistent with a role in male sexual development (Raymond et al., 1998, 1999; Smith et al., 1999; De Grandi et al., 2000; Kettlewell et al., 2000; Moniot et al., 2000).

Here, we have investigated the role of mab-3 in regulating male sexual development by analyzing its expression, its regulation by the sex determination pathway, and its role in nervous system differentiation and function. While the function of mab-3 in sexual differentiation has been clearly established in the male intestine, where it acts as a transcriptional repressor of yolk protein genes (Yi and Zarkower, 1999), the other functions and regulation of mab-3 are poorly understood. We find that mab-3 is a direct target of TRA-1A transcriptional regulation. MAB-3 acts synergistically with the bHLH transcription factor LIN-32 to promote sensory ray neuroblast differentiation. We find that spatial control of sensory ray formation is accomplished by restricting the region of the lateral hypodermis that expresses both MAB-3 and LIN-32. We find that mab-3 has additional functions, and that, like doublesex, it is required for male-specific neuronal gene expression and male sexual behavior. These results demonstrate that mab-3 has diverse functions in male sexual development.

Culture and genetic manipulation of C. elegans were performed by standard methods as described previously (Sulston and Hodgkin, 1988). mab-3 mutants were of genotype mab-3;him-5(e1490).

Plasmids were coinjected with pRF4 at a concentration of 50-100 ng/μl. pRF4 contains the dominant rol-6(su1006) marker (Mello et al., 1991). Except where noted, at least three independent transgenic lines were analyzed in each experiment. To compare expression of srd-1::gfp and lov-1::gfp reporters in wild type and mab-3(null) males, heritable extrachromosomal arrays of the reporter and pRF4 were established in strains of genotype mab-3(null);him-5(e1490). Transgenic hermaphrodites from these strains were mated with him-5(e1490) males, and mab-3/+; him-5 Rol hermaphrodites from these crosses were self-fertilized to generate a mix of one quarter Mab (mab-3/mab-3) and three quarters wild-type (mab-3/+ and +/+) transgenic males that were examined for reporter gene expression. For srd-1::gfp, we made 10 lines in mab-3(e1240);him-5(e1490) and five lines in mab-3(mu2);him-5(e1490). None expressed srd-1 in the ray 9 neuron. We crossed two of the e1240 lines with him-5 males and tested F₂ self-progeny; in both cases, all of the non-Mab males expressed the reporter and none of the Mab siblings had detectable expression. For lov-1::gfp, we generated six transgenic lines in mab-3(e1240;him-5(e1490), all of which expressed the reporter in the tail but not in the CEM cells. Two lines were mated with him-5 males, and in both cases all non-Mab transgenic males expressed the reporter in the CEM cells and no Mab siblings had detectable expression.

For heatshock experiments, laid eggs were collected for 8-10 hours from hermaphrodites carrying extrachromosomal arrays of the relevant heatshock transgenes. Eggs were allowed to hatch at 20°C for approx. 12-16 hours and then raised at 20°C with 45 minutes heatshock treatments at 33°C every 12-14 hours until adulthood. The heterogeneous staging of animals and repeated heatshocks are intended to ensure that a significant proportion of animals receive treatment during the appropriate stages of development. Heatshock constructs contained the relevant cDNAs fused to the hsp16-41 promoter (Stringham et al., 1992) in pPD49.78 (gift from A. Fire). Plasmids were pDZ124 for MAB-3, EM#226 for LIN-32 (Zhao and Emmons, 1995) and pKM1034 for HLH-2 (Harfe et al., 1998).

For gel mobility shift experiments, probes were generated by PCR amplification with primers MAB PROF1 and WY52 at low dATP concentration (20 μM) and in the presence of 10 μCi [³²P] dATP. For the wild-type probe, the template was –1497pro::gfp (pDZ147) and for the 6 bp mutant probe (Mutant 1) the template was pDZ147M1. Labeled PCR products were purified on a non-denaturing 6% polyacrylamide gel. TRA-1A protein was generated by in vitro transcription and translation of a full-length tra-1 cDNA (pDZ118) using a T7-based coupled reticulocyte lysate (TNT Coupled Reticulocyte Lysate System, Promega). After in vitro translation of protein, ZnSO₄ was added to 50 μM. Gel mobility shift assays were performed as described (Pollock and Treisman, 1990), with incubation for 20 minutes at room temperature before electrophoresis on 4% acrylamide gels/0.5×TBE. Competitor DNAs were generated by PCR using primers MAB PROF1 and WY52 and relevant templates, pDZ147M2, pDZ147M3 or pDZ147M4 (mutations are detailed in Fig. 4). PCR products were eluted and purified from non-denaturing polyacrylamide gels. Competition experiments were performed with 10 fmol of ³²P-labeled PCR fragments as probe, and 0, 5, 50 or 500 ng of unlabeled PCR fragments as competitor. Mutant –1497pro::gfp plasmids were made by site-directed mutagenesis with the following primers (sequences follow): DZ1000 for Mutant 1; WY49 for Mutant 2; WY50 for Mutant 3; and WY51 for Mutant 4.

