The Snail-like CES-1 protein of C. elegans can block the expression of theBH3-only cell-death activator gene egl-1 by antagonizing the function of bHLH proteins

The elimination of unwanted cells by programmed cell death or apoptosis is a fundamental feature of animal development and the mechanisms of programmed cell death have been conserved through evolution (reviewed by Ellis et al., 1991; Strasser et al., 2000). For example, more than 50% of all neurons formed by neurogenesis in vertebrates are removed by programmed cell death before adulthood (reviewed by Oppenheim, 1991). By removing neurons that are superfluous or that have not made appropriate connections with their targets, programmed cell death plays an important role in the formation of a functional vertebrate nervous system. What triggers neurons to undergo programmed cell death in the developing nervous system is largely unknown. One inducer of neuronal death appears to be the removal of trophic factors, such as nerve growth factor (NGF). NGF deprivation has been shown to induce programmed cell death in a cell non-autonomous manner in a number of models of neuronal death, such as the death of sympathetic neurons (reviewed by Deshmukh and Johnson, 1997). The death of these neurons is not only dependent on NGF deprivation but on the synthesis of macromolecules, i.e. on an active transcription and translation machinery. The programmed death of neurons by trophic factor removal therefore also depends on intrinsic, cell-autonomous programs and/or the fate of the cell.

Programmed cell death is also a fundamental feature of neurogenesis in invertebrates, including the nematode Caenorhabditis elegans. The nervous system of a wild-type C. elegans hermaphrodite is composed of 302 neurons of at least 118 different types(Sulston and Horvitz, 1977; Sulston et al., 1983). During the development of a C. elegans hermaphrodite, 131 somatic cells die by programmed cell death, a process that is determined by the essentially invariant C. elegans cell lineage. Most of these 131 cells are sisters of cells that differentiate into neurons and, when prevented from undergoing programmed cell death, often adopt a neuronal fate themselves(Ellis and Horvitz, 1986; Avery and Horvitz, 1987). Hence, more than a quarter of the cells that are initially generated and that have the potential to form neurons are eliminated by programmed cell death during the development of the C. elegans nervous system.

Genetic analyses of programmed cell death in C. elegans have been instrumental for our current understanding of the conserved, molecular mechanisms of this essential process (reviewed by Horvitz, 1999). Four genes, egl-1 (egl, egg-laying defective), ced-9(ced, cell-death defective), ced-4 and ced-3, have been identified, which, when mutated, can block programmed cell death during development. These four genes act in a simple genetic pathway, in which egl-1 negatively regulates ced-9, ced-9 negatively regulates ced-4 and ced-4 positively regulates ced-3, the activity of which is required for programmed cell death. The ced-3gene encodes a pro-caspase, ced-4 an Apaf1-like adaptor, ced-9 a Bcl-2-like cell-death inhibitor and egl-1 a pro-apoptotic member of the Bcl-2 family, a BH3-only protein. Genetic and cell biological observations indicate that CED-9, CED-4 and proCED-3 proteins are present in most if not all cells during C. elegans embryogenesis. It has been proposed that in cells that live, CED-9 blocks the activity of CED-4 and, hence, the CED-4-dependent activation of proCED-3. Conversely, in cells destined to die, EGL-1 negatively regulates CED-9 thereby allowing the activation of proCED-3 (Chen et al.,2000).

Little is known thus far about how the central cell-death pathway and, in particular, the activity of its most upstream component, the BH3-only protein EGL-1, is regulated to specifically cause the death of the 131 cells that are destined to die during C. elegans development. EGL-1 is likely to be regulated by cell-specific factors because genes have been identified that act upstream of egl-1 and that, when mutated, block specific cell deaths. For example, mutations in the genes ces-2 and ces-1(ces, cell-death specification) block specifically the death of the NSM sister cells, and of the NSM sister cells and the I2 sister cells respectively (Ellis and Horvitz,1991). Furthermore, a mutation in the gene tra-1(tra, transformer) prevents the death of the hermaphrodite-specific neurons (HSNs) in males (Conradt and Horvitz, 1999). In addition to being involved in the specification of the male-specific death of the HSNs, the tra-1 gene has a more general role during C. elegans development. tra-1 functions as the terminal regulator of all somatic sexual fates in C. elegans(reviewed by Goodwin and Ellis,2002). The cell-specific pathways that regulate EGL-1 activity therefore may not be cell-death specific pathways but rather pathways that play additional, important roles during development. tra-1 encodes a zinc-finger DNA-binding protein, TRA-1A, which directly represses the transcription of egl-1 in the HSNs of hermaphrodites but not in the HSNs of males. At least in the HSNs, in which the life-versus-death decision is determined by somatic sex, EGL-1 activity is therefore regulated at the transcriptional level (Conradt and Horvitz,1999).

In contrast to the male-specific death of the HSNs, the death of the NSM sister cells appears to be determined solely by lineage. Whether the signal that triggers this particular death also does so by regulating egl-1transcription has not yet been determined. As described above, two genes, ces-1 and ces-2, have been identified that, when mutated,can block the death of the NSM sister cells and therefore cause a cell-death specification or Ces phenotype (Ellis and Horvitz, 1991). A loss-of-function (lf) mutation in the gene ces-2 blocks the death of the NSM sister cells, indicating that ces-2 is required for their programmed death. The death of the NSM sister cells is also blocked by a gain-of-function (gf) mutation in ces-1, which suggests that ces-1 has cell-death protective activity and that it can function to prevent the death of the NSM sister cells. A ces-1(lf) mutation causes no obvious phenotype; however, it suppresses the ability of the ces-2(lf) mutation to block the death of the NSM sister cells, suggesting that ces-2 causes the NSM sister cells to die by negatively regulating the cell-death protective activity of ces-1. ces-2 encodes a DNA-binding protein most closely related to the proline- and acid-rich (PAR) subfamily of basic leucine-zipper (bZIP)transcription factors of vertebrates(Metzstein et al., 1996). ces-1 encodes a DNA-binding protein most similar to members of the Snail family of zinc-finger transcription factors(Metzstein and Horvitz, 1999). cis-regulatory regions upstream of the ces-1 transcription unit include a potential CES-2 binding site, which suggests that CES-2 might be a direct, negative regulator of ces-1 transcription. The ces-1(gf) mutation is located adjacent to this potential CES-2 binding site, which suggests that this mutation results in NSM sister cell survival by causing overexpression of ces-1 in the NSM sister cells. This hypothesis is supported by the observation that overexpression of ces-1 from extrachromosomal arrays carrying the wild-type ces-1 locus causes NSM sister cell survival(Metzstein and Horvitz,1999).

Members of the Snail family of zinc-finger DNA-binding proteins act predominantly as repressors of transcription (reviewed by Hemavathy et al., 2000). It has therefore been proposed that in ces-1(gf) animals, CES-1 blocks the death of the NSM sister cells by blocking the transcription of a proapoptotic gene (Metzstein and Horvitz,1999). As ces-1 acts upstream of the cell-death activator gene egl-1, egl-1 is a candidate target of ces-1(Conradt and Horvitz, 1998). In this paper we present data indicating that the basic helix-loop-helix (bHLH)proteins HLH-2 and HLH-3 are at least partially required for the death of the NSM sister cells and that a heterodimer composed of HLH-2 and HLH-3,HLH-2/HLH-3, acts as a direct activator of egl-1 transcription. Furthermore, we describe studies that suggest that in ces-1(gf)animals, CES-1 acts as a direct repressor of egl-1 transcription by antagonizing the function of HLH-2/HLH-3.

