Similarity of DNA binding and transcriptional regulation by Caenorhabditis elegans MAB-3 and Drosophila melanogaster DSX suggests conservation of sex determining mechanisms

Sexual differentiation is controlled by cascades of regulatory genes. In nematodes and insects, where they have been most extensively characterized, these can be divided into hierarchies composed of three classes of genes (reviewed in Cline and Meyer, 1996). The earliest acting class of genes are coordinately acting switch genes that control both sexual differentiation and dosage compensation, the process by which expression of genes on the sex chromosomes is equalized. Examples of such coordinate regulators include sdc-2 in Caenorhabditis elegans and sex-lethal (Sxl) in Drosophila melanogaster (Cline, 1984; Nusbaum and Meyer, 1989). These coordinate regulators control the activity of a second class of switch genes that function exclusively in sex determination and affect all sexually dimorphic somatic features. Mutations in these genes, for example tra-1 in C. elegans or transformer (tra) in Drosophila, can lead to full sex reversal (Sturtevant, 1945; Hodgkin, 1987). The first two classes of globally acting sex determination genes control the activity of a third class of genes, downstream regulators that direct more limited aspects of sexual development. Examples of downstream sexual regulators include mab-3 in C. elegans, and fruitless (fru), dissatisfaction (dsf) and doublesex (dsx) in Drosophila (Baker and Ridge, 1980; Shen and Hodgkin, 1988; Ryner et al., 1996; Finley et al., 1997). Rather than causing complete sex-reversal, mutations in these downstream sexual regulators can cause incorrect sexual differentiation in specific tissues in one or both sexes.

Sex determination, unlike most major developmental processes, is notable for the lack of evolutionary conservation at the molecular level. For example, while alternative splicing plays a central role in Drosophila sex determination, there is no indication that this process regulates sex determination in C. elegans or mammals (Cline and Meyer, 1996; Schafer and Goodfellow, 1996). Indeed, the only molecular similarity in sex-determining genes that has been observed between phyla is between the mab-3 gene of C. elegans and the dsx gene of Drosophila (Raymond et al., 1998). These two genes occupy analogous positions in their respective sex determination genetic pathways. Both are downstream regulators that are controlled by the cascade of globally acting switch genes and are required for a subset of sexually dimorphic features. Mammals may use similar genes in sex determination: a gene related to dsx and mab-3 is expressed exclusively in human testis and maps to 9p24.3, a chromosomal region required in two copies for testis development (Raymond et al., 1998).

mab-3 and dsx perform similar roles in sexual differentiation. mab-3 functions in two male tissues. In the peripheral nervous system of the male tail it directs the development of sensory rays (V rays) that are required for mating, and in the male intestine it prevents expression of vitellogenins (yolk proteins) (Shen and Hodgkin, 1988). dsx is alternatively spliced to encode male- and female-specific proteins that act in each sex to drive development away from an intersexual state (Burtis and Baker, 1989). dsx controls a wider range of sex-specific features than mab-3, but these include sensory bristle formation in the peripheral nervous system and regulation of yolk protein expression in the fat body, features analogous to those controlled by mab-3 (Baker and Ridge, 1980).

The proteins encoded by mab-3 and dsx are structurally related, as both contain a DNA binding motif called the DM domain (Erdman and Burtis, 1993; Raymond et al., 1998). The DM domain chelates zinc and contains conserved cysteine and histidine residues, but it is otherwise distinct from classical ‘zinc fingers’ (Erdman et al., 1996). The DM domain was identified in DSX as the minimal DNA binding element of the protein, and it is required for dsx function in vivo (Erdman and Burtis, 1993). DSX has one DM domain, while MAB-3 has two. Missense mutations in either of the MAB-3 DM domains eliminate mab-3 activity, demonstrating that both domains are essential in vivo (Shen and Hodgkin, 1988; Raymond et al., 1998). MAB-3 and DSX are also functionally similar: the protein encoded by the male splice form of the dsx mRNA, DSX^M, can replace MAB-3 in the nematode nervous system, restoring the formation of sensory rays to mab-3 mutant males (Raymond et al., 1998).

