Mechanisms and control of programmed cell death in ... - Nature

Oncogene (1998) 17, 3215 ± 3223 ã 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc

Mechanisms and control of programmed cell death in invertebrates Andreas Bergmann1, Julie Agapite1 and Hermann Steller*,1 1

Massachusetts Institute of Technology, Howard Hughes Medical Institute, Departments of Biology and Brain and Cognitive Science, 77 Massachusetts Avenue, Bldg. 68-430, Cambridge, Massachusetts 02139, USA

Apoptosis is a morphologically distinct form of programmed cell death that plays important roles in development, tissue homeostasis and a wide variety of diseases, including cancer, AIDS, stroke, myopathies and various neurodegenerative disorders (see Thompson (1995) for review). It is now clear that apoptosis occurs by activating an intrinsic cell suicide program which is constitutively expressed in most animal cells, and that key components of this program have been conserved in evolution from worms to insects to man. Genetic studies of programmed cell death in experimentally highly accessible invertebrate model systems have provided important clues about the molecular nature of the death program, and the intracellular mechanisms that control its activation. This review summarizes some of the key ®ndings in this area, but also touches on some of the many unresolved questions and challenges that remain. Keywords: C. elegans; Drosophila; apoptotsis; programmed cell death; ced genes

Introduction Essentially all animal cells have the ability to kill themselves by activating an intrinsic cell suicide program when they are no longer needed or have become seriously damaged (Jacobson et al., 1997). The execution of this program leads to a morphologically distinct form of cell death termed apoptosis (Kerr et al., 1972; Wyllie et al., 1980). It is now generally accepted that apoptosis is of central importance for development and homeostasis of metazoan animals. The roles of apoptosis include the sculpting of structures during development, deletion of unneeded cells and tissues, regulation of growth and cell number, and the elimination of abnormal and potentially dangerous cells (Jacobson et al., 1997). In this way, apoptosis provides a stringent and highly e€ective `quality control mechanism' that limits the accumulation of harmful cells, such as self-reactive lymphocytes, virus-infected cells, and tumor cells (Reed, 1995; Thompson, 1995; Naik et al., 1996; Morin et al., 1996; White, 1996). On the other hand, inappropriate apoptosis is associated with a wide variety of diseases, including AIDS, neuro degenerative disorders, and ischemic stroke (Martinou et al., 1994; Thompson, 1995; Pettman and Henderson, 1998). Because it is now clear that apoptosis is the result of an active, genedirected process, it should be eventually possible to

*Correspondence: H Steller

manipulate this form of death by developing drugs that interact with cell death proteins. During the last few years, there has been considerable progress in understanding the intracellular mechanism of apoptosis, and some of the key biochemical steps in this process have been de®ned. However, much remains to be learned about the precise mechanism of death, and its regulation by di€erent signaling pathways. For example, we still know extremely little about how speci®c cells are selected for death, and how cells that should live are protected from inappropriate activation of the death program. Progress in this area has been slow because it is still dicult and time-consuming to perform decisive mechanistic studies of apoptosis in vertebrate animals, and because it appears that many key molecules have not yet been identi®ed. Fortunately, the basic mechanism of apoptosis has been conserved throughout animal evolution, from C. elegans to Drosophila to humans (Steller, 1995; White, 1996; Hengartner, 1996; Jacobson et al., 1997; Pettman and Henderson, 1998). Therefore, studies on the mechanism and regulation of apoptosis in simple genetic model organisms can provide an intellectual framework in which mammalian apoptosis can be analysed. This review summarizes some of the key ®ndings and concepts that have emerged from studies of programmed cell death in invertebrate systems. Mechanism of apoptosis: lessons learned from Caenorhabditis elegans During the last few years, rapid progress has been made in identifying some of the molecules that are responsible for the regulation and execution of apoptosis. In assessing this work, it is useful to distinguish between the e€ector components of the cell intrinsic death program, i.e. the molecules that carry out apoptosis (as de®ned by characteristic morphological and biochemical changes), and regulators that control the activation of this program. There is strong evidence that the death program is constitutively expressed in most cells, but the activation of this program may require new protein synthesis (Steller, 1995; Shaham and Horvitz, 1996; Jacobson et al., 1997). Much of our current knowledge about the nature of cell death e€ectors was originally derived from genetic studies in the nematode Caenorhabditis elegans (Ellis and Horvitz, 1986; Ellis et al., 1991; Hengartner and Horvitz, 1994a; Hengartner, 1996). In particular, this work demonstrated the importance of an unusual class of cysteine proteases, termed caspases (for cysteine aspartic acid-speci®c protease; Alnemri et al., 1996) in apoptosis: loss-of-function mutations in

Molecular genetics of apoptosis A Bergmann et al

3216

ced-3, which encodes a caspase in C. elegans, lack programmed cell death (Yuan et al., 1993). Subsequently, a large number of mammalian caspases have been isolated, and evidence for their involvement in apoptosis has come from a variety of inhibition and overexpression studies (Takahashi and Earnshaw, 1996; Salvesen and Dixit, 1997; Thornberry et al., 1997; Nicholson and Thornberry, 1997). Caspases are synthesized as inactive zymogens which are widely expressed in both dying and living cells. In order to generate the active enzyme, the zymogen has to be proteolytically processed. Based on the known 3-D structure of two caspases (caspase-1 and -3), it is thought that the active protease consists of two heterodimers that are produced by internal proteolytic cleavages of the pro-enzyme. This leads to the removal of an inhibitory N-terminal pro-domain and the generation of a large (p20) and small (p10) subunit. Signi®cantly, these cleavages occur after aspartate residues at substrate recognition consensus sites for caspases, and mature caspases can often cleave and activate their own as well as other caspase zymogens in vitro. Thus, a caspase cascade in which `initiator' caspases activate downstream `executioner' caspases has been proposed (Takahashi and Earnshaw, 1996; Salvesen and Dixit, 1997; Thornberry et al., 1997). It is thought that the executioner caspases are responsible for cell killing by cleaving critical cellular target proteins, such as nuclear lamins and other structural proteins (Rao et al., 1996; Kothokota et al., 1997; Tan and Wang, 1998). One of the central question that has remained is how such a caspase cascade is initially activated in response to death-inducing stimuli. A series of recent studies suggest that a caspase cascade may `self-start' by autoproteolysis following aggregation of two or more zymogen molecules. If pro-caspase molecules are brought together into close contact, the weak proteolytic activity of the zymogen molecules can be sucient to form active enzyme by trans-proteolysis (Yang et al., 1998; Srinivasula et al., 1998; Muzio et al., 1998; MacCorkle et al., 1998; Steller, 1998). Once active enzyme is formed, it may contribute to the proteolytic activation of additional zymogen molecules and thereby start a proteolytic cascade that leads to apoptosis. At least in general terms, this simple `proximity' model for the activation of caspases resembles the activation of the complement protease cascade and tyrosine kinase activation by intermolecular cross-phosphorylation (Ullrich and Schlessinger, 1990; Davie et al., 1991). These observations raise an important question. How do death-inducing factors promote the aggregation of pro-caspases? The prodomain of caspases appears to be the target for regulatory proteins, such as CED-4 in C. elegans (Yuan and Horvitz, 1992) and mammalian APAF-1 (apoptosis protease-activating factor-1) which shares signi®cant amino acid homology with CED-4 (Zou et al., 1997). The ced-4 gene acts genetically upstream of ced-3 in C. elegans (Figure 1), and CED-4 can physically interact with pro-CED-3 and certain mammalian pro-caspases (Shaham and Horvitz, 1996; Seshagiri and Miller, 1997; Chinnaiyan et al., 1996; Wu et al., 1997). APAF-1 can bind to the prodomain of pro-caspase-9 and activate it in the presence of cytochrome C and dATP in a cell free system (Li et

