Abstract
Programmed cell death (PCD) is regulated by multiple evolutionarily conserved mechanisms to ensure the survival of the cell. Here we describe pvl-5, a gene that likely regulates PCD in Caenorhabditis elegans. In wild-type hermaphrodites at the L2 stage there are 11 Pn.p hypodermal cells in the ventral midline arrayed along the anterior-posterior axis and 6 of these cells become the vulval precursor cells. In pvl-5(ga87) animals there are fewer Pn.p cells (average of 7.0) present at this time. Lineage analysis reveals that the missing Pn.p cells die around the time of the L1 molt in a manner that often resembles the programmed cell deaths that occur normally in C. elegans development. This Pn.p cell death is suppressed by mutations in the caspase gene ced-3 and in the bcl-2 homolog ced-9, suggesting that the Pn.p cells are dying by PCD in pvl-5 mutants. Surprisingly, the Pn.p cell death is not suppressed by loss of ced-4 function. ced-4 (Apaf-1) is required for all previously known apoptotic cell deaths in C. elegans. This suggests that loss of pvl-5 function leads to the activation of a ced-3-dependent, ced-4-independent form of PCD and that pvl-5 may normally function to protect cells from inappropriate activation of the apoptotic pathway.
APOPTOSIS, programmed cell death, is a morphologically distinct form of physiological cell death that is essential for normal development and homeostasis in metazoans. Apoptotic cell death is used to sculpt structures or organs during development, to remove unused or unwanted cells or structures, to control cell numbers, and to remove damaged, infected, or otherwise potentially dangerous cells (Jacobson et al. 1997; Vaux and Korsmeyer 1999; Meier et al. 2000). In most cases, apoptotic cell deaths are physiological “suicides” in which the cell activates an intrinsic death program that leads to its orderly destruction and engulfment by other cells.
During development of the nematode Caenorhabditis elegans hermaphrodite, 1090 somatic cells are born, of which 131 undergo programmed cell death and engulfment by neighboring cells (Sulston and Horvitz 1977; Kimble and Hirsh 1979; Sulston et al. 1983). Genetic analysis of these programmed cell deaths has identified a pathway of genes regulating the death or survival of specific cells (Metzstein et al. 1998). Loss-of-function mutations in the genes ced-3, ced-4, and egl-1 and a gain-of-function mutation in the gene ced-9 lead to a suppression of most or all programmed cell deaths during normal development, indicating that ced-3, ced-4, and egl-1 are required for normal somatic programmed cell deaths in the worm. Conversely, animals carrying a loss-of-function mutation in ced-9 show massive ectopic cell death and die early during development, indicating that the normal function of ced-9 is to protect cells from activation of the apoptotic pathway during development. Genetic epistasis analysis has shown that these genes act in a pathway in the following order: egl-1, ced-9, ced-4, ced-3.
Molecular characterization of these genes showed that ced-3 encodes a caspase that presumably becomes active in cells fated to die and initiates the cell death program by the proteolysis of specific target proteins (Yuan et al. 1993). ced-4 encodes an adaptor protein with sequence and functional similarity to the mammalian Apaf-1 protein (Yuan and Horvitz 1992; Cecconi et al. 1998). CED-4 can physically interact with both CED-3 and CED-9, as well as with itself (Spector et al. 1997; Wu et al. 1997a,b; Chen et al. 2000). The oligomerization of CED-4 is believed to result in the cleavage of inactive CED-3 into its active form via CED-3 autocatalysis (Chinnaiyan et al. 1997; Seshagiri and Miller 1997; Wu et al. 1997a; Yang et al. 1998). ced-9 encodes a protein homologous to the mammalian antiapoptotic protein Bcl-2 (Hengartner and Horvitz 1994b), and bcl-2 can function for ced-9 in C. elegans (Hengartner et al. 1992; Vaux et al. 1992). CED-9 is believed to function in a protective manner by binding to CED-4 and preventing CED-4 oligomerization, thereby preventing CED-3 autoactivation (Chinnaiyan et al. 1997; Seshagiri and Miller 1997; Wu et al. 1997a; Yang et al. 1998). Finally, egl-1 encodes a proapoptotic bcl-2 family member of the “BH3 domain only” subtype (Conradt and Horvitz 1998). In cells fated to die, the cell death pathway is activated when EGL-1 binds to membrane-tethered CED-9, displacing CED-4 from CED-9, thereby allowing CED-4 to oligomerize and CED-3 to become active (del Peso et al. 2000).
While many cells die during development due to activation of a programmed cell death (PCD) pathway, other cells die by necrosis. Necrotic cell deaths in mammalian cells occur in response to physiological insult or injury. Necrotic cell death is morphologically distinct from PCD (reviewed in Kerr et al. 1972). During PCD, dying cells shrink, their membranes form blebs, and their chromatin undergoes extensive fragmentation until, ultimately, the dying cell is engulfed by neighboring cells (Sulston and Horvitz 1977). During necrosis, cells become enlarged and burst, spilling their contents into the extracellular space, often resulting in an inflammatory response. Necrotic cell death in C. elegans induced by gain-of-function mutations in deg-1, deg-3, and mec-4 resembles the necrotic cell deaths seen in mammalian cells. These degenerin genes encode ion channels, and mutations in these genes appear to perturb the ionic balance in cells, resulting in their death (Driscoll and Chalfie 1992).
Genetic analysis in C. elegans has also identified genes required for the orderly engulfment of cells dying by either necrosis or apoptosis (Gumienny and Hengartner 2001). Mutations that result in the persistence of cell corpses have identified seven genes that define two partially redundant pathways required for the engulfment of corpses: ced-1, ced-6, ced-7 and ced-2, ced-5, ced-10, ced-12 (Ellis et al. 1991; Chung et al. 2000). In addition to removal of corpses, the phagocytic machinery also promotes the killing of cells that are programmed to die, as mutations in ced-1, ced-6, and ced-7 result in increased survival of some cells that are destined to die (Hoeppner et al. 2001; Reddien et al. 2001).
