Two C. elegans genes control the programmed deaths of specific cells in the pharynx

During the development of both vertebrates and invertebrates, many cells are bom that neither divide nor form part of the adult, but instead die (Saunders, 1966; Cowan et al. 1984; Truman, 1984). Programmed cell death removes cells that do not function or that function only transiently (Finlayson, 1956; Sulston et al. 1983), influences morphogenesis (e.g. Hinchliffe, 1981; Sternberg and Horvitz, 1981), regulates the size of neuronal populations (Hamburger and Oppenheim, 1982) and helps create sexual dimorphism (e.g. Hinchliffe, 1981; Sulston et al. 1983). One animal in which programmed cell death has been studied in detail is the nematode Caenorhabditis elegans. Of the 1090 somatic cells generated during the normal development of a C. elegans hermaphrodite, 131 die. The same cells die in every animal, each at its own precise time, and all show similar changes in morphology as they undergo cell death (Sulston and Horvitz, 1977; Sulston et al. 1983).

Genetic studies of C. elegans have defined ten genes that function in the programmed deaths of these 131 cells. Eight of these genes are required for the degradation of dead cells. Mutations in the genes ced-1, ced-2, ced-5, ced-6, ced-7, ced-8 and ced-10 prevent dead cells from being engulfed and degraded by their neighbors (ced=cell death;Hedgecock et al. 1983; Ellis et al. 1991), and mutations in the gene nuc-1 prevent the DNA of a dead cell from being digested (nuc=nuclease;Sulston, 1976; Hedgecock et al. 1983). Two genes, ced-3 and ced-4, are required for the deaths of all 131 cells (Ellis and Horvitz, 1986). Mutations in either of these genes prevent all of these cell deaths from occurring. Mosaic analyses indicate that ced-3 and ced-4 act within dying cells, possibly as part of a cellular suicide program (Yuan and Horvitz, 1990). These results suggest that ced-3 and ced-4 either encode or regulate the expression of cytotoxic substances.

How are the genes that act in programmed cell death regulated? Cell deaths occur throughout C. elegans development (Sulston and Horvitz, 1977; Sulston et al. 1983), and cells that are prevented from dying by mutations in ced-3 or ced-4 develop into many different cell types (Ellis and Horvitz, 1986; Avery and Horvitz, 1987). Because of the diversity in origin and cell type among dying cells, there could well be regulatory genes that control the fates of some dying cells but not of others. Such regulatory genes might choose between life and death for specific cells by controlling the activities of ced-3 and ced-4 in specific cell types. To understand how cell death is initiated in the correct cells during development, it is important to identify such regulatory genes and determine how they interact with the genes that cause cells to die.

We have identified and characterized two genes that control the decision by specific cells to live or die: ces-1 and ces-2 (ces=cell death specification). Mutations in these genes can prevent specific cell deaths without affecting the deaths of other cells. We show that these genes act together to determine the fates of two cells in the pharynx, the sisters of the NSM neurons, probably by regulating the cell death genes ced-3 and ced-4.

Techniques for culturing C. elegans are described by Brenner (1974). All strains were grown at 20 °C unless otherwise indicated. The wild-type parent of all strains described here is C. elegans variety Bristol strain N2 (Brenner, 1974). Genetic nomenclature is described by Horvitz et al. (1979).

We used the following mutations as genetic markers: LGI, dpy-5(e61), dpy-14(el88), unc-13(e51), unc-13(e1091), lin-10(el439), daf-8(el393), fer-l(hcl), sup-17 (n1258), unc-29(e1072), ced-l(el752), unc-59(e1005), lev-10(xl7), unc-54(e1092), let-208(el719), let(e2000); LGIII, ced-4(nll62); LGIV, ced-2(el752), him-8(el489), unc-31(e928), ced-3(n717); LGV, him-5(el490). We used the following chromosomal rearrangements: nDf23 I, nDf24 I, nDf25 I, eDf3 I, eDf4 I, eDf6 I, eDf9 I, eDf10 I, eDf13 I, eDf14 I, nDp4(I;unknown), sDpl(I;f). The mutations unc-54(e1092) and let(e2000) (previously known as let(r202)), are described by Anderson and Brenner (1984), both sup-17(nl258) and nDp4 have been characterized by J. Thomas (personal communication), the mutations unc-13(e1091) and unc-31(e928) are from the strain collection established by Brenner (1974), and the other mutations are described by Hodgkin et al. (1988).

Techniques for staining worms with anti-serotonin antiserum are described by Desai et al. (1988).

Procedures for using Nomarski microscopy to observe living animals are described by Sulston and Horvitz (1977). Because it is difficult to count cells in the pharynx of an animal that is moving and feeding, we anesthetized the worms by mounting them in a drop of M9 salt solution (Sulston and Brenner, 1974) containing 30 mM NaN₃ (Avery and Horvitz, 1987). We usually studied the survival of cells in worms during the L2, L3 or L4 larval stages, although occasionally we scored adults or LI larvae.

The mutations n703 and n732 were isolated by N. Tsung and C. Trent (Trent, 1982), who used the technique of formaldehyde-induced fluorescence (Sulston et al. 1975; Horvitz et al. 1982) to identify animals that had unusual patterns of serotonin expression from among the broods of 4594 F₂ worms isolated after mutagenesis of the wild type with ethyl methanesulfonate.

We used the technique for mutagenesis with ethyl methanesulfonate (EMS) described by Brenner (1974). In the reversion of ces-1(n703), the number of F₁ animals scored equals the number of haploid genomes screened. However, in the screen for new ces mutants, we screened 21000 F₂ animals from among the descendants of about 27000 F₁ animals from an EMS mutagenesis, which corresponds to 54 000 mutant haploid genomes. Some of the chromosomes from F₁ animals would have been homozygous in more than one F₂ animal scored, so we estimated the probability of scoring one particular haploid genome present among the F₁ to be l-(53,999/54000)¹⁰⁵⁰⁰=0.18. Thus we screened (54000)(0.18)=9720 haploid genomes for recessive mutations. Similar considerations indicate that we screened about 24000 haploid genomes for dominant mutations.

