|Home | About | Journals | Submit | Contact Us | Français|
Programmed cell death, or apoptosis, is a highly conserved cellular process that has been intensively investigated in nematodes, flies and mammals. The genetic conservation, the low redundancy, the feasibility for high-throughput genetic screens and the identification of temporally and spatially regulated apoptotic responses make Drosophila melanogaster a great model for the study of apoptosis. Here, we review the key players of the cell death pathway in Drosophila and discuss their roles in apoptotic and non-apoptotic processes.
Apoptosis, the major form of programmed cell death (PCD), is a physiological cell suicide process characterized by cell shrinkage, nuclear fragmentation, chromatin condensation and DNA fragmentation.1 As a type of non-traumatic cell death, apoptosis neither releases cell contents nor causes inflammation. Apoptosis is essential for normal development and homeostasis of metazoan organisms. It helps sculpt our bodies, removes unwanted cells and eliminates potentially dangerous cells.2 Misregulation of apoptosis is linked to many developmental defects and diseases such as tumor formation, autoimmune diseases and neurodegenerative disorders.3
PCD plays indispensable roles through most developmental stages in Drosophila. The earliest PCD starts at embryonic stage 11 and becomes widespread and prominent thereafter.4 The temporal and spatial pattern of embryonic cell death matches certain developmental events during embryogenesis such as dorsal closure, head involution, germ band retraction and central nerve cord condensation. Apoptosis near each segment border helps establish the precise pattern of segmentation.5 In eye imaginal discs, cell death occurring between 24 and 30 hours after pupal formation (APF) is required to remove supernumerary interommatidial cells to establish the precise hexagonal lattice in the Drosophila eye.6,7 During metamorphosis, PCD removes larval structures that are no longer needed, such as hindgut, salivary glands and larval muscles.8 PCD removes epithelial cells during maturation of the wings in newly eclosed flies.9,10 PCD eliminates nurse cells in late oogenesis and entire egg chambers under starvation conditions in mid-oogenesis.11-14 Apoptosis also plays a role in removing mis-specified cells during development.15
The components of the apoptotic pathway are conserved from flies to humans (Fig. 1). Caspases are the main executioners of the apoptotic process. These are Cys-proteases that cleave specifically after Asp residues16 (although exceptions exist17). In Drosophila, there are seven caspase genes.16 Critical for the apoptotic response are Dronc,18 a Caspase-9-like initiator caspase, and Dcp-119 and DrICE,20,21 caspase-3-like effector caspases (Fig. 1 and see below). In the absence of apoptotic signals, inhibitor of apoptosis proteins (IAPs) such as Drosophila IAP1 (Diap1)22 bind to caspases and inhibit their activity23,24 (Figs. 1 and and2A).2A). In the presence of apoptotic stimuli, the pro-apoptotic proteins Reaper,25 Hid26 and Grim27 (RHG proteins) bind to Diap1,24,28,29 and release the caspases. This release is largely mediated through ubiquitin-mediated degradation of Diap130-32 (Fig. 2B). Released Dronc binds to the adaptor protein Ark to form the apoptosome.18,33-36 The active apoptosome subsequently cleaves and activates the effector caspases Dcp-1 and DrICE17 (Figs. 1 and and2B).2B). These caspases then trigger a series of downstream cleavage reactions resulting in removal of apoptotic cells.
In the following, we describe the genetic approaches that were utilized to identify and characterize these genes. We will discuss their mechanistic interactions and also present their involvement in non-apoptotic processes.
Three different types of systematic genetic screens for identification of cell death genes have been performed in the past. In the first screen, performed by Steller and colleagues, homozygous deficiencies were screened for loss of PCD in embryos. This approach resulted in the identification of the H99 deficiency, which deletes approximately 300 kb of genomic DNA in the 75C1 region on chromosome arm 3L.25 Homozygous H99 mutant embryos lack all developmental cell death and block most irradiation-induced cell death. Subsequent analysis revealed that the H99 region contains the genes reaper (rpr), head involution defective (hid, also known as Wrinkled) and grim, collectively referred to as the RHG genes which are required for apoptosis.25-27 Overexpression of any of these genes is sufficient to induce cell death. The RHG genes encode proteins without any enzymatic activity. They contain a stretch of 12 partially conserved residues at the N-terminus, called the RHG motif37 (Fig. 3A) (or IBM for IAP-binding motif), which is essential for the function of these proteins (see below). Two mammalian proteins, Smac and HtrA2, also contain an IBM and function similarly to the RHG proteins (Fig. 1).38-43
Subsequently, three additional RHG genes in Drosophila were identified. jafrac2 was isolated as an IAP-interacting protein,44 sickle (skl) was identified using database searches for proteins with an RHG motif,45-47 and dOmi/HtrA2 was cloned based on sequence similarity to human Omi/HtrA2.48-50 These proteins also contain an RHG motif (Fig. 3A) and overexpression induces ectopic cell death.