al., 1997; Zou et al., 1997). These observations suggest that CED-4/APAF-1 may regulate the activation of a caspase cascade by complex formation, possibly in a manner similar to the `proximity model'. However, the validity of this model remains to be critically tested in vivo, since all the cited physical interactions have so far been only demonstrated under unphysiological conditions. Another important question that emerges from these observations is how cells are protected from inappropriate activation of a caspase cascade. Procaspases are widely expressed at high levels and have low but signi®cant protease activity (approximately 1% of the mature enzyme). Despite this potentially dangerous cargo, many cells can avoid the activation of a caspase cascade and death for many years (or even our lifetime, as in the case of many neurons). Therefore, ecient mechanisms must exist that inhibit caspase activation. How is the killing activity of CED-4/APAF-1 restricted to doomed cells? Again, this appears to be mediated by post-translational mechanisms since these proteins are expressed in both living and dying cells (Shaham and Horvitz, 1996; Zhou et al., 1997). In C. elegans, the activity of ced-9 is required to prevent inappropriate cell deaths, and ced-9 acts genetically upstream of ced-4 and ced-3 (Figure 1; Hengartner et al., 1992; Shaham and Horvitz, 1996). Signi®cantly, ced-9 encodes a protein homologous to mammalian Bcl-2-like molecules that regulate apoptosis in mammals (Hengartner and Horvitz, 1994b; White, 1996). The human Bcl-2 gene was initially identi®ed as a translocation breakpoint common in many B cell lymphomas. As a result of this translocation, Bcl-2

Figure 1 Schematic outline of the cell death pathways in C. elegans and Drosophila. Both organisms provide a di€erent genetic paradigm for the induction of apoptosis through caspase activation. In C. elegans, caspase activation is accomplished by releasing the inhibitory activity of CED-9 by protein/protein interaction with EGL-1 via the BH-3 domain. As a result, CED-4 is activated and can turn on the caspase CED-3. In Drosophila, IAPs may directly bind to pro-caspases via their BIR domains and block the activation. RPR, HID and GRIM may dissociate the IAP/caspase complex by competing for binding to the BIR domain. This might allow aggregation of two or more pro-caspase molecules followed by trans-proteolysis to generate the mature enzyme. Activation of RPR, HID and GRIM occurs in response to many death-inducing stimuli, including steroid hormones (ecdysone), developmentally controlled cell death, lack of peptide survival factors, activation of death receptors, failure of di€erentiation, viral infection and many others. Activation of the Ras/MAPK pathway by extracellular survival factors leads speci®cally to inhibition of the death activator HID. For more details see text

Molecular genetics of apoptosis A Bergmann et al

comes under the control of the immunoglobulin heavy chain enhancer and is constitutively expressed in B cells (Bakhshi et al., 1985; Tsujimoto et al., 1985; Cleary et al., 1986). The resulting protection against apoptosis apparently permits the survival and accumulation of aberrant B cells that will ultimately give rise to malignancies. In recent years, a large number of additional Bcl-2-like genes have been identi®ed (White, 1996). Like Bcl-2, many of these new genes have anti-apoptotic activities but some members, such as Bax, can actually stimulate apoptosis (Reed, 1997). One important property of Bcl-2-like proteins is their ability to form homo- and heterodimers with other family members, and these interactions can modulate activity (Oltvai et al., 1993; White, 1996). Despite intense e€orts, the precise mechanism by which Bcl-2like proteins regulate cell death remains unknown. Over the years, several di€erent biochemical functions have been proposed (see, for example, Hockenberry et al., 1993; Oltvai et al., 1993; White, 1996; Vander Heiden et al., 1997; Reed, 1997). A model that currently enjoys considerable popularity is based on the observation that Bcl-2-like proteins can have poreforming activities in synthetic lipid membranes (Minn et al., 1997; Schendel et al., 1997; Antonnsson et al., 1997). Because BCL-2 is present in the outer mitochondrial membrane, it has been suggested that Bcl-2 may block apoptosis by inhibiting the release of apoptosis-inducing factors, such as cytochrome C, from mitochondria (Kluck et al., 1997; Kim et al., 1997; Yang et al., 1997). Cytochrome C may be an important cofactor for the activation of caspases by APAF-1. Hence, it is attractive to speculate that Bcl-2 might block apoptosis by preventing the massive release of cytochrome C into the cytosol. However, this model is not easily reconciled with the phenotype of the ced-97 ced-37 double mutant in C. elegans. If ced-9 function was required to prevent cytochrome C release, the double mutant should die due to lack of mitochondrial function. However, ced-97 ced-37 animals live and contain many extra cells that would have normally died (Hengartner et al., 1992). These `undead' cells show no signs of mitochondrial defects that should have resulted from the postulated massive mitochondrial release of cytochrome C prior to caspase activation. Consequently, this model of Bcl-2 function remains subject to considerable controversy. An alternative model for Bcl-2 function has been based on the observation that CED-9 can form a ternary complex with CED-3 and CED-4 upon overexpression in mammalian cells (Chinnaiyan et al., 1996) and the observation of an equivalent complex between a Bcl-2 family member, Bcl-xL, with APAF-1 and caspase-9 (Pan et al., 1998). In this model, CED-9/ Bcl-xL inhibits CED-4/APAF-1 from activating CED3/caspase-9 by direct physical association with CED-4/ APAF-1. This model is attractive since it is consistent with the genetic data from C. elegans (Shaham and Horvitz, 1996). Thus, in living cells CED-9, CED-4, and CED-3 are thought to be present in an inactive complex. How then do pro-apoptotic stimuli lead to the activation of CED-3? The recent identi®cation of the C. elegans gene, egl-1 has led to insight into this question (Conradt and Horvitz, 1998). This gene appears to be pro-apoptotic, since loss of egl-1 function prevents most, if not all