L. Avery and H. Ellis (personal communications) had observed linkage of ces-1(n703) to markers on LGI. The following cross confirms this linkage, and demonstrates that ces-1(n703) is located to the left ot sup-17: 30/34 Une non-Sup recombinant progeny of unc-13(e51) sup-17/ ces-1(n703) heterozygotes segregated ces-1(n703). The other dominant alleles of ces-1 also map to the left of sup-17’. 12/13 Une nonSup recombinant progeny of unc-13(e51) sup-17/ ces-1 (n1895) heterozygotes segregated ces-1 (n1895), and 11/12 Une nonSup recombinant progeny of unc-13(e51 ) sup-17/ ces-1(n1896) segregated ces-1(n1896). The following cross demonstrates that ces-1 (n703) is probably located to the right of both daf-8 and fer-1, and to the left of unc-29: 7/8 homozygous Daf non-Fer Une recombinant progeny from daf-8 ces-1 (n703)/ lin-10 fer-1 unc-29 hermaphrodites carried ces-1 (n703). The mutations ces-1 (n1895) and ces-1(n1896) are also located to the right of daf-8 and to the left of unc-29: 0/1 Unc-29 Fer non-Daf non-Unc-13 recombinant progeny of ces-1(n1895)/ unc-13(e51) daf-8 fer-1 unc-29 heterozygotes segregated ces-1(n1895) and 2/2 Unc-29 non-Fer non-Daf non-Unc-13 recombinants segregated ces-1 (n895). From an equivalent experiment using ces-1 (n1896), we found 0/2 Unc-29 Fer non-Daf non-Unc-13 recombinant progeny segregated ces-1(n1896), and 6/6 Unc-29 non-Fer non-Daf non-Unc-13 progeny segregated ces-1 (n1896).

We mapped the location of ces-1 (n703 n1434) by examining homozygous recombinant F₂ progeny of daf-8 fer-1 unc-29/ lin-10 ces-1(n703 n1434) heterozygotes: 0/8 Lin Daf Fer Une recombinants, 0/1 Lin non-Daf Fer Une recombinants, and 0/2 Lin non-Daf non-Fer Une recombinants had surviving NSM sisters. Because ces-1(n703) maps much closer to fer-1 than it does to unc-29 (only 1/33 recombination events in this interval separated fer-1 and ces-1 (see above, also unpublished data)), one or both of the two lin-10 daf-8(+) fer-1 (+) unc-29 recombinant chromosomes probably carried the ces-1(n703) mutation. Neither of these animals showed the Ces-l(n703) phenotype, so the suppressor mutation n1434 must either map to the left of n703, or near to n703 but on its right. We also tested one of the lin-10 daf-8 fer-1 unc-29 chromosomes, the lin-10 daf-8(+) fer-1 unc-29 chromosome, and the lin-10 daf-8(+) fer-1 (+) unc-29 chromosomes for suppression of ces-2. Only the two lin-10 daf-8(+) fer-1(+) unc-29 recombinants carried a suppressor of ces-2, presumably the n1434 mutation. These results show that n1434 is located to the right of daf-8. Overall, these experiments show that n1434 maps very near to n703.

Similarly, to map the location of n1406, we examined homozygous recombinant F₂ progeny of daf-8 fer-1 unc-29/ lin-10 ces-1 (n703 nl406) heterozygotes: 0/6 Lin Daf Fer Une and 0/3 Lin non-Daf non-Fer Une recombinants had surviving NSM sisters. We also tested each of the three lin-10 daf-8(+) fer-1 (+) unc-29 chromosomes for suppression of ces-2(n732), and all three chromosomes carried a suppressor of ces-2, presumably nl406. These results indicate that, if n1406 is located to the right of n703, it must be close to the n703 mutation and that, if nl406 is located to the left of n703, it must lie in the small region that extends from just left of the gene daf-8 to ces-1 (n703).

In preliminary experiments, ces-2(n732) showed weak linkage to the gene dpy-5 on LGI (data not shown). The following cross confirms this assignment: 16/16 Ced non-Unc recombinant progeny of ces-2/ ced-1 unc-54 heterozygotes segregated ces-2(n732).-Furthermore, ces-2 is located to the right of unc-54 and possibly to the right of let-208: 4/4 Une non-Let recombinant progeny of ces-2/unc-54 let-208 heterozygotes segregated ces-2(n732). To complete the mapping of ces-2, we examined male progeny from the following crosses involving deletions in this region: ces-2 males mated with eDf9/ let(e2000) hermaphrodites yielded 13/20 Ces cross progeny, ces-2 males mated with eDf6/let(e2000) hermaphrodites yielded 8/16 Ces cross progeny, ces-2 males mated with eDf4/let(e2000) hermaphrodites yielded 18/35 Ces cross progeny, ces-2 males mated with eDfl3/let(e2000) hermaphrodites yielded 1/36 Ces cross progeny, and ces-2 males mated with eDfl4/let(e2000) hermaphrodites yielded 1/22 Ces cross progeny. Note that deletions that fail to complement ces-2(n732) show a semidominant effect on the survival of the NSM sisters (see below), as does ces-2(n732) itself; however, the NSM sisters survive much less often in these heterozygotes than in ces-2(n732)/Df animals. Because of the semidominant behavior of n732 and the deficiencies, these complementation tests are not conclusive, but indicate that ces-2 probably lies in the region deleted by eDf4, eDf6, and eDf9, and does not lie in the region deleted by either eDfl3 or eDf4.

The mutation nl952 maps to a different location than does ces-1. From lin-10 ces-1(n703 n1406) unc-29/ + + + ; ces(n1952)/ + heterozygotes, 2/6 Ces(nl952) progeny segregated Lin Une animals. We tested one of these two strains by crossing with the wild type, and recovered the ces(n1952) mutation from this lin-10 ces-1 (n703 nl406) unc-29; ces(nl952) strain, which confirmed its genotype. These results indicate that ces(n1952) is not located between lin-10 and unc-29, and so cannot be an allele of ces-1, and suggest that ces(nl952) may not be linked to ces-1.

The mutation nl952 is also not an allele of ces-2. Because ces-2(n732) is tightly linked to the unc-54 gene, we tested ces(nl952) for linkage to unc-54. We examined at 25°C four Ces(nl952) progeny of unc-54/+; ces(nl952)/+; him-8/+ heterozygotes. One of these progeny was unc-54/unc-54, one was unc-54/+, and two were +/+. These results indicate that ces(n1952) is not closely linked to unc-54, and thus is not closely linked to ces-2 either.

To determine if ces-l(n703 nl4O6) and ces-1 (n703 nl434) fail to complement for the suppression of ces-2, we mated lin-10 ces-1 (n703 n1434) ced-1 ces-2/ + + + ces-2 males with unc-13(e51) lin-10 ces-1 (n703 n1406) ced-1 ces-2 hermaphrodites at 25°C. Among the non-Unc cross progeny, only 1/12 Lin worms showed the Ces-2 phenotype of NSM sister survival (these animals were ces-1(n703 n1406)/ces-1 (n703 n1434)), but 16/26 non-Lin worms showed the Ces-2 phenotype (these animals were ces-1(n703 n1406)/+). Therefore ces-1(n703 n1406) and ces-1(n703 n1434) fail to complement for the suppression of ces-2.

