ELT-5 and ELT-6 are required continuously to regulate epidermal seam cell differentiation and cell fusion in C. elegans

The epidermis (or hypodermis) of the nematode Caenorhabditis elegans performs critical functions during development, including establishing the basic body form, providing the substrate for cell and axon migrations, secreting the cuticle, and producing most of the additional cells that arise during post-embryonic development (Hedgecock et al., 1987; Priess and Hirsh, 1986; Sulston and Horvitz, 1977; White, 1988). Most of the body is covered by the ‘major’ epidermal cells (Gendreau et al., 1994), which are invariably produced by fewer cell divisions than the ‘minor’ epidermal cells, the small syncytial cells at the extreme ends of the worm. Three major epidermal cell types, the seam, syncytial and P cells comprise the epidermal epithelium in the embryo.

When epidermal cells are born, they assemble into three rows on either side of the worm, corresponding to dorsal, lateral and ventral positions (Sulston et al., 1983). The dorsal, and the anterior and posterior ventral cells fuse to form several syncytia (Podbilewicz and White, 1994). Neither the P cells, the central cells in the ventral row, nor seam cells, the lateral cells, fuse during embryogenesis. The seam cells are particularly critical: they play an organizing role in embryonic and larval morphogenesis, contribute to larval growth by generating additional epidermal and nervous tissue post-embryonically, and regulate the body form during formation of the alternative larval form, the dauer larva (Singh and Sulston, 1978; Sulston and Horvitz, 1977; Wissmann et al., 1999; Wissmann et al., 1997). At hatching, there are ten bilateral pairs of seam cells, nearly all of which are blast cells that, during post-embryonic development, produce many progeny, including more seam cells, neurons and syncytial epidermis (Sulston and Horvitz, 1977). Late in the final (L4) larval stage, the seam cells undergo homotypic fusion to form the bilateral seam syncytia of adults (Podbilewicz and White, 1994).

Comprehensive genetic screens have shown that specification of the embryonic epidermis and its patterning into the three major types are genetically complex processes (Chanal and Labouesse, 1997; Terns et al., 1997). Most of the epidermis, including all of the seam and P cells, arises from the AB ‘founder cell’ (Sulston et al., 1983). A combination of inductive interactions and asymmetric cell divisions in the early embryo (Gendreau et al., 1994; Hutter and Schnabel, 1994; Hutter and Schnabel, 1995; Kaletta et al., 1997; Lin et al., 1998; Mango et al., 1994; Mello et al., 1994; Moskowitz et al., 1994) dictates which early AB-derived blastomeres will ultimately produce epidermis. These early events restrict expression of the ELT-1 GATA transcription factor to precursors of the major AB-derived epidermal cells and their descendants. elt-1 is required to specify the major epidermal fate (Page et al., 1997); in elt-1(−) mutants, all cells that would normally become major epidermal cells instead adopt the fates of their closest non-epidermal relatives, such as muscle and neurons. LIN-26, a transcription factor expressed in non-neuronal ectodermal cell types (Labouesse et al., 1994) is a potential target of ELT-1. Epidermal cells degenerate or become neurons in lin-26 mutants. Another likely target of ELT-1 is ELT-3, a GATA factor expressed in all major epidermal cells except the seam cells (Gilleard et al., 1999). elt-3 null mutants show no obvious defects, suggesting that elt-3 may be functionally redundant with another gene (Gilleard and McGhee, 2001). The exclusion of ELT-3 specifically from seam cells suggests that it may play a role in distinguishing non-seam from seam epidermis.

Although little is known about how the epidermis is patterned into three distinct types and how fusion is regulated in the embryo, a good deal is known about the mechanisms that distinguish individual epidermal cells according to their anteroposterior position and developmental stage, particularly during post-embryonic development (for reviews, see Ambros, 1997; Kenyon et al., 1997). For example, two Hox genes, lin-39 and mab-5, interact to regulate fusion of the posterior daughters of some P cells in hermaphrodite and male larvae (Salser et al., 1993). In addition, in mab-5(−) male larvae, two posterior seam cells, V5 and V6, which normally produce the sensory rays, instead produce alae (Salser and Kenyon, 1996). Though several other genes (Wrischnik and Kenyon, 1997) are also known to determine the precise fate of particular seam and P cells based on positional or temporal (Bettinger et al., 1996) information, genes that regulate seam cell development per se have not been reported.

In the present study, we have sought to identify genes that impart the seam fate as opposed to the P or syncytial epidermal fates. As the ELT-1 GATA factor is required to specify all major epidermal cells, and the ELT-3 GATA factor is expressed in non-seam epidermis, we hypothesized that there may be another GATA factor that functions in the seam epidermis. Indeed, we report here that two GATA factors, ELT-5 and ELT-6, function in seam cell development. Removal of elt-5 and -6 function results in profound defects in many aspects of seam cell development, including fusion with neighboring epidermal syncytia, failure to express several seam-specific markers, and inappropriate expression of elt-3. elt-5 and -6 are expressed in seam cells and many other cells, under the control of separable, cell type-specific enhancer regions. elt-5 and -6 apparently form an operon, and appear to be transcribed both monocistronically and dicistronically, depending on the specific cell expressing them. Our findings demonstrate that elt-5 and -6 are required continuously for embryonic and post-embryonic seam cell development and for repression of epidermal cell fusion.

C. elegans Bristol variety N2 was used as the wild-type strain. Maintenance of strains was as described previously (Brenner, 1974).

The strain JR672 contains an integrated array (wIs54, made by M. Fukuyama and J. Kasmir in our laboratory) expressing a seam-specific GFP marker (SCM), a derivative of a construct described previously (Terns et al., 1997). The strain SU93 (jcIs1) contains an integrated array expressing a JAM-1::GFP fusion protein in adherens junctions (Mohler et al., 1998). (We use all upper cases to refer to translational fusions, and lower-case italics followed by upper cases, as in elt-3::GFP, to refer to transcriptional fusions.) Both JR672 and SU93 are available from the Caenorhabditis Genetics Center. Strain JG5 (vpIs1), which carries elt-3::GFP on an integrated transgenic array (Gilleard et al., 1999), was obtained from J. Gilleard. The following strains carrying seam-specific reporters on extrachromosomal arrays were obtained from P. Sengupta (Miyabayashi et al., 1999): PY1215 (nhr-72::GFP), PY1269 (nhr-73::GFP), PY1267 (nhr-74::GFP), PY1214 (nhr-77::GFP), PY1282 (nhr-81::GFP), PY1324 (NHR-82::GFP) and PY1397 (nhr-89::GFP). JR1736 was created by co-injecting an nhr-75::GFP construct (a gift from P. Sengupta) and an unc-119(+) rescuing construct into unc-119(ed4) hermaphrodites.