somatic programmed cell deaths. Furthermore, genetic evidence suggests that egl-1 functions by negatively regulating the bcl-2 like gene ced-9 (Figure 1). Interestingly, the EGL-1 protein contains a 9 amino acid region similar to the BCL-2 homology domain 3 (BH3 domain) and may be a member of the BH3 only family of cell death activators (see Conradt and Horvitz, 1998, and references therein). Members of this family interact with BCL-2 like proteins, and EGL-1, likewise, has been shown to physically interact with CED-9 (Conradt and Horvitz, 1998). These results suggest a model in which EGL-1 promotes cell death by binding CED-9 and as a result displacing CED-4 and CED-3 from the inactive CED-9/CED-4/CED-3 complex. This model is further supported by the observation that BH3 containing proteins or peptides can block CED-4 binding to BCL-2 like death inhibitors (Chinnaiyan et al., 1997; Ottilie et al., 1997). EGL-1 function must be tightly regulated to avoid the inappropriate activation of CED-3 and death in cells that should live. The C. elegans cell death speci®cation genes (ces genes) are thought to be cell type speci®c regulators of the basic cell death machinery. ces-1 and ces-2 control the cell death decision of two speci®c cells, the NSM sisters, in the pharynx (Ellis and Horvitz, 1991). Presumably, other ces-like genes exist in other cell types that also regulate death versus life decisions. Since egl-1 is the most upstream component of the central cell death machinery identi®ed and appears to function downstream of ces-2 and ces-1, it is possible that it serves as the integration point of signals that specify the cell death fate (Conradt and Horvitz, 1998). Because the ces-2 gene encodes a basic leucine zipper (bZIP) transcription factor (Metzstein et al., 1996), it seems possible that egl-1 regulation occurs at the level of ces2-dependent transcription. Genetic analysis in C. elegans has been instrumental in formulating the current models for the mechanism underlying programmed cell death and probably holds the key to many additional unanswered questions. Control of apoptosis: lessons learned from Drosophila melanogaster A major gap in our current understanding of apoptosis is how speci®c signaling pathways regulate the induction of apoptosis. Not surprisingly, the control of programmed cell death is remarkably complex. Many di€erent signals that may originate either from within or outside a cell in¯uence the decision between cell death and survival. These include steroid hormones, lack of peptide survival factors, activation of death-inducing cell surface receptors, viral infection, oxidative stress, excitotoxicity, ischemia, unfolded proteins, and unrepaired DNA breaks (such as caused by ionizing radiation) (Figure 1; Oppenheim, 1991; Truman et al., 1992; Ra€, 1992; Steller and Grether, 1994; Agapite and Steller, 1997; Nagata, 1997; Pettmann and Henderson, 1998). Some new insights into these questions have recently come from genetic studies of apoptosis in Drosophila. In Drosophila, large numbers of cells die during embryonic and imaginal development as well as

3217

Molecular genetics of apoptosis A Bergmann et al

3218

metamorphosis (Truman et al., 1992; Abrams et al., 1993; Steller, 1995; McCall and Steller, 1997). These deaths occur with the characteristic morphology of apoptosis and result from the activation of a caspase pathway (McCall and Steller, 1997). Signi®cantly, in Drosophila, like in vertebrates, the regulation of apoptosis is highly plastic and involves a wide variety of stimuli originating from both within a cell as well as from its environment. For example, the neuronal cell number in the Drosophila visual system is regulated by adjusting both the rate of cell proliferation and cell death through competitive interactions between cells in the developing retina and brain (Steller et al., 1987; Selleck and Steller, 1991; Campos et al., 1992). Cells which fail to participate in the formation of appropriate neuronal circuits in the Drosophila visual system undergo apoptotic cell death (Steller and Grether, 1994). Therefore, Drosophila is particularly well-suited for studying the mechanisms regulating the control of apoptosis. Genetic control of apoptosis in Drosophila During Drosophila embryogenesis, large numbers of cells die in an overall reproducible pattern, and these deaths can be quickly and reliably detected in live embryos that are stained with the vital dye acridine orange (Abrams et al., 1993). To identify genes involved in apoptosis, White et al. (1994) examined the pattern of cell death in embryos homozygous for previously identi®ed chromosomal deletions. From this screen, one deletion, Df(3L)H99 (in the following referred to as H99) at chromosomal location 75C1,2 was found to be essential for virtually all cell deaths that normally occur during embryogenesis. In addition, these deletion mutant embryos were also protected against the ectopic cell deaths that are normally induced upon radiation, or in developmental mutants. However, very high doses of X-irradiation were capable of inducing a small amount of cell death in mutant embryos. Significantly, the morphology and kinetics of the few cell deaths observed under these circumstances was indistinguishable from that seen in wild type. This suggested that the cell death genes in H99 were speci®cally required for the activation, but not for the execution of the cell death program. The molecular characterization of H99 led to the identi®cation of three genes that play a central role in the activation of apoptosis in Drosophila. These genes are reaper (rpr), head involution defective (hid) and grim (Figure 1; White et al., 1994; Grether et al., 1995; Chen et al., 1996a). All three genes encode novel proteins. Careful examination of the deduced protein sequences of REAPER, HID and GRIM revealed a small region of similarity at the N-terminus (Grether et al., 1995; Chen et al., 1996a). As discussed in more detail below, this N-terminal region can physically interact with a domain of baculoviral IAP proteins. RPR also has some weak homology to the mammalian `death domain' of the type-1 tumor necrosis factor receptor (TNFR1) and Fas (Golstein et al., 1995). The signi®cance of this is not yet clear since mutation of residues conserved between RPR and Fas does not a€ect the cell killing activity of RPR (Chen et al., 1996b).