C. elegans was cultured and maintained as described by Brenner on NGM medium at 20°C unless otherwise noted(Brenner, 1974). The Bristol strain N2 was used as the standard wild-type strain. Mutations used in this study are listed below and are described by Riddle et al.(Riddle et al., 1997), except where noted otherwise. LGI: unc-87(e1216), ces-1(n703), ces-1(n703 n1434), ces-2(n732ts), hlh-2(bx108)(Portman and Emmons, 2000). LGIII: cnd-1(ju29)(Hallam et al., 2000), bcIs1(P_egl-1gfp; this study). LGIV: ced-3(n717), bcIs25(P_tph-1gfp; this study). LGV: egl-1(n1084 n3082)(Conradt and Horvitz, 1998), unc-76(e911), bcIs37(P_egl-1his24-gfp; this study). LGX: lin-15(n765), bcIs24(P_tph-1gfp; this study), bcIs30(P_tph-1gfp; this study).

Standard molecular biology protocols were used unless otherwise noted. Primers used throughout this study were based on sequences determined by the C. elegans Sequencing Consortium(The C. elegans Sequencing Consortium, 1998). Plasmid pBC105 was generated by amplifying his24-gfp from plasmid pJH2.19 (M. Dunn and G. Seydoux, personal communication), using the polymerase chain reaction (PCR) and appropriate primers. The PCR product was digested with SmaI and NcoI and used to replace the SmaI-NcoI fragment of plasmid pBC99(Conradt and Horvitz, 1999). Plasmid pBC08 (Conradt and Horvitz,1998) was used to generate plasmids pBC119, pBC13, pBC11, and pBC149. Plasmid pBC148, which is based on pBluescript and contains a 2.9 kb PstI-XhoI fragment of pBC08, including Region B, was used to mutagenize the four Snail-binding sites/E-boxes in Region B. PCR-mediated mutagenesis was performed. The sequence of the resulting plasmids carrying four Snail^-/E-box^- sites (5′-CATATA-3′)(pBC170) or four Snail^-/E-box⁺ sites(5′-CATATA-3′) (pBC180) was confirmed by sequence analysis using an automated ABI sequencer (Applied Biosystems). A 1.4 kb HpaI-PstI fragment, which includes the four Snail-binding sites/E-boxes, of pBC170 and pBC180 was used to replace the HpaI-PstI fragment of pBC08 and to generate plasmids pBC181(Snail^-/E-box^-) and pBC182(Snail^-/E-box⁺). The plasmid for the expression in E. coli of dsRNA of hlh-2 (pBC132A) was constructed by subcloning the corresponding cDNA from plasmid pKM1199 into the BamHI site of vector L4440 (Krause et al.,1997; Timmons et al.,2001). The plasmid for the expression of dsRNA of hlh-3in vitro was obtained by subcloning the corresponding cDNA from pKM1195 into the BamHI site of pBluescript to generate pBC226. Plasmids for the expression of hlh-4, hlh-6, hlh-12 or hlh-14 in E. coli were obtained by PCR-amplifying coding regions of the genes from genomic DNA using the following primers: 5′-gaa ggg atc ctg ttc tga aac aac atc ttcc aac g-3′ and 5′-gaa ggg atc cgc agt tga tgg ttg ata gaa ata tg-3′ for hlh-4, 5′-gaa ggg atc caa ttc cac att cca act tcc-3′ and 5′-gaa ggg atc ccc aaa ctg atg agc tga aaa tt-3′ for hlh-6, 5′-gaa ggg atc cag cca cct ctt aca taa ttc-3′ and 5′-gaa ggg atc cat ata aac att ggt ttg ggg-3′ for hlh-12, 5′-gaa ggg atc cct gag ctc aga ttt tca g-3′ and 5′-gaa ggg atc ctg cgt tct ctc tca ttt ctg-3′ for hlh-14. PCR fragments were digested with BamHI and ligated into pBluescript to generate plasmids pBC227 (hlh-4), pBC228 (hlh-6), pBC229(hlh-12) and pBC230 (hlh-14).

Germline transformation was performed as described by Mello and Fire(Mello and Fire, 1995). For transformation with the P_egl-1his24-gfp reporter, ced-3(n717); lin-15(n765) animals were injected with plasmid pBC105 (P_egl-1his24-gfp)(2.5 ng/μl) and the co-injection marker pL15-EK (50 ng/μl), which rescues the lin-15(n765) multivulva or Muv phenotype(Clark et al., 1994). Injected animals were shifted to 25°C, and non-Muv F1 animals were picked to establish transgenic lines. The line carrying the extrachromosomal array bcEx78 was used to integrate the array into the genome. ced-3(n717); lin-15(n765); bcEx78animals were mutagenized using ethyl methanesulfonate (EMS)(Brenner, 1974) and transgenic F2 animals were selected that gave rise to 100% non-Muv progeny at 25°C. The strain carrying the integration bcIs37 V was backcrossed five times to ced-3(n717); lin-15(n765). For transformation with the P_tph-1gfp reporter, we injected lin-15(n765) animals with plasmid pBY668(P_tph-1gfp)(Rohrig et al., 2000) (50 ng/μl) and pL15-EK (50 ng/μl). Transgenic lines were selected and maintained at 25°C. The line carrying the array bcEx113 was used for integration. The strains carrying the integrations bcIs24, bcIs25and bcIs30 were backcrossed four times to N2. For NSM sister cell death rescue, egl-1(n1084 n3082) unc-76(e911); lin-15(n765) bcIs24animals were injected with the egl-1 rescuing plasmids pBC08 (2.0 ng/μl), pBC119 (1.5 ng/μl), pBC13 (1.1 ng/μl), pBC11 (1.0 ng/μl),or pBC149 (1.1 ng/μl) and the co-injection marker p76-16B (75 ng/μl),which rescues the uncoordinated or Unc phenotype of unc-76(e911) animals(Bloom and Horvitz, 1997).(egl-1 rescuing fragments are toxic and were therefore injected at such low concentrations.) Non-Unc F1 animals were selected to establish transgenic lines. ces-2(n732); bcIs25;egl-1(n1084 n3082) unc-76(e911) animals were injected with the egl-1 rescuing plasmids pBC08 (2 ng/μl), pBC181(2 ng/μl) or pBC182 (2 ng/μl) and p76-16B (75 ng/μl). bcIs1is an integrated array of plasmid pBC99(P_egl-1gfp) (2.0 ng/μl)(Conradt and Horvitz, 1999) and the co-injection marker pL15-EK (50 ng/μl).

For electrophoretic mobility shift assays, probes were generated by PCR amplification in the presence of 10 μCi [³²P]-dATP using the primers 5′-aac tca tcc acg tca cca aa-3′ and 5′-ttg tcc act cgt tta cca ca-3′ and plasmids pBC08 (wild-type), pBC181(Snail^-/E-box^-) or pBC182(Snail^-/E-box⁺) as templates. The labeled PCR products were purified on a 6% acrylamide/TBE gel. A GST-CES-1 zinc-finger fusion protein construct, expressing GST fused to amino acids 117-270 of CES-1(referred to as GST-CES-1), was made as described by Metzstein et al.(Metzstein et al., 1999). GST-CES-1 was produced and purified from E. coli as described by Ip et al. (Ip et al., 1992) (4.25 ng represents 1×10^-13 moles). Expression plasmids for His₆-HLH-2 (pKM1199) and His₆-HLH-3 (pKM1195) fusion proteins were provided by M. Krause(Krause et al., 1997). His₆-HLH-2 and His₆-HLH-3 fusion proteins were produced in E. coli strain BL21-CodonPlus(DE3)-RIL (Stratagene) and purified as described (Portman and Emmons,2000). 5.0 ng and 3.2 ng of His₆-HLH-2 and His₆-HLH-3 fusion protein, respectively, represents 1×10^-13 mol. The purity and concentration of the fusion proteins were assessed by SDS-PAGE and the BioRad protein assay (BioRad). EMSAs were performed as described for HLH-2 and HLH-3, and CES-1(Krause et al., 1997; Metzstein and Horvitz, 1999). Binding was quantified using a phosphoimager (Fujifilm BAS-2500) and appropriate software (Aida Image Analyzer V. 3.21).

