The Caenorhabditis elegans GATA factor elt-1 is essential for differentiation and maintenance of hypodermal seam cells and for normal locomotion

The GATA transcription factors are a gene family encoding zinc-finger DNA-binding proteins that are involved in the regulation of differentiation and tissue-specific gene expression in a wide range of organisms. Remarkable levels of functional conservation have been observed between GATA factors of different phyla, such is the critical role they play in the control of development (Patient and McGhee, 2002). There are just six members of the GATA transcription factor family in vertebrates, all of which are essential and utilised multiple times during development (Patient and McGhee, 2002). Caenorhabditis elegans, although having a simpler body plan than vertebrates, has 11 family members, which have central roles in endodermal and epidermal development (Michaux et al., 2001). There is a large degree of redundancy in the C. elegans gene family. For example end-1 and end-3 (Zhu et al., 1997), med-1 and med-2 (Maduro et al., 2001) represent mutually redundant gene pairs and elt-3 (Gilleard et al., 1999), elt-7 and elt-4 (Fukushige et al., 2003) appear not to be essential for normal development in the laboratory. Only three of the family members appear to be essential genes in their own right: elt-1, elt-2 and elt-5 (Fukushige et al., 1998; Koh and Rothman, 2001; Page et al., 1997).

The only C. elegans GATA factor with the classic double-GATA-type zinc-finger domain structure that characterises all the vertebrate GATA factors is elt-1; the rest of the C. elegans family are single-finger polypeptides (Spieth et al., 1991). The only known function of elt-1 is in the early C. elegans embryo where it triggers a partially redundant cascade of GATA factors that activate epidermal differentiation (Gilleard and McGhee, 2001; Koh and Rothman, 2001; Page et al., 1997). The elt-1 gene is expressed in all hypodermal precursors and is essential for the formation of most hypodermal cell types (Page et al., 1997). Embryos homozygous for elt-1(zu180), a nonsense mutation which leads to a truncated polypeptide lacking the carboxyl GATA-type zinc finger, arrest early in development with a failure to produce hypodermis (Page et al., 1997). Lineages that give rise to hypodermal cells show abnormalities in elt-1(zu180) mutants at least two cell divisions before hypodermal cells are born (Page et al., 1997). Hence elt-1 is necessary to specify the fates of nearly all hypodermal cells by virtue of an essential function during early embryogenesis. The elt-3 gene is another hypodermal-specific GATA factor whose expression is initiated in the dorsal and ventral, but not the lateral, hypodermis immediately after the cell division that gives rise to their formation (Gilleard et al., 1999). This gene is sufficient to activate hypodermal differentiation and is thought to act downstream of elt-1 in the dorsal and ventral hypodermis, although its function is apparently not essential under laboratory conditions (Gilleard and McGhee, 2001). The elt-5 and elt-6 genes are expressed in the lateral epidermis (seam) and together are essential both for the differentiation of seam cells and in preventing their premature fusion into the dorsal and ventral hypodermal syncytium during subsequent development (Koh and Rothman, 2001). In this paper, we investigate the function of elt-1 throughout development subsequent to its early role in the specification of hypodermal cell fates.

All C. elegans culture and genetic crossing methods were as previously described (Brenner, 1974). `Starvation arrested' L1 larvae were produced by alkaline hypochlorite treatment of gravid adult worms (Sulston and Hodgkin, 1988). Transformation of C. elegans by germline microinjection and the detection of lacZ expression were performed using standard procedures (Fire et al., 1990; Mello and Fire, 1995). Programmed cell death was visualised using two independent techniques. First, TUNEL staining was performed as previously described (Wu et al., 2000) using the APO-Direct™ kit (Pharmingen). Second, a ced-1::GFP reporter gene was used to visualise cell corpses during engulfment. ced-1 is a transmembrane receptor that mediates cell engulfment following programmed cell death (Zhou et al., 2001). The pZZ67 plasmid is a reporter construct in which ced-1::GFP expression is under the control of the col-10 promoter and consequently allows visualisation of cell corpses during engulfment in the hypodermis. Plasmid pZZ67 (kindly supplied by Zeng Zhou, IT, Cambridge, MA) was co-injected with marker plasmid pRF4 into ced-5(n1812) IV; nuc-1(e1392) X hermaphrodites, a genetic background in which cell corpses persist, and transgenic extrachromosomal array lines JG175 and JG176 were established. Scanning electron microscopy (SEM) was performed as previously described (Thein et al., 2003).

The following worm strains and alleles were used: JR667 (wIs51) SCM::GFP, JR1000 (wIs51; jcIs1) SCM::GFP and ajm-1::GFP (supplied by Joel Rothman, University of California, Santa Barbara, CA); IA105 (ijIs12) dpy-7::GFP (supplied by Iain Johnstone, University of Glasgow, UK); TP12 (kaIs12) col-19::GFP (supplied by Tony Page, University of Glasgow, UK); NW1229 evIs111 F25B3.3::GFP (supplied by Joe Cullotti, Samuel Lunewald Research Institute, Toronto, Canada); VH15 (rhIs4) glr-1::GFP (supplied by Christina Schmid/Harold Hutter, Max-Planck-Institut, Heidelberg, Germany); GY401 (gly-4:GFP) (supplied by Fred Hagen, University of Rochester, NY); JM53 caIs4 hsp16-2::elt-1, JM55 caIs6 hsp16-2::elt-1, JM59 caIs10 hsp16-2::elt-3, JM60 caIs11 hsp16-2::elt-3 (supplied by Jim McGhee, University of Calgary, Canada); CB1392 nuc-1(e1392); MT8793 ced-5 (n1812); nuc-1 (e1392); NL2099 rrf-3 (pk1426); DR1567 daf-2 (m577); JG5 (vpIs1) elt-3::GFP; JG186 rrf-3 (pk1426); wIs51; JG190 daf-2 (m577); wIs51.

