Cell. Vol 44, 817-829, March 28, 1986, Copyright

0 1986 by Cell Press

Genetic Control of Programmed Cell Death in the Nematode C. elegans
Hilary M. Ellis,’ and H. Robert Horvitz
Department of Biology Massachusetts Institute of Technology Cambridge, Massachusetts 02139 The phenomenon of programmed cell death raises a number of questions. Why are cells generated only to die? By what mechanisms do they die? How is it determined during development which cells die? C. elegans is well suited for studies that attempt to answer these questions. This nematode has fewer than 1000 somatic cells, and fixed patterns of cell divisions, migrations, and deaths generate individuals of invariant anatomy (Sulston and Horvitz, 1977; Kimble and Hirsh, 1979; Sulston et al., 1983). Thus, specific developmental events can be examined reproducibly and at the resolution of single cells. In addition, the short generation time (3 days at 20%) and large brood size of C. elegans facilitate genetic manipulations (Brenner, 1974; Herman and Horvitz, 1980). We describe here the isolation and characterization of mutations that prevent the initiation of programmed cell death in C. elegans, causing cells that would normally die to survive instead. These mutations define two genes, ted-3 and ted-4, that may be involved in determining which cells express the fate of programmed cell death.

Summary The wild-type functions of the genes ted-3 and ted-4 are required for the initiation of programmed cell deaths in the nematode Caenorhabditis elegans. The reduction or loss of ted-3 or ted-4 function results in a transformation in the fates of cells that normally die; in ted-3 or ted-4 mutants, such cells instead survive and differentiate, adopting fates that in the wild type are associated with other cells. ted-3 and ted-4 mutants appear grossly normal in morphology and behavior, indicating that programmed cell death is not an essential aspect of nematode development. The genes ted-3 and ted-4 define the first known step of a developmental pathway for programmed cell death, suggesting that these genes may be involved in determining which cells die during C. elegans development. Introduction
Cell death occurs during the normal development both of vertebrates (Saunders, 1966; Cowan et al., 1984) and invertebrates (Truman, 1984). During the development of the nematode Caenorhabditis elegans, the generation of the 959 somatic nuclei of the hermaphrodite is accompanied by the generation and subsequent deaths of 131 cells (Sulston and Horvitz, 1977; Sulston et al., 1983). As in other organisms, naturally occurring or “programmed” cell death in C. elegans is particularly common during the development of the nervous system; approximately 20% of all presumptive neural cells die (Sulston and Horvitz, 1977; Horvitz et al., 1982a; Sulston et al., 1983). Morphological and genetic studies have indicated that essentially all programmed cell deaths in C. elegans probably involve the same mechanism. First, nearly all dying cells undergo the same sequence of morphological changes (Sulston and Horvitz, 1977; Sulston et al., 1983). Second, a number of mutations affecting the process of cell death have been isolated. These mutations, which have defined the genes ted-7 (for cell death abnormal), ted-2, and nut-7 (nuclease deficient) (Sulston, 1976; Hedgecock et al., 1983), affect all programmed cell deaths. Thus, all cell deaths appear to involve the same genetic (and, hence, molecular) processes. For these reasons we can regard programmed cell death in C. elegans as a specific cell fate and may be able to use cell death as a model system for examining mechanisms that specify cell fates. f Present address: Department of Biology, Urwersity San Diego, La Jolla, California 92093. of Califorma,

Results Isolation of Cell Death Mutants by Direct Screening

Programmed cell deaths can easily be identified in living nematodes using Nomarski differential interference contrast microscopy (Sulston and Horvitz, 1977). The first sign of the impending death of a cell is a slight increase in its refractility. The nucleus of the dying cell becomes increasingly refractile until it resembles a flat button; this stage persists for lo-30 min. Subsequently, the nucleus of the dying cell decreases in refractility, begins to appear crumpled, and then gradually disappears. This process is completed in less than 1 hr (Sulston and Horvitz, 1977; Sulston et al., 1983). In ted-7 and ted-2 mutants (Hedgecock et al., 1983), dying cells remain in the highly refractile stage for many hours, often persisting in this state through several larval stages. Thus in ted-7 and ted-2 animals, deaths that occur over a wide range of developmental times can be observed at once. To isolate mutations that perturbed the normal pattern of programmed cell deaths, we treated ted-7 hermaphrodites with the mutagen ethyl methanesulfonate (EMS) (Brenner, 1974) and used Nomarski microscopy to screen their F2 progeny. The progeny of approximately 4000 Fl hermaphrodites were examined. Among the mutants we obtained in this way were two strains in which the characteristic cell deaths of ted-7 were not seen (Figure 1). From these strains we isolated two recessive mutations, n777 and n778, that define the gene ted-3 on linkage group IV (Figure 2). (See Experimental Procedures for methods of isolation, mapping, and complementation testing.)

ted-3 Mutations

Eliminate

Programmed

Cell Death

Embryonic and postembryonic cell deaths normally seen in ted-7 animals are seen only rarely in ted-7; ted-3 animals (Table 1). The direct observation in ted-7; ted-3

Cell
818

Iu

.lmy

I

nDfZ7

i Ill and IV

Figure 2. A Partial Genetic

Map of Linkage Groups

Figure 1. Absence

of Cell Deaths in ted-3 Animals ted-7 larva. Arrows of a newly hatched that shown in (a). be seen in both (a) larva. Bar = 10 u.

(a) Nomarski photomicrograph of a newly hatched indicate dying cells. (b) Nomarski photomicrograph ted-7; ted-3 larva. Plane of focus is approximately Arrowheads indicate several of the nuclei that can and (b). No cell deaths are seen in the ted-7; ted-3

or ted-3 animals of the postembryonic lineages of blast cells that generate the ventral nerve cord (Pl-P12), the postdeirid sensilla (V5L.pa and ER.pa), and neurons located in the lateral hypodermis (QL and QR) indicated that programmed cell deaths do not occur in these lineages. In each case the lineages observed were identical with those of wild-type animals except that cells that normally die in wild-type animals survived in ted-7; ted-3 or ted-3 animals (Figure 3). After determining the postdeirid lineage in several ced3 animals, we observed the postdeirid cells through two further larval stages until the animals molted into adults. In no case did we observe the division of a cell death survivor, nor were any delayed deaths observed. Cell death survivors from the Q neuroblast and ventral cord lineages also do not appear to divide; in ted-3 animals extra postembryonically generated neuronal cells are found in the ventral cord, postdeirids, and lateral hypodermis (Figure 4) and the number of extra cells is consistent with the number of cells that would normally die in the lineages of the Q, VS.pa, and P blast cells. In addition, in our studies of ted-7; ted-3 or ted-3 animals we have never seen extra divisions in these lineages. The observation that the lineages of ted-3 animals are normal, except in the failure of cells to die, suggests that ted-3 mutations prevent the

expression of programmed cell deaths. Presumably the ted-3 mutation acts prior to ted-7 to block the initiation of programmed cell death. Mutations in ted-3 block the programmed cell deaths that occur during embryonic development as well as those that occur during postembryonic development. We have screened ted-7; ted-3 hermaphrodites of various stages, both embryonic and postembryonic, for the presence of dying cells. Although occasional dying ceils were seen, no particular cell died in all or even many hermaphrodites (Table 1). In addition, in newly hatched ted-3 larvae a number of extra embryonically generated cells can be observed. For example, ted-3 mutants have several extra cells in the anterior bulb of the pharynx (Figure 5; also see below). Again, the number of these cells is consistent with the survival of cells that would normally die in lineages that contribute to this structure. ted-3 mutations appear to block all programmed cell deaths in hermaphrodites. Programmed cell deaths are also blocked in ted-3 males. However, we have noted one exception, the male-specific linker cell. Unlike most programmed cell deaths in C. elegans (Horvitz et al., 1982a), the death of the linker cell requires the presence of another ceil (U.lp or U.rp) (Sulston and White, 1980) and thus could be termed a “murder” (Horvitz et al., 1982a). In three out of five ted-3 males in which the fate of the linker cell was followed, the linker cell had died by the L4 molt. (In 7 out of 7 wild-type males observed, the linker cell had died by the L4 molt.) Thus, although ted-3 may play some role in the death of the linker cell, normal ted-3 function is not essential for linker cell death. We have not examined ted-3 animals for the deaths of the male-specific cells B.a(l/r)apaav and B-gamma.a(l/r)d, both of which appear to be “murdered” (Sulston and White, 1980).
Cell Death “Survivors” Do Not Divide