A fundamental unresolved question in apoptosis research is how distinct death-inducing stimuli converge to activate a common apoptotic program. Insight into this question has come from studying the expression of the Drosophila cell death genes reaper and grim. Unlike cell death e€ectors, such as caspases, reaper and grim are speci®cally expressed in cells that are doomed to die, anticipating death by several hours (White et al., 1994; Chen et al., 1996a; Robinow et al., 1997). The expression of rpr is also induced in response to a variety of other death-inducing stimuli, including X-irradiation, block of cellular di€erentiation, and steroid hormone regulated deaths (Nordstrom et al., 1996; Robinow et al., 1997; A.F. Lamblin and H.S., unpublished). This indicates that the integration of di€erent death inducing signals occurs, at least in part, by a transcriptional mechanism. Based on these observations, a model has been proposed according to which the genes rpr and grim integrate di€erent incoming death inducing signals to activate a common cell death program (Figure 1; White et al., 1994; McCall and Steller, 1997). A detailed analysis revealed that distinct promoter elements in the rpr promoter are required for the regulation of rpr expression in response to di€erent stimuli (AF Lamblin and HS, unpublished). As outlined below in more detail, the activity of hid appears to be regulated di€erently and involves both transcriptional and posttranslational control elements. Although each gene appears to be capable to induce apoptosis independently of the each other (see below), there is now increasing evidence that at least in some cellular paradigms these genes are acting in concert. Robinow et al. (1997) demonstrated that rpr and grim transcripts accumulate prior to apoptosis in a certain subset of neurons, the n4 cluster, whereas hid transcripts do not. Likewise, during metamorphosis in response to pulses of the steroid hormone ecdysone expression of hid and rpr is upregulated, whereas grim expression is not (Jiang et al., 1997). More strikingly, a synergistic interaction between rpr and hid has been shown during apoptosis of the midline glia (Zhou et al., 1997). Cell killing by reaper, hid and grim Ectopic expression of either reaper, hid or grim induces apoptosis in cultured cells, and in many di€erent cell types of transgenic animals (Grether et al., 1995; White et al., 1996; Pronk et al., 1996; Chen et al., 1996a; Zhou et al., 1997; McNabb et al., 1997; Bergmann et al., 1998). A particularly useful approach has been the transgenic expression of cell death genes under control of the eye speci®c GMR promoter in the Drosophila compound eye. Transgenic expression of GMR-rpr, GMR-hid and GMR-grim induces apoptosis that results in eye ablation (Figure 2; Grether et al., 1995; White et al., 1996; Chen et al., 1996a). Signi®cantly, coexpression of the anti-apoptotic baculovirus protein p35 completely suppresses these phenotypes and allows for essentially normal ommatidial di€erentiation (Grether et al., 1995; White et al., 1996; Chen et al., 1996a). Because p35 speci®cally inhibits caspase proteases (Bump et al., 1995; Xue and Horvitz, 1995), this result suggests that reaper, hid and grim kill by activating a caspase pathway (Figure 1).

Molecular genetics of apoptosis A Bergmann et al

3219

Figure 2 The eye ablation phenotypes caused by HID and RPR can be modi®ed by Ras and DIAP1. A Drosophila wild-type eye is shown in (a). Eye speci®c expression of the death activator HID in (b) and RPR in (d) (GMR-hid and GMR-rpr) results in strong ablation of the ¯y eye compared to wild-type (a). Note that the eye ablation caused by GMR-hid is much stronger than the one by GMR-rpr. The GMR-hid eye phenotype is strongly suppressed by activating mutations of Ras (RasV12) indicating that the Ras pathway exerts an anti-apoptotic activity (c). Loss-of-function mutations in the gene diap1 enhance the GMR-rpr eye phenotype (e) and also the GMR-hid and GMR-grim eye phenotypes (data not shown). This enhancement by diap1 mutations indicate a protective function of DIAP1 in the regulation of apoptosis. See text for more details

The eye phenotypes caused by the GMR-hid and GMR-rpr transgenes provide an exquisitely sensitive and rapid assay in genetic cell death screens aimed in identifying mutants in genes involved in the regulation and execution of apoptosis. These so-called modi®er screens have been very successful in de®ning the downstream components of Receptor Tyrosine Kinase signaling in Drosophila (Simon et al., 1991; Dickson et al., 1996; Karim et al., 1996), and cell death modi®er screens have been performed recently in several laboratories. Two interesting classes of known genes that have been recovered in the screens will be discussed in more detail below. Other mutants which de®ne potentially novel genes are currently under intensive study in our laboratory (JA, K McCall and HS, unpublished). The Drosophila gene hid links Ras-dependent cell survival and apoptosis In a genetic screen for modi®ers of the GMR-hid induced eye ablation phenotype, mutations in genes which regulate the activity of the Ras pathway in Drosophila were recovered as strong suppressors (Figure 2). The observed genetic interaction suggests an anti-apoptotic activity of the Ras pathway (Bergmann et al., 1998; Kurada and White, 1998; Sawamoto et al., 1998). Such an activity of the Ras pathway has been demonstrated before in ¯ies (Miller and Cagan, 1998; Sawamoto et al., 1998) as well as in vertebrates (reviewed in Downward, 1998). Miller and Cagan (1998) showed that apoptosis of secondary and tertiary pigment cells after laser ablation of primary pigment cells in the compound eye is blocked by simultaneous expression of an activated Ras allele, RasV1. Moreover, overexpression of Argos, a secreted inhibitor of Ras signaling, in post-mitotic cells resulted in induction of apoptosis (Sawamoto et al., 1998). Further analysis showed that activation of the Ras pathway leads speci®cally to inactivation of HID, whereas RPR and GRIM are not a€ected (Bergmann et al., 1998). This observation is interesting since hid expression is not only observed in cells which die, but also in cells which live (Grether et al., 1995), in contrast to rpr and grim which are exclusively expressed in dying cells (see above; White et al., 1994; Chen et al., 1996a). Thus, in the surviving hidexpressing cells very ecient mechanisms must operate

to prevent them from hid-induced apoptosis. Survival factors which act through the Ras pathway provide such an activity. The hid-inhibiting activity of Ras is largely mediated by activation of MAP kinase (MAPK), since the MAPK phosphorylation consensus sites in HID are critical for the inhibition (Bergmann et al., 1998). In addition to the post-translational regulation of HID activity by MAPK, activation of the Ras pathway also down-regulates hid expression (Kurada and White, 1998). This double mode of regulation makes perfect sense in that a cell that is committed to live acts in an acute response by inactivating existing HID protein levels and in a second response ensures that HID is no longer synthesized by down-regulating its transcription. In summary, these ®ndings link Ras-dependent survival signaling and apoptosis by directly inactivating a critical component of the intrinsic cell death machinery. Moreover, since oncogenic mutations in Ras are associated with 30% of all human tumors (Bos, 1989) and since cancer cells have acquired some resistance to apoptotic stimuli (Ho€man and Liebermann, 1994), this work might also have important implications for tumorigenesis and metastasis. Drosophila inhibitors of apoptosis proteins A second important class of genes, mutations of which have been isolated in cell death screens are Drosophila homologs of IAPs (Inhibitor of Apoptosis Proteins; Hay et al., 1995; JA, K McCall, and HS, unpublished). IAPs were originally discovered in baculovirus strains by their ability to functionally substitute for p35 in blocking apoptosis (Crook et al., 1993; Birnbaum et al., 1994; Clem and Miller, 1994). Subsequently, a number of related genes were discovered in Drosophila and mammals (Hay et al., 1995; Rothe et al., 1995; Liston et al., 1996; Duckett et all., 1996; Uren et al., 1996; Ambrosini et al., 1997). IAPs share several structural motifs: they contain at least one and usually two or three tandem baculovirus IAP repeat (BIR) motifs, and most have a carboxyterminal RING ®nger domain. Several, but not all IAPs have been shown to inhibit apoptosis. For example, loss-of-function mutations of diap1, one of the two known Drosophila IAPs, enhance reaper, grim and hid-induced cell death (Figure 2), and overexpression of diap1 suppresses apoptosis (Hay et al.,