In order to determine the elt-1 5′ end, from which to design a reporter gene construct, the longest available elt-1 EST clone, yk397a9, was obtained from Yugi Kohara (National Institute of Genetics, Mishima, Japan). Sequencing of the insert revealed this transcript included an SL1 splice leader sequence and was 229 bp longer than the previously published elt-1 transcript (Spieth et al., 1991). The A of the initiator ATG of the yk397a9 transcript is at genomic position 9,615,139 (chromosome IV). An elt-1::GFP/lacZ reporter gene construct was made in which a 6123 bp elt-1 genomic fragment, extending 4380 bp upstream of the ATG, formed a translational fusion between the fifth exon of elt-1 and GFP/lacZ in the C. elegans expression vector pPD96-04. A fill-in PCR technique was used for the cloning. The 6123 bp elt-1 genomic fragment was PCR amplified from cosmid W09C2 using sense primer W04, 5′-acgtactgcagTTAGTCAAATAGACAAGCTGTATCG-3′ and antisense primer W03, 5′-tacctttgggtcctttggccaatcccGTTGTGTCGTTGGATAGAAGTAACTCGG-3′. W04 contained a Pst-1 restriction enzyme tag (lower case, restriction site underlined) and W03 included a sequence tag complementary to the sense strand immediately downstream of the SmaI site in vector pPD96-04 (lower case). A fill-in PCR was performed on the 6123 bp elt-1 fragment mixed with SmaI-digested pPD96-04 vector. This produced an in-frame fusion fragment which was then circularised by self-ligation following PstI digestion. The resulting construct pelt-1gfp2.17 was injected at 100 ng/μl into DP39 unc-119(ed4) adult hermaphrodites along with the rescuing plasmid pDP#MM016B (Maduro and Pilgrim, 1995) at 20 ng/μl. Several transgenic lines were generated carrying extrachromosomal arrays. The transgene was chromosomally integrated using γ-irradiation to produce three independent lines JG31 unc-119(ed4) vpIs7[pDP#MM016B, pelt-1gfp2.17], JG32 unc-119(ed4) vpIs8 [pDP#MM016B, pelt-1gfp2.17] and JG33 unc-119(ed4) vpIs9 [pDP#MM016B, pelt-1gfp2.17]. All three lines had indistinguishable expression patterns.

Two elt-1 RNAi `feeding' constructs were made that contained non-overlapping fragments corresponding to different regions of the elt-1 cDNA transcript. For the first construct, a 1179 bp EcoRI fragment was released from the pyk397a EST clone (–22 bp to +1157 bp) and cloned into EcoRI-digested L4440 `RNAi feeding' vector to produce plasmid pPM41. For the second construct, a 448 bp fragment was amplified from pyk397a with primers elt1RNAi5, 5′-ATGGAATTCAGACCCGTAATCG-3′) and elt1RNAi6, 5′-acgtctcgagTATATCACAGAAATATGAGAGG-3′. The sense primer, elt1RNAi5, spanned the EcoRI site (underlined) in the elt-1 cDNA sequence and the antisense primer, elt-1RNAi6, had a 5′ tag (lower case) containing an XhoI site (underlined). EcoRI/XhoI double digestion resulted in a 421 bp fragment (from +1157 bp to +1578 bp), which was cloned into the corresponding sites of the L4440 vector to produce construct pPM88. Plasmids pPM41 and pPM88 were separately transformed into competent E. coli HT115 (DE3) cells. Feeding plates for elt-1 RNAi were produced by seeding a bacterial lawn on NGM culture plates containing 2 mM isopropyl-1-thio-β-D-galactopyranoside (IPTG) and 50 μg/ml ampicillin to induce expression of dsRNA. Negative-control plates contained 2 mM IPTG and 50 μg/ml ampicillin and were seeded with HT115(DE3) cells carrying the vector L4440 alone. To investigate the embryonic effects of elt-1 RNAi, adult worms were placed on elt-1 RNAi feeding plates at 20°C and the phenotype of embryos produced was examined at regular time points from the start of feeding. To investigate the postembryonic effects of elt-1 RNAi, starvation-arrested L1 larvae were placed onto elt-1 RNAi and negative-control feeding plates, allowed to develop and examined at regular intervals.

