A C. elegans E/Daughterless bHLH protein marks neuronal but not striated muscle development

The E proteins are helix-loop-helix (HLH) transcription factors that were first identified in a screen for factors that bind to a human immunoglobulin enhancer element (Murre et al., 1989a). The mammalian E proteins have since been shown to be encoded by at least three genes (E2A, E2-2, and HEB), each of which is expressed in a wide variety of tissues (Henthorn et al., 1990; Hu et al., 1992). This broad tissue distribution reflects the role of E proteins as heterodimeric partners for other tissue-specific HLH proteins. For example, vertebrate E proteins dimerize with MyoD family members in the regulation of skeletal myogenesis (Murre et al., 1989b; Lassar et al., 1991) and with achaete-scute family members in the regulation of neurogenesis (see Jan and Jan, 1993). E proteins may also be able to function efficiently as homodimers in some tissues. E2A gene product homodimers can be readily detected in B-cells (Shen and Kadesch, 1995), B-cell extracts can promote disulfide-bond formation between E protein homodimers (Benezra, 1994), and gene knockout experiments in the mouse demonstrate that E2A is essential for B cell development (Zhuang et al., 1994; Bain et al., 1994).

In Drosophila, the product of the daughterless (da) gene is the lone known relative of the vertebrate E proteins (Cronmiller et al., 1988). Like its vertebrate counterparts, da is broadly expressed (Caudy et al., 1988b; Murre et al., 1989a; Cronmiller and Cummings, 1993) and has multiple developmental roles. Best understood of these are its roles in neurogenesis and sex determination. Genetic and molecular data have shown that Daughterless (DA) can dimerize with members of the achaetescute complex (AS-C) to form a transient, but essential, transcription factor regulating differentiation in the peripheral nervous system (Caudy et al., 1988a,b; Murre et al., 1989b; Cabrera and Alonso, 1991; Van Doren et al., 1991; Vaessin et al., 1994). Heterodimers between DA and a single member of the AS-C, the T4 gene product, also function early in embryogenesis as a numerator element in the X/A ratio, thereby helping to set the sex-determination pathway to levels appropriate for the chromosomal constitution of the cells (Cronmiller and Cline, 1986, 1987; Cline 1988; Parkhurst et al., 1990, 1993). da mutants also have embryonic striated muscle defects, suggesting a role for da in somatic myogenesis (Caudy et al., 1988a).

Based on these studies in vertebrates and Drosophila, it seemed likely that E/DA-like proteins would be important in nematode development. Several vertebrate and Drosophila HLH proteins that heterodimerize with E/DA proteins have been found to have C. elegans homologs. These include a single MyoD family member (Krause et al., 1990) and several achaete-scute (AS) family members (Zhao and Emmons, 1995; this paper). We report here the characterization of a single gene, called hlh-2, encoding a C. elegans E/DA-related protein (designated CeE/DA). In addition, we describe a new AS-like gene, hlh-3, and show that its product is likely to be a dimerization partner for CeE/DA in neurogenesis.

The existence of a C. elegans E/DA protein, as identified by sequence of random cDNA clones, was pointed out to us by R. McCombie. The physical mapping position of hlh-2 and hlh-3 has been determined by fingerprinting of Lambda phage clones (A. Coulson, personal communication) and by direct genomic sequencing of the corresponding cosmid clones (Waterston and Sulston, 1995). The hlh-3 cDNA (accession number U78953) was identified in screens of a C. elegans Lambda gt11 embryonic expression library for components capable of binding ³²P-labelled human E12 protein (see Armand et al., 1994 for methods).

HLH-1 and HLH-2 fusion proteins were generated in the pRSET Expression System (Invitrogen) by induction in bacteria or in vitro translation of T7 RNA polymerase products. Bacterial fusion proteins were purified as described in Shirakata et al. (1993) using BL21(LysS) cells. Following denaturation, fusion proteins were purified by affinity chromatography on Ni-Sepharose columns. HLH-3 fusion protein was generated in pQE-9 (Qiagen). Resuspended cell pellets were stirred overnight in 6 M guanidine HCL, 25 mM NaPO₄, pH 8.0, and fusion proteins were purified using acid elution after binding to a nickel-NTA agarose resin (Qiagen).