The two sexes of C. elegans and Drosophila are highly dimorphic. For example, in C. elegans about 30% of hermaphrodite cells and 40% of male cells are overtly sexually specialized (reviewed in Hodgkin, 1988). From these differences it is evident that the sex determination regulatory pathways must control the sex-specific activity of many structural genes. However, the only structural genes known to be directly regulated by a sex determination pathway are the Drosophila yolk protein (yp) genes. Transcription of the yp genes in the fat body is regulated directly by dsx (reviewed by Bownes, 1994). DSX^F acts as an activator of yp transcription, while DSX^M represses yp transcription (Coschigano and Wensink, 1993; An and Wensink, 1995a,b; Li and Baker, 1998). The targets of dsx regulation in other tissues are unknown. Conserved positive regulatory elements, called VPE1 and VPE2 sequences, have been identified in the C. elegans vitellogenin promoters (MacMorris et al., 1994). However, the regulatory factors that act via these elements have not been identified, and the control of tissue-, stage- and sex-specific expression of the vitellogenin genes is poorly understood.

Here we investigate the conservation of sex-determining mechanisms by examining the DNA binding and transcriptional regulatory properties of MAB-3. We find that despite containing different numbers of DM domains (two versus one), MAB-3 and DSX bind to similar DNA sequences. The vitellogenin gene promoters in C. elegans contain potential MAB-3 binding sites upstream of the start site of transcription. We find that mutating the MAB-3 binding site in the vit-2 promoter abolishes both binding of MAB-3 in vitro and sex-specific transcriptional regulation of a vit-2 reporter gene in vivo, suggesting that MAB-3 acts as a direct repressor of vitellogenin transcription. Thus DSX in insects and MAB-3 in nematodes encode DM domain transcription factors that directly regulate expression of yolk protein genes by binding to similar DNA sites, suggesting that at least this aspect of sex determination has been conserved in evolution.

Culture and genetic manipulation of C. elegans were performed by standard methods as described previously (Sulston and Hodgkin, 1988). mab-3 mutants were of phenotype mab-3;him-5(e1490). The mab-3 alleles used, e1240 and mu15, are null alleles (Shen and Hodgkin, 1988; Raymond et al., 1998). The him-5 mutation increases the frequency of X chromosome nondisjunction, and thus the proportion of males in a population (Hodgkin et al., 1977).

Plasmids for in vitro transcription are based on pT7NTagplink, which contains a T7 RNA polymerase promoter, the 5′ untranslated region of the human beta globin gene, and a c-myc epitope tag (Bardwell and Treisman, 1994). The MAB-3 expression plasmid pDZ130 contains the MAB-3 coding region as a EcoRI/XbaI fragment generated by PCR from C. elegans cDNA. The resulting protein has a c-myc epitope tag (MEQKLISEEDLNM) followed by the sequence AGSEF and the full MAB-3 coding region, minus the initiation methionine. The protein coding sequences in pDZ130 were confirmed by DNA sequencing. The DSX^F expression vector pDZ134 was made by inserting a EcoRI(filled)/SalI fragment of the DSX^F cDNA (gift of K. Burtis) into BamHI(filled)/XhoI cut pT7plink. The resulting protein contains the c-myc epitope tag followed by the sequence GSIPSEAGI and the full DSX^F coding sequence. MAB-3 and DSX proteins were generated by coupled transcription/translation in vitro with the TNT T7 coupled reticulocyte lysate system (Promega) according to the manufacturer’s instructions.