To test ces-2(n732) and ces(n1952) for complementation, we mated ces-2(n732) males with ces(n1952) hermaphrodites at 25°C. Among the male cross progeny, only 1/17 worms showed the Ces phenotype, approximately the frequency at which ces-2/ + animals show this phenotype.

Triple mutants with the ces genes and ced-1; ced-2

In ced-1 or ced-2 mutants, some cell corpses are not engulfed (Hedgecock et al. 1983), and, in ced-1; ced-2 double mutants, the number of unengulfed corpses increases significantly (Ellis et al. 1991). For example, the NSM sister corpses are found in a ced-1; ced-2 double mutant, but not in ced-1 or ced-2 single mutants (Ellis et al. 1991). We therefore constructed ces-1(n703) ced-1; ced-2 and ced-1 ces-2(n732); ced-2 strains by selecting for recombinant progeny from the appropriate triple heterozygotes in which (1) the extra neurons characteristic of ces-1 and ces-2 mutants were present, and (2) the number of cell corpses was otherwise characteristic of the ced-1; ced-2 double mutant. The presence of both ced-1 and ced-2 in the ces-1 ced-1; ced-2 strain was verified by failure to complement ced-1 and ced-2, and the presence of both ced-1 and ced-2 in the ced-1 ces-2; ced-2 strain was verified by the presence of NSM sister corpses at 15 °C (at which temperature the NSM sisters die in ces-2 animals, as in the wild type).

In all gene dosage studies, animals were anesthetized with NaN₃ and examined using Nomarski microscopy to determine how often the NSM sisters and the 12 sisters survive. The location of the NSM sisters at the rear of the anterior bulb is distinctive, but the 12 sisters are located extremely close to other cells that also die. As noted in the text, the number of pharyngeal cells varies slightly in the wild type, so we first determined the variability in the number of cells near the 12 neurons to determine if this variability would introduce significant errors into our assay for 12 sister survival. In control wild-type animals at 15 °C, neither NSM sister survived in 95 animals scored, but an extra neuron was found near an 12 neuron in 21 of these animals. In 20 of the 21 cases, the extra cell was found on the left side. Most pharyngeal development is left/right symmetrical, and this asymmetry suggests that the extra cell might be the surviving sister of MCL, since the sister of the right-side cell MCR normally survives, whereas the sister of the left-side cell MCL normally dies (Sulston et al. 1983). At 20°C the NSM sisters again all died, but in 19/205 animals an extra cell was present near an 12 neuron. Similarly, at 25°C the NSM sisters all died, but in 7/100 animals an extra cell was found near an 12 neuron. In both cases, extra cells were found more often on the left side than on the right side. The extra cells found near the 12 neurons in the wild type could be (1) surviving 12 sisters, (2) other surviving cells, such as II sisters or 12 aunts or, when found on the left side of the pharynx, MCL sisters, or (3) neurons generated by extra cell divisions. In the wild type, the cold-sensitive survival of the MCL sister, and the rare survival of unknown cells occur infrequently, and so should not introduce significant error into the measurement of 12 sister survival in the various ces strains that we have examined.

To determine if a deletion of the ces-1 gene results in the dominant phenotype caused by ces-1 (n703), we examined 20 unc-13(e1091) lin-11/ nDf23 animals, and 20 unc-13(e1091) Un-11/ nDf24 animals. The NSM sisters always died in these animáis, and 79/80 12 sisters clearly died. An extra neuron was located near the right 12 neuron in one animal; as noted above, an extra cell is sometimes found in this position in the wild type.

To measure the effects of altering ces-1 gene dosage, we examined animals of several genotypes in order to determine how often the 12 sisters and NSM sisters survived. We only present data for the 12 sisters because in all strains with one copy of ces-1(n703) the NSM sisters survive about 90% of the time, so that NSM sister survival cannot be used to measure the effects of small changes in gene dosage (Table 1, Fig. 4, data not shown). The three deficiencies of the ces-1 gene behave similarly in trans to ces-1(n703), consistent with the genetic map data (Fig. 1), which indicates that each deficiency should completely eliminate ces-1 function. Furthermore, in the ces-1(n703)/ + animals, it appears to make no difference if the ces-1 mutation is derived from the mother or from the father, and the results also show little dependence on which marker mutations were used (Table 1). That ces-1 (n703)/+ animals have more surviving 12 sisters than do ces-1 (n703)/Df animals indicates that the ces-1 (+) allele enhances the effect of one copy of ces-1(n703).

We also examined animals carrying a wild-type copy of the ces-1 gene on a duplication. The attached duplication nDp4 covers the ces-1 gene. Animals homozygous for nDp4 are very sick and can be identified easily, and animals with a single copy of nDp4 are defective in egg-laying (J. Thomas, personal communication). In the nDp4/+ animals that we examined, which had three wild-type copies of the ces-1 gene, 30/30 NSM sisters died and, in nDp4/nDp4 animals, which had four copies of ces-1, 20/20 NSM sisters died. The free duplication sDp1 also covers ces-1, and 12/12 NSM sisters died in dpy-5 unc-13 (e51)/ dpy-5 unc-13(e51)/ sDp1 animals, which have three copies of ces-1. These results suggest that increasing ces-1 gene expression even two-fold is not enough to cause the NSM sister survival seen in ces-1 (gf) mutants.

In addition, we used the attached duplication nDp4 to study the survival of the 12 sisters in animals with two copies of ces-1(n703) and one copy of ces-1(+). Among the progeny of ces-1(n703) males crossed with unc-13(e1091) ces-1(n703); nDp4 homozygotes, 32/72 12 sisters survived, and among the healthy non-Unc progeny of unc-13(e1091) ces-1 (n703); nDp4/+ heterozygotes, 75/107 12 sisters survived. We also studied animals with two copies of ces-1 (+) and one copy of ces-1 (n703). Among the progeny of wild-type males crossed with unc-13(e1091) ces-1(n703); nDp4 homozygotes, 5/32 12 sisters survived, and among the progeny of ces-1 (n703) males crossed with unc-13(e51); nDp4 homozygotes, 11/4112 sisters survived. Thus, the duplication nDp4 appears to lower, not enhance, the effect of two copies of ces-1(n703) on the survival of the 12 sisters. Experiments involving these large duplications are difficult, because they cause general sickness in animals, and the pharyngés are sometimes distorted and difficult to score. Furthermore, we cannot prove that the effect caused by these duplications on 12 sister survival is a consequence of the extra copy of ces-1 (+) that they contain. The fact that with one copy of ces-1(n703) the ces-1 (+) allele enhances the effect of the ces-1(n703) mutation, whereas with two copies of ces-1(n703) the ces-1(+) allele appears to suppress ces-1 (n703), suggests that the regulation of ces-1 activity is complicated. Perhaps the ces-1 (+) product both increases ces-1(n703) expression by a trans-acting autoregulation, and also competes with ces-1(n703) product for some limiting molecule necessary for ces-1 function. In this case, the overall effect of increasing ces-1 (+) activity could depend upon the level of ces-1(n703) activity.