Although reaper, hid and grim show many similarities in inducing apoptosis, they are not functionally identical. hid is essential for cell death during eye development51 and is directly inhibited by Ras/MAPK.52,53 Compared to hid or reaper, grim is more effective at inducing cell death in the embryonic central nervous system.27 Apoptosis in post-embryonic neuroblasts specifically requires the function of reaper.54 Ectopic expression of sickle can induce apoptosis in embryos and enhance reaper- and grim-induced apoptosis in the Drosophila eye.46,47 Recent reports suggest that dOmi/HtrA2, a Drosophila homolog of a mitochondrial serine protease, displaces Dronc from the BIR2 domain of Diap1 and promotes cell death through degradation of Diap1 in the vicinity of mitochondria.48,49 These and other reports55,56 suggest that mitochondria may play a role in apoptosis in Drosophila.
It has been observed that Drosophila embryos fail to induce apoptosis in response to X-ray irradiation after embryonic stage 12. This was found to be due to epigenetic regulation of an irradiation-responsive enhancer region (IRER) located upstream of rpr, which is required for the expression of Rpr and Hid in response to irradiation.57 The IRER becomes enriched for trimethylated Lysine 9 and Lysine 27 on histone H3 (H3K9 and H3K27, respectively) and forms a heterochromatin-like structure during the sensitive-to-resistant transition. This study provides evidence for epigenetic regulation of pro-apoptotic genes in Drosophila.57 In addition, it has been shown that pro-apoptotic genes are also subject to post-transcriptional regulation by microRNAs.58,59
The second type of cell death screen was a modifier screen taking advantage of the eye-ablation phenotype caused by expression of reaper or hid under eye-specific GMR control (GMR-reaper, GMR-hid) (Fig. 4A). This type of genetic screening resulted in the isolation of mutations in the IAP (inhibitor of apoptosis proteins) genes diap1 and dBruce.22,29,60-62
IAPs are conserved from yeast to humans. The first IAP was found in Baculovirus.63 Although not all IAPs inhibit cell death, anti- apoptotic IAPs protect cells from PCD through binding to and inhibiting caspases.64 The interaction with caspases is mediated by BIR (baculovirus IAP repeat) domains of which IAPs contain between one and three (Fig. 3B). The BIR domain also interacts with the RHG motif of the RHG proteins.24 Some IAPs contain a C-terminal RING (really interesting new gene) domain (Fig. 3B), conferring E3-ubiquitin ligase activity to IAPs65 (Fig. 2; see also The Role of Ubiquitylation of Control of PCD). The RING domain targets RHG proteins,66,67 caspases,61,68-70 and IAPs themselves30-32,71-73 for ubiquitylation (Fig. 2) (reviewed in ref. 74).
The Drosophila genome encodes four IAP family members: Diap1, Diap2, dBruce and Deterin (Fig. 3B). Diap1 is the most important anti-apoptotic IAP in Drosophila. The first diap1 mutant recovered was the thread1 (th1) allele. th1 is a weak allele-specific mutation that blocks branching of the antenna, hence the name. This phenotype is likely caused by a non-apoptotic function of diap1 (see section Non-Apoptotic Functions of Caspases). Subsequently, strong loss-of-function alleles of diap1 were recovered as dominant enhancers of GMR-reaper and GMR-hid.22,29 The enhancer alleles correspond to loss-of-function alleles. When homozygous, these diap1 mutants display global embryonic cell death in the embryo due to increased caspase activity, demonstrating that Diap1 is an essential anti-apoptotic protein in Drosophila.28,29,60 Diap1 contains two BIR domains (Fig. 3B) which bind to different caspases and RHG proteins24 (Fig. 2A and B). The loss-of-function mutations are present in either of the BIR domains and affect the interaction with caspases. Thus, the encoded mutant proteins are unable to bind and inhibit caspases, leading to inappropriate caspase activation and unrestrained cell death.
Interestingly, two classes of gain-of-function alleles of diap1 have been recovered. In contrast to loss-of-function alleles, which are enhancers of GMR-reaper- and GMR-hid-induced eye phenotypes, the class I gain-of-function alleles were recovered as suppressors of both GMR-reaper and GMR-hid. Class I diap1 alleles also affect the BIR domains.29 However, in contrast to the loss-of-function alleles, in which the BIR mutation affects the interaction with caspases (see above), the BIR domain mutations of class I gain-of-function alleles cause loss of the interaction with the RHG proteins, rendering the mutant proteins resistant to RHG-induced apoptosis.29
The class II gain-of-function alleles of diap1 affect its C-terminally located RING domain (Fig. 3B) which carries an E3-ubiquitin ligase activity65 (see also The Role of Ubiquitylation of Control of PCD). In living cells, the RING domain ubiquitylates the caspases Dronc and DrICE,61,70,75 ensuring cell survival (Fig. 2A). However, in dying cells, binding of the RHG proteins to Diap1 changes the substrate specificity of the RING domain, leading to its own (auto-) ubiquitylation and degradation.30-32 Class II gain-of-function alleles behave genetically differently than class I alleles. They were recovered as suppressors of GMR-hid, but act as enhancers of GMR-reaper.60,61 In addition, class II alleles behave as gain-of-function alleles only in a heterozygous condition in GMR-hid background; when homozygous, they act like loss-of-function alleles.