As described above, postembryonic cell death “survivors” do not divide. In addition, we have directly followed the fate in ted-3 animals of a cell that dies during wild-type embryogenesis, Ab.alapapaa. This cell is the grand-

C. elegans Cell Death Mutants 819

Table 1. Elimination

of Cell Death by ted-3 and ted-4 Mutations Average Embryonic Number of Deaths Observed Deaths Postembryonic Ventral Cord’ 8.7 n = 28 0.04 n = 50 0.03 n = 35 0 n = 23 N.D 0.04 n = 27 10 Deaths Postdeiridt 0.93 n = 29 0 n = 15 0 n = 24 0.05 n = 20 0.13 n = 30 0 n = 21 1 c!* 1.6 n = 28 0 n = 24 0 n = 21 0.06 n = 17 N.D. 0 n = 21 2

Genotype ted-1 ted-1; ted-1; ted-1; ted-3 (n717j ted-3 (n718) ted-3 (n1040)

Head of Ll 28.0 n = 23 0.3 n = 21 0.5 n = 26 7.0§ n = 21 3.0 n = 22 0.6 n = 23 #

ted-1; ted-3 (n1129) ted- 1; ted-4 (n 1162) Total Number of Deaths That Occur rn Wild Type11

Ammals of the indicated genotypes were observed using Nomarski differential interference contrast microscopy. The numbers of dead cells present in the head of first larval stage (Ll) animals, the ventral cord, and right anteriolateral hypodermis (where progeny of the QR blast cell are generated) of L2 animals, and the right postdeirid of L3 animals were determined. * In wild-type animals, 10 programmed cell deaths are generated in the ventral cord during the Ll stage by the blast cells W and Pl.a-P12.a. T The lineages of VSR.pa and V5L.pa, which generate the postdeirid sensilla, each include one programmed cell death. Only the right postdeirid was observed, smce on the left side of the animal, QL generates cell deaths in the region of the postdeirid. * On the left side of the animal, QL generates two programmed cell deaths posterior to the gonad, near the region of the postdleirid. On the right side of the animal, QR generates two programmed cell deaths in the lateral hypodermis anterior to the gonad. Only deaths in the right anterior lateral hypodermis were observed in order to avoid confusion with the death generated by the left postdeirid lineage. 5 ted-3 (n7040) appears to affect these embryonically generated cell deaths less severely than postembryonically generated deaths, suggesting that the embryonic deaths require lower amounts of ted-3 activity. 11 The values in this row indicate the total number of deaths that occur in particular regions of the wild-type anrmal prior to the trme at which mutant ammals were examined. As can be seen by comparing this row with the first row of this table, not all deaths last long enough to have been observed rn ted-1 animals. # We have not determined which of the deaths that occur during wild-type embryonic development can be visualized within the head region we have examined in ted-1 mutant animals, N.D. indicates not determined.

daughter of Ab.alapap, which generates a lineage identical with that generated by another cell, AB.alappp, except that the cell homologous to AB.alapapaa (i.e., AB.alapppaa) divides to generate two neurons. Thus it seemed plausible that in ted-3 animals the cell death “survivor” AB.alapapaa might divide. We followed the lineage that generates this cell in four ted-3 animals and observed the cell for 1 to 2 hr after its birth. AB.alapapaa neither divided nor died in any of these animals. In one ted-3 embryo in which AB.alapapaa was followed for a longer period of time (3.5 hr), this cell eventually became detached from the embryo. (One or two such detached cells are often observed in ted-3 embryos floating between the embryo and the eggshell.) Thus, those cell death survivors that we have directly followed do not divide. In addition, there appears to be no gross proliferation of cell number in ted-3 mutants. These observations suggest that cell death in C. elegans does not function to terminate cell lineages by the elimination of stem or blast cells that would otherwise continue to divide.

Cell Death Survivors Differentiate At least some cell death survivors assume a differentiated fate that we can identify. For example, the histochemical technique of formaldehyde-induced fluorescence (FIF) can be used to identify dopamine- or serotonin-containing neurons (Sulston et al., 1985; Horvitz et al., 1982b). Animals carrying mutations in ted-3 often have an extra dopaminergic neuron in each of their postdeirids (Figure 6). In wild-type animals, the postdeirid contains one dopaminergic neuron, V5.paaa. The sister of V5.paaa, VS.paap, divides to generate a programmed cell death, Vhpaapp, and a nondopaminergic neuron, V5.paapa, which has a characteristic morphology, as viewed with Nomarski microscopy, that is distinguishable from the other cells of the postdeirid (Figure 7). In ted-3 animals the postdeirid lineage is identical with that of the wild-type except that the final division generates two neuronal cells. Since the anterior daughter of this division appears identical by Nomarski criteria with the nondopaminergic neuron of the wild-type postdeirid (Figure 7). and since the posterior daughter seems identical with the normal dop-

Cell 820

WILD TYPE Pn I

V5ipa

ted-3

Figure 3. Lineages Generated Wild-Type and ted-3 Animals

by the Q, P, and V5pa

Blast Cells in

Each vertical line represents a cell, and each branch represents a cell division. Cells are named according to their ancestry, for example, V5.pa is the anterior daughter of the posterior daughter of V5. X indicates a programmed cell death. During the first larval stage (Ll) of wiidtype animals QL and QR (the Q blast cells present on the left and right sides of the animal, respectively), each divide to generate four neural cells and a programmed cell death. In wild-type hermaphrodites, each of the 12 P blast cells (collectively Pn) divides during the Ll to generate a neuroblast, Pn.a and a hypodermal cell, Pnp. All 12 Pn.a cells undergo the same pattern of divisions, generating five descendants. Pl.aap, PP.aap, and PS.aap-Pl2.aap (but not PS.aap-PB.aap) and Pll.aaap and P12.aaap (but not Pl.aaap-PlO.aaap) undergo programmed cell death. P12.p divides once during the Ll generating a posterior daughter that dies. During the L2 V5L.pa and V5R.pa each divide to generate four neural cells and a programmed cell death. In ted-3 animals the Q, P, and Vdpa lineages are identical with those of wild-type animals, except that cells that normally die survive instead.

