Introduction

Apoptosis is a physiological mechanism by which unneeded, harmful, or damaged cells commit suicide. It plays an important role in development and homeostasis, and its mis-regulation has been associated with various pathologies.1, 2, 3 Although we understand the molecular events occurring during apoptosis relatively well, less is known about the regulatory signals that trigger this cell fate.

Here we describe a systematic approach to identify new genes that regulate germ cell apoptosis in Caenorhabditis elegans. RNA interference (RNAi) was used to knockdown expression of 16 757 worm genes to find antiapoptotic factors that prevent excess germ cell death. Knockdown of 21 genes reproducibly resulted in a moderate to strong increase in germ cell apoptosis in wild-type animals. By epistasis analysis, we show that a large subset of our RNAi candidates require p53 and a functional DNA damage response pathway to activate germ cell apoptosis. Furthermore, we characterize mutations in two genes identified in our screen, pmk-3 and bmk-1, and show that these mutants indeed have more germ cell deaths than wild-type worms. The identification of new C. elegans germline apoptosis (gla) genes, many of which have mammalian homologues, should shed light on the different pathways that regulate apoptosis during metazoan development.

Results

C. elegans hermaphrodites possess two U-shaped gonads connected at a common uterus (Figure 1a). Each gonad has a distal-to-proximal polarity, the most distal part being capped by the somatic distal tip cell (DTC). The DTC expresses a growth factor, LAG-2, which promotes the mitotic proliferation of the adjacent germ cells. In the distal arms, germ cell nuclei are not completely enclosed by membrane, and thus form a large syncytium. Nevertheless, because each nucleus acts independently of its neighbour, we will refer to these syncytial nuclei as germ cells. Once germ cells escape the influence of the DTC, they enter meiosis (transition zone in Figure 1a) and progress into the pachytene stage of meiosis I; exit from pachytene requires the activation of the RAS/MAPK pathway.4, 5 Germ cells can then further differentiate into oocytes, or die by apoptosis. Germ cell apoptosis depends on the caspase CED-3 and the Apaf-1 homologue CED-4, and is blocked by the Bcl-2-like protein CED-9.6 Animals with a conditional loss-of-function (lf) mutation in ced-9 have five times more apoptotic corpses in their germ line than wild-type worms.6 This result can readily be phenocopied by inactivating ced-9 with RNAi (Figure 1b).7

Figure 1
figure 1

RNAi and acridine orange can be combined for a high-throughput, genome-wide screen for apoptotic germ cell death. (a) Schematic diagram of an adult hermaphrodite C. elegans gonad. Apoptotic cells are normally observed at and distal to the bend. (b) Knockdown of ced-9 by RNAi can phenocopy a conditional loss-of-function mutation in ced-9. For each genotype, three independent experiments are shown; the checked, black, and white bars represent mean±S.D. for each individual experiment (n=20 gonads), whereas the dashed bars represent mean±S.D. of the three experiments. See Materials and methods for a description of the statistics used in this study. (c) Left column shows representative pictures of a gonad after RNAi feeding observed by differential interference contrast (DIC) microscopy. Apoptotic germ cells can be seen as highly refractile disks (arrowheads). Right column shows that acridine orange (AO) stains specifically apoptotic corpses in the germ line. For all the pictures, anterior is to the left and dorsal to the top. Scale bar, 8 μm

Full size image

To identify other antiapoptotic genes, we inactivated 86% of the ∼19 500 predicted C. elegans protein-coding genes using a RNAi feeding library.8 To visualize apoptotic germ cells by fluorescent microscopy, the vital dye acridine orange (AO) was used. Ingestion of AO-stained Escherichia coli by C. elegans specifically labels engulfed apoptotic corpses in the germ line (Figure 1c).6 Using this assay combined with the RNAi library, we identified 21 gene inactivations that cause a Gla phenotype, that is, more germ cell apoptosis than observed in wild-type worms (Table 1). Sequence analysis of these candidates revealed that knockdown of a large variety of genes, belonging to several functional classes, induce germ cell apoptosis (Table 1; sequences can be retrieved at Wormbase; www.wormbase.org). Many of the genes identified in our screen encode proteins with known antiapoptotic functions in other systems, thereby validating our approach. For example, as has previously been shown,9, 10 we found that knocking down RAD50 or the RecA-like protein RAD51, two proteins involved in homologous recombination and DNA double-strand break repair, results in a Gla phenotype (Table 1). Similarly, mice carrying a mutation in either Rad50 or Rad51 show increased apoptosis.11, 12 Moreover, it has recently been reported that inactivation of the p53 inhibitor iASPP, and its worm counterpart ape-1, induces apoptosis in a p53-dependent manner.13 We independently identified ape-1 in our RNAi screen, and confirmed that its inactivation requires p53 to cause a Gla phenotype (Table 1 and see below).