The vit-2::GFP reporter pCR2 was made by inserting a 366 bp XbaI/SacI(filled) fragment containing the 247 bp minimal vit-2 promoter from pFPO-gfp-V3′(FUB1) (gift of M. MacMorris and T. Blumenthal) into XbaI/SmaI cut pPD95.70 (gift of A. Fire). The resulting plasmid has the minimal 247 bp vit-2 promoter followed by the first 14 amino acids of vit-2, a synthetic intron and the GFP coding sequence. Mutant reporters M1 and M2 were generated from the wild-type vit-2::GFP reporter pCR2 using the GeneEditor^TM mutagenesis system (Promega). Both reporters are highly expressed in hermaphrodites (not shown), confirming that vit-2 promoter strength has not been impaired by the mutations. Mutagenic primers were the following: WY22, 5′ GGCTCTCACCGAATGGTTCAATTTGTTT-CTGAT 3′ (for vit-2 M1), WY42, 5′ CAAAATTAATAGAC-AGGGCTCTCACCGAAGAGTTAAATTTGTTTCTGATAAGGGT-CACAAAGCG 3′ (for vit-2 M2).

Full sequences of double stranded oligonucleotides used for DNA binding experiments in Fig. 3 were as follows (one strand is shown): Probe A: 5′ AATTCACAACTACAATGTTGCAATCAGCTAGCC 3′; Probe B: 5′ AATTCTCTCACCGAATGTTGCAATTTGTTTCTC 3′; Probe C: 5′ AATTCTCTCACACAATGTTGCAATTTGTTTCTC 3′; Probe D: 5′ AATTCTCTCACCGAATGTTGCGATTTGTTTCTC 3′; Probe E: 5′ AATTCTCTCACCGAATGGTTCAATTTGTTTCTC 3′; Probe F: 5′ AATTCTCTCACCGAATATTGCAATTTGTTTCTC 3′.

Transgenic nematodes were generated by standard methods (Mello et al., 1991), by coinjecting the reporter plasmid and the dominant transformation marker pRF4 (containing the mutant rol-6 allele su1006) each at 100 ng/μl. Reporters were analyzed in lines containing heritable extrachromosomal DNA arrays, and each construct was assayed in at least five independent transgenic lines.

Binding site selection from random oligonucleotides was performed as described previously (Pollock and Treisman, 1990) except for the following changes. After in vitro translation of proteins, ZnSO₄ was added to 50 mM. For gel-mobility shift assays in Fig. 3, probes were made by annealing oligonucleotides and filling overhanging ends with Klenow DNA polymerase in the presence of [α-³²P]dATP, and unincorporated nucleotides were removed by chromatography through Sephadex G-50 resin. For other gel-mobility shift assays, probes were labeled with [α-³²P]dATP during PCR amplification as described (Pollock and Treisman, 1990). DSX^F was used in all experiments; DSX^F and DSX^M have identical DNA binding properties (Erdman and Burtis, 1993; Cho and Wensink, 1996, 1997; Erdman et al., 1996).

Cloned oligonucleotides from site selections were sequenced using the Thermosequenase labeled terminator cycle sequencing kit (Amersham Life Science) according to the manufacturer’s instructions. Sequences were aligned manually to derive consensus binding sequences. The potential MAB-3 binding sites in the vitellogenin promoters were identified using the MacVector program, allowing 4-bp mismatches. C. briggsae vit gene promoter sequences are described by Zucker-Aprison and Blumenthal (1989).

The C. briggsae mab-3 gene was identified by searching the C. briggsae DNA database with WU-BLAST 2.0 (http://www.isrec.isb-sib.ch/software/WUBLAST_form.html) using the C. elegans mab-3 sequence. The C. briggsae MAB-3 sequence was predicted by conceptual splicing of the genomic sequence with the splice sites used by the C. elegans gene, all of which are conserved. Database searches of the Wormpep database (http://www.sanger.ac.uk/Projects/C_elegans/blast_server.shtml) with the BLASTP program identified the following ten apparent DM domain-containing sequences in C. elegans: Y53C12B.5a (mab-3), F10C1.5, C34D1.1, C34D1.2, T22H9.4 (two DM domains), Y43F8A.G, F13G11.1, C27C12.6, Y67D8A_390.A and K08B12.2.