Although structurally and functionally conserved homologs of these nematode proteins function in programmed cell death in mammals and other invertebrates such as Drosophila, it appears that in other species the regulation of apoptosis may be more complicated than the apparently linear pathway controlling C. elegans cell death. In mammals, for example, the initiation of programmed cell death is regulated both extracellularly by binding of ligands and activation of death receptors and intracellularly by the release of proapoptotic factors from the mitochondria in response to cytotoxic stress and other conditions (Budihardjo et al. 1999; Earnshaw et al. 1999). Also, in both mammals and Drosophila, the activity of caspases is regulated by the binding of inhibitor of apoptosis proteins, which must be removed or destroyed for the efficient activation of the apoptotic pathway (Salvesen and Duckett 2002). There is currently little evidence for regulation of programmed cell death in C. elegans by these types of mechanisms, although recent work suggests that proapoptotic factors released from the mitochondria can regulate programmed cell death in C. elegans (Parrish et al. 2001; Wang et al. 2002). However, even in C. elegans not all cells that undergo programmed cell death follow the linear program described above. For example, in the gonad of the adult hermaphrodite, large numbers of oocytes die by a ced-3- and ced-4-dependent, but egl-1- and ced-9-independent PCD pathway (Gumienny et al. 1999).
Although much is known about the mechanism of apoptosis, comparatively little is known about the developmental regulation of programmed cell death in C. elegans. In the worm, most cell deaths occur in an invariant pattern, suggesting that death represents the adoption of a distinct cell fate by these cells. Transcriptional regulation of egl-1 is believed to control the death or survival of cells during development (Conradt and Horvitz 1999), and two genes encoding transcription factors, ces-1 and ces-2, are known to regulate specific cell deaths (Metzstein et al. 1996; Metzstein and Horvitz 1999). Whether other mechanisms protect specific cells from activation of this pathway during development is unclear. Here we describe pvl-5, a gene that appears to inhibit abnormal cell death in the epithelial cells that will give rise to the adult vulval structure.
The vulva, part of the egg-laying apparatus of the adult C. elegans hermaphrodite, is formed from a small number of vulval precursor cells (VPCs) found in the ventral hypodermis of the developing larva (reviewed in Greenwald 1997; Kornfeld 1997). The VPCs are a subset of 12 cells that are the posterior daughters (Pn.p cells) of the P neuroectoblast cells. At hatching, the 12 P cells are present as two rows of 6 cells. The 12 P cells undergo short-range migrations to align along the ventral midline (Podbilewicz and White 1994). In the L1 stage, each P cell divides to give an anterior Pn.a neuroblast and a posterior Pn.p hypodermal cell. The cells P1.p, P2.p, P9.p, P10.p, and P11.p fuse with the surrounding hypodermal cellular syncytium, while the P12.p cell divides once to generate a hypodermal daughter and a daughter that dies by programmed cell death (Sulston and Horvitz 1977; Podbilewicz and White 1994). The remaining 6 unfused Pn.p cells (P3.p–P8.p) are all competent to adopt vulval fates and are the VPCs. These 6 cells adopt distinct vulval cell fates on the basis of input from several extracellular signaling pathways (Greenwald 1997; Kornfeld 1997). Three of the cells (P3.p, P4.p, and P8.p) adopt an uninduced fate, which is to divide once to make daughter cells that fuse with the hypodermal syncytium in the L3 stage. The other 3 cells (P5.p, P6.p, and P7.p) adopt induced cell fates and divide to generate the 22 cells that form the adult vulval structure.
Genetic screens have identified several mutations that affect the generation of the VPCs. In unc-83 and unc-84 mutants, the migration of P nuclei is affected, resulting in fewer Pn.p cells in the ventral midline (Ferguson and Horvitz 1985). Mutations in the gene lin-26, which encodes a zinc finger transcription factor, cause a Pn.p (hypodermal) to Pn.a (neuronal) cell fate transformation, resulting in a Vulvaless phenotype (Labouesse et al. 1994). lin-39 is a Hox gene expressed in the midbody region, including the VPCs. In lin-39 null mutants the VPCs adopt the fate of their anterior and posterior non-VPC sister cells and fuse with the syncytium in the L1 stage (Clark et al. 1993; Wang et al. 1993). Finally, in lin-24(n432) and lin-33(n1043) gain-of-function mutants, some of the Pn.p cells undergo cell death and the surviving cells exhibit abnormal cell division patterns (Ferguson and Horvitz 1985). The cell deaths in lin-24 and lin-33 mutants are genetically and morphologically distinct from the programmed cell deaths seen during normal development and appear to be degenerative in nature. Together with the semidominant nature of these alleles, this suggests that these lin-24 and lin-33 alleles may encode mutant gene products that are toxic to the Pn.p cells.
pvl-5, a gene that affects the generation of Pn.p cells, was identified in a screen based on the protruding vulva (Pvl) phenotype. In pvl-5 mutant animals there are <12 Pn.p cells in the ventral midline; consequently, often fewer than three VPCs are induced, leading to defective vulval formation (Eisenmann and Kim 2000). Here we show that in pvl-5 mutants, the Pn.p cells undergo abnormal cell death that is suppressible by ced-3 loss-of-function and ced-9 gain-of-function mutations, suggesting that the programmed cell death pathway is activated in Pn.p cells in the pvl-5 mutant. However, the Pn.p cell death is not suppressed by loss of ced-4 function, suggesting that the cell death machinery is activated by a novel mechanism that is independent of ced-4. These results suggest that pvl-5 normally functions in the Pn.p epithelial cells to prevent the inappropriate activation of the machinery responsible for cell death in the Pn.p cells, thereby promoting the survival of these cells.
MATERIALS AND METHODS
Genetic methods and alleles:
The culture and genetic manipulation of C. elegans was carried out as described in Brenner (1974). Wild-type animals were C. elegans N2 Bristol strain. Experiments were performed at 20° unless otherwise indicated. The genes and alleles are as referenced in Riddle et al. (1997) unless noted. LGI, ced-1(e1735), unc-75(e950); LGII, pvl-5(ga87) (Eisenmann and Kim 2000), pvl-5(de4) (this study), dpy-10(e128), rol-6(e187), unc-4(e120), bli-2(e768), arDp2; LGIII, ced-4(n1162, n1894, n2274), dpy-17(e164), ced-6(n2095), ced-7(n1892), ced-9(n1950, n1950n2161) (Hengartner et al. 1992), unc-69(e587) unc-47(e307), lon-1(e185); LGIV, ced-2(e1752), ced-3(n717), ced-3(n2433) (Shaham et al. 1999), ced-5(n1812), ced-10(n1993), unc-17(e245), unc-30(e191), dpy-20(e1282); LGV, egl-1(n1084 n3082) (Conradt and Horvitz 1998), unc-61(e228). arDp2 is a free duplication of a small region near the center of chromosome II. jcIs1 indicates the integrated array containing ajm-1::GFP and a dominant rol-6 co-injection marker (Mohler et al. 1998). qIs56 indicates the integrated array containing lag-2::GFP and unc-119(+) (Siegfried and Kimble 2002).