For RNAi experiments using feeding as the method of delivering dsRNA,plasmids pBC132A, pBC124A, pBC188, pBC189, pBC190 and pBC191 were transformed into E. coli strain HT115(Timmons et al., 2001). NGM plates containing 6 mM IPTG, 50 μg/ml ampicillin and 12.5 μg/ml tetracycline were inoculated with transformed HT115 bacteria. The expression of dsRNA was induced overnight at room temperature. The plates were subsequently inoculated with L4 larvae. Animals were cultured at 15°C and their progeny was analysed. For RNAi experiments using injection as the method of delivering dsRNA, plasmids pBC226, pBC227, pBC228, pBC229 and pBC230 were used to PCR-amplify their inserts flanked by the T3 and T7promoter using appropriate primers. PCR products were used as templates for in vitro transcription using the T3 and T7 polymerases(Promega). Reactions contained 100 U T3 or T7 polymerase, 2 mM each rATP, rCTP, rGTP, rUTP, 10 mM DTT in a final volume of 100 μl in DEPC-H₂O buffered with transcription buffer. After incubating for 2 hours at 37°C, five units DNase was added and the reaction incubated for another 20 minutes at 37°C. The RNA was purified by phenol/chloroform extraction and resuspended in 20 μl DEPC-H₂O. Single stranded sense RNA was annealed with an equal amount of the corresponding antisense RNA at 37°C for 20 minutes, centrifuged at 4°C for 10 minutes and injected into the gonad of young adult worms, which were subsequently incubated at 25°C. The progeny of injected animals was analysed.

For the analysis of P_egl-1his24-gfpexpression in the NSMs and NSM sister cells, the NSMs and NSM sister cells were identified in L1 larvae using Nomarski microscopy as described(Ellis and Horvitz, 1991) and analysed for his24-gfp expression using epifluorescence. Percent NSM sister cell survival was determined in L3 or L4 larvae carrying the P_tph-1gfp reporter using a combination of Nomarski microscopy and epifluorescence. To obtain the number of surviving NSM sister cells per animal, the total number of GFP-positive cells in the anterior pharynx of an animal was determined and subtracted by two. Percent NSM sister cell survival represents the percent of NSM sister cells that survived and n the maximum number of NSM sister cells that could have survived in the number of animals analysed.

Embryos were prepared in 10 μl H₂O on poly L-Lysine coated slides and allowed to develop until the 1.5-fold stage in a moist chamber. They were fixed with 5% paraformaldehyde and stained as described(Krause et al., 1990; Krause et al., 1997). HLH-2 was detected using a polyclonal anti-HLH-2 antibody raised in rabbits(provided by Mike Krause) and GFP was detected using a monoclonal anti-GFP antibody (Clontech). Immunofluorescence was viewed using a Leica TCS NT confocal microscope.

About 400 minutes after the first cleavage of a C. elegans embryo,two bilaterally symmetric cells, ABaraapapaa and ABaraapppaa, each divide. The ventral daughters of these cells survive and differentiate into serotonergic,neurosecretory motoneurons, called NSMs, which are located in the anterior pharynx. Their dorsal sisters, the NSM sister cells, however, die by programmed cell death shortly after they are born (at about 420-430 minutes)(Sulston et al., 1983). The programmed death of the NSM sister cells is blocked by lf mutations in the C. elegans cell-death activator genes egl-1, ced-3 and ced-4, and by a gf mutation in the C. elegans cell-death inhibitor gene ced-9, indicating that these specific cell deaths are dependent on the activation of the central cell-death pathway(Horvitz, 1999).

In the case of another specific cell death, the death of the HSNs in males,it has been shown that the decision between life and death is specified by egl-1 expression (Conradt and Horvitz, 1999). To determine whether egl-1 expression also specifies the cell-death fate of the NSM sister cells, we monitored the expression of egl-1 in these cells, using an integrated Pegl-1his24-gfp transgene. Pegl-1his24-gfp expresses a nuclearly localized fusion of the His24 protein and the green fluorescent protein (GFP) under the control of the cis-regulatory regions of the egl-1 gene. The NSM sister cells normally die during a stage called the 1.5-fold stage of embryogenesis. At this stage of development it is difficult to identify cells on the basis of their position within the embryo. We therefore analysed the expression of Pegl-1his24-gfp in the background of the ced-3 lf mutation n717. ced-3(n717) blocks programmed cell death, including the death of the NSM sister cells (Ellis and Horvitz,1986). However, because ced-3 acts downstream of egl-1 genetically, ced-3(n717) should not interfere with the expression of the Pegl-1his24-gfp transgene. This experimental design enabled us to analyse the expression of Pegl-1his24-gfp in the NSMs and in surviving `undead' NSM sister cells in animals of the first larval stage of development (L1 larvae), in which these cells are identifiable by position in the anterior pharynx using Nomarski differential interference contrast (Nomarski microscopy)(Ellis and Horvitz, 1991). In ced-3(n717); Pegl-1his24-gfp larvae, we observed GFP in 88%of the NSM sister cells, which normally die, and in 0% of the NSMs, which normally survive (n=51) (Fig. 1A). This indicates that the activity of EGL-1 is regulated at the transcriptional level in the NSMs and NSM sister cells and, hence, that the cell-death fate of these cells is specified by egl-1 expression.

ces-1 acts upstream of egl-1 and a gf mutation of ces-1, n703, specifically blocks the death of the NSM sister cells. To determine whether ces-1(n703gf) affects egl-1expression in the NSM sister cells, we analysed the expression of Pegl-1his24-gfp in ces-1(n703gf); ced-3(n717) animals. In this background, we detected GFP in 2% of the NSM sister cells and in 0% of the NSMs (n=51)(Fig. 1B). The ces-1(gf) mutation therefore prevents the death of the NSM sister cells by blocking the transcription of egl-1 in these cells.

The egl-1 lf mutation n1084 n3082 blocks programmed cell death, including the death of the NSM sister cells. A 7660 bp genomic fragment of cosmid C01G9, pBC08, which includes the egl-1 transcription unit,1036 bp of its upstream region and 5575 bp of its downstream region, when introduced as an extrachromosomal array, can rescue the cell-death defect of egl-1(n1084 n3082) animals, including the death of the NSM sister cells (Fig. 2)(Conradt and Horvitz, 1998). We therefore conclude that pBC08 contains the cis-regulatory region or regions required for the expression of egl-1 in the NSM sister cells. To determine which region or regions of pBC08 are specifically required for the expression of egl-1 in these cells, we analysed subclones of pBC08 for their ability to rescue the death of the NSM sister cells in egl-1(n1084 n3082) animals(Fig. 2A).