MAB PROF1: 5′ CGCAAGCTTCGCAGAGATCACACGATTC-GCGGA 3′WY52: 5′ CCCCTCCTGTGAAACGGGGCGGGTCCC 3′DZ1000: 5′ CTCTAATTATCGTCGTGCTGCAGCTTCTATCCAA-TCGC 3′WY49: 5′ CTCTAATTATCGTCGTGTGACATCTTCTATCCAATC-GC 3′WY50: 5′ CTCTAATTATCGTCGTGTGAGGCATTCTATCCAA-TCGC 3′WY51: 5′ CTCTAATTATCGTCGTGTTGGGTCTTCTATCCAATC-GC 3′

The rescuing mab-3::gfp reporter gene –1497pro-mab-3::gfp (pDZ162) contains mab-3 genomic sequences from 1497 bp 5′ of the site of splice leader SL1 addition (1508 bp 5′ to the MAB-3 start codon), fused in frame to the gfp coding region. These sequences were inserted as a KpnI/AscI PCR fragment into pPD117.01 (gift from A. Fire). In gene –1497pro-mab-3::gfp, the gfp stop codon is followed by a 1755 bp EcoRI/NheI PCR fragment containing mab-3 genomic sequences including the 3′ UTR and poly(A) addition signals, extending about 400 bp beyond the site of poly(A) addition (Raymond et al., 1998). Promoter-only reporters (e.g., –1497pro::gfp) included the indicated amounts of mab-3 genomic sequences upstream of the site of SL1 addition and the first four codons of mab-3, fused in frame to gfp in the vector pPD95.67 (gift from A. Fire).

The lin-32::gfp reporter plasmid, pLIN32GFP, contains lin-32 genomic sequences from 3.2 kb upstream of the initiation codon to the last codon, fused in frame to gfp in pPD95.67. Other reporter constructs used are srd-1::GFP (Troemel et al., 1995), plov-1::gfp1 (Barr and Sternberg, 1999) and ppkd-2::gfp1 (Barr and Sternberg, 1999).

mab-3 mutations cause defects in V ray formation and male intestinal differentiation, but, like dsx, mab-3 might play other sex-specific roles not yet identified. For example, mab-3 expression in the nervous system might regulate mating behavior. To identify such roles and to investigate how mab-3 expression is regulated, we first assayed the expression of reporter genes containing mab-3 promoter and coding sequences fused to a gfp cassette. We initially included 1497 bp of genomic DNA 5′ of the mab-3 AUG, as this amount is sufficient for phenotypic rescue of mab-3 mutants (Raymond et al., 1998). The reporter gene –1497pro-mab-3::gfp contains the mab-3 promoter from –1497 and the entire mab-3 genomic coding region, and is fused in frame to a gfp cassette (Fig. 1A). We also constructed a series of reporter genes containing only mab-3 promoter sequences and the first four codons of mab-3, fused to a gfp cassette. These are denoted –1497pro::gfp, −1266pro::gfp, and so on (Fig. 1B).

The –1497pro-mab-3::gfp reporter rescues both of the known mab-3 phenotypes: defective V ray formation in the male tail (Fig. 1C, top); and ectopic expression of yolk in the male intestine (Fig. 1C, bottom). Thus, its expression is likely to approximate that of the endogenous gene. Rescue of yolk expression was not complete, with yolk eventually accumulating in mutant adult males expressing MAB-3::GFP, and thus the fusion protein may be less active than native MAB-3. Early in development, –1497pro-mab-3::gfp was expressed in head neurons of late embryos and L1 larvae of both sexes (not shown). Later in larval development, starting in L3, it was expressed in the lateral hypodermis (seam) in both sexes (Fig. 2A-D), and male-specifically in the nervous system, both in the head and tail, including the sensory ray neuroblasts R1-R9 in the tail (Fig. 2C; data not shown). Initially there was some expression in the intestine of both sexes, but expression becomes male-specific by the adult stage, when vitellogenin transcription begins in hermaphrodites.