Another IAP family member in Drosophila is dBruce, which contains a single BIR domain and an E2-ubiquitin-conjugating (Ubc) domain (Fig. 3B). dBruce appears to specifically regulate rpr- and grim-, but not hid-dependent cell death.62 Steller's laboratory recently reported that dBruce acts as a substrate for the Cullin-3-based E3-ligase complex, which is required for caspase activation during spermatid individualization.76 It was also reported that dBruce serves as a negative regulator of autophagy during nutrient-rich times.77 Autophagy is a “self-eating” process that features the engulfment of part of the cytoplasm inside double-membrane vesicles called autophagosomes.78,79
Although overexpression of Diap2 also suppresses GMR-reaper and GMR-hid,22 diap2 mutants fail to show any detectable phenotype in developmental apoptosis.80 Instead, it was reported that Diap2 is required for NF-kB-related innate immune responses in Drosophila.80-83 However, Meier's group showed that diap2 mutants respond very sensitively to mild stress.84 They also reported that Diap2 specifically binds to DrICE and controls DrICE activity.
Deterin is a Survivin-like IAP member in Drosophila.85 The function of Deterin is unknown due to the lack of mutants.
Surprisingly, in the aforementioned genetic screens, mutant alleles of caspases and the adaptor protein ark (see section Adaptor Protein Ark) were not recovered. There are a number of possible reasons that would explain this. Because there are seven caspase genes in Drosophila16 (see section Caspases), the mutational loss of one caspase may be compensated for by another caspase; thus, there may be functional redundancy. Furthermore, the embryonic deficiency screen will only identify deficiencies deleting genes for which there is an early zygotic requirement that is not fulfilled by maternally-expressed protein. As was determined later, all caspase genes and the ark gene have a strong maternal contribution, explaining the failure to identify these genes in the embryonic deficiency screen. Moreover, a prerequisite of the modifier screen is that a 50% reduction in the dose of a gene is sufficient to visibly modify the GMR-hid and/or GMR-reaper small eye phenotype. This applies to diap1 and dBruce (and several other genes not discussed here). However, a 50% reduction of the dose of a caspase gene may not be sufficient to modify GMR-reaper and/or GMR-hid, and a stronger reduction is necessary to recover mutants in these genes. Therefore, a third type of genetic screening was performed in which homozygous mutant clones obtained by the ey-FLP/FRT-system were induced in the GMR-hid background (referred to as GheF screen for GMR-hid/ey-FLP).86 This approach indeed resulted in the isolation of dronc, drICE and ark alleles as strong recessive suppressors of GMR-hid86-88 (Fig. 4). Other genes recovered in the GheF screens, such as vps25,89 vps36,90 and D-cbl,91 are only indirectly involved in the control of cell death and thus are not discussed in this review.
Ark (Apaf-1-related killer, also known as Dark, D-Apaf-1 and Hac-1), is the Drosophila homolog of mammalian Apaf-1 and C. elegans CED-4.33-35,92 33 mutant alleles of ark were recovered in the GheF screen.88 These alleles were very informative for a genetic structure/function analysis. Ark is an essential pro-apoptotic protein; most cell death is blocked in ark mutants (Fig. 5D).88,93,94 Ark, like Apaf-1, contains a CARD (caspase activation and recruitment domain) domain, a NOD (nucleotide oligomerization domain) domain, and WD40-repeats (Fig. 3E). Unlike Apaf-1, Ark contains a unique C-terminal domain of 180 residues. Ark provides the structural backbone of the apoptosome, a large protein complex required for apoptosis. 16 Ark proteins form the apoptosome, organized in two stacked rings of eight subunits each.36 The apoptosome has a wheel-shaped architecture with a central CARD ring as the hub and the WD40 repeats as the spokes.36 This arrangement is different from the mammalian Apaf-1 apoptosome which forms a single wheel structure composed of seven subunits of Apaf-1.95 In addition, cytochrome c, which is required for assembly of the Apaf-1 apoptosome, does not seem to be required for the formation of the Ark apoptosome.36,96,97 Another difference concerns the function of the WD40 domains. In Apaf-1, the WD40 repeats are auto-inhibitory and block the interaction of the CARD domains of Apaf-1 and Caspase-9.98,99 Binding of cytochrome c to the WD40 repeats releases the CARD of Apaf-1 and allows interaction with the CARD of Caspase-9. Because Cytochrome c is not essential for apoptosis in Drosophila, the WD40 repeats of Ark may have a different function. Indeed, mutations affecting the WD40 domain cause a loss-of-function phenotype,88 suggesting that the WD40 domain is pro-apoptotic in this organism. Cryo-EM studies suggest that the WD40 repeats may be required to stabilize the Ark apoptosome.36 Interestingly, many of the ark mutations recovered in the GheF screen affect the NOD domain.88 These mutations may also affect apoptosome formation, as Cryo-EM studies indicate that the NOD domain is necessary for formation of the double ring in the Ark apoptosome.36 This suggests that the NOD domain is essential for the pro-apoptotic activity of Ark due to its role in the formation of the apoptosome.92
Caspases are synthesized as inactive zymogen precursor proteins consisting of a prodomain, and large and small catalytic subunits (Fig. 3C) (reviewed in ref. 16). Caspases can be classified into two types: initiator (apical) caspases and effector (executioner) caspases.16 Initiator caspases contain a long N-terminal prodomain which harbors protein-protein interaction domains such as the CARD and DED (death effector domain) (Fig. 3C). Effector caspases have a small prodomain (Fig. 3C). Upon activation, proteolytic cleavage separates the small from the large subunit, and often cleaves off the prodomain.16 Two large and two small subunits form the active protease (Fig. 3D).
