Mutational Analysis of the Caenorhabditis elegans Cell-Death Gene ced-3

Corresponding author: H. Robert Horvitz, Howard Hughes Medical Institute, Department of Biology, Room 68-425, Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA 02139.

Communicating editor: R. K. HERMAN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mutations in the gene ced-3, which encodes a protease similar to interleukin-1ß converting enzyme and related proteins termed caspases, prevent programmed cell death in the nematode Caenorhabditis elegans. We used site-directed mutagenesis to demonstrate that both the presumptive active-site cysteine of the CED-3 protease and the aspartate residues at sites of processing of the CED-3 proprotein are required for programmed cell death in vivo. We characterized the phenotypes caused by and the molecular lesions of 52 ced-3 alleles. These alleles can be ordered in a graded phenotypic series. Of the 30 amino acid sites altered by ced-3 missense mutations, 29 are conserved with at least one other caspase, suggesting that these residues define sites important for the functions of all caspases. Animals homozygous for the ced-3(n2452) allele, which is deleted for the region of the ced-3 gene that encodes the protease domain, seemed to be incompletely blocked in programmed cell death, suggesting that some programmed cell death can occur independently of CED-3 protease activity.

THE gene ced-3 functions cell-autonomously to promote programmed cell death in the nematode Caenorhabditis elegans (ELLIS and HORVITZ 1986 ; YUAN and HORVITZ 1990 ; SHAHAM and HORVITZ 1996A ). ced-3 appears to act downstream of both ced-4S, a positive regulator of cell death (ELLIS and HORVITZ 1986 ; SHAHAM and HORVITZ 1996A , SHAHAM and HORVITZ 1996B ) that encodes a protein similar to the human cell-death protein Apaf-1 (ZOU et al. 1997 ), and ced-9, a negative regulator of cell death and a member of the bcl-2 gene family (HENGARTNER et al. 1992 ; SHAHAM and HORVITZ 1996A ).

ced-3 encodes a member of the CED-3/ICE (interleukin-1ß converting enzyme) family of cysteine proteases (YUAN et al. 1993 ; XUE et al. 1996 ). These proteases have been named caspases, for cysteine aspases, because they have active-site cysteines and cleave after aspartate residues (ALNEMRI et al. 1996 ). Caspase genes encode precursor proteins that are activated by cleavage at specific aspartate residues. Such cleavage generates C-terminal and central polypeptides that associate to form a heterodimeric protease and an N-terminal polypeptide not present in the active protease and called the prodomain (THORNBERRY et al. 1992 ; FAUCHEU et al. 1995 ; XUE et al. 1996 ). X-ray crystallographic studies of caspase-1 (ICE) suggest that the active protease consists of two interacting heterodimers (WALKER et al. 1994 ; WILSON et al. 1994 ). A similar X-ray structure has been determined for caspase-3 (CPP32; ROTONDA et al. 1996 ). All caspases examined cleave after particular aspartate residues (THORNBERRY et al. 1992 ; XUE et al. 1996 ), a substrate specificity shared with only one other known eukaryotic protease, granzyme B/fragmentin 2, which is thought to function in cell death mediated by cytotoxic T cells (SHI et al. 1992A , SHI et al. 1992B ; HEUSEL et al. 1994 ).

Several observations indicate roles for mammalian and fly caspases in apoptosis. First, some caspases are activated during apoptosis. For example, caspase-8 (MACH/FLICE) binds FADD, a Fas-associated protein, and is activated in cells undergoing apoptosis following Fas stimulation (BOLDIN et al. 1996 ; MUZIO et al. 1996 ). Second, overexpression of mammalian caspases can result in cell death in culture (MIURA et al. 1993 ; FERNANDES-ALNEMRI et al. 1994 , FERNANDES-ALNEMRI et al. 1995A , FERNANDES-ALNEMRI et al. 1995B ; KUMAR et al. 1994 ; WANG et al. 1994 ; FAUCHEU et al. 1995 ; MUNDAY et al. 1995 ; TEWARI et al. 1995 ), and overexpression of two Drosophila melanogaster caspases, DCP-1 and drICE, can cause cell death both in flies and in cultured cells (FRASER and EVAN 1997 ; SONG et al. 1997 ). Third, inhibition of caspase activity can prevent apoptosis. The caspase inhibitor p35 can prevent programmed cell death in C. elegans (XUE and HORVITZ 1995 ) and in D. melanogaster (HAY et al. 1994 ). Similar effects can be seen in mammalian cells. For example, expression in neurons of the viral caspase-1 inhibitor protein crmA (RAY et al. 1992 ) prevents the deaths of these neurons following trophic factor deprivation (GAGLIARDINI et al. 1994 ). Caspase-1 inhibitors also prevent the in vivo deaths of chick embryo neurons that die when unable to innervate their target muscles (MILLIGAN et al. 1995 ). In addition, disruption of the caspase-1 gene in mice results in a defect in Fas-mediated cell death (KUIDA et al. 1995 ), suggesting that caspase-1 might normally have a role in programmed cell death. As another example, caspase-3 is likely to be a key component of cytoplasmic extracts that trigger morphological changes similar to those observed during physiological cell death in isolated nuclei (NICHOLSON et al. 1995 ), and inhibitors of caspase-3 prevent Fas-induced cell death (ENARI et al. 1996 ). Disruption of the caspase-3 gene as well as disruption of the caspase-3 activator caspase-9 in mice results in a significant reduction in cell death in the brain, indicating that both genes are normally important for mediating cell death (KUIDA et al. 1996 , KUIDA et al. 1998 ; HAKEM et al. 1998 ). These results suggest that caspases are involved in cell death not only in C. elegans but in vertebrates and insects as well.

To define those regions of caspase proproteins required for protease activation and/or enzymatic activity and to understand better the function of caspases in programmed cell death, we have analyzed the effects of mutations in ced-3 on programmed cell death in vivo.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

General methods and strains:
The techniques used for culturing C. elegans were as described by BRENNER 1974 . All strains were grown at 20°. The wild-type strain used was C. elegans variety Bristol strain N2. Genetic nomenclature follows the standard C. elegans system (HORVITZ et al. 1979 ). The mutations used were described previously by BRENNER 1974 , TRENT et al. 1983 , HEDGECOCK et al. 1983 , ELLIS and HORVITZ 1986 , CLARK et al. 1988 , ELLIS et al. 1991 , HENGARTNER et al. 1992 , and YUAN et al. 1993 , or were isolated by us and members of our laboratory. These mutations are listed below:

• Linkage Group (LG) I: sem-4(n1378), ced-1(e1735)

• LGIII: ced-9(n1950 n2077, n1950 n2161), ced-11(n2744), ced-5(n2098)

• LGIV: ced-3(n717, n718, n1040, n1129, n1163, n1164, n1165, n1286, n1949, n2424, n2425, n2426, n2427, n2429, n2430, n2432, n2433, n2436, n2438, n2439, n2440, n2442, n2443, n2444, n2445, n2446, n2447, n2449, n2452, n2454, n2719, n2720, n2721, n2722, n2830, n2854, n2859, n2861, n2870, n2871, n2877, n2883, n2885, n2888, n2889, n2921, n2922, n2923, n2924, n2998, n3001, n3002), dpy-4(e1166), sDf21

• LGV: egl-1(n487).