The mutation ces-2(n732) is not completely recessive. From a cross of ces-2(n732) males with unc-13(e51) lev-10 hermaphrodites at 25 °C, 4 % of the cross progeny had a surviving NSM sister (n = 100 animals), and from a cross of wild-type males with ces-2(n732); unc-31 hermaphrodites at 25°C, 7% of the cross progeny had a surviving NSM sister (n=100 animals). Similarly, the NSM sisters sometimes survive in animals heterozygous for a deletion of ces-2. At 25 °C, the NSM sisters survived in 12/30 eDf4/let(e2000) animals (a total of 13/60 cells survived), and the NSM sisters survived in 9/18 eDf9/let(e2000) animals (13/36 cells survived). By contrast, the nearby deletions eDf13 and eDf14 do not remove ces-2, and in both eDfl3/let(e2000) and eDfl4/let(e2000) animals at 25°C all of the NSM sisters died (n=30 animals each). As noted above, the NSM sisters always die in the wild type as well. (The deficiency eDf10 might show a partial failure to complement ces-2(n732), and eDf3/eDf10 heterozygotes have a low rate of NSM sister survival; data not shown).

We examined animals carrying different ces-1 dominant alleles in trans. Among the cross progeny of ces-1 (n703) males mated with unc-13(e51) ces-1(nl895) hermaphrodites 161/200 12 sisters survived, among the cross progeny of ces-1 (n703) males mated with unc-13(e51) ces-1 (n1896) hermaphrodites 159/200 12 sisters survived, among the cross progeny of ces-1(nl896) males mated with unc-13(e51) ces-1 (n1896) hermaphrodites 168/22012 sisters survived, and among the cross progeny of ces-1(n1895) males mated with unc-13(e51) ces-l(n1895) hermaphrodites 160/200 12 sisters survived. Finally, among the cross progeny of ces-l(n1895) males mated with unc-13(e51) ces-l(n1896) 193/234 12 sisters survived, and among the cross progeny of ces-1 (n1896) males mated with unc-13(e51) ces-l(n1895) hermaphrodites 214/300 12 sisters survived. These values are all similar.

Preliminary observations using Nomarski microscopy suggested that both ces-l(n703 nl4O6) and ces-l(n703 n1434) animals (n=10) appear wild-type. We further examined ces-l(n703 n1406) animals in the following ways. At 25°C, two unc-13(e51) lin-10 ces-1 (n703 nl406) animals each had 80 cells in the pharynx, the number found in the wild-type. Furthermore, at 25°C, nine lin-10 ces-l(n703 nl406) unc-29 animals, examined less thoroughly, all had the wild-type number of neurons and epithelial cells in the pharynx, and at 20°C, 19/20 lin-10 ces-1 (n703 n1406) unc-29 animals had the wild-type number of neurons and epithelial cells in the anterior pharynx, and 1/20 had a single extra neuron near the 12 cell. We also examined the Une male progeny of lin-10 ces-1(n703 nl406) unc-29/ + + + males crossed with nDf24 unc-13(e1091) lin-11 hermaphrodites. At 25°C, 3/4 such lin-10 ces-1(n703 n1406) unc-29/ nDf24 animals had the wild-type number of neurons and epithelial cells in the pharynx, and 1/4 may have been missing one neuron in the posterior pharynx; at 20°C, 6/6 such animals appeared wild-type. Thus, it appears that n1406 and n1434 mutants may indeed have a wild-type phenotype; even ces-1(n703 n1406)/Df worms at 25°C, which should have very little ces-1 activity, appear wildtype.

Because ces-1 (If) mutations result in a wild-type phenotype, we used ces-1 (If) mutations closely linked to marker mutations in the constructions described in this section. Furthermore, we studied several isolates of most of these strains to ensure that the ces-1 (If) allele had not been lost by a rare recombination event in any particular construction. All markers segregated at expected frequencies in these crosses (data not shown), so we do not believe that ces-1 (If) alleles have any lethal interactions with the other genes that we used.

Double mutants between the ces-1 (If) mutations and ced-3 or ced-4 mutations were built as follows. From lin-10 ces-1(n703 nl434)/++; ced-3/ + heterozygotes we isolated Ced progeny and, from these Ced progeny, we isolated Lin offspring. These animals are lin-10; ced-3 based on their phenotypes, and probably are homozygous for ces-1(n703 n1406), which is tightly linked to lin-10. In four separate Lin Ced-3 isolates the NSM sisters always survived. A lin-10 ces-1(n703 nl406); ced-4 strain was constructed similarly, and three separate isolates all showed the same phenotype. We built unc-13(e51) lin-10 ces-1(n703 nl406); ced-3 using an equivalent procedure, and tested three separate Une Lin Ced isolates for NSM sister survival. The strain unc-13(e51) lin-10 ces-1 (n703 n1406); ced-4 was built similarly, and one isolate tested. The survival of the NSM sisters in the above strains was tested using Nomarski optics for all isolates, and with anti-serotonin staining for one isolate of each strain.

To construct the lin-10 ces-1(n703 nl434) ced-1 ces-2 strain, we isolated Lin Ced recombinant progeny from a lin-10 ces-1(n703 nl434) + +/+ + ced-1 ces-2 heterozygote. Our results indicated that ces-2 is suppressed by ces-1 (If), so to ensure that ces-2 was present on the lin-10 ced-1 chromosome, we mated ces-2 males into the Lin Ced strain and selected progeny in which both NSM sisters survived. Because ces-2 is recessive for this trait, these animals must be ces-2 homozygotes. We then re-isolated Lin Ced worms from among the progeny of the putative lin-10 ces-1 (n703 n1434) ced-1 ces-2/+ + + ces-2 heterozygotes. The NSM sisters die in these Lin Ced animals, which indicates that the ces-1 (If) allele is also homozygous in these worms. Genetic mapping (see above) proves that the activity that suppresses ces-2 is tightly linked to ces-1, and so is not caused by one of the other markers in this strain. Therefore, this strain must be of genotype lin-10 ces-1 (n703 n1406) ced-1 ces-2. As a final test that ces-2 is homozygous in this strain, we showed that 8/8 putative ltn-10 ces-1(n703 n1406) ced-1 ces-2 animals behave as ces-2 homozygotes when tested for complementation with ces-2.

