C. elegans EIF-3.K Promotes Programmed Cell Death through CED-3 Caspase

Programmed cell death (apoptosis) is essential for the development and homeostasis of metazoans. The central step in the execution of programmed cell death is the activation of caspases. In C. elegans, the core cell death regulators EGL-1(a BH3 domain-containing protein), CED-9 (Bcl-2), and CED-4 (Apaf-1) act in an inhibitory cascade to activate the CED-3 caspase. Here we have identified an additional component eif-3.K (eukaryotic translation initiation factor 3 subunit k) that acts upstream of ced-3 to promote programmed cell death. The loss of eif-3.K reduced cell deaths in both somatic and germ cells, whereas the overexpression of eif-3.K resulted in a slight but significant increase in cell death. Using a cell-specific promoter, we show that eif-3.K promotes cell death in a cell-autonomous manner. In addition, the loss of eif-3.K significantly suppressed cell death-induced through the overexpression of ced-4, but not ced-3, indicating a distinct requirement for eif-3.K in apoptosis. Reciprocally, a loss of ced-3 suppressed cell death induced by the overexpression of eif-3.K. These results indicate that eif-3.K requires ced-3 to promote programmed cell death and that eif-3.K acts upstream of ced-3 to promote this process. The EIF-3.K protein is ubiquitously expressed in embryos and larvae and localizes to the cytoplasm. A structure-function analysis revealed that the 61 amino acid long WH domain of EIF-3.K, potentially involved in protein-DNA/RNA interactions, is both necessary and sufficient for the cell death-promoting activity of EIF-3.K. Because human eIF3k was able to partially substitute for C. elegans eif-3.K in the promotion of cell death, this WH domain-dependent EIF-3.K-mediated cell death process has potentially been conserved throughout evolution.


Introduction
Programmed Cell death is an evolutionarily conserved cellular process that eliminates unnecessary, damaged, or harmful cells [1,2]. Inappropriate regulation of this process can lead to developmental disorders, tumorigenesis, or degenerative pathologies in C. elegans, flies, mice, or humans [3].
Molecular and genetic studies in C. elegans have led to the identification and characterization of the evolutionarily conserved genes egl-1, ced-3, ced-4, and ced-9, which constitute the core cell death pathway [4][5][6]. The proteins encoded by these genes act in an inhibitory cascade. EGL-1(a BH3-containing protein) promotes cell death by antagonizing the cell death inhibitory function of CED-9, a homolog of BCL-2 [5,7]. CED-9 inhibits cell death by antagonizing the Apaf-1-like protein CED-4, which promotes death by activating CED-3 [8,9]. CED-3 belongs to a cysteine protease family known as caspase [10]. It has been proposed that the binding of EGL-1 to CED-9 on the mitochondrial outer membrane transmits a pro-apoptotic signal that results in the CED-4-dependent activation of the cytoplasmic CED-3 caspase, thereby triggering apoptosis [8,11]. Recent structural evidence suggests that eight CED-4 molecules form a funnel-shaped structure with four-fold symmetry, with each unit being defined by an asymmetric CED-4 dimer [12]. The cavity of this octameric structure provides space for two CED-3 molecules and facilitates their autocatalytic activation. Additionally, the auto-activation of the CED-3 zymogen is negatively regulated by the CED-3 orthologs CSP-2 and CSP-3, which lack caspase activity [13,14], revealing that the regulation of CED-3 activity during programmed cell death is complex.
Additional factors that regulate the cell killing process during C. elegans development have been reported. MAC-1, an AAA family ATPase, can bind to CED-4 in vitro and prevent programmed cell death [15]. ICD-1 and TFG-1, which are similar to human bNAC and TRK-fused gene, respectively, suppress CED-4-dependent, but CED-3-independent, cell death [16,17]. In contrast to these cell-death inhibitors, WAN-1, which is a mitochondrial adenine nucleotide translocator and is associated with CED-4 and CED-9 in vitro, can induce ectopic cell death dependently on the core cell death proteins [18]. It is not clear whether additional component(s) may exist to promote the cell killing process upstream of CED-3. Moreover, some cell death effectors that act downstream of (or in parallel to) CED-3, such as CED-8 [19] and WAH-1 [20], or are CED-3 substrates, such as DCR-1 [21], are important for the timing or progression of programmed cell death.
The eukaryotic translation initiation factor 3 (eIF3) plays essential roles in the initiation of translation [22]. The mammalian eIF3 complex contains 10-13 subunits, including five highly conserved core subunits and five to eight less conserved non-core subunits [23,24]. The 28 kDa human eIF3k protein was originally identified as a non-core subunit of the eIF3 complex [25]. An in vitro reconstitution experiment showed that eIF3k is not required for the formation of the active eIF3 complex [26]. Interestingly, eIF3k is conserved among metazoans, including C. elegans, D. melanogaster, M. musculus, and H. sapiens, but is absent in S. cerevisiae, suggesting a specialized role for eif-3.K in multicellular organisms [25,27]. In addition, human eIF3k is associated with dynein [27], cyclin D3 [28], the 5-HT 7 receptor [29], and keratin K18 [30], suggesting the involvement of eIF3k in processes that are unrelated to translation. Recently, we reported an apoptosis-promoting function for eIF3k in simple epithelial cells [30]. Upon apoptotic stimuli, keratin K18 is cleaved by caspase 3, resulting in the collapse of K8/K18 intermediate filaments into apoptotic bodies and the sequestration of caspase 3 in kerain-containing inclusions [31]. eIF3k binds to keratin inclusions, which in turn leads to the release of keratin-associated caspase into the cytosol to facilitate the execution of apoptosis [30]. Keratin K8/K18 is the major intermediate filament in epithelial cells [31]. It is not clear whether eIF3k may potentiate apoptosis in other cell types, such as neurons or muscle cells, where intermediate filaments other than keratin are present. In addition, it is unclear whether the apoptosispromoting function of eIF3k has been conserved throughout evolution.