The Drosophila genome encodes seven caspase genes: dronc, dredd, strica, drICE, dcp-1, decay and damm18-20,100-104 (reviewed in ref. 16). Based on the length of the prodomain, Dronc, Dredd and Strica qualify as initiator caspases (Fig. 3C). The other four are candidate effector caspases. In the following, we discuss the individual caspases.
Dronc is an essential initiator caspase in Drosophila. As the only fly caspase containing a CARD domain, Dronc interacts with Ark through a CARD/CARD interaction.18 Thus, dronc is functionally most similar to caspase-9, an essential human initiator caspase (Fig. 1). Consistently, Dronc can function as an initiator caspase to cleave and activate the effector caspase DrICE in vitro.17
dronc mutants recently became available from the GheF screen and from imprecise P-element excisions.86,105-109 dronc mosaics suppress GMR-hid and GMR-reaper-induced cell death completely and produce a normal eye86,106 (Fig. 4D). However, homozygous dronc mutations cause pupal lethality, suggesting that dronc is essential for normal development. Only at a very low rate (less than 1%) are homozygous mutant dronc adults recovered, but they die within 2−3 days after eclosion.86 Downregulation of dronc causes decreased cell death during development.23,110 dronc mutant embryos derived form germline clones lack most, but not all, developmental cell death (Fig. 5B) and contain additional cells in the nervous system.86 These studies demonstrate that dronc is essential for most PCD from embryonic stages until adult stages. dronc is also essential for irradiation-induced apoptosis. Furthermore, dronc mutations revert the ovary degeneration defect induced by loss of diap1, which indicates that dronc genetically acts downstream of diap1.86 dronc was reported to be required for ecdysone-induced cell death of salivary glands, but not for death of the larval midgut.105,106 However, a later report demonstrated that dronc is required for the proper onset of cell death in the larval midgut beginning at late third instar in response to a higher dose of ecdysone.107
Recent studies on the structure of Dronc protein and the Ark apoptosome provided more information about Dronc activation.36,111-113 Yan et al. showed that autocatalytic cleavage after Glu352 of Dronc strongly promotes its dimerization and its catalytic activity.111 Dronc forms a stable dimer in solution after autocatalytic cleavage at Glu352. The mutant DroncE352A (Glu352 is replaced by Ala), which cannot be cleaved, remains a monomer in solution as does the wild-type zymogen. The structure of the Dronc zymogen reveals an unproductive active site conformation, consistent with the idea that Dronc must be cleaved to become active.111 However, it is not known whether autocatalytic cleavage of Dronc occurs intramolecularly or in trans between transiently dimerized zymogens.
Nevertheless, Dorstyn et al. recently reported that auto-cleavage of Dronc is not essential for its initial activation and its catalytic activity to DrICE.112 They determined that Dronc autoprocessing occurs at E352 between the large and small subunits and also at D135 (Asp135) following the CARD domain. Dronc mutants that are uncleavable at these sites still present catalytic activity towards the substrates and are still apoptotic.112 A more recent in vitro study also found that cleavage of Dronc at Glu352 is neither necessary nor sufficient for Dronc activation, although cleavage was shown to stabilize the active Dronc dimer,113 which supports the notion of Dorstyn's work.112
Dorstyn et al. also found that Ark alone is not sufficient to activate Dronc and that an additional factor can significantly enhance Ark-mediated Dronc activitation.112 The identification of this factor would be interesting because cytochrome c, which is crucial for activation of human caspase-9, does not seem to be important for formation of the Ark apoptosome and apoptotic function in Drosophila.96,97
It is conceivable that cells need to promptly and irreversibly trigger caspase activity to perform self-killing upon apoptotic stimuli. However, a recent report from the Ryoo group suggested a novel negative regulatory loop between Dronc and Ark, which may be useful for cells when non-apoptotic processes (see section Non-Apoptotic Functions of Caspases) require moderate levels of caspase activity.114 They found that Dronc and Ark mutually suppress each other's protein levels in vivo. Ectopic expression of either one will result in decreased protein levels of the other one. In addition, the mutual suppression between Dronc and Ark is promoted by Diap1, probably through its RING activity.114 This is an interesting and novel finding for regulation of the apoptosome in Drosophila. It also raises some questions as to how strong this feedback-inhibition would be in normal tissue without overexpression of Dronc or Ark, and whether this inhibition accounts for the moderate caspase activity present during caspase-dependent non-apoptotic processes (see section Non-Apoptotic Functions of Caspases).