The elt-5 and -6 cDNA sequences were determined by sequencing EST clones yk474 and yk113, respectively. (yk clones were generous gifts from Y. Kohara.) These sequences differ in several respects from the Worm Genome Consortium’s predictions for the corresponding open reading frames (ORFs) F55A8.1 and F52C12.5 (C. elegans Sequencing Consortium, 1998). Most notably, the predicted ORF F55A8.1 does not include the last intron (3) and exon (4) of the elt-5 cDNA, and the predicted ORF F52C12.1 does not include exons 1 and 2, and introns 1 and 2 of the elt-6 cDNA. The elt-5 and -6 cDNA sequences have been reported to GenBank (Accession Numbers AF353302 and AF353303, respectively).

An elt-6-specific primer (5′ GAAGCACGGCTTTCAGTTG 3′) from Exon 3 was used to PCR-amplify the 5′ end of the elt-6 gene from an embryonic cDNA pool prepared by J. Zhu (Zhu et al., 1997). The PCR products were cloned into vector pCR2.1 (Invitrogen). Nine clones, determined by PCR to be approximately the expected size or somewhat longer, were sequenced. Four of these were trans-spliced with SL1, and two carried 105-120 nucleotides of sequence upstream of the elt-6 trans-splice site. The remaining three sequences started a few bases after the trans-splice site, and did not provide useful data. A cDNA clone, yk391a6, also starts ∼120 bp upstream of the elt-6 trans-splice site.

Molecular cloning procedures were performed according to standard methods (Sambrook et al., 1989). Expression constructs were made with PCR products containing tagged restriction sites amplified from genomic DNA (fragments C, D, E and G), the cosmid GGC8 (fragments A and B), or the cDNA clone yk113 (fragment F). The PCR fragments are as follows (the numbers in parentheses correspond to the base position relative to the base A of the elt-5 ATG; see Fig. 5 below).

A: 3.4 kb upstream of the elt-5 ATG (−3380, +4);

B: 1.2 kb upstream of the elt-5 ATG (−1211, +4);

C: between the elt-5 ATG and poly(A) site (+8, +4398);

D: between the elt-5 ATG and the elt-6 ATG (+8, +4512);

E: between the elt-6 ATG and poly(A) site (+4518, +8158);

F: between the elt-6 ATG and poly(A) site (+4518, +8158);

G: 3.1 kb upstream of the elt-6 ATG (+1425, +4512).

The coding region of GFP (fragment H) was amplified from pGFPm5 (a gift from R. Zeller).

Translational fusions were made by inserting the coding region of GFP (fragment H) into the 5′ end of elt-5 or elt-6 in frame shortly after the ATG. Three transcriptional fusions (pKK41, pKK44 and pKK7) and six translational fusions (pKK39, pKK52, pKK38, pKK47, pKK25 and pKK49) containing various combinations of PCR fragments were made as follows (numbers in parentheses refer to the concentrations of injected DNA in μg/ml).

pKK39: A, H, D and E in pPD96.04 (2);

pKK52: A, H and C in pBluescript (50);

pKK38: A, D, H and E in pPD96.04 (2);

pKK41: A and D in pPD96.04 (50);

pKK44: G in pPD95.67 (50);

pKK7: A in pPD96.62 (20);

pKK47: A, H and E in pPD96.04 (50);

pKK25: B, H and E in pPD96.04 (50);

pKK49: nhr-74 promoter, H and F in pBluescript (50).

The GFP and β-galactosidase-coding regions in the vector pPD96.04 were replaced by PCR fragments in pKK39, pKK38, pKK47 and pKK25. The nhr-74 promoter in pKK49 was subcloned from the nhr-74::GFP construct obtained from P. Sengupta (Miyabayashi et al., 1999). All pPD constructs were gifts of A. Fire. Each reporter construct, as well as a mixture of two constructs, pKK25 and pKK49, was co-injected with pRF4 (rol-6^D, 200 μg/ml) into N2 hermaphrodites or with pDP#MM016B (unc-119(+), 200 μg/ml) into unc-119(ed4) hermaphrodites. Strains containing pKK25, pKK47 or pKK49 were crossed with the unc-119(ed4);jsIs1 strain to obtain strains carrying each of the constructs and JAM-1::GFP.

The translational fusions proved to be somewhat toxic. It was difficult to obtain lines with pKK38 and pKK39, two constructs that contain both the elt-5 and -6 genes, owing to embryonic or larval lethality caused by these sequences. To circumvent this problem, these constructs were injected at low concentrations. The lines obtained showed low overall levels of expression. pKK52, an elt-5 reporter construct, and pKK41, an elt-6 reporter construct, contained the elt-5- but not the elt-6-coding region. We were able to obtain several lines with these constructs that gave relatively high expression levels. The expression patterns of the two elt-5 reporters containing 3.4 kb upstream and the entire elt-5 gene, pKK39 and pKK52, appeared identical except that pKK52, but not pKK39, showed low levels of expression in the descendants of the vulval precursor cells (K. K. and J. H. R., unpublished). In addition, two elt-6 reporters, pKK41 and pKK44, showed occasional expression in some larval and adult intestinal cells, which was not observed with any other reporter (not shown).