Probe DNA oligonucleotides were ³²P-labeled with Klenow or endlabeled with T4 kinase. Oligonucleotides used for gel shift assays were as follows: muscle creatine kinase gene E-box AGCTTCCAA-CACCTGCTGCAAGCT (Lassar et al., 1989); CeMyoD selected E-box oligonucleotide sequence (WM1) CCCCCGTCAGCT-GACGCCTGA; achaete-scute T5 E-box AGGTAGTCACGCAGG-TGGGATCCCTAGGCCC (Van Doren et al., 1991); mutant T5 E-box AGGTAGTCACGAAGGAGGGATCCCTAGGCCC. Dimerization and DNA binding reactions were for 15 minutes at 37°C in 20 μl of 0.15 M KCl, 25 mM Hepes-KOH (pH 7.9), 2.5 mM MgCl₂, 20 mM dithiothreitol, 2.5% (v/v) glycerol, 0.1% (v/v) Nonidet P-40, 50 ng/μl poly (dI-dC), 50 ng/μl bovine serum albumin (Boehringer-Mannheim). Reaction material was separated in 5% (w/v) native polyacrylamide gels at 100 V for 1.5 hours at room temperature. Electrophoresis buffer was 106 mM Tris, 89 mM boric acid and 0.1% (v/v) Nonidet P-40.

Bacterially produced CeE/DA:pRSET-B fusion protein (2 mg/ml) was used to immunize two rabbits. Serum samples taken at 4, 7, 9 and 11 weeks following the initial inoculation were positive in immunocytochemistry of fixed animals. The patterns of immunolocalization were identical for all sera; pre-immune serum from either rabbit showed no staining.

Two fixation methods were used for immunocytochemistry. One method used 5 M paraformaldehyde in 0.5× PBS as previously described (Krause et al., 1990). The second fixation method is a slight modification of that described by Tabara and colleagues (1996) for in situ hybridization to whole embryos. Embryos were handled in a silanized Eppendorf tube and fixed in buffered paraformaldehyde as described. Following fixation, embryos were either dehydrated in ethanol for storage or used immediately after three washes with TTBS (0.1 M Tris, pH 7.5, 0.15 M NaCl, 0.1% Tween-20).

Antibody-positive nuclei were identified after orienting the DAPI-staining pattern to published nuclear positions for given developmental stages or tissues. In some cases, individual antibody-positive nuclei could be unambiguously identified; this was not possible in regions with where positive and negative nuclei were tightly clustered and present in multiple focal planes (e.g. head region of bean-stage embryo).

Transgenic strains were generated by microinjection (Mello and Fire, 1995) using the pRF4 plasmid as a marker to identify transformed animals. Test plasmid constructs (10-100 ng/μl) in combination with pRF4 DNA (50 ng/μl) were injected; in some cases, N2 genomic DNA was included in injections (Kelly et al., 1997).

Reporter plasmids were as follows. For hlh-2::lacZ, an approx. 3.5 kb restriction fragment 5′ to the gene was fused to lacZ (using pPD21.28; Fire et al., 1990) using a BamHI site just downstream of the predicted initiator methionine. hlh-2:: GFP chimeras were produced by inserting a GFP cassette into one of two restriction sites within a 7.7 kb AgeI-NcoI genomic fragment containing the entire coding region of hlh-2. The GFP cassette was inserted at the SpeI site (intron 2) of hlh-2 in strain PD8079 and at the BamHI site (exon 1) in stains PD8096 and PD8097. These GFP reporters include approx. 3.8 kb 5′ and 700 bp 3′ flanking sequences. The hlh-3::GFP reporter gene was generated by inserting a GFP cassette, in frame, into a unique PstI site within the coding region of hlh-3 contained on a 7 kb BamHI-HindIII genomic clone. A single stable line of this reporter construct was generated by gamma irradiation in a smg-1(r861) unc-54(r293) background.

Sense or antisense RNA was transcribed in vitro using either T3 or T7 RNA polymerase; reactions were stopped and used directly for injections (treatment of the RNA transcription reaction with DNAse had no effect). For inhibition of hlh-1, RNA was injected at 2.8 μM. Approximately 15% of progeny were phenotypically similar to the CeMyoD null mutant allele, cc450. Some variability in severity of the phenocopy may reflect mosaicism in inhibition among cells of the embryo. For inhibition of the hlh-2 gene encoding CeE/DA, antisense RNA was injected at 2.2 μM. Approximately 90% of progeny generated within a day of the injection arrested, yielding variable phenotypes.