Isolation and characterization of ced-3 mutants:
We isolated the ced-3 alleles n2859, n2861, n2870, n2877, n2883, n2885, n2888, n2889, n2921, n2922, n2923, n2924, n3001, and n3002 as suppressors of the maternal-effect lethality caused by the massive ectopic programmed cell death of embryos homozygous for the loss-of-function allele ced-9(n1950 n2161) (HENGARTNER et al. 1992 ). Specifically, unc-69(e587) ced-9(n1950 n2161)/qC1 animals were mutagenized using ethyl methanesulfonate (EMS) and allowed to produce self-progeny (BRENNER 1974 ). Unc-69 F₁ animals were then placed 10 to a plate, and F₂ animals that grew to adulthood were picked and used to establish a suppressed strain. The presence of a ced-3 mutation in the strain was confirmed by a complementation test using the ced-3(n717) allele, followed by mapping to establish linkage to unc-30(e191) on chromosome IV. The ced-3 alleles n2719, n2720, n2721, n2722, n2830, and n2998 were isolated by Gillian Stanfield (personal communication) in our laboratory as suppressors of phenotypes of persistent corpses or abnormal corpse morphology of ced-5 or ced-11 mutants, respectively. ced-3 mutations suppress the ced-5 and ced-11 defects by preventing programmed cell death and thus not allowing corpse formation. The ced-3 alleles n2424, n2429, n2432, n2436, n2439, n2440, n2442, n2443, n2444, n2445, n2446, n2447, n2449, n2452, n2454, n2854, and n2871 were isolated by Michael Hengartner (personal communication) in our laboratory as suppressors of the maternal-effect lethality of animals homozygous for ced-9 loss-of-function mutations using a protocol similar to that described above. The ced-3 alleles n717, n718, n1040, n1129, n1163, n1164, n1165, n1286, n1949, n2426, n2430, and n2433 were described previously (ELLIS and HORVITZ 1986 ; YUAN et al. 1993 ) as were the ced-3 alleles n2425, n2427, and n2438 (HENGARTNER and HORVITZ 1994B ; see RESULTS). These alleles were isolated in four separate screens as suppressors of mutations that block the engulfment of dead cells (see above), as suppressors of the maternal-effect lethality conferred by ced-9 loss-of-function mutations (see above), as mutations that prevent programmed cell deaths, or as mutations that fail to complement ced-3(n717) for suppression of the egg-laying defect of egl-1(n487) animals. The egl-1(n487) mutation causes the HSN neurons required for egg laying to undergo programmed cell death. These ectopic deaths are suppressed by mutations in ced-3 (TRENT et al. 1983 ; ELLIS and HORVITZ 1986 ). All 52 ced-3 alleles described in this study were identified as ced-3 alleles on the basis of complementation tests and linkage to chromosome IV.

To quantitate cell survival in ced-3 mutants we scored for the presence of extra cells in the anterior region of the pharynx (HENGARTNER et al. 1992 ). We observed one extra cell in ~5% of wild-type animals. Except for those alleles noted in Table 3, all ced-3 mutants we analyzed were backcrossed at least twice to wild-type N2 animals to remove any non-ced-3 mutations that might be present in the strain.

Determination of allele sequences:
To characterize coding regions and exon/intron junctions from mutant strains, we amplified the ced-3 genomic coding region using the polymerase chain reaction (PCR) and a set of four primer pairs. Specifically, primers SHA2 and PCR2 were used to amplify exons 1–3, primers PCR3 and PCR4 were used to amplify exon 4, primers PCR5 and 650 were used to amplify exons 5–7, and primers BD1 and 1200 were used to amplify exon 8. The sequences of these primers are shown in Table 1. DNA was amplified as follows. One to 10 worms were placed in 3 µl PCR lysis buffer (60 µg/ml proteinase K in 10 mM Tris pH 8.2, 50 mM KCl, 2.5 mM MgCl₂, 0.45% Tween 20, and 0.05% gelatin) and frozen at -70° for 20–30 min. Samples were allowed to incubate at 60° for 1 hr followed by a 15-min incubation at 95°. The entirety of each sample was then used as the DNA source in a standard PCR reaction using one of the primer pairs described above. Samples were run on a 1.4% agarose gel, purified using ß-agarase (New England Biolabs, Beverly, MA), and resuspended in 20 µl of TE buffer. Sample sequences were determined using the fmol sequencing kit (Promega, Madison, WI), following instructions of the manufacturer for ³³P-labeling and using the primers listed in Table 1, except for primers PWR.30, PWR.32, and PWR.40. Samples were run on standard polyacrylamide sequencing gels (Life Technologies, Gaithersburg, MD). Gels were dried and exposed to X-ray film for 1–5 days. For each allele we determined the entire sequence of the ced-3 open reading frame as well as of all exon/intron junctions. Sequences of sites at which a potential mutation was identified were redetermined for both strands.

DNA flanking the ced-3(n2452) deletion site was isolated using the CLONTECH (Palo Alto, CA) Advantage cDNA PCR kit, using primers PWR.30 and PWR.32, and following the instructions of the manufacturer. Sequences of the resulting DNA were determined using primer PWR.40 and an ABI sequencer (Applied Biosystems, Foster City, CA).

Southern hybridization and RT-PCR experiments:
Southern analysis of ced-3(n2452) and wild-type genomic DNA was performed as described by SAMBROOK et al. 1989 using the restriction enzymes EcoRV, HindIII, XhoI, and XbaI (New England Biolabs, Beverly, MA) and a full-length ced-3 cDNA as a probe. RNA for reverse transcriptase PCR (RT-PCR) was prepared as follows. Worms grown on one or two 9-cm plates were added to a liquid culture containing S medium and antibiotics as described by SULSTON and HODGKIN 1988 . Frozen bacteria were added to the culture as a food source. Cultures were harvested after 5–7 days, and mRNA was prepared using the FastTrack mRNA preparation kit (Invitrogen, San Diego). RT-PCR was performed using the RNA GeneAmp kit (Perkin Elmer-Cetus, Norwalk, CT). The resulting bands were purified as described in the previous section. The sequences of the ced-3(n2440), ced-3(n717), and ced-3(n2854) products were determined using an ABI sequencer (see below).