Vitellogenins are synthesized in the adult hermaphrodite intestine, exported into the body cavity, taken up by the ovary, and incorporated into maturing oocytes (Kimble and Sharrock, 1983). A short (247 bp) segment of the vit-2 promoter is sufficient to recapitulate the tissue-, stage- and sex-specific expression of the intact vit-2 gene, suggesting that the primary control of vitellogenin expression is by transcriptional regulation (MacMorris et al., 1992). Loss-of-function mutations in mab-3 abrogate the sexual regulation of vitellogenin expression, causing the vitellogenins to be expressed abundantly in the adult male intestine (Shen and Hodgkin, 1988).

To determine whether the deregulation of vitellogenin expression in mab-3 mutants results from a failure of transcriptional control, we tested the expression of a transgene containing 247 bp of vit-2 promoter sequence fused to the green fluorescent protein (GFP) coding region. The vit-2::GFP reporter is expressed in wild-type hermaphrodites, but not in wild-type males (Fig. 1, top panels). In contrast, mab-3 mutants express the vit-2::GFP reporter at high levels in both sexes (Fig. 1, bottom panels). Eliminating mab-3 activity abolishes the sex-specificity of vit promoter activity, but stage-specificity and tissue-specificity are unaffected (Fig. 1, and data not shown). The misexpression of vit-2::GFP in mab-3 mutant males suggests that mab-3 regulates vitellogenin transcription, either directly or indirectly. To test whether the regulation is direct, we first investigated the DNA binding specificity of MAB-3.

We found previously that the male-specific isoform of DSX, DSX^M, can restore V rays to mab-3 mutant males almost as well as MAB-3 (Raymond et al., 1998). This suggests that, despite containing different numbers of DM domains, DSX and MAB-3 can bind to similar DNA sequences in vivo and can regulate some of the same genes. To test this possibility more directly, we determined the preferred MAB-3 in vitro DNA binding site. As a control, we also determined the preferred binding site of DSX (Erdman et al., 1996).

To identify high-affinity DNA binding sites of MAB-3 and DSX, we used a selection method in which oligonucleotides containing functional binding sites are isolated from a large pool of random oligonucleotides by immunoprecipitation of protein/DNA complexes (Pollock and Treisman, 1990). The proteins used for selection were made by in vitro translation and contained a short amino-terminal c-myc sequence recognized by the 9E10 monoclonal antibody. Four rounds of DNA binding and immunoprecipitation with 9E10 were performed with each protein. Pools of oligonucleotides from the fourth round of selection were tested for binding in gel-mobility shift assays, and the shifted oligonucleotides were cloned and sequenced. DNA sequences of the selected oligonucleotides were compared to identify potential binding site preferences.

The binding site consensus sequences derived from the selected oligonucleotides are summarized in Fig. 2. 63 of 65 sequenced oligonucleotides bound by MAB-3 contained a close match to the 13-nucleotide consensus AATGTTGCGA(T/A)NT, where N is any nucleotide. Most positions showed a very strong preference for one or two bases. The preferences at positions 3-8 are particularly strong, suggesting that these positions may be critical for high-affinity binding. All of 58 sequenced oligonucleotides bound by DSX contained a close match to the 13-nucleotide consensus (nT)(nG)(T/A)ACAATGT(A/T)(nC)C (nT = not T). The central nine base pairs form a punctuated palindrome around the central base pair, consistent with binding of DSX as a dimer to two half sites (Erdman et al., 1996). Our DSX site selection used full-length protein made by in vitro translation, but the consensus derived here is very similar to the 13-nucleotide consensus (G/A)NNAC(A/T)A(T/A)GTNN(C/T) reported previously for the bacterially expressed DM domain of DSX using an affinity chromatography method (Erdman et al., 1996). The most strongly preferred core nucleotides (ACAATGT) in the DSX site selected here and the previously reported site are identical. There are minor differences in flanking nucleotide preferences, possibly due to the use of full-length versus truncated protein or to differences in stringency between the two selection methods.