A reporter containing just the upstream mab-3 sequences fused to gfp (Fig. 1B. –1497pro::gfp) showed similar but distinct expression. Like −1497pro-mab-3::gfp, the −1497pro::gfp reporter was expressed in the larval intestine, with stronger expression in males, and became completely male-specific by the adult stage (Fig. 2E; data not shown). Expression of –1497pro::gfp was much stronger in the intestine than that of –1497pro-mab-3::gfp, possibly because the MAB-3::GFP fusion protein made by –1497pro-mab-3::gfp was less stable than GFP alone (Fig. 2E,F). –1497pro::gfp was also expressed in neurons in adult males, including, most prominently, the sensory neuron ADF in the head (Fig. 2E), V ray neurons (Fig. 2F) and one hook neuron (Fig. 2G). Unlike –1497pro-mab-3::gfp, −1497pro::gfp expression was not detectable in the lateral hypodermis or the early sensory ray lineages, suggesting that additional regulatory elements may lie within the mab-3 coding region. Since the promoter-only reporter was expressed sex-specifically in the intestine and in neurons of the head and tail, we conclude that the regulation of mab-3 expression is primarily transcriptional.

To identify regulatory sequences in the mab-3 promoter, we examined the expression of six reporters with deletions of sequences 5′ to the mab-3 coding region (Fig. 1B). Deleting mab-3 promoter sequences from −1266 to −1497 greatly reduced expression in the intestine without affecting expression in neurons of the head and tail (Fig. 1B; compare Fig. 2E with 2H, and Fig. 2F with 2I). Deletion beyond −566 eliminates neural expression (Fig. 1B). These data indicate that there are at least two regulatory elements in the mab-3 promoter: one required for intestinal expression and one required for neuronal expression.

The mutant phenotype of mab-3 and genetic epistasis in the intestine (Shen and Hodgkin, 1988) suggest that mab-3 acts downstream of tra-1. Consistent with this, −1497pro::gfp was expressed in the nervous system and tail of adult XO animals, in which tra-1 is inactive, but not in adult XX animals, in which tra-1 is active (Fig. 3A). In XX tra-1(null) mutants, −1497pro::gfp was expressed exactly as in wild-type XO males, suggesting that TRA-1A represses, directly or indirectly, the transcription of mab-3 in XX animals.

To determine whether the transcriptional regulation of mab-3 by TRA-1A is direct, we first searched for potential binding sites in the promoter of mab-3. These searches identified a single consensus TRA-1A binding site (Zarkower and Hodgkin, 1993) located at –1300, within the region required for intestinal mab-3 reporter expression. There are also several TRA-1A consensus binding sites located 5′ to –1497 (Clarke and Berg, 1998), but we have not investigated these further. The −1300 site is highly conserved between C. elegans and C. briggsae (Fig. 3B), suggesting that it may be functionally significant. Although the site in the C. briggsae mab-3 promoter is about 400 bp closer to the mab-3 coding region (Fig. 3B), conservation of flanking sequences between the two species (not shown), suggests that these elements are homologous. In a gel mobility shift assay with mab-3 promoter sequences from –1449 to –1213, TRA-1A bound to a wild-type fragment (Fig. 3C lane 2) but not to one in which the TRA-1A site was mutated at 6 positions (Fig. 3C, lane 3).

To test whether this TRA-1A binding site is required for mab-3 regulation in vivo, we assayed expression of a –1497pro::gfp reporter containing the same 6 bp TRA-1A site mutation described above. In contrast to the wild-type –1497pro::gfp reporter, which is expressed only in males (Fig. 3D, top panels), the mutant reporter was expressed in the intestine in both sexes (Fig. 3D, bottom panels). These data suggest that TRA-1A directly represses the transcription of mab-3 in the intestine. Regulation by tra-1 in the nervous system does not require this TRA-1A binding site, as the mutant reporter was expressed male specifically in neurons (data not shown). We have not identified other TRA-1A-binding sites within the −1497 promoter region, and most of the neurons that express mab-3 are present only in males. Thus, regulation of mab-3 expression in the nervous system by TRA-1A is likely to be indirect.

Deregulation of mab-3 reporter gene expression by mutation of the TRA-1A binding site strongly suggests that mab-3 transcription in the intestine is directly regulated by TRA-1A.

However, we also considered the possibility that a different factor binds to a site overlapping the TRA-1A site, and that this factor, rather than TRA-1A, regulates mab-3 transcription. To test this possibility, we made three additional mutant reporters, each with two base pairs altered in the TRA-1A site (Fig. 4). One mutant reporter is predicted to have severely reduced TRA-1A-binding affinity based on in vitro binding studies (Zarkower and Hodgkin, 1993), while the other two are predicted to retain TRA-1A binding. To compare the affinity of the mutant reporters for TRA-1A in vitro, we performed gel mobility shift assays using a labeled wild-type probe and either wild-type or mutant unlabeled competitor. The TRA-1A/DNA complex was efficiently competed by unlabeled wild-type probe and not by the 6 bp mutant (Mutant 1, Fig. 4A). As predicted, Mutant 2, which lacks a critical pair of G residues (Zarkower and Hodgkin, 1993), also did not compete efficiently for TRA-1 binding, while Mutants 3 and 4 competed nearly as well as wild type (Fig. 4A). In vivo, the effects of the 2 bp mutations on mab-3::gfp regulation paralleled their effects on DNA binding in vitro. Mutant 2, which binds TRA-1A poorly, was expressed in both sexes, while Mutants 3 and 4, which bind TRA-1A well, retained normal male-specific expression. Since the regulation of mab-3::gfp in vivo correlates precisely with the ability of TRA-1A to bind in vitro, we conclude that TRA-1A directly represses mab-3 transcription in the hermaphrodite intestine. In the male intestine, where TRA-1A is inactive, MAB-3 directly represses yolk protein gene transcription (Yi and Zarkower, 1999). Thus, in this tissue mab-3 serves as a direct molecular link between the terminal regulator of the global sex determination pathway, tra-1, and the sex-specific structural genes, the vitellogenins.