Plasmid constructions:
Construct A was made by partially digesting the pJ40 plasmid containing ced-3 genomic sequences (YUAN et al. 1993 ) with BglII (to delete sequences 3' of the BglII site in the ced-3 coding region) and self-ligating. Construct B was made by digesting pJ40 with MluI and ApaI, filling in overhangs with the Klenow enzyme, and ligating the vector-containing fragment to the lacZ moiety of pPD22.04 (FIRE et al. 1990 ) digested with BamHI and ApaI, and filled in with the Klenow enzyme. Construct C was made by digesting pJ40 with the enzymes BglII and ApaI and ligating the vector-containing fragment to the lacZ moiety of pPD22.04, which had been cut using the enzymes BamHI and ApaI. Construct D was made by digesting pJ40 with the enzymes SalI and ApaI, and ligating the vector-containing fragment to the lacZ moiety of vector pPD22.04, which had been cut using the enzymes SalI and ApaI. Construct E was made by digesting the heat-shock vector pPD49.78 (MELLO and FIRE 1995 ) with NheI, digesting the ced-3 cDNA plasmid pS126 (SHAHAM and HORVITZ 1996A ) using SpeI and then partially digesting it using BglII, and digesting pPD22.04 with BamHI and SpeI, followed by ligation of the three components. Construct F was made in the identical manner as construct E, except that the smaller pS126 fragment was used. Construct G was made in the identical manner as construct C, except that the green fluorescent protein (GFP) vector Tu#62 (CHALFIE et al. 1994 ; M. CHALFIE, personal communication) was used instead of the lacZ vector. Construct H was produced in the identical manner as construct E, except that the the GFP vector Tu#62 was used instead of the lacZ vector. Constructs in Table 2 were made by mutating pS126 to introduce the appropriate changes using an in vitro mutagenesis kit (Amersham, Arlington Heights, IL) and following the instructions of the manufacturer. The mutant ced-3 cDNAs were digested with SpeI and SmaI and ligated to the P_mec-7-containing vector pPD52.102 (MELLO and FIRE 1995 ) digested with NheI and EcoRV.

Germline transformation:
Our procedure for microinjection and germline transformation followed that of FIRE 1986 and MELLO et al. 1991 . DNA for injections was purified using the QIAGEN system for DNA purification (QIAGEN, Inc., Chatsworth, CA) according to the instructions of the manufacturer. The concentrations of all plasmids used for injections were between 50 and 100 µg/ml. All constructs were coinjected with the pRF4 plasmid containing the rol-6(su1006) gene as a dominant marker. Animals carrying the pRF4 plasmid exhibit a Rol phenotype. All transformation experiments were into wild-type or ced-9(n2812); ced-3(n717) animals. Approximately 30 animals were injected in each experiment, and ~50–100 F₁ Rol animals were picked onto separate plates. F₁ animals segregating Rol animals were established as lines containing extrachromosomal arrays (WAY and CHALFIE 1988 ).

Splicing mutants of ced-3:
We examined in more detail three of the four ced-3 alleles (n2854, n717, n2440, and n3002) that are likely to affect splicing. The allele n2854 contains the sequence AGGCG|gattt in the donor region of intron 5 of ced-3 (Table 3) instead of the AGGCG|gttcg present in the wild type. To characterize the ced-3 transcripts made in animals carrying this ced-3 mutation, we isolated RNA from mutant animals (see above), prepared cDNAs from the RNA, and amplified this DNA using PCR and ced-3-specific primers. The sequence of the resulting band was then determined. Interestingly, the only product isolated from the ced-3(n2854) mutant was spliced at a position upstream of the normal splice site to give a deletion of 3 bp with respect to the wild-type message, resulting in the deletion of glycine 360 in the open reading frame of ced-3. Why this splicing pattern occurred is not understood. The ced-3(n717) mutation changes a conserved acceptor site G to an A in intron 7. To characterize the product(s) made in n717 animals, we isolated RNA from mutants and used this RNA for a Northern blot probed with a ced-3 cDNA. The size and level of the message were not discernibly different from those of the wild-type message (data not shown). We then prepared cDNAs from the ced-3(n717) RNA and amplified this DNA using PCR and ced-3-specific primers. Sequence determination of the resulting bands suggested that splicing occurred at positions -1, -2, -3, 0, +1, +2, and +3 (-, upstream; +, downstream) of the wild-type splice site (data not shown). The mutation in ced-3(n2440) changes the sequence CCGCAAGTT to CCGTAAGTT, altering codon 401 from a glutamine to an ochre stop codon. However, we noticed that this change also creates a potential splice-donor site (CC|gtaagtt), which might be used instead of the intron 6 splice donor immediately downstream of the mutation site. To determine if this splice-donor site is used, we determined the sequence of ced-3 cDNAs prepared from ced-3(n2440) mutant RNAs (see above). Only one class of RNAs was discernible and used the predicted new donor site. The product produced by this splice is out of frame and is predicted to form a truncated protein with 13 amino acids downstream of amino acid 400. Thus, the mutation in ced-3(n2440) is likely not to be a nonsense mutation.

ced-3 reporter constructs and expression patterns:
The construction of reporter transgenes is described above. All lacZ and GFP reporter transgenes we examined were expressed in many cells throughout the animal primarily during embryogenesis, starting at about the 200-cell stage, and during the first larval period (L1; data not shown). Very weak expression was seen in a small number of cells after the L1 stage (data not shown). Expression was detected both in cells that normally die and in those that normally live (data not shown), consistent with previous experiments suggesting that ced-3 activity is present both in cells that do and in cells that do not die (SHAHAM and HORVITZ 1996A ). Because the expression patterns we observed were somewhat variable from strain to strain, perhaps because the expression constructs were present on unstable extrachromosomal arrays, we did not pursue a detailed characterization of these patterns.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Cys-358, asp-221, and asp-374 are important for CED-3-induced cell death:
To determine whether CED-3 protease activity is required for programmed cell death, we used site-directed mutagenesis to generate mutant ced-3 cDNAs that should lack either CED-3 protease activity or the CED-3 precursor cleavage sites. We expressed these cDNAs in the ALM neurons using the promoter of the gene mec-7 (P_mec-7; SAVAGE et al. 1989 ); such overexpression of a wild-type ced-3 cDNA results in the programmed deaths of these cells (SHAHAM and HORVITZ 1996A ). In transgenic animals containing extrachromosomal arrays of wild-type P_mec-7ced-3 constructs, ~50% of ALMs die (SHAHAM and HORVITZ 1996A ). By contrast, in animals containing mutant transgenes in which the presumptive active site cysteine 358 (YUAN et al. 1993 ; SHAHAM and HORVITZ 1996A ; XUE et al. 1996 ) was altered to either alanine (C358A) or serine (C358S), nearly all ALMs survived (Table 2; Figure 1). Similarly, in animals containing transgenes with D221E or D374A mutations, which alter sites of CED-3 proprotein processing (XUE et al. 1996 ), nearly all ALMs survived. On the other hand, ALM killing seemed normal in animals containing D131A or D371A mutated transgenes, which alter aspartate residues at which the CED-3 proprotein is not processed in vitro. These results suggest that both proprotein processing and protease activity of CED-3 are required for programmed cell death.

View larger version (8K): In this window In a new window Download PPT slide   Figure 1. Schematic diagram of the CED-3 protein. Boxes indicate regions of the CED-3 protein. Vertical lines separating boxes indicate sites at which the CED-3 proprotein is processed in vitro. Specific amino acid residues are indicated by a letter either above or below the boxes followed by a position number. The last residue present in ced-3(n2452) animals is indicated as S179. Mutations examined by in vitro mutagenesis are indicated in parentheses adjacent to the relevant residues. The serine-rich region (between amino acid residues R93 and S205), the region deleted in ced-3(n2452) animals, and the QACRG sequence surrounding the active-site cysteine 358 are indicated.