In this work, we characterized the function of eif-3.K in C. elegans and showed that its apoptosis-promoting function has indeed been conserved throughout evolution. Furthermore, we identified a new function for the WH domain of EIF-3.K in the promotion of programmed cell death.

Cell Death Assays
Cell corpse numbers in embryos or germline of indicated mutants were scored as previously described [41]. Extra surviving cells in the anterior pharynx were scored at the late L3 or early L4 larval stage, as previously described [41]. To assay extra surviving cells in the ventral cord, the integrated nIs106 (P lin-11 gfp) transgene was utilized [38]. The nIs106 transgene was crossed to ced-2 or ced-7 single mutants or ced-2; eif-3.K or ced-7; eif-3.K double mutants. The extra Pn.aap cells in the P2, P9-P12-derived regions of the transgenic mutants were scored at the L4 stage by the fluorescence microscopy as previously described [38]. The TUNEL assay was carried out using an in situ cell-death detection kit (Roche) as previously described [42]. To assay the UV-C radiation-induced cell death in the germline, adult worms (24 h post the L4 stage) were exposed to 254 nm UV-C light at150 J/m 2 using a Stratalinker UV crosslinker (Stratagene, model 2400) as previously described [43], and the cell corpses in the gonadal arms were scored 24 hours after the treatment.

Transgenic Animals
Germline transformation experiments were performed as previously described [46]. For the rescue experiment or structure function analysis of EIF-3.K, the indicated constructs (50 mg/ml) were injected into eif-3.K(gk126) animals with the coinjection marker pTG96 plasmid. The pTG96 plasmid contains sur-5::GFP that is expressed in almost all somatic cells [47].

Heat Shock Experiments
To overexpress the wild-type or mutant eif-3.K cDNA or human eIF3k cDNA, young adults carrying the respective transgene were allowed to lay eggs overnight, and the laid embryos were cultured at 20uC (non-heat shock) or at 33uC (heat shock) for 1 hr, which was followed by a 20uC recovery for at least 1.5 hrs. The embryos were scored for cell corpses at the comma and 1.5-fold stages under DIC optics.
For immunostaining, embryos and worms were collected off plates and treated with hypochlorite (10 N NaOH and NaOCl) to enrich embryos. Embryos were then washed with ddH 2 O for three times and fixed in fixation buffer (2% paraformaldehyde, 90% methanol, 10% EGTA, 1 M spermine, 100 mM spermidine, and 0.5 M PIPES) overnight at -80uC as described by Guenther and Garriga [49]. After fixation, embryos were thawed, washed with Tris-Triton buffer (100 mM Tris-HCl pH7.4, 1% Triton X-100, and 1 mM EDTA) and blocked with 5% bovine serum albumin in PBS. Treated embryos were incubated with purified antibodies against EIF-3.K overnight at 4uC. After washing with wash buffer (1X PBS, 1% BSA, 0.5% Triton X-100, 0.05% NaN 3 , and 1 mM EDTA), embryos were then incubated with rhodamine-conjugated donkey secondary antibodies against rabbit (Jackson Immune Research Laboratories). After incubation for 2 hr at room temperature, antibodies were washed off using wash buffer three times for 5 min each, with DAPI included in the first wash. For MitoTrackerH staining, embryos were collected from worms grown in the dark on NGM agar plates containing MitoTrackerH Red 580 (1 mg/mL, Molecular Probes). Stained embryos were mounted with VECTASHIELDH mounting medium H-1000 (Vector Laboratories) and observed using confocal laser scanning microscopy (Leica TCS SP2 Confocal Spectral Microscope).
For western blot analysis, total protein extracts of indicated genotypes were resolved by 12% SDS-PAGE and transferred to nitrocellulose membranes. The blot was incubated with affinitypurified EIF-3.K antibodies (1:2500) and monoclonal antia tubulin antibodies (Abcam). ECL detection system (pierce) was used for detection.

Bacteria-mediated RNAi
Induction RNA interference (RNAi) experiments were carried out using a bacterial feeding protocol [50]. L4 larvae were transferred to the control (pPD129.36) or indicated RNAi plates and cultured at 20uC. F1 embryos laid approximately 48 hours later were picked for phenotypic analysis. The eif-3.K RNAi clone was obtained from the Ahringer RNAi library.

Results
The Loss of eif-3.K Causes Reduced Cell Death in Both Somatic and Germline Cells We obtained the cDNA for eif-3.K through reverse transcription polymerase chain reaction (RT-PCR) using total RNA from mixed-stage worms. The predicted full length amino acid sequence of EIF-3.K is 35% identical and 57% similar to that of human eIF3k ( Figure 1A). We next characterized the eif-3.K mutant allele gk126, which was isolated by the C. elegans Gene Knockout Consortium. This allele contains a 538 base pair (bp) long deletion from 119 bp upstream of the start ATG codon to the second exon in the eif-3.K locus ( Figure 1B). No EIF-3.K protein was detected in the eif-3.K(gk126) mutant by western blotting or immunostaining analyses using purified anti-EIF-3.K antibodies (Figure 2), suggesting that the gk126 allele is null. Because eif-3.K(gk126) and eif-3.K(RNAi) mutant worms were viable and had normal development and growth rates (Table S1), we concluded that eif-3.K is not an essential component of the general translation machinery in C. elegans.