A 797 bp fragment corresponding to the elt-1 cDNA sequence (+151 bp to +948 bp) was amplified from pyk397a with primers elt1RNAi1, 5′-acgtgagctcTTGATCCTGACACAAACTCCATC-3′ and elt1RNAisst1, 5′-actggagctcGTAGAGGCCGCATGCGTTGCAGAGG-3′). Both primers included a 5′ tag (lowercase) that contained a SacI restriction site (underlined). The fragment was cloned in both forward and reverse orientations into the SacI site of pPD49-78 to produce constructs pPM46 and pPM48 respectively. These two plasmids were co-injected into C. elegans hermaphrodite gonads at a concentration of 20 ng/μl each along with co-injection marker ttx-3::GFP (Hobert et al., 1997) at 50 ng/μl to produce transgenic lines JG136 and JG137. These lines contain extrachromosomal arrays in which RNA is transcribed from the 797 bp elt-1 fragment in both forward and reverse orientation under the control of the hsp-16-2 heat-shock promoter to allow induction of elt-1 dsRNA in response to heat shock.

Previous immunolocalisation studies have shown that ELT-1 is expressed in hypodermal precursor cells from the 28-cell stage of embryogenesis and in all major hypodermal cells shortly after they are born (Page et al., 1997). During early morphogenesis the level of ELT-1 protein expression declines in the dorsal and ventral hypodermis but remains higher in the lateral seam cells. However, the expression of ELT-1 after the comma stage of embryogenesis has not been examined. To determine the expression pattern of the elt-1 gene during later embryogenesis and postembryonic development, a reporter construct was made in which gfp::lacZ expression was placed under the control of 4380 bp of elt-1 5′ flanking sequence. Reporter gene expression was first seen around the 20-30 cell stage and was maintained during subsequent early embryogenesis in cells corresponding to hypodermal precursor cells (Fig. 1A). Just prior to morphogenesis, reporter gene expression could be seen in all the major hypodermal cell nuclei. During formation of the comma stage, embryo reporter gene expression declined in the dorsal and ventral hypodermis but remained high in the lateral seam cells (Fig. 1B). Hence, the expression of the reporter gene during early embryogenesis exactly matches the previously described ELT-1 immunolocalisation studies. The seam-cell expression of the reporter gene continued throughout morphogenesis to the pretzel stage although at lower levels. In addition, at the threefold stage of development high levels of expression were seen in a group of cells adjacent to the pharynx in a position consistent with neuronal cells of the retrovesicular ganglion (Fig. 1C). During postembryonic development relatively low levels of expression were maintained in seam cells from the L1 to the adult stage (Fig. 1D) as were high levels of ELT-1 in neuronal cells of the retrovesicular ganglion (Fig. 1E,F,G). ELT-1 expression was also clearly apparent in several different groups of neuronal cell bodies and axons; those of the ventral cord were the most prominent (Fig. 1E,F). Expression was also seen in several neurons in which the axon extended from the ventral cord and encircled the pharynx just in front of the posterior bulb (Fig. 1G).

The reporter gene also showed postembryonic expression in sex-specific structures. In hermaphrodites, ELT-1 expression was seen in the vulval muscles (vm1 and vm2) (Fig. 1H) and in males, in a subset of the lateral seam, the SET cells, that give rise to the sensory rays (Fig. 1I).

RNAi was used to investigate possible roles for elt-1 during embryogenesis in addition to its early function in cell fate specification. Two independent, non-overlapping elt-1 cDNA fragments were used to express elt-1 dsRNA in the bacterial strain HT115(DE3) and the RNAi phenotypes obtained were precisely the same for both fragments. Embryos produced by adult hermaphrodites fed on elt-1 RNAi plates for more than 10 hours produced an early arrest phenotype at very high penetrance (>90%). This RNAi phenotype was a precise phenocopy of elt-1(zu180) homozygotes, which arrest as early embryos with a maximum of 16 major hypodermal cells as opposed to the 71 present in wild-type embryos (Fig. 2A,B). In order to observe weaker loss-of-function phenotypes, embryos laid earlier after the onset of elt-1 dsRNA feeding were examined. A proportion of embryos laid between 6-10 hours developed to the L1 stage with lumpy-dumpy (Lpy-Dpy) phenotype (Fig. 2C). Hypodermal cells were examined in these Lpy-Dpy larvae using ajm-1::GFP and SCM::GFP markers that allow adherens junctions in hypodermal cell membranes and the nuclei of lateral seam cells to be visualised respectively (Fig. 2D) (Mohler et al., 1998; Terns et al., 1997). In elt-1 RNAi Lpy-Dpy larvae, there were gaps in the line of seam cells delineated by the ajm-1::GFP with an associated absence of SCM:GFP nuclear expression. In severely affected individuals, the remaining seam cells appeared to be disorganised and misaligned, often with weak or absent SCM:GFP expression (Fig. 2E). Examination of these larvae by Normarski optics revealed regions where the alae were missing, supporting the conclusion that some seam cells were absent in these larvae (data not shown). Careful examination of the Lpy-Dpy larvae did not reveal any nuclei expressing the SCM:GFP marker that lacked a surrounding ajm-1::GFP fluorescing membrane. Hence, no evidence of abnormal fusion of seam cells into the hypodermal syncytium was detected (Koh and Rothman, 2001).