Anti-ELT-5 antibodies were raised against ELT-5-specific peptides, pep5A (EDPMDQDVKQEESERSDIPTC) and pep5B (TETRPESAEQQHHEC), and anti-ELT-6 antibodies were raised against ELT-6-specific peptides, pep6A (RKRKPTKESVNRHLEC) and pep6B (CLEQMSESGSEEKYP; custom synthesized by Sigma Genosys). The peptides were selected from regions of little or no homology to minimize potential crossreactivity. Two rabbits, UCSB47 and UCSB48, were immunized against the ELT-5-specific peptides (performed by Cocalico Biologicals). Sera from the two rabbits gave essentially the same staining pattern. Anti-ELT-5 sera were used directly, as affinity purification did not improve staining. Two additional rabbits, UCSB49 and UCSB50, were immunized against the ELT-6-specific peptides. Sera from UCSB49 were used after affinity purification using a peptide-coupled column (AminoLink Plus kit by Pierce), according to the manufacturer’s protocol. Sera from UCSB50 gave similar but slightly weaker staining.

Embryos were fixed and stained for immunofluorescence as described (Sulston and Hodgkin, 1988). Anti-ELT-5 sera and affinity-purified anti-ELT-6 antibodies were used at a 1:25 to 1:100 dilution. Monoclonal antibody (mAb) MH27, which recognizes epithelial adherens junctions (Priess and Hirsh, 1986; Waterston, 1988), was a generous gift of R. Waterston.

Two non-overlapping regions of elt-5, corresponding to amino acids 58-215 and 224-363, were amplified by PCR from the cDNA clone yk474 and served as the templates for elt-5 dsRNAs. dsRNAs directed to both elt-5 regions gave the same phenotypes. The region of elt-6 corresponding to amino acids 94-286 was PCR amplified from the cDNA clone yk113 and served as the template for elt-6 dsRNA. RNA was synthesized and purified using the MEGAscript kit (Ambion) according to the manufacturer’s protocol. elt-5 or -6 dsRNA, or a mixture of the two, was injected into young hermaphrodites as described (Fire et al., 1998). Progeny laid at least 10 hours after injection were analyzed.

To obtain strong embryonic effects, elt-5 RNA was injected at 2 mg/ml. To obtain weaker effects, which allowed us to observe post-embryonic phenotypes, elt-5 RNA was injected at 0.2 mg/ml. Throughout the paper, ‘elt-5(RNAi)’ animals refers to animals treated with high levels of elt-5 dsRNA.

All Nomarski and some fluorescence images were acquired with an Optronics VI-470 camera on a Nikon Microphot SA microscope. Other fluorescence images were acquired with a BioRad 1024 Confocal Microscope.

In an effort to learn how the C. elegans epidermis becomes patterned into the three major epidermal cell types during embryogenesis, we have attempted to identify factors that impart seam cell-specific identity (Terns et al., 1997). Because GATA factors, namely ELT-1 and -3, are involved in other aspects of epidermal development, we examined several GATA factor-encoding genes predicted from the C. elegans genomic sequence (C. elegans Sequencing Consortium, 1998) for a possible role in embryonic seam cell development. We found that RNA-mediated interference (RNAi) of a GATA factor-encoding gene we have named elt-5 (erythroid-like transcription factor 5) results in penetrant embryonic and early larval lethality and causes morphological defects that include the lack or malformation of seam-specific cuticular specializations, called alae. This observation led us to investigate the role of elt-5 and its paralog, elt-6 in seam cell development.

elt-5 and elt-6 are adjacent genes encoding single-finger GATA factors. Their encoded DNA binding domains are 76% identical and are ∼60% identical to those in other GATA factors (Fig. 1). As is typically the case for GATA factors, ELT-5 and -6 are not significantly similar to the other GATA factors outside the DNA-binding domains. However, the two proteins are 46% identical overall, implying that one of the genes arose by duplication of the other. The putative polyadenylation site of the upstream gene, elt-5, and the trans-splice site of the downstream gene, elt-6, are separated by only ∼130 base pairs, characteristic of genes that reside on the same operon (Blumenthal and Steward, 1997). However, attempts to determine whether the two genes indeed form an operon have not produced conclusive evidence. As most, if not all, downstream genes are trans-spliced to the SL2 leader (Spieth et al., 1993; Zorio et al., 1994), we looked for, but failed to find, evidence for SL2 trans-splicing of transcripts from the downstream gene, elt-6 (see Materials and Methods). All seven elt-6 cDNA clones examined were either trans-spliced to the SL1 leader or were not trans-spliced. This suggests that elt-6 is often transcribed from its own transcription initiation site near the SL1 splice site rather than co-transcribed with elt-5 as a dicistronic transcript. This interpretation is consistent with data obtained with a reporter construct (see below). However, based on the effects of elt-5 dsRNA on reporter constructs, it seems likely that at least a fraction of mature elt-6 message is generated from elt-5/elt-6 dicistronic transcripts (see below).

We used the technique of RNAi (Fire et al., 1998; Guo and Kemphues, 1995) to assess the developmental function of elt-5 and -6. We found that nearly all (90/95) progeny of hermaphrodites injected with high levels of elt-5 dsRNA (see Materials and Methods) arrest late in embryogenesis (pretzel stage) or as early L1 larvae. The arrested L1 larvae are invariably uncoordinated (Unc phenotype), lumpy (Lpy phenotype) and slightly dumpy (Dpy phenotype), suggesting defects in epidermal development (Fig. 2B). In addition, the entire buccal capsule, the cuticular structure of the mouth (Wright and Thomson, 1981), invariably fails to attach to the anteriormost region of the head (‘pharynx unattached’ or Pun phenotype; Fig. 2B), suggesting defects in the buccal epidermis. A small fraction (5/95) of embryos arrest earlier, apparently with ruptures at the head or ventral midline (not shown).

Injection of elt-6 dsRNA at high levels did not produce any observable phenotype, and co-injection of both elt-5 and elt-6 dsRNAs at high levels did not result in an enhanced phenotype compared with elt-5 dsRNA alone. These results, however, do not necessarily indicate that the observed phenotypes are due to elimination of elt-5 function alone. In fact, elt-5 dsRNA affects expression of both elt-5 and -6 in seam cells (see below), and the observed phenotypes of elt-5(RNAi) animals may arise from inhibition of both genes or of elt-5 alone. For simplicity, we will use the notation elt-5/6 to refer to the function of either the elt-5 gene alone or of both the elt-5 and -6 genes. Based on its effects on reporter gene constructs, elt-5 dsRNA at high levels appears to abolish its function (as well as elt-6 function in some tissue types; see below); thus, these phenotypes are likely to reflect a strong loss-of-function or null phenotype.