pvl-5(de4) was isolated in an F1 noncomplementation screen in which EMS-mutagenized him-5(e1490) males were mated into pvl-5(ga87) rol-6(e187); unc-3(e151) animals. F1 Egl/Pvl nonRol animals were cloned. The newly identified allele was made homozygous, backcrossed twice with dpy-10; him-5, and tested for failure to complement pvl-5(ga87). Three-factor mapping places de4 between bli-2 and unc-4 as 4/17 Unc nonBli and 2/5 Bli nonUnc recombinants contain de4. de4, like ga87, shows failure to complement with the deficiency mnDf39. Only a single new allele was isolated from 41,000 EMS-mutagenized haploid genomes and 38,000 trimethylpsoralen/UV-mutagenized haploid genomes.
The general strategy for creating double mutants containing pvl-5(ga87) and a ced mutation was to create animals of the genotype pvl-5(ga87)/+; ced/linked marker and then to identify animals from their progeny that were homozygous for pvl-5(ga87) and that never segregated the linked marker. The presence of the ced mutation was verified by counting cell corpses in the embryo or L1 larvae of the putative double mutant.
ced-1(e1735); pvl-5(ga87) unc-4(e120); ced-5(n1812) dpy-20(e1282) was built by first constructing ced-1(e1735)/+; pvl-5(ga87) unc-4(e120)/+ +; ced-5(n1812) dpy-20(e1282)/+ +. Unc nonDpy animals with persistent corpses in the adult hermaphrodite germline were cloned. Dpy Unc progeny from these animals were then cloned. The average number of Pn.p cells in pvl-5(ga87); dpy-20(e1282) is 7.3 ± 1.5 (n = 53), indicating that dpy-20(e1282) does not suppress the pvl-5 phenotype. Brood size was determined by picking five L4 hermaphrodites to NGM plates seeded with OP50. The worms were transferred to a fresh plate every 24 hr for 4 days. The progeny on each plate were counted 3 days afterward.
Observation of P cells, Pn.p cells, and cell corpses:
To obtain synchronized L1 larvae, embryos were allowed to hatch overnight on plates with no food. The partially synchronized L1 larvae were transferred to NGM plates supplemented with OP50 and allowed to feed for 6–9 hr. Pn.p cell counts were done by observing hypodermal cell nuclei in the ventral cord of synchronized mid-L2 larvae using Nomarski optics on a Zeiss Axioplan2 microscope. Cell corpses in the head were observed and counted as described in Ellis et al. (1991). Germline corpses in the adult hermaphrodite were visualized by staining with SYTO-12 as described in Gumienny et al. (1999). To examine the migration of the P cells, a synchronous population of pvl-5(ga87); jcIs1 L1 larvae was obtained as described above. The position and behavior of the ajm-1::GFP-expressing P cells was observed every hour using fluorescence microscopy on a Zeiss Axioplan2 microscope. For lineage analysis, two larvae were picked onto an agarose pad and the behavior and divisions of the P or Pn.p cells were followed as described (Sulston and Horvitz 1977). Four pvl-5(ga87) L1 larvae were observed from the mid-L1 larval stage through the L1/L2 molt for a few hours into the L2 larval stage. The remaining two animals were observed starting at 9 hr posthatching, by which time the P cells had already descended into the ventral midline and divided to generate the Pn.p cells.
RNA interference:
The ced-3 plasmid for RNA interference (RNAi) by feeding was generated by PCR amplification of N2 genomic DNA with primers CED35 (GCTCCTGACAATTCGAGACTTTGCC) and CED33 (TTAGACGGCAGAGTTTCGTGCTTCC). The ced-4 plasmid for RNAi by feeding was generated by PCR amplification of N2 genomic DNA using the primers CED45 (CAGTAAAATGTCAACTCGCCTCG) and CED43 (CTGGATTTCCACTGCTTAGTTCG). Both PCR products were ligated into the NotI site of the feeding vector pPD129.36 and transformed into the T7 polymerase-expressing Escherichia coli strain HT115 (Timmons et al. 2001). Bacteria were grown to an OD of 0.6–0.8, induced with 1 mm isopropyl thiogalactoside (IPTG) for 4 hr, and then concentrated and spread onto NGM plates containing 1 mm IPTG, 50 μg/ml carbenicillin, and 12 μg/ml tetracycline. L3 larvae were transferred directly onto these plates at 20° and their progeny were scored at the L2 stage for the number of Pn.p cells. The efficacy of RNAi was tested by scoring neuronal deaths in ced-1 L1 larvae. An average of 16.6 corpses (n = 24) were seen in the heads of ced-1 L1 larvae fed on E. coli transformed with empty feeding vector. ced-3(RNAi) and ced-4(RNAi) larvae had fewer neuronal corpses in the head at the L1 stage, 8.0 (n = 34) and 8.4 (n = 29), respectively.
Antibody staining and immunofluorescence:
Double staining of animals with the anti-LIN-31 and MH27 antibodies was carried out as described by Eisenmann et al. (1998) and stained animals were observed using fluorescence microscopy on a Zeiss Axioplan2 microscope.
RESULTS
pvl-5 mutants have defects in the number of Pn.p cells:
pvl-5(ga87) was originally identified in a screen for mutants in which vulval development is abnormal, resulting in a protruding vulva (Pvl) phenotype (Eisenmann and Kim 2000). In wild-type animals, there are 11 large Pn.p hypodermal nuclei in the ventral midline at the start of the L2 larval stage. In pvl-5(ga87) mutants there are an average of 7 Pn.p cells in the ventral midline at this time (range 4–10; Table 1) . Consequently, there are often fewer than six VPCs in pvl-5(ga87) animals and <3 Pn.p cells that can be induced to adopt vulval fates (Figure 1B) . This underinduction defect can lead to a protruding vulva (Pvl) and/or egg-laying defective (Egl) phenotype (Table 2) .