Most missense alleles of ced-3 affect residues conserved with other caspases:
To define additional residues important for ced-3 function, we isolated 14 new ced-3 alleles. We then characterized the phenotypes caused by and the molecular lesions of these alleles and of 38 previously existing ced-3 alleles induced in vivo. One of the 52 alleles we examined, n1949, was isolated as an inhibitor of normal programmed cell death; 4 alleles (n1163, n1164, n1165, and n1286) were isolated in noncomplementation screens for suppression of the ectopic cell death of the HSN neurons in egl-1 mutant animals; 10 alleles (n717, n718, n1040, n1129, n2719, n2720, n2721, n2722, n2830, and n2998) were isolated as suppressors of mutations causing defects in either the engulfment or morphology of cell corpses; and 37 alleles (n2424, n2425, n2426, n2427, n2429, n2430, n2432, n2433, n2436, n2438, n2439, n2440, n2442, n2443, n2444, n2445, n2446, n2447, n2449, n2452, n2454, n2854, n2859, n2861, n2870, n2871, n2877, n2883, n2885, n2888, n2889, n2921, n2922, n2923, n2924, n3001, and n3002) were isolated as suppressors of the lethality conferred by the weak loss-of-function mutation ced-9(n1950 n2161) (see MATERIALS AND METHODS).

To quantify the severity of the defects in programmed cell death of different ced-3 mutants, we counted the number of extra surviving cells present in the anterior region of the pharynx, as has been previously described (HENGARTNER et al. 1992 ). To determine the molecular nature of the ced-3 alleles studied, we used PCR to amplify coding regions and exon/intron boundaries from each mutant strain and determined the DNA sequences of these regions (see MATERIALS AND METHODS). In the case of the allele n2452 we failed to amplify sequences downstream of intron 3 of the ced-3 gene, which suggests that a deletion might be present (see below; Figure 1 Figure 2 Figure 3).

View larger version (133K): In this window In a new window Download PPT slide   Figure 2. Positions of ced-3 mutations. An alignment of sequences of CED-3 proteins from the nematodes C. elegans, C. briggsae, and C. remanei (YUAN et al. 1993 ) with sequences of mouse and human caspase-1 (CERRETTI et al. 1992 ; THORNBERRY et al. 1992 ; MOLINEAUX et al. 1993 ), human caspase-4 (FAUCHEU et al. 1995 ), mouse caspase-2 (KUMAR et al. 1994 ), human caspase-2L and caspase-2S (WANG et al. 1994 ), human caspase-3 (FERNANDES-ALNEMRI et al. 1994 ; TEWARI et al. 1995 ), human caspase-6 (FERNANDES-ALNEMRI et al. 1995A ), human caspase-7 (FERNANDES-ALNEMRI et al. 1995B ), human caspase-5 (MUNDAY et al. 1995 ), amino acids 226–479 of human caspase-8 (BOLDIN et al. 1996 ; MUZIO et al. 1996 ), amino acids 243–479 of human caspase-10 (FERNANDES-ALNEMRI et al. 1996 ), human caspase-9 (SRINIVASULA et al. 1996 ), human caspase-11 (WANG et al. 1996 ), mouse caspase-12 (VAN DE CRAEN et al. 1997 ), human caspase-13 (HUMKE et al. 1998 ), D. melanogaster drICE (FRASER and EVAN 1997 ), D. melanogaster DCP-2 (INOHARA et al. 1997 ), and D. melanogaster DCP-1 (SONG et al. 1997 ). Shaded regions represent residues conserved between C. elegans CED-3 and any of the aligned proteins. CED-3 missense mutations are indicated with the allele name and the altered residue above the wild-type residue or by the allele name followed by a colon and the altered residue with the mutation site identified by a vertical line. Nonsense mutations are indicated with the allele name and an x above the altered residue. The sites of exon borders in the C. elegans sequence are indicated by arrowheads above the alignment. The N-terminal end of the deletion in ced-3(n2452) animals, which is located in intron 3, is indicated as an arrow; all sequences to the right of the vertical bar are absent. Mutations in splice donor or acceptor sites are indicated as Do or Ac, respectively, followed by the allele name. CED-3 cleavage sites are indicated with arrows above the C. elegans CED-3 sequence. *, animals carrying the n2445 allele probably contain a protein extended 26 amino acids beyond the wild-type prote

View larger version (28K): In this window In a new window Download PPT slide   Figure 3. ced-3(n2452) is a deletion. (A) Southern blot of ced-3(n2452) and wild-type genomic DNA digested with XhoI and probed with a full-length ced-3 cDNA. Note the absence of the 1.7-kb and 20.2-kb bands and the appearance of a new band of 4.7 kb in the ced-3(n2452) lane. (B) Southern blot of ced-3(n2452) and wild-type genomic DNA digested with HindIII and probed with a full-length ced-3 cDNA. Note the absence of the 3.7-kb and 4.8-kb bands and the appearance of a 4.9-kb band in the ced-3(n2452) lane. (C) A schematic diagram of the ced-3 genomic region in wild-type and ced-3(n2452) animals. Shaded box corresponds to sequences that are deleted in ced-3(n2452) animals. The ced-3 mRNA is indicated below the horizontal bar. Numbers correspond to the sizes in kilobases of intervals flanked by restriction sites. (D) Sequences surrounding the ced-3(n2452) deletion breakpoints. Underlined residues are ced-3 intron 3 residues. Bold residues lie 17.3 kb downstream of the ced-3 intron 3 breakpoint. Italicized residues cannot be accounted for by the known sequence of the region and indicate an insertion of 12 bp.

All 52 alleles analyzed were isolated in mutant screens that used EMS as a mutagen. Of these alleles, 44 contained a single GC AT transition, the mutation induced most often by EMS (COULONDRE and MILLER 1977 ; ANDERSON 1995 ). Two alleles (n2432 and n2445) contained a single AT TA transversion, one allele (n2883) contained a single AT GC transition, one allele (n2446) contained a single GC TA transversion, three alleles (n2424, n2830, and n2854) contained several altered nucleotides, and one allele (n2452) contained a deletion of 17,229 bp (see below). Of the alleles containing single point mutations, 38 had missense mutations, 6 had nonsense mutations, and 4, including the n2440 allele, probably affected splicing. [Although n2440 converts a glutamine codon to an ochre stop codon, our studies of this allele identified a single class of ced-3(n2440) RNA generated by a new splice-donor site located just upstream of the n2440 mutation; this RNA presumably encodes an altered protein unaffected by the stop codon described above; see MATERIALS AND METHODS.]

Of the 30 distinct sites affected by these missense mutations, 29 are conserved with at least one other non-nematode caspase, even though CED-3 is no more than 34% identical to any of these caspases (Figure 2). The nonconserved serine-rich region of the CED-3 protein (amino acids 93–205; YUAN et al. 1993 ) is not affected by any of these missense mutations, and its functional importance remains unknown. Interestingly, seven of the alleles we studied (n1040, n2439, n2449, n2424, n718, n2719, and n2830) alter residues within the CED-3 prodomain, suggesting that the prodomain is essential for CED-3 function during programmed cell death.