To investigate a possible post-embryonic role for elt-5, we injected hermaphrodites with lower levels of elt-5 dsRNA. Such injections resulted in a mixture of weakly and strongly affected progeny. Weakly affected larvae appeared normal at hatching; they were neither Lpy nor Unc, and their buccal capsules were properly attached. Many of these larvae, however, became lethargic and sickly at later stages, were molting defective, and arrested at various stages of larval development. A small fraction of the weakly affected larvae also showed other gross morphological abnormalities that were suggestive of epidermal defects, including Roller (Rol) and protruding vulva (Pvl) phenotypes (not shown). The post-embryonic developmental defects were examined in more detail, as described later. All elt-5(RNAi) animals described in the remainder of this paper were obtained from mothers injected with high levels of elt-5 dsRNA.

The gross phenotypes we observed suggested defects in epidermal structure and/or development. To characterize the epidermis in elt-5(RNAi) embryos, we visualized their epithelial adherens junctions with monoclonal antibody MH27 (Priess and Hirsh, 1986; Waterston, 1988). In a wild-type embryo, all epidermal cells, including the row of ten lateral seam cells on each side, are clearly outlined by MH27 staining (Fig. 2C). In contrast, although most of the epidermal pattern appeared normal in elt-5(RNAi) embryos, the rows of seam cells often showed gaps in MH27 staining (Fig. 2D). In addition, seam cells were occasionally displaced from the linear row of lateral cells. For example, Fig. 2F shows an elt-5(RNAi) embryo in which a seam cell, V1 (asterisk), is ventrally misplaced such that neighboring seam cells, H2 and V2, contact each other (compare with Fig. 2E). The misalignment and gaps in the pattern were never observed in P cells.

There are at least four possible explanations for the gaps in the seam rows seen in elt-5(RNAi): (1) seam cells are misspecified as non-epidermal cells at birth; (2) seam cells are misspecified as epidermal syncytial cells at birth, leading them to fuse with other syncytial epidermal cells; (3) seam cells, although correctly specified initially, lose their identity and later adopt a syncytial-type identity; or (4) seam cells retain their seam identity, but fusion is misregulated. (For simplicity, cells that normally become seam cells in wild-type will be called seam cells, irrespective of their ultimate identity in elt-5(RNAi) animals.) In an effort to distinguish between these possibilities, we examined the epidermal pattern over time using the JAM-1::GFP marker, which reveals the MH27 adherens junction pattern in living embryos (Mohler et al., 1998). These studies demonstrated that the gaps in the seam row probably result from fusion of existing seam cells with the surrounding epidermal syncytium. In wild-type embryos, most dorsal and ventral syncytial cells complete their fusion between the 1.5- and twofold stage of elongation (Podbilewicz and White, 1994). In elt-5(RNAi) embryos at the same stages, only occasional lateral cells lacked JAM-1::GFP expression. By hatching, however, many lateral cells (32%, n=61 larvae) lacked the adherens junction marker. This progressive disappearance of the adherens junction marker from the lateral row results from cell fusion; we were able to observe ongoing dissolution of the adherens junctions between individual lateral cells and the adjacent epidermal syncytium as they were caught in the act of fusion (see Fig. 4H, arrowhead). These ongoing fusions were observed in some cases as late as in newly hatched animals.

The relatively late onset of fusion (i.e. after the time that the normal epidermal syncytial cells fuse) (Podbilewicz and White, 1994) suggests that many seam cells that ultimately fuse with surrounding syncytia are initially specified correctly in elt-5(RNAi) embryos. This notion was supported by observing expression of SCM, a seam cell-specific marker (Terns et al., 1997). In wild-type embryos, SCM is expressed in all seam cells from the twofold stage through adulthood (Fig. 2G). We found that SCM expression is visible in all seam cells of elt-5(RNAi) embryos at the threefold stage (n=52), although arrested embryos and larvae show somewhat reduced expression (not shown). These observations revealed that seam cells are not misspecified as syncytial cells at the time of their birth in elt-5(RNAi) embryos. Consistent with the view that cells that have been specified as seam cells subsequently fuse with the syncytium, we often observed several nuclei in the epidermal syncytium, surrounding the region in which a seam cell had fused, ‘ectopically’ expressing SCM, albeit at low levels (Fig. 2H, arrowheads). Presumably fusion of SCM-expressing seam cells into neighboring syncytia allows release of some GFP molecules, which are taken up by nearby syncytial nuclei.

Morphological observations indicated that seam development in elt-5(RNAi) animals is abnormal even in the seam cells that do not fuse. Seam cells normally produce the alae, bilateral ridges of specialized cuticle superjacent to the seam cells of L1 larvae, dauer larvae, and adults. Wild-type alae are clearly evident at the L1 stage as two parallel ridges running along the body on each side (Fig. 3A,E). We found that the alae in elt-5(RNAi) larvae were invariably missing or malformed: 86% of elt-5(RNAi) larvae showed no visible alae, and 14% had partial and/or defective alae (n=56). For example, Fig. 3B,F show an elt-5(RNAi) larva that lacks alae over most of its length and that contains a stretch of ala with several irregularly shaped branches. By correlating the position of missing alae with the JAM-1::GFP pattern, it was evident that seam cells failed to produce alae irrespective of whether or not they had fused (Fig. 3D).