Table 1 Gene candidates that increase germline apoptosis following RNAi
Full size table

Why do so many different gene inactivations cause germ cell death? In an effort to address this issue, we performed epistasis analysis with our RNAi candidates. First, we showed that in all cases a strong lf mutation in the caspase ced-3 completely suppresses the Gla phenotype, thus confirming the apoptotic nature of the cell deaths detected by AO and differential interference contrast (DIC) (Table 1). Apoptosis in the hermaphrodite C. elegans germ line can be triggered by genotoxic treatment, such as γ-irradiation.10 It has been shown that these DNA damage-induced germ cell deaths are dependent on the worm p53 homologue cep-114, 15 and the checkpoint gene hus-1.16 To test if the gene candidates found in our screen activate the DNA damage response pathway, their expression was also inactivated by RNAi in cep-1 and hus-1 mutants. The basal level of germ cell death in these mutant backgrounds is lower than in wild-type worms (in Table 1, compare control(RNAi) for each mutant). The reason for this difference is unknown; one possibility is that in the absence of HUS-1 or CEP-1, endogenous DNA damage (e.g. aberrant meiotic intermediates) remains undetected and therefore cannot cause apoptosis. We found that cep-1(lf) and hus-1(lf) completely suppress the Gla phenotype of all but five genes, suggesting that p53 is a key sensor of germ cell stresses in C. elegans and that many of our RNAi candidates indirectly affect germ cell apoptosis through the activation of quality control checkpoint pathways (Table 1).16, 17 For the five gene knockdowns that are not suppressed by cep-1 or hus-1, a reduction in the number of apoptotic germ cells was observed in comparison to wild-type worms (Table 1, p53-independent class). As already mentioned, these mutant strains have initially less germ cell deaths than the wild-type strain; this could explain the differences observed. Epistasis analysis with ced-9(gf), a mutation that completely blocks cep-1-dependent germ cell death, but does not affect physiological germline apoptosis,6, 10 confirmed the above results (Table 1).

As RNAi does not always faithfully reproduce the known lf phenotypes, we sought to confirm our RNAi results with genetic mutants. As a first step towards this aim, we obtained and characterized deletion alleles for two of our RNAi candidates, pmk-3 and bmk-1. Neither mutant has any obvious growth defects (data not shown),18 but both pmk-3(lf) and bmk-1(lf) mutants displayed an increased number of germ cell corpses when compared with wild-type worms (Figure 2a). PMK-3 is one of the three p38 MAPK isoforms in C. elegans, which are all present on the same operon.18 The worm p38 PMK-1 has been implicated in stress responses and innate immunity, whereas the role of PMK-2 and PMK-3 is still unclear.18, 19, 20 Contrary to the pmk-3(lf) Gla phenotype, knockdown of either pmk-1 or pmk-2 does not increase germ cell death (Figure 2b), suggesting a specific involvement of the PMK-3 isoform of p38 in the protection against germ cell apoptosis. bmk-1 encodes a C. elegans homologue of the BimC kinesin-like motor protein involved in spindle formation.21 Double-mutant analysis revealed that ced-3 is epistatic to both pmk-3 and bmk-1 (Figure 2b). In contrast, cep-1(lf), hus-1(lf), and ced-9(gf) abrogated bmk-1(lf)-induced germ cell death, but not the Gla phenotype of pmk-3(lf) worms (Figure 2b). These results corroborate our RNAi data (Table 1), and imply that loss of bmk-1 activates p53-dependent germ cell death, whereas pmk-3 acts in a p53-independent germ cell apoptotic pathway.