To construct an unc-13(e51) lin-10 ces-1(n703 n1406) ced-1 ces-2 strain, we isolated three Une Lin Ced recombinants from unc-13(e51) lin-10 ces-1(n703 n14O6) + + /+ + + ced-1 ces-2 heterozygotes. The NSM sisters die in all three of these recombinant strains, so if ces-2 is homozygous, the ces-1 (If) allele must also be homozygous so that it suppresses ces-2. We showed that 5/5 putative unc-13 lin-10 ces-1 (n703 n1406) ced-1 ces-2 worms from each of these three strains behave as ces-2 homozygotes when used in complementation tests with ces-2, so we conclude that each strain is homozygous for ces-2. Thus each strain is of the genotype unc-13 lin-10 ces-1 (n703 n1406) ced-1 ces-2.

To construct an unc-13(e51) lin-10 ces-1(n703 nl406) ced-1 ces-2; ced-3 strain, we isolated an Une Lin Ced-3 animal from an unc-13(e51) lin-10 ces-1(n703 nl406) ced-1 ces-2/+ + + + ces-2; ced-3/ + heterozygote. The ces-1(lf) allele is still present in this strain because, in a cross of ces-2 males with the putative unc-13(e51) lin-10 ces-1 (n703 n1406) ced-1 ces-2; ced-3 hermaphrodites, the F₁ animals segregate Une animals in which the NSM sisters die (except those F₂ worms homozygous for ced-3, in which all cells, including the NSM sisters, live).

Finally, we constructed a lin-10 ces-1(n703 n1406) unc-29; n1952 strain by first isolating animals in which one or both NSM sisters survived from among the progeny of lin-10 ces-1 unc-29/ + + + n1952/+ heterozygotes. From one of these n1952 homozygotes, we isolated Lin Une progeny. The deaths of the NSM sisters in this strain were confirmed using Nomarski optics.

The mutations n703 and n732 were isolated by N. Tsung and C. Trent (Trent, 1982). In the pharyngés of wildtype animals only the two bilaterally symmetric NSM neurons contain serotonin (Horvitz et al. 1982); by contrast, in the pharyngés of both n703 and n732 animals, there are four serotonergic cells (Trent, 1982). Nomarski microscopy revealed that there are two extra neurons in this region of the pharynx (H. Ellis, personal communication). We mapped the n703 and n732 mutations, and showed that each defines a new gene on linkage group I (Materials and methods, Fig. 1). Because these genes appear to be involved in the specification of which cells live and which cells die (see below), we have named them ces-l(n703) and ces-2(n732), where ces stands for cell death specification.

In wild-type animals, the two NSM cells differentiate into serotonergic neurons, and the sisters of the NSM neurons die. By contrast, in ced-3 mutants, the sisters of the NSM neurons along with many other cells fail to die, and there are four cells in the pharynx that contain serotonin (Ellis and Horvitz, 1986). Some cells that are prevented from dying by a mutation in ced-3 adopt the fate of a near relative, which suggests that the two extra serotonergic cells in ced-3 animals are the surviving NSM sisters (Ellis and Horvitz, 1986). The NSM sisters might similarly survive in ces-1 and ces-2 mutants, which would account for the two extra neurons and the total of four serotonergic cells in the pharyngés of these animals.

To see if the NSM sisters fail to die in ces-l(n703) and ces-2(n732) animals, we examined mutant larvae using Nomarski microscopy. We observed in both ces-1 and ces-2 mutants an extra neuron just posterior and dorsal to each NSM neuron (Fig. 2). This position is exactly that of the NSM sisters in wild-type animals before they die (Sulston et al. 1983). Furthermore, staining with anti-serotonin antisera (Desai et al. 1988) revealed that the extra serotonergic cells in ces-1 and ces-2 mutants strongly resemble those found in ced-3 animals, in which cell deaths do not occur; in all three mutants, the nerve processes of these cells are similar in morphology to those of the NSM neurons (data not shown). While viewing ces-1 mutants with Nomarski microscopy, we also identified two additional extra neurons located anterior to the 12 neurons, one on each side of the pharynx (Fig. 2). These results suggest that the NSM sisters indeed fail to die in ces-1 and ces-2 animals, and that two additional cells (possibly the 12 sisters) fail to die in ces-1 worms.

It remained possible that these extra neurons resulted from extra cell divisions, rather than from the survival of cells that normally die. To explore this possibility, we directly examined cells that die in this region of the pharynx. In ced-1; ced-2 double mutants, the corpses of dead cells are not quickly degraded and instead persist for hours; these corpses can be assayed reliably (Hedgecock et al. 1983; Ellis et al. 1991). (Note that in C. elegans genetic nomenclature a semicolon separates genes located on different chromosomes; Horvitz et al. 1979). In particular, two specific corpses, which by position are likely to be the dead NSM sisters, are easily visible in ced-1; ced-2 double mutants (Fig. 3A). If the NSM sisters do not die in ces-1 and ces-2 animals, then the corpses of the NSM sisters should be missing in ces-1(n703) ced-1; ced-2 and in ced-1 ces-2(n732); ced-2 triple mutants.

We constructed the appropriate triple mutants (see Materials and methods) and examined newly hatched animals to see which cell corpses were present and which were missing. The putative NSM sister corpse found posterior to each NSM neuron in ced-1; ced-2 animals is present 100-fold less often in ces-1 (n703) ced-1; ced-2 animals (Table 2, Fig. 3B). Often a cell corpse anterior to the 12 neuron is also missing in ces-1 (n703) ced-1; ced-2 worms, which suggests that the extra neuron anterior to the 12 in ces-1 animals is also a cell that fails to die. Thus, two pairs of corpses are missing in ces-1 pharyngés, and these corpses correspond in position to the two pairs of extra cells found in these mutants (Fig. 3).

In ced-1 ces-2; ced-2 mutants, only the putative NSM sister corpse is affected by the presence of the ces-2 mutation. This corpse is present about seven times more often in ced-1; ced-2 worms than in ced-1 ces-2(n732ts); ced-2 animals raised at 25°C (Table 2). Furthermore, these data reveal that the ces-2(n732) mutation is strongly temperature-sensitive. Thus, in ces-2 mutants at 25 °C, one pair of pharyngeal corpses is missing, and these missing corpses correspond in position to the extra pair of serotonergic neurons found in these animals.