neuron, it seems very likely that the extra dopaminergic neuron in ted-3 postdeirids is &paapp, the cell that dies in the wild-type postdeirid lineage (Figure 8). While V5paapp nearly always survives and appears neuronal in ted-3(n717) and ced3(n718) animals, only 42 of 79 postdeirids in ted-3(n777) animals and 26 of 56 postdeirids in ted-3(n778) animals had an extra dopaminecontaining neuron. In other words, in ted-3 animals the choice between death and survival is nearly invariant, but the production of dopamine in the surviving cell is variable. We do not know whether the cell death survivors that fail to make dopamine represent a cell type that differs from the dopaminergic neuron Kpaaa, or whether these survivors simply fail to express one aspect of the differentiated phenotype of the dopaminergic neuron. ted-3 animals also often have extra dopaminergic neurons in their heads with positions and morphologies similar to the dopaminergic ventral lateral cephalic neurons, CEPVR and CEPVL. Since in wild-type animals CEPVR and CEPVL are generated as “sisters” of deaths (Sulston et al., 1983), it may be that in ted-3 animals dopaminergic aminergic

neurons rather than deaths are generated as sisters of CEPVR and CEPVL. The production of dopamine by these cell death survivors also is variable; for example, of 16 ted-3(n777) animals observed, six had an extra dopaminergic neuron in the region of the ventral cephalic neuron on one side and five had an extra dopaminergic neuron in this region on both sides. FIF histochemical staining of ted-3 hermaphrodites has also indicated the presence of two extra serotonergic neurons, one per side, in the anterior bulb of the pharynx (Figure 6). These supernumerary serotonergic neurons are morphologically similar to the normal pharyngeal serotonergic neurons, the NSMs (“neurosecretory motor neurons”) (Albertson and Thomson, 1976; Horvitz et al., 1982b). Since in wild-type animals the NSMs are generated as sisters of deaths (Sulston et al., 1983) it may be that in ted-3 animals serotonergic neurons rather than deaths are generated as sisters of the NSMs. The presence of serotonin in these putative supernumerary NSMs (at least after the uptake of exogenous serotonin; see Figure 6) is less variable than that of dopamine in the supernumerary dopaminergic neurons described above. Specifically, 29 of 34 NSMs in ted-3(n777) animals and 36 of 38 NSMs in ted-3(n778) animals were adjacent to an extra serotonergic cell. Sexually dimorphic cell deaths that occur during the embryonic development of C. elegans are also affected by mutations in ted-3. In wild-type males the cells that in hermaphrodites become the two hermaphrodite-specific neurons (HSNs) die; in wild-type hermaphrodites the four cells that in males become the cephalic companion neurons (CEMs) die (Sulston et al., 1983). ted-3 males have an extra cell on each side that by position and morphology (as viewed with Nomarski microscopy) appears to be an HSN (Figure 7). Similarly, ted-3 hermaphrodites have four extra cells that by position and morphology appear to be the CEMs (Figure 7). In addition, sensory endings with the positions and morphologies of the CEM sensory endings have been observed in serial section electron micrographs of aced-3 hermaphrodite (J. G. White, personal communication). Cell Death Survivors Function In the absence of programmed cell death, cells that would normally die instead survive and differentiate. Can these cell death survivors function? The following observations suggest that they can. The HSN neurons are required for normal egg-laying (Trent et al., 1983); for this reason eg/l(n487) hermaphrodites, which lack HSNs, are egg-laying defective (Trent et al., 1983). Since, as noted above, the HSN homologs normally die during the embryonic development of C. elegans males, it seemed plausible that the phenotype of egl-7 animals might be caused by the inappropriate death of the HSNs in egl-7 hermaphrodites. We constructed a ted-3; egl-7 strain and found that this strain had HSNs and laid eggs normally. The presence of HSNs in ted-3; egl-7 hermaphrodites indicated that the absence of HSNs in egl-7 hermaphrodites was caused by cell death, presumably of the HSNs. Furthermore, the egglaying competent phenotype of the ted-3; egl-7 strain sug-

C. elegans Cell Death Mutants 821

Figure 4. Extra Cells in ted-3 Animals Photomrcrographs of (a,c,e) wild-type and (b,d,f) ted-3 animals stained with the fluorescent DNA-binding dye DAPI (diamidinophenolindo le). Anterior IS to the right and ventral is down. Lines indicate nuclei of (a,b) ventral cord neurons, (c,d) neurons generated by QR. and (e.f) postd eirid neural cells. Bar = 10 tr. At the anterior end of the ventral cord of (a) wild-type animals, the neuronal nuclei form a characteristic pattern of two nuclei, followed by a space and then four nuclei. In (b) ted-3 animals, this first pair of nuclei is followed by a set of five nuclei. In (c)wild-type animals QR genorates t wo neurons typodermis. In the region of the anterior lateral hypodermis, while in (d) ted-3 animals, QR generates three neurons in the region of the anterior lateral f’ The (e) wild-type postdeirid sensillum consists of four cells; the (f) ted-3 postdeirid has five cells.

Figure 5. Nomarski and ted-3 Animals

Photomicrographs

of the Pharynx

o f Wild-Type

Anterior is to the left. (a) and (c)show the pharynx ot a wild-l lype animal in two planes of focus. (b) and (d) show the pharynx of act ?d-3 animal in planes of focus equivalent to those shown for weld-type. IIn (a) arrows indicate the nuclei of the NSM neuron (to the right) and ther I2 neuron. In (b) three neuronal nuclei are seen in the equrvalent pla ne of focus in a ted-3 animal. In (c) the arrow indicates the nucleus of the neuron MCL, and arrowheads Indicate two nuclei of the multinucls rate muscle cell ml. In (d) an extra neuronal nucleus is seen in the equiv ralent plane of focus in a ted-3 animal. Bar = 10 K.

Cell 822

Figure 6. Wild-Type and ted-3 Animals Stained Using the Technique of Formaldehyde-Induced

Fluorescence

(FIF)

(a) Wild-type animals have one dopaminergic neuron on each side embedded in the lateral hypodermis. while (b) ted-3 animals have two dopaminergic neurons per side. Similarly, (c) w&f-type animals have one serotonergic neuron on each side of the anterior bulb of the pharynx, while (d) ted-3 animals have two serotonergic neurons per side. The animals in (c)and(d) were incubated in a solution containing 5 mglml serotonin prior to staining by FIF. In both wild-type and cec%3 animals, the NSM cell bodies can be visualized by FIF only after they have been exposed to exogenous serotonin. The ted-3 animal in (d) is younger than the ofher animals shown. Bars = 10 P.

gests that cell death survivors can function. servations of egl-7 and ted-7; eg/-7 embryos indicated that the HSNs undergo programmed in egl-7 hermaphrodites (data not shown), thus these conclusions.

Direct obhave also cell death supporting

ted-3 Mutants Appear Behaviorally Normal Despite the absence of programmed cell deaths and the presence of many supernumerary cells, ted-3 mutants appear normal in morphology and behavior as viewed with a dissecting microscope. ted-3 animals are not uncoordinated in locomotion; they respond as does the wild type by moving forward when gently touched on the tail and backwards when touched on the head (Chalfie et al., 1985), lay eggs at a normal rate, and are stimulated to lay eggs in response to exogenous serotonin and imipramine, which stimulate egg-laying by wild-type hermaphrodites (Trent et al., 1983). We have also shown that animals carrying mutations in ted-3 move up a gradient of NaCl (Table 2) as does the wild type (Ward, 1973). In addition, male mating ability is not significantly decreased in ted-3 strains. Specifically, when six wild-type males were crossed with six dpy-77(e224) hermaphrodites, 1159 cross progeny were produced. Six ted-3(n777) males produced 934 cross progeny, and six ted-3(n778) males produced 1181 cross progeny in this assay.

