Translocation of C. elegans CED-4 to Nuclear Membranes During Programmed Cell Death

See allHide authors and affiliations

Science  25 Feb 2000:
Vol. 287, Issue 5457, pp. 1485-1489
DOI: 10.1126/science.287.5457.1485


The Caenorhabditis elegans Bcl-2–like protein CED-9 prevents programmed cell death by antagonizing the Apaf-1–like cell-death activator CED-4. Endogenous CED-9 and CED-4 proteins localized to mitochondria in wild-type embryos, in which most cells survive. By contrast, in embryos in which cells had been induced to die, CED-4 assumed a perinuclear localization. CED-4 translocation induced by the cell-death activator EGL-1 was blocked by a gain-of-function mutation in ced-9 but was not dependent onced-3 function, suggesting that CED-4 translocation precedes caspase activation and the execution phase of programmed cell death. Thus, a change in the subcellular localization of CED-4 may drive programmed cell death.

Programmed cell death is important in regulating cell number and cell connections and for sculpting tissues during metazoan development (1). When misregulated, programmed cell death can contribute to various disease states, including cancer, autoimmune disease, and neurodegenerative disease (2). Many of the central components of the cell death machinery have been identified through genetic studies of the nematode Caenorhabditis elegans (3). Loss-of-function mutations in any of the genes egl-1, ced-3, or ced-4 or a gain-of-function mutation in the gene ced-9 block programmed cell death. Loss-of-function mutations in ced-9 cause sterility and maternal-effect lethality as a consequence of ectopic cell death and can be suppressed by ced-3 andced-4 mutations but not by egl-1 mutations, suggesting that ced-9 acts upstream of ced-3 andced-4 and downstream of egl-1. CED-9 is a member of the Bcl-2 family of cell-death regulators (4), and the EGL-1 protein contains a BH3 (Bcl-2 homology 3) domain and can physically interact with CED-9 (5). ced-3 encodes a caspase (6), while CED-4 is similar to mammalian Apaf-1, an activator of caspases (7). CED-4 can bind CED-9 and CED-3 in vitro, in yeast, and in mammalian cells (8), and the interaction of CED-9 and EGL-1 may influence CED-4 activity (9). These observations suggest a model (3) in which CED-3 causes programmed cell death; CED-4 activates CED-3; CED-4 is directly inhibited by CED-9 (10); and EGL-1 initiates cell death by directly inhibiting CED-9. To determine when and where these cell-death proteins act, we have explored physical interactions among them using immunohistochemistry.

To study the expression and subcellular localization of CED-9 and CED-4 in C. elegans, we generated polyclonal antibodies that recognize these proteins (11). Affinity-purified antibodies to CED-9 (anti–CED-9) specifically recognized bacterially expressed CED-9 and a 32-kD protein corresponding to CED-9 on a Western blot of wild-type (WT) (N2) embryo lysates; this protein was absent inced-9(n2812) embryo lysates (12,13). The ced-9(n2812) allele contains an amber stop mutation at codon 46 and is probably a molecular and genetic null allele (4). Fixed embryos stained with anti–CED-9 revealed that CED-9 was present in all cells during C. elegans embryogenesis (Fig. 1A), beginning as early as the two-cell stage. CED-9 levels peaked at approximately the 200-cell stage and slowly diminished, becoming undetectable around the time of hatching. CED-9 protein was not observed in larvae or adults. On the subcellular level, CED-9 exhibited a weblike, cytoplasmic staining pattern. CED-9 staining was highly similar to the staining of Mitotracker Red (14), which specifically labels mitochondria (Fig. 1).

Figure 1

CED-9 and CED-4 are localized to mitochondria in WT embryos. (A) CED-9 expression in a WT embryo of ∼ 30 to 50 cells. (B) Mitotracker Red localization in the same embryo as in (A). (C) Merged image of (A) and (B). (D) CED-4 expression in a WT embryo of ∼200 cells. (E) Mitotracker Red localization in embryo in (D). (F) Merged image of (D) and (E).

Anti–CED-4 recognized bacterially expressed CED-4 and detected a 63-kD protein on Western blots of N2 embryo lysates; this protein was absent in ced-4(n1162) embryo lysates (12). The ced-4(n1162) allele contains an ochre stop mutation at codon 79 and is probably a molecular and genetic null allele (15). Embryos stained with anti–CED-4 displayed a weblike pattern in all cells (Fig. 1D), very similar to the patterns of CED-9 and Mitotracker Red. CED-4 staining appeared at approximately the 100-cell stage, before the first programmed cell death, persisted through embryogenesis, and like CED-9, was not detected in larvae and adults. Of the 131 developmental cell deaths inC. elegans hermaphrodites, 113 occur during embryogenesis and the remainder occur during larval development. Although we have not detected CED-4 or CED-9 in larvae, ced-4 andced-9 mutants are defective in larval programmed cell deaths, suggesting that the CED-4 and CED-9 proteins act postembryonically.