Although MAB-3 has two DM domains and DSX has only one, the DNA binding sites selected by the two proteins do bear some similarity. MAB-3 selected an asymmetrical site, while DSX prefers an inverted repeat (Erdman et al., 1996), but the sites selected by the two proteins are similar at eight base pairs (Fig. 3A). As a further test of the similarity of the MAB-3 and DSX binding sites, we assayed binding to several short double stranded oligonucleotides (Fig. 3B). These included a DSX binding site from the yp1 promoter (probe A), a potential MAB-3 site from the vit-2 promoter (probe B, see below), and several oligonucleotides based on the binding site consensus sequences described above (probes C-F).

A DSX binding site from the Drosophila yp1 gene (DSX A site) (Coschigano and Wensink, 1993) contains both MAB-3 and DSX consensus binding sequences, and both proteins bind efficiently (Fig. 3C, lanes 1 and 2). The C. elegans vit-2 promoter contains a potential MAB-3 binding site (see below), and this sequence is bound efficiently by MAB-3 (lane 3). However, the site in vit-2 lacks an AC dinucleotide that is preferred by DSX, and is bound inefficiently by DSX (lane 4). As predicted by the consensus sequences, adding the AC dinucleotide results in efficient binding by both proteins (lanes 5 and 6). The potential MAB-3 site in the vit-2 promoter differs at position 9 from an ideal MAB-3 consensus site (containing an A, rather than the preferred G), but optimizing this position to G has little effect (lane 7). This is consistent with the relatively weak preference for a G at that position (Fig. 2). At other positions a particular nucleotide is absolutely preferred, and changing one or two of the nucleotides at those positions can severely reduce binding by MAB-3 (Fig. 3C, lanes 9 to 12). Taken together, these results confirm that the binding site selections provided optimal consensus DNA binding sites, as well as accurate indications of which positions are most important for efficient binding in vitro. They also confirm that MAB-3 and DSX can bind to very similar short DNA sequences.

To investigate the role of MAB-3 in regulating vitellogenin transcription, we searched for potential MAB-3 binding sites in the proximal promoters of the C. elegans vit genes. In all six vit genes, possible MAB-3 binding sites occur within the proximal 350 bp of the promoter (Fig. 4). Five of the vit genes have single proximal MAB-3 consensus binding sites, while vit-6 has four sites. These sites range from 8/13 to 12/13 bp matches to the in vitro consensus DNA binding site (Materials and methods). Most of the sites do not contain a G residue at position 9, but as shown above this position is not critical for DNA binding (Fig. 3). The most proximal sites in vit-3, vit-4 and vit-5 are poorer matches to the consensus than those in vit-1, vit-2 and vit-6. However, vit-3, vit-4, vit-5 and vit-6 contain additional sequences further upstream that more closely match the MAB-3 binding site consensus (Fig. 4B). Since MAB-3 is conserved between C. elegans and the related nematode C. briggsae (see below), functional MAB-3 binding sites might also be expected to be evolutionarily conserved between these two nematodes. Complete promoter sequences for the all of the C. briggsae vit genes are not available, so a full comparison is not currently possible. Nevertheless, the C. briggsae vit-2 and vit-6 homologs have potential MAB-3 binding sites in positions and orientations equivalent to those in C. elegans (Fig. 4B). In summary, searches for MAB-3 binding sites reveal that all of the C. elegans vit genes are candidates for direct transcriptional regulation by MAB-3.

We next tested the functional importance of the MAB-3 binding site in the vit-2 promoter. The vit-2 promoter was chosen because it is the best characterized of the vit promoters and contains only one potential MAB-3 binding site near the transcriptional start site. MAB-3 binds the minimal 247 bp vit-2 promoter in vitro in a gel-mobility shift assay (Fig. 5, lanes 1 and 2). To test whether MAB-3 binds the site identified by sequence analysis, we made two mutant promoters, vit-2 M1 and vit-2 M2, which have two and five nucleotide changes in the potential MAB-3 binding site, respectively. MAB-3 does not bind to either of the mutant promoters (Fig. 5, lanes 3 and 4), demonstrating that the site is required for MAB-3 binding in vitro.