The cells of the dorsal and ventral hypodermis were examined in elt-1 RNAi Lpy-Dpy larvae using the dpy-7::GFP marker expressed from the ijIS12 transgene. This particular dpy-7::GFP reporter strain expresses GFP at very high levels in all hypodermal cells of the L1 (dorsal, ventral and lateral seam cells) with minimal mosaicism (Fig. 2F). Examination of numerous elt-1 RNAi Lpy-Dpy larvae revealed that there was minimal disruption of the dpy-7::GFP expression pattern (Fig. 2G). The cells of the dorsal and ventral hypodermis appeared unaffected except for a minor degree of misalignment because of the overall change of larval morphology. By contrast, gaps in the line of lateral seam-cell nuclei were visible with this marker as with the SCM:GFP and ajm-1::GFP markers (Fig. 2G). The mean number of dpy-7::GFP-positive nuclei in elt-1 RNAi Lpy-Dpy larvae was 65.4 (n=12; s.d.=4.1) compared with a mean of 70.2 (n=12; s.d.=1.3) in wild-type controls. The numerical reduction in dpy-7::GFP-expressing cells can be entirely accounted for by the observed seam-cell deficit and so there was no detectable loss of dorsal/ventral hypodermal cells. Therefore the elt-1 RNAi weak loss-of-function phenotype leads to a marked reduction of seam cells but no associated loss of dorsal/ventral hypodermal cells.

The GATA transcription factor elt-3 is expressed in dorsal and ventral hypodermal cells, but not the seam cells, immediately after they are born and is a potential downstream target of elt-1 (Gilleard and McGhee, 2001). Ectopic expression of elt-1 is sufficient to activate elt-3::GFP expression but because the majority of hypodermal cells in elt-1(zu180) mutant embryos fail to be specified, it has not been previously determined whether elt-1 function is necessary for the activation of elt-3 expression in differentiated hypodermal cells. Consequently we examined the expression of an elt-3::GFP reporter gene in elt-1 RNAi Lpy-Dpy larvae in which the number and organisation of dorsal/ventral hypodermal cells is normal as demonstrated by the dpy-7::GFP expression pattern. RNAi elt-1 Lpy-Dpy larvae showed greatly reduced levels of elt-3::GFP expression in the majority of dorsal/ventral hypodermal cells compared with negative controls (Fig. 2H,I). This suggests that elt-1 has an essential role in the activation of elt-3 expression in dorsal/ventral hypodermal cells after its early role in the general specification of hypodermal cell fates.

In order to investigate the function of elt-1 during postembryonic development, L1 hatchling larvae were placed on lawns of HT115(DE3) bacteria expressing elt-1 dsRNA. Development from L1 to L4 progressed normally but many adult worms were flaccid, often with the gonad herniating through the vulva (Fig. 3A). The vulva of such worms was anatomically normal as judged by careful observation of differential-interference microscopy and SEM (Fig. 3B,C). A time course performed on worms carrying the kaIs12 col-19::GFP adult-specific marker (Thein et al., 2003) revealed that worms lost structural integrity approximately 2-4 hours after the L4 to adult moult (Fig. 3D). This phenotype occurred when elt-1 RNAi feeding was initiated at the L1, L2 and L3 stages but not when initiated at the late L4 or adult stages. The same phenotype was produced 2-4 hours after the L4-adult moult when elt-1 RNAi was applied to the rrf-3(pk1426) RNAi-hypersensitive mutant strain (Simmer et al., 2002). RNAi of elt-1 was also applied by inducible expression of a 797 bp elt-1 dsRNA fragment from a transgene in strains JG136 and JG137. When a 2 hour heat shock was applied to synchronous cultures of transgenic worms at the L1 to L3 stages, the `burst-vulva' phenotype was observed shortly after the L4-adult moult but not when the heat shock was applied at the late L4 or adult stages.

The effect of elt-1 RNAi on the development and integrity of the different tissues in which the elt-1 reporter gene is expressed was examined using a number of different GFP markers. There was no discernable abnormality of the ventral nerve cord, retrovesicular ganglion or vulval muscles as visualised by the expression of glr-1::GFP (Wacker et al., 2003), F25B3.3::GFP (Altun-Gultekin et al., 2001) and gly-4::GFP (Fred Hagen, unpublished observation), respectively. By contrast, application of elt-1 RNAi during postembryonic development resulted in a loss of SCM::GFP nuclear-localised seam-cell marker expression (Terns et al., 1997) (J. Rothman, unpublished observation) (Fig. 4). There was an associated loss of ajm-1::GFP expression, which allows visualisation of hypodermal cell membranes (Mohler et al., 1998) (Fig. 4F) and examination under differential-interference microscopy confirmed loss of seam-cell nuclei. When elt-1 RNAi was applied from the L1 stage onwards, seam-cell loss was first apparent in L2 larvae and was most obvious in the rrf-3(pk1426) mutants (Fig. 4A). The total number of seam cells continued to decline throughout subsequent development until very few seam cells were present in adult hermaphrodites (Fig. 4A-E). Application of postembryonic elt-1 RNAi also resulted in adult males with abnormal tail morphologies in which the bursal rays were either absent or stunted and examination of the SCM::GFP marker revealed seam-cell loss in adult males (Fig. 4H-J). The SCM::GFP marker is also expressed in the SET cells, which are derived from the V5 seam cell and give rise to the bursal rays (Sulston and Horvitz, 1977), and these cells were also significantly depleted by elt-1 RNAi (Fig. 4H).