Figure 2
figure 2

pmk-3(ok169) and bmk-1(ok391) mutants exhibit a Gla phenotype. (a) Time-course of germ cell death in pmk-3(ok169) and bmk-1(ok391) mutants. Worms were synchronized and scored for germ cell apoptosis by DIC every 12 h post-L4 stage. Data shown represent mean±S.D. of three experiments (n=20 gonads for each experiment). (b) ced-3(lf) prevents germ cell death in pmk-3 and bmk-1 mutants, whereas cep-1(lf), hus-1(lf), and ced-9(gf) block only bmk-1(lf)-induced germ cell death. Worms were scored 24 h post-L4 stage

Full size image

Discussion

In this paper, we describe the results of a genome-wide RNAi screen for genes that affect apoptosis in the C. elegans adult hermaphrodite germ line. We identified 21 genes that reproducibly induce germ cell death when knocked down by RNAi, confirmed that the additional germ cell deaths were apoptotic in nature, and used various cell death mutants to determine which death-inducing pathways were activated in each case.

Interestingly, knockdown of most of the genes identified in our screen appears to cause germ cell death indirectly, through activation of p53-dependent checkpoint or quality control mechanisms. The C. elegans p53 protein CEP-1 has been shown to participate in a conserved signalling pathway in response to DNA damage,16 suggesting that at least some of these gla genes will be involved in the maintenance of genome stability. Consistent with this hypothesis, the p53-dependent class contains genes such as rad-50 and rad-51, which are required for DNA repair and meiotic recombination, as well as genes predicted to protect from reactive oxygen species, such as gst-5.

We also identified in our screen ape-1, the worm orthologue of the p53-binding protein iASPP. Mammalian iASPP is the most phylogenetically conserved inhibitor of p53 identified so far.13 It has been proposed that the primary function of iASPP is to inhibit the proapoptotic activity of p53 in the absence of genotoxic stress. As previously reported,13 we found that ape-1(RNAi) induces germ cell apoptosis in a p53-dependent manner, consistent with the hypothesis that C. elegans iASPP APE-1 might also act as a p53 inhibitor by binding and regulating CEP-1's pro-apoptotic function. However, we found that ape-1(RNAi) resulted in cep-1-dependent apoptosis only in the presence of a functional upstream DNA damage response pathway (Table 1). This indicates that releasing CEP-1 from APE-1 inhibition is not sufficient by itself to trigger p53-dependent germ cell death. It is possible that effective activation of the apoptotic response might require not only release from APE-1, but also modification of CEP-1 by checkpoint regulators of the DNA damage response pathway. Further genetic and biochemical analysis is clearly required in order to understand the role of C. elegans iASPP in regulating p53-mediated apoptosis.

In mammals, p53 has been shown to mediate apoptosis in response not only to DNA damage, but also other stresses, such as oncogene activation and hypoxia.22 In contrast, we have been unable so far to identify a DNA damage-independent death pathway for CEP-1 in C. elegans: all the cep-1-dependent RNAi candidates that we report here also showed a significantly reduced Gla phenotype in animals mutant for the gene hus-1 (Table 1), which acts early in the DNA damage response pathway. We suspect that the ability of p53 to respond to oncogene activation might be a more recent evolution than its ability to respond to DNA damage, as tumour development is not much of a threat for simple, short-lived invertebrates such as C. elegans. Alternatively, our screen might simply have been unable to generate stresses that activate other p53-dependent pathways.

Our screen led to the identification of a small number of genes that, upon knockdown, induce germ cell apoptosis in a p53-independent manner. One of these, ced-9, encodes the C. elegans homologue of mammalian Bcl-2.23 CED-9 acts downstream of p53 in the regulation of germ cell apoptosis,16 readily explaining the observed epistasis. The mechanism of action of the other four genes – two predicted RNA-binding proteins, a p38 MAPK, and the C. elegans homologue of the S. cerevisiae Ring-finger protein Bre1p– is more obscure. We have previously shown that about half of all potential oocytes undergo apoptosis in the adult C. elegans germ line, and have suggested that many of these deaths serve a homeostatic function: to eliminate nuclei that served transiently as nurse cells and thereby reduce the nuclear/cytoplasmic ratio of the mature syncytial gonad.6 It is tempting to suggest that these four genes function in the regulation of this physiological germ cell death process. Alternatively, knockdown of these four genes might simply activate another, p53-independent quality control pathway that results in germ cell apoptosis.