A C. elegans sequence capable of encoding a protein related to vertebrate E and Drosophila Daughterless (E/DA) was first identified by random cDNA clone sequencing efforts (McCombie et al., 1992); the 430 bp E/DA-related clone (CEESI77R.b) was kindly provided to us by R. McCombie. Using the EST clone as a probe we identified several longer cDNA clones from a mixed-stage library (Barstead and Waterston, 1989). These cDNA clones and corresponding genomic clones define a single genetic locus, hlh-2 (Fig. 1A) (GenBank accession numbers U30248 and U13614).

The hlh-2 gene is composed of five exons spanning approximately 2.5 kb (Fig. 1B). All splice sites match those typically seen in C. elegans, with the possible exception of the 3′ splice acceptor sequence of intron 3 (TTCGAG). Intron 3 is notable because it interrupts the HLH region and its splicing could potentially regulate the function of the encoded protein. We have not directly investigated potential differential splicing at this site. The message is trans-spliced to the SL1 leader sequence (Krause and Hirsh, 1987) at a sequence located 7 bp upstream of the predicted initiation codon. Two adjacent poly(A) addition sites were demonstrated by cDNA sequence and are located 14 and 15 bp downstream of the consensus sequence TATAAA (see Krause, 1995).

Although only a single E/DA gene has been identified in Drosophila, there are at least three E protein-encoding genes in vertebrates. We have looked for additional E/DA-encoding genes in C. elegans but have been unable to identify any genes, other than hlh-2, that are closely related in sequence. Searches have included library screens by hybridization, Southern blot hybridizations under normal or reduced stringency with whole gene or bHLH region-only probes, RT-PCR with degenerate primers from the conserved regions of the E/DA HLH domain, and current C. elegans sequence databases. Although we find no evidence for additional related genes in C. elegans, negative results from these screening methods do not rule out the possible existence of other related genes; it remains an open question that will be definitively answered with the completion of the remaining 40% of the nematode genome-sequencing project (Waterston and Sulston, 1995).

The hlh-2 gene encodes a predicted protein product, which we designate CeE/DA, of 399 amino acids and 43 kDa. A comparison of CeE/DA to vertebrate E proteins and Drosophila DA shows a high level of homology (67% identity) across the basic-HLH (b-HLH) region of the proteins, including the signature amino acid residues characteristic of this subclass of HLH proteins (Fig. 1C). Outside the bHLH region there is no apparent homology between CeE/DA and its vertebrate or Drosophila counterparts. The N-terminal portion of the C. elegans protein is generally acidic.

C. elegans embryogenesis can be divided into two distinct phases, the first being cell proliferation and the second being differentiation and morphogenesis. The proliferative phase, in which almost all of the 558 embryonic cells are generated, spans about 5 hours and is marked by fairly short cell-cycle times (approx. 30 minutes on average). The differentiation and morphogenesis phase spans the remaining 8 hours of embryogenesis. CeE/DA is abundant in early proliferative embryos but is progressively restricted to fewer cells by the differentiation and morphogenesis phase, as described below.

Two rabbit polyclonal antibodies were raised against a single bacterially expressed CeE/DA fusion protein (see Materials and Methods) and used to immunolocalize CeE/DA; both antisera give identical results when used for immunohistochemistry and will together be referred to as anti-CeE/DA antibodies. At all developmental stages, CeE/DA antibody staining is nuclear (except in the germline, see below), consistent with the protein functioning as a transcription factor.

There are four lines of evidence that suggest that the antibody signal we observe in fixed embryos is specific for CeE/DA protein. First, serum from each of two rabbits gave identical immunolocalization results, whereas the pre-immune serum for each detected nothing. Second, the antibody signal in fixed embryos is eliminated by competition with purified, bacterially expressed CeE/DA fusion protein. Third, anti-CeE/DA antibodies detect protein derived from a heat shock promoter-driven hlh-2 cDNA construct in transgenic animals after these strains are stressed (data not shown). The high levels of induced CeE/DA protein in these transgenic strains are distributed in cells expected from the known expression pattern of the heat shock promoter and are nuclear localized. Fourth, hlh-2 reporter genes are expressed in a pattern that mimics the somatic distribution of CeE/DA protein, as determined by immunolocalization, during mid and late stages of embryogenesis (see below). Because reporter genes in C. elegans are essentially never expressed maternally, the early ubiquitous distribution pattern of CeE/DA cannot be confirmed with reporter gene assays.