Genetic characterization has suggested that DrICE is an important (probably the most important) apoptotic effector caspase in Drosophila. As an effector caspase, DrICE does not contain a long N-terminal prodomain. The initiator caspase Dronc can cleave and activate DrICE.17,23 DrICE also cleaves Dronc in a positive feedback loop to amplify caspase activity.115 It was reported that DrICE is required for apoptosis in S2 cells.20,21,115 It is upregulated upon ecdysone induction116,117 and RNAi-mediated depletion of drICE delays salivary glands removal,82 which suggest that drICE is necessary for efficient cell death induced by ecdysone. In contrast, drICE mutants do not seem to affect normal cell death in salivary glands, possibly because of redundancy in vivo.118 However, recent investigations of drICE mutants indeed prove that drICE is an essential apoptotic caspase. drICE mutations suppress hid-, rpr- and grim-induced apoptosis87,109,118 (Fig. 4E), as well as apoptosis induced by loss of diap1,87,109,118 which confirms that drICE acts downstream of these factors. Cell death is decreased, but not completely abolished, in drICE mutant embryos (Fig. 5E), larval and pupal tissues. drICE mutations also alleviate irradiation-induced apoptosis. Pupae mutant for the null allele drICEΔ1 possess many abnormal masses in the head, abdomen, wing disc and leg disc.118 drICEΔ1 adults show developmental defects in aristae, genitalia, analia and abdomen.118 Moreover, DrICE also functions in cell death during oogenesis11,13,14,119 and the non-apoptotic process of sperm individualization118 (for details see section Non-Apoptotic Functions of Caspases).
A recent study showed that Diap2 seems to play a role in inhibiting DrICE activity through both physical binding and ubiquitylation.84 As reported, Diap2 can bind to DrICE through its BIR3 domain. Asp100 is one of the cleavage sites of Diap2 processed by caspases (possibly by DrICE). The cleavage process is essential for Diap2 to bind to DrICE. Diap2 also robustly ubiquitylates DrICE in vivo. The BIR3 domain, the RING domain and Asp100 of Diap2 are required for inhibition of DrICE-mediated cell death.84
Dcp-1, another effector candidate caspase, possesses catalytic activity and can induce DNA fragmentation in HeLa cells.19 dcp-1 plays an essential role in nurse cell deaths during mid-oogenesis under condition of starvation.11 However, dcp-1 might not be as crucial as dronc and drICE for normal development since dcp-1 null mutants are viable without apparent adult defects.11,87,109,118
dcp-1 and drICE seem to function redundantly in some cells because dcp-1 and drICE double mutant embryos show further reduction of cell death compared to drICE single mutant embryos87 (Fig. 5F). Consistently, while the single mutants are homozygous viable,11,87,118 dcp-1 drICE double mutants are pupal lethal, similar to dronc mutants.87 The redundancy of dcp-1 and drICE has been reported for PCD during embryogenesis (Fig. 5F), PCD in larval eye disc, aristae development, cell death during late oogenesis and viability of animals.14,82,87,109,118
Strica, a candidate initiator caspase, contains a unique long serine- and threonine-rich prodomain101 (Fig. 3C). Overexpression of strica can induce cell death in Drosophila S2 cells. Although Strica-induced cell death could be downregulated by Diap1, Strica physically interacts with Diap2.101 In a systematic RNAi screen, strica (as well as ark, dronc and drICE) seems to be required for the timely removal of larval salivary glands during Drosophila metamorphosis.82 However, persistence of larval salivary glands is not observed in mutants homozygous for the null allele strica4.14 During oogenesis, it has been suggested that Strica acts redundantly with Dronc14 (see section Non-Apoptotic Functions of Caspases). Defects during mid- and late oogenesis appear only in dronc strica double mutants but not in either single mutant.14 Conversely, RNAi knock-down of both dronc and strica presents stronger suppression for GMR-hid than knocking down either single gene.82
Dredd contains two DED (death effector domain) domains in the N-terminal prodomain (Fig. 3C) and is most similar to human caspase-8.102 Expression of rpr, hid or grim increases the level of dredd mRNA in apoptotic cells, and dredd mutations dominantly suppress rpr-, hid- or grim-induced cell death.102 However, Dredd seems to be more involved in a non-apoptotic process, the activation of the innate immune response after infection by Gram-negative bacteria120,121 (see section Non-Apoptotic Functions of Caspases). It has been reported that Dredd is required for cleavage and activation of Relish,122 one of the NF-kB family members, which is essential for the expression of antimicrobial genes.123 Moreover, dredd mutants also show defects in spermatid individualization124 (see section Non-Apoptotic Functions of Caspases).
Homozygous decay mutants are viable and fertile without any obvious morphological abnormalities, and developmental apoptosis is normal in the mutants.109 This may be due to redundancy with strica, as RNAi assays show that decay and strica might act redundantly downstream of hid-induced cell death.82
Currently, the only information about damm, an effector caspase, is that ectopic expression of damm in the Drosophila eye results in a rough eye phenotype.104
Ubiquitylation refers to the covalent attachment of ubiquitin (Ub), a 76 amino acid polypeptide, to target proteins. Historically, ubiquitylation has been mostly studied as a process that targets proteins for degradation by the 26S proteasome.125 However, in the last decade, non-degradative functions of ubiquitylation involved in endosomal trafficking, chromatin modification, transcription, DNA repair and NFκB signaling have been uncovered.125 Ubiquitylation also plays a critical role in the cell death pathway. Interestingly, it serves both pro-apoptotic and anti-apoptotic roles.