From the results described above and previous work (Yi and Zarkower, 1999), the role of mab-3 in sexual differentiation of the intestine is now fairly clear, but its role in sensory ray development is less clear. mab-3 activity is required for differentiation of the 6 pairs of V ray neuroblasts, R1-R6, acting downstream of genes that determine which region of the lateral hypodermis (the body seam) forms ray neuroblasts. These genes include the Hox gene mab-5 (Kenyon, 1986) and the basic helix-loop-helix (bHLH) transcription factor gene lin-22, a homolog of the Drosophila hairy gene (Wrischnik and Kenyon, 1997). As shown in Fig. 5, MAB-5 promotes the formation and differentiation of the V ray neuroblasts R1-R6 in the posterior lateral hypodermis, which is derived from the V5 and V6 cells (Kenyon, 1986). LIN-22 has the opposite role, preventing sensory ray formation in the anterior lateral hypodermis, which is derived from the V1-V4 cells (Wrischnik and Kenyon, 1997). Thus, the combined activities of MAB-5 and LIN-22 result in sensory ray formation only in the posterior lateral hypodermis, in the V5 and V6 lineages. Since mab-3 is required for differentiation of the ray neuroblasts, a simple model would be that these upstream regulators control ray formation by determining where mab-3 is expressed.

However, this is unlikely, since our reporter gene analysis indicates that mab-3 is expressed in the lateral hypodermis not only where rays form (V5 and V6 lineages), but also where they do not (V1-V4 lineages). Rather, it is likely that mab-3 acts together with another gene whose expression is restricted to the posterior body seam under the control of MAB-5 and LIN-22 (‘gene X’ in Fig. 5). We therefore investigated the relationship between mab-3 and other genes involved in sensory ray formation.

A good candidate for a gene acting with mab-3 in ray formation is lin-32, which encodes a bHLH transcription factor (Zhao and Emmons, 1995). Like mab-3, lin-32 is required for differentiation of the V ray neuroblasts. It also is required in the T ray neuroblasts and is involved in neurogenesis elsewhere in both sexes (Zhao and Emmons, 1995; Zhao, 1996; Emmons, 1999). To investigate the relationship of mab-3 and lin-32, we first compared the expression of gfp reporters for both genes in wild-type animals and in mutants with altered V ray formation. A reporter with 3.2 kb of lin-32 promoter and the two coding exons fused to gfp partially rescued V ray formation in a lin-32(lf) mutant (not shown). This reporter, denoted lin-32::gfp, was expressed in all 18 ray neuroblast (Rn) cells in L3 males (Fig. 6A) beginning slightly later than mab-3::gfp, but unlike mab-3::gfp, it is not expressed in the lateral hypodermis (compare with Fig. 2A,C). Thus, in wild-type males lin-32 expression only overlapped that of mab-3 in the cells that form rays, as predicted if lin-32 is ‘gene X’.

Next we tested whether mab-5 is required for expression of mab-3 or lin-32 in this region. In mab-5(lf) mutants, the posterior hypodermal cells V5 and V6 do not form ray neuroblasts, instead they form an ectopic body seam, and T rays form normally. In mab-5(lf) L3 males, the –1497pro-mab-3::gfp reporter is still expressed in the ectopic body seam and in the remaining T-derived ray neuroblast cells, R7-9. (Fig. 6B). This indicates that while mab-5 is required for the formation of V ray neuroblasts in the V5 and V6 cell lineages of the tail, it is not required for expression of mab-3 in those lineages. However, mab-5(lf) mutations do eliminate lin-32::gfp expression in the descendants of V5 and V6, without affecting expression in R7-9 (Fig. 6C). Thus, lin-32, but not mab-3, requires mab-5 activity for expression in the lineages that form V rays. This suggests that mab-5 directs V ray differentiation at least in part by positively regulating lin-32 expression.