Molecular genetics of apoptosis A Bergmann et al

3220

1995). The other Drosophila IAP homolog, diap2, functions similarly to diap1 when misexpressed in ¯y eyes. However, since mutations in the diap2 gene have not yet been described, its wild-type function is not clear. Ectopic expression of human IAPs can suppress apoptosis in several paradigms (Ducket et al., 1996; Liston et al., 1996; Uren et al., 1996; Ambrosini et al., 1997; Chu et al., 1997; Deveraux et al., 1997), and the expression of survivin, a human IAP homolog, correlates with oncogenic transformation (Ambrosini et al., 1997). Finally, mutations in NAIP, another human IAP gene, are thought to contribute to spinal muscular atrophy (SMA) which involves inappropriate neuronal apoptosis (Roy et al., 1995). It has been shown that IAPs function by directly binding to and inhibiting caspases from activation (Deveraux et al., 1997). Interestingly, IAPs appear to interact only with a certain subset of caspases. For instance, human XIAP as well as cIAP-1 and cIAP-2 can bind to and inhibit caspase-3, -7 and -9 but fail to interact with caspase-8, -1 and -6 (Deveraux et al., 1997, 1998; Roy et al., 1997). Conversely, NAIP fails to bind to any of these caspases. The interaction between IAPs and caspases requires the BIR domain of IAPs. The RING domain might modify this interaction. A detailed analysis showed that the second domain of the three BIR domains of XIAP is sucient for inhibiting caspases (Takahashi et al., 1998). Both baculovirus and Drosophila IAPs physically interact with RPR, HID and GRIM (Vucic et al., 1997a, 1998). These interactions are mediated through the BIR motifs of the IAPs and the N-terminal 14 residues of RPR, HID and GRIM, the region of sequence similarity shared by the three proteins (see above). Based on these biochemical observations (inhibition of caspase activation by IAPs as seen for XIAP, and direct interaction of RPR/HID/GRIM with IAPs) we propose the `double inhibition' model of Figure 1 to understand the role of the death activators RPR/HID/GRIM and the death inhibiting IAPs in the regulation of the onset of apoptosis. According to this model, IAPs may directly bind to and inhibit caspase activation. RPR, HID and GRIM trigger apoptosis by preventing IAPs from blocking caspase activation. Since the interaction of both RPR/HID/GRIM and caspases with IAPs requires the BIR domain of IAPs (see above), RPR/HID/GRIM may dissociate the IAP/ caspase protein complex simply by competition for binding to the BIR domain of IAPs. However, the `double inhibition' model in Figure 1 still awaits direct experimental proof. Although there are no vertebrate homologs of rpr, hid and grim known to date there are good reasons to believe that homologous genes exist in vertebrates. As mentioned above, RPR, HID and GRIM interact physically with the highly conserved IAP molecules (Vucic et al., 1997a, 1998). Also, the conserved Ras/ MAPK pathway regulates hid-activity (Bergmann et al., 1998; Kurada and White, 1998). In addition, RPR and GRIM can induce apoptosis in mammalian cultured cell lines (McCarthy and Dixit, 1998). Similarly, HID is also able to induce apoptosis when transfected in HeLa cells (N. Haining and H.S., unpublished). Furthermore, RPR can induce morphological changes associated with apoptosis in Xenopus

embryonic extracts and interacts with speci®c proteins in these extracts (Evans et al., 1997; Thress et al., 1998). Overall, it is likely that lessons learned from studying the major cell death activators in Drosophila will bene®t our understanding of the control of apoptosis in general. Drosophila caspases As previously discussed, a critical step in apoptosis is the activation of an unusual family of cysteine proteases, called caspases (Figure 1). At least three caspases have been identi®ed in Drosophila, DCP-1, drICE and DCP-2/DREDD (Song et al., 1997; Fraser and Evan, 1997; Inohara et al., 1997; Chen et al., 1998). DCP-1 and drICE are structurally and biochemically similar to CED-3 and mammalian caspase-3. They can induce apoptosis in both insect and mammalian cells, and apoptotic-like events in a cell free system (Song et al., 1997; Fraser and Evan, 1997). Consistent with their function as apoptotic e€ectors, they are widely expressed throughout development. Whereas no mutations have been reported in the drICE gene, loss of zygotic dcp-1 function causes larval lethality and melanotic tumors, demonstrating that this gene is essential for normal development (Song et al., 1997). Interestingly, no obvious defects in cell death were observed in these animals, presumably due to the maternal component and/or the existence of additional caspases with similar properties. In order to eliminate the maternal contribution, ¯ies carrying germ-line clones homozygous for dcp-1 loss-of-function mutations were generated (McCall and Steller, 1998). Females with mutant germ lines are sterile because of defects in oogenesis. Signi®cantly, this is due to a failure of nurse cells to transfer (`dump') their cytoplasm into the oocyte. Normally during Drosophila oogenesis, nurse cells transfer their cytoplasmic contents and organelles through cytoplasmic bridges into the oocyte and then die (Foley and Cooley, 1998). In the absence of dcp-1 function, this transfer is defective and nurse cell death is delayed, indicating that properly controlled apoptosis of the nurse cells is a necessary event for oocyte development. It is possible that DCP-1 is directly required to alter the actin cytoskeleton in nurse cells. The process of nurse cell dumping is associated with alterations in the actin cytoskeleton which appears to generate the motor force necessary for ecient cytoplasmic transfer (Knowles and Cooley, 1994). In dcp-17 egg chambers, the ®lamentous actin bundles fail to reorganize (McCall and Steller, 1998). Perhaps DCP-1 cleaves and thereby alters the activity of actin-binding proteins. For example, it was recently reported that gelsolin is cleaved and activated by caspase-3 in mammals (Kothokota et al., 1997). The DCP-2/DREDD protease is most similar to caspase-8 but has some unusual properties (Inohara et al., 1997; Chen et al., 1998). Its catalytic site is unique among caspases since it contains a Glutamic acid residue at a position which is normally occupied by a Glycine residue (Chen et al., 1998). This change might impose a novel cleavage speci®city for DCP-2/DREDD and might re¯ect a preference for novel substrates that remain yet to be identi®ed. Surprisingly, DCP-2 is not