There are fewer Pn.p cells at the L2 larval stage in pvl-5 mutants
pvl-5 phenotypes. (A, C, and E) Wild-type N2. (B, D, and F) pvl-5(ga87) mutant. (A) Wild-type vulva at the mid-L4 stage. (B) An underinduced vulva at the mid-L4 stage. (C and D) Double staining with α-LIN-31 and MH27 antibodies. Pn.p nuclei stained with α-LIN-31 appear as bright oval spots along the ventral midline of L2 larvae. Staining with MH27 antibody, which recognizes the protein AJM-1 found at the adhesion junctions of epithelial cells, appears as a line ventral to the nuclei of unfused epithelial cells. (C) Eleven Pn.p nuclei (arrows and arrowheads) including six vulval precursor cells (arrows) in the ventral midline of a wild-type L2 larvae. (D) There are fewer Pn.p nuclei (arrows and arrowheads) and only three vulval precursor cells (arrows) in pvl-5(ga87) L2 larvae. (E) Wild-type migration of the posterior arm of the hermaphrodite gonad. (F) Abnormal migration of the posterior arm of the hermaphrodite gonad. (G) Wild-type male tail with nine rays on each side. (H) Missing rays in pvl-5(ga87) him-5 males (arrows). The arrowhead indicates the possible fusion of rays 8 and 9. Except in G and H, anterior is to the left and dorsal to the top. G and H show a ventral view.
pvl-5(ga87) and pvl-5(de4) mutant phenotypes
To demonstrate this defect in Pn.p and VPC numbers, we stained pvl-5(ga87) animals with an antibody to the transcription factor LIN-31, which is expressed in the nuclei of P1.p–P11.p (Miller et al. 1993; Tan et al. 1998), and with the antibody MH27, which recognizes AJM-1, a component of epithelial cell junctions (Mohler et al. 1998). In wild-type animals at the late L2 stage, 11 nuclei express LIN-31, and the central five (P4.p–P8.p) or six (P3.p–P8.p) of these cells retain their cell junctions (the remainder fuse with the hypodermal syncytium after their birth in the late L1; Figure 1C). In pvl-5(ga87) animals, fewer cells show lin-31 and ajm-1 expression. For example, the animal in Figure 1D shows only six lin-31 expressing cells and, of these, only three have ajm-1 expression. This further illustrates the loss of Pn.p cells in this mutant strain and indicates that Pn.p-like cells are not found elsewhere in the body in pvl-5 mutants.
The defect in the number of Pn.p cells in pvl-5(ga87) appears to be fully recessive at 15°, 20°, and 25° (Table 1 and data not shown). Also, when placed in trans to the deficiency mnDf39, the defect in the number of Pn.p nuclei does not get significantly worse (Tables 1 and 2). In addition, pvl-5(ga87)/pvl-5(ga87)/+ animals have wild-type numbers of Pn.p nuclei at 15°, 20°, and 25° (using the free duplication arDp2; Table 1 and data not shown). Together, these results suggest that the ga87 phenotype may represent the strong loss-of-function or null phenotype for this locus.
A second allele of pvl-5, de4, was identified after screening >70,000 haploid genomes in several F1 noncomplementation screens (see materials and methods). We do not know why additional alleles of this locus are so difficult to identify (see discussion). pvl-5(de4) animals have a slightly higher average of 9 Pn.p cells per animal (range 6–11; Table 1), suggesting that de4 is a weaker allele of pvl-5 than ga87. All of our analysis on pvl-5 was carried out using pvl-5(ga87) animals.
In addition to having fewer Pn.p cells, pvl-5(ga87) mutants exhibit several other phenotypes. First, ∼30% of pvl-5(ga87) progeny show embryonic lethality (Table 2). The multicellular embryos appear misshapen and circular as opposed to the oval appearance of wild-type embryos (data not shown). We do not know the reason for the embryonic lethality observed in pvl-5 mutants; however, we have observed excess cell deaths in pvl-5(ga87) early larvae (see below), suggesting that loss of cells could contribute to this phenotype. Second, pvl-5(ga87) animals display defects in somatic gonad morphology. In wild-type hermaphrodites, the gonad is a symmetrical two-armed structure that arises due to stereotyped migration of the anterior and posterior gonad arms (Kimble and Hirsh 1979). We found that 44% of pvl-5(ga87) animals displayed abnormal migration of one or both gonad arms (Table 2; Figure 1F). Less frequently (6%), animals displayed complete absence of a gonad arm (data not shown). The migration of the gonad arms is mediated by the distal tip cells (DTCs), which express the Notch pathway ligand lag-2 (Henderson et al. 1994). The missing gonad arm observed in pvl-5(ga87) is not likely due to the absence of DTCs, since two lag-2::GFP-expressing cells were always observed in the gonads of pvl-5(ga87) animals (data not shown). In addition to the gonad migration defect, a small percentage of pvl-5(ga87) adults have an everted gonad (Spew phenotype; Table 2). As a consequence of these embryonic, vulval, and gonadal defects, pvl-5 mutants have a reduced brood size compared to wild type (Table 2). Approximately 20% of pvl-5(ga87) animals exhibited severely reduced fecundity with fewer than five viable progeny/hermaphrodite. Finally, pvl-5(ga87) males exhibit defects in the number of rays. The male tail is characterized by the presence of nine pairs of sensilla known as rays found in association with an acellular structure called the fan (Sulston et al. 1980). pvl-5(ga87) males have an average of 7.7 ± 1.6 rays per side compared to an average of 8.9 ± 0.3 (N = 30) rays per side in him-5 males. In addition to missing rays, occasional ray fusion defects are also observed (Figure 1H).
The Pn.p cells undergo abnormal cell death in pvl-5 mutants:
There are several possible explanations for the presence of fewer Pn.p cells in pvl-5 mutants at the L2 stage. The decrease in Pn.p cell numbers could be due to defects in the generation, migration, or survival of the P cells (the Pn.p mothers) or defects in Pn.p cell generation, differentiation, or survival. To better understand the Pn.p cell defect in pvl-5 mutants, we examined pvl-5(ga87) animals expressing ajm-1::GFP, a translational fusion protein that marks the boundaries of all epithelial cells, including the P and Pn.p cells (Mohler et al. 1998; Koppen et al. 2001). In wild-type larvae 12 ajm-1::GFP-expressing Pn.p cells are present in the ventral midline at 9 hr posthatching (mid L1 stage; Mohler et al. 1998). We also observed 12 ajm-1::GFP-expressing cells in pvl-5(ga87) mutants at this time, suggesting that the Pn.p cells are born correctly in these animals (data not shown). Consistent with this result, there is no increase in the number of neurons in the ventral cord as would be predicted if there were a Pn.p to a Pn.a cell fate transformation like that seen in lin-26 mutants (data not shown). Therefore, the defect in pvl-5 mutants does not appear to be in the generation, migration, or survival of the P cells or in their division to generate Pn.p hypodermal daughters in the L1 stage, suggesting that the loss of Pn.p cells occurs after these events.