We next examined whether eif-3.K(RNAi) or eif-3.K(gk126) embryos have defective programmed cell death. A time course analysis of embryonic cell corpses using differential interference contrast (DIC) microscopy showed that eif-3.K(RNAi) or eif-3.K(gk126) embryos had fewer cell corpses than wild-type embryos throughout embryogenesis ( Figure 3A). To determine whether this decrease in cell corpse number corresponded with a reduction in cell death or was simply due to abnormal corpse morphology, we further analyzed the embryos using the TUNEL (Terminal deoxynucleotidyl transferase dUTP nick end labeling) assay. The degradation of DNA in dying cells is a hallmark of apoptosis and can be detected in situ using TUNEL staining [42,51]. As shown previously [42], wild-type embryos had very few TUNEL-positive corpses ( Figure 3B); however, embryos lacking the nuc-1 gene, which codes for a protein similar to DNAse II that is involved in DNA degradation [42], had many more TUNEL-positive corpses ( Figure 3B). The eif-3.K embryos, like the wild-type embryos, had few TUNEL-positive corpses; however, eif-3.K; nuc-1 double mutant embryos had fewer TUNEL-positive corpses than nuc-1 single mutant embryos, indicating that apoptotic.
DNA degradation is compromised in the eif-3.K mutants. This result, together with the observed decrease in cell corpse number ( Figure 3A), indicates that apoptosis is compromised in eif-3.K mutants during embryogenesis.
Like somatic cells, germline cells also undergo apoptosis in C. elegans [52]. Because few germ cell corpses can be observed in the wild-type adult gonad at any given time due to the prompt removal of cell corpses by the gonadal sheath cells [52], we utilized ced-1(e1735) mutant worms, in which cell corpses are not efficiently removed and therefore accumulate, to increase our chances of detecting cell corpses. We found that ced-1(e1735); eif-3.K(gk126) double mutants had significantly fewer germ cell corpses than ced-1(e1735) single mutants at all adult stages ( Figure 3C). Therefore, eif-3.K is also important for programmed cell death in germline cells.

The Loss of eif-3.K Enhances Cell Survival in Sensitized Mutants
We next examined whether a loss of eif-3.K function could prevent cell death and result in an accumulation of surviving cells. Two assays were used to score the surviving cells in various regions of the animal [53]. First, superfluous surviving cells that were present in the anterior pharynx were scored using DIC optics. As was previously shown [54], in the presence of a strong loss-offunction mutation in the pro-apoptotic gene ced-3(n717), which blocks nearly all cell deaths, resulted in approximately 10 additional surviving cells in the anterior pharynx ( Figure 3D). Animals harboring the weak ced-3(n2427) mutation had only 1.2 additional surviving cells ( Figure 3D). We found that the eif-3.K(RNAi) or eif-3.K(gk126) single mutant animals had 0.2 or 0.5 extra surviving cells, similar to the wild-type animals ( Figure 3D), indicating that the loss of eif-3.K could not detectably block apoptosis in these cells; however, the eif-3.K(RNAi) or eif-3.K(gk126) mutation did enhance cell survival in the weak ced-3(n2427) mutant animals. The ced-3(n2427); eif-3.K(RNAi) or ced-3(n2427); eif-3.K(gk126) double mutants had 2.7 or 2.1 additional surviving cells in the anterior pharynx ( Figure 3D). This is significantly more than ced-3(n2427), eif-3.K(RNAi) or eif-3.K(gk126) single mutants.
Moreover, the RNAi-mediated inactivation of eif-3.K also significantly enhanced cell survival in the worms lacking the ced-8 gene ( Figure 3D), which controls the timing of programmed cell death [19]. These results show that the loss of eif-3.K enhances cell survival in sensitized mutants.
We further analyzed the identities of the surviving cells in these mutants. The extraneous surviving cells observed in ced-3(n2427) single mutants and ced-3(n2427); eif-3.K(gk126) double mutants appeared similar and included sisters of muscle cells m1and m2 and neurons I1, I2, and MC ( Figure S1). It is possible that these cells are more likely to survive than others when the apoptotic machinery is compromised. Consistent with this hypothesis, m1 and m2 sister cells were occasionally observed to survive in the wild-type or eif-3.K(gk126) animals ( Figure S1).We also compared the identities of surviving cells in the ced-3(n2427) animals that were treated with either the eif-3.K or control RNAi. Compared to the control RNAi, eif-3.K RNAi enhanced the survival of the niece of the epithelial cell e1, the sister of the neuron I1 and those cells that were also enhanced by the eif-3.K(gk126) mutation in the ced-3(n2427) single mutants, including sisters of m1, m2, I1, I2, and MC cells ( Figure S1). Because the eif-3.K null allele did not enhance the total number of extra surviving cells in the strong ced-3(n717) mutants ( Figure 3D), eif-3.K likely functions with ced-3 in the same genetic pathway to promote most, if not all, programmed cell death. Additionally, because the identities of apoptotic cells can be inferred from the cell fates of their differentiated sister cells [55], our observations suggest that eif-3.K exerts a cell deathpromoting function in multiple cell types, including neuron, muscle and epithelial cells during development.