The brood size of ted-3 hermaphrodites may be somewhat lower than that of wild-type hermaphrodites; the average brood size of eight wild-type hermaphrodites was 308, while the average brood size of eight ted-3(n777) hermaphrodites was 200. However, this reduction in brood size may reflect the presence of background mutations in the ted-3 strain. Consistent with this interpretation, the average brood size of eight ced-3(n777)/ced-3(n777) dpy4(e7766) hermaphrodites was 259. Additional ted-3 Alleles We have identified two additional alleles of ted-3. ced3(n7729) was obtained on the basis of its failure to complement ted-3(n777) for the suppression of the egg-laying defect caused by egl-l(n487) (see Experimental Procedures). ted-3(n7040) was isolated in our laboratory by Nancy Tsung (personal communication), who used Nomarski microscopy to screen 6000 F2 progeny of mutagenized hermaphrodites for animals with extra pharyngeal cells near the positions of the NSMs. The general phenotype of both of these mutants is the same as that of ted-3(n777) and ted-3(n778) mutants. However, n7040 and n7729 are of lower expressivity than n777 or n778. For example, while cell deaths are rarely seen in the heads of first larval stage ted-7; ted-3(n777) or ted-7; ted-3(n778) animals, in ted-7; ted-3@7729) and ted-7; ted-3(n7040)

C elegans Cell Death Mutants 823

Figure 7. Nomarski Photomicrographs of Postdeirid Cells, CEMs, and HSNs, in WIldType and ted-3 Animals In (a) and (b) anterior is to tine right. In (c), (d), (e), and (f) anterior IS to the left. (a) The right postdeirid of a wild-type animal. (b) The right postdeirid of a ted-3 animal. The postdeirid lineages of the animals shown in (a) and (b) were followed using Nomarski optics; lines indicate the nucleus of the nondopaminergic neuron of the postdeirid, and arrowheads indicate the nuclei of the dopaminergic neuron and two structural cells of the postdeirid sensillum. The arrow in (b) indicates lthe cell death survivor. (c) The arrow indicates the ventral CEM of a wild-type male. (d) The iarrow indicates the ventral CEM in a ted-3 hermaphrodite. (e) The HSN neuron, indicated by an arrow, in a weldtype hermaphrodite. (f) A cell with the morphology and position (ventral and about midway between anterior and posterior) of an HSN in a ted-3 male. In (e) the arrowhead indicates the vulva, a hermaphrodite-specific feature. In (f) the arrowhead indicates a coelomocyte. The position of this cell differs lbetween male and hermaphrodttes (Sulston and Horvitz, 1977). The appearance of a coelomocyte approximately midway along the length of the animal (next to the HSN) indicates that the animal in (f) IS a male Bar = 10 1

first larval stage animals an average of three and seven deaths, respectively, are seen (Table 1).

ted-3 Mutations Result in a Loss or Reduction of Gene Function
Several observations suggest that the phenotype of ted-3 mutants results from a loss or reduction of wild-type function. First, direct screening by Nomarski microscopy for ted-3 mutants has produced ted-3 alleles at a frequency of approximately 2.5 x 1O-4 per mutagenized haploid genome (two mutants among the progeny of approximately 4000 Fl animals), which is close to the frequency expected for the generation of null mutations in the average C. elegans gene (Brenner, 1974; Greenwald and Horvitz, 1980). Second, three observations indicate that a deficrency of the ted-3 locus, nDf27, and ted-3(n777) behave genetically in the same manner. First, nDf27ked-3(n777) animals are similar in phenotype to ted-3(n777)/ced3(n777) animals, as observed with Nomarski optics. Second, nDf277ced-3(n777); egl-7 animals are fully suppressed for the egl-7 egg-laying defect, just as are ted-3(n777)/ ted-3(n777); egl-7; animals. Third, nDf27, like ted-3(n777), is a semidominant suppressor of egl-7/ + heterozygotes (Table 3).

Amber mutations provide an additional criterion for the identification of the phenotype resulting from the loss of function of a gene. However, we have not yet identified an amber allele of ted-3. We have constructed double mutant strains between each ted-3 allele and the amber suppressor sup-5 (Wills et al., 1983). Ten or more hermaphrodites of each strain were scored using Nomarski microscopy for the presence of extra cells in the pharynx and postdeirids and for the presence of the normally mate-specific CEM neurons. None of the ted-3 alleles was suppressed.

ted-4 Stations Cell Deaths

Also Eliminate

Programmed

ted-4(n7762) was isolated in our laboratory by Chand Desai (personal communication) as a suppressor of the egg-laying defect of the HSN-deficient mutant egl-l(n7084). The phenotype of ted-4(n7762) animals (see below) appears identical with that of ted-3 mutants. However, ced4(n7762) complements ted-3 mutations and maps to linkage group III (Figure 2; see Experimental Procedures for map data). In ted-4 animals, as in ted-3 animals, cells that would normally die instead survive. The lineages of the VS.pa and Q blast cells in ted-4 animals were identical with the

Cell 924

WILD TYPE

ted- 3

ND
Figure 9. The Lineage of V5.pa in Wild-Type and ted-3 Animals In wild-type animals, V5.pa generates four cells of the postdeirid sensillum, a dopaminergic neuron (D), a nondopaminergic neuron (N), two structural cells with compact nuclei (c), and a programmed cell death (X). In ted-3 animals, this cell death fails to occur. As explained in the text, it seems very likely that the extra domaminergic neuron in ted-3 animals is the cell that normally dies.

Figure 9. Nomarski in ted-4 Animals

Photomicrographs

of CEMs and Postdeirid Cells

(a) Arrow indicates the ventral CEM in a ted-4 hermaphrodite. The postdeirid lineage of the animal shown in (b) was followed using Nomarski optics. (b) Arrow indicates the cell death survivor, line indicates the nondopaminergic neuron, and arrowheads indicate the dopaminergic neuron and the two structural cells of the postdeirid. Bar = lob.

Table 2. Chemotaxis

Assays Average Number of Animals Accumulating Without Attractant 0.08 f 0.08 0.50 f 0.23 0.08 f 0.08

Genotype Wild Type ted-3(n717) the-3(e7 124)

With Attractant 7.0 f 1.4 7.9 f 1.3 0.75 + 0.35

Chemotaxis assays were performed as described in Experimental Procedures. For each genotype, 24 adult hermaphrodites were tested in each of 12 separate experiments. the-3 is a chemotaxis-defective mutant (Lewis and Hodgkin, 1977). Attractant was placed in the center of the assay plate, and animals to be tested were placed around the edge of the plate. The values indicated are the average number of hermaphrodites that accumulated at the center of the plate after 15 min plus or minus the standard error of the mean @EM).

Table 3. Suppression Genotype

of egCl(n487)/+

Heterozygotes % Egl 71 30’ 33 84 n n n n = = = = 300 212 90 207

egl-l(n487)/+ ted-3(n717)/+; egl-l(n487)/+ nDf27/+; egf-l(n487)/+ ted-4(n1162)/+; egl-l(n487)/+ The number ed with late * The ted-3 suppressors

of animals of the indicated genotype that became bloatstage eggs (Egl-) was determined. alleles n718, ni040, and n7729 are also semidominant of egl-l heterozygotes.