We examined whether the expression and localization of CED-9 and CED-4 were affected by mutations that disrupt programmed cell death. Loss-of-function mutations in ced-3, ced-4, andegl-1, genes required for programmed cell death, did not affect either the expression pattern or mitochondrial localization of CED-9 protein. The expression and localization of CED-4 protein was also unaffected by loss-of-function mutations inced-3 and egl-1. To determine the expression pattern and localization of CED-4 in the absence of functional CED-9 protein, we stained ced-9(n2812);ced-3(n717) double-mutant embryos with anti–CED-4. Because ced-9(n2812) embryos derived from homozygous ced-9(n2812) hermaphrodites arrested before the appearance of visibly recognizable corpses and before CED-4 expression, these embryos could not be studied directly for CED-4 localization. Becauseced-3(n717) did not affect the localization of CED-4 but does suppress the lethality ofced-9(n2812), we instead used this double mutant to analyze CED-4 in the absence of CED-9. Inced-9(n2812);ced-3(n717) embryos, CED-4 was not localized to mitochondria but rather was associated with nuclear membranes (Fig. 2, A to C), as visualized by double staining embryos with anti–CED-4 and antibodies directed against C. elegans lamin (16). We obtained similar results using the ced-9 loss-of-function alleles n1950 n2161 or n1950 n2077 in combination with ced-3(n717). Mitotracker Red staining was not altered in ced-9(n2812);ced-3(n717) embryos, indicating that the shift in CED-4 localization represents a movement of CED-4 protein rather than a change in the morphology and/or localization of mitochondria (17).

Figure 2

CED-9 is required for the localization of CED-4 to mitochondria. (A) CED-4 expression in aced-9(n2812); ced-3(n717) loss-of-function (lf) embryo of ∼150 cells. (B) Lamin localization in the same embryo as in (A). (C) Merged image of (A) and (B). (D) CED-4 expression in WT embryo of ∼200 cells. (E) Lamin staining of embryo in (D). (F) Merged image of (D) and (E).

To confirm this localization of CED-4 protein to nuclear membranes inced-9(lf) embryos, we performed subcellular fractionations of embryo lysates (18). Both CED-9 and CED-4 were present predominantly in the organelle and membrane fraction, which includes the mitochondria [for example, (19)], in WT embryos (Fig. 3). By contrast CED-4 was present almost exclusively in the nuclear fraction inced-9(n2812);ced-3(n717) embryos. Thus, in WT embryos, in which most cells survive, both CED-9 and CED-4 appeared to be predominantly mitochondrial. However, inced-9(n2812);ced-3(n717) embryos, in which ectopic cell death was presumably initiated but blocked by the ced-3mutation, CED-4 was redistributed from mitochondria to nuclei. Thus, CED-9 protein is necessary to localize CED-4 to mitochondria.

Figure 3

CED-4 fractionates primarily with membranes and organelles from WT embryos and with nuclei fromced-9(lf) embryos. Western blot of subcellular fractionation (18) of lysates from WT and ced-9(n2812);ced-3(n717) double-mutant embryos, separated into nuclear, organelle and membrane, and cytosolic fractions. Histone H4, a nuclear fraction marker. HSP90, C. elegans heat-shock protein, a cytosolic marker.

These data suggest that stimuli that induce programmed cell death would induce a redistribution of CED-4 to nuclear membranes and that it might be possible to block programmed cell death by blocking CED-4 relocalization. We tested these predictions by ectopically inducing programmed cell death in embryos.

The binding of EGL-1 protein to CED-9 may directly inhibit CED-9 function and trigger programmed cell death by releasing CED-4 from a CED-9–CED-4 complex (5, 9). To determine whether EGL-1 protein can affect the localization of CED-9 or CED-4, we expressed EGL-1 protein globally from an egl-1 cDNA under the control of two C. elegans heat-shock promoters (Phsp egl-1) (20) in the presence of the ced-1(e1735) mutation, which reduces cell-corpse engulfment and allows the quantification of cells that have undergone programmed cell death (21). Animals carrying heat-shock vectors without the egl-1 cDNA insert developed normally, but transgenic animals carrying Phsp egl-1 arrested during embryogenesis after heat-shock treatment. The few hatched L1 larvae contained many more cell corpses than vector-only animals, indicating extensive programmed cell death (Table 1). Localization of CED-9 was unaffected in these animals. By contrast, overexpressed EGL-1 triggered the translocation of CED-4 from mitochondria to nuclei (Fig. 4A).