Some of the missense mutations we studied could be assigned to residues that have been implicated in a specific caspase function based on the caspase-1 and caspase-3 X-ray structures. These residues seem likely to have a similar role in CED-3 function. Specifically, the alleles n2427, n2438, and n2830 (G474R) alter a glycine residue of CED-3 that, on the basis of its corresponding site in caspase-1, is probably located on the heterodimer-heterodimer interface (WILSON et al. 1994 ). CED-3 multimerization might be defective in these mutants. The alleles n2429 and n2883 (S314L and S314P, respectively), n2870 (R429K), n2721 and n2720 (H315Y), n2871 (R359Q), and n2433 (G360S) all alter residues with equivalents in caspase-1 located within 4 Å of the bound substrate, suggesting that these residues might be important for the CED-3 active site.

The allele n2871(R359Q) encodes a protein with a QACQG pentapeptide containing the active site cysteine. This sequence is present in at least five other caspases that possess proteolytic activity. Interestingly, we found that the CED-3(n2871) protein expressed in Escherichia coli lacked proteolytic activity (data not shown), suggesting that the QACQG sequence is functional only in specific sequence contexts. Similarly, a number of other mutations also introduce into the CED-3 protein amino acids normally found in other caspases: n1040(L27F), n2439(L30F), n3001(R242C), n2425-(G277D), n2889(E318K), n2924(E318K), n2923(A347V), n2870(R429K), and n1163(S486F).

The phenotypic characterizations of and the sequence alterations caused by the 52 ced-3 alleles are presented in Table 3 and described below.

ced-3 alleles define a graded series of function:
As shown in Table 3, the ced-3 alleles we analyzed define a graded series based on the number of extra cells present in the anterior pharynx. To determine if this assay was consistent with other measurements of ced-3 killing activity, we compared our results from Table 3 to results from two other tests of ced-3 activity.

First, we examined the ability of eight different ced-3 alleles to suppress the maternal-effect lethality of animals homozygous for the strong ced-9(n1950 n2077) loss-of-function allele, which contains a nonsense mutation at codon 160 of the 280-codon ced-9 open reading frame (HENGARTNER and HORVITZ 1994A ). Although all eight ced-3 alleles suppressed the lethality caused by the weak ced-9(n1950 n2161) mutation (data not shown), not all suppressed the stronger ced-9(n1950 n2077) mutation. Specifically, as shown in Table 4, none of the ced-9(n1950 n2077) progeny of ced-9(n1950 n2077)/+; ced-3(n2424, n2923, n2446, n2449, n2425)/+ animals was fertile. By contrast, ced-9(n1950 n2077) progeny of ced-9(n1950 n2077)/+; ced-3(n2447, n2443, n717)/+ animals were increasingly fertile (~10, 30, and 50%, respectively). Fertility correlated with the extra cell counts shown in Table 3: animals homozygous for the alleles n2424, n2923, n2446, n2449, and n2425 had no extra cells in the anterior pharynx, and n2447, n2443, and n717 animals had increasing numbers of extra cells (0.8, 1.8, and 11.2 extra cells, respectively).

Using a second assay of ced-3 activity, we observed that ced-9(n1950 n2077); ced-3(n2447 or n2443) animals were severely egg-laying defective, presumably as a consequence of the deaths of the HSN neurons required for egg laying (data not shown; HENGARTNER et al. 1992 ), whereas ced-9(n1950 n2077); ced-3(n717) animals were egg-laying competent. None of these ced-3 mutants is egg-laying defective in the absence of the ced-9 mutation. These results suggest that the ced-3 alleles n2447 and n2443 are less defective in ced-3 function (i.e., allow more cell death to occur) than is the allele n717, consistent with the results presented in Table 3 and Table 4.

Taken together, our observations support the hypothesis that the 52 ced-3 alleles we examined define a graded series of ced-3 activities as listed in Table 3.

ced-3(n2452) animals lack the protease region of CED-3:
The n2452 allele is deleted for the region of the ced-3 gene that encodes the p17 and p15 subunits, which form the CED-3 protease (Figure 1). As shown in Figure 3, Southern blots of ced-3(n2452) genomic DNA digested with XhoI or HindIII and probed with a full-length ced-3 cDNA revealed the absence of fragments present in wild-type genomic DNA. Specifically, as shown in Figure 3A, both a 1.7-kb XhoI fragment internal to the ced-3 gene and an adjacent 20.2-kb XhoI fragment present in wild-type animals were absent in ced-3(n2452) animals, whereas a 2.6-kb XhoI fragment containing the ced-3 promoter and first three exons remained intact in the mutant. Furthermore, a novel 4.7-kb XhoI fragment appeared in ced-3(n2452) animals. Similarly, as shown in Figure 3B, both a 4.8-kb and a 3.7-kb HindIII fragment present in wild-type animals were missing in ced-3 (n2452) animals, and a novel 4.9-kb band appeared in the mutant. Similar results were observed using other enzymes (data not shown). On the basis of these results we propose that ced-3(n2452) is a deletion that removes all coding sequences downstream of intron 3 of ced-3 (see Figure 1 Figure 2 Figure 3).

In support of this interpretation, we were able to amplify a wild-type-sized DNA fragment from ced-3(n2452) animals using PCR and primers located upstream of exon 1 and at the 5' end of intron 3 (primers SHA2 and PCR2, Table 1; data not shown). However, we could not amplify any DNA fragments from ced-3(n2452) animals using PCR and primer pairs located at the 3' end of intron 3 and in intron 4 (primers PCR3 and PCR4, Table 1), in introns 4 and 7 (primers PCR5 and 650, Table 1), or in intron 7 and downstream of the ced-3 stop codon (primers BD1 and 1200, Table 1; data not shown). Furthermore, we were able to amplify a 3.2-kb genomic fragment of DNA from ced-3(n2452) animals using primers PWR.30 and PWR.32 located 59 nucleotides upstream of the ced-3 intron 3 splice-donor site and 20.3-kb downstream of the ced-3 intron 3 splice-donor site, respectively. Partial sequence of this 3.2-kb DNA fragment was determined using the primer PWR.40. The resulting sequence was consistent with a deletion of 17,229 bp downstream of position 4008 in the ced-3 genomic sequence (YUAN et al. 1993 ; Figure 3). This deletion lacks the ced-3 region encoding amino acids 180 to the end of the protein—the region necessary for CED-3 protease activity (Figure 1 Figure 2 Figure 3). This deletion also removes two other putative genes (C48D1.1 and F58D2.2) and disrupts a third putative gene (F58D2.1).

Programmed cell death may occur in the absence of CED-3 protease function:
ced-3(n2452) animals had 9.5 ± 1.5 extra cells in the anterior pharynx (Table 3). By contrast, numerous other ced-3 mutants contained significantly more extra cells in the anterior pharynx than did ced-3(n2452) animals. For example, ced-3 (n2433) animals contained 12.4 ± 1.0 extra cells in the anterior pharynx (P < 0.001 by the unpaired Student's t-test). Mutant animals carrying a number of ced-4 mutations or the gain-of-function ced-9(n1950) allele similarly contained significantly more extra cells in the anterior pharynx than did ced-3(n2452) animals. ced-4(n1162) animals, for example, had 11.9 ± 1.1 extra cells and ced-9(1950) animals had 12.5 ± 0.8 extra cells. These observations suggest that some cells die by programmed cell death in ced-3(n2452) animals. If so, the protease activity of CED-3 might not be necessary to cause all programmed cell deaths in C. elegans.