To address the possibility of a later defect in the Pn.p cells we performed lineage analysis of the P cells in pvl-5(ga87) larvae. Lineage analyses of L1 and L2 larvae showed directly that the P cells migrate into the ventral midline and divide correctly to generate 12 Pn.p hypodermal cells in pvl-5 mutant animals (Figure 2) . However, around the time of the L1 molt some of the Pn.p cells were observed to lose their characteristic hypodermal nuclear morphology. Such nuclei became more refractile, resembling cells undergoing apoptosis. However they differ from the typical “button-shaped” apoptotic cell corpse morphology seen in C. elegans in that they were more oval and often had a slight depression in the center (Figure 3A) . These cell corpses were resolved rapidly in a process that was usually completed within 30 min. Occasionally the cell escaped dying and recovered; however, the resultant nucleus was smaller in size and showed abnormal morphology (data not shown). We have also observed cells with a large swollen nucleus reminiscent of cells undergoing necrosis (Figure 3B). (Note that here and throughout we refer to Pn.p “cells,” when we are actually observing only the nuclei of these cells, some of which may have already fused with the surrounding syncytium.) We observed 21 Pn.p cell deaths in this analysis; 13 were observed directly and 8 were inferred from the absence of Pn.p nuclei when the animals were observed 12–16 hr later. These inferred deaths must have occurred several hours after the L1 molt. Of these deaths, 11/21 (52%) occurred in the VPCs (P3.p–P8.p), which are 6/11 (55%) of the Pn.p cells, suggesting that these Pn.p cells are not biased toward death. Therefore, these results indicate that in pvl-5 mutants the decreased number of Pn.p cells is not due to a defect in the generation of the Pn.p cells, but is due to some Pn.p cells undergoing abnormal cell death subsequent to their birth in the L1 stage.
Lineage analysis of P and Pn.p cells in pvl-5(ga87). Lineage analysis of P1–P11 and/or P1.p–P11.p for six pvl-5(ga87) animals shows that some Pn.p cells undergo cell death at the L1/L2 molt or thereafter. A solid line ending in an X indicates the death of a cell that was directly witnessed in the course of lineage analysis. A dotted line with an X indicates that a Pn.p nucleus was missing at that position 12–16 hr after the lineage analysis was performed, and we infer that a cell death occurred. The L1/L2 molt is indicated by a horizontal line along the vertical line, which depicts developmental time.
Morphology of dying Pn.p cells in pvl-5(ga87). (A) The arrow indicates a refractile Pn.p nucleus in a pvl-5(ga87) L1 larva compared to a wild-type Pn.p nucleus (arrowhead). (B) The arrow indicates a large swollen Pn.p nucleus in pvl-5(ga87).
The pvl-5 Pn.p cell death defect is suppressed by ced-3(lf) and ced-9(gf):
The morphology of the dying cells as observed during lineage analysis suggested that the cells could be undergoing either apoptosis, as indicated by the appearance of refractile ovate corpses (Figure 3A), or necrotic cell death, as evidenced by the presence of large swollen nuclei (Figure 3B). To determine if the Pn.p cell deaths observed in pvl-5(ga87) mutants are due to inappropriate activation of the programmed cell death pathway, we built double mutants between pvl-5(ga87) and mutations affecting the core components of the PCD pathway, ced-3, ced-4, ced-9, and egl-1 (Metzstein et al. 1998). Loss-of-function mutations in ced-3, ced-4, and egl-1 or a gain-of-function mutation in ced-9, n1950 can prevent most programmed cell deaths in the worm. We found that two different ced-3 strong loss-of-function mutations (Ellis and Horvitz 1986; Shaham et al. 1999) and ced-9(n1950gf) were each able to suppress the Pn.p cell death defect in pvl-5 animals (Table 3) . In the pvl-5(ga87); ced-3(n717) double-mutant strain the average number of Pn.p nuclei at the L2 larval stage was 10.3, and in pvl-5(ga87); ced-9(n1950) it was 10.2, compared with 7.0 in pvl-5(ga87) alone (P < 0.001 for all; two-tail t-test). We also compromised ced-3 function in a pvl-5 mutant background by dsRNA interference and found that this treatment partially suppressed the Pn.p cell death defect in pvl-5(ga87) mutants (Figure 4) . Mutations in ced-3 resulted in the suppression of the vulval phenotypes seen in pvl-5(ga87) animals, suggesting that the Pn.p cells in pvl-5(ga87); ced-3 mutants were capable of adopting induced cell fates like wild-type Pn.p cells (data not shown). The gonad migration defect caused by pvl-5(ga87) is also suppressed completely by mutations in ced-3 and partially by ced-9(n1950gf) (Table 3). The ray defects seen in pvl-5(ga87) are also partially suppressed by mutations in ced-3, as pvl-5(ga87); ced-3(n717) males have an average of 8.7 ± 0.8 (N = 30) rays compared to 7.7 ± 1.6 in pvl-5(ga87) alone. The observation that mutations compromising the function of the ced-3 caspase can suppress the Pn.p cell deaths in pvl-5(ga87) indicates that the Pn.p cells are most likely undergoing ced-3-dependent programmed cell deaths when pvl-5 function is reduced or inactivated. This suggests that the normal role of pvl-5 may be to prevent ced-3-dependent cell death in the Pn.p cells, in order to ensure their survival.