The pattern of CeE/DA immunolocalization during C. elegans embryogenesis is complex (Fig. 2). CeE/DA can first be detected in both nuclei of 2-cell embryos; the fragility of single-cell embryos in our fixation method precludes us from making definitive conclusions about its presence or absence in the nucleus of singlecell embryos. Staining persists apparently in all nuclei of the early embryo for the first 150-200 minutes of development (100-200 cells). By the 100-cell stage, the six founder cell lineages (AB, MS, C, D, E, P₄) have been established and two of these, D and E, are clonally committed to produce only body wall muscle and gut cells, respectively.

By 270 minutes of development (approx. 350 cells) a dramatic change in antibody staining has occurred in which persistent staining is seen in progressively fewer blastomeres. Most, but not all, blastomeres that initially retain CeE/DA antibody staining at this stage are neurons or their immediate precursors (Fig. 2I,J,K). There are a few neuronal precursors that are located away from the neuronal clusters in the embryo (for example the postembryonic neuroblast W), for which we have not detected antibody staining. Therefore, although persistent antibody staining is largely restricted to neurons or their precursors, not every such cell is antibody-positive.

CeE/DA-antibody staining is transient for the majority of these cells, with the staining progressively lost as differentiation and morphogenesis occur (see Fig. 2). This is most clearly evident at the 1.5-fold stage of embryogenesis, in which a lateral view of the embryo shows staining in the head, ventral nerve cord and tail. As the embryo begins elongating, the level of CeE/DA-antibody staining decreases in these cells. Note that most of these cells are postmitotic.

Although the majority of cells lose CeE/DA-antibody staining during the later half of embryogenesis, a small percentage of cells remain antibody-positive through the remainder of embryogenesis and after hatching. There are 14 of these continually staining cells in the head and seven more in the tail region (Fig. 2; see also Fig. 4). Of the 14 head cells, 5 are pharyngeal. The pharyngeal nuclei have been identified, with the generous help of Dr Leon Avery, as two pharyngeal muscle nuclei (pm5L and R) and three pharyngeal gland cell nuclei (g1P, g2L and R) as shown in Fig. 3.

The remaining nine CeE/DA antibody-positive cells in the head are outside of the pharynx and are located in the neuronal cluster between the nerve ring and the posterior pharyngeal bulb. There are four bilateral pairs of stained nuclei and one positive nucleus lying along the ventral mid-line. Using hlh-2::GFP reporter strains (see below) and DiI staining, three of the bilateral pairs of neurons have been identified, with the generous help of Cori Bargmann, as ADL (L and R) and ASH (L and R) and RIC (L and R). The large number of neurons in this area makes it difficult to identify unambiguously each of the remaining three CeE/DA antibody-positive cells. The seven tail cells with nuclei that remain CeE/DA antibody-positive throughout embryonic development include the two Q neuroblasts and five cells that we have tentatively identified as DVA (an interneuron), the bilateral pair of intestinal muscle cells, the anal depressor muscle and the anal sphincter muscle. The two intestinal and two anal muscle cells are postmitotic and are non-striated muscles.

There is a single gene, hlh-1, identified in C. elegans that is related to the vertebrate MyoD family (MyoD/Myf-5/Myogenin/MRF-4); hlh-1 is expressed in bodywall (striated) muscle cells and their precursors during embryogenesis (Krause et al., 1990). Stable accumulation of the hlh-1 gene product, CeMyoD, is first detected in proliferating blastomeres at about the time they become restricted to the bodywall muscle lineage; these cells and their descendants remain CeMyoD-positive throughout the animal’s life (Krause et al., 1990). Surprisingly, CeE/DA antibodies do not stain differentiating bodywall muscle cells or their immediate precursors. This is best seen at the 1.5-fold stage of embryogenesis, by which time the bodywall muscle cells have all been born and are arranged in four quadrants along the length of the embryo. At this stage, the posterior region of the dorsal two bodywall muscle quadrants lies in a region of the embryo that is devoid of neuronal cells, allowing one to easily distinguish between neuronal and bodywall muscle staining (Fig. 4). There are no CeE/DA antibody-positive cells in this region of the embryo at this stage, nor have we ever detected CeE/DA in differentiated bodywall muscle cells or their immediate precursors.

For lineages that are clonally committed to the bodywall muscle fate early in development, such as the C and D lineages, there is a period of development during which both CeE/DA and CeMyoD are co-localized in the nucleus. However, this co-localization ends with the progressive disappearance of CeE/DA in these cells during subsequent divisions, with complete loss of the CeE/DA signal occurring prior to the terminal cell division and any signs of overt muscle differentiation. In the early myogenic blastomeres of the MS lineage, the accumulation of CeMyoD occurs at about the time that the ubiquitious distribution of CeE/DA has faded and CeE/DA is restricted to mostly neuronal precursors. Double-labeling of hlh-1::lacZ embryos at the comma to two-fold stage with antibodies for CeE/DA and beta-galactosidase shows that the patterns for these two antibodies are mutually exclusive (data not shown).