Secondly, we scored superfluous surviving cells in the ventral cord in larvae. In contrast to the extra surviving cells we observed in the anterior pharynx, which are generated during embryogenesis [56], extra surviving cells in the ventral cord are generated during larval development [57]. In strong ced-3(n717) mutants, five cells P2.aap and P9-P12.aap in the ventral cord survive [38]. These Pn.aap cells are known to differentiate into VC motor neuron-like cells and express the P lin-11 gfp reporter (Table 1) [38]. We scored extra surviving Pn.aap cells using the P lin-11 gfp transgene as a marker and found that only 2% of eif-3.K(gk126) worms exhibited extra Pn.aap cells (Table 1). However, the eif-3.K(gk126) mutation increased the average number of extra surviving Pn.aap cells in ced-3(n2427) mutants from 2.6 to 3.6 ( Table 1). A previous study showed that strong mutations in genes essential for the removal of apoptotic cells, such as ced-2 or ced-7, block cell death, albeit at low efficiency, as apoptotic cell removal is involved in the death of cells [38,58]. The frequency of extra Pn.aap cell survival in these mutants can be enhanced by a weak mutation in the core programmed cell death genes ced-3, ced-4, or egl-1 [38]. Therefore, we tested whether the loss of eif-3.K enhanced the frequency of superfluous Pn.aap cell survival in strong ced-2 or ced-7 mutants. We found that the eif-3.K(gk126) mutation increased the frequency of Pn.aap cell survival in ced-2 (n1994) or ced-7(n1996) mutants from 27% to 100% and 83% to 93%, respectively (Table 1). In addition, the average number of extra surviving Pn.aap cells in ced-2(n1994) or ced-7(n1996) mutants also increased from 0.3 to 2.4 and 1.6 to 2.1, respectively (Table 1). These observations support the idea that the cell death machinery is compromised in the eif-3.K mutants. The observed decreased cell death in eif-3.K mutants ( Figures 3A-3C) as well as the enhanced cell survival observed in the anterior pharynx ( Figure 3D) and ventral cord (Table 1) of sensitized mutants shows that eif-3.K is a positive mediator of programmed cell death.
The Loss of eif-3.K Partially Suppresses the Ectopic Cell Deaths Induced by the Overexpression of egl-1 or ced-4 We next tested whether eif-3.K genetically interacts with the core programmed cell death genes ced-3, ced-4, and egl-1. Previous studies have shown that cell-specific expression of these three genes under the control of the P mec-7 promoter, which is expressed in six touch neurons (AVM, ALMR/L, PVM, and PLMR/L), promotes these neurons to undergo programmed cell death [5,35]. We tested if the programmed cell death of these cells required the activity of eif-3.K. To facilitate the scoring of touch neurons, the P mec-4 gfp reporter in the integrated transgene bzIs8, which labels the six touch neurons with GFP, was used as a cell viability marker [39]. We found that the eif-3.K(gk126) mutation partially suppressed the apoptosis of the touch neurons that was induced by the overexpression of egl-1 or ced-4 ( Figures 4A and 4B). For example, the overexpression of egl-1 or ced-4 resulted in the death of approximately 50% or 29% of the PVM neurons, respectively. The loss of eif-3.K reduced cell death to 24% and 3%, respectively. A significant reduction in cell death was also observed in other touch neurons, except for PLMR/L with egl-1 overexpression ( Figure 4A and 4B). Therefore, the efficient apoptosis of touch neurons induced by the overexpression of egl-1 or ced-4 requires eif-3.K. This result showed that eif-3.K functions downstream of or in parallel to egl-1 or ced-4 to promote cell death.
In contrast, eif-3.K(gk126) failed to suppress the apoptosis of touch neurons induced by ced-3 overexpression. The overexpression of ced-3 resulted in the death of approximately 34% of the PVM neurons in the wild-type animals, similar to the percentage (40%) of cell death observed in the eif-3.K(gk126) mutants ( Figure 4C). When activated CED-3 (acCED-3) was expressed in touch neurons via the P mec-7 acCED-3 transgene [40], 40% and 43% of the PVM neurons were killed in the wild-type and eif-3.K(gk126) worms, respectively ( Figure 4D). This result showed that eif-3.K also fails to inhibit apoptosis caused by the overexpression of activated CED-3. Similar results were observed in other touch neurons expressing either the P mec-7 ced-3 or the P mec-7 acCED-3 transgene (Figures 4C and 4D).