Seam cells could disappear during postembryonic development in response to elt-1 RNAi by a number of processes. The possibility that seam cells were undergoing programmed cell death was examined in two ways: first, worms were examined by TUNEL staining following application of elt-1 RNAi; second, elt-1 RNAi was applied to a strain carrying a ced-1::GFP transgene, which allows the membranes of engulfed cell corpses to be visualised in the cytoplasm of the phagocytosing cell (Zhou et al., 2001). Both techniques were applied to N2 and ced-5(n1812); nuc-1(e1392) worms, a genetic background in which cell corpses persist (Wu et al., 2000), but no evidence of programmed cell death was observed in the lateral hypodermis of elt-1 RNAi worms. The fate of seam cells was also carefully monitored by simultaneously observing the SCM::GFP and ajm-1::GFP markers in a large number of elt-1 RNAi and control worms under high-power magnification. These markers allow the process of seam-cell division and fusion to be accurately monitored; during cell fusion the ajm-1::GFP becomes punctuate and the SCM::GFP gradually disappears. During normal development, seam cells divide once between each moult, the anterior daughter fuses to the hypodermal syncytium and the posterior daughter remains as the new seam cell. Despite careful examination of many elt-1 RNAi larvae, no evidence of inappropriate fusion of seam cells into the dorsal/ventral hypodermis was seen. This would have been visible as seam cells undergoing fusion prior to cell division or SCM::GFP-expressing nuclei being present in the main hypodermal syncytium as has been described for the elt-5/6 RNAi phenotype (Koh and Rothman, 2001); neither of these phenomena were observed. Occasional degenerate-looking seam cells were observed in elt-1 RNAi worms, but not in control worms, in which the SCM::GFP marker was dispersed and the nuclei not discernable by differential-interference microscopy suggesting that at least some seam cells were undergoing degeneration (Fig. 4G). Degeneration of seam cells has been previously reported; loss of function of the guanine nucleotide exchange factor pfx-1 results in seam-cell degeneration (Pellis-van Berkel et al., 2005) and seam cells atrophy and die in lin-6(e1466) mutants because of a dysfunction of DNA replication after hatching (Singh and Sulston, 1978).

The effect of elt-1 RNAi on cuticle structure was investigated by SEM (Fig. 5). The L4 cuticle appeared normal (data not shown). However, the adult cuticle showed gross abnormalities of the lateral alae, consistent with seam-cell loss (Fig. 5A,B). In many adult worms, the alae were largely absent whereas in some worms there were discrete gaps in the alae, presumably reflecting the loss of individual seam cells (Fig. 5B). To further investigate the effects of elt-1 RNAi on cuticle structure, the expression of the col-19::GFP transgene was examined (Thein et al., 2003). This GFP marker is incorporated throughout the wild-type adult cuticle allowing the visualisation of both annulae and alae (Fig. 5C). Although the cuticle underlying the missing alae of elt-1 RNAi worms appeared normal by SEM (Fig. 5B), there was a clear disruption of the col-19::GFP expression pattern underlying the regions of missing alae (Fig. 5D). This suggests that the loss of seam cells has resulted in abnormal assembly of the deeper layers of the cuticle in the lateral regions as well as a loss of the alae themselves. One further observation was that patches of amorphous material were sometimes present on the surface of the cuticle in positions where alae were absent, often forming a shape reminiscent of seam cells (Fig. 5E). These patches contained disorganised GFP-tagged COL-19 (Fig. 5F) suggesting they might be the result of aberrant secretions from abnormal or degenerating seam cells.

In order to look at the requirement for elt-1 function during dauer larval development, elt-1 RNAi was performed on the temperature-sensitive daf-2(m577) mutant that produces dauer larvae at the restrictive temperature (25°C). L1 larvae from daf-2(m577) were placed on elt-1 RNAi plates and allowed to develop at 25°C. After approximately 30-35 hours, the majority of the larvae arrested development and had a dauer-like appearance. The larvae had greatly reduced motility, pharyngeal pumping ceased, the hypodermal and gut cytoplasm had a marked granular appearance and the gonad morphology was typical of dauer larvae. However, the larvae had a notably wider diameter than normal dauer larvae and tended not to take on the rod-like conformation that is typical of dauers (Fig. 6A,B). In addition, a large number of larvae appeared to take longer to shed the L2d cuticle (Fig. 6C). At 40 hours of development, by which time the majority of the control daf-2(m577) worms had successfully shed the L2d cuticle, many of the elt-1 RNAi worms had failed to do so. However, by 60 hours, ecdysis was fully complete in the elt-1 RNAi daf-2(m577) dauers with no other sign of defective moulting. SEM was performed on daf-2(m577) elt-1 RNAi worms at 40 hours and 60 hours of development from the L1 stage as well as on normal daf-2(m577) dauer larvae. In the majority of 40 hour elt-1 RNAi dauer larvae, the surface of the cuticle had a uniform appearance with no sign of alae consistent with the retention of the L2d cuticle (Fig. 6D,E). By contrast, alae were visible in elt-1 RNAi dauer larvae by 60 hours, owing to successful shedding of the L2d cuticle, but were often severely abnormal with gaps (Fig. 6F). Examination of elt-1 RNAi daf-2(m577) L1 larvae carrying the wIs51 transgene revealed a marked reduction in the number of SCM::GFP expressing cells in both L2d larvae and the dauer larvae (Fig. 6G,H,I). This, together with the loss of alae, demonstrates that elt-1 function is required to maintain seam-cell fates during L2d and dauer larval development.