We have used in situ hybridization to assay hlh-2 gene expression in embryos. The results from early-stage embryos are difficult to define precisely because of sporadic hybridization of the hlh-2 probe; we can only conclude that the first reproducibly detectable expression in these assays occurs at about the 50-cell stage of embryogenesis in a group of unidentified, anteriorly located blastomeres. Later in embryogenesis, there is a consistent pattern of expression that is confined to the anterior and ventral regions of the embryo. By the ‘bean’ stage, the anterior-ventral expression determined by in situ hybridization is very similar to the distribution of CeE/DA protein as determined by antibody staining (Fig. 4).

In addition to the 21 cells that are CeE/DA antibody-positive at hatching, there are several additional cells detected immunologically during subsequent development (Fig. 5). One prominent set of cells that becomes CeE/DA antibody-positive during the L3 stage are the 16 developing vulval and uterine muscle cells (non-striated) (Fig. 5D). These nuclei remain antibody-positive in the mature vulva, although staining intensity appears to decrease.

Another prominent pair of postembryonic, CeE/DA antibody-positive nuclei are the distal tip cells (DTC) (Fig. 5E). The DTC nuclei are CeE/DA antibody-positive from the start of gonad elongation in larval development and remain positive in adulthood. Very faint antibody staining can also be detected in the syncytial gonad; because of the low level it is unclear if this actually reflects the in vivo distribution of CeE/DA. If real, however, this staining is cytoplasmic (not nuclear), suggesting that the subcellular localization of CeE/DA may be regulated.

In C. elegans, microinjection of antisense (or sense) RNA is capable of inhibiting maternal genes in a sequence-specific manner, resulting in embryos that phenocopy genetic loss-of-function mutations at the corresponding locus (Guo and Kemphues, 1995, 1996; Lin et al., 1995). In some cases, zygotically transcribed genes can also be inhibited by RNA injections. For example, injecting hlh-1 RNA can phenocopy hlh-1 null mutants (data not shown). However, we observed a lower percentage of hlh-1 phenocopies (approx. 15%) than is typically seen for maternal gene inhibition by this technique. Moreover, affected animals displayed a range in the severity of the phenocopy, reminiscent of a genetic allelic series.

Because no genetic mutations have yet been identified at the hlh-2 locus, we have used RNA inhibition as a preliminary assay of CeE/DA function. Both antisense and sense RNA complimentary to regions of the hlh-2 gene result in abnormal progeny from injected hermaphrodites (data not shown). The frequency of affected embryos from these experiments was approximately 90%, which is reminiscent of maternal gene product inhibition (the distribution of CeE/DA demonstrates a maternal contribution, see above). The phenotype of these abnormal progeny was variable, ranging from embryonic lethality to viable larvae with slight morphological defects. Most of the arrested embryos had developed beyond stages of cellular differentiation, allowing us to score for the presence of pharynx, neurons and hypodermis by morphology, for bodywall muscle by movement, and for gut by autofluorescent granules. Although recognizable, each of these tissues (with the exception of the gut) was routinely affected in morphology or function in the arrested embryos and severe defects in morphogenesis and elongation of the embryo were common. Embryos that escaped this early lethality often developed into larvae displaying two common phenotypes; vacuoles in the head of the animal indicative of cell degeneration, and morphological defects (a bulge) in the region of the anus. The head and tail regions affected are coincident with sites at which CeE/DA persists postembryonically, suggesting that these RNA-induced phenotypes might reflect the reduction or elimination of CeE/DA activity in a fraction of the CeE/DA-positive cells.

Studies of vertebrate E and Drosophila DA proteins have demonstrated both in vitro and in vivo interactions with cell type-specific members of the b-HLH super family, including Achaete-scute (AS) members in neurons (Cabrera and Alonso, 1991) and MyoD members in striated muscle (Murre et al., 1989b; Lassar et al., 1991). We have addressed the question of CeE/DA heterodimerization preference using C. elegans homologs of both the AS and MyoD families.