Ub conjugation is a stepwise process requiring the activity of three different types of enzymes.125 Free Ub is covalently bound to and activated by an E1 ubiquitin-activating (Uba) enzyme. The Drosophila genome contains only one ubiquitin E1, termed Uba1.126 From the E1, Ub is transferred to E2 ubiquitin-conjugating (Ubc) enzymes. Hect (homology to E6 AP C-terminus) ligases and RING ligases are the two main types of E3 ubiquitin-ligases. Hect ligases form a covalent intermediate with Ub as it is transferred from the E2 to the substrate. RING ligases do not covalently bind Ub unless they are targeted for degradation themselves (see below). RING ligases mediate the transfer of the E2 enzyme to the substrate. Several IAPs contain a RING ubiquitin-ligase domain (see section Dominant Modifier Screens: Identification of IAPs in Drosophila and Fig. 3B). Ub is attached to substrates as either a single protein (mono-Ub) or as a poly-Ub chain of monomers that are linked to each other at lysine (K) residues at position 48 (K48) or 63 (K63) of the ubiquitin moiety.127 K48 poly-Ub linkages target substrates for protein degradation via the 26S proteasome, while mono-Ub and K63 poly-Ub linkages are involved in a variety of non-degradative processes.127
The first ubiquitylation substrate of Diap1's RING domain was found to be Diap1 itself.30-32 This ubiquitylation step is triggered by the RHG proteins in dying cells and induces proteasome-mediated degradation of Diap1.30-32 In this capacity, ubiquitylation is pro-apoptotic. UbcD1 has been identified as the E2 enzyme involved in Diap1 turnover.30,72 Another factor involved in Diap1 turnover is Morgue.71,73 Morgue contains a UEV (ubiquitin-conjugating enzyme variant) domain and an F-box. The UEV domain is an E2 domain, but it lacks a critical cysteine residue in the active site. The F-box is a domain found in many Cullin-RING E3 ligases. The precise role of Morgue is unclear, but morgue mutants accumulate Diap1 protein, suggesting that Morgue is required for Diap1 turnover.71,73
Further support for the importance of ubiquitin-mediated degradation of Diap1 for PCD comes from the observation that mild reduction in the total levels of activated Ub protects cells from PCD. This was observed in mutants carrying weak alleles of the sole E1 ubiquitin-activating enzyme, Uba1, which also exhibited increased levels of Diap1.128,129 However, complete loss of Uba1, and thus complete loss of ubiquitin-conjugation, causes a different phenotype (see below).
While ubiquitylation leads to proteasome-mediated degradation of Diap1,130,131 the role of ubiquitylation for caspase regulation is less clear. Before the RING domain was identified as a ubiquitin ligase, it was believed that binding of the BIR domains of IAPs was sufficient for caspase inhibition.132 However, this model does not explain the strong apoptotic phenotype of class II diap1 alleles, which affect the RING domain (Dominant Modifier Screens: Identification of IAPs in Drosophila) but maintain the Diap1/Dronc interaction,23,60 implying that binding of Diap1 to Dronc is not sufficient for Dronc inhibition. It was then shown that the RING domain of Diap1 can indeed ubiquitylate Dronc61,70 and DrICE,75 suggesting that ubiquitylation may play a critical role for caspase regulation.
Complete loss of Uba1 function, and thus complete loss of ubiquitin conjugation, also leads to inappropriate Dronc activation and cell death.128,129 In strong Uba1 mutant clones, even overexpression of Diap1 cannot block Dronc activation (Lee TV and Bergmann A, unpublished). This supports the idea that binding of Diap1 to Dronc is not sufficient for Dronc inhibition and that ubiquitylation of Dronc is required for its inactivation.
However, it is unclear whether ubiquitylation of Dronc promotes its proteasomal degradation. Convincing in vivo evidence for proteasome-mediated degradation of ubiquitylated Dronc as an anti-apoptotic mechanism has not been provided to date. In fact, we have analyzed proteasome mutants to address this question. If Dronc is subject to proteasome-mediated degradation, then it should accumulate in proteasome mutants. However, while Diap1 protein clearly accumulates in two different proteasome mutants, Dronc protein does not (Lee TV and Bergmann A, unpublished). This is a surprising observation, as Dronc is a fairly short-lived protein with a half-life of 3 hours.31 Thus, Dronc may be degraded by a different mechanism, such as lysosomal degradation or compensatory autophagy.133 In any case, ubiquitylation of Dronc in living cells may serve a non-degradative mechanism. At this time, it is unclear how ubiquitylation inhibits Dronc activity. However, in Uba1 and diap1ΔRING mutants, Dronc protein is processed into the catalytic subunits in the absence of an apoptotic signal suggesting that ubiquitylation inhibits processing of Dronc (Lee TV and Bergmann A, unpublished). In addition, Diap1 has been shown to ubiquitylate Dronc in a unique conformation that is not recognized by the proteasome.134 This indicates that ubiquitylation negatively regulates Dronc function independent of the proteasome.
While these observations have been made for unprocessed monomeric Dronc in living cells, the mechanism regulating Dronc levels is altered when Dronc is processed and present in the apoptosome.135 K48-linked poly-Ub chains are observed on apoptosome-associated Dronc, which is thus targeted for protein degradation.114 This may be a mechanism to quickly remove any inappropriately processed Dronc caspase or to keep Dronc activity low in non-apoptotic processes (see section Non-Apoptotic Functions of Caspases) when high caspase activity is not desired.