Undead NSM sister cells are located close to and posterior to their sisters, the NSMs, in the anterior pharynx and can therefore be identified due to their position using Nomarski microscopy(Ellis and Horvitz, 1991). In addition, like their sisters, undead NSM sister cells differentiate into serotonin-expressing neurons. The NSMs are the only serotonin-expressing neurons in the anterior pharynx of C. elegans. Hence, in wild-type animals, two serotonin-expressing cells can be detected in this part of the pharynx, the two NSMs, and in animals, in which the NSM sister cells survive,four serotonin-expressing cells, the two NSMs and the two undead NSM sister cells (Ellis and Horvitz,1991). The Ptph-1gfp (tph, tryptophan hydroxylase) transgene expresses gfp under the control of the cis-regulatory regions of the tph-1 gene, which encodes an enzyme required for serotonin biosynthesis and which is expressed in serotonergic neurons, including the NSMs(Sze et al., 2000). To test whether the Ptph-1gfp transgene is also expressed in undead NSM sister cells (and therefore can be used to conveniently identify undead NSM sister cells), we analysed the expression of this reporter in wild-type animals, in which the NSM sister cells die, and in animals homozygous for egl-1(n1084 n3082), in which the NSM sister cells survive. In wild-type animals, we observed GFP in only two cells located in the anterior part of the pharynx, the NSMs (0% NSM sister cell survival, n=414). In most egl-1(n1084 n3082) animals,however, we observed GFP in the two NSMs and in two additional cells located more posterior, the undead NSM sister cells (96% NSM sister cell survival, n=120). Throughout our studies, NSM sister cell survival was therefore analysed using a combination of Nomarski microscopy and P_tph1gfp/epifluorescence.

As mentioned above, in wild-type animals, 100% of the NSM sister cells die;however, 96% of them survive in egl-1(n1084 n3082) animals. In transgenic egl-1(n1084 n3082) animals carrying extrachromosomal arrays of the egl-1 rescuing fragment pBC08, between 35% and 5% of the NSM sister cells survived (6/6 lines), indicating that pBC08 rescues the death of the NSM sister cells(Fig. 2B). (Rescue was defined as 40% or less NSM sister cell survival.) Similarly, subclone pBC119, which includes 3447 bp of the region downstream of the egl-1 transcription unit, rescued the cell-death defect of egl-1(n1084 n3082)animals (8/8 lines). By contrast, subclones pBC13 (806 bp of downstream region) and pBC11 (377 bp of downstream region) failed to rescue (0/4 lines and 0/5 lines). pBC13 lacks 2641 bp of pBC119 that includes a 352 bp region,called Region B, which is conserved between the egl-1 locus of C. elegans and C. briggsae (Fig. 2A,C) (see below). As sequence conservation between these two Caenorhabditis species suggests functional relevance(Heschl and Baillie, 1990), we tested whether addition of Region B alone can restore the rescuing activity of pBC11 (pBC149). We found that pBC149 rescued NSM sister cell death in four out of seven lines (4/7). Conversely, pBC08 lacking only Region B (pBC165) failed to rescue (0/6 lines). These results indicate that Region B of the egl-1 locus is required to rescue the death of the NSM sister cells in egl-1(n1084 n3082) animals. As the death of the NSM sister cells is dependent on the expression of egl-1 in the NSM sister cells, Region B most probably contains the cis-regulatory regions necessary for the transcription of egl-1 in these cells.

An alignment of the sequence of Region B of the C. elegans egl-1locus with the sequence of the corresponding region of the C. briggsae egl-1 locus, revealed extensive sequence conservation at the nucleotide level. Over their entire length, the sequences are 76% identical(Fig. 2C). Sequence inspection revealed that Region B contains four closely spaced Snail-binding sites,DNA-binding sites for members of the Snail family of zinc-finger transcription factors (Hemavathy et al.,2000). The core motif of three of these binding sites is completely conserved in C. briggsae (binding sites I, II and IV; 6/6 bases identical), and one of them has one base change in C. briggsae(binding site III; 5/6 bases identical). The sequence of binding sites I and II of C. elegans is a perfect match to the sequence of the core motif of a consensus Snail-binding site (5′-CACCTG-3′) whereas the sequences of binding sites III and IV have one mismatch to the consensus sequence (5′-CATCTG-3′ and 5′-CAGCTG-3′, respectively).

CES-1 is a member of the Snail family of DNA-binding proteins and has previously been shown to bind to Snail-binding sites in vitro(Metzstein and Horvitz, 1999). In addition, in ces-1(gf) animals, CES-1 prevents the death of the NSM sister cells by blocking egl-1 expression. To determine whether in ces-1(gf) animals, CES-1 might block the transcription of egl-1 directly by binding to the four Snail-binding sites in Region B, we tested whether CES-1 can bind to these sites in vitro, using electrophoretic mobility shift assays (EMSAs). A bacterially produced,affinity-purified CES-1 fusion protein consisting of the C-terminal half of CES-1 (which includes all five zinc fingers of CES-1) fused to glutathione S-transferase (GST) could bind and shift a radioactively labeled 390 bp DNA fragment consisting of wild-type Region B, including the four Snail-binding sites (Fig. 3, lanes 1-6). Long exposures and the use of various fragments of Region B indicate that CES-1 can bind to at least three of the four Snail-binding sites in vitro (data not shown). Binding of CES-1 to Region B was severely reduced after the introduction of point mutations that destroy the core motif of the four Snail-binding sites (5′-CACCTG-3′ to 5′-CATATA-3′)(Fig. 3, lanes 7-12). Using an amount of CES-1 that binds about 50% of the wild-type probe (set to 100%binding), CES-1 binding was reduced to on average 6% (n=2)(Fig. 3, compare lanes 4 and 10). Intact Snail-binding sites are therefore required for the ability of CES-1 to bind to Region B in vitro. The specificity of the observed binding was furthermore confirmed by competition experiments, using the wild-type Region B as a probe and a wild-type or mutant Snail-binding site as competitor(data not shown).

These results suggest that in ces-1(gf) animals, in which the NSM sister cells survive, CES-1 might bind to the Snail-binding sites in Region B of the egl-1 locus thereby directly repressing egl-1transcription in the NSM sister cells.

Region B of the egl-1 locus is required to rescue the death of the NSM sister cells in egl-1(lf) mutants and is therefore most likely to be necessary for the expression of egl-1 in these cells. We therefore sought to identify transcriptional activators that are required for the death of the NSM sister cells and that act through Region B. The core motif of a Snail-binding site (5′-CACCTG-3′) also represents an E-box motif(5′-CANNTG-3′), the DNA-binding site for members of the family of bHLH DNA-binding proteins, many of which function as transcriptional activators (reviewed by Massari and Murre,2000). Indeed, it has been suggested that members of the Snail family of transcription factors can functionally antagonise bHLH proteins by competing for binding to Snail-binding sites/E-boxes(Fuse et al., 1994; Kataoka et al., 2000; Nakayama et al., 1998). We therefore set out to test whether C. elegans bHLH proteins are involved in the specification of the NSM sister cell death in vivo. We were particularly interested in the C. elegans homologues of neuronal bHLH proteins, which can be divided into two families: the Achaetescute complex-related proteins and the Atonal-related proteins. The Atonal-related proteins are further subdivided into three groups: the Neurogenin group, the NeuroD group and the ATO group (reviewed by Hassan and Bellen, 2000; Lee, 1997). As tissue-specific bHLH proteins form DNA-binding heterodimers with ubiquitously expressed E proteins or Daughterless-like proteins, we also analysed C. eleganshomologues of this class of bHLH proteins. The C. elegans genome contains at least 35 bHLH genes, including one daughterless-like, five achaete-scute-like, one atonal-like and one NeuroD-like gene (reviewed by Ledent and Vervoort, 2001). Using existing mutants and RNA-mediated interference (RNAi)(Fire et al., 1998), we tested whether these genes are involved in specifying the death of the NSM sister cells.