In screening a C. elegans cDNA expression library with labeled human E12 protein, we isolated a clone that showed extensive homology to the b-HLH regions encoded by achaetescute genes isolated from other organisms (Fig. 6); the genomic locus corresponding to this clone is designated hlh-3. Preliminary expression studies using a hlh-3::GFP reporter gene construct suggest that the hlh-3 promoter can function during embryogenesis in many of the neurons that show persistent CeE/DA localization (Fig. 6D). This potential coincident expression with hlh-2 made HLH-3 a good candidate for a protein that would heterodimerize with CeE/DA.

We have tested CeE/DA:HLH-3 dimerization potential by gel shift assays using an oligonucleotide containing an E-box sequence (CAGGTG) from the promoter of the Drosophila achaete (or T5) gene (Villares and Cabrera, 1987; Van Doren et al., 1991); this site has been shown to have high affinity for heterodimers of the Drosophila proteins DA and AS-C T5. We find that apparent CeE/DA and HLH-3 heterodimers can be detected using this sequence, whereas neither protein binds this sequence efficiently as a homodimer (Fig. 7A). Mutating two positions of this E-box consensus (to AAGGAG) completely eliminates binding by CeE/DA:HLH-3 heterodimers.

Because we have been unable to detect CeE/DA in differentiating or mature bodywall muscle cells (striated), we were interested in determining if CeE/DA could interact with CeMyoD, as assayed by gel shifts. Previous studies of CeMyoD had demonstrated an ability for this protein to form efficient homodimers that could be used in vitro to select a high affinity, symmetrical, E-box sequence-containing site (CGTCAGCTG ACG, K. Blackwell, personal communication). This selected E-box sequence, when concatenated, is capable of promoting bodywall muscle-specific expression when reintroduced into the animal (data not shown).

We have assayed CeMyoD:CeE/DA dimerization by gel mobility shifts of oligonucleotides containing either of two E-box sequences; one contains the CeMyoD homodimerselected, symmetrical E-box sequence (CAGCTG) and the other the canonical vertebrate MCK right E-box (CACCTG). The results show a strong preference for CeMyoD homodimers in all assays in which it was included, regardless of which E-box sequence was used. Heterodimers between CeE/DA and CeMyoD could be detected using the MCK right site E-box sequence; however, it required that CeE/DA be present in large excess of CeMyoD (Fig. 7B). Even in the presence of a 100-fold molar excess of CeE/DA to CeMyoD, CeMyoD homodimers can be detected.

We have identified hlh-2, a C. elegans gene that encodes a protein (CeE/DA) related to vertebrate E and Drosophila Daughterless (DA) b-HLH transcription factors. Like Drosophila, C. elegans may have only one such gene; a variety of approaches to identify additional related sequences in the nematode have been negative. While some aspects of CeE/DA tissue distribution and function appear to be evolutionarily conserved, there are also examples in which it differs from its vertebrate and Drosophila counterparts.

In vertebrates, the E protein genes are expressed in a wide variety of tissue types reflecting roles in multiple cell fate determination events during development. Although the exact tissue distribution of the various vertebrate E protein isoforms has not yet been determined, mRNAs encoding E proteins are found in most tissues. Putative cell-type specific heterodimerization partner proteins for E factors have been identified in striated muscle, nerve, skin and skeletal tissue (Lassar et al., 1991; Johnson et al., 1990; Li et al., 1995; Cserjesi et al., 1995). Similarly, Drosophila da appears to function throughout development (Cronmiller and Cline, 1986, 1987; Caudy et al., 1988a,b; Cronmiller and Cummings, 1993; Vaessin et al., 1994). The da gene has both maternal and zygotic functions; the protein is present continually and in nearly all tissues, and da has defined roles in sex determination, neurogenesis and oogenesis.

Given these precedents, we were surprised to find that in C. elegans the CeE/DA protein has a highly restricted pattern in late development. Although ubiquitous in all early blastomeres, CeE/DA becomes progressively less prevalent during subsequent embryogenesis. Both in situ hybridization and reporter gene studies suggest that transcriptional regulation could be responsible, at least in part, for controlling the dynamic changes observed in embryonic CeE/DA distribution. It is possible that alternative and/or additional mechanisms, such as degradation of maternal RNA, differential RNA stability or protein stabilization (e.g. due to heterodimerization status), could also contribute to the overall pattern of CeE/DA distribution.