The pvl-5(ga87) Pn.p cell death phenotype is suppressed by ced-3(lf) and ced-9(gf) but not by ced-4(lf) and egl-1(lf)
ced-3 RNAi, but not ced-4 RNAi, suppresses the Pn.p cell deaths in pvl-5(ga87). pvl-5(ga87) L3 or young L4 hermaphrodite larvae were fed on bacteria expressing either ced-3 or ced-4 dsRNA. The number of Pn.p nuclei in individual F1 animals at the mid-L2 stage was scored. The average number of Pn.p nuclei in pvl-5(ga87); ced-3 RNAi animals was 9.0 (N = 90), indicating partial suppression of the cell deaths. The average number of Pn.p nuclei in pvl-5(ga87); ced-4 RNAi animals is 7.3 (N = 149), indicating failure to suppress the pvl-5(ga87) cell death phenotype.
ced-4(lf) and egl-1(lf) fail to suppress the Pn.p cell death defect in pvl-5:
ced-4 encodes a protein homologous to Apaf-1 that is believed to be required for the activation of the caspase CED-3 by an induced proximity mechanism (Seshagiri and Miller 1997; Wu et al. 1997a; Yang et al. 1998). To determine if the ced-3-dependent Pn.p cell deaths seen in pvl-5(ga87) were also dependent on ced-4, we built double mutants between pvl-5(ga87) and the ced-4(n1162) loss-of-function mutation, which creates a stop codon at amino acid 40 of CED-4 (Yuan and Horvitz 1992). Interestingly, ced-4(n1162) was not able to suppress the cell death defect in pvl-5 mutants (Table 3). This is surprising given that ced-4 loss-of-function mutations suppress all known programmed cell deaths in C. elegans. To verify this result we created double-mutant strains with two other ced-4 strong loss-of-function alleles (Ellis and Horvitz 1986; Yuan and Horvitz 1990; Table 3) and also compromised ced-4 function in a pvl-5 mutant background by RNAi (Figure 4). In all of these cases we saw that reduction of ced-4 function failed to suppress the pvl-5 cell death phenotype. Therefore, in pvl-5 mutants the Pn.p cells appear to be undergoing abnormal cell death by a novel pathway that is ced-3 dependent but ced-4 independent.
egl-1 encodes a proapoptotic BH3-only protein and is the most upstream component of the core programmed cell death pathway in C. elegans (Conradt and Horvitz 1998). EGL-1 is postulated to inhibit CED-9 function, thereby allowing the activation of CED-3 (Conradt and Horvitz 1998; del Peso et al. 1998). To determine if the Pn.p cell deaths observed in pvl-5 mutants require egl-1 to initiate the PCD pathway we scored the number of Pn.p cells in a double-mutant strain containing pvl-5(ga87) and the loss-of-function allele egl-1(n1084 n3082), which is known to suppress most programmed cell deaths in C. elegans. However, egl-1(n1084 n3082) failed to suppress the Pn.p cell deaths caused by pvl-5(ga87) (Table 3). Together, these data suggest that the Pn.p cell deaths seen in pvl-5 mutants may be initiated by a novel mechanism that is independent of egl-1 and ced-4.
ced-9 and pvl-5 both function to prevent Pn.p cell death:
As mentioned above, ced-9 normally acts to prevent cells from undergoing apoptosis, since ced-9(gf) mutants have fewer cell deaths, while ced-9(lf) animals display extensive PCD leading to embryonic lethality. It has been observed that in viable ced-9(lf) larvae derived from ced-9(lf)/+ mothers the Pn.p cells survive while their sister cells, the Pn.a neurons, undergo programmed cell death (Hengartner et al. 1992). If the wild-type function of pvl-5 is to prevent the activation of the programmed cell death process in Pn.p cells, then pvl-5 could be protecting the Pn.p cells from programmed cell death in ced-9(lf) mutant animals. In addition, since not all Pn.p cells die in pvl-5(ga87) mutant animals even when that mutation is in trans to a deficiency, it is possible that pvl-5 and ced-9 might both function in the Pn.p cells to inhibit cell death. To test this idea, we built a pvl-5(ga87); ced-9(lf) double-mutant strain to determine if the Pn.p cell death phenotype was worse than in pvl-5(ga87) alone. Indeed, we found that in pvl-5(ga87); ced-9(lf) animals derived from pvl-5(ga87); ced-9(lf)/+ mothers there are fewer Pn.p cells (average of 5.0, range, 4–9) than in pvl-5(ga87) alone (Table 3). This suggests that pvl-5 and ced-9 may both act to prevent the inappropriate death of the Pn.p cells.
Mutation of both cell engulfment pathways partially suppresses Pn.p cell death in pvl-5 mutants:
There are two genetic pathways for the engulfment of corpses in C. elegans. One pathway consists of the genes ced-1, ced-6, and ced-7 and the other ced-2, ced-5, ced-10, and ced-12 (Ellis et al. 1991; Chung et al. 2000). Double mutants in which both engulfment pathways are compromised exhibit more persistent corpses than mutants in which either single pathway is affected. In addition to clearing cell corpses, phagocytosis by the surrounding cells ensures the death of the cells triggered to undergo PCD, since in the absence of engulfment some cells destined to die by the activation of CED-3 are able to survive (Hoeppner et al. 2001; Reddien et al. 2001).
We sought to determine if the Pn.p cell deaths in pvl-5(ga87) were suppressible by mutations in any of the “engulfment” genes. We saw that the average number of Pn.p cells remained approximately seven in double-mutant strains containing pvl-5(ga87) and loss-of-function mutations in ced-1, ced-5, ced-6, ced-7, and ced-10 (Table 4) . Consistent with this failure to suppress the Pn.p cell deaths, we saw no evidence of persistent Pn.p cell corpses in these double mutants. In the pvl-5(ga87); ced-2(e1752) double mutant the average number of Pn.p cells is slightly higher (8.6), suggesting ced-2 may partially suppress the pvl-5(ga87) defect (Table 4). All of the engulfment gene single-mutant strains had wild-type numbers of Pn.p nuclei (data not shown; n = 25 for all). We also built a triple-mutant strain in which both engulfment pathways were compromised in a pvl-5(ga87) mutant background. In this pvl-5(ga87); ced-1(e1735); ced-5(n1812) mutant strain, the average number of Pn.p cells increased to 10.0 per animal. The increase in nuclei appears to be due to the presence of “undead” cells in the ventral midline. It is likely that these undead cells do not divide correctly and contribute to vulval formation, since the Egl and Pvl phenotypes caused by pvl-5(ga87) are still present in the triple-mutant strain (data not shown). This is in contrast to the suppression of the vulval defects seen in pvl-5; ced-3 mutants. This suggests that promoting the persistence of the Pn.p hypodermal cells by compromising the corpse engulfment pathways is not sufficient to suppress the changes resulting from the activation of the cell death pathway.