death does not require wild-type ted-3 or ted-4 function. Extra embryonically generated cells are seen in ted-4 animals; their number is consistent with the survival of cells that would normally die. FIF-histochemical staining of ted-4 animals shows that they often have an extra dopaminergic neuron in each of their postdeirids and in the region of CEPVR and CEPVL. As in ted-3 mutants, the production of dopamine by the cell death survivor in the postdeirid is variable. Of 60 postdeirids in ted-4 animals, 27 had an extra dopaminergic neuron. FIF staining of ted-4 animals also reveals on each side of the anterior bulb of the pharynx the presence of an extra serotonergic neuron that is similar in position and morphology to the NSM. These observations indicate that the phenotype of ced4 mutants is the same as that of ted-3 mutants. There is, however, one difference between the behavior of ted-3 mutations and ted-4(n7762). While ted-3 mutations are semidominant suppressors of the egg-laying defect of egl-l(n487) heterozygotes, ted-4 mutations are recessive suppressors of this defect (Table 3). (ted-3(n777); egl7(n487)/+ animals and ted-4(n7762); egl-l(n487)/+ animals and ted-4(n7762); egl-l(n487) animals are egg-laying competent.) A Developmental Pathway for Programmed Cell Death Mutations in ted-7 and ted-2 prevent the engulfment of dying cells by their neighbors and cause the DNA in dying cells neither to condense nor to be degraded (Hedgecock et al., 1983). The phenotype of ted-7; ted-2 double mutants has been shown previously to be identical with that of either mutant strain alone (Hedgecock et al., 1983). Thus the order of action of these two genes cannot yet be defined. Mutations in nut-7 result in the absence of an endonuclease, slowing the degradation of the DNA of dying cells (Sulston, 1976; Albertson et al., 1978). We have con-

wild-type lineages except that cells that normally die instead survived. ted-4 animals have extra neuronal cells in the ventral cord, postdeirids (Figure 9) and lateral hypodermis. The number of extra cells is consistent with the number of cells that normally die in the P, Vdpa, and Q lineages. Like ted-3 mutations, ted-4(n7762) blocks essentially all embryonic and postembryonic deaths. In ced7; ted-4 hermaphrodites no single cell has been observed to die consistently (Table 1). However, the linker cell was observed to die in two of six ted-4 males and in one of six ted-4(n7762); ted-3(n777) males. Thus, the linker cell

C. elegans Cell Death Mutants 825

OF DEATH

DEGRADATION

Figure 10. A Proposed Genetic Pathway for Programmed In C. elegans

Cell Death

The wild-type functions of ted-3 and ted-4 are required for the initiation of programmed cell death. In the absence of ted-3 or ted-4 activity, cells that normally die instead survive and may differentiate. In the wild type, dying cells are usually engulfed by neighboring cells. The wild-type ted-l and ted-2 products are required for engulfment or for an earlier step in the death pathway that must precede engulfment. The activrties of ted-7 and ted-2 are not, however, required for cell death per se (Hedgecock et al., 1983). The wild-type nut-7 gene product IS required to degrade the DNA of dying cells. Engulfment appears to be a necessary prerequisite for degradation of the DNA of dying cells, suggesting that the not-I nuclease may be expressed by the engulfing cell (Hedgecock et al., 1983). This pathway appears to be expressed by essentially all cells that undergo programmed cell death during C. elegans development.

Furthermore, in the dominant mutant met-4(e7677) six mechanosensory neurons, the microtubule neurons, die (Chalfie and Sulston, 1981). Hedgecock et al. (1983) found that these cell deaths were not affected by mutations in ted-7 or ted-2. Similarly, we have observed that these deaths still occur in ted-3(n777); met-4(e7677) animals. In addition, the deaths of the ventral hypodermal (Pn.p) cells in the dominant mutant /in-24(n432) (Ferguson and Horvitz, 1985) are not suppressed in /in-24(n432) ted-3(n777) animals (H. Ellis, E. Ferguson, and P Sternberg, unpublished observations). The cell deaths caused by mec4(e7677) and /in-24(n432) differ in morphology from cell deaths that occur during the development of the wild type (e.g., Chalfie and Sulston, 1981). These observations suggest that the mutations met-4(e7611) and /in-24(n432) result in the cell-specific expression of a cytotoxic product. Discussion Recessive mutations in the C. elegans genes ted-3 or ted-4 cause cells that would normally die instead to survive and, in at least some cases, to differentiate. Genetic studies indicate that the ted-3 phenotype is likely to result from the absence of wild-type ted-3 function. The recessive nature of the phenotype of ted-4(n7762) animals suggests that this mutation also results in a reduction of wildtype function. Since these mutations result in a reduction or loss of gene activity, it is unlikely that the differentiated phenotypes of cell death survivors in ted-3 or ted-4 animals are produced de novo. Rather, mutations in these genes may be revealing underlying developmental potentials. These underlying potentials may reflect evolutionarily more primitive states of the lineage. For example, the presence of HSN neurons in ted-3 and ted-4 males suggests that the male homologs of the HSNs, which normally die, have an undertying potential to differentiate into HSNs. Thus, sexual dimorphism for the presence of these neurons appears to be generated by programmed cell death. Similarly, the presence of the normally male-specific CEM neurons in ted-3 and ted-4 hermaphrodites suggests that sexual dimorphism for the presence of these neurons is also generated by programmed cell death. Other cell death survivors may also have fates that correspond to those of homologous cells. For example, the cell P2.aap in ted-3 hermaphrodites has characteristics that suggest that P2.aap, which normally dies in wild-type animals, has an underlying potential that is similar to that of its spatial homologs PS.aap-PB.aap. Specifically, during the first larval stage of the hermaphrodite, 12 neuroblasts (Pl.a-P12.a) undergo identical patterns of cell divisions to generate five neural cells each (Sulston, 1976). The fates of these cells are strongly correlated with their lineage histories. For example, all 12 Pn.apa cells become AS motoneurons. However, only PB.aap-P8.aap become VC neurons; the other six Pnaap cells undergo programmed cell death. Reconstruction from serial section electron micrographs of the anatomy of the anterior region of the ventral cord of a ted-3 hermaphrodite has indicated that P2.aap in this animal had many of the properties of its nor-

strutted a ted-7; WC-7 strain and have found that ted-7 is epistatic to nut-7; in ted-7; nut-7 animals the DNA of dying ceils is neither condensed nor degraded. Similarly, ted-2 is epistatic to nut-7. Strains that carry mutations in ted-3 and in one of the three genes ted-7, ted-2, or nut-7, and strains that carry mutations in ted-4 and in one of the four genes ted-1, ted-2, ted-3, or nut-7, have also been constructed. Mutations in ted-3 and ted-4 are epistatic to mutations in ted-7, ted-2, and nut-7; no cell deaths are seen in ted-7; ted-3, ted-7; ted-4, ted-2 ted-3, ted-4; ced2, ted-3; nut-7, or ted-4; nut-7 animals. Since the phenotype of ted-4; ted-3 animals is the same as that of either ted-3 or ted-4 mutants, we cannot yet define the order of action of these genes. The epistasis of ted-7 over nut-7, and of ted-3 and ted-4 over all three other cell death genes, suggests the genetic pathway for programmed cell death shown in Figure 10. The phenotypes of these cell death mutants are consistent with the pathway suggested on the basis of genetic interactions. In ted-3 and ted-4 mutants no signs of cell death are seen in cells that would normally die; no increase in refractility is seen with Nomarski microscopy and no condensation of the DNA is seen when ted-3 animals are stained with diamidinophenolindole (DAPI). In contrast, cell death begins normally in ted-7 and ted-2 mutants, but the dying cells are not engulfed and remain in a highly refractile state for an abnormally long time. In addition, the nuclei of dying cells in ted-7 and ted-2 mutants are only slightly condensed, and the DNA of these cells is not degraded (Hedgecock et al., 1983). Finally, in nut-7 mutants dying cells become refractile, are engulfed, and their cytoplasm is removed, leaving the DNA highly condensed; this DNA is visible after staining with DAPI (Sulston, 1976). Not all cell deaths appear to involve this pathway for programmed cell death. For example, as discussed above, the linker cell death can occur in ted-3 and ted-4 mutants,

Cell 826

N x

N0

tl N is
Reitera-

Figure 11. ted-3 and ted-4 Mutants Suggest an Underlying tive Nature to the Postdeirid Lineage