Next we investigated whether lin-22 prevents anterior V ray formation by limiting the domain of lin-32 expression to the posterior. As predicted by the model, in lin-22(lf) L3 males, lin-32::gfp was expressed in the anterior lateral hypodermis as well as in the 18 ray neuroblast cells (Fig. 6D). This suggests that the ectopic rays in lin-22 mutants may result from ectopic lin-32 expression. Similar results with lin-32 have been obtained by Zhao (1996). From the comparison of mab-3 and lin-32 expression it appears that lin-22 and mab-5 specify the domain in which V rays form at least in part by defining the cells of the lateral hypodermis in which lin-32 is expressed with mab-3. The combined expression of mab-3 and lin-32 then acts to promote V ray differentiation.

mab-3 and lin-32 are both crucial for V ray formation, but two lines of evidence suggest that their functions are distinct, with lin-32 an essential determinant of sensory ray formation and mab-3 playing more of a supporting role. First, mab-3(null) mutations severely reduce the number of V rays formed, but do not eliminate ray formation completely (Shen and Hodgkin, 1988). In contrast, even non-null lin-32(lf) mutations can have more severe effects on ray formation (Zhao and Emmons, 1995). Second, as shown above, lin-32 expression, regulated positively by mab-5 and negatively by lin-22, is strictly correlated with ray formation, while mab-3 is expressed throughout the lateral hypodermis of both sexes. Together these results suggest that lin-32 plays an instructive role in ray formation, while mab-3 plays more of a permissive role, possibly acting to enhance the activity of lin-32.

To test the idea that mab-3 enhances lin-32 activity we asked whether the requirement for mab-3 in V ray formation can be overcome by elevated expression of LIN-32. We expressed cDNAs encoding MAB-3 and LIN-32 from a heatshock promoter and assayed rescue of the mab-3 null allele e1240. As expected, HS-MAB-3 restored V ray formation to mab-3(e1240) males, increasing the V ray number from 5% to 40% of wild type (Fig. 7A,B,E). HS-LIN-32 also suppressed mab-3(e1240), to 30% of the wild type V ray number (Fig. 7C and E). MAB-3 overexpression, however, did not suppress lin-32(lf), suggesting that the two genes do not have identical functions in the V ray lineage (Fig. 7E). We also tested HLH-2, a C. elegans E/daughterless homologue (Krause et al., 1997). HLH-2 is expressed in the V ray lineage and thus might serve as a dimerization partner for LIN-32 (J. M. R., data not shown; D. Portman and S. Emmons, personal communication). HS-HLH-2 can suppress mab-3(e1240), but less efficiently than HS-LIN-32 (15% of wild type), and produces rays with abnormal morphology (Fig. 7D,E). Collectively these data suggest a model in which LIN-32 is a crucial determinant of V ray formation and MAB-3 acts to enhance LIN-32 activity in specifying V ray neuroblast cell fate. Normally MAB-3 is required for V ray differentiation, but if sufficient LIN-32 activity is present, MAB-3 can become dispensable. lin-32 is probably not the only crucial target of mab-5 in the V ray lineages, however, as expression of LIN-32 and HLH-2 together did not restore V rays to mab-5 mutants (Fig. 7E).

Expression of mab-3 reporters in male neurons outside the sensory ray lineages suggests that mab-3 might have additional functions in the male nervous system. In the V ray lineages, mab-3 clearly is important for neuroblast differentiation. Other male sensory neurons that express mab-3, while present in mab-3 mutants, might not function normally. To test this possibility, we examined the expression in mab-3 mutant males of three genes with known or suspected roles in male mating behavior. The first gene, srd-1, encodes a putative chemosensory receptor protein that is expressed male-specifically in the ADF amphid neuron of the head and one of the ray 9 neurons of the tail, and in both sexes in the ASI amphid neuron (Troemel et al., 1995). Since mab-3 reporters also are expressed male-specifically in ADF and the T ray neuroblasts, the srd-1 gene is a good candidate to be regulated by mab-3. In mab-3(null) adult males, srd-1::gfp was expressed normally in ASI and ADF in the head (Fig. 8A), but no expression was detectable in the ray 9 neuron in the tail (Fig. 8B,C). srd-1::gfp was expressed in ray 9 in mab-5 mutant males (Fig. 8D), demonstrating that the loss of expression in this cell in mab-3 mutant males is not simply a consequence of disrupted V ray formation.

We also tested the expression of lov-1, a gene encoding a putative cell-surface protein required for normal male mating behavior (Barr and Sternberg, 1999), and its close homolog pkd-2. Both genes are expressed in the CEM neurons of the male head, and in the hook neuron HOB and ray neurons of the tail (Barr and Sternberg, 1999). pkd-2::gfp expression appeared normal except in rays 1-6, which are missing in mab-3(null) (not shown). However, lov-1::gfp expression was not detectable in the CEM cells of mab-3 mutant males (Fig. 8E,F). bv-1::gfp is expressed normally in R7-9 of mab-3 mutant males (Fig. 8G). Thus mab-3 is required in both the head and the tail of males for expression of neuronal genes.