We next examined if overexpression of ced-3 using the heat shock promoter P hsp was able to rescue the cell death defects caused by the eif-3.K(gk126) mutation. Heat shock-induced ced-3 overexpression rescued the defect at the comma and 1.5-fold stages, but it also slightly elicited ectopic cell killing at the comma stage ( Table 2). The eif-3.K(gk126) embryos carrying the P hsp ced-3 transgene had approximately 7.0 cell corpses at the comma stage under non-heat shock condition. The heat shock-induced overexpression of ced-3 in the transgenic embryos at the same developmental stage increased the cell corpse number to 10.7, which was significantly more than the 8.4 cell corpses that were observed in the wild-type embryos carrying the control P hsp gfp transgene under the same conditions (Table 2). This result supports the model that ced-3 acts downstream of eif-3.K to execute programmed cell death. To determine whether eif-3.K promotes programmed cell death, we tested if the overexpression of eif-3.K caused cells that would normally live to undergo programmed cell death by overexpressing eif-3.K under the control of the heat shock promoter P hsp in the wild-type animals. The overexpression of eif-3.K, but not of the control gfp, slightly but significantly increased the cell corpse number at the 1.5-fold stage, despite a lack of significant ectopic killing at the comma stage (Table 2). Nonetheless, this ectopic killing at the 1.5-fold stage supports for a cell death-promoting function for eif-3.K. In addition, this ectopic killing was significantly suppressed by the strong ced-3(n717) or ced-4(n1162) mutations (Table 2). This result, in combination with the reciprocal experiment in which the loss of eif-3.K suppressed the efficient apoptosis of touch neurons in the presence of ced-4 The inactivation of eif-3.K by RNAi or by genetic deletion reduced cell corpse numbers throughout embryogenesis. Cell corpses of the indicated genotype or RNAi treatment were scored at the comma, 1.5-fold, 2fold, 3-fold and 4-fold embryonic stages. The eif-3.K(gk126) embryos were compared to the wild-type embryos, and the eif-3.K(RNAi) embryos were compared to control(RNAi) embryos at each stage. All comparisons were performed using the unpaired t test (*P,0.05, **P,0.001). Data are presented as the mean 6 standard deviation. Error bars represent S.D. Greater than 20 embryos per stage were analyzed. (B) The loss of eif-3.K reduces TUNEL staining in nuc-1 embryos. The number of cells exhibiting TUNEL staining was determined in embryos of the indicated genotypes at the 1.5-fold stage. The eif-3.K(gk126) embryos were compared to the wild-type embryos, and the nuc-1(e1392)single mutants were compared to the eif-3.K(gk126); nuc-1(e1392) double mutants. Comparisons were performed using the unpaired t test (*P,0.05, **P,0.001). Data are presented as the mean 6 standard deviation. Error bars represent S.D. Greater than 20 embryos of each genotype were analyzed. n.s. indicates no significant difference. (C) A loss of eif-3.K reduces cell corpse number in the germline. Cell corpses in the germline of the ced-1 single mutants (white columns) and eif-3.K; ced-1 double mutants (black columns) were counted at the indicated times after entry into the adulthood. The y axis represents the average number of cell corpses scored in each gonadal arm. The eif-3.K(gk126); ced-1(e1735) double mutants were compared to the ced-1(e1735) single mutants at the same developmental stage using the unpaired t test (*P,0.05, **P,0.001). Data are presented as the mean 6 standard deviation for .20 gonadal arms. Error bars represent S.D. (D) A loss of eif-3.K increases the number of extra surviving cells in weak ced-3 mutants. Cells that failed to undergo programmed cell deaths in the anterior pharynx were scored in the indicated animals. All comparisons were performed using the unpaired t test (*P,0.05, **P,0.001). Data are presented as the mean 6 standard deviation for .20 larvae. Error bars represent S.D. n.s. indicates no significant difference. doi:10.1371/journal.pone.0036584.g003 overexpression (Figure 4 B), suggests a mutual requirement for eif-3.K and ced-4. Additionally, because the loss of eif-3.K failed to suppress the efficient apoptosis of touch neurons in the presence of ced-3 overexpression, our result suggests a unidirectional requirement of eif-3.K for ced-3 to achieve effective ectopic cell death under overexpression conditions.

eif-3.K Promotes Cell Death in a Cell-Autonomous Fashion
To determine whether eif-3.K promotes programmed cell death in a cell-autonomous fashion, we tested if the P mec-7 eif-3.K transgene, in which eif-3.K is expressed under the P mec-7 promotor in touch neurons, could trigger touch neuron apoptosis in the wildtype animals. Although the overexpression of eif-3.K transgene resulted in a low frequency of individual touch neuron apoptosis, approximately 19.2% of trasngenic worms had at least one missing touch neuron ( Figure 4E). In contrast, only 1.9% of the wild-type animals carrying the control P mec-7 gfp transgene had one missing touch neuron ( Figure 4E). This result not only reinforced the cell death-promoting function for eif-3.K but also showed that eif-3.K executes this function in a cell-autonomous fashion.

The Loss of eif-3.K Significantly Reduces Ectopic Cell Deaths in icd-1 Mutants
The inactivation of icd-1 (inhibitor of cell death-1) by RNAi results in ectopic cell death that can be blocked by the loss of ced-4 but not ced-3, revealing that the cell death in icd-1(RNAi) embryos is ced-4-dependent but ced-3-indpendent [17]. The observation that eif-3.K is required for cell death induced by the overexpression of ced-4 but not ced-3 ( Figures 4B-4D) prompted us to test whether eif-3.K could suppress ectopic cell death resulting from the loss of icd-1. Like the icd-1(RNAi) embryos described previously [17], icd-1(tm2873) embryos had additional cell corpses compared to the wild-type embryos at the comma and 1.5-fold stages ( Figure 5A). Although the eif-3.K(gk126) mutation significantly reduced the cell corpse number in icd-1(tm2873) embryos at both stages, the cell corpse number was not reduced to the extent observed in the eif-3.K(gk126) mutants alone ( Figure 5B). Instead, the cell corpse number for the double mutant was between that observed in the eif-3.K(gk126) and icd-1(tm2873) single mutants. This result is consistent with the model that eif-3.K acts in parallel with icd-1 to promote cell death; however, we cannot rule out the possibility that icd-1 may in part prevent programmed cell death in an eif-3.Kdependent manner.