The only C. elegans daughterless-like gene, hlh-2, has so far only been defined by weak lf mutations, such as bx108, which were identified in a screen for enhancers of the phenotype caused by a weak lf mutation of lin-32, the only C. elegans atonal-like gene(Portman and Emmons, 2000). bx108 is a missense mutation in the helix-loop-helix dimerization domain of HLH-2 and is predicted to affect the ability of the protein to form heterodimers with other bHLH proteins such as LIN-32(Portman and Emmons, 2000). However, bx108 has so far not been shown to cause a phenotype in an otherwise wild-type background. The inactivation of hlh-2 by RNAi results in embryonic lethality, indicating that hlh-2 is essential for development (Krause et al.,1997). To determine whether reducing hlh-2 function has an effect on the survival of the NSM sister cells, we analysed hlh-2(bx108) animals and `escapers' of hlh-2(RNAi) for the survival of NSM sister cells using the Ptph-1gfp reporter. We found that hlh-2(bx108)leads to the survival of up to 5% of the NSM sister cells, an effect that is temperature sensitive (Table 1A). Furthermore, we found that 15% of the NSM sister cells survived in hlh-2(RNAi) embryos that escaped early developmental arrest and developed to a stage, at which the Ptph-1gfpreporter is expressed (Table 1A).

To confirm that the additional GFP positive cells observed represent undead NSM sister cells and not other pharyngeal cells that have acquired a serotonergic fate, we analysed ces-1(n703gf); hlh-2(RNAi); Ptph-1gfp animals. In ces-1(n703gf); Ptph-1gfp animals, the NSM sister cells survive and therefore four GFP positive cells can be observed in the anterior pharynx, the NSMs and the undead NSM sister cells. If the additional GFP positive cells observed in hlh-2(RNAi);Ptph-1gfp embryos are not undead NSM sister cells, we would expect to detect more than four GFP positive cells in ces-1(n703gf); hlh-2(RNAi); Ptph-1gfp embryos. We found that 34%of the hlh-2(RNAi); Ptph-1gfp animals analysed had more than two GFP positive cells (n=45) and that only 2% of the ces-1(n703gf); hlh-2(RNAi);Ptph-1gfp animals (n=42) had more than four GFP positive cells. This indicates that the majority of additional GFP positive cells in hlh-2(RNAi); Ptph-1gfp embryos represent undead NSM sister cells.

The C. elegans atonal and NeuroD-like genes lin-32 (lin, lineage abnormal) and cnd-1 (cnd,C. elegans NeuroD), respectively, have been defined by non-lethal, strong lf mutations (Zhao and Emmons,1995; Hallam et al.,2000). To determine whether these two genes are involved in specifying the NSM sister cell death, we analysed NSM sister cell survival in animals carrying lf mutations in either gene. In cnd-1(ju29)animals, 0% of the NSM sister cells survived (n=60), indicating that the C. elegans NeuroD homologue is probably not required for the NSM sister cell death. Similarly, lf mutations of lin-32 did not affect NSM sister cell survival in an otherwise wild-type background or in a hlh-2(bx108) background (data not shown).

The five C. elegans achaete-scute-like genes (hlh-3, hlh-4,hlh-6, hlh-12 and hlh-14) have so far not been defined by mutations. For this reason, we used RNAi to analyse their potential role in the specification of the NSM sister cell death. No effect on NSM sister cell survival was observed in hlh-4(RNAi), hlh-6(RNAi), hlh-12(RNAi) or hlh-14(RNAi) animals(Table 1B and data not shown). hlh-3(RNAi), however, caused 7% of the NSM sister cells to survive (n=178), indicating that hlh-3 is at least partially required for the death of the NSM sister cells(Table 1B). Moreover, hlh-3(RNAi) [but not hlh-4(RNAi), hlh-6(RNAi), hlh-12(RNAi) or hlh-14(RNAi)] increased NSM sister cell survival in hlh-2(bx108) animals from 4% to 30%. hlh-2 and hlh-3 therefore might act together to cause the death of the NSM sister cells.

To determine whether reducing hlh-2 and hlh-3 function causes a general block in programmed cell death or specifically results in the survival of the NSM sister cells, we analysed the survival of other cells that normally die in hlh-2(bx108); hlh-3(RNAi)animals using Nomarski microscopy. During the development of the anterior pharynx, 16 cells undergo programmed cell death and mutations that block programmed cell death in general, such as egl-1(n1084 n3082), block many of these cell deaths. egl-1(n1084 n3082) animals therefore have on average about 12 extra cells in this part of the pharynx (Conradt and Horvitz,1998). We found that hlh-2(bx108); hlh-3(RNAi) animals have on average 1.3 extra cells(n=16). 57% of these extra cells were undead NSM sister cells, as confirmed by the position of their nuclei and by Ptph-1gfpexpression, and 23% possibly were undead m2 sister cells, as determined by the position of their nuclei. We were unable to determine the identity of 20% of the extra cells. Therefore, the majority of the surviving cells are NSM sister cells. Reducing the activity of hlh-2 and hlh-3, hence,results in the survival predominantly of the NSM sister cells.

The ability of the ces-2 lf mutation n732 to cause NSM sister cell survival depends on a functional ces-1 gene, which suggests that ces-1 acts downstream of ces-2. To determine whether the ability of hlh-2(RNAi) and hlh-3(RNAi) to cause NSM sister cell survival similarly depends on a functional ces-1 gene, we tested whether the ces-1 lf mutation n703 n1434 can block the NSM sister cell survival observed in hlh-2(RNAi) and hlh-3(RNAi) animals. In hlh-2(RNAi)animals, 14% of the NSM sister cells survived and in ces-1(n703 n1434lf) animals treated with control RNA, 0% survived (dsRNA delivered by feeding) (Timmons et al.,2001) (Fig. 4A). In ces-1(n703 n1434lf); hlh-2(RNAi) animals,14% of the NSM sister cells survived. Similarly, in hlh-3(RNAi) animals, 6% of the NSM sister cells survived and in ces-1(n703 n1434lf) animals treated with control RNA, 1%survived (dsRNA delivered by injection)(Fire et al., 1998). In ces-1(n703 n1434lf); hlh-3(RNAi) animals,5% of the NSM sister cells survived (Fig. 4B). These data indicate that a functional ces-1 gene is not required for the ability of hlh-2(RNAi) or hlh-3(RNAi) to cause NSM sister cell survival, which suggests that ces-1 is not acting downstream of hlh-2 and hlh-3. hlh-2 and hlh-3 therefore act downstream of or in parallel to ces-1 to kill the NSM sister cells.