We assessed the range of seam cell characteristics that require elt-5/6 by analyzing several markers of seam-specific fate. As noted earlier, SCM, a marker of seam fate, is expressed in the seam cells of elt-5(RNAi) embryos, implying that seam-specific differentiation is initiated in these mutants. Eight genes that encode nuclear hormone receptors (NRs) are also apparently expressed exclusively in seam cells (Miyabayashi et al., 1999). We found that expression of reporters for some, but not all of these NR genes was diminished or abolished in elt-5(RNAi) embryos (Table 1). Expression of three NR reporters, nhr-75::GFP, nhr-81::GFP, and NHR-82::GFP, was undetectable in all elt-5(RNAi) embryos examined (Fig. 4A,B). In contrast, expression of two NR reporters, nhr-73::GFP and nhr-74::GFP, was only slightly affected, both in terms of the fraction of expressing animals and the level of GFP signal (Fig. 4C,D). The remaining three NR reporters, nhr-72::GFP, nhr-77::GFP, and nhr-89::GFP, gave intermediate results; these were expressed much less frequently in elt-5(RNAi) embryos than in wild-type embryos, and expression was barely detectable in only a few cells for those embryos showing any expression (Fig. 4E,F).

The foregoing observations indicate that elt-5/6 is essential for many, but not all, aspects of seam cell differentiation. To investigate the possibility that they may also participate in specifying seam identity, we examined the expression of the elt-3::GFP reporter, which accurately reflects expression of endogenous ELT-3 (Gilleard et al., 1999). At the 1.5-fold stage, elt-3::GFP is expressed in all major epidermal cells except seam cells (Fig. 4G). In contrast, we found that seam cells often express elt-3::GFP ectopically in elt-5(RNAi) embryos (Fig. 4H). As most seam cells in elt-5(RNAi) embryos are still unfused at this stage, the expression of elt-3::GFP cannot be simply the consequence of seam fusion, but apparently reflects a partial transformation in fate of seam cells into non-seam cells. In nearly all (19/20) 1.5-fold stage embryos examined in detail, at least one unfused seam cell, and an overall average of 27% of the unfused seam cells, expressed elt-3::GFP. Of particular note, the V3 seam cell was the most likely to express elt-3::GFP (80% of V3s examined). However, V3 was not the most likely to fuse. One explanation for this behavior is that, of all the seam cells, V3 is the most closely related to P cells: its sister and all its cousins are P cells (Sulston et al., 1983). It is therefore possible that in the absence of elt-5/6, V3 often adopts the fate of its sister and cousins (i.e. the P cell fate) and therefore both expresses elt-3 and remains unfused.

These findings indicate that elt-5/6 is required to maintain the identity of seam cells. Its role in repressing cell fusion may reflect conversion to a syncytial epidermal fate in the absence of seam-specifying information.

To assess the expression patterns of elt-5 and -6, we created several transcriptional and translational reporter constructs (Fig. 5). An elt-5 translational fusion construct (pKK52), in which the elt-5 promoter drives expression of a GFP::ELT-5 fusion protein, shows a complex and dynamic expression pattern that can be divided into three major components. First, pKK52 expression begins at the 28-cell stage in all four granddaughters and 16 great-great granddaughters of the MS and AB founder cells, respectively (Fig. 6A); this expression continues in many, possibly all, of their descendants until around the time of hatching. Second, expression becomes more pronounced in seam cells about 1 hour after their birth. This seam expression remains strong throughout embryonic and larval development (Fig. 6B), but becomes slightly reduced in adults. Third, robust expression is also seen in several cells in the head region, at least some of which are cells in the nervous system (neurons and/or support cells), beginning at approximately the comma stage (Fig. 6B) and continuing through adulthood. For simplicity, we will refer to this component of the expression pattern as nervous system expression, although we have not determined the precise identity of these cells.

An elt-6 transcriptional reporter (pKK41) is expressed in the same groups of cells as the elt-5 translational reporter (pKK52), but the relative expression levels are different. Whereas the elt-5 reporter is strongly expressed in both seam cells and the nervous system during the comma through pretzel stages (Fig. 6B), the elt-6 reporter is strongly expressed only in the nervous system (Fig. 6E). Only weak expression of the elt-6 reporter is apparent in seam cells and in the AB and MS descendants during embryogenesis, but the seam expression becomes stronger during larval development (not shown). Strong expression of the elt-6 reporter in the nervous system continues throughout larval development (Fig. 6F).

To confirm the expression patterns obtained with the reporter constructs, we raised antibodies against peptides specific for ELT-5 and -6. For each protein, two peptides were selected from regions of little or no similarity between the two proteins. Anti-ELT-5 staining is readily detected in the nuclei of seam cells during mid- to late-embryogenesis (Fig. 6C,D). At these stages, many unidentified cells in the head region also stained, consistent with the pattern seen for the GFP reporters (Fig. 6B). This staining is eliminated in elt-5(RNAi) embryos (not shown). We have been unable to obtain consistent and reliable staining of early embryos, larvae and adults, and therefore have not confirmed the reporter expression pattern at these stages. Anti-ELT-6 staining is most readily seen in several cells in the head and is faint in seam cells (Fig. 6G,H), consistent with the GFP reporter data. elt-6 dsRNA eliminates all nuclear staining (not shown).

As described above, despite the apparent operon-like organization of the elt-5 and -6 genes, we could not find evidence that elt-6 cDNA is trans-spliced to SL2. Previous studies have shown that some downstream genes in apparent operons are transcribed monocistronically under the control of their own promoters (Gilleard et al., 1997). To test whether elt-6 can be transcribed monocistronically, we constructed an elt-6 reporter (pKK44), which includes the unusually large (>2 kb) last intron and the last exon of elt-5 upstream of the elt-6 ATG (see Fig. 5). pKK44 is expressed strongly in some cells in the nervous system but is expressed in neither seam cells nor in early AB and MS descendants (not shown). Because pKK44 lacks the 5′ end and the first three exons of elt-5, this result indicates that elt-6 messages can be transcribed alone.

The result with pKK44 also suggests that enhancer sequences for seam, AB and MS expression are separable from those for expression in the nervous system. Indeed, we found that an elt-5 transcriptional reporter (pKK7) that includes only the 3.4 kb sequences upstream of the elt-5 ATG (see Fig. 5) is expressed in the early AB and MS lineages and in seam cells, but is not expressed in the nervous system. This pattern is complementary to the expression pattern of pKK44, demonstrating that separable enhancer regions regulate elt-5 and -6 expression in different groups of cells. Whereas enhancers for seam, AB and MS are contained in the 3.4 kb region upstream of elt-5, those for the nervous system reside in the 3.1 kb region upstream of elt-6 (perhaps within the last, large intron of elt-5).