Both the male-specific neuronal expression of mab-3 and the defects in lov-1::gfp and srd-1::gfp expression in mab-3 mutants suggest a potential requirement for MAB-3 activity in male-specific sensory neuron function. We therefore tested the ability of mab-3 mutant males to interact with hermaphrodites. The lack of V rays in mab-3 mutant males complicates this task, as V rays are crucial mediators of early steps in mating, and thus mab-3 mutant males are incapable of copulation for mechanical reasons. As a result, it is not possible to assess the ability of mab-3 males to perform the different sub-behaviors of copulation. Instead we focused on earlier steps of attraction to hermaphrodites.

Direct observation of mab-3 mutant males in the presence of hermaphrodites suggests a deficit in attraction to and interaction with hermaphrodites (not shown). To evaluate this deficit more quantitatively, we used the leaving assay of Lipton and Emmons (personal communication) to measure the interaction of males with paralyzed hermaphrodites. In this assay, him-5(e1490) (wild-type) males placed in the presence of paralyzed hermaphrodites on a small bacterial lawn remained on the lawn (probability of leaving, P_L=0.006/hr), while those plated with paralyzed males left at a much higher rate (P_L=0.040/hr) (Table 1). This difference in leaving rate between males plated with hermaphrodites and those plated with males suggests that wild type males detect hermaphrodite-specific cues that retain them on the bacterial lawn (Lipton and Emmons, unpublished). mab-3 mutant males are insensitive to these cues and failed to discriminate between the two sexes, leaving males (P_L=0.169/hr) and hermaphrodites (P_L=0.176/hr) at equivalent high rates. The rate at which mab-3 mutant males left worms of both sexes is higher than the rate at which wild-type males left males. This might indicate a general sensory defect unrelated to sexual attraction. However, since mab-3 males respond normally to volatile attractants and repellents, food, and mechanical stimuli (data not shown), a general sensory deficit is unlikely. Another possibility is that mab-3 males have defects not only in attraction to hermaphrodites, but also in detection of non-sex specific nematode cues.

To exclude the possibility that the failure of mab-3 males to interact with hermaphrodites is due to the absence of V rays, we determined leaving probabilities for mab-5 mutants, which also lack V rays and have other tail defects but have normal development of sensory neurons in the head. Although mab-5 males left hermaphrodites at a significantly higher rate (P_L=0.044/hr) than did wild-type males (P_L=0.006/hr), they did distinguish between males (P_L=0.089/hr) and hermaphrodites (P_L=0.044/hr). Because mab-5 males are unable to copulate but still are preferentially attracted to hermaphrodites, it is likely that the interaction has contact-independent components. The observation that mab-3 males did not respond to hermaphrodites under conditions in which mab-5 males were selectively retained indicates that sensory structures other than the V rays contribute to detection and response to the hermaphrodite cue and that mab-3 is required for this process. To determine whether loss of lov-1 expression might contribute to the defective interaction of mab-3(null) males with hermaphrodites, we tested lov-1(sy582Δ) males (Barr and Sternberg, 1999) in the leaving assay. lov-1(sy582Δ) males left paralyzed males (P_L=0.043/hr) and hermaphrodites (P_L=0.004/hr) at rates indistinguishable from wild type, indicating that lov-1 is dispensable for hermaphrodite detection in this assay. This result suggests that lov-1 carries out a MAB-3-regulated function separate from hermaphrodite detection or redundant with other targets of MAB-3 regulation.

Genes that are transcriptionally regulated by TRA-1A serve as a bridge between the global sex-determination cascade and the terminal effectors of sexual differentiation, and defining the molecular basis of these interactions is essential to a detailed understanding of sexual development. Other than mab-3, however, egl-1 is the only somatic target of TRA-1A regulation reported (Conradt and Horvitz, 1999). Regulation of these two genes hints at the varied strategies that may be used by tra-1 to control sexual development. egl-1 is a general apoptosis regulator whose transcription is directly repressed by TRA-1A in a single pair of cells, the HSNs, preventing their death in XX animals (Conradt and Horvitz, 1998, 1999). Similarly, TRA-1A regulates vitellogenin expression by direct transcriptional repression of mab-3 in the XX intestine (Yi and Zarkower, 1999). In other cells, such as the V ray and T ray neuroblasts and ADF, TRA-1A appears to regulate mab-3 expression indirectly, and in the lateral hypodermis, mab-3 expression is independent of TRA-1A. Both mab-3 and egl-1 are negatively regulated by TRA-1A, demonstrating that the protein is a transcriptional repressor. It is not known whether TRA-1A can both repress and activate transcription, like the related transcription factors CI and GLI (Johnson and Scott, 1998), a point that should eventually be clarified by the identification of additional direct TRA-1A targets.