These results strongly suggest that the NSM sisters do not die in ces-1 and ces-2 mutants, and that two additional cells, located anterior to the 12 neurons, fail to die in ces-1 animals. Direct observation of the cell lineage of developing ces-1 embryos by J. Sulston (personal communication) has confirmed this hypothesis: in ces-1 embryos the NSM sisters and the 12 sisters fail to die. Although the cell lineages of ces-2 embryos have not been directly observed, the two extra cells in ces-2 mutants resemble the surviving NSM sisters in ces-1 animals in position, morphology, and the ability to produce serotonin; so we feel confident that they also are NSM sisters that fail to die.

Although mutations in ces-1 and ces-2 prevent some cells from undergoing cell death, these mutations do not affect most of the cells that die during C. elegans development. First, none of the other 18 cells that normally die during the development of the pharynx (Sulston et al. 1983) seems affected by these two mutations. We used Nomarski optics to determine the total number of cells in the pharyngés of three wildtype, three ces-1, three ces-2 (25 °C) and three ced-3 larvae. Although the wild-type animals studied by Sulston et al. (1983) had 80 nuclei in the pharynx, we observed a small amount of natural variability among the three wild-type animals that we scored (80, 81, and 83 cells, for an average of 81). Some of this variability appears to be caused by variable survival of the pharyngeal MCL sister (see Materials and methods). In addition, in the ces-2 pharyngés, one or both NSM sisters survived (81, 81, and 82 pharyngeal cells, for an average of 81) and, in all three ces-1 pharyngés, both 12 sisters and both NSM sisters survived (85, 86, and 84 cells, for an average of 85). By contrast, when all cell deaths are prevented by a mutation in ced-3, the pharynx has 20-21 extra cells (101 cells observed in all three animals). Based on the cell lineage, if all cells that form the pharynx lived, there would be 22 extra cells present, for a total of 102 cells (Sulston et al. 1983).

A second observation also indicates that ces-1 and ces-2 mutations do not affect all dying cells. In ces-1 ced-1; ced-2 and ced-1 ces-2; ced-2 triple mutants only the corpses of the NSM sisters (and of the 12 sisters in ces-1 animals) are missing; all other cell corpses appear to be present. In particular, we have observed that the nine cell deaths in the ventral nervous system (Sulston and Horvitz, 1977) occur in ces-1 and ces-2 mutants, and that many cell corpses are found in the heads of ces-1 ced-1; ced-2 mutants and ced-1 ces-2; ced-2 mutants, just as they are in ced-1; ced-2 animals. Thus, mutations in ces-1 and ces-2 prevent specific cell deaths in the pharynx, but do not affect any other dying cells we have examined.

The mutation n703 results in a gain of ces-1 function To learn how the ces-1 mutation n703 prevents the deaths of specific cells, we studied the effects of different doses of the ces-1 gene on the survival of the 12 sisters, which are much more sensitive to changes in ces-1 dosage than are the NSM sisters. We used Nomarski microscopy to directly count surviving cells, and three different deficiencies -nDf23, nDf24, and nDf25 (Fig. 1) - to decrease the level of ces-1 activity.

Two experiments show that none of these deletions of the ces-1 gene behaves like the dominant ces-1(n703) mutation, which suggests that n703 does not cause a loss of ces-1 function. First, in ces-1(n703)/ + animals, the 12 sisters survive 48% of the time (n=688 animals), and the NSM sisters survive 91 % of the time (n=226 animals). By contrast, these four cells always die in nDf23/+ animals (n=20) and nDf24/+ animals (n=20), just as they do in the wild type (n=100). Second, we performed gene dosage experiments using different alleles of ces-1 in trans to the mutation n703, as shown in Fig. 4. The frequency that the 12 sisters survive in these strains decreases in the order: ces-l(n703)/ces-1 (n703)> ces-1 (n703)/+> ces-1 (n703) / Df.

The n703 mutation is stronger than the wild-type allele of ces-1, whereas deletions of the ces-1 gene are weaker than the wild-type allele. These observations show that n703 has a gain of ces-1 function.

Experiments using the duplication nDp4, which contains the ces-1 gene, indicate that n703 does not act by causing overexpression of ces-1. Although the 12 sisters survive 92% of the time in ces-1 (n703)/ces-1(n703) animals (n=212 cells), they only survive 60% of the time in ces-1 (n703)/ces-1 (n703)/ + worms (n=179 cells) (see Materials and methods, Fig. 4 and Table 1). Similarly, the 12 sisters survive 48% of the time in ces-1 (n703)/ + heterozygotes (n=1376 cells), but only 22% of the time in ces-1 (n703)/+/ + animals (n=74 cells). Thus in some circumstances extra wildtype copies of ces-1 antagonize the n703 mutation. This result indicates that n703 does not simply cause higher levels of ces-1 gene expression or ectopic ces-1 expression, and suggests that n703 results in a novel ces-1 function.

Because n703 results in a gain of ces-1 gene function, we sought mutations that cause a loss of ces-1 function by isolating suppressors of n703. A second mutation within the ces-1 gene could suppress the dominant effects of n703 by eliminating ces-1 function. To find such new ces-1 alleles, we mutagenized ces-1(n703) males marked with a closely linked lin-10 mutation, and mated them with non-ces-1 animals marked with a closely linked unc-13 mutation (Fig. 5). Because +/nDf23, +/nDf24, and +/nDf25 animals live and appear wild-type (see above), we know that this screen can recover mutations that completely eliminate ces-1 function, just as these deficiencies do. We screened about 9600 Ft progeny, using Nomarski optics to determine if the NSM and 12 sisters were alive or dead in each animal. From this screen, we isolated two mutations that are cri-dominant suppressors of ces-1(n703). These mutations are called n1406 and n1434. Complementation tests show that these mutations are allelic (see Materials and methods).

We believe that these suppressors are located within the ces-1 gene for two reasons. First, both mutations are cis-dominant suppressors of ces-1 (n703). Second, both revertant mutations are tightly linked to ces-1 (n703) - each of these mutations is located within 0.1 map units of ces-1(n703), and we have not recovered the n703 allele from either revertant chromosome (see Materials and methods). These mutations suppress ces-2(n732) as well (see below), and genetic mapping shows that this suppressor activity is also located in the interval between the genes daf-8 and unc-29, where ces-1 is located. Measurements of the ces-1 activities of the revertant chromosomes show that both behave as if they have reduced levels of ces-1 function (Fig. 4). At 15° and 20 °C the revertant chromosomes appear to have more ces-1 activity than deficiencies have, so nl406 and nl434 might not entirely eliminate ces-1 gene function at these temperatures.