To determine the effectiveness and specificity of the elt-5 and -6 dsRNAs, we injected them into strains carrying various elt-5 and -6 GFP reporter genes (Table 2). Expression of an elt-5 translational fusion (pKK39) was reduced to undetectable levels in elt-5(RNAi) embryos, suggesting that RNAi results in a strong loss-of-function or null phenotype. Furthermore, although elt-6 dsRNA did not cause an observable phenotype, we confirmed that elt-6 dsRNA was also effective; expression of an elt-6 translational fusion (pKK47) was eliminated by elt-6 dsRNA. In contrast, consistent with only moderate similarity between the two genes (∼60% identity overall), elt-5 dsRNA did not significantly affect expression of the elt-6 translational fusion, pKK47, which lacks the elt-5 coding region, and elt-6 dsRNA likewise did not alter expression of an elt-5 fusion (pKK39).

Previous studies have suggested that in some cases RNAi targeted against one gene in an operon can inhibit expression of another in the same operon (Bosher et al., 1999). We therefore explored the possibility that elt-5 dsRNA interferes with expression of both elt-5 and -6 and found that, indeed, elt-5 dsRNA blocks expression of the elt-6 fusion pKK41 (which contains the elt-5-coding region) in seam cells and in early AB and MS descendants, but not in the nervous system. (This probably does not relate to the relative insensitivity of the nervous system cells to RNAi (Tavernarakis et al., 2000), as we were able to eliminate expression of elt-6 in these cells using elt-6 dsRNA (data not shown).) Because pKK41 does not include any portion of elt-6, this is not due to cross-hybridization: instead, this result implies that elt-5 dsRNA eliminates expression of elt-6 in some (although not in all) cell types. However, elt-6 dsRNA did not affect expression of an elt-5 fusion (pKK39) that includes both elt-5- and elt-6-coding regions, suggesting that a fraction of the elt-5 transcripts are monocistronic. In summary, our results suggest that in seam cells and in early AB and MS descendants elt-5 is either transcribed alone or co-transcribed with elt-6, while in the nervous system elt-5 and -6 are each transcribed monocistronically (although there may be some dicistronic transcripts in the nervous system as well). Collectively, our findings suggest that the apportionment of monocistronic versus dicistronic transcription of this pair of genes is regulated in a tissue-specific manner.

Given the strong similarity of the ELT-5 and -6 proteins, we explored the possibility that they are functionally interchangeable. To test whether elt-6 can rescue the absence of elt-5, we created a GFP::ELT-6 fusion, driven by the 3.4 kb upstream region of elt-5 (pKK47, Fig. 5), which allowed us to drive expression of ELT-6 in seam cells and AB and MS descendants even in the presence of elt-5 dsRNA. Most elt-5(RNAi) animals expressing pKK47 (52 of 57 animals carrying the construct, as assessed by GFP expression) appeared wild type at hatching, and many (48/57) grew up to be viable adults. These rescued animals appeared to develop normally with the exception that many of them lacked a vulva (not shown), as will be described in more depth in a separate publication. These results demonstrate that elt-5 and -6 are functionally interchangeable through most of development and that expression of ELT-6 in seam cells and in AB and MS descendants is sufficient to rescue elt-5(RNAi) animals to viability.

We next asked whether the elt-5(RNAi) phenotypes we observed were attributable to the activity of the gene in seam cells, in AB and MS descendants, or both. Because all seam cells are derived from the AB lineage, it is not possible to completely separate the two components. We therefore created two rescuing constructs, pKK25 and pKK49, with partially overlapping, but largely complementary expression patterns (Fig. 5). In construct pKK25, expression of a GFP::ELT-6 translational fusion is driven by the 1.2 kb upstream promoter of elt-5. This construct is expressed in AB and MS descendants from early embryogenesis until about the time of hatching and is expressed in seam cells during embryogenesis only as part of the broad AB expression pattern. By the late L1 stage, however, no GFP is detectable. We created another construct, pKK49, in which the GFP::ELT-6 fusion protein is driven by the nhr-74 promoter. This promoter, which is expressed only in seam cells from about the comma stage through adulthood (Miyabayashi et al., 1999), was chosen because it is not significantly affected by elt-5 dsRNA (Table 1). The sum of the expression patterns of the two constructs closely resembles that of pKK47, the GFP::ELT-6 fusion construct containing the 3.4 kb elt-5 promoter. Indeed, we found that most (31/44) elt-5(RNAi) embryos carrying both pKK25 and pKK49 were rescued and grew up to become fertile adults.

While each of these constructs alone was not sufficient for viability, they allowed us to assess the temporal and tissue requirements for elt-5/6. All (n=54) elt-5(RNAi) animals expressing GFP::ELT-6 only in seam cells (from construct pKK49) arrested as embryos or larvae by the early L1 stage, the stage at which elt-5(RNAi) animals normally arrest. The arrested animals showed the Pun phenotype (Fig. 7A), which is presumably the cause of lethality. However, the alae defects normally seen in elt-5(RNAi) larvae were rescued (31/33 larvae showed normal alae; Fig. 7B), implying a cell-autonomous function of elt-5/6 in alae formation. In addition, the Lpy phenotype was partially rescued: the animals were only slightly lumpy, indicating that elt-5/6 activity in seam cells contributes to proper morphogenesis.

In contrast, most (36/42) elt-5(RNAi) larvae expressing GFP::ELT-6 in AB and MS descendants (from construct pKK25) developed beyond the early L1 stage, but many (32/42) had missing or malformed alae. Those that developed beyond early L1 (n=36) arrested growth by the L3-L4 molt. Of particular significance, by the time these larvae arrested, nearly all of their seam cells had fused with the surrounding epidermal syncytia (Fig. 7C,D). Some larvae (5/36) showed no visible seam cell boundaries at all, and the rest contained between one and four seam cells with distinct boundaries (overall the mean number of unfused seam cells in 36 larvae was 1.8). In addition, the arrested larvae were defective in molting: though they apparently initiated the molting process, they were unable to shed their old cuticle completely (Fig. 7E,F). Many of these were still encased in the old cuticle, and because their mouth was blocked they were unable to eat; this is presumably the cause of lethality (Fig. 7E). The old cuticle was often wrapped around the body of these larvae, forming a constriction, and sometimes the space between the old and new cuticle filled up with waste material (Fig. 7F).