Our analysis of the role of mab-3 in regulation of srd-1 and lov-1 expression indicates additional complexity in downstream sexual regulation. srd-1 expression in ray 9 in the tail is dependent on mab-3, but its expression in ADF in the head is not. However, since srd-1 expression in ADF is sexually dimorphic (Troemel et al., 1995), it must somehow be regulated by tra-1. This regulation might be via a downstream regulator other than mab-3, or alternatively TRA-1A might directly repress srd-1 in ADF. Likewise, expression of lov-1 in the CEM cells of the male head is dependent on mab-3, but expression in tail neurons is not. Expression of the close lov-1 homolog pkd-2 in the CEM cells is independent of mab-3, despite the fact that lov-1 and pkd-2 reporters have virtually identical expression patterns (Barr and Sternberg, 1999). Clearly tra-1 controls the expression of terminal products of sexual differentiation via a complex network of downstream regulators, and in at least some cases, tra-1 controls a given downstream target gene in different cells by very different means.

The identification of TRA-1A-binding sites in mab-3 and egl-1 suggests a general mechanism for TRA-1A repression of sexually dimorphic transcription. In mab-3 reporters, a TRA-1A site 1.3 kb upstream is necessary for transcriptional repression by TRA-1A in the intestine, and sequences outside the promoter region are not required for this regulation. A TRA-1A site essential for repression of egl-1 is instead located 5.6 kb downstream of the egl-1 transcriptional unit (Conradt and Horvitz, 1999). This demonstrates that TRA-1A can repress transcription through binding sites on either side of the coding region of genes and at (for C. elegans) a considerable distance. In both cases, there is significant conservation between C. elegans and the related nematode C. briggsae of sequences flanking the TRA-1A-binding site. These are sequences that are not required for TRA-1A binding, suggesting that other factors bind nearby. A simple model consistent with the available genetic and molecular data is that TRA-1A binding represses transcription by interference with transcriptional activators that bind to nearby regulatory sequences. Binding of TRA-1A might prevent either the association of these regulators with DNA or their activity once bound. This model also provides a means for TRA-1A, which is presumed to be expressed in most or all cells, to control the expression of a given gene in one cell or group of cells but not elsewhere. If a target gene contains multiple tissue-specific enhancer elements, TRA-1A will only repress transcription directed by those elements adjacent to TRA-1A binding sites. TRA-1A may also employ other repression mechanisms, such as interaction with transcriptional co-repressor complexes, but this has not been tested.

mab-3, in concert with other genes including lin-32, is required for V ray differentiation in the male tail. Genetic epistasis analysis and comparison of mutant phenotypes (Shen and Hodgkin, 1988; Zhao and Emmons, 1995; Wrischnik and Kenyon, 1997) indicate that mab-3 and lin-32 act later in ray neuroblast differentiation than mab-5 and lin-22, but whether their expression is regulated by these genes and how they interact functionally has been unclear. Using reporter genes, we found that lin-32 expression was regulated positively by mab-5 and negatively by lin-22, while mab-3 expression, in contrast, appeared to be independent of these genes. Thus, the crucial determinant of where V rays form appears to be lin-32 rather than mab-3.

Several lines of evidence suggest that mab-3 acts to enhance the activity of lin-32 to promote ray formation. First, mab-3(null) mutant males, while severely defective in V ray formation, do produce a small number of V rays, and thus mab-3 is not absolutely essential for ray formation. Second, lin-22 mutations caused the ectopic expression of lin-32 in the anterior lateral hypodermis (Fig. 6), but this causes ectopic ray formation only if mab-3 is also present (Shen and Hodgkin, 1988; Wrischnik and Kenyon, 1997). Third, we found that ectopic expression of LIN-32 could restore V ray formation to mab-3(null) mutants. This result must be interpreted with caution as it involves overexpression, but it suggests that mab-3 is dispensable for V ray formation if sufficient LIN-32 is present. The reciprocal is not the case: MAB-3 overexpression did not suppress ray defects in lin-32 mutants. This result argues against models in which mab-3 and lin-32 perform the same function in ray formation. In such models the total activity of MAB-3+LIN-32, rather than the activity of one protein or the other, is crucial for ray formation. Ectopic HLH-2 expression also restored V ray formation to mab-3 mutants, but less efficiently, perhaps by increasing the concentration of a complex with LIN-32.