Mutation of both cell engulfment pathways suppresses the pvl-5(ga87) Pn.p cell death phenotype
pvl-5(ga87) causes ectopic cell deaths in addition to Pn.p cell death:
The pvl-5(ga87) strain was isolated because of its defects in vulval formation. We have shown that this defect is due to the inappropriate cell death of the Pn.p cells that are required for vulval induction. We also wanted to determine if other cells in the worm exhibited inappropriate cell death in a pvl-5(ga87) mutant background. First, we looked for extra cell corpses in the heads of newly hatched L1 larvae. We found an average of 15.5 cell corpses in ced-1 mutants (Table 5) . However, we found that the average number of persistent corpses was higher (average of 19.3 ± 3.1) in ced-1; pvl-5(ga87) animals (Table 5). Although we see some corpses with an irregular morphology (Figure 5A) , most of the corpses exhibited a round, refractile morphology similar to that of other cells that normally undergo programmed cell death (Figure 5B). We believe that the irregular morphology of some of the corpses might reflect the shape of an already differentiated cell undergoing cell death. The Pn.p cell deaths observed in pvl-5(ga87) mutants were also found to be independent of ced-4. ced-4(n1162) suppresses all normally occurring cell death, resulting in almost no persistent corpses in the heads of ced-1; ced-4 L1 larvae. In contrast, there are 2–6 persistent corpses in ced-1; pvl-5(ga87); ced-4 L1 larvae (Table 5; Figure 5). However, these ectopic cell deaths were suppressible by loss of ced-3 function as ced-1(e1735); pvl-5(ga87); ced-3(n717) L1 larvae had an average of 0.3 ± 0.5 corpses in the head. Thus, the ectopic cell deaths in the head, like the Pn.p cell deaths, require the activity of the caspase ced-3 but occur independently of ced-4 function.
Ectopic corpses are observed in the heads of pvl-5(ga87) L1 larvae
Ectopic ced-4-independent cell death in the head region of pvl-5(ga87) L1 larvae. (A) Arrow indicates an atypical cell corpse in ced-1(e1735); pvl-5(ga87) unc-4(e120). (B) Arrows indicate persistent corpses in ced-1(e1735); pvl-5(ga87) unc-4(e120); ced-4(n1162) dpy-17(e164). (C) No persistent corpses in ced-1(e1735); ced-4(n1162) dpy-17(e164). All animals are L1 larvae.
Second, we examined germ cell deaths in the gonads of adult hermaphrodites. Germ cell deaths in the germline also use the evolutionarily conserved “core” cell death machinery; however, this machinery is regulated differently in the germline than in the soma (Gumienny et al. 1999). We stained the gonads of adult pvl-5 hermaphrodites with the vital dye SYTO-12 64, 76, and 88 hr posthatching, but found no significant difference in the numbers of germ cell corpses in the pvl-5(ga87) background compared to wild type (data not shown). The ventral nerve cord is another region that contains cells that are normally destined to undergo PCD. Similarly, we did not notice any extra ventral cord neuronal deaths in pvl-5 mutants, and these animals have the same number of neurons as wild-type animals (data not shown). However, as noted above, in pvl-5(ga87) males we observe missing rays and ray fusion defects that are suppressible by loss of ced-3 function, suggesting that ced-3-dependent ectopic cell deaths may also be occurring in the male tail. Therefore, in some (ventral hypodermis, larval head, and perhaps male tail) but not all tissues, small numbers of ectopic cell deaths are seen in pvl-5(ga87) mutants. These results suggest that pvl-5 may function as a negative regulator of cell death in cells other than the Pn.p cells, but it clearly does not act in all cells undergoing ced-3-dependent programmed cell death.
DISCUSSION
Here we describe further characterization of C. elegans animals carrying the mutation pvl-5(ga87). This allele was originally identified in a screen for mutants having a protruding vulva (Pvl) phenotype, and it was shown that in pvl-5(ga87) animals there are too few Pn.p cells present in the ventral midline at the time of vulval induction, leading to defects in vulval formation in many animals (Eisenmann and Kim 2000). We show that in these mutant animals the 12 Pn.p cells are born correctly in the L1 stage, but that a fraction of these cells disappear around the time of the L1 molt. Several lines of evidence indicate that these cells are dying due to inappropriate activation of the cell death machinery. First, some of the Pn.p cell deaths morphologically resemble other programmed cell deaths in C. elegans, and the disappearance of the cell nuclei occurs with comparable kinetics. Second, loss-of-function mutations in the caspase gene ced-3 partially suppress these abnormal Pn.p cell deaths, and ced-3 is required for all known programmed cell deaths in C. elegans (Ellis and Horvitz 1986). Third, the disappearance of the Pn.p cells in pvl-5(ga87) animals is also suppressed when both cell engulfment pathways are simultaneously compromised by mutation. The simplest hypothesis to explain these morphological and genetic results is that the abnormal cell deaths seen in pvl-5(ga87) mutants result from the activation of a ced-3-dependent cell death process. Thus, the normal function of pvl-5 may be to protect these cells from inappropriate activation of the cell death machinery during normal development. pvl-5 may also function in other cells in a cell death protective fashion, because we observed extra cell corpses in the heads of newly hatched pvl-5(ga87) L1 larvae and missing rays in pvl-5(ga87) male tails and these defects are suppressed in pvl-5(ga87); ced-3 animals. On the basis of the involvement of ced-3 and ced-9, two of the three core components of the PCD machinery, we believe that the abnormal deaths seen in pvl-5 mutants are apoptotic in nature; however, further ultrastructural analysis is needed to confirm the nature of these cell deaths.
In addition to ced-3, all known programmed cell deaths in the worm also required the Apaf-1 homolog ced-4 (Ellis and Horvitz 1986). Surprisingly, we found that the Pn.p cell deaths seen in pvl-5 mutants are not suppressed by loss-of-function mutations in ced-4 or by RNAi of ced-4 function. We believe this is the first example of ced-3-dependent, ced-4-independent cell death in C. elegans. If, as suggested above, pvl-5 normally functions to protect cells from inappropriate activation of the cell death pathway, this result suggests that pvl-5 may act at a step upstream of ced-3 but downstream or in parallel to ced-4. For example, the pvl-5 gene product could act to suppress basal activity of CED-3. In this model, increased basal activity of CED-3 in a pvl-5 mutant could lead to sufficient CED-3 function to cause programmed cell death in some, but not all, Pn.p cells (and some other cells), independently of the activation of CED-3 by CED-4 oligomerization. The fact that additional Pn.p cell deaths are observed in the pvl-5(ga87); ced-9(lf) double mutant is consistent with the model of pvl-5 acting downstream and in parallel to the main cell death pathway. Alternatively, cell death in the Pn.p cells could utilize a factor distinct from CED-4 for the activation of CED-3. In this model, pvl-5 would encode a negative regulator of this ced-4-like activity, along with ced-9, such that loss of pvl-5 activity would result in ced-3-dependent, ced-4-independent Pn.p cell deaths. In either case, pvl-5 is likely to represent a novel regulator of cell death in C. elegans.