In ces-1 (n703 n1406) and ces-1 (n703 n1434) mutants, the NSM and 12 sisters die, as they do in wild-type animals. There are no apparent differences from the wild type elsewhere in the pharynx, at either 20° or 25°C. Furthermore, in these mutants serotonin is produced by the NSM neurons at apparently normal levels (data not shown). In addition, ces-1 (n703 n1406)[ nDf24 animals are also phenotypically wild-type, both in pharyngeal anatomy and in the appearance and behavior of these animals as viewed with a dissecting microscope (see Materials and methods). Thus we do not know what function, if any, the ces-1 gene plays in the normal development of the animal.

We believe that n732 acts by lowering but not eliminating ces-2 gene function at higher temperatures. In two different experiments, n732 has effects similar to but weaker than those of a deletion of the ces-2 gene. First, both ces-2(n732) and deletions of the ces-2 gene show semi-dominance for the survival of the NSM sisters. These cells always die in the wild type (n=800 cells), but at 25°C, 2.3% of them survive in ces-2(n732)/+ animals (n=400 cells) and 27% survive in Df/ + animals (n=96 cells, see Materials and methods). Second, in both ces-2(n732)/ces-2(n732) animals and ces-2(n732)/eDf6 animals at 25 °C the NSM sisters usually survive (Fig. 6). However, the NSM sisters survive more often in ces-2(n732)/eDf6 animals than in ces-2(n732) homozygotes, which suggests that n732 does not completely eliminate ces-2 function, even at high temperatures.

To isolate more mutations that affect the decision of the NSM sisters to live or die, we developed a general screen. Looking for animals with extra serotonergic cells, as was done to isolate ces-1(n703) and ces-2(n732), is time-consuming. Such a screen requires the examination of fixed and stained animals, and mutants must therefore be recovered from among previously cloned siblings. Instead, we used Nomarski optics to examine living worms, seeking mutants in which the NSM sisters survived. We’screened 21000 F₂ worms, which represents about 9700 haploid genomes scored for recessive mutations (see Materials and methods).

From this mutagenesis, we recovered three new ces mutations -two ces-1 dominant alleles, nl895 and nl896, and one recessive, temperature-sensitive mutation, n1952. We also isolated six mutations that prevent the deaths of not only the NSM sisters, but of all other dying cells as well. These six mutations will be described elsewhere.

The penetrance of NSM sister survival is low for the recessive mutation ces(nl952)\ 25 % of the animals had at least one surviving NSM sister at 25 °C, and only 2 % at 20°C (n=100 animals in each case). Because of its low penetrance, we have not yet fully characterized this mutation, but preliminary results suggest that nl952 might define a new gene specifically involved in the deaths of the NSM sisters. First, n1952 complements ces-2(n732) and is not linked to ces-1 or ces-2 (see Materials and methods). Second, the only abnormality that we have observed in these animals is the presence of two extra pharyngeal cells. These extra cells are serotonergic neurons, and based upon their positions appear to be the surviving sisters of the NSM neurons (data not shown).

We believe the two new dominant mutations are alleles of ces-1 for three reasons. First, the ces-1 dominant mutation n703 and the new mutations n1895 arid n1896 result in the same phenotype: the NSM and 12 sisters fail to die, but other cell deaths appear unaffected (data not shown). Second, all three mutations map between the genes daf-8 and sup-17 (see Materials and methods), which places them all within an interval of about 0.1 map units. Third, all trans heterozygotes involving these mutations appear identical: between 76% and 81% of the 12 sisters survive (n>=200 cells in all cases, see Materials and methods); furthermore, all three mutations behave similarly in studies of ces-1 gene dosage (Fig. 7).

The NSM sisters die in the two ces-1 loss-of-function mutants, whereas they live in ces-2, ced-3 and ced-4 mutants. This difference in phenotype allowed us to study the interactions between ces-1 and these other genes that affect the deaths of the NSM sisters. We constructed double mutants between ces-1(n703 n1406) and mutations in each of the other genes, and determined if the NSM sisters lived or died.

In animals carrying ces-l(n703 n1406) and a mutation in either ced-3 or ced-4, the NSM sisters live, just as they do in ced-3 and ced-4 mutants. These results indicate that ces-1 function is not required for NSM sister survival in ced-3 or ced-4 animals. Since loss-of-function mutations in ces-1 and these ced genes result in opposite effects on the NSM sisters (death vs. life, respectively), it seems likely that ces-1 and the ced genes do not control sequential steps in a pathway (e.g. of biosynthesis) but rather that one negatively regulates the other. If so, our results imply that ces-1 acts before ced-3 and ced-4 to decide whether the NSM sisters should five or die. By contrast, the NSM sisters die in ces-1(n703 n1406) ces-2 and ces-1(n703 n1406); n1952 animals, just as they do in ces-1(n703 n1406) mutants. These results show that ces-1 function is probably required for ces-2(n732) or ces(n1952) to prevent the deaths of the NSM sisters, suggesting that ces-1 acts after ces-2 and ces(n1952). However, since neither n732 nor n1952 results in a complete loss-of-function, this conclusion must be regarded as tentative.

We repeated several of these experiments using the second ces-1 loss-of-function allele. In each case, ces-1(n703 nl434) behaved like ces-1 (n703 n1406). Specifically, ces-1 (n703 n1434) was suppressed by ced-3 or ced-4 mutations, and suppresses the ces-2 mutation. We also constructed the mutant ces-1 (n703 n1406) ces-2; ced-3. In this animal, the NSM sisters survive, which shows that these cells die in ces-1 (If) ces-2 animals by the normal process of programmed cell death, which depends on ced-3 function.

We examined a group of 200 ces-1(n703)/ + animals at 20°C, in which 93 % of the NSM sisters survived. If mutations in ces-1 act independently on each NSM sister, then (93 %)²=86.5 % of the worms should have two surviving NSM sisters, 2(93 %)(7 %)=13 % of the worms should have only a single NSM sister surviving, and (7%)²=0.5% of the worms should have no NSM sisters surviving. Among these 200 animals, we saw 87.5% with two NSM sisters, 11% with one NSM sister, and 1.5% with no NSM sisters. This result suggests that ces-1(n703) might act independently on these two cells. Similar data indicate the independence of the 12 sister deaths in ces-1 (n703)/ + animals and of the NSM sister deaths in ces-2 animals (àll of these experiments were done at 15°, 20° and 25°C; data not shown). In each case, the data fit the hypothesis that these two genes act independently on each of the affected cells.