To test the in vivo relevance of the MAB-3 binding site in the vit-2 promoter, we constructed vit-2::GFP reporters containing the two mutant vit-2 promoters and assayed their expression in transgenic worms relative to the wild-type promoter. Wild-type males do not express the wild-type vit-2 reporter (Fig. 1). In contrast, both of the mutant reporters are expressed in the wild-type male intestine (Fig. 6). The M1 reporter is expressed at relatively low levels in males, mainly in the posterior intestinal cells, but at high levels in hermaphrodites, suggesting that it may retain some MAB-3 binding in vivo (Fig. 6, top panel, and data not shown). The M2 reporter, with five nucleotide changes, is highly expressed in males throughout the intestine (Fig. 6, bottom panel). Neither mutation affects the tissue specificity or stage specificity of vit-2 reporter gene expression (not shown). Disrupting the MAB-3 binding site in the minimal vit-2 promoter results in a loss of sex specificity of vit-2 transcription, the same effect as eliminating mab-3 function. We conclude that MAB-3 is a direct transcriptional repressor of vitellogenins in the male intestine.

Sex determination pathways appear to evolve rapidly compared to other major developmental regulatory pathways, and there is some evidence that upstream sex-determining genes may evolve more rapidly than downstream genes (reviewed in Marin and Baker, 1998). The conservation between nematodes and insects of MAB-3 and DSX, not only in sequence, but also DNA binding and function, suggests that these two sex-determining genes are ancient (Raymond et al., 1998). If so, then mab-3 might be expected to be more highly conserved between nematode species than the upstream sex-determining genes. In addition, by comparing related MAB-3 genes it might be possible to identify conserved functional domains of the proteins.

We compared the C. elegans MAB-3 protein to that of the related nematode C. briggsae. Fig. 7 shows an alignment of the C. elegans protein sequence, Ce-MAB-3, predicted from the mab-3 cDNA sequence, with the C. briggsae protein sequence, Cb-MAB-3, predicted from a genomic DNA sequence identified by database searching. C. elegans contains ten predicted DM domain-containing genes (Materials and methods), but only mab-3 and one other gene (T22H9.4) have two DM domains, and there is little conservation between mab-3 and the other C. elegans DM domain genes outside the DM domains (Raymond et al., 1998; W. Y. and C. Raymond, data not shown). From the presence of two highly similar DM domains, the similarity of other regions of the protein, and the apparent conservation of all splice sites between the two genes (data not shown), we conclude that the C. briggsae gene we identified is the homolog of C. elegans mab-3.

Overall the two predicted MAB-3 proteins are 64% identical. Several regions of the proteins are particularly similar. These include the two DM domains, which are 91% identical between Ce-MAB-3 and Cb-MAB-3, as well as two other regions, one of which is proline- and serine-rich (Fig. 7). These domains may have conserved functions such as mediating transcriptional repression or protein-protein interaction. The DSX proteins also contain a proline- and serine-rich region (Burtis and Baker, 1989), raising the possibility that these domains are a more widely conserved feature. MAB-3 is more similar between these two nematode species than are the upstream sexual regulators TRA-1 and TRA-2, which are 44% and 43% identical, respectively (de Bono and Hodgkin, 1996; Kuwabara, 1996). MAB-3 also is slightly more conserved between these nematode species than the upstream regulator FEM-2, which has a region that is 63% identical between species and an amino-terminal domain that is present only in the C. briggsae protein (Hansen and Pilgrim, 1998). Thus our data support the view that downstream sex-determining genes evolve more slowly than upstream genes.