The elt-1 RNAi daf-2(m577) dauer larvae slowly degenerate and die over a period of several days. Furthermore, when the temperature was shifted from 25°C to 15°C at either 40 hours or 60 hours of development from the L1 hatchling stage, the elt-1 RNAi daf-2(m577) dauer larvae did not resume development but degenerated after several days.

Although elt-1 RNAi did not cause any gross visible morphological abnormalities until after the L4-adult moult, a small number of larvae with abnormal motility were observed during the RNAi feeding experiments. Abnormal motility was seen in L2, L3, L4 and adult stages and involved a loopy, uncoordinated movement with exaggerated sinusoidal waveform (Fig. 7A,B). This motility phenotype, which was seen with both independent elt-1 dsRNA fragments, had a low penetrance but this increased to approximately 10% of worms when elt-1 RNAi was applied to rrf-3(pk1426) mutants. The same phenotype was also produced, at a penetrance of approximately 5-10% when elt-1 dsRNA was expressed from the transgene under the control of the heat-shock promoter (transgenic line JG136). Movement was quantified as bends per minute and worms with the uncoordinated phenotype showed increased motility relative to controls (Fig. 7C). These uncoordinated worms responded normally to touch and no other locomotory or behavioural abnormalities were seen.

The effect of overexpression of elt-1 on motility was examined using transgenic lines JM53 and JM55, which contain integrated transgenes that allow forced expression of ELT-1 under the control of the hsp16-2 heat-shock promoter (Gilleard and McGhee, 2001). Transgenic lines JM59 and JM60 allow forced expression of another hypodermal GATA factor, ELT-3, using the same heat-shock promoter and these were used as controls. Experiments with all four transgenic lines, as well as N2 worms, were performed in parallel. Induction of ELT-1 expression by a 2 hour heat shock of 33.5°C to strains JM53 and JM55 resulted in dramatically reduced motility for over 90% of worms within 2 hours of the application of heat shock. The effect of elt-1 overexpression on the nature of the worm motility was difficult to determine because the transgenic lines have a roller phenotype owing to the presence of the pRF4 marker transgene. However, the affected worms clearly moved more slowly than the JM59 and JM60 or N2 controls. Heat-shock inductions consisting of two applications of 33.5°C for 2 hours, separated by a 1 hour recovery interval, resulted in complete paralysis of over 90% of JM53 and JM55 worms but had little effect on JM59 and JM60 controls (Fig. 7D). All stages of worms (L1 to adult) were affected. It is acknowledged that there could be a variety of detrimental effects induced by widespread ectopic expression of elt-1, particularly because worms did not recover and resume development over time. However, affected worms maintained normal active pharyngeal pumping for several hours after the onset of immobility, which suggests a paralysis phenotype is produced before the occurrence of other more generalised effects.

Several genes have previously been identified as key regulators of seam-cell differentiation and development. The elt-5 and elt-6 genes are a pair of GATA transcription factors with potentially overlapping functions that are expressed monocistronically and possibly also as a dicistronic transcript (Koh and Rothman, 2001). The elt-5 gene is expressed in the 16 great-great-granddaughters of the MS and AB blastomeres and in most of their descendents. It is subsequently expressed at high levels in the embryonic lateral seam cells about an hour after they are born. RNAi of elt-5, which also down regulates elt-6, leads to a loss of seam cells in the embryonic hypodermis and this appears to be due to premature fusion of the seam cells into the dorsal/ventral hypodermis (Koh and Rothman, 2001). Recently ceh-16, a homologue of the Drosophila homeobox gene engrailed, has also been shown to be essential for seam-cell differentiation in the C. elegans embryo (Cassata et al., 2005). ceh-16 gene expression is visible in the lateral hypodermal seam cells very soon after the final embryonic precursor cell division and null mutations of this gene cause a loss of seam cells in the embryo. The ceh-16 gene appears to have a dual role in the production and maintenance of seam cells; as well as repressing the fusion of seam cells with the dorsal/ventral hypodermis, it is also directly necessary for seam-cell differentiation independently of cell fusion. A regulatory cascade has been proposed in which ceh-16 activates elt-5/6, which in turn activate other seam-cell markers and repress cell fusion (Cassata et al., 2005).

Although elt-1 has previously been shown to be essential in early embryogenesis to specify all hypodermal cell fates, our results show that seam cells respond to elt-1 RNAi in a markedly different way to dorsal/ventral hypodermal cells. Lpy-Dpy larvae produced in response to elt-1 RNAi consistently have a severe seam-cell deficit but a normal number of dorsal/ventral hypodermal cells. One possible explanation for this is that elt-1 may have a specific role in seam cells later in embryogenesis after hypodermal cell fates have been specified. For example, it could act downstream of ceh-16 in the differentiation or maintenance of seam-cell fates. This hypothesis is supported by the observation that elt-1::GFP is expressed in the seam cells, but not the dorsal/ventral hypodermal cells, during mid and late embryogenesis and this aspect of reporter gene expression is supported by previous immunolocalisation studies (Page et al., 1997). However, we cannot entirely exclude the possibility that the difference in elt-1 function between the two cell types is quantitative rather than qualitative; i.e. seam cells could be more sensitive to reductions in elt-1 function than are the dorsal/ventral hypodermal cells.