Our results are most consistent with a model in which mab-3 plays a permissive role in V ray formation in concert with lin-32. In wild-type males, mab-5 directly or indirectly activates lin-32 expression only in the V5- and V6-derived neuroblasts R1-R6. The combined expression of mab-3 and lin-32 in R1-R6 results in their differentiation into V rays. In the anterior body seam (V1-V4 lineages), mab-3 is expressed but lin-32 is not, because it is repressed by lin-22, and this prevents sensory ray formation. The repression of lin-32 by lin-22 could be direct, or it may be mediated by mab-5, as lin-22 mutants ectopically express mab-5 in the V1-V4 lineages (Wrischnik and Kenyon, 1997). Co-expression of mab-3 is necessary for full lin-32 activity, but this requirement can be bypassed by elevating the level of LIN-32 expression. This model predicts that ectopic LIN-32 expression in the anterior body seam should result in ectopic rays, which has been shown to be the case (Zhao and Emmons, 1995). The regulation of lin-32, an achaete-scute homolog, by lin-22, a hairy homologue, suggests that the regulatory relationship of these genes may be conserved between flies and worms (Skeath and Carroll, 1991; Orenic et al., 1993; Wrischnik and Kenyon 1997)

mab-3 might potentiate the activity of lin-32 by any of several mechanisms, which are not mutually exclusive. One possibility is that MAB-3 and LIN-32 physically interact to generate a more active form of LIN-32. A second possibility is that MAB-3 regulates a gene that affects the activity of LIN-32. It could repress an inhibitor of LIN-32 or activate an enhancer of LIN-32 activity. A third possibility is that MAB-3 and LIN-32 may regulate some of the same downstream targets, and that more LIN-32 is required to achieve proper regulation of these genes when MAB-3 is absent. Mechanistic studies and searches for regulatory targets of MAB-3 and LIN-32 should help address these possibilities.

An intriguing finding of the work reported here is that mab-3 reporters were expressed in a number of sensory neurons in the male head and the tail whose formation is not prevented by mab-3 mutations. All of these cells are good candidates for mediating male mating behavior, and indeed some have been shown to be required for specific aspects of male mating (Ward et al., 1975; Liu and Sternberg, 1995). This raises the possibility that mab-3 plays additional behavioral roles in the male nervous system.

We tested the possible role of mab-3 in male mating behavior using two approaches. First, we investigated whether mab-3 is required for the expression of genes implicated in male mating behavior. Of three genes assayed, mab-3 was required for normal expression of two: lov-1 in the head and srd-1 in the tail. Second, we investigated whether mab-3 males exhibit defective interaction with hermaphrodites. While wild-type males show a strong preference for hermaphrodites over males, we found that mab-3 males were not attracted to either sex and rapidly left. This defect cannot result entirely from lack of V rays, as mab-5 mutants still show a preference for hermaphrodites over males. Taken together, these results strongly suggest that mab-3 is required in the nervous system for expression of genes that mediate early, and perhaps also later, steps of male mating behavior. In this regard, mab-3 further resembles doublesex, which is required for male courtship behavior in Drosophila (Villella and Hall, 1996). Additional assays will be needed to distinguish whether mab-3 mutant males are defective in taxis to hermaphrodites, sustained interaction with hermaphrodites once located, or both. It also will be important to determine in which cells mab-3 is required for which aspects of male behavior. The finding that mab-3, like dsx, is required for male mating behavior further suggests that these two genes may be conserved from an ancient sexual regulator.

It is now clear that tra-1 coordinates sexual development and behavior via a group of downstream regulatory genes including egl-1 and mab-3. These genes provide an interface between the global sex-determination pathway, with tra-1 at its terminus, and the expression of the genes responsible for terminal differentiation and function of sexually dimorphic cells throughout the animal. mab-3 serves as a direct link between tra-1 and terminal differentiation in the intestine, and as an indirect link in the nervous system, playing key roles in both the formation and the function of male neurons. Even by regulating both mab-3 and egl-1 expression, tra-1 directs the sexually dimorphic development of only a small proportion of cells. An important goal for the future will be to identify the genes that link tra-1 to sexually dimorphic development elsewhere in the animal.

We thank members of the Zarkower laboratory and the University of Minnesota Center for Developmental Biology for many useful discussions. We thank Andrew Fire, Cori Bargmann, Maureen Barr, Michael Krause and Cynthia Kenyon for strains, plasmids and reagents; and Jonathan Hodgkin, Marc Bickle, Scott Emmons, Douglas Portman and Jonathan Lipton for sharing results prior to publication. We thank Electra Coucouvanis, Jeff Simon, Scott Emmons, Douglas Portman, Jonathan Lipton and Vivian Bardwell for critical reading of the manuscript. Some C. elegans strains were obtained from the Caenorhabditis Genetics Center, which is funded by a grant from the NIH National Center for Research Resources. This work was supported by grants from the NIH (D. Z.), the Minnesota Medical Foundation (D. Z.), the STAGE program of the University of

Minnesota (W. Y.), and the University of Minnesota Graduate School (D. Z.).