In light of the models presented above, two of our conclusions regarding pvl-5(ga87) bear commenting upon. First, our interpretation of the pvl-5 epistasis data is predicated on the fact that ga87 represents a reduction- or loss-of-function mutation in the pvl-5 gene. The completely recessive nature of the ga87 mutations (unlike mutations in the lin-24 and lin-33 genes), combined with the demonstration that none of the pvl-5(ga87) mutant phenotypes are enhanced in trans to a deficiency, is consistent with ga87 causing a reduction of pvl-5 function. However, we have found it difficult to isolate additional alleles of pvl-5. We performed several F1 noncomplementation screens with ga87, utilizing both EMS (the mutagen used for the screen in which ga87 was identified) and UV/trimethylpsoralen, and after screening >70,000 F1 animals we identified only a single additional allele, de4, which has a phenotype weaker than that of ga87. For the typical C. elegans gene, additional alleles are often found at a frequency of one per 3000 mutagenized haploid genomes (Anderson 1995). This suggests that (1) pvl-5 is a small gene, (2) the pvl-5 gene is not particularly mutable using the methods we tried, (3) the phenotype of pvl-5(ga87)/Df does not reflect the phenotype of pvl-5(ga87) over a null allele for the pvl-5 locus and pvl-5(ga87)/pvl-5(null) animals may be inviable or have a phenotype different from that for which we screened (Egl and/or Pvl), or (4) ga87 and de4 are not simple reduction- or loss-of-function mutations, but may represent rare, recessive, neomorphic alleles that display unpredictable heteroallelic interactions. We currently do not know which of these explanations pertains.
Second, although the pvl-5 cell death phenotype is not suppressed by ced-4 mutations, suggesting a site of action downstream of ced-4, the phenotype is suppressed by the ced-9 gain-of-function mutation n1950. This is surprising because current models suggest that this gain-of-function mutation, which encodes a G169E substitution in the conserved BH1 domain of CED-9 (Hengartner and Horvitz 1994a), likely acts by creating a mutant CED-9 protein that continues to interact with CED-4 protein even in the presence of EGL-1 (Spector et al. 1997; Conradt and Horvitz 1998; del Peso et al. 1998). If this is the mode of action of the CED-9 protein encoded by n1950, it indicates that the Pn.p cell deaths in pvl-5(ga87) animals should be dependent on ced-4 function. This interpretation predicts that ced-4 mutations would also suppress the pvl-5 mutant phenotype, which we have not observed using either RNAi or several putative ced-4 loss-of-function mutations. If, as described above, the Pn.p cell deaths use a factor other than CED-4 for the activation of CED-3 and if this factor is inhibited by CED-9 as CED-4 is, this could explain the difference between the ced-4 and ced-9(gf) genetic suppression results. Alternatively, we note that, in addition to its function as an inhibitor of CED-4 oligomerization, CED-9 may also function directly as an inhibitor of CED-3 (Xue and Horvitz 1997). If the n1950 mutation also affects this putative function of CED-9, then perhaps the CED-9 mutant protein may inhibit CED-3 even in the absence of pvl-5 activity, which would allow ced-9(n1950) to suppress pvl-5(ga87) in a manner independent of ced-4 activity. Further molecular and genetic analysis of pvl-5 and its interaction with other known cell death components should help us resolve this genetic conundrum in the future.
Abnormal cell death of the Pn.p cells is also seen in lin-24 and lin-33 gain-of-function mutants (Ferguson and Horvitz 1985). These cell deaths are genetically and morphologically distinct from those caused by activation of the egl-1-ced-9-ced-4-ced-3 pathway. For example, the lin-24 and lin-33 Pn.p cell deaths are not suppressed by mutations in ced-3, ced-4, or ced-9, but are reported to be suppressible by mutations in the engulfment genes ced-2, ced-5, and ced-10 (Hengartner 1997). This is reasonable, since both apoptotic and necrotic corpses are rapidly cleared by engulfment by neighboring cells (Chung et al. 2000). Given these results, it has been proposed that these lin-24 and lin-33 mutations may create toxic gene products that kill the Pn.p cells in which they are expressed, suggesting that lin-24 and lin-33 may not normally function in regulation of cell death in the Pn.p cells. The pvl-5(ga87) mutation described here appears to be genetically distinct from the lin-24 and lin-33 mutations because it is recessive in nature, and the Pn.p cell deaths it causes are suppressed by mutations in ced-3 and ced-9. Therefore we think that the “toxic gene product” model is less likely to apply to the pvl-5 mutant phenotype and that the pvl-5 locus may normally function in the regulation of cell death in the Pn.p cells.
Interestingly, in the nematode Pristionchus pacificus, the Pn.p cells that are not part of the vulval equivalence group die by ced-3-dependent programmed cell death (Sommer and Sternberg 1996; Sommer et al. 1998). This cell death is suppressed in those Pn.p cells expressing the homeodomain transcription factor lin-39 (Sommer et al. 1998). This is in contrast to C. elegans, wherein the Pn.p cells that do not express lin-39 fuse with the hypodermal syncytium. This is a distinct but equally effective mechanism for removing these Pn.p cells from the equivalence group. Therefore, in another nematode species the Pn.p cells are preprogrammed to die unless protected from programmed cell death. pvl-5 could represent a factor necessary for the difference in the fate of Pn.p cells between these two species.
Acknowledgments
We thank Kellee Siegfried and Judith Kimble for kindly providing lag-2::GFP; Andy Fire for providing pPD129.36; Javier Wagmaister for insightful discussions; and Lakshmi Natarajan, Elizabeth Szyleyko, and Julie Gleason for critical reading of the manuscript. We acknowledge an anonymous reviewer for insightful comments on the manuscript. We thank Usha Thiagrajan, Shreeram Joshi, and Vani Joshi for support and encouragement. Some nematode strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the National Institutes of Health National Center for Research Resources. This work was supported by the National Science Foundation under grants IBN-9817123 and IBN-0235922.
Footnotes
Communicating editor: B. Meyer
- Received July 27, 2003.
- Accepted March 19, 2004.
- Genetics Society of America