To determine whether HLH-2 and HLH-3 can bind to the Snail-binding sites/E-boxes in Region B of the egl-1 locus, we performed EMSAs,using bacterially produced, affinity-purified His-tagged HLH-2 and HLH-3 fusion proteins. Using the 390 bp DNA fragment consisting of wild-type Region B, including all four Snail-binding sites/E-boxes as a probe, we found that homodimers of HLH-2 and heterodimers of HLH-2 and HLH-3 but not homodimers of HLH-3 could bind and shift the probe (Fig. 5A). Upon longer exposures and the use of various fragments of Region B, we could determine that HLH-2/HLH-3 heterodimers can bind to at least three of the four Snail-binding sites/E-boxes in vitro (data not shown). The amount of protein required to detect binding to Region B and to bind 50%of the probe was 10-fold and fivefold higher for CES-1 than for HLH-2/HLH-3,respectively. This suggests that at least in vitro, the binding affinity of HLH-2/HLH-3 for these Snail-binding sites/E-boxes is higher than the binding affinity of CES-1 [compare Fig. 3 (lane 3) with Fig. 5A (lane 13), and Fig. 3 (lane 4) with Fig. 5A (lane 15)]. The binding of HLH-2/HLH-3 heterodimers was much reduced when a DNA fragment consisting of mutant Region B, in which the four Snail-binding sites/E-boxes had been mutated (5′-CACCTG-3′ to 5′-CATATA-3′) was used as a probe(Fig. 5B, lanes 7-12). Using amounts of HLH-2/HLH-3 that are sufficient to bind 50% of the wild-type probe(100% binding), binding to mutant Region B was reduced to on average 4%(n=3) (compare Fig. 5B, lanes 4 and 10). HLH-2/HLH-3 binding to Region B in vitro therefore is dependent on functional Snail-binding sites/E-boxes. We confirmed the specificity of the binding observed by competition experiments (data not shown). These results suggest that HLH-2 and HLH-3 might cause the NSM sister cells to die by acting as a direct activator of egl-1 transcription in the NSM sister cells.

If HLH-2 acts as a direct activator of egl-1 transcription, HLH-2 should be present in the NSM sister cells at the time their cell-death fate is specified, i.e. in the NSM sister cells of embryos at the 1.5-fold stage of development (about 400-430 minutes after the first cleavage). HLH-2 has been shown to be broadly distributed throughout the embryo during the proliferative phase of embryogenesis (until about 350 minutes). During later stages of embryogenesis, the distribution of HLH-2 becomes progressively restricted(Krause et al., 1997). Using an antibody specific for HLH-2 (Krause et al., 1997), we stained 1.5-fold stage ced-3(n717) embryos and found that at that stage of development HLH-2 is still present in a number of nuclei predominantly in the head and tail region (Fig. 6). However, owing to the lack of a marker for the early NSM lineage, we were unable to determine whether the HLH-2-positive cells include the NSM sister cells. We therefore stained ced-3(n717) embryos carrying an integrated Pegl-1gfp transgene [Pegl-1gfp;ced-3(n717)], which expresses the gfp gene under the control of the cis-regulatory regions of egl-1. egl-1appears to be specifically expressed in cells destined to die during embryogenesis (R. Schnabel and B.C., unpublished), including, as shown above,the NSM sister cells. In a wild-type background, about 98 cells have undergone programmed cell death by the time an embryo reaches the 1.5-fold stage(Sulston et al., 1983). In a ced-3(n717) background, most if not all of these 98 cells survive (Ellis and Horvitz,1986). As egl-1 acts upstream of ced-3,ced-3(n717) does not interfere with the expression of the Pegl-1gfp transgene. In a Pegl-1gfp; ced-3(n717)embryo at the 1.5-fold stage, a large number of undead, GFP-positive cells can therefore be detected (Fig. 6). When staining these embryos for HLH-2, we found that a total of four cells expressed Pegl-1gfp and were positive for HLH-2(Fig. 6). All four cells are located in the head region where the developing pharynx is found. To determine whether two of these four cells may represent the NSM sister cells, we stained embryos of genotype ces-1(n703gf); Pegl-1gfp;ced-3(n717). As shown above, ces-1(n703gf)prevents the deaths of the NSM sister cells by blocking the expression of egl-1 in these cells. In a ces-1(n703gf); ced-3(n717) background, the distribution of HLH-2 and the expression of Pegl-1gfp in 1.5-fold stage embryos was overall unchanged; however, a total of only two cells expressed Pegl-1gfp and were positive for HLH-2 (Fig. 6). The two cells that are still positive for HLH-2 but no longer express Pegl-1gfp may well be the NSM sister cells. This result suggests that HLH-2 is most probably present in the NSM sister cells at the time their cell-death fate is determined.

Our results suggest that in the NSM sister cells, a repressor of egl-1 transcription, CES-1, as well as an activator of egl-1transcription, HLH-2/HLH-3, act on egl-1 transcription through the same DNA-binding sites in Region B of the egl-1 locus. To test this hypothesis in vivo, we sought experimental approaches that would allow us to analyse the effect of bHLH binding to Region B in the absence of CES-1 binding.

As shown above, CES-1 and HLH-2/HLH-3 bind to wild-type Region B with the four intact Snail-binding sites/E-boxes (referred to as `wild-type' sites) in vitro but fail to efficiently bind to mutant Region B with the four Snail-binding sites/E-boxes mutated to 5′-CATATA-3′ (referred to as`Snail^-/E-box^-' sites)(Fig. 3, Fig. 5B). In contrast to the consensus sequence for Snail binding (5′-CACCTG-3′), bases 3 and 4 of the consensus sequence for bHLH binding can be variable(5′-CANNTG-3′) (Massari and Murre, 2000). We therefore tested whether mutating the four Snail-binding sites/E-boxes in Region B to 5′-CATATG-3′(referred to as `Snail^-/E-box⁺' sites) would disrupt CES-1 binding but still allow the binding of bHLH proteins such as HLH-2/HLH-3 in vitro. As shown in Fig. 3,CES-1 binding to a probe consisting of Region B, in which the four Snail-binding sites/E-boxes have been mutated to Snail^-/E-box⁺, is severely reduced when compared with CES-1 binding to the wild-type probe (Fig. 3, compare lanes 1-6 with lanes 13-18). Using an amount of CES-1 protein that is sufficient to bind about 50% of the wild-type probe (100%binding), the introduction of the Snail^-/E-box⁺ mutation reduced binding to on average 6% (n=2) (compare Fig. 3, lanes 4 and 16). On the contrary, using an amount of HLH-2/HLH-3 that was sufficient to bind about 50%of the wild-type probe (100% binding), the introduction of the Snail^-/E-box⁺ mutation only reduced binding to on average 17% (n=3) (Fig. 5B; compare lanes 4 and 16). At least in vitro, the introduction into Region B of Snail^-/E-box⁺ mutations therefore affects the binding of CES-1 more dramatically than the binding of HLH-2/HLH-3.

To determine the effect of bHLH binding and CES-1 binding to the four Snail-binding sites/E-boxes in Region B in vivo, we introduced the Snail^-/E-box^- and Snail^-/E-box⁺mutations into Region B of the rescuing fragment pBC08 and analysed the ability of the resulting fragments to rescue NSM sister cell death in ces-2(n732ts); egl-1(n1084 n3082) animals. n732ts is a temperature sensitive lf mutation of ces-2: 13%of the NSM sister cells survive in ces-2(n732ts) animals raised at 15°C, the permissive temperature; and 73% survive in animals raised at 25°C, the non-permissive temperature(Fig. 7)(Ellis and Horvitz, 1991). ces-2(n732ts) has been proposed to cause NSM sister cell survival as a result of ces-1 overexpression in the NSM sister cells(Metzstein and Horvitz, 1999). Culturing transgenic ces-2(n732ts); egl-1(n1084 n3082) animals at 15°C and 25°C therefore allowed us to analyse the ability of the fragments to rescue the NSM sister cell death defect caused by egl-1(n1084 n3082) in the presence of slightly or strongly elevated levels of CES-1 protein in the NSM sister cells,respectively.