To confirm that cells can undergo programmed death in ced-3(n2452) animals we examined the number of cell corpses present in the heads of ced-3(n2452) L1 animals. To facilitate our analysis, animals were scored in a ced-1(e1735) background, which results in the persistence of cell corpses. As shown in Table 5, half of the ced-1(e1735); ced-3(n2452) animals we examined contained at least one cell corpse. These results support the notion that programmed cell death can still occur in the absence of CED-3 protease activity. Interestingly, even animals harboring stronger ced-3 alleles such as ced-3(n717) or ced-3(n718) contained some corpses.

It is conceivable that the ced-3(n2452) strain we studied contains a second mutation that bypasses a need for CED-3 protease activity to induce programmed cell death. To explore this possibility, we backcrossed the ced-3(n2452) strain to the wild-type N2 strain four times. Also, we generated strains in which the backcrossed ced-3(n2452) chromosome IV had undergone recombination with a non-ced-3(n2452) chromosome IV on either side of the ced-3 gene (see MATERIALS AND METHODS). Four-times backcrossed ced-3(n2452) animals had 9.5 ± 1.5 (n = 15) extra cells; ced-3(n2452) dpy-4(e1166) animals in which the right arm of the original ced-3(n2452) chromosome IV was replaced with an arm from a dpy-4(e1166) strain by recombination had 9.7 ± 1.4 (n = 18) extra cells; and unc-32(e189) ced-3(n2452) in which the left arm of the original ced-3(n2452) chromosome IV was replaced with an arm from an unc-32(e189) strain by recombination had 9.3 ± 2.0 (n = 15) extra cells. These results indicate that if a modifier mutation exists in the original ced-3(n2452) strain, it is likely to be located within four map units to the right of ced-3 and two map units to the left of ced-3. It remains possible that the disruption of C48D1.1, F58D2.2, or F58D2.1 in the ced-3(n2452) strain bypasses a need for CED-3 protease activity, although none of these genes has a sequence suggesting a direct or indirect role in programmed cell death (our unpublished observations).

The N-terminal nonprotease prodomain of CED-3 might be important for programmed cell death:
Whether the phenotype of ced-3(n2452) animals represents the phenotype caused by a true ced-3 null allele is unclear (see DISCUSSION). For example, it is possible that the prodomain of the CED-3 protein, which might be functional in ced-3(n2452) animals, could affect programmed cell death.

We have obtained data that suggest that the prodomain of CED-3 can prevent programmed cell death when fused to a heterologous protein. Specifically, while examining the expression patterns of translational fusions of ced-3 to either lacZ or GFP reporter genes containing nuclear localization signals (NLS), we observed that wild-type animals expressing transgenes encoding the prodomain of CED-3 displayed extra cells in the anterior pharynx. (We have not characterized the expression patterns of these transgenes in detail; see MATERIALS AND METHODS.) First, we expressed genomic regions of ced-3 containing sequences 2.5 kb upstream of the start codon (P_ced-3) and terminating at different locations within the ced-3 coding region, fused to lacZ (Figure 4, transgenes B–D). Second, we expressed a C. elegans heat-shock promoter fused to a truncated ced-3 cDNA fused, in turn, to lacZ (Figure 4, transgenes E and F).

View larger version (23K): In this window In a new window Download PPT slide   Figure 4. ced-3-reporter fusion transgenes can inhibit programmed cell death. The structure of each construct is indicated graphically, with the names A–H indicated to the left. CED-3 amino acid positions are indicated under the ced-3 portion of each construct. lacZ and GFP fragments indicated contain an SV40 T antigen-derived nuclear localization signal (FIRE et al. 1990 ). The average number of extra cells in the anterior pharynx of animals of a given transgenic line is indicated. For heat-shock promoter transgenes the number of extra cells was scored in L3 larvae heat-shocked at 33° for 1 hr at the 200-cell embryo stage. No significant extra cell survival was detected in nonheat-shocked animals (data not shown). GFP expression in line 1 of construct G was significantly reduced compared with line 2, which presumably accounts for the observed lower number of extra cells. Each number represents an independent transgenic line. n, the number of animals observed for each strain. The range of extra cells observed in the anterior pharynx of each strain is indicated.

Transgenes C and D partially inhibited programmed cell death (Figure 4). For example, two lines of animals transgenic for transgene C had on average 1.5 or 2.4 extra cells in the anterior pharynx. Extra cells were also seen in animals carrying transgene F, indicating that ced-3 promoter sequences are not required to generate extra cells. The presence of extra cells also was not dependent on the lacZ moiety, because transgenes containing fusions to GFP also caused extra cells to be present (Figure 4, transgenes G and H). Finally, we demonstrated that reporter transgenes equivalent to transgenes C and D but lacking a NLS could also promote the presence of extra cells (data not shown), suggesting that the NLS is not required for this phenotype. In short, the expression of CED-3 residues 1–151 resulted in extra cells.

To test the hypothesis that the extra cells we observed resulted from an inhibition of programmed cell death, we introduced transgene D into animals homozygous for the ced-9(lf) alleles n1950, n2161, or n2812. Transgene D suppressed the lethality of both strains, further showing that this transgene can inhibit programmed cell death. These observations indicate that the expression of CED-3 prodomain fusion constructs can interfere with programmed cell death and raise the possibility that this region of the CED-3 proprotein might normally interact with components of the cell-death machinery. It has been suggested that CED-3 activation is mediated by binding of CED-4 to the CED-3 prodomain (CHINNAIYAN et al. 1997 ). If so, our CED-3 prodomain fusion constructs might directly compete with wild-type CED-3 for binding to CED-4, resulting in inhibition of programmed cell death.

Transgenes B and E, which encode residues 1–71 and 1–94 of CED-3, respectively, did not inhibit programmed cell death (Figure 4). This observation suggests that the region between residues 94–151 is important for protection. Interestingly, transgenes B and E both interrupt a region (caspase recruitment domain, amino acids 1–86) postulated to be required for interaction with the CED-4 protein, and it is possible that our CED-3 prodomain fusions interact with CED-4. That transgene A, containing identical N-terminal-encoding ced-3 sequences as transgene C but without a reporter fusion, was unable to protect against programmed cell death suggests that this region on its own might produce an unstable protein or require fusion to a heterologous protein to prevent cell death. Alternatively, this transgene might not have been expressed at adequate levels to prevent cell death.

Interestingly, ced-3(n2452) animals harboring transgene D (Figure 4) contained significantly more surviving cells in the anterior pharynx (11.7 ± 1.2, P < 0.001) than ced-3(n2452) animals alone (9.5 ± 1.5), suggesting that transgene D might affect a CED-3 protease-independent mode of programmed cell death.