Molecular genetics of apoptosis A Bergmann et al

ubiquitiously expressed like most other caspases. Instead it is speci®cally expressed in cells which are doomed to die and requires the cell death genes in the H99 de®ciency for this expression (Chen et al., 1998). In addition, unlike most other caspases, ectopic expression of DCP-2/DREDD is not sucient to induce apoptosis on its own, nor was processing of the protein observed (Chen et al., 1998). Coexpression with RPR or GRIM resulted in processing of DCP-2/ DREDD. However, this processing is insensitive to p35 inhibition and other anti-caspase peptides. Therefore, dcp-2/dredd may encode a novel caspase-like protein with interesting properties. Unfortunately, a dcp-2/ dredd mutant is currently not available which would allow to examine its precise role in vivo. Inhibition of pathological cell death A variety of diseases, including cancer, AIDS, stroke, myopathies and various neurodegenerative disorders, are associated with the abnormal regulation of apoptosis (Thompson, 1995). Consequently, there has been considerable interest in exploring the possibility of using apoptosis-based technologies and drugs for therapeutic purposes. However, it is uncertain whether the prevention of apoptosis in degenerative disorders provides a bene®t for the organism, since rescued cells may no longer be functional. Drosophila o€ers unique opportunities to study the speci®c contribution of apoptosis in disease and provides interesting models for human diseases that are associated with abnormal apoptosis. For example, mutations in rhodopsin or associated proteins can cause retinitis pigmentosa in humans and mice, and a very similar condition in ¯ies (Chang et al., 1993; Kurada and O'Tousa, 1995; Gregory-Evans and Bhattacharya, 1998). In the Drosophila model, it has been shown that the retinal degeneration due to mutations in either rhodopsin or a rhodopsin phosphatase gene occurred by apoptosis and could be blocked by caspase inhibitors (Davidson and Steller, 1998). Signi®cantly, this suppression of apoptosis restored visual function to otherwise blind ¯ies. These ®ndings demonstrate that `undead cells' can serve a useful purpose for the organism, and these results provide a strong rationale to further explore anti-apoptotic therapies in mammalian models for retinitis pigmentosa, as well as other neurodegenerative disorders. In a di€erent model, it has been shown that inactivation of the Drosophila homolog of the tumor suppressor APC (adenomatous polyposis coli) causes retinal neuronal degeneration which results from apoptotic cell death, a condition remarkably similar to that found in humans with germline APC mutations (Ahmed et al., 1998). The degenerating neurons in APC mutants can be rescued by expression of the caspase inhibitor p35 without any obvious change in cell fate (Ahmed et al., 1998). However, it was not demonstrated that the rescued neurons are functional. Finally, Jackson et al. (1998) and Warrick et al. (1998) described Drosophila models for the glutamine repeat neurological disorders Huntington's disease and

Machado-Joseph disease, respectively. In both studies, polyglutamine-expanded alleles of the human genes expressed in the Drosophila compound eye cause massive degeneration of photoreceptor neurons by apoptotic cell death. Although coexpression of the caspase inhibitor p35 failed to inhibit the neuronal degeneration in both systems, these models provide powerful genetic systems for dissecting the neuronal degeneration induced by glutamine-containing proteins. These approaches may reveal potential targets for therapeutic intervention. Final remark Although some components of the cell death machinery, most notably caspases, have been found in C. elegans, insects and mammals, genetic studies of apoptosis in worms and ¯ies have identi®ed apparently distinct pathways and mechanisms for the regulation of caspase activation (Figure 1). While this may re¯ect genuine di€erences in the overall design and control of apoptosis, it is more likely that these apparent discrepancies re¯ect gaps in our current knowledge. Rather, we prefer to believe that both pathways are of central importance for the control of apoptosis. Historically, the regulation of caspases by ced-9/Bcl-2 family proteins has received considerably more attention. However, there are now good reasons to think that the IAP-pathway is equally important, and that both pathways operate together in both vertebrates and invertebrates. There is evidence that Bcl-2 family members can inhibit apoptosis in invertebrate systems (Vucic et al., 1997b) suggesting that insects may have Bcl-2-like genes. Likewise, we expect that mammalian homologs of the Drosophila cell death activators REAPER, HID and GRIM will be identi®ed, since they interact functionally and physically with proteins that are well-conserved in evolution: REAPER, HID and GRIM interact genetically and physically with IAPs, and HID interacts with MAP kinase (Figures 1 and 2). Finally, it appears that many important components of the apoptotic pathway remain to be identi®ed. Future genetic studies of programmed cell death in worms and ¯ies should greatly advance our knowledge in these areas, and should continue to provide the conceptual framework for elucidating the mechanisms and control of apoptosis in humans.

Acknowledgements We would like to thank our collegues in the Steller laboratory for advice and comments, especially Nicholas Haining, Lakshmi Goyal, Zhiwei Song, Lei Zhou, HF Lamblin and K McCall. Andreas Bergmann is supported by post-doctoral fellowships from the Human Frontier Science Program Organization (HFSPO) and the Anna Fuller Fund in Molecular Oncology. Hermann Steller is an investigator of the Howard Hughes Medical Institute (HHMI).

3221

Molecular genetics of apoptosis A Bergmann et al

3222

References Abrams JM, White K, Fessler LI and Steller H. (1993). Development, 117, 29 ± 43. Agapite J and Steller H. (1997). Control of Neuronal Cell Death. In Molecular and Cellular Approaches to Neural Development. Cowan WM, Jessell TM and Zipursky SL (eds). Oxford University Press, New York, pp. 264 ± 289. Ahmed Y, Hayashi S, Levine A and Wieschaus E. (1998). Cell, 93, 1171 ± 1182. Alnemri ES, Livingston DJ, Nicholson DW, Salvesen G, Thornberry NA, Wong WW and Yuan J. (1996). Cell, 87, 171. Ambrosini G, Adida C and Altieri DC. (1997). Nat. Med., 3, 917 ± 921. Antonsson B, Conti F, Ciavatta A, Montessuit S, Lewis S, Maritinou I, Bernasconi L, Bernard A, Mermod JJ, Mazzei G, et al. (1997). Science, 277, 370 ± 373. Bakshi A, Jensen JP, Goldman P, Wright JJ, McBride OW, Epstein AL and Korsmeyer SJ. (1985). Cell, 41, 889 ± 906. Bergmann A, Agapite J, McCall KA and Steller H. (1998). Cell, 95, 331 ± 341. Birnbaum MJ, Clem RJ and Miller LK. (1994). J. Virol., 68, 2521 ± 2528. Bos JL. (1989). Cancer Res., 49, 4682 ± 4689. Bump NJ, Hackett M, Hugunin M, Seshagiri S, Brady K, Chen P, Ferenz C, Franklin S, Ghayur T, Li P, et al. (1995). Science, 269, 1885 ± 1888. Campos AR, Fischbach K-F and Steller H. (1992). Development, 114, 355 ± 366. Chang GQ, Hao Y and Wong F. (1993). Neuron, 11, 595 ± 605. Chen P, Rodriguez A, Erskine R, Thach T and Abrams JM. (1998). Dev. Biol., 201, 202 ± 216. Chen P, Nordstrom W, Gish B and Abrams JM. (1996a). Genes Dev., 10, 1773 ± 1782. Chen P, Lee P, Otto L and Abrams JM. (1996b). J. Biol. Chem., 247, 25735 ± 25737. Chinnaiyan AM, Orth K, O'Rourke K, Duan H, Poirier GG and Dixit VM. (1996). J. Biol. Chem., 271, 4573 ± 4576. Chinnaiyan AM, O'Rourke K, Lane BR and Dixit VM. (1997). Science, 275, 1122 ± 1126. Chu ZL, McKinsey TA, Liu L, Gentry JJ, Malim MH and Ballard DW. (1997). Proc. Natl. Acad. Sci. USA, 94, 10057 ± 10062. Cleary ML, Smith SD and Sklar J. (1986). Cell, 47, 19 ± 28. Clem RJ and Miller LK. (1994). Mol. Cell. Biol., 14, 5212 ± 5222. Conradt B and Horvitz HR. (1998). Cell, 93, 519 ± 529. Crook NE, Clem RJ and Miller LK. (1993). J. Virol., 67, 2168 ± 2174. Davidson FF and Steller H. (1998). Nature, 391, 587 ± 591. Davie EW, Fujikawa K and Kisiel W. (1991). Biochemistry, 30, 10363 ± 10370. Devereaux QL, Takahashi R, Salvesen GS and Reed JC. (1997). Nature, 388, 300 ± 304. Deveraux QL, Roy N, Stennicke HR, van Arsdale T, Zhou Q, Srinivasula SM, Alnemri ES, Salvesen GS and Reed J. (1998). EMBO J., 17, 2215 ± 2223. Dickson B, van der Straten A, Dominguez M and Hafen E. (1996). Genetics, 142, 163 ± 171. Downward J. (1998). Curr. Opin. Genet. Dev., 8, 49 ± 54. Duckett CS, Nava VE, Gedrich RW, Clem RJ, Van Dongen JL, Gil®llan C, Shiels H, Hardwick JM and Thompson CB. (1996). EMBO J., 15, 2685 ± 2694. Ellis RE and Horvitz HR. (1986). Cell, 44, 817 ± 829. Ellis RE, Yuan J and Horvitz HR. (1991). Annu. Rev. Cell Biol., 7, 663 ± 698. Ellis RE and Horvitz HR. (1991). Development, 112, 591 ± 603. Evans EK, Kuwana T, Strum SL, Smith JJ, Newmeyer DD and Kornbluth S. (1997). EMBO J., 16, 7372 ± 7381.