Although the Lpy-Dpy larvae produced in response to elt-1 RNAi have the normal number of dorsal/ventral hypodermal cells, there is a marked downregulation of elt-3::GFP expression in these cells. The elt-3 gene is a GATA transcription factor that is expressed in the dorsal/ventral hypodermis, but not the lateral seam, from the early comma stage onwards. It has been proposed to act downstream of elt-1 based on the respective expression patterns of the genes and the fact that ectopic expression of elt-1 can activate elt-3 expression (Gilleard and McGhee, 2001; Gilleard et al., 1999). However, because hypodermal cells are not specified in elt-1(zu180) embryos, the role of elt-1 in the activation of elt-3 expression during hypodermal cell differentiation has not been directly investigated. The RNAi phenotype shows that elt-1 function is essential for elt-3 expression even when dorsal/ventral hypodermal cells have been correctly specified in the early embryo. Since there are six TGATAA sites in a 240 bp element just upstream of the elt-3 gene it is very likely that elt-1 is a direct regulator of elt-3 in differentiating hypodermal cells (Gilleard et al., 1999).

Consistent with its expression pattern during larval development, elt-1 RNAi applied during postembryonic development leads to a severe loss of seam cells but has no discernable effect on the rest of the hypodermis. The ajm-1::GFP and SCM::GFP are useful markers for studying the fate of seam cells during postembryonic development (Cassata et al., 2005; Koh and Rothman, 2001; Kostrouchova et al., 2001; Pellis-van Berkel et al., 2005). For example, inappropriate seam-cell loss because of abnormal cell fusion to the dorsal/ventral hypodermis in response to elt-5/6 RNAi is clearly visible using these markers (Koh and Rothman, 2001). Careful examination of ajm-1::GFP and SCM::GFP expression revealed no evidence of inappropriate seam-cell fusion into the dorsal/ventral hypodermis in response to elt-1 RNAi. However a number of degenerating seam cells were observed in which the SCM::GFP marker was dispersed throughout the cytoplasm and the nuclei of these cells was not discernable using differential-interference optics. Hence, although we have not definitively determined the mechanism by which seam cells are lost, our results suggest that inappropriate cell fusion is unlikely to be a major cause of seam-cell loss in response to elt-1 RNAi and at least some cells undergo degeneration. This is similar to the fate of seam cells in pxf-1 mutants, which has been described using the same markers (Pellis-van Berkel et al., 2005).

The elt-1 RNAi phenotype provides a useful insight into the function of seam cells during postembryonic development. The loss of alae in the adult cuticle confirms the role of seam cells in producing this structure, which has previously been shown by laser ablation studies (Singh and Sulston, 1978). The apparently normal appearance of the underlying cuticle is also consistent with these previous studies and presumably this is derived from the dorsal/ventral hypodermis. RNAi of elt-1, applied during larval development, has a severe effect on the integrity of adult worms within a few hours of the L4-adult moult. Adult hermaphrodites show a `burst-vulva' phenotype, in which the uterus herneates through the vulva. This is likely to be a direct consequence of seam-cell loss because the lateral seam anchors the vulval and uterine cells in position by virtue of the utse cell connection (Michaux et al., 2001; Newman et al., 2000; Sharma-Kishore et al., 1999). This hypothesis is supported by the description of a `burst-vulva' phenotype for pxf-1 mutants in which seam cells are also lost (Pellis-van Berkel et al., 2005). In contrast to the catastrophic effect on adult worms, elt-1 RNAi-induced seam-cell loss has no observable effect on the morphology or viability of larval stages. This supports previous evidence that seam cells are not essential for normal larval growth and development; lin-5(e1348) mutants develop to a normal adult size despite progressively losing seam cells throughout postembryonic development (Singh and Sulston, 1978).

The role of seam cells in the moulting process is less clear and has been the subject of some speculation. There are several genes which are expressed in the lateral seam that have been shown, either by mutation or RNAi, to be essential for normal moulting; nhr-25 (Chen et al., 2004; Silhankova et al., 2005), nhr-23 (Kostrouchova et al., 2001), acn-1 (Brooks et al., 2003), pxf-1 (Pellis-van Berkel et al., 2005) and lrp-1 (Yochem et al., 1999). However, these are also expressed in the dorsal/ventral hypodermal syncytium and, in the case of nhr-25, lrp-1 and pxf-1, the experimental evidence suggests that it is primarily their function in the hypodermal syncytium and not the lateral seam that is required for normal moulting (Pellis-van Berkel et al., 2005; Silhankova et al., 2005; Yochem et al., 1999). The only evidence that has directly implicated seam cells as being essential for normal larval moulting is the RNAi phenotype of elt-5/elt-6 (Koh and Rothman, 2001). These genes are expressed at high levels in seam cells and RNAi produces a highly penetrant moulting defect in the larval stages. Tissue-specific rescue experiments of the elt-5 RNAi phenotype with transgenically expressed elt-6 suggested that elt-5/6 function was required in the seam for normal moulting (Koh and Rothman, 2001). This in turn led to the suggestion that seam cells may play an essential role in larval moulting. However, a striking feature of our results is that the severe seam-cell loss induced by elt-1 RNAi has little or no effect on moulting. No sign of abnormal moulting was detected in the L2-L3 or L3-L4 moults despite dramatic seam-cell loss in the L2 and L3 stages. Similarly, despite a highly penetrant and severe seam-cell loss in L4 larvae, adult worms only very occasionally retained fragments of L4 cuticle (only two such worms were seen out of many hundreds examined). The delayed ecdysis observed during the L2d-dauer moult following elt-1 RNAi was not accompanied by retention of L2d cuticle fragments or constrictions of the dauer cuticle and so was not a true moulting defect as previously described by others (Chen et al., 2004; Kostrouchova et al., 2001; Pellis-van Berkel et al., 2005; Yochem et al., 1999). This delay in the L2d ecdysis is probably simply a consequence of the physical abnormalities of the elt-1 RNAi dauer larvae. There is other evidence to suggest that seam-cell loss does not necessarily lead to moulting defects. Examination of lin-5(e1348) and lin-6(e1466) mutants, which both lose large numbers of seam cells during development, did not reveal moulting defects (data not shown); this was consistent with early descriptions of lin-5(e1348) mutants which did not report abnormal moulting (Singh and Sulston, 1978). Also, loss of function of the PlexinA gene, plx-1, leads to seam-cell loss, although not as dramatic as that induced by elt-1 RNAi, with no reported moulting defects (Fujii et al., 2002).