Figure 4

Overexpression of EGL-1 induces CED-4 translocation from mitochondria to nuclear membranes inced-9(+) embryos but not inced-9(n1950) embryos. (A) CED-4 localization after heat shock in a ced-9(+) embryo carrying Phsp egl-1. (B) CED-4 localization after heat shock in a ced-9(n1950) embryo carrying Phsp egl-1. (C) CED-4 localization in a ced-4(n3040) embryo was diffusely cytoplasmic.

Table 1

EGL-1 induces ectopic cell death that can be suppressed by the ced-9 gain-of-function mutationn1950. Corpses were counted in the heads of transgenic L1 animals subjected to heat shock (20) and represent the mean ± SD and range observed (n, number of animals). The egl-1(n1084 n3082) allele is referred to asegl-1(lf), whereasegl-1(n3082) indicates a transgene engineered to contain only the n3082 5-bp deletion and not then1084 lesion in the egl-1 3′ regulatory region.

View this table:

We next introduced the extrachromosomal array carrying Phsp egl-1 into two strains in which programmed cell death is blocked. Theced-3(n717) mutation suppressed programmed cell death induced by EGL-1 overexpression (Table 1) but did not affect CED-4 translocation from mitochondria to nuclear membrane (22). This observation supports the idea that the release of CED-4 is not merely a consequence of cell death but rather precedes the execution of programmed cell death. Likeced-3(n717), theced-9(n1950) gain-of-function mutation blocked the ectopic death induced by egl-1 overexpression (Table 1). However, unlike ced-3(n717),ced-9(n1950) also blocked the translocation of CED-4 (Fig. 4B), suggesting that this mutant form of CED-9 either is unable to interact with EGL-1 or is unable to release CED-4. We tested the interaction of CED-9(G169E) protein, which is encoded by theced-9(n1950) mutation, with EGL-1 protein both in vitro and in yeast two-hybrid experiments and were unable to detect any difference between the interactions of the CED-9 and CED-9(G169E) proteins with the EGL-1 protein (23). It is possible that these in vitro studies failed to reveal a defect in the interaction between EGL-1 and CED-9 sufficient to produce the gain-of-function phenotype observed in vivo inced-9(n1950) animals. Alternatively, inced-9(n1950) animals, EGL-1 may form a ternary complex with CED-9 and CED-4 without causing the release of CED-4. We also generated an egl-1 heat-shock construct bearing the egl-1(n3082) mutation, which results in a truncated EGL-1 protein, a disruption of CED-9 binding, and a strong cell-death defective phenotype. Transgenic animals carrying this construct had significantly fewer corpses than animals bearing the WTegl-1 construct (Table 1). CED-4 localization was predominantly mitochondrial, but in occasional animals a few cells displayed nuclear CED-4 localization. Thus, overexpression of thisegl-1(n3082) gene resulted in a weak partial induction of both programmed cell death and CED-4 translocation.

Overexpression of egl-1 was sufficient to trigger both cell death and CED-4 translocation. Is egl-1 necessary for the CED-4 translocation that occurs in the absence of CED-9? We stainedced-9(n2812);ced-3(n717); egl-1(n1084 n3082) embryos and determined that CED-4 protein was nuclear, just as in the ced-9(n2812);ced-3(n717) embryos. Thus, in the absence of CED-9 protein, EGL-1 is not required to release CED-4 from mitochondria to nuclei, indicating that EGL-1 promotes CED-4 translocation by antagonizing the activity of CED-9.