The restriction of CeE/DA largely to neuronal cells and their precursors during late embryogenesis is consistent with expectations that CeE/DA has a role in neurogenesis as in other animals, where E/DA proteins serve as necessary dimerization partners for members of the achaete-scute complex (AS-C)- related genes (Johnson et al., 1990; Jan and Jan, 1993; Akazawa et al., 1995; Ishibashi et al., 1995). At least five genes in C. elegans have been identified that are related to the achaete-scute genes of Drosophila (Zhao and Emmons, 1995; Waterston and Sulston, 1995; this work and unpublished observations). Here, we show that the product of the C. elegans AS-like gene, hlh-3, is capable of interacting with CeE/DA in vitro. In addition, preliminary analysis of a hlh-3::GFP reporter gene expression pattern shows it to overlap with the embryonic distribution of CeE/DA. It seems likely that CeE/DA functions as a heterodimeric partner for HLH-3 and perhaps for other AS-like genes in vivo to promote neurogenic determination and/or differentiation. These neurogenic functions may be the basis for the head and tail defects observed in a fraction of hlh-2 antisense RNA-treated animals.

At hatching, CeE/DA is present in a subset of muscle cells in C. elegans. All of the CeE/DA antibody-positive muscles (pm5L and R, anal depressor, anal sphincter, intestinal, uterine and vulval) are non-striated (Albertson and Thompson, 1975; Ardizzi and Epstein, 1987; Waterston, 1988) and do not contain detectable levels of CeMyoD (Krause et al., 1990; M. Krause, unpublished). We have no candidate HLH partner proteins for CeE/DA in these non-striated muscle cells of C. elegans. CeE/DA could function in these tissues as a homodimer (reminiscent of models proposed for B cell development in mammals), or there could be one or more novel bHLH proteins that serve as dimerization partners for CeE/DA in these cells.

The developmental profile of CeE/DA distribution in C. elegans striated bodywall muscle cells is surprising in view of work on other systems. E proteins have long been ascribed a role in striated myogenesis. Heterodimers between E and MyoD family members can be demonstrated in vertebrate skeletal myogenesis (Lassar et al., 1991) and Daughterless:Nautilus heterodimers are suspected in Drosophila myogenesis (Caudy et al., 1988b; B. Paterson, unpublished). C. elegans has a single MyoD family homologue, CeMyoD, which accumulates in striated muscle cells and is required for proper myogenesis (Krause et al., 1990; Chen et al., 1992, 1994). Although CeMyoD is not required solely for cells to adopt their proper fate, it plays an important developmental role in these cells and can be detected early in the lineage of blastomeres destined to making only striated muscle cells.

Although CeE/DA is clearly present in some of the early bodywall muscle-cell lineages, CeE/DA is undetectable in differentiating bodywall muscle cells. It is possible that CeE/DA could interact with CeMyoD early in development to commit these lineages (D lineage for example) to the bodywall musclecell fate. We do not favor this explanation, however, for three reasons. First, rescue of the hlh-1 mutant phenotype can be achieved by expressing wild-type hlh-1 relatively late in embryogenesis using the bodywall muscle myosin heavy chain gene (unc-54) promoter (A. Fire, unpublished). The expression of this construct occurs at a time when CeE/DA is no longer immunologically detectable in bodywall muscle, suggesting these two factors need not be present together in cells for them to undergo myogenesis. Second, the 14 bodywall muscle cells born postembryonically from the M cell lineage are CeMyoD-positive but never express detectable levels of CeE/DA during any postembryonic period. In contrast, many of the non-striated muscle cells derived from M postembryonically have readily detectable levels of CeE/DA. If CeMyoD and CeE/DA were required embryonically for bodywall myogenesis, one might expect a similar requirement in postembryonic bodywall muscle lineages, yet CeE/DA appears to be absent. Third, we have been unable to detect a strong interaction between CeE/DA and CeMyoD in vitro, whereas such interactions are easily demonstrated between their vertebrate homologues that are together thought to be essential for striated myogenesis.

If CeE/DA is not interacting with CeMyoD in vivo, could another factor serve as the partner for CeMyoD? One possibility is that a second E/DA-related protein serves as a heterodimer partner for CeMyoD. Multiple search strategies have failed to identify any additional E/DA-like factors in C. elegans; therefore it seems reasonable to consider the possibility that CeMyoD homodimers, not CeE/DA:CeMyoD heterodimers, might act as an essential transcription factor for proper myogenesis. This model is consistent with the distribution patterns of both CeMyoD and CeE/DA and is supported by our in vitro dimerization studies, which confirm and extend earlier studies showing a strong propensity for CeMyoD to form homodimers. Moreover, CeMyoD homodimers are capable of high affinity, E-box-specific DNA binding, as shown by their resistance to competition from either mutant E-box oligonucleotides or excessive amounts of CeE/DA.