It therefore appears that the presence of large numbers of seam cells is not required for normal moulting. This leads to the question of why severe moulting defects are induced by elt-5/6 RNAi. There are two possible explanations: first, elt-5/6 function may be required in other tissues outside the lateral seam in which they are expressed at low levels; second, and possibly more persuasively, moulting defects could be due to the nature of the seam-cell loss induced by elt-5/6 RNAi: seam cells are not simply lost, but undergo a fate change to a dorsal/ventral hypodermal cell fate. This leads to ectopic expression of elt-3, and presumably other genes normally expressed only in the dorsal/ventral hypodermis, and to inappropriate fusion of seam cells into the hyp-7 syncytium. This may well produce a more generalised and severe effect on cuticle synthesis than would a simple loss of seam cells. Therefore, the moulting abnormalities caused by elt-5/6 RNAi may be related to more widespread cuticle abnormalities rather than being directly related to seam-cell loss per se.

The elt-1 gene is also required to maintain seam-cell fates during dauer larval development because daf-2(m577) L2d and dauer larvae have a seam-cell deficit. The associated lack of alae demonstrates that seam cells are necessary for alae formation in the dauer cuticle as in the adult. The elt-1 RNAi dauer larvae have many of the characteristic features of normal dauers but have an increased diameter. Furthermore, those elt-1 RNAi dauer larvae that had a severe seam-cell loss were `fatter' than those with minimal seam-cell loss, which occasionally had localised regions of increased diameter (data not shown). During normal dauer formation, larvae reduce in diameter by about 10% and our results provide strong evidence to support a previous suggestion that seam cells may be responsible for this diametric shrinkage (Singh and Sulston, 1978).

Wild-type worms move in well-coordinated sinusoidal undulations of a fixed amplitude. RNAi of elt-1 induces abnormal locomotion in all stages from L1 to adult, which involves a loopy uncoordinated movement with increased amplitude of the sinusoidal waveform. These uncoordinated worms are also hypermotile in terms of the rate at which they initiate body bends. Overexpression of elt-1 using the hsp16-2 promoter produces the reciprocal phenotype of reduced motility. The hypermotility phenotype suggests an abnormality in the neuronal circuitry that regulates locomotion and the elt-1 reporter gene expression pattern supports this. There are several major classes of motor neuron that innervate C. elegans body wall muscle: DA,DB, DD neurons and VA,VB,VD neurons innervate the dorsal and ventral body wall muscles, respectively (Chalfie and White, 1988; White et al., 1976). Each of these neuronal classes has multiple members that are situated along the length of the ventral cord and the elt-1 reporter gene is expressed in at least some of these (Fig. 1E,F). Four bilaterally symmetrical pairs of interneurons, AVA, AVB, AVD and PVC, provide input into the ventral cord motor neurons and are involved in the regulation of locomotory behaviour (Chalfie and White, 1988). Again, the elt-1 reporter is expressed in cell bodies and axons consistent with the position of these interneurons (Fig. 1G). The elt-1 GATA transcription factor could have a critical role in either the development or the function of locomotory neurons. Examination of the expression pattern of the elt-1::GFP reporter gene, or the pan-neuronal marker F25B3.3::GFP, in elt-1 RNAi worms did not reveal any obvious loss or abnormalities in ventral cord or associated interneurons. Furthermore, application of RNAi after the L1 stage still produced the uncoordinated phenotype. Hence the phenotype could be due to effects on neuronal function or aspects of neuronal remodelling that occur after the L1 stage. A similar hypermotility phenotype has been previously associated with disruption of the Gαo-Gαq signalling network that regulates neurotransmitter release and it is possible elt-1 is involved in regulating some aspect of this network (Mendel et al., 1995; Segalat et al., 1995).

We are grateful to the following people who kindly donated strains and reagents: Christina Schmid, Fred Hagen, Iain Johnstone, Tony Page, Joel Rothman, Jim McGhee, Yugi Kohara and Zhen Zhou. Other strains were kindly provided by Theresa Stiernagle from the Caenorhabditis Resource Centre (CGC), which is supported by the National Institutes of Health National Center for Research Resources. We are grateful to Margaret Mullin and Laurence Tetley (University of Glasgow) for help with the scanning electron microscopy.