Thus, we observed that CED-4 was mitochondrial in living cells and nuclear in cells that had initiated programmed cell death, so that the subcellular localization of CED-4 appeared to correlate with the cell-death status of a cell. We next studied the localization of CED-4 in six ced-4 missense mutants: n2860,n2879, n3040, n3043, n3100, and n3141. In five of the six mutants, CED-4 was mitochondrially localized in the presence of CED-9 and was associated with the nuclear membrane in the absence of CED-9, as in the WT. Inced-4(n3040) embryos, however, CED-4 displayed a diffuse, cytoplasmic localization both in the presence and in the absence of CED-9 (Fig. 4C), distinct from the weblike mitochondrial pattern of WT CED-4. ced-4(n3040), which causes a proline-to-leucine substitution at amino acid 23 (P23L) in a region that lacks any known protein motifs, results in as strong a cell-death defect as does ced-4(n1162), which contains an early ochre nonsense mutation. This P23L substitution reduces the interaction between CED-9 and CED-4 by about 75% in the yeast two-hybrid assay (24). The failure of CED-4(P23L) to associate with either mitochondria or nuclear membranes suggests that CED-4 is actively recruited not only to mitochondria (presumably through interaction with CED-9) but also to the nucleus. Alternatively, CED-4 may first have to interact with CED-9 to be competent to translocate to nuclear membranes. That WT CED-4 associated with nuclear membranes in the absence of CED-9 argues against this latter model.

CED-9 localization to mitochondria in C. elegansembryos is not surprising, given that the mammalian CED-9–like cell-death protectors Bcl-2 and Bcl-XL both localize to mitochondria (25). Although Bcl-XL and the CED-4–like protein Apaf-1 have been reported to physically interact (26), Moriishi et al.(27) recently reported that they could find no interaction between Apaf-1 and any known anti-apoptotic Bcl-2 family member. Furthermore, there is no evidence for the localization of Apaf-1 to mitochondria. Apaf-1 was isolated as a cytosolic activator of caspases (7), and overexpressed CED-4 is cytosolic in mammalian cells (8). Therefore, the mitochondrial localization of CED-4 is unexpected.

Our data suggest a model in which the activity of CED-4 is regulated by its subcellular localization. Specifically, we propose that in living cells, CED-9 prevents CED-4 activity by sequestering CED-4 to mitochondria. In cells triggered to undergo programmed cell death, EGL-1 binding to CED-9, possibly as a consequence of increasedegl-1 transcription (28), causes CED-4 release from CED-9 and allows the translocation of CED-4 to the nuclear region. There CED-4 activates the CED-3 procaspase, thereby causing programmed cell death.

How might we reconcile our findings with the report of Moriishiet al. (27) describing their failure to detect interactions between Apaf-1 and Bcl-2 family members? One possibility is that CED-9 has anti-apoptotic activity independent of its interaction with CED-4 and that this activity corresponds to the anti-apoptotic activity of Bcl-2 and Bcl-XL. For example, CED-9 can directly inhibit the CED-3 caspase (29), although it has not been shown that this inhibition acts physiologically and the region of CED-9 involved is not present in Bcl-2 or Bcl-XL. Furthermore, at least some CED-4 is localized to the nuclear membrane at the permissive temperature in ced-9(n1653ts) embryos (22), suggesting that this mutant CED-9 protein can protect against cell death even when CED-4 is localized to the nucleus; however, we suspect that the level of nuclear CED-4 in these embryos is lower than in cells that are dying, so this level may simply be insufficient to trigger programmed cell death.

The death-promoting proteins Bax and BAD, which like EGL-1 contain BH3 domains, translocate to mitochondria and bind anti-apoptotic Bcl-2 family members in response to apoptotic signals (30). Whether and how this translocation promotes cell death is unknown. Our results suggest that Bax and BAD may act to release Apaf-1 or another CED-4–like protein, allowing it to activate caspase processing. Some caspase precursors, specifically procaspases-2, and -3, are present in mitochondria and upon activation translocate to nuclei (31). It is possible that this movement of caspases involves the translocation of a complex that includes a CED-4–like protein. By analogy, the translocation of a CED-4–CED-3 complex from mitochondria to the nuclear envelope could provide access for the active caspase to both the nucleus and the cytosol, thereby fulfilling the roles of the multiple, differentially localized mammalian caspases.

The release of CED-4 from mitochondria resulted in the translocation of CED-4 to another distinct subcellular compartment rather than in the dispersal of CED-4 throughout the cell. This result, combined with our finding that the CED-4(P23L) mutant protein was diffusely cytoplasmic, suggests that CED-4 is recruited to nuclear membranes, possibly by interacting with another protein or protein complex. The identification of such a CED-4 receptor should help us understand the mechanism of action of CED-4 in the execution of programmed cell death.

  • * These authors contributed equally to this work.

  • Present address: Whitehead Institute, Department of Biology, WI-525, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.

  • Present address: Max Planck Institute for Neurobiology, Am Klopferspitz 18A, D-82152 Planegg-Martinsried, Germany.


View Abstract

Navigate This Article