In Drosophila, DA and AS proteins heterodimerize and function to promote neurogenesis. One function of these DA:AS heterodimers is to activate the gene encoding Delta, a ligand for Notch, and initiate the lateral inhibition signalling cascade that prevents neighboring cells from also adopting the neurogenic cell fate (Kunisch et al., 1994). Homologs of AS, Delta and Notch have been identified and studied in C. elegans; their role in intercellular signaling during gonadogenesis in the worm appears mechanistically very similar to their neurogenic role in Drosophila (see Greenwald and Rubin, 1992; Zhao and Emmons, 1995). In C. elegans, the Delta-like gene lag-2 is expressed in the distal tip cells (DTCs) and is necessary to activate glp-1 in the germline (Lambie and Kimble, 1991; Henderson et al., 1994; Tax et al., 1994; Crittenden et al., 1994; Fitzgerald and Greenwald, 1995). Reporter gene studies suggest that the lin-32 (AS homolog) promoter is expressed in the DTCs (S. Emmons, personal communication) and we detect CeE/DA in the DTCs throughout gonadogenesis. It is tempting to speculate that CeE/DA:LIN-32 heterodimers are involved in the activation and/or maintenance of lag-2 expression in the DTCs. CeE/DA may similarly be involved in other GLP-1/LIN-12 signaling pathways, perhaps underlying other sites of CeE/DA accumulation during C. elegans development.

We have observed that apparently identical cell types can express different combinations of transcription factors, apparently reflecting ontogeny rather than just terminal differentiation. The pharynx has 20 muscle cells arranged as eight groups (pm1-8) (Albertson and Thompson, 1975). Five of these eight groups (pm3-7) comprise three identical cells, reflecting the threefold rotational symmetry of the pharynx. pm5 is one such group of three identical cells, each of which is binucleate as a result of the fusion of two cells during late embryogenesis (Sulston et al., 1983). Of the six pm5 muscle nuclei, only two are CeE/DA-positive. Moreover, these two nuclei are not in the same cell; each is one of a pair of adjacent nuclei in a single cell in which one nucleus is CeE/DA-positive and the other CeE/DA-negative. Therefore, the pm5 muscle group has two cells in which half of the nuclei are CeE/DA-positive, and one cell with neither of the nuclei CeE/DA-positive.

The cell lineage suggests a mechanism for how localization of CeE/DA to one nucleus of binucleate pm5 cells is achieved. The two CeE/DA-positive nuclei are descendants of the AB blastomere, whereas the remaining four pm5 nuclei are MS descendants (Sulston et al., 1983). We assume that the two AB-derived nuclei of the cells accumulated CeE/DA prior to, and during, pharyngeal organogenesis, and that after cell fusion the nuclear CeE/DA must be stable and unable to exchange with its co-syncytial MS-derived nucleus.

However, this possible mechanism does not account for the heterogeneity among nuclei of the pharyngeal gland cell (g1). In this case, heterogeneity may be achieved by localized differences within the cell caused by its response to intercellular signals, as has been observed to cause transcriptional heterogeneity among syncytial nuclei of the Drosophila embryo (reviewed in St Johnson and Nusslein-Volhard, 1992) or vertebrate muscle cells (Burden, 1993).

Regardless of the cause or function of nuclear heterogeneity with respect to CeE/DA accumulation, the observation that such heterogeneity can exist and persist over time demonstrates again that common cytoplasm does not dictate nuclear homogeneity. As more transcription factors are localized, it is likely that additional examples of heterogeneity among syncytial nuclei will be seen in C. elegans. This may be especially true for cells spanning large areas of the animal, such as the hyp 7 cell, which arises by the fusion of many cells from different branches of the cell lineage.

We thank Leon Avery for help in identifying CeE/DA-positive cells in pharynx, Cori Bargmann for help in identifying CeE/DA-positive head neurons, Scott Emmons for sharing information on DTC expression and Lihsia Chen for comments on the manuscript. Thanks also to Robert Littlejohn and Rey Lee-Llacer for technical assistance during the course of this work. This paper is dedicated to the memory of Hal Weintraub. His love and contagious enthusiasm for science was remarkable and instilled a sense of excitement in those of us that were fortunate enough to spend time with him. This work was supported in part by NIH grants GM15052 to J. Y., CA55248 to M. C., GM37602 to I. G. and GM37706 to A. F.; I. G. is an Associate Investigator of the Howard Hughes Medical Institute.