EIF-3.K is Widely Expressed throughout Embryogenesis and Localized to the Cytoplasm
To determine the localization pattern of EIF-3.K, we raised antibodies against a recombinant EIF-3.K protein (see Experimental Procedures). Using affinity-purified EIF-3.K antibodies and western blot analysis, we detected a band of apparent molecular mass 27 kDa from wild-type worm extracts by western blot analysis (Figure 2A). This protein was absent in extracts from the eif-3.K(gk126) mutants (Figure 2A), confirming that the 27 kDa protein is the product of the eif-3.K gene. We used the purified EIF-3.K antibodies to stain embryos and larvae. EIF-3.K was widely expressed in embryos and larvae and was localized to the cytoplasm (Figures 2B-2D). EIF-3.K did not appear to be associated with mitochondria, where several cell death regulators such as CED-9, WAH-1, and WAN-1 are located [7,18,20], because EIF-3.K did not co-localize with MitoTracker Red, a marker of mitochondria ( Figure S2).
The WH Domain of EIF-3.K is Necessary and Sufficient for its Cell Death-Promoting Activity EIF3.K contains two distinct domains, the HAM (HEAT Analogous Repeats) and WH (Winged Helix) domains, which have been implicated in protein-protein and protein-RNA interactions, respectively [59]. To test the importance of these domains for EIF-3.K function, we deleted the region corresponding to the HAM or WH domains, respectively (eif-3.KDHAM or eif-3.KDWH constructs). We then tested the ability of the mutant construct to rescue the cell death defects in eif-3.K(gk126) embryos by expressing the mutant construct under the control of the ubiquitous let-858 promoter P let-858 [44]. To our surprise, the HAM domain, comprising more than one-third of the EIF-3.K protein, was dispensable for eif-3.K activity, as P let-858 eif-3.KDHAM completely rescued the cell death defect in the eif-3.K(gk126) embryos (Table 3). In contrast, P let-858 eif-3.KDWH failed to rescue the defect (Table 3), suggesting an essential role for the WH domain in the cell death-promoting function of eif-3.K. Because the expression level of the P let-858 eif-3.K transgene was lower than that of the endogenous eif-3.K, as detected by western blotting or immunostaining analysis (data not shown), the stronger heat shock Average numbers of fluorescent cells caused by expression from P lin-11 gfp in P2, 9, 10, 11, and 12-derived regions were determined using DIC microscopy equipped with an ultraviolet light source. Greater than 20 larvae of each genotype were analyzed. b The percentages of animals that had at least one fluorescent cell in the P2, 9, 10, 11, and 12-derived regions were determined. doi:10.1371/journal.pone.0036584.t001 promoter P hsp was subsequently used to increase the expression of the mutant eif-3.K construct in an effort to confirm our results. The heat shock-induced expression of the wild-type and mutant eif-3.K genes resulted in slightly higher protein expression levels ( Figure  S3). The overexpressed proteins exhibited a similar localization as the endogenous EIF-3.K protein, suggesting that these proteins localize normally ( Figure S3). Similar to the results obtained using the P let-858 promoter, expression under the heat shock promoters revealed that eif-3.KDHAM, but not eif-3.KDWH, rescued the eif-3.K mutant phenotype at the comma and 1.5-fold stages (Table 3). The percentage of animals missing specific touch neurons are shown for the wild-type (white columns) or eif-3.K(gk126) (black columns) embryos carrying the P mec-7 egl-1 (A), P mec-7 ced-4 (B), P mec-7 ced-3 (C), or P mec-7 acCED-3 (D) transgenes. The eif-3.K(gk126) transgenic worms were compared to the analogous wild-type transgenic worms. Comparisons were performed using the unpaired t test (*P,0.05, **P,0.001). Data are presented as the mean 6 standard deviation. Error bars represent S.D. n.s. indicates no significant difference. More than 100 animals were scored for each strain.
(A)The percentage of animals missing specific touch neurons or missing at least one touch neuron are shown for wild-type control P mec-7 gfp transgenic animals (white columns) or P mec-7 eif-3.K transgenic animals (black columns). More than 100 animals were scored for each strain. doi:10.1371/journal.pone.0036584.g004 These data show that the WH domain, but not the HAM domain, is necessary for the cell death-promoting function of EIF-3.K. We next tested if the WH domain of EIF-3.K is sufficient to rescue the cell-death defect caused by the eif-3.K mutation. We expressed the WH domain alone using either the P let-858 or P hsp promoters in the transgenes P let-858 WH or P hsp WH, respectively. We found that either promoter rescued the cell death defect in the eif-3.K mutants (Table 3). Moreover, when the P hsp WH transgene was expressed in the wild-type animals, it induced ectopic cell deaths. Furthermore, superfluous cell corpses were observed at the comma and 1.5-fold stages (Table 3). Therefore, the WH domain is both necessary and sufficient for the cell death-promoting activity of eif-3.K.