Taken together, the results obtained with these two rescuing constructs (pKK49 and pKK25) imply that elt-5/6 activity in seam cells is essential for proper alae formation, suppression of seam fusion and molting. They further suggest that early L1 lethality is attributable either to elt-5/6 activity in AB and MS descendants or to expression in very early seam cells, before pKK49 gives robust expression in seam cells ∼2 hours after their birth. Finally, they reveal an apparently continuous requirement for elt-5/6 in maintaining seam identity and/or repression of their fusion throughout post-embryonic development.

We have shown that ELT-5 and -6 (or ELT-5 alone) are required in seam cells to promote a number of differentiated seam cell characteristics. These include production of alae, expression of several seam-specific genes, repression of a gene that is expressed in non-seam epidermis and maintenance of seam cells in an unfused state. These factors appear to be part of a regulatory system that links ELT-1, which broadly specifies the general epidermal fate, to the genes imparting the unique characteristics to seam cells.

One of the conspicuous defects in elt-5(RNAi) animals is the inappropriate fusion of seam cells with the epidermal syncytia. Cell fusion occurs in many cell types in a tightly regulated fashion in C. elegans (Podbilewicz and White, 1994); however, its function is not well understood and only a small number of genes that regulate it have been found (Alper and Kenyon, 2001; Clark et al., 1993; Kenyon, 1986; Wang et al., 1993). Our results demonstrate that elt-5/6 plays a major role in repressing both embryonic and larval epidermal cell fusion. The later fusion of the seam cells into a single lateral syncytium at the L4 molt (Podbilewicz and White, 1994) must involve a mechanism that bypasses the fusion-repressive function of ELT-5/6. However, this repressive function presumably remains active in L4 larvae and adults, as the seam cells, while fusing with each other, remain separate from the other epidermal syncytia. Thus, the homotypic fusion of seam cells appears to involve a regulatory process distinct from that controlling fusion with the epidermal (hyp) syncytia.

Several findings suggest that the inappropriate fusion of seam cells with the epidermal syncytia is not simply the result of a failure to repress fusion per se. In particular, the failure of seam cells to express several seam-specific markers and the observation that elt-3::GFP is inappropriately expressed in unfused seam cells imply a (partial) loss of seam identity. Our finding that some seam-specific markers are largely unaffected, however, suggests that a combination of regulatory inputs from elt-5/6 and at least one other factor is required to specify and maintain the entire identity of seam cells. Such a factor might simply act redundantly with ELT-5/6, or it may activate a unique set of target genes that, together with the ELT-5/6 targets, contribute to the full suite of genes that define seam cell identity. It is noteworthy that the ELT-1 protein is present in major epidermal cells after their epidermal cell fate is specified, and that ELT-1 expression is stronger in seam cells than in syncytial or P cells (Page et al., 1997). One interesting possibility is that ELT-1 may function in seam cell specification or differentiation as well as in general epidermal cell specification. Alternatively, elt-5/6 may be essential for all aspects of seam cell specification and the residual seam characteristics observed in elt-5(RNAi) animals might be the result of incomplete inactivation of the elt-5/6 function. Although we have shown that ELT-5/6 expression is undetectable, based on both GFP reporter expression or immunostaining, it is possible that undetectably low levels of ELT-5/6 can provide partial function in elt-5(RNAi) animals.

A model for seam epidermal development in the C. elegans embryo is presented in Fig. 8. elt-1 is required to specify all major epidermal cells, which then adopt one of three fates: seam, P or syncytial. Our present results suggest that elt-5/6, together with an unknown factor (Factor X in Fig. 8), may direct seam cells to adopt their appropriate identity. We have found that a transcriptional elt-5::GFP reporter is expressed in many cells in elt-1(−) mutants (K. K. and J. H. R., unpublished), suggesting that at least the widespread AB and MS expression of the reporter does not require ELT-1; however, it seems likely that elt-5/6 expression in seam cells per se requires ELT-1. Although we have not tested this possibility directly, their respective mutant phenotypes support the hypothesis that elt-1 functions upstream of elt-5/6 in seam cells. (As suggested above, elt-1 may also function in parallel with elt-5/6 in seam cells.) Given its expression pattern, elt-3 may function in syncytial and P cell specification or differentiation. However, deletion of elt-3 does not produce any noticeable phenotype, suggesting pervasive functional redundancy of this gene (Gilleard and McGhee, 2001). ELT-3 and another factor (Factor Y in Fig. 8) may specify a general non-seam fate, and other factors may then apportion syncytial and P cell identities to the appropriate cells.

Specification of the epidermis and its subdivision into the different epidermal cell types by a sequential cascade of GATA factors is reminiscent of the regulatory events that specify the endoderm (E founder cell) and major mesoderm (MS founder cell) precursors in very early C. elegans embryos. In the latter case, a pair of redundant GATA factors, MED-1 and -2, dictate the identity of the EMS cell, and therefore both of its daughters, E and MS (Maduro et al., 2001). Following division of EMS, the fate of the E cell is, in turn, regulated by the END-1 and -3 GATA factors (Zhu et al., 1997; M. Maduro and J. H. R., unpublished). A cascade of three other GATA factors (ELT-2, -4, and -7) is subsequently activated in the E lineage (Fukushige et al., 1998; Fukushige and McGhee, personal communication; K. Strohmaier and J. H. R., unpublished). In contrast, it seems unlikely that additional GATA factors act downstream of ELT-3 and ELT-5/6 in the epidermis: the expression patterns of all 11 GATA factors predicted from the genomic sequence are now known and only ELT-1, -3, -5 and -6 are present in the epidermis.