We have identified two genes, ces-1 and ces-2, that affect the decisions of specific cells to live or die. Gain-of-function mutations in ces-1 and loss-of-function mutations in ces-2 (both n732 and deficiencies) prevent the sisters of the NSM neurons from dying. These ces-1 mutations also prevent the sisters of the 12 neurons from dying. However, other cell deaths in these mutants occur normally, and there are no other obvious abnormalities in phenotype. In particular, the other cells that form the pharynx, many of which are closely related to the NSM sisters and the 12 sisters, all appear normal. Thus mutations in the ces-1 and ces-2 genes can affect the fates of the NSM and 12 sisters in the pharynx without changing other aspects of development.

Our observations indicate that a normal function of both the ces-1 and ces-2 genes is to control the deaths of the NSM sisters. First, since a loss of ces-2 function prevents the NSM sisters from dying, ces-2 normally causes these cells to die. Second, since a loss of ces-1 function results in the NSM sisters dying in a ces-2 mutant animal, ces-1 can cause the NSM sisters to live.

Only one other C. elegans gene, egl-1, is known to affect specifically the decisions of particular cells to undergo programmed cell death (Trent et al. 1983; Ellis and Horvitz, 1986). In hermaphrodites, the two HSN neurons control egg-laying, whereas in males these cells undergo programmed cell death. Mutations in egl-1 cause the HSN neurons to die in hermaphrodites as well as in males, possibly by transforming the sexual identity of these cells. It is conceivable that egl-1 acts in the process of sex determination rather than in the direct specification of cell death. This reservation does not apply to the functions of ces-1 and ces-2, which act similarly in both sexes.

Two observations suggest that ces-1 and ces-2 control the decision of the NSM sisters to live or die by controlling genes that act in all cell deaths. First, the NSM sisters die in ces-1 (If) animals, but in ces-1 (If); ced-3 or ces-1 (If); ced-4 animals these cells live. As we discuss above, this result indicates that ces-1 acts prior to ced-3 and ced-4 and is consistent with models in which ces-1 decides if the NSM sisters should live or die and ced-3 and ced-4 are then required to kill these cells. Second, the NSM sisters die in ces-1 (If) ces-2 animals, suggesting that ces-2 acts with or through ces-1 to regulate the genes involved in all programmed cell deaths. The fact that in ces-1 (If) ces-2; ced-3 animals the NSM sisters live demonstrates that in ces-1(lf) ces-2 mutants these cells die by normal programmed cell death, which requires ced-3 function.

Mutations in the ces genes could affect the activities of ced-3 and ced-4 directly, by controlling the initiation of cell death, or indirectly, by transforming the NSM sisters into cells that normally live, presumably the NSM neurons themselves. It seems unlikely that ces-1 normally acts to determine NSM identity: although ces-1 gain-of-function mutations cause the NSM sisters to survive and develop like NSM neurons, ces-1 loss-of-function mutations eliminate only the surviving NSM sisters and not the NSM neurons themselves. In fact, in mutants with a loss of ces-1 function, the NSM neurons appear to survive, differentiate and produce serotonin normally. Mutations in ces-2 also have no known effects on the NSM neurons, and genetic results (see above) indicate that ces-2 acts via ces-1. Thus, these genes probably control the initiation of programmed cell death rather that the acquisition of the NSM cell fate. When the deaths of the NSM sisters are blocked, as in ced-3 or ced-4 animals, these cells appear identical to the surviving NSM sisters found in the ces mutants, consistent with the hypothesis that the ces genes are involved in directly controlling the deaths of these two cells.

What are the phenotypes of animals with a complete loss of function for ces-1 or ces-2? It is possible that a complete loss of ces-1 function results in a wild-type phenotype, because the two ces-1 alleles that we isolated in a screen for mutations that reduce or eliminate ces-1 activity both result in a wild-type phenotype. Both alleles resemble deficiencies in several, although not all, of our gene dosage tests. By contrast, only one allele of ces-2 exists. The n732 mutation is extremely temperature-sensitive, and gene dosage experiments suggest that this mutation reduces only partially the activity of the ces-2 gene. It is possible that a complete loss of ces-2 function would affect other aspects of development besides the deaths of the NSM sisters. Screens for new mutations in this gene should help determine the phenotype that results from a complete loss of ces-2 function.

If a loss of ces-1 function results in a wild-type phenotype, there must be other genes that cause the NSM sisters to decide to die. One possibility is that there are two regulatory pathways, each of which controls the deaths of the NSM sisters (Fig. 8A). One process acts to prevent cell death and includes ces-2 and ces-1; cell death is prevented effectively only in mutants with a gain of ces-1 function or a loss of ces-2 function, whereas in animals with wild-type or inactive ces-1 genes the second pathway successfully initiates cell death. Alternatively, ces-2 and ces-1 might both act upstream of an unknown regulatory gene that directly regulates ced-3 and ced-4 to initiate the deaths of the NSM sisters (Fig. 8B).

In some organisms, hormonal signals play an important role in deciding if or when a cell should die. For example, in the moth Manduca sexta, a decrease in the level of ecdysteroids initiates many programmed cell deaths (Truman and Schwartz, 1982). Cells that would normally die continue to live if provided with ecdysteroids, and when ecdysteroid levels decline these cells die; since other cells do not die, factors other than hormone levels must specify which cells are capable of dying. Are the genes ces-1 and ces-2 involved in a similar system? Our data indicate that the decision to live or die of each of the cells affected by ces-1 and ces-2 mutations appears to be independent of that of the other cells. These observations suggest that ces-1 and

ces-2 do not act to control the level of a systemic hormonal factor. Although it remains possible that these genes act in response to such a factor or in a system involving signalling between adjacent cells, several lines of evidence suggest that in general in C. elegans the decision to die occurs within dying cells or their parents (reviewed by Yuan and Horvitz, 1990). We therefore suspect that ces-1 and ces-2 encode factors that act in a pathway operating entirely within the NSM sisters to control the deaths of these cells.

We thank Nancy Tsung and Carol Trent for providing the mutations n703 and n732, John Sulston for his generous help in studying the embryonic lineage of ces-1 (n703) animals, Phil Anderson and Jonathan Hodgkin for providing strains, and Hilary Ellis and Leon Avery for sharing unpublished observations. We are also grateful to Erik Jorgensen, Patricia Kuwabara and Eric Lambie for suggestions concerning this manuscript. This work was supported by US Public Health Research Grants GM24663 and GM24943. R. E. E. was supported by a National Science Foundation Graduate Fellowship, and by a National Institutes of Health Training Grant. H. R. H. is an Investigator of the Howard Hughes Medical Institute.