Human eIF3k can Partially Substitute for C. elegans EIF-3.K The human eIF3k mediates apoptosis in simple epithelial cells, likely by binding to keratin K18 via its HAM domain [30]; however, the HAM domain of C. elegans EIF-3.K appears dispensable for its function in cell death. We tested whether the expression of human eIF3k by P hsp was able to rescue the cell death defect caused by the eif-3.K mutation. We found that human eIF3k partially rescued the defective apoptosis in the eif-3.K(gk126) mutants (Table 3). This result indicates that the pro-apoptotic function of EIF-3.K has been conserved through evolution from C. elegans to humans and that the mechanisms by which human eIF3k and C. elegans EIF-3.K promote apoptosis may, in part, be similar.
In C. elegans, the loss of eif-3.K caused reduced programmed cell death ( Figures 3A-3C) and enhanced cell survival in sensitized mutants ( Figure 3D and Table 1). In contrast, the overexpression of eif-3.K by the heat shock promoter or a touch neuron-specific promoter resulted in ectopic cell death (Table 2 and Figure 4E). These results demonstrate that eif-3.K promotes programmed cell death. Our results also show that eif-3.K is essential for the efficient cell death that is induced by the overexpression of egl-1 or ced-4, but not ced-3, as the loss of eif-3.K partially suppresses the cell death that is induced by the overexpression of egl-1 or ced-4 only ( Figures 3A-3D). In addition, the observation that ced-3 overexpression can rescue the cell death-defective phenotype of eif-3.K mutants and that the ced-3 strong mutation can suppress cell death caused by heat shock-induced eif-3.K overexpression ( Table 2) further reinforces the notion that eif-3.K requires ced-3 to promote programmed cell death. Furthermore, the wide range in the identity and type of extraneous surviving cells that are affected by the eif-3.K mutation ( Figure S1) suggests that eif-3.K may be involved in the majority of programmed cell death. This is consistent with the ubiquitous expression of EIF-3.K in embryos and larvae (Figures 2C and 2D). In addition to physiological cell deaths, DNA damage caused by genotoxic stress such as UV or IR radiation also induces cell deaths in the germline [43,64]. We found that the eif-3.K(gk126) mutation significantly reduced UVinduced cell deaths in the germline ( Figure S5), indicating that eif-3.K also mediates DNA damage-induced cell death.
During C. elegans development, the activity of the executioner caspase CED-3 is under both positive and negative regulation. Previous studies have shown that CED-4 facilitates the autocleavage of pro-CED-3 to generate the active CED-3 caspase during the promotion of cell death [12,65,66], while the CED-3 paralogs CSP-2 and CSP-3 associate with the CED-3 zymogen and inhibits its auto-activation, thereby protecting cells from inappropriate apoptosis [13,14]. Our observation that neither EIF-3.K nor its WH domain bind to CED-3 or CED-4 in a yeast 2hybrid system ( Figure S4) suggests that EIF-3.K may not promote cell death through a direct association with either protein. In addition, since CED-3 and CED-4 are the only known proteins involved in CED-3 activation from pro-CED-3 [9,12], EIF-3.K likely does not affect this activation process directly. It is possible that EIF-3.K may promote programmed cell death after CED-4induced CED-3 activation. Human eIF3k has been proposed to promote apoptosis by facilitating the release of active caspases from an inhibitory compartment of intermediate filament-containing inclusions into the cytosol, thereby allowing the released caspase better access to its cytosolic substrates [30]. Although the mechanism by which eIF3k may affect the release of caspases from intermediate filament-containing inclusions is not clear, the binding of eIF3k to intermediate filaments is known to be important for the release process [30]. Similarly, C. elegans EIF-3.K may promote programmed cell death by affecting the distribution of active CED-3, thus facilitating the substrate cleavage and the subsequent execution of cell death. Alternatively, EIF-3.K might promote programmed cell death in parallel with CED-4 by antagonizing CSP-2 or CSP-3, thus facilitating CED-3 autoactivation from the zymogen in germline or somatic cells, respectively [13,14]. To test the latter possibility, we used bzIs8 (P mec-4 gfp ), which labels six touch neurons, as marker to monitor the survival of touch neurons and tested the effect of the eif-3.K mutation on the missing cell phenotype of the csp-3(lf) animals. As previously shown [14], in csp-3(lf) animals six touch neurons were lost randomly at a frequency from 2% to 10% (Table S2) and 24% of animals lost at least one touch neuron (Table S2). The eif-3.K(gk126) mutation strongly suppressed this missing cell defect in csp-3 mutants (Table S2). In addition, loss of csp-2 resulted in increased germline cell deaths [13], and this phenotype can also be suppressed by the eif-3.K mutation ( Figure S6). These results suggest that EIF-3.K may promote cell death downstream of or in parallel to csp-2 or csp-3 in the germline and somatic cell deaths, respectively. Human eIF3k co-localizes with keratin and requires keratin for its apoptosis-promoting function in simple epithelial cells [30]. Upon apoptotic stimuli, keratin K18 is cleaved by caspase 3 at VEVD 238 of the L1-2 linker region or DALD 397 of the C terminal (tail) domain, resulting in a collapse of keratin filaments [67]. C. elegans contains eleven genes that encode cytoplasmic intermediate filaments, including ifa-1, mua-6, ifa-3, ifa-4, ifb-1, ifb-2, ifc-1, ifc-2, ifd-1, ifd-2, and ifp-1 [68]. No detectable change in either the localization or the level of the EIF-3.K protein was observed by immunostaining analysis in mua-6(rh85) or ifb-1(ju71) mutant embryos or in embryos treated with ifa-1, mua-6, ifa-4, ifb-1, ifc-1, or ifd-1 interfering RNAs (data not shown). Potential CED-3 cleavage sites (DXXD) are found in IFA-1 (DAED), IFC-2 (DNRD), IFD-1(DNRD and DVDD), and IFP-1 (DSVD). The RNAi-mediated inactivation of ifa-1, but not ifc-2, ifd-1, or ifp-1, reduced the number of cell corpses at the comma stage. For example, the ifa-1(RNAi) worms had on average 6.961.8 cell corpses, which is similar to the number of cell corpses observed in eif-3.K mutants; however, whether IFA-1, IFC-2, IFD-1, or IFP-1 are direct targets of the CED-3 caspase or are involved in programmed cell death needs to be evaluated. It is not yet clear whether EIF-3.K localizes to intermediate filaments or mediates programmed cell death through intermediate filaments in C. elegans, similar to human eIF3k. Previously, the pro-apoptotic function of human eIF3k was identified and assayed in simple epithelial cells [30] in which keratin K8/K18 is the major intermediate filament. It will be interesting to determine whether human eIF3k, like C. elegans eIF3.K, can promote apoptosis in muscle or neuron cells, as human eIF3k is widely expressed in many tissues, including the brain and muscle [27], where no or very little keratin is expressed [69]. In addition, because eIF3k is also present in D. melanogaster, an organism that lacks intermediate filaments, it will be interesting to see if eIF3k plays a role in apoptosis in D. melanogaster as well.