We observed post-embryonic developmental defects when low doses of elt-5 dsRNA were used or when the embryonic defects seen at high doses were rescued by forced expression of ELT-6 in selected cells under a partial elt-5 promoter. Under the latter condition, almost all seam cells fuse with the surrounding syncytium during larval development, suggesting that ELT-5/6 may be continuously and non-redundantly required to specify or maintain the identity of seam cells post-embryonically.

We also found that elt-5(RNAi) animals rescued for embryonic or early L1-stage lethality by expression of ELT-6 under a partial elt-5 promoter had difficulty completing molting and arrested before reaching adulthood. These defects were rescued by additional, seam-specific expression of ELT-6. These results suggest that seam cells play an essential role in molting. Previous studies have identified two genes essential for proper molting. lrp-1 loss-of-function mutant larvae show molting problems similar to those we observed in elt-5(RNAi) larvae (Yochem et al., 1999). LRP-1 is similar to gp330/megalin, a member of the low-density lipoprotein receptor family. LRP-1 is expressed on the apical surface of the main body epidermal syncytium, hyp7, and genetic mosaic analysis has indicated that it is required in hyp7. RNAi of another gene, nhr-23, which encodes CHR3 (a nuclear hormone receptor), causes similar problems in molting (Kostrouchova et al., 1998). CHR3 is highly similar to Drosophila DHR3, which is induced by ecdysone, a molting hormone and plays an essential role in metamorphosis. CHR3 is expressed in all epidermal cells including seam cells during late embryogenesis and larval stages. It seems likely that the molting process in C. elegans larvae is regulated by a network of genes that acts in both seam and syncytial cells. It will be interesting to learn how elt-5 and -6 act in such a network and if any of the many known seam-specific nuclear hormone receptors also participate in the regulation of molting.

elt-5 and -6 are spaced ∼130 bp apart and are organized in an apparent operon; however, our results suggest that they do not form a typical operon. We could not find evidence that elt-6 messages are trans-spliced to SL2 or to other non-SL1 leaders. Most downstream genes in operons are trans-spliced to SL2 exclusively or to a mixture of SL1 and SL2 (Spieth et al., 1993; Zorio et al., 1994). A few exceptions in which the downstream genes are trans-spliced exclusively to SL1 involve pairs of genes with no intercistronic sequences (Williams et al., 1999). Further evidence that elt-6 need not be transcribed as part of a polycistronic transcript comes from our identification of an apparent transcription initiation site ∼120 bp upstream of the elt-6 trans-splice site. Nevertheless, it seems likely that a fraction of elt-6 messages are transcribed dicistronically. In support of this possibility, we found that elt-5 dsRNA inhibited expression of reporters for both elt-5 and -6 in seam cells, and AB and MS descendants (elt-6 dsRNA did not affect elt-5 reporter expression). Recent evidence suggests that for certain dicistronic genes, RNAi targeted against one gene can interfere with both (Bosher et al., 1999). Accumulation of dicistronic elt-5 and -6 pre-mRNA may allow elt-5 dsRNA to inhibit elt-6 gene activity. (The reason elt-6 dsRNA does not inhibit elt-5 gene activity may be because a substantial portion of elt-5 is transcribed monocistronically.) Intriguingly, such an effect would appear to be tissue specific: while elt-6 messages may be co-transcribed with elt-5 messages in seam cells and descendants of AB and MS, allowing elt-5 dsRNA to interfere with them, this does not appear to be the case in the nervous system. As ELT-6 is most strongly expressed in the nervous system, our inability to detect SL2 trans-splicing on elt-6 cDNAs may simply reflect the relative abundance of monocistronic versus dicistronic elt-6 transcripts; monocistronic transcripts may be the predominant form in the nervous system.

We found that the elt-5(RNAi) defects can be rescued by expression of elt-6, implying that the two gene products can function interchangeably. In addition, we have found that widespread expression of either ELT-5 or -6 can result in an excess of cells expressing a seam-specific marker (SCM; K. K. and J. H. M., unpublished), suggesting that either elt-5 or -6 alone may be sufficient to initiate seam differentiation in non-seam cells. It is unclear, however, if the two genes in fact perform redundant functions under normal conditions. The low level of elt-6 expression in seam cells and descendants of AB and MS may be insufficient to provide the essential functions in those cells. Although the two genes can apparently function interchangeably, they may be responsible for different activities depending on where they are most abundantly expressed, which might account for the preservation of both genes during evolution.

Although we have focused on the role of elt-5/6 in seam cell development, several findings indicate that these genes may have additional functions in other cell types. First, they are expressed in several different cell types. Second, we observed a vulvaless phenotype in elt-5(RNAi) animals rescued by GFP::ELT-6 (as will be described in a subsequent report). Third, seam-specific expression of GFP::ELT-6 using the nhr-74 promoter did not rescue all embryonic defects in elt-5(RNAi) animals. Although we cannot rule out the possibility that earlier expression of ELT-6 in seam cells than that directed by the nhr-74 promoter might fully rescue the embryonic elt-5(RNAi) defects, a plausible interpretation is that elt-5/6 performs an unknown essential function in early AB and MS descendants. It will be interesting to investigate whether elt-5/6 performs distinct functions in different types of cells, and if so, how specificity for particular cell types is achieved.

We have recently found that elt-5 corresponds to the previously identified gene egl-18 (K. K., C. G. Wood and J. H. R., unpublished).

We thank K. Foltz for helpful comments on the paper, M. Maduro for bringing the elt-5 and -6 sequences to our attention, E. Newman-Smith for help with initial RNAi experiments, J. Zhu for the embryonic cDNA pool, D. Braun for help with injections, and the other members of the Rothman laboratory for helpful discussions. We are grateful to M. Palfreyman and P. Sengupta for the seam-specific reporter constructs and strains, J. Gilleard for the elt-3::GFP strain, J. Hardin for the adherens junction marker JAM-1::GFP, A. Fire for expression and heat-shock vectors, and M. Fukuyama and J. Kasmir for wIs54. We thank Y. Kohara and the Worm Genome Consortium for providing clones and sequences, and the C. elegans Genetic Stock Center for strains. This work was supported by grants from the NIH (HD37487) and the March of Dimes to J. H. R.