Using Ptph-1gfp to analyse NSM sister cell survival, we found that the wild-type fragment pBC08 rescued the death of the NSM sister cells in ces-2(n732ts); egl-1(n1084 n3082) animals grown at 15°C (5/5 independent lines) but not at 25°C (0/5 independent lines) (Fig. 7A). This suggests that an activator of egl-1 transcription, which is required to activate egl-1 transcription in the NSM sister cells, can function through cis-regulatory regions contained in pBC08, and that high levels of CES-1 in the NSM sister cells can block this activator. The ability of this activator to activate egl-1 transcription is dependent on functional Snail-binding sites/E-boxes in Region B of the egl-1 locus, as the Snail^-/E-box^- fragment (pBC181) failed to rescue NSM sister cell death in animals grown at 15°C (0/3 independent lines) or at 25°C (0/3 independent lines) (Fig. 6B). Finally, the Snail^-/E-box⁺ fragment(pBC182) rescued the NSM sister cell death in animals grown at 15°C (6/6 independent lines) and also in animals grown at 25°C (6/6 independent lines) (Fig. 7C). The ability of the activator to activate egl-1 transcription is therefore dependent on functional E-boxes but not on functional Snail-binding sites in Region B of the egl-1 locus. This indicates that the activator of egl-1 transcription in the NSM sister cells is most likely an E-box binding bHLH protein. In light of our in vivo data demonstrating a requirement for the genes hlh-2 and hlh-3 for NSM sister cell death and our in vitro data demonstrating binding of a HLH-2/HLH-3 heterodimer to the Snail-binding sites/E-boxes in Region B, we propose that a heterodimer of HLH-2 and HLH-3 activates egl-1 transcription in the NSM sister cells to cause these cells to undergo programmed cell death.

The results obtained with the Snail^-/E-box⁺ fragment furthermore indicate that the ability of high levels of CES-1 to block the activator of egl-1 transcription is dependent on functional Snail-binding sites in Region B of the egl-1 locus. As CES-1 and HLH-2/HLH-3 therefore act through overlapping DNA-binding sites, the NSM sister cells might survive in ces-2(n732ts) animals grown at 25°C because high levels of CES-1 protein successfully compete with HLH-2/HLH-3 for binding to Region B in the egl-1 locus.

We found that reducing the activity of hlh-2 or hlh-3 by RNAi results in the survival of 15% and 7% of the NSM sister cells,respectively. Reducing the activity of hlh-2 and hlh-3simultaneously by a weak hypomorphic mutation and by RNAi leads to the survival of on average 30% of the NSM sister cells. hlh-2 and hlh-3 are therefore at least partially required for the death of the NSM sister cells. Furthermore, a heterodimer composed of HLH-2 and HLH-3,HLH-2/HLH-3, can bind to Snail-binding sites/E-boxes in the egl-1locus in vitro, which are required to kill the NSM sister cells in vivo, and HLH-2 appears to be present in the NSM sister cells at the time their cell-death fate is determined. Hence, we propose that HLH-2/HLH-3 most probably acts as a direct activator of egl-1 transcription in the NSM sister cells.

hlh-2 and hlh-3 have so far not been defined by null mutations. It is therefore unclear whether the low penetrance of the Ces phenotype observed is due to additional factors that can kill the NSM sister cells in the absence of hlh-2 and hlh-3 or residual hlh-2 and hlh-3 activity in the NSM sister cells. It has been shown that C. elegans neurons are more resistant to the inactivation of gene function by RNAi than other cell types(Timmons et al., 2001). As this also appears to be the case for the NSMs and NSM sister cells (J.H. and B.C., unpublished), we favour the possibility that the low penetrance of the Ces phenotype observed is the result of the incomplete inactivation of hlh-2 and hlh-3 in the NSM sister cells. The inactivation by mutation of the Snail-binding sites/E-boxes in the egl-1 locus,through which we propose HLH-2/HLH-3 activates egl-1 transcription in the NSM sister cells, results in the complete failure to kill NSM sister cells. This observation suggests that, if additional factors exist that can kill the NSM sister cells in the absence of hlh-2 and hlh-3,they are most likely to be additional E-box binding proteins, i.e. additional bHLH proteins.

So far one lf mutation of ces-2, three gf mutations of ces-1 (which were found to carry the identical molecular lesion) and one lf mutation of ces-3, a gene that acts upstream of or in parallel to ces-1 and that remains to be characterized at the molecular level,have been identified in genetic screens for mutants, in which the NSM sister cells survive (Ellis and Horvitz,1991). However, these screens failed to recover mutations in hlh-2 and hlh-3. The failure to identify mutations in these two genes is possibly the result of one or more of the following observations. First, the fact that ces-2 and ces-3 have so far only been defined by one lf allele suggests that the screens performed to date were not saturating. Second, hlh-2(RNAi) results in embryonic lethality (Krause et al.,1997). However, previous screens were performed in a way that did not allow the identification of mutations in essential genes. Finally, it is possible that the complete inactivation of hlh-2 and hlh-3results in a NSM sister cell survival phenotype with low penetrance,decreasing the likelihood of identifying mutations in these genes accordingly.

bHLH proteins play key roles in the specification and execution of cell fates, including neuronal fates, in various organisms (reviewed by Lee, 1997; Massari and Murre, 2000). For example, in C. elegans, the daughterless-like gene hlh-2 and the atonal-like gene lin-32 function at multiple steps during the establishment of a specific neuronal sublineage,which gives rise to two neurons and one support cell(Portman and Emmons, 2000). To our knowledge, however, it has not yet been demonstrated that bHLH proteins play a direct role in the activation of programmed cell death. Our results indicate that the daughterless-like and achaete-scute-like genes hlh-2 and hlh-3 of C. elegans are at least partially required for the execution of the programmed cell death fate of the NSM sister cells. Furthermore, hlh-2 and hlh-3 may not be required for the specification and execution of the neuronal fate of the NSMs,as the Ptph-1gfp reporter continues to be expressed in the NSMs in animals, in which hlh-2 and/or hlh-3 have been at least partially inactivated. If hlh-2 and hlh-3 function in the specification or execution of the NSM fate as well, the requirement for their function for this process is less stringent than the requirement for their function in promoting the programmed death of the NSM sister cells.

It has been proposed that the NSM sister cells survive in ces-2(lf) or ces-1(gf) animals as a result of ces-1overexpression in the NSM sister cells(Metzstein and Horvitz, 1999). We have shown that CES-1 can bind to Snail-binding sites/E-boxes in the egl-1 locus in vitro, which are required for the ability of the ces-2 lf mutation n732 to block the death of the NSM sister cells in vivo. This suggests that in ces-2(lf) and ces-1(gf)animals, CES-1 blocks the death of the NSM sister cells by directly repressing egl-1 transcription.

Based on our finding that an activator of egl-1 transcription in the NSM sister cells, most probably HLH-2/HLH-3, acts through the identical Snail-binding sites/E-boxes in the egl-1 locus, we propose a molecular model for how CES-1 can repress egl-1 transcription(Fig. 8). We propose that in wild-type animals, in which CES-1 most likely is absent or present at low levels in the NSM sister cells, HLH-2/HLH-3 binds to the Snail-binding sites/E-boxes in the egl-1 locus thereby activating egl-1transcription in the NSM sister cells, which results in the death of these cells. In ces-2(lf) or ces-1(gf) animals, in which CES-1 levels in the NSM sister cells most likely are elevated, sufficient CES-1 protein is present to successfully bind to the Snail-binding sites/E-boxes thereby preventing HLH-2/HLH-3 from binding to these sites and from activating egl-1 transcription. This molecular model implies that in wild-type animals, CES-1 might play no role in specifying the cell-death fate of the NSM sister cells. This is supported by the fact that the ces-1 lf mutation n703 n1434 does not cause a phenotype in the NSM sister cells. Our model also suggests that in ces-1(gf) animals, CES-1 acts by blocking the function of HLH-2/HLH-3, which is supported by our finding that genetically ces-1 acts upstream of or in parallel to hlh-2 and hlh-3.