In wild-type animals, V5paa divides to generate a dopaminergic neuron (D) (M.paaa) and a daughter (Vdpaap) that, in turn, divides to generate a cell (V5.paapp) that dies (X) and a cell (V5paapa) that becomes a nondopaminergic neuron (N). In ted-3 and ted-4 animals, V5.paa divides to generate a dopaminergic neuron and a daughter, which divides to generate a dopaminergic neuron rather than a cell death and a nondopaminergic neuron. This observation suggests that in wild-type animals, V5.paapp has a cryptic potential fate, the ability to become a dopaminergic neuron, and that V5paa and V5.paap both have the potential to generate dopaminergic daughters. Thus, V5paa and V5.paap can be considered to be of the same cell type (A).

mal spatial homologs, the VC neurons (J. G. White, personal communication). Thus, in the case of P2.aap, cell death may act to modify spatial repeats of a specific lineage pattern. In other instances cell death survivors may express fates normally expressed by their sisters. For example, the extra serotonergic neurons in ted-3 and ted-4 mutants seem likely to be the sisters of the two NSMs. Similarly, the extra dopaminergic neurons often found in ted-3 and ted-4 animals in the region of the ventral cephalic neurons are probably the sisters of the ventral cephalic neurons. In contrast, the cell death survivor in the postdeirid in many instances expresses the fate normally associated with its mother’ sister. The dopaminergic phenotype of s this cell suggests an underlying stem cell or “reiterative” (Chalfie et al., 1981) nature to this lineage. Such reiterative cell lineages, in which a cell divides to produce one or more descendant cells like itself, occur in many animals, including C. elegans. In ted-3 or ted-4 mutants, the postdeirid neuroblast VEi.paa divides to generate a daughter (Vdpaaa) that becomes a dopaminergic neuron, as it does in the wild type, and a daughter (VS.paap) that also generates a dopaminergic daughter, the cell death survivor Kpaapp. Thus, V5paa and its daughter Vdpaap are alike in that they share the potential to generate dopaminergic progeny cells. This observation suggests that in wild-type animals as well, VS.paa and V5.paap share the potential to produce dopaminergic progeny, a potential normally masked by the death of V5.paapp (Figure 11). The underlying reiterative nature of the postdeirid lineage was previously suggested on the basis of observations of mutants defective in the gene uric-86 (Chalfie et al., 1981). The absence of programmed cell death in ted-3 or ced4 animals results in the presence of many extra cells. For example, during the normal development of C. elegans approximately 20% of the neural cells generated subsequently die. Thus, ted-3 and ted-4 animals may have up

to 20% more neural cells than wild-type animals. It is striking, however, that mutations in ted-3 and ted-4 do not have any obvious deleterious effects on behavior. Although ted-3 or ted-4 mutations may cause behavioral defects that we have not detected, it seems clear that the absence of cell death is not grossly detrimental to the nematode. Thus, we are left with the question of what selective advantage may be associated with the extensive programmed cell death that occurs during C. elegans neurogenesis. Even small effects on behavior might provide a selective advantage. Alternatively, the elimination of unnecessary cells that would otherwise differentiate may decrease energy consumption and hence increase fitness. The phenotypes of strains carrying mutations in pairwise combinations of the genes involved in cell death have suggested that these genes act sequentially in a developmental pathway for cell death. ted-3 and ted-4 define the first known step in this pathway. Since relatively few mutants affecting cell death have been identified, it seems likely that other genes in this pathway remain to be defined. Furthermore, in addition to genes involved in all cell deaths, there presumably are genes involved in the specification of particular cell deaths. For example, egl-7 may be one such gene; dominant mutations in egl-7 appear to result in the expression in hermaphrodites of the normally male-specific fate of HSN cell death (Trent et al., 1983), but do not affect other cell deaths (H. Ellis, unpublished observations). The identification of a genetic pathway used for most C. elegans programmed cell deaths helps us to distinguish three distinct classes of cell deaths. First, almost all of the cell deaths that occur during wild-type development involve all four ted genes; deaths of this class appear to be suicides, that is, they are cell autonomous and can be induced by mutations (e.g., egl-7) that lead to the ectopic expression of the complete cell death pathway. Second, at least one death that occurs during normal development (that of the male linker cell) involves ted-7 and ted-2, but can take place without normal ted-3 and ted-4 function; this death (as well as those of a few others that we have not tested for ted-3 or ted-4 involvement but that seem likely to be of the same class) appears to be a murder, that is, it occurs only in the presence of another cell. Third, certain mutation-induced deaths do not require the functions of any of the four ted genes; these deaths (such as those caused by the dominant mutations met-4(e7677) and /in-24(n432) are morphologically abnormal and apparently result from the cell-specific synthesis of cytotoxic products. Although ted-3 and ted-4 appear to act at the same step in the developmental pathway for cell death, genetic evidence suggests that these genes may differ functionally. The suppression of egCI/ + heterozygotes by ced3/ + heterozygotes but not by ted-4/ + heterozygotes indicates that ted-3 function is relatively sensitive to the dosage of the wild-type gene product (a P-fold reduction of wild-type product can lead to a mutant phenotype), while ted-4 function may be relatively less sensitive to dosage. A number of observations have suggested that most

C. elegans Cell Death Mutants a27

cell deaths, like most other cell fates in C. elegans, are specified in a cell autonomous manner (see Horvitz et al., 1982a, for discussion). Although the specification of which cells die is likely to be cell autonomous, cell extrinsic factors could still be involved in cell death. In particular, the experiments described in this paper do not indicate whether ted-3 and ted-4 act within dying cells (or their precursors) or act extrinsically to the dying cells. For example, at least one of these genes could act to control a hormonal signal that triggers cell death; such hormonal influences on neuronal and muscle cell deaths have been identified in the moth Manduca sexta (Truman and Schwartz, 1982a, 1982b). The reduction or loss of ted-3 or ted-4 function results in a transformation of cell fates; in ted-3 or ted-4 mutants, cells that would normally die instead adopt fates normally associated with other cells. Other “homeotic” mutations, which caL.se specific transformations in developmental fates, have been identified in Drosophila (for a review see Lawrence and Morata, 1983) and in C. elegans (Greenwald et al., 1983; Horvitz et al., 1983). Such mutations appear to be good candidates for defining genes involved in the specification of cell fates. It remains possible, though, that the functions of ted-3 and ted-4 are simply necessary for the “differentiation” of the fate of cell death. For example, the products of ted-3 and ted-4 may directly “kill” cells rather than control the activities of other genes that act to kill cells. Nonetheless, the homeotic nature of ted-3 and ted-4 suggests that these genes may be involved in the determination of a specific cell fate, programmed cell death.