In addition to inhibiting Dronc function, Ub conjugation also functions to negatively regulate effector caspases. In mammals, XIAP has been shown to polyubiquitylate Caspase-3.69,136 In vitro studies have shown that Caspase-3 and Caspase-7 are mono-ubiquitylated by cIAP2.68 Meanwhile in Drosophila, DrICE is resistant to inactivation by Diap1 if it is unable to be ubiquitylated. Using a diap1ΔRING mutant or a mutant form of DrICE that cannot be ubiquitylated, it was recently shown that DrICE is regulated by Diap1-mediated ubiquitylation. Interestingly, any of the nine Lysine residues on the protein surface of DrICE can be sites of Ub-conjugation, and all of these sites have to be mutated to render DrICE resistant to Diap1 inactivation.75 Intriguingly, while necessary, the RING domain of Diap1 is not sufficient to inhibit effector caspases. For degradation of DrICE, Diap1 has to recruit an additional E3-ligase that functions in the N-end rule pathway to degrade DrICE.75,137 Thus, Diap1 is using two distinct Ub-conjugation systems to effectively inhibit effector caspases. Furthermore, Diap2 functions as an E3-ligase, too. Mutants for diap2 contain increased DrICE activity and, after genotoxic stress, are more sensitive to cell death.84,96 This suggests that living cells may use multiple methods to ubiquitylate DrICE to prevent its activation.
However, much like Dronc ubiquitylation, ubiquitylation of DrICE does not target it for protein degradation but rather inhibits DrICE function by a non-degradative mechanism.75 There are several models describing how non-degradative ubiquitylation can prevent caspase activation and cell death. One possibility is that the Ub chains can sterically block the caspase catalytic site, thereby preventing substrate interaction. Poly-Ub chains may also induce allosteric conformation impairment of DrICE's catalytic pocket, resulting in decreased function. It has been proposed that Ub-conjugation acts as a “mixed” inhibitor displaying both competitive and non-competitive properties.75 Another possible explanation is that the addition of mono- or polyubiquitin chains prevents the dimerization of caspases, thus preventing their activation. Recent analyses have shown that in the case of Dronc, dimerization is required for activation.113 Finally, the addition of ubiquitin onto caspases may promote interaction with Ub-binding domain (UBD) containing proteins, named Ub receptors.127 It is possible that these Ub-receptors may sequester caspases within the cell or prevent their processing, thereby inhibiting their activation and thus, cell death.
PCD performs indispensable functions during oogenesis. There are two types of cell death during oogenesis. One is developmental cell death occurring in nurse cells during late oogenesis. At the late stages of oogenesis, nurse cells in each egg chamber, which contains one oocyte and 15 nurse cells, transport (dump) their cytoplasmic contents into the oocyte. The dumping process ends as neighboring cells engulf the empty nurse cells.138
Developmental cell death during late oogenesis resembles apoptosis, as the nurse cell nuclei stain positively for the traditional apoptotic markers Acridine Orange and TUNEL.12,139,140 Remarkably, germline overexpression of the caspase inhibitors p35 and Diap1 only partly blocks nurse cell death in late oogenesis, showing that this process of developmental PCD can occur in a caspase-independent manner.13,14,119 It is also possible that cell death in late oogenesis may depend on caspases not inhibited by Diap1 and p35 or that other mechanisms may exist to degrade Diap1 and p35 even when they are overexpressed. Along the same lines, this cell death process is not affected in flies that are single mutant for any of the caspases dredd, dronc, dcp-1 or strica, or even for the H99 deficiency.13,14,140 However, double mutants for the initiator caspases dronc and strica show a partial defect in late oogenesis cell death: up to 21% of egg chambers at the final developmental stage retain nurse cell nuclei in these mutants.14 The double mutants of the effector caspases dcp-1 and drICE show a similar phenotype. This indicates that both initiator caspases, dronc and strica, and effector caspases dcp-1 and drICE, play redundant roles in late oogenesis.14 Finally the adaptor protein Ark appears to play a role in late stage developmental nurse cell death. ark mutants maintain nurse cell nuclei and even un-dumped nurse cells in many egg chambers at late stages, suggesting that ark functions in both dumping and engulfment processes.141
Another type of cell death during oogenesis refers to the degeneration of entire egg chambers, including follicle cells, nurse cells and oocytes, in response to starvation, developmental defects and various stresses.142 This type of insult-induced degeneration can occur in both region 2 of the germarium and in stages 7−8.143 TUNEL-positive staining and apoptotic morphology indicates that this process is apoptosis-like.13,138,143,144 In stages 7−8 in mid-oogenesis, caspases seem to be required for this process since Diap1 or p35 expression alleviates starvation-induced PCD during mid-oogenesis.13,14,119 This is consistent with observations that diap1 mutants cause degeneration of egg chambers and this degeneration can be reverted by dronc mutations.86,145 Single mutants of any of the three initiator caspases, dredd, dronc or strica, do not affect cell death in mid-oogenesis, while dronc and strica double mutants display a large block in cell death during stages 7−8, suggesting dronc and strica may function redundantly in this process.14 Finally, the effector caspase Dcp-1 is necessary for the process, as starved flies that carry a loss-of-function mutation in dcp-1 fail to complete cell death during mid-oogenesis.11,13,14 In these female flies, the process of cell death appears to begin but is not completed; abnormal egg chambers collect in the ovarioles with whole nurse cells but too few follicle cells.13
A recent report indicates that autophagy, and not apoptosis, may be occurring in region 2 of the germarium and in mid-oogenesis in response to low nutrient conditions.77 Dcp-1 appears to play a role in inducing this autophagic cell death in response to starvation and the IAP protein dBruce serves as a negative regulator of autophagy during nutrient-rich times.77
Clearly, there are many questions that remain to be answered regarding the control of cell death during Drosophila oogenesis. Research indicates that caspases play a significant role in early and in mid-oogenesis cell death in response to starvation. At the same time, it is unclear if the cell death process controlled by caspases is characterized by apoptosis, autophagy, or by another cell death process, and the door is certainly left open for other players. Caspases also have an effect, albeit small, on the developmental programmed cell death that occurs at the end of oogenesis. Therefore, there are additional unknown mechanisms that play a role in the control of this cell death process.