Mutations in the conserved QACRG active-site pentapeptide of CED-3 are weakly dominant-negative:
While examining cell survival in animals heterozygous for the 52 EMS-induced ced-3 alleles, we noticed that 4 of these alleles (n2871, n2433, n2440, and n2430) were weakly semidominant (Table 3). Animals heterozygous for these alleles showed weak cell survival (0–2 extra cells per animal with more than half of the animals having at least one extra cell). These values are statistically different from those of wild-type animals (unpaired Student's t-test: P < 0.004 for n2440 and n2433, P < 0.001 for n2430 and n2871). To test for another semidominant effect of the ced-3(n2871) allele, we examined the viability of animals homozygous for the weak ced-9 allele n1950 n2161 and heterozygous for ced-3(n2871). unc-69(e587) ced-9(n1950 n2161) animals do not produce viable progeny (HENGARTNER and HORVITZ 1994A ). However, unc-69(e587) ced-9(1950 n2161); ced-3(n2871) dpy-4(e1166)/+ + produce viable animals. Of 84 such progeny examined, 48 were non-Dpy. Only 3 or 4 of the non-Dpy progeny would be expected to be homozygous for ced-3(n2871) as a result of recombination between ced-3 and dpy-4. Thus, the majority of the non-Dpy progeny observed were likely heterozygous for ced-3. Furthermore, 10 of 10 non-Dpy progeny scored were not homozygous for ced-3(n2871) as assessed by examination of their Ced phenotype. These results confirm that the ced-3(n2871) allele has semidominant effects.

To test whether the semidominant phenotype conferred by these alleles was caused by a haplo-insufficiency of the ced-3 locus, we examined animals either heterozygous for the deficiency sDf21, which spans ced-3, or heterozygous for the ced-3(n2452) deletion allele. As shown in Table 3, neither sDf21/+ animals nor ced-3(n2452)/+ animals showed significant extra cell survival in the anterior pharynx, suggesting that the semidominant phenotype conferred by the ced-3(n2871, n2433, n2440, n2430) alleles was caused not by haplo-insufficiency but rather by a dominant-negative interaction.

Two of the four ced-3 alleles with semidominant effects, ced-3(n2871) and ced-3(n2433), alter the arginine (R359) and glycine (G360) residues, respectively, in the highly conserved pentapeptide QACRG, which surrounds the active site of CED-3 and is characteristic of most caspases (Table 3, Figure 2; YUAN et al. 1993 ). To determine if other active-site CED-3 mutants might have dominant-negative effects, we expressed in wild-type animals a ced-3 transgene containing a heat-shock promoter fused to a ced-3 cDNA encoding an active-site cysteine-358-to-alanine mutant protein (see Figure 4 legend for heat-shock protocol). We found that in three separate lines containing this mutant transgene, animals contained on average 2.9 (n = 8), 4.9 (n = 9), and 2.6 (n = 9) extra cells in the anterior pharynx. This result further indicates that ced-3 alleles containing mutations affecting the conserved QACRG motif can prevent programmed cell death in a dominant-negative fashion.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The C. elegans CED-3 proprotein consists of an N-terminal prodomain with no known catalytic function and central and C-terminal regions that are cleaved from the proprotein and associate to form an active protease (YUAN et al. 1993 ; XUE et al. 1996 ). In this study we have characterized the effects of site-directed mutations introduced in vitro into the ced-3 gene, as well as ced-3 mutations induced in vivo by the chemical mutagen EMS and identified in genetic screens, to address the roles of the CED-3 protease and N-terminal prodomain in programmed cell death. This study has allowed a detailed analysis of the effects of specific amino-acid substitutions on CED-3 caspase function in vivo. Specifically, key residues in the prodomain and in the protease subunits of CED-3 have been identified as important for CED-3 function.

The protease activity of CED-3 and the processing of the CED-3 proprotein are important for programmed cell death in C. elegans:
In this article we demonstrate that a mutation in the presumptive active-site cysteine 358 of CED-3, previously shown to perturb CED-3 protease activity in vitro (XUE et al. 1996 ), also perturbed the in vivo killing activity of ced-3 (Table 2). In addition, we describe seven EMS-induced ced-3 mutant strains (n2429, n2883, n2870, n2721, and n2720, n2871, and n2433), each of which is defective in programmed cell death, and each of which contains an alteration in a residue (serine 314, histidine 315, arginine 359, glycine 360, or arginine 429) that, on the basis of comparison to the caspase-1 crystal structure, seems likely to lie within 4 Å of the bound substrate and be an integral element of the CED-3 active site. Together, these findings indicate that CED-3 protease activity is important for programmed cell death.

Our data also indicate that processing of the CED-3 proprotein is important for programmed cell death, because site-directed mutations in two residues at which the CED-3 proprotein is processed in vitro (aspartate 221 and aspartate 374; XUE et al. 1996 ) abolished the killing activity of CED-3 in vivo. The processing of the CED-3 proprotein could be important only for generating active CED-3 protease. Alternatively, such processing could be important for releasing the N-terminal prodomain of CED-3, which might have a role in programmed cell death (see RESULTS).

CED-3 protease activity might not be essential for all programmed cell deaths in C. elegans:
The ced-3 allele n2452 eliminates CED-3 protease function, because ced-3(n2452) animals contain a deletion that removes all sequences present in the mature protease (Figure 1 Figure 2 Figure 3; XUE et al. 1996 ). Nonetheless, some programmed cell deaths still occur in ced-3(n2452) animals. Specifically, we showed that animals carrying the ced-3(n2452) mutation contain an average of 9.5 extra surviving cells in the anterior pharynx, whereas at least 17 strains homozygous for other ced-3 alleles (n2442, n2721, n2889, n717, n2439, n2720, n2719, n2454, n2432, n2830, n718, n3002, n2883, n2440, n2871, n2430, and n2433) as well as strains homozygous for strong ced-4 alleles or for the ced-9(n1950) gain-of-function allele (HENGARTNER et al. 1992 ; SHAHAM and HORVITZ 1996B ; S. SHAHAM and H. R. HORVITZ, unpublished results) contain more extra surviving cells (up to an average of 12.5; unpaired Student's t-test: P < 0.01 for the number of extra surviving cells in the least cell-death defective strains listed). These results suggest that on average ~3 cells undergo programmed cell death in the anterior pharynx of ced-3(n2452) animals. Furthermore, in ced-3(n2452) animals containing a ced-1(e1735) mutation, which allows the visualization of cell corpses, corpses were observed in the head, providing additional evidence that programmed cell death occurs in this ced-3 mutant strain. We also have shown that ced-3(n2452) animals harboring an additional mutation affecting programmed cell death, ced-8(n1891), contain an average of 11.9 extra cells in the anterior pharynx (S. SHAHAM, G. STANFIELD and H. R. HORVITZ, unpublished observations), supporting the hypothesis that cell deaths do occur in ced-3(n2452) animals, because it appears that these deaths can be prevented by the ced-8(n1891) mutation.