What might the function of ces-1 be in wild-type animals? The ces-1-like gene SLUG of mammals is required to protect hematopoietic progenitor cells from radiation-induced programmed cell death in vivo (Inoue et al., 2002). One speculative model therefore is that the cell-death protective activity of ces-1 might be required to protect the NSMs, the fate of which is to survive, from programmed cell death. However, in ces-1(lf) animals,the NSMs survive like in wild-type animals, which suggests that either ces-1 is dispensable for the survival of the NSMs or that its function is redundant. Hence, the normal function of ces-1 remains enigmatic.

It has previously been shown that the function of bHLH proteins can be antagonized by Snail-like proteins. For example, the Snail-like protein Escargot of Drosophila can repress the E-box-dependent transcriptional activation of a reporter gene by the bHLH proteins Daughterless and Scute in transfection assays(Fuse et al., 1994). In addition, it has been shown that the murine Snail-like protein mSna competes with a homodimer composed of the Daughterless-like protein E47 and with a heterodimer composed of E47 and the Achaetescute-like protein MASH-2 for binding to E-boxes in vitro and in cultured cells(Nakayama et al., 1998). Finally, the murine Snail-like protein Smuc can block the binding of a heterodimer composed of the Daughterless-like protein E12 and the bHLH protein MyoD to E-boxes in vitro and represses E12/MyoD-dependent activation of a reporter gene in transfection experiments(Kataoka et al., 2000). Our results now demonstrate that by binding to the Snail-binding sites/E-boxes in cis-regulatory regions of the egl-1 locus in vivo, the Snail-like protein CES-1 of C. elegans blocks the ability of HLH-2/HLH-3 to activate the cell-death activator gene egl-1 and to execute the cell-death fate of the NSM sister cells.

The cell death of at least two types of neurons in C. elegans is dependent on the transcriptional activation of the BH3-only gene egl-1: the male-specific death of the HSN neurons and the death of the NSM sister cells. Transcriptional regulation might therefore be an important mechanism through which the activity of EGL-1 is regulated. The activity of at least four of the 10 mammalian BH3-only proteins identified to date is regulated at the transcriptional level (reviewed by Puthalakath and Strasser,2002). For example, after DNA damage, the BH3-only genes Noxa and Puma/Bbc3 are transcriptionally upregulated in a p53-dependent manner in thymocytes and fibroblasts(Han et al., 2001; Nakano and Vousden, 2001; Oda et al., 2000; Yu et al., 2001). Furthermore,the BH3-only genes Hrk/DP5 and Bim have been found to be upregulated in cultured neurons after NGF withdrawal, a process that appears to be dependent on the activation of the c-jun N-terminal kinase (JNK)(Harris and Johnson, 2001; Imaizumi et al., 1999; Putcha et al., 2001; Whitfield et al., 2001). In sympathetic neurons, it could further be shown that the upregulation of Bim after NGF-withdrawal is required for their programmed death(Putcha et al., 2001; Whitfield et al., 2001). Hence, the observation that NGF-withdrawal induced death of sympathetic neurons in culture is dependent on the synthesis of macromolecules can be explained by the requirement for Bim expression. The transcriptional activation of BH3-only genes therefore is crucial for the programmed death of neurons not only in C. elegans, in which neuronal death is specified by cell lineage, but also for the programmed death of neurons in mammals, in which neuronal death is predominantly triggered by cell non-autonomous signals.

The cell-death function of ces-2 and ces-1 has been conserved through evolution. The homologue of ces-2 in humans is the proto-oncogene HLF (HLF, hepatic leukaemia factor)(Inaba et al., 1996). The oncogenic form of HLF, the E2A-HLF fusion protein, found in patients carrying the t(17; 19) (q22;p13) chromosomal translocation, is composed of the trans-activation domain of the Daughterless-like bHLH protein E2A and the DNA-binding domain of the CES-2-like protein HLF. E2A-HLF has been shown to block the programmed death of pro-B cells thereby allowing their leukaemic transformation and it has been proposed that it does so by inappropriately activating the transcription of a gene with anti-apoptotic function. The ces-1-like gene SLUG of humans was found to be a target of E2AHLF and the overexpression of SLUG can block the programmed death of leukaemic pro-B cells (Inukai et al.,1999). Furthermore, hematopoietic progenitor cells from mice lacking a functional SLUG gene are more sensitive to DNA-damage induced programmed cell death, supporting an anti-apoptotic role for SLUG in hematopoietic lineages(Inoue et al., 2002). A mammalian counterpart of the genetic pathway involved in the specification of the NSM sister cell death in C. elegans might therefore play an important role in the regulation of the programmed death of pro-B cells in mammals. We have shown that the Snail-like CES-1 protein can prevent the death of the NSM sister cells in C. elegans by antagonizing the function of HLH-2 and HLH-3 thereby directly blocking the transcription of the BH3-only gene egl-1 (Fig. 8). It is therefore feasible that homologues of C. elegans HLH-2 and HLH-3 in mammals act downstream of or in parallel to SLUG to activate a mammalian BH3-only gene in pro-B cells. Indeed,the Daughterless-like bHLH protein E2A, a mammalian homologue of HLH-2, is expressed in B-cell lineages and has been shown to have tumour suppressor activity (Massari and Murre,2000). Furthermore, a number of BH3-only genes have been shown to be expressed in hematopoietic cells in mammals, including Hrk/DP5 and Bim(Puthalakath and Strasser,2002). Whether Hrk/DP5 or Bim are targets of the conserved, HLF- and SLUG-dependent cell-death pathway in pro-B cells and whether E2A plays a role in their activation remains to be determined.

The BH3-only gene egl-1 of C. elegans is required for most if not all of the 131 programmed cell deaths that occur during C. elegans development. The studies described resulted in the identification of two new factors, hlh-2 and hlh-3, that are involved in specifying the death of two of these 131 cells, the death of the NSM sister cells. The death of the NSM sister cells is specified by lineage. Like their counterparts in higher organisms, hlh-2 and hlh-3, which encode a Daughterless-like and an Achaete-scute-like bHLH protein, respectively, have been implicated in cell fate determination. The lineage-dependent signal required to kill the NSM sister cells might therefore be transduced by hlh-2 and hlh-3.

We thank Scott Cameron, Mike Krause, Mark Metzstein, Heinke Schnabel, Simon Tuck and Nicole Wittenburg for comments concerning this manuscript; Scott Cameron, Bob Horvitz and Mark Metzstein for insightful discussions; Doug Portman for providing hlh-2(bx108); Yishi Jin for cnd-1(ju29); Mike Krause for plasmids pKM1195 and 1199, and for anti-HLH-2 antibodies; Melanie Dunn and Geraldine Seydoux for plasmid pJH2.19; and Ralf Baumeister for plasmid pBY668. We are grateful to Claudia Huber for technical support, Helga Zepter for secretarial support and Helma Tyrlas for DNA sequence determinations. This work was supported by funding from the Leukemia and Lymphoma Society of America, the Max-Planck Society and the Deutsche Forschungsgemeinschaft.