Foley K and Cooley L. (1998). Development, 125, 1075 ± 1082. Fraser AG and Evan GI. (1997). EMBO J., 16, 2805 ± 2813. Golstein P, Marguet D and Depraetere V. (1995). Cell, 81, 185 ± 186. Gregory-Evans K and Bhattacharya S. (1998). Trends Genetics, 14, 103 ± 108. Grether ME, Abrams JM, Agapite J, White K and Steller H. (1995). Genes & Development, 9, 1694 ± 1708. Hay BA, Wassarman DA and Rubin GM. (1995). Cell, 83, 1253 ± 1262. Hengartner MO, Ellis RE and Horvitz HR. (1992). Nature, 356, 494 ± 499. Hengartner MO and Horvitz HR. (1994a). Curr. Opinion Genetics Devel., 4, 581 ± 586. Hengartner MO and Horvitz HR. (1994b). Cell, 76, 665 ± 676. Hengartner MO. (1996). Curr. Opinion Genetics Devel., 6, 34 ± 38. Hockenberry, DM, Oltvai ZM, Yin XM, Milliman CL and Korsmeyer SJ. (1993). Cell, 75, 241 ± 251. Ho€man B and Liebermann DA. (1994). Oncogene, 9, 1807 ± 1812. Inohara N, Koseki T, Hu Y, Chen S and Nunez G. (1997). Proc. Natl. Acad. Sci. USA, 94, 10717 ± 10722. Jackson GR, Salecker I, Dong X, Yao X, Arnheim N, Faber PW, MacDonald ME and Zipursky SL. (1998). Neuron,21, 633 ± 642. Jacobson MD, Weil M and Ra€ MC. (1997). Cell, 88, 347 ± 354. Jiang C, Baehrecke EH and Thummel CS. (1997). Development, 124, 4673 ± 4683. Karim F, Chang HC, Therrien M, Wassarman DA, Laverty T and Rubin GM. (1996). Genetics, 143, 315 ± 329. Kerr JFR, Wyllie AH and Currie AR. (1972). Br. J. Cancer, 26, 239 ± 257. Kim CN, Wang X, Huang Y, Ibrado AM, Liu L, Fang G and Bhalla K. (1997). Cancer Res., 57, 3115 ± 3120. Kluck RM, Bossy-Wetzel E, Green DR and Newmeyer DD. (1997). Science, 275, 1132 ± 1136. Knowles BA and Cooley L. (1994). Trends Genetics, 10, 235 ± 241. Kothakota S, Azuma T, Reinhard C, Klippel A, Tang J, Chu K, McGarry TJ, Kirschner MW, Koths K, Kwiatkowski DJ and Williams LT. (1997). Science, 278, 294 ± 298. Kurada P and White K. (1998). Cell, 95, 319 ± 329. Kurada P and O'Tousa JE. (1995). Neuron, 14, 571 ± 579. Lamblin AF and Steller H. (1998). In press. Li P, Nijhawan D, Budihardjo I, Srinivasula SM, Ahmad M, Alnemri ES and Wang X. (1997). Cell, 91, 479 ± 489. Liston P, Roy N, Tamai K, Lefebvre C, Baird S, ChertonHorvant G, Farahani R, McLean M, Ikeda JE, MacKenzie A and Korneluk RG. (1996). Nature, 379, 349 ± 353. MacCorkle RA, Freeman KW and Spencer DM. (1998). Proc. Natl. Acad. Sci. USA, 95, 3655 ± 3660. Martinou JC, Dubois-Dauphin M, Staple JK, Rodriguiz I, Frankowski H, Missotten M, Albertoni P, Talbot D, Catsicas S, Pietra C and Huarte J. (1994). Neuron, 13, 1017 ± 1030. McCall KA and Steller H. (1997). Trends Genetics, 13, 222 ± 226. McCall KA and Steller H. (1998). Science, 279, 230 ± 234. McCarthy JV and Dixit VM. (1998). J. Biol. Chem., 273, 37 ± 41. McNabb SL, Baker JD, Agapite J, Steller H, Riddiford LM and Truman JW. (1997). Neuron, 19, 813 ± 823. Metzstein MM, Hengartner MO, Tsung N, Ellis RE and Horvitz HR. (1996). Nature, 382, 545 ± 547.