Characterization of sex-determining genes from several species has revealed almost no molecular similarity between phyla, but the observation that mab-3 and dsx encode proteins with a related motif, the DM domain, suggests that some evolutionary conservation may exist. Here we have described additional evidence that at least some features of sex determination in nematodes and insects are conserved. We have found that MAB-3 and DSX bind to similar DNA sequences, that C. elegans vit genes have potential MAB-3 binding sites in their promoter regions, and that a MAB-3 binding site is essential for male-specific repression of vit-2 transcription. This strongly suggests that MAB-3 is a direct regulator of vit-2 and possibly of other vit genes. Thus MAB-3 and DSX not only have related protein sequences, but they both directly control yolk protein gene transcription by binding to similar DNA sequences.

Although evidence is limited, upstream sexual regulators appear to diverge more rapidly than downstream regulators; for example, the upstream regulators tra-1 and tra-2 are among the most rapidly diverging nematode genes (de Bono and Hodgkin, 1996; Kuwabara, 1996). Similarly, in dipterans, the upstream regulator Sxl appears to have become involved in sex determination more recently than downstream genes like dsx (reviewed by Marin and Baker, 1998). The reasons for higher conservation of downstream genes are unclear, but among the proposed factors are the potentially higher pleiotropy of downstream genes due to a larger number of affected biochemical pathways (Waxman and Peck, 1998), and relatively recent recruitment of upstream genes to optimize regulation of more ancient downstream genes (Wilkins, 1995). mab-3 and dsx both control sex-specific peripheral sensory organ formation, and ectopic DSX^M can at least partially replace MAB-3, restoring male sensory ray formation to a mab-3 mutant (Raymond et al., 1998). Thus the two proteins can perform similar functions in vivo. Based on these results and the relatively high conservation of MAB-3 between nematode species, we speculate that the interface between sex-determining regulatory pathways and sex-specific structural genes may be more highly conserved than other aspects of sex determination.

We have found that MAB-3 binds to DNA sequences similar to those bound by the DSX proteins of Drosophila. This result is striking for two reasons. First, while both proteins contain the DM domain, MAB-3 has two DM domains and DSX has only one. Second, while the yolk proteins of nematodes and fruitflies are functionally analogous, they are molecularly unrelated (Wahli, 1988; Spieth et al., 1991). It therefore appears from our results that the regulatory proteins have been conserved, but entirely different structural gene families have been selected. We speculate that the pleiotropy of MAB-3 and DSX, which need to regulate multiple aspects of sexual development, may constrain their evolution, while yolk proteins may be less constrained.

Consistent with the symmetrical nature of the DSX DNA binding site, DSX proteins have been shown to bind DNA as dimers. Two dimerization domains have been identified in DSX, and the proteins have been shown to dimerize both in vitro on DNA, and in vivo (An et al., 1996; Erdman et al., 1996). The MAB-3 DNA binding site is asymmetrical, suggesting that MAB-3 may bind DNA as a monomer, although this has not been shown. The similar sizes of the MAB-3 and DSX DNA binding sites are consistent with this view, as is the requirement for both DM domains for mab-3 activity in vivo. Thus both DSX and MAB-3 might form complexes with DNA using two DM domains, with DSX doing so by forming dimers and MAB-3 binding as a monomer.