Did our screen identify all the genes that can induce germ cell death upon inactivation? Most certainly not. First, the RNAi collection we used contained only about 85% of all predicted C. elegans genes. Thus, about a seventh of the C. elegans genome was not tested in our screen. Second, because of the intrinsic variability and limited efficiency of RNAi,24 we likely missed a significant number of positive genes in our primary and secondary screens due to false negative results. Third, the particular feeding protocol used in our screen (feeding of Po L1 larvae, screening of Po adults) would have precluded us from identifying any genes that are required not only for germ cell survival but also for larval growth, as well as genes that need to be fed for more than one generation for the RNAi phenotype to become evident.

Indeed, we recovered in our screen only a fraction of the genes previously described to affect germ cell apoptosis. For example, we identified only two of the more than 12 DNA damage response genes previously shown to induce germ cell apoptosis upon RNAi-mediated knockdown.9, 10, 25, 26 However, many of these genes needed to be fed for several generations at high dsRNA concentrations for the apoptosis phenotype to become evident. We also failed to identify in our screen daz-1 and the ste13/ME31B/RCK/p54 homologue cgh-1; both genes have been shown to induce germ cell apoptosis through as yet unknown mechanisms upon inactivation.27, 28 We did however identify in our screen CPB-3, one of four C. elegans CPEB family members. In clam and Xenopus oocytes, CPEB has been shown to interact functionally and physically with p47/p54, the orthologue of C. elegans CGH-1.29 Whether CPB-3 and CGH-1 also interact in C. elegans, and whether they cooperate to control germ cell apoptosis, possibly by regulating the translation of mRNA targets, remains to be determined.

Previous genome-wide RNAi screens in C. elegans have been used to determine new gene functions8, 24 as well as to identify genes involved in body fat regulation30 and genome stability.31, 32 Our results demonstrate that systematic RNAi screens can also successfully be used to identify genes that affect apoptosis. Indeed, based on our RNAi data, we identified two new genetic mutants that have excess germ cell death, confirming our reverse genomic approach. Further analysis of the apoptotic genes identified in our screen will provide useful genetic entry points into the study of mechanisms that control germ cell survival, genomic stability and of the genetic networks that regulate p53 activation.

Materials and Methods

Strains

C. elegans was cultivated using standard methods.33 All strains were grown at 20°C. The strains used in this study were the following: wild-type N2 Bristol, hus-1(op244), cep-1(gk138), ced-9(n1653), ced-9(n1950gf), pmk-3(ok169), ced-3(n717), bmk-1(ok391). ok169 and ok391 were backcrossed to the wild-type strain three times before phenotypical analysis. The deletion alleles were detected by PCR using the following primer sequences: for ok169 5′-CCCATTTTTCACTGCGTCTCAATCG-3′ and 5′-TCTGCTTCTCCAGGGATTAACGGTG-3′, and for ok391 5′-ATTTGCTGCGAACCTTGACT-3′ and 5′-GCCGCGAATCATTGTATTTC-3′.

RNAi germ cell apoptosis screen

We carried out RNAi as described.34 In total, 30–50 synchronized L1 worms were placed on NGM agarose plates seeded with E. coli producing double-stranded RNA (dsRNA). Worms were grown for 3 days, then stained for 1 h in the dark by adding 500 μl of M9 buffer containing AO (Molecular Probes; 0.02 mg/ml) to the plate. Worms were then transferred to fresh agar plates, and allowed to destain for 1 h in the dark. AO staining was assessed by fluorescent microscopy. RNAi candidates that induced a Gla phenotype were tested again in duplicate. The presence of increased germ cell corpses was confirmed by Nomarski optics 24 h post-L4 stage. For all clones that produced a Gla phenotype, plasmids were isolated and the DNA sequence encoding the dsRNA was sequenced to confirm the identity of the inactivated gene.

Statistical analysis

In Figure 1b, the checked, black, and white bars represent mean±S.D. for each experiment (n=20 gonads), to illustrate variation within each experiment. The dashed bar represent mean ±S.D. for the three experiments, to illustrate variation amongst experiments. This latter statistical representation is also used in Table 1 and Figure 2. Genes were considered to cause increased cell death only if they reproducibly caused significant increases in germ cell apoptosis (cutoff for inclusion in Table 1 was P<0.0001 versus wild-type). In Table 1, the cutoff value to establish epistatic groups was chosen as P<0.001 (RNAi versus respective mutant control). Although the P-value for cep-1(gk138); pmk-3(RNAi) (P=0.003) is above the cutoff value, we decided to group pmk-3 into the p53-independent class based on the clear phenotype of the double mutant (Figure 2b).