Molecular genetics of apoptosis A Bergmann et al

Miller DT and Cagan RL. (1998). Development, 125, 2327 ± 2335. Minn AJ, Velez P, Schendel SL, Liang H, Muchmore SW, Fesik SW, Fill M and Thompson CB. (1977). Nature, 385, 353 ± 357. Morin PJ, Vogelstein B and Kinzler KW. (1996). Proc. Natl. Acad. Sci. USA, 93, 7950 ± 7954. Muzio M, Stockwell BR, Stennicke HR, Salvesen GS and Dixit VM. (1998). J. Biol. Chem., 273, 2926 ± 2930. Nagata S. (1997). Cell, 88, 355 ± 365. Naik P, Karrim J and Hanahan D. (1996). Genes Dev., 10, 2105 ± 2116. Nicholson WD and Thornberry NA. (1997). Trends Biochem. Sci., 257, 299 ± 306. Nordstrom W, Chen P, Steller H and Abrams JM. (1996). Developmental Biology, 180, 227 ± 241. Oltvai ZN, Millman CL and Korsmeyer SJ. (1993). Cell, 74, 609 ± 619. Oppenheim, RW. (1991). Annu. Rev. Neurosci., 14, 453 ± 501. Ottilie S, Wang Y, Banks S, Chang J, Vigna NJ, Weeks S, Armstrong RC, Fritz LC and Oltersdorf T. (1997). Cell Death Di€., 4, 526 ± 533. Pan G, O'Rourke K and Dixit VM. (1998). J. Biol. Chem., 10, 5841 ± 5845. Pettmann B and Henderson CE. (1998). Neuron, 20, 633 ± 647. Pronk G J, Ramer K, Amiri P and Williams LT. (1996). Science, 271, 808 ± 810. Ra€ MC. (1992). Nature, 356, 397 ± 400. Rao L, Perez D and White E. (1996). J. Cell Biol., 135, 1441 ± 1455. Reed JC. (1995). Curr. Opin. Oncol., 7, 541 ± 546. Reed JC. (1997). Nature, 387, 773 ± 776. Robinow S, Draizen TA and Truman JW. (1997). Dev. Biol., 190, 206 ± 213. Rothe M, Pan MG, Henzel WJ, Ayres TM and Goeddel DV. (1995). Cell, 83, 1243 ± 1252. Roy N, Deveraux QL, Takahashi R, Salvesen GS and Reed JC. (1997). EMBO J., 16, 6914 ± 6925. Roy N, Mahadevan MS, McLean M, Shutler G, Yaraghi Z, Farahani R, Baird S, Besner-Johnston A, Lefebvre C, Kang X, Salih M, Aubry H, Tamai K, Guan X, Ioannou P, Crawford TO, de Jong PJ, Surh L, Ikeda J, Korneluk RG and MacKenzie A. (1995). Cell, 80, 167 ± 178. Salvesen GS and Dixit VM. (1997). Cell, 91, 443 ± 446. Sawamoto K, Taguchi A, Yamada C, Jin M and Okano H. (1998). Cell Death and Di€erentiation, 5, 262 ± 270. Schendel SL, Xie Z, Montal MO, Matsuyama S, Montal M and Reed JC. (1997). Proc. Natl. Acad. Sci. USA, 94, 5113 ± 5118. Selleck S and Steller H. (1991). Neuron, 6, 83 ± 99. Seshagiri S and Miller LK. (1997). Curr. Biol., 7, 455 ± 460. Shaham S and Horvitz HR. (1996). Genes Dev., 10, 578 ± 591. Simon MA, Bowtell DDL, Dodson GS, Laverty TR and Rubin GM. (1991). Cell, 67, 701 ± 716. Song Z, McCall KA and Steller H. (1997). Science, 275, 536 ± 540. Srinivasula SM, Ahmad M, Fernandes-Alnemri T and Alnemri ES. (1998). Mol. Cell., 1, 949 ± 957. Steller H, Fischbach K-F and Rubin GM. (1987). Cell, 50, 1139 ± 1153.

Steller H and Grether ME. (1994). Neuron, 13, 1269 ± 1274. Steller H. (1995). Science, 267, 1445 ± 1449. Steller H. (1998). Proc. Natl. Acad. Sci. USA, 95, 5421 ± 5422. Takahashi A and Earnshaw WC. (1996). Curr. Opin. Genet. Dev., 6, 50 ± 55. Takahashi R, Deveraux Q, Tamm I, Welsh K, Assa-Munt N, Salvesen GS and Reed JC. (1998). J. Biol. Chem., 273, 7787 ± 7790. Tan X and Wang JYJ. (1998). Trends Cell. Biol., 8, 116 ± 120. Thompson CB. (1995). Science, 67, 1456 ± 1462. Thornberry NA, Rano TA, Peterson EP, Rasper DM, Timkey T, Garcia-Calvo M, Houtzager VM, Nordstrom PA, Roy S, Vaillancourt JP, et al. (1997). J. Biol. Chem., 272, 17907 ± 17911. Thress K, Henzel W, Shillinglaw W and Kornbluth S. (1998). EMBO J., 17, 6135 ± 6143. Truman JW, Thorn RS and Robinow S. (1992). J. Neurobiol., 23, 1295 ± 1311. Tsujimoto Y, Gorham J, Cossman J, Ja€e E and Croce CM. (1985). Science, 229, 1390 ± 1393. Ullrich A and Schlessinger J. (1990). Cell, 61, 203 ± 212. Uren AG, Pakusch M, Hawkins CJ, Puls KL and Vaud DL. (1996). Proc. Natl. Acad. Sci. USA, 93, 4974 ± 4978. Vander Heiden MG, Chandel NS, Williamson EK, Scumacher PT and Thompson CB. (1997). Cell, 91, 627 ± 637. Vucic D, Kaiser WJ, Harvey AJ and Miller LK. (1997a). Proc. Natl. Acad. Sci. USA, 94, 10183 ± 10188. Vucic D, Seshagiri S and Miller LK. (1997b). Mol. Cell. Biol., 17, 667 ± 676. Vucic D, Kaiser WJ, Harvey AJ and Miller LK. (1998). Mol. Cell Biol., 18, 3300 ± 3309. Warrick JM, Paulson HL, Gray-Board GL, Bui QT, Fischbeck KH, Pittman RN and Bonini NM. (1998). Cell, 93, 939 ± 949. White E. (1996). Genes Dev., 10, 1 ± 15. White K, Grether M, Abrams J, Young L, Farrell K and Steller H. (1994). Science, 264, 677 ± 683. White K, Tahaoglu E and Steller H. (1996). Science, 271, 805 ± 807. Wu D, Wallen HD, Inohara N and Nunez G. (1997). J. Biol. Chem., 272, 21449 ± 21454. Wyllie AH, Kerr JFR and Currie AR. (1980). Int. Rev. Cytol., 68, 251 ± 306. Xue D and Horvitz HR. (1995). Nature, 377, 248 ± 251. Yang J, Liu X, Bhalla K, Kim CN, Ibrado AM, Cai J, Peng TI, Jones DP and Wang X. (1997). Science, 275, 1129 ± 1132. Yang X, Chang HY and Baltimore D. (1998). Molecular Cell, 1, 319 ± 325. Yuan J and Horvitz HR. (1992). Development, 116, 309 ± 320. Yuan J, Shaham S, Ellis HM and Horvitz HR. (1993). Cell, 75, 641 ± 652. Zou H, Henze, WJ, Liu X, Lutschg A and Wang X. (1997). Cell, 90, 405 ± 413. Zhou L, Schnitzler A, Agapite J, Schwartz LM, Steller H and Nambu JR. (1997). Proc. Natl. Acad. Sci. USA, 94, 5131 ± 5136.

3223