We have found that repression of vit-2 transcription in the XO intestine requires both functional MAB-3 and an intact MAB-3 binding site in the vit-2 promoter. We propose that MAB-3 is a direct transcriptional repressor of vit-2. What is the role of mab-3 in regulating other vit genes? Since null mutations in mab-3 deregulate all the vitellogenins, mab-3 must play a role in regulating all six vit genes. The simplest model is that MAB-3 directly represses transcription of the vit genes. Supporting this possibility, we have found potential MAB-3 binding sites upstream of the transcriptional initiation sites of all six vit genes, suggesting that mab-3 may regulate these genes directly. The potential MAB-3 sites in the vit promoters are not perfect matches to the in vitro MAB-3 binding consensus or to the MAB-3 binding site in the vit-2 promoter (Fig. 4), but the requirements in vivo are likely to be less stringent. In DNA binding assays in vitro, a single nucleotide change is sufficient to virtually eliminate binding of mab-3 to an oligonucleotide (e.g. Fig. 3) and mutating two nucleotides in the vit-2 promoter (probe E, Fig. 3C; vit-2 M1, Fig. 5) completely eliminates binding. However, the vit-2 M1 reporter is not fully deregulated in vivo, suggesting that MAB-3 is still capable of binding to the mutant promoter and partially repressing its transcription. One possibility is that DNA binding in the in vitro assay requires higher DNA/protein affinity than is required in vivo for regulation. A second possibility is that MAB-3 may bind cooperatively with other factors in vivo, and these factors may facilitate the efficient binding of MAB-3 to a less than optimal site. In either case it appears that the sequence requirements for regulation by MAB-3 in vivo are less strict than for MAB-3 DNA binding in vitro, and so MAB-3 might directly regulate all of the vit genes. Alternatively, it is possible that mab-3 regulates some of the vit genes indirectly.

Sexual regulation of vitellogenin expression probably involves more than just repression of vit gene transcription by MAB-3 in males and repression of mab-3 activity by tra-1 in hermaphrodites. Genetic experiments suggest that tra-1 promotes vitellogenin expression not only by negatively regulating mab-3, but also by a mab-3-independent mechanism (Shen and Hodgkin, 1988). If tra-1 activity promoted vit gene expression solely by repressing mab-3 activity, then animals deficient in mab-3 would no longer require tra-1 for high-level vit gene expression. This is not the case. Rather, while mab-3(null) XX animals express vitellogenins abundantly, tra-1;mab-3(null) XX animals express reduced levels of vitellogenins, with the amount of vitellogenin expression depending on the the tra-1 allele tested. When the tra-1 allele is the null allele e1099, tra-1;mab-3(null) XX animals accumulate reduced levels of vitellogenins relative to mab-3(null) XX animals (Shen and Hodgkin, 1988). These results suggest that tra-1 can promote vit gene expression not only by repressing mab-3 activity, but also through another mechanism that does not require mab-3. It is unclear whether this mab-3-independent activation of vit expression by tra-1 is direct or indirect, but we have not found potential TRA-1 binding sites in any of the vit gene promoters, and TRA-1A does not bind the minimal vit-2 promoter in vitro (data not shown). We speculate that one or more of the other nine C. elegans DM domain proteins might also have a role in regulating vitellogenin expression, possibly under the control of tra-1.

From the experiments presented here and from previous work (Sievert et al., 1997; Raymond et al., 1998; Shearman and Frommer, 1998), the downstream genes mab-3 and dsx do appear likely to be ancient sex-determining genes. At present it is unclear whether this is an isolated example or whether other downstream sexual regulators also are conserved in evolution. Intriguingly, mammals have DM domain-containing genes whose expression and chromosomal locations are consistent with a role in sex determination (Raymond et al., 1998; C. Raymond and D. Z., unpublished data), but this possibility has not been directly tested. Several other downstream regulators have been identified in Drosophila (Finley et al., 1997; Marin and Baker, 1998), but little is known in C. elegans about sexual regulators other than mab-3 that act downstream of tra-1. Such genes must exist, since mab-3 mutations only affect a fraction of sexually dimorphic cell lineages (Shen and Hodgkin, 1988; Hodgkin et al., 1989). It will be important to establish whether any other features of sexual dimorphism are controlled by conserved downstream regulatory genes.

We thank colleagues in the Developmental Biology Center of the University of Minnesota for helpful discussions, M. MacMorris, T. Blumenthal, R. Treisman, K. Burtis and A. Fire for clones and vectors and Christopher Raymond for clones and assistance. We thank Julie Ahringer, Vivian Bardwell, Jonathan Hodgkin and Ann Rougvie for critical reading of the manuscript. This work was supported by grants from the Minnesota Medical Foundation, University of Minnesota Graduate School and the NIH (GM53099).