C. elegans EIF-3.K and human eIF3k both contain two conserved domains, the WH and HAM domains. The HAM domain, but not the WH domain, of human eIF3k interacts with keratin 18 in a yeast two-hybrid system and thus may be important for eIF3k localization to keratin [30]; however, in C. elegans, the HAM domain is dispensable for the cell death-promoting function of EIF-3.K and the WH domain alone is sufficient to promote cell death (Table 3). This result suggests that the WH domain may promote programmed cell death by an IF-independent mechanism. The WH domain has been implicated in DNA or RNA binding [59], but how it may promote programmed cell death needs further study. The result that human eIF3k can partially rescue the cell death defect in the eif-3.K mutants suggests that the eIF3k family may promote apoptosis through a conserved mechanism, which may be dependent upon the WH domain. Figure S1 The identification of extraneous surviving cells in the mutants. The y axis represents the percentage of animals with specific superfluous surviving cells (x axis). The extra surviving cells are named after their sister or niece cells, such as ''e1 sister cell'' and ''I2 niece cell''. M4, MC, NSM, I1 and I2 are neurons. e1 is an epithelial cell, and m1 and m2 are muscle cells. L: left, R: right. The identities of extraneous surviving cells were determined as previously described [41]. More than 20 worms for each genotype were scored. (TIF) Figure S2 EIF-3.K is not associated with mitochondria. A wild-type embryo was co-stained with anti-EIF-3.K antibodies (A) and MitoTracker (B). The merged image is shown in C. Scale bar = 10 mm. (TIF) Figure S3 Deletion of the WH domain does not affect the expression pattern or stability of EIF-3.K. The wild-type embryo (A) and eif-3.K mutant embryo (B) with no transgene, and the eif-3.K mutant embryos carrying the transgene P hsp eif-3.K (C), P hsp eif-3. KDHAM (D), or P hsp eif-3. KDWH (E) were heat shocked and co-stained with anti-EIF-3.K antibodies (red) and DAPI (blue). Representative images of anti-EIF-3.K antibody staining (upper panel) and merged images of anti-EIF-3.K antibody and DAPI staining (lower panel) are shown. Scale bar = 10 mm. (TIF) Figure S4 Neither EIF-3.K nor the WH domain alone interacts with CED-3 or CED-4 in a yeast 2-hybrid assay. Pairs of constructs expressing the indicated fusion proteins were transformed into the yeast strain MaV203. The resulting transformants were streaked on SC-Trp-Leu-His or SC-Trp-Leu plates containing 30 mM 3 AT. Growth on the SC-Trp-Leu-His+30 mM 3 AT plate indicates an interaction between the fusion proteins. The E2F1and RB pair was used as positive control [70]. ''-'' in the lower panel indicates no insert was present in the AD fusion construct. (TIF) Figure S5 Loss of eif-3.K reduced DNA damage-induced apoptosis. Apoptotic germ cell corpses were scored in the wildtype (black columns) and eif-3.K(gk126) (white columns) young adult worms 24 hr following exposure to 150 J/m 2 UV-C radiation. The eif-3.K(gk126) mutants were compared to the Transgenic animals were subjected to heat-shock (+) or left at 20uC (2). b Transgenic embryos were scored for the number of cell corpses 1.5 hrs after heat shock. (see Materials and methods). Data are presented as the mean 6 standard deviation from two independent stably transmitting lines. Greater than 20 embryos were analyzed from each line. For the P let-858 expressing transgene, eif-3.K mutant embryos carrying the transgene were compared to eif-3.K mutant without the transgene. For the P hsp expressing transgene, the transgenic embryos after heat shock were compared to the corresponding transgenic embryos without heat shock. All comparisons were performed using the unpaired t test (*P,0.05, **P,0.001). doi:10.1371/journal.pone.0036584.t003

Supporting Information
wild-type using the unpaired t test (**P,0.001). More than 20 gonadal arms were scored for each genotype.