Histochemistry Formaldehyde-induced fluorescence (FIF) was used to identify serotonin- or dopamine-containing cells (Sulston et al., 1975; Horvitz et al., 1982b). Diamidinophenolindole (DAPI) (Sigma) was used to stain nuclei. Animals were fixed in 6:3:1 ethanol:chloroform:acetic acid (Carnoy’ solution) and stained with 1.0 Kg/ml DAPI and 1.0 nglml s phenoxypropanol in M9 buffer (W. Frxsen, Ph.D. thesis, M.I.T., Cambridge, Massachusetts, 1965). Behavioral Assays Chemotaxis The chemotaxis assays used were modified from those described by Ward (1973). A 100 m m tissue culture dish was spread evenly with 5 ml of a thick slurry of Sephadex G-200-40 gel beads (Sigma) swelled overnight in 0.1 M HEPES and 2.5% sucrose (pH 7.2). Five microltters of 2 M NaCl was placed in the center of the dish and 4t/z to 5 hr later, 24 adult hermaphrodites were placed around the edge of the dish. After 15 min, the number of worms that had accumulated within a radius of 1.6 cm from the center of the dish was determmed. Male Mating The efficiency of male mating was tested as described by Hodgkin et al. (1979). Six L4 males and six L4 dpy-ll(e224) hermaphrodites were placed on a petri dish seeded with a small (~1 cm) circle of Escherichia coli, and 24 hr later the males were removed. The hermaphrodites were transferred to fresh plates each day and total cross progeny (nonDpy animals) were counted. Pharmacological Tests of Egg Laying Exogenously added serotonin or imipramine stimulate gravid wild-type adults to lay eggs (Horvitz et al., 1982b; Trent et al., 1983). The response of adult ted-3 hermaphrodites to serotonin and imipramine was tested as described by Trent et al. (1963). Gravid adults were placed in microtiter wells containing serotonin (Sigma) (5 mglml) or imipramine (Sigma) (0.75 mg/ml), and 60 min (serotonin) or 90 min (imipramine) later the number of eggs laid by each hermaphrodite was counted. All behavioral assays were performed at 20% Genetic Protocols Mapping ted-3 was tested for linkage to markers on all linkage groups and weak linkage to dpy-73(e784j IV was observed (data not shown). ted-3 was more precisely mapped to a position between uric-30 and uric-26 on linkage group IV (Figure 2). The results of a three-factor cross confirmed the assignment of ted-3 to linkage group IV. 21/21 Dpy non-Uric and O/26 Uric non-Dpy recombinant progeny ot uric-26 dpy-4/ ted-3 heterozygotes segregated ted-3 animals. A second three-factor cross established that ted-3 mapped between uric-30 rand dpy-4; 5/6 Dpy non-Uric and 1110 Uric non-Dpy recombinant progeny of uric-30 dpy-4/ ted-3 heterozygotes segregated ted-3 animals. In these crosses, the Ced-3 phenotype was scored using Nomarski optics to detect the presence of extra cells. A four-factor cross ordered ted-3 with respect to uric-26. Heterozygotes of genotype ted-3 uric-26/imc-30 dpy-4; egl-7 were constructed and from their progeny Ced3 no’ n-Uric-26 and Unc26 non-Ced3 recombinants were picked and scored for the segregation of uric-30 or dpy-4 animals. (ted-3 mutants are indistinguishable from the wild-type as viewed with a dissechng mrcroscope. However, the ability of mutations in ted-3 to suppress the egg-laying defect of egl-7 animals allowed us to score ted-3 in an eg/-? background using a dtssecting microscope.) Three out of three Ced-3 non-Uric-26 recombinants segregated dpy-4 but not uric-30 animals, while 313 Uric-26 non-Ced-3 recombinants segregated uric-30 but not dpy-4 animals, ted-4 was tested for linkage to markers on all linkage groups and was found to be linked only to linkage group Ill; 2/15 Ced-4 non-Uric animals segregating from ted-4/uric-32 heterozygotes gave Uric progeny (frequency of recombination [p] approximately equal to 0.07, since the frequency of animals segregating Uric-32 progeny should be 2pl(l+p), that is, 213 if ted-4 and uric-32 were unlinked and 2p if they were closely linked). A three-factor cross suggested that ted-4 mapped to the left of dpy-77; 313 Lon non-Dpy recombinant progeny of dw-77 /on-l/ted-4 heterozygotes segregated ted-4 animals. A second threefactor cross indicated that ted-4 mapped between uric-93 and dpy-77; 116 Dpy non-Uric and 616 Uric non-Dpy recombinant progeny of uric-93 dpy-77ked-4 heterozygotes segregated ted-4 progeny. An additional

Experimental

Procedures

General Methods and Strains The techniques used for the culturing and ethyl methanesulfonate (EMS) mutagenesis of C. elegans were as described by Brenner (1974). All strains were grown at 20% unless otherwise indicated. The wild-type parent of all strains described here was C. elegans variety Bristol, strain N2 (Brenner, 1974). The genetic markers used are as follows: LG I dpy-5(e67j, ted-l(e7735), the-3(e7724j; LG II dpy-7O(e728), b/i-2(e768); LG Ill unc93(e7500), uric-79(e7068), dw-77(ei64), unc32(e189); LG IV ted-2(e7752), dpyG’ (e784). /im3(e7477/, dpy-2O(e7362), uric-22(e66j, uric-3O(e797), um26(e205), lev-7(x22), tra-3(e7767), dpy4(e7166); LG V dpy-ll(e224j. egl-l(n487, n7084j; LG X dpy-3(e27), nut-7(e7392), /on-2(e678). These markers have been described by Brenner (1974) or by Swanson et al. (1964). This paper conforms to the standard system of C. elegans genetic nomenclature (Horvitz et al.. 1979). Cell Lineage The use of Nomarski differential interference contrast microscopy for the determination of cell lineages has been described by Sulston and Horvitz (1977). All lineages discussed in this manuscript were observed in at least three animals. The nomenclature used to designate individual cells reflects their lineage history (Sulston and Horvitz, 1977). Blast cells are designated by upper case letters, which may be followed by numbers to indicate a set of analogous blast cells, for example, Vl through V6. L (left) or R (right) are added to blast cell names to indicate identical pairs of postembryonic cells found on the left and right sides of the animal, for example, V5L and V5R. The progeny of a blast cell are represented by adding to the blast cell name a lower case letter indicating the relative positions of the daughters immedrately following division; for example, V5.a is the anterior daughter of V5.

Cell a28

three-factor cross indicated that ted-4 mapped between uric-79 and dpy-77 (Figure 2); 5/5 Uric non-Dpy and 117 Dpy non-Uric recombinant progeny of uric-79 dpy-Weed-4 heterozygotes segregated ted-4 animals. In these crosses, the Ced-4 phenotype was scored using Nomarski optics to detect the presence of extra cells. Compfementation Tests Putative ted-3 alleles (ted-3’ were tested for failure to complement ) ted-3(n777). Heterozygotes of genotype ted-3(n777)/ced-3’ were constructed and scored using Nomarski microscopy for the presence of cell death survivors. ted-4(n1162) was tested for complementation of ted-3(n777). ced4(n7762); egl-l(n7084) males were crossed with ted-3(n777) dpy4(e7766); egl-l(n487) hermaphrodites. The resulting cross progeny (non-Dpy) hermaphrodites were egg-laying defective (non-Ced-3) and 12112 cross progeny scored using Nomarski microscopy did not have a Ced-3 phenotype. Complementation Screens ted-3(n7729) was isolated on the basis of its failure to complement ced3(n(nn7). L4eg/-l(n487) males were mutagenized with EMS and crossed with ted-3 dpy-4; egC7; dpy3 hermaphrodites. The majority of the hermaphrodite cross progeny were non-Dpy and egg-laying defective (genotype ted-3 dpy-4/+ +; egC7; dpy-3/+) rare animals were phenotypically wild-type and candidates for harboring a new ted-3 allele (genotype ted-3 dpy-4/ted-3” +; egl-7; dpy-3/+). Putative new ted-3 alleles were isolated from these heterozygotes on the basis of their failure to segregate dw-4 progeny and were phenotypically and genetically characterized. In this experiment approximately 12,000 Fl cross progeny were examined, and one new ted-3 allele, n7129, was generated. Generation of Deficiencies for cedd To generate the deficiency nDf27 IV, L4 males on petri dishes were irradiated for 2.5 min at a dose rate of approximately 3000 radlmin in a Gamma Cell 220 Co source (Atomic Energy of Canada, Ltd.) (Greenwald and Horvitz, 1960) and then crossed with L4 lev-7(x22) dpy4(e7766) hermaphrodites. The recessive mutation lev-7(x22) results in resistance to the antihelmenthic drug levamisole (Lewis et al., 1960). Rare levamisole-resistant non-Dpy Fl animals (four of 5000 Fl animals examined) were selected. These animals were candidates for carrying deficiencies of the lev-7 locus. Two strains segregated approximately one-quarter inviable zygotes and were tested to determine whether they contained a lethal levamisole-resistant allele that failed to complement other loci in the region (i.e., to ascertain whether the lethal levamisole-resistant allele was a deficiency). One of these strains contained a lethal lev-I allele (nDf27) that failed to complement uric-26, ted-3, uric-30, uric-22(H. Ellis, unpublished observations), and dpy20 (E. Ferguson, personal communication) and complemented lin-3 (E. Ferguson, personal communication), tra-3 and dpy-4 (Figure 2). The second strain, which was subsequently lost, also contained a deficiency of the lev-7 locus. Acknowledgments We thank Ed Hedgecock for sharing with us his strains, thoughts, and unpublished data. We thank John Sulston for encouragement and advice, and Nancy Tsung and Chand Desai for isolating some of the mutants we have studied. We also thank Leon Avery, Marty Chalfie, Chip Ferguson, Bill Fixsen, Ed Hedgecock, Jonathan Hodgkin, and Carol Trent for discussions and comments on the manuscript. This research was supported by U.S. Public Health Service grants GM24663 and GM24943 and Research Career Development Award HD00369 to H. R. Horvitz. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “ advertisement” in accordance with 18 U.S.C. Section 1734 ‘ solely to indicate this fact. Received References Albertson, D., and Thomson, N. (1976). The pharynx of Caenorhabdifis elegans. Phil. Trans. Roy. Sot. (Lond.) B 275, 299-325. Albertson, D., Sulston, J. E., and White, J. G. (1976). Cell cycling and September 30, 1985; revised December 23, 1985.