We cannot preclude the possibility that simply eliminating CED-3 protease activity would result in the absence of all programmed cell deaths and that in ced-3(n2452) animals there is loss of a death-protective function in addition to the loss of CED-3 protease activity. Such a death-protective function could be provided by the CED-3 protein, by an alternative product of the ced-3 gene, or by the product of a gene closely linked to ced-3 and also disrupted in the ced-3(n2452) strain (see RESULTS). In each of these cases, cell death would occur in the absence of CED-3 protease activity, again suggesting that this activity is not absolutely essential for all programmed cell deaths.

We previously showed that the ectopic overexpression of ced-4 can induce programmed cell death to a limited extent in animals homozygous for the strong ced-3 (n2433) mutation, which substitutes a serine for glycine at codon 360 (SHAHAM and HORVITZ 1996A ). That cells can still die in ced-3(n2433) animals suggests three possibilities: (1) the ced-3(n2433) mutation does not fully eliminate CED-3 protease activity, (2) the CED-3 protein contains a nonprotease killing activity, or (3) a CED-3-independent activity is capable of killing cells in the absence of CED-3 activity. The latter two possibilities are also suggested by our finding that in the ced-3(n2452) deletion mutant cell death can occur in the absence of CED-3 protease activity.

How might some ced-3 alleles prevent programmed cell death more than the protease deletion mutant ced-3(n2452)?
First, as noted above, it is possible that the CED-3 protein contains a nonprotease killing activity that is not disrupted in the ced-3(n2452) mutant. In this case, the ced-3(n2452) mutation would not be a ced-3 null allele, and stronger, null, ced-3 alleles would disrupt both the protease and the nonprotease CED-3 killing activities. Second, if ced-3(n2452) is a null allele, two possibilities seem plausible. On the one hand, ced-3(n2452) might eliminate not only a ced-3 killing function but also a ced-3 protective function; both ced-4 (SHAHAM and HORVITZ 1996B ) and ced-9 (HENGARTNER and HORVITZ 1994B ) appear to have both killing and protective functions. On the other hand, the proteins encoded by strong ced-3 alleles, such as ced-3(n2433), may not only be defective in CED-3 activity but may also interfere with a CED-3-independent activity required for programmed cell death. Consistent with this hypothesis is our observation that the ced-3(n2433) allele can prevent cell death in a weakly semidominant fashion, suggesting that the ced-3(n2433) product can actively interfere with the cell-death process. Perhaps CED-3(n2433) can inhibit a second caspase capable of inducing programmed cell death. The mutant CED-3 protein might titrate an activator of or serve as a substrate for such a second caspase. In the latter case, the mutant CED-3 protein would act as a competitive inhibitor, similar to the proposed action of the baculovirus p35 protein in preventing programmed cell death in C. elegans (XUE and HORVITZ 1995 ). Alternatively, heterodimer formation between the CED-3 mutant protein and a second caspase might prevent protease activity of the second caspase. Three caspase-related genes have been identified in C. elegans (SHAHAM 1998 ). If the product of any of these genes participates in programmed cell death, it might be inhibited by mutant CED-3 products as described above.

The null phenotype of ced-3:
The issues discussed above highlight the importance of unambiguously establishing the phenotype caused by a complete absence of ced-3 function. It seems likely that a mutant completely lacking ced-3 function is viable and deficient in programmed cell death, as are the large number of ced-3 mutants we have characterized. In support of this hypothesis, four ced-3 alleles (n1163, n1164, n1165, and n1286) were isolated in noncomplementation screens—which could have identified lethal ced-3 alleles—and all four when homozygous result in animals that are viable and deficient in programmed cell deaths. Furthermore, when ced-3 function is inhibited using the method of RNA-mediated interference (FIRE et al. 1998 ), the resulting animals are also deficient in programmed cell deaths (P. W. REDDIEN and H. R. HORVITZ, unpublished observations).

However, 50 of the 52 ced-3 alleles we studied contain mutations that could retain some ced-3 function: missense mutation, nonsense mutations that could potentially allow synthesis of a fragment of the CED-3 protein, or splicing mutations. Furthermore, we examined ced-3 mRNA size and level in 11 of these 50 ced-3 mutants (n717, n718, n1040, n1129, n1163, n1165, n1286, n1949, n2426, n2430, and n2433) and found that all 11 produced ced-3 mRNA of roughly the same size and abundance as in the wild type (data not shown). By contrast, nonsense, missense, and splicing mutants defective in other C. elegans genes often contain little or no RNA (PULAK and ANDERSON 1993 ). Thus, it is conceivable that most of the existing ced-3 mutants generate CED-3 protein products that have either reduced or abnormal ced-3 activity.

Nonetheless, two ced-3 alleles (n2452 and n2888) do seem like good candidates for being null alleles, based upon their molecular lesions. As discussed above, ced-3(n2452) animals are deleted for the entire C-terminal region of the 503-amino-acid CED-3 protein beyond amino acid 180. ced-3(n2888) animals have an arginine-to-opal nonsense mutation at codon 154, presumably resulting in a truncated CED-3 protein lacking the regions necessary for protease activity. Thus, assuming that CED-3 protease activity is essential for all ced-3 function, ced-3(n2452) and ced-3(n2888) are both most likely null alleles. Isolation of complete deletions of the ced-3 locus should help resolve this issue.

Whatever the ced-3 null phenotype, it is clear from our studies that semidominant mutations in ced-3 can prevent programmed cell death. It is therefore possible that similar mutations in human caspases also result in cell survival. Such cell survival, as in the case of bcl-2, could promote tumor formation. Thus, not only recessive mutations, but also dominant mutations in human caspase genes might predispose carriers to the development of cancer.

	FOOTNOTES

¹ Present address: Department of Biochemistry and Biophysics, University of California, San Francisco, CA 94143-0448.
² Present address: Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, Texas 75235-9038.

	ACKNOWLEDGMENTS

We thank Linda Huang and members of the Horvitz laboratory for helpful comments about the manuscript and Gillian Stanfield and Michael Hengartner for isolating some of the ced-3 alleles we studied. S.S. was supported by a William Keck Foundation fellowship and a fellowship from the Glaxo Research Institute. B.D. was supported by the MIT Undergraduate Research Opportunity Program (UROP) and by the Howard Hughes Medical Institute. P.W.R. was supported by a National Science Foundation fellowship and a National Institutes of Health training grant. H.R.H. is an Investigator of the Howard Hughes Medical Institute.

Manuscript received June 16, 1999; Accepted for publication August 27, 1999.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ALNEMRI, E. S., D. J. LIVINGSTON, D. W. NICHOLSON, G. SALVESEN, and N. A. THORNBERRY et al., 1996  Human ICE/CED-3 protease nomenclature. Cell 87:171[Medline].

ANDERSON, P., 1995  Mutagenesis. Methods Cell Biol. 48:31-58[Medline].

BOLDIN, M. P., T. M. GONCHAROV, Y. V. GOLTSEV, and D. WALLACH, 1996  Involvement of MACH, a novel MORT1/FADD-interacting protease, in Fas/APO-1- and TNF receptor-induced cell death. Cell 85:803-815[Medline].

BRENNER, S., 1974  The genetics of Caenorhabditis elegans.. Genetics 77:71-94