DNA replication In a mutant blocked in cell division in the nematode Caenorhabditis elegans. Dev. Biol. 63, 165-178. Brenner, S. (1974). The genetics of Caenorhabdifis 77; 71-94. elegans. Genetics

Chalfie, M., and Sulston, J. E. (1981). Developmental genetics of the neurons of Caenorhabditis elegans. Dev. Biol. 82, mechanosensory 358-370. Chalfie, M.,Horvitz, H. R., and Sulston, J. E. (1981). Mutations that lead to reiterations in the cell lineages of C. elegans. Cell 24, 59-81. Chalfie, M., Sulston, J. E., White, J. G., Southgate, E., and Thomson, J. N. (1985). The neural circuit for touch sensitivity in Caenorhabditis elegans. J. Neurosci. 5, 956-964. Cowan, W. M., Fawcett. J., O’ Leary, D., and Stanfield, B. (1984). Regressive events in neurogenesis. Science 225, 1258-1265. Ferguson, E. L., and Horvitz, H. R. (1965). Identification and characterization of 22 genes that affect the vulva1 cell lineages of the nematode Caenorhabditis elegans. Genetics 770, 17-72. Greenwald, I. S., and Horvitz, H. R. (1980). uric-93(e7500): a behavioral mutant of Caenorhabditis elegans that defines a gene with a wild-type null phenotype. Genetics 96, 147-164. Greenwald, I. S., Sternberg, l? W., and Horvitz, H. R. (1983). The /in-72 locus specifies cell fates in Caenorhabditis elegans. Cell 34, 435-444. Hedgecock, E., Sulston, J. E., and Thomson, N. (1963). Mutations affecting programmed cell deaths in the nematode Caenorhabditis e/egans. Science 220, 12774280. Herman, R., and Horvitz, H. R. (1980). Genetic analysis of Caenorhabditis elegans. In Nematodes as Biological Models, B. Zuckerman, ed. (New York: Academic Press), pp. 227-264. Hodgkin, J., Horvitz, H. R., and Brenner, S. (1979). Nondisjunction mutants of the nematode Caenorhabditis elegans. Genetics 97, 67-94. Horvitz, H. R., Brenner, S., Hodgkin, J., and Herman, R. (1979). A uniform genetic nomenclature for the nematode Caenorhabditis elegans. Mol. Gen. Genet. 775, 129-133. Horvitz, H. A., Ellis, l-t. M., and Sternberg, P W. (1962a). Programmed cell death in nematode development. Neurosci. Comment. 7, 56-65. Horvitz, H. R., Chalfie, M., Trent, C., Sulston, J., and Evans, P. (1982b). Serotonin and octopamine in the nematode C. elegans. Science 276, 1012-1014. Horvitz, H. R., Sternberg, i? W., Greenwald, I, S., Fixsen, W., and Ellis, H. M. (1983). Mutations that affect neural cell lineages and cell fates during the development of the nematode Caenorhabditis elegans. Cold Spring Harbor Symp. Quant. Biol. 48, 453-463. Kimble, J. E., and Hirsh, D. (1979). The postembryonic cell lineages of the hermaphrodite and male gonads in Caenorhabditis elegans. Dev. Biol. 70, 396-417. Lawrence, P. A., and Morata, G. (1983). The elements of the bithorax complex. Cell 35, 595-601. Lewis, J., and Hodgkin, J. (1977). Specific neuroanatomical changes in chemosensory mutants of the nematode Caenorhabditis elegans. J. Comp. Neurol. 772, 489-510. Lewis, J., Wu, C. H., Berg, H., and Levine, J. (1980). Genetics of levamisole resistance in the nematode Caenorhabditis elegans. Genetics 95, 905-928. Saunders, 604-612. J. (1966). Death in embryonic systems. Science 754,

Sulston, J. E. (1976). Post-embryonic development in the ventral cord of Caenorhabditis elegans. Phil. Trans. Roy. Sot. (Lond.) B 275, 287-297. Sulston, J. E., and Horvitz. H. R. (1977). Post-embryonic cell lineages of the nematode Caenorhabdifis elegans. Dev. Biol. 82, 110-156. Sulston, J. E., and White, J. G. (1980). Regulation and cell autonomy during postembryonic development of Caenorhabditis elegans. Dev. Biol. 78, 577-597. Sulston, J. E., Dew, M., and Brenner, S. (1975). Dopaminergic neurons in the nematode Caenorhabdifis elegans. J. Comp. Neuroi. 763, 215-226. Sulston, J. E., Schierenberg, E., White, J. G., and Tnomson, N. (1963).

C. elegans Cell Death Mutants 829

The embryonic cell lineage of the nematode Caenorhabditis Dev. Biol. 100, 64-119.

elagans.

Swanson, M. M., Edgeley, M. L., and Riddle, D. L. (1984). Caenorhabditis elegans. In Genetic Maps, S. O’ Brien, ed. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory), pp. 244-258. Trent, C., Tsung, N., and Horvitz, H. R. (1983). Egg-laying defective mutants of the nematode Caenorhabditis elegans. Genetics 704,619-647. Truman, J. W. (1984). Cell death in invertebrate Rev. Neurosci. 7, 171-188. nervous systems. Ann.

Truman, J. W., and Schwartz, L. M. (1982a). Insect systems for the study of programmed neuronal death. Neurosci. Comment. 7, 66-72. Truman, J. W, and Schwartz, L. M. (1982b). Programmed cell death in the nervous system of a moth. Trends Neurosci. 5, 270-273 Ward, S. (1973). Chemotaxis by the nematode Caenorhabditis elegans: Identification of attractants and analysis of response by use of mutants. Proc. Natl. Acad. Sci. USA 70, 817-821. Wills, N., Gesteland, Ft. F., Karn, J., Barnett, L., Bolten, S., and Waterston, R H. (1983). The genes sup-7X and sup-5 111 C. elegans supof press amber nonsense mutations via altered transfer RNA. Cell 33, 575-583.