Although caspases have been best characterized as players in the apoptotic pathway, it has emerged in the last years that they also play essential roles in several non-apoptotic processes such as innate immunity, proliferation (see section Caspases Function in Compensatory Proliferation), cell differentiation, cell migration, and cell shape (reviewed in ref. 146). Mutations in many apoptotic genes such as diap1, hid, dronc and drICE affect the branching of aristae, indicating the function of caspases in cell shape.14,118,147
As aforementioned, Diap2, the caspase-8 ortholog Dredd and the caspase-8 adaptor dFADD play essential roles in the immune deficiency (IMD) pathway in innate immunity.80,81,83,120,122,148-152 The Drosophila innate immune system functions through two pathways that regulate distinct classes of NF-kB proteins. The Toll pathway responds to infection by Gram-positive bacteria and fungi by activating Dorsal and Dif (Dorsal-related immunity factor), while the IMD pathway responds to Gram-negative bacteria by activating Relish.153 dredd mutant flies fail to process Relish and are deficient in the production of antimicrobial peptides, thus they are highly susceptible to Gram-negative bacterial infection.120
Dendrite pruning requires dronc activity.154,155 During the pupal stage of Drosophila development, large numbers of dendrites and axonal projections are removed without the death of the neuron cell body in order to remodel the larval nervous system into the adult nervous system. This pruning process partially resembles apoptosis in that it involves disruption of the cytoskeleton and the clearance of cell debris by phagocytosis. This process requires UbcD1, the E2 ligase involved in Diap1 proteolysis.155 It was found that local activation of Dronc caused by downregulation of Diap1 is responsible for the pruning of C4da dendrites during metamorphosis. It is not known why this caspase activation remains localized to the dendrites and does not result in apoptosis of the cell, but one speculation is that dendrite-specific trafficking of proteins is involved in this process.155
The migration of border cells during oogenesis is also regulated by Diap1 and Dronc in a non-apoptotic function.156 The border cells are a group of follicle cells that originate at the anterior pole of the follicle and that migrate to the center of the egg chamber to rest at the border between the nurse cells and the oocyte. The expression of a dominant negative construct of the small GTPase Rac prevents this cell migration. Overexpression of Diap1, but not p35, along with dominant negative Rac can suppress this migration defect. Because p35 can inhibit only DrICE and Dcp-1, but not Dronc, whereas Diap1 can inhibit all three of these caspases, this suggests that Dronc activity is required for this process. The expression of dominant negative dronc also rescued the border cell migration defect caused by dominant negative Rac.156
During Drosophila spermatogenesis, sperm are generated within a male germline syncitium. To complete spermatogenesis, spermatid individualization occurs by the removal of cytoplasm and the enveloping of each single sperm with its own plasma membrane. Similar to the process of dendrite pruning, this process involves elimination of cell components and cytoskeleton remodeling without apoptosis of the cell. Multiple components of the canonical cell death pathway, including Dronc, DrICE and Ark, as well as the initiator caspase Dredd and its adaptor dFadd, were found to be required for spermatid individualization.76,118,124,157-159 p35 and Diap1 overexpression specifically in the male germline resulted in the disruption of individualization, suggesting that caspase activity is required. The RNAi-mediated knockdown of dcp-1, an executioner caspase, resulted in a similar phenotype. Flies homozygous for a hypomorphic ark allele show a high degree of male sterility. Male germline specific expression of an RNAi against ark and a dominant negative dronc construct resulted in disruption of the individualization process. Hid, but not Reaper, was found to be required for this process. Also, localized active Dronc in the spermatids is missing in hid mutant flies, suggesting that Hid-mediated Dronc activity is required.
Caspases also function in cell differentiation. Drosophila macrochaete are produced from a single sensory organ precursor (SOP) cell. Through the process of lateral inhibition, only one cell in a field will become a SOP. If this process fails, the result can be the existence of two macrochaete where there should only be one. Flies bearing ark mutations or expressing dominant-negative dronc contain one extra sensory organ precursor (SOP) cell on each side, suggesting that caspase activity is involved in the control of SOP cell formation in the scutellum.160,161 In fact, it was recently shown that dronc caspase activity negatively regulates the formation of scutellum SOP cells.160 Interestingly, although cytochrome c does not appear to play a major role in Drosophila PCD, cytochrome c-d mutant flies have an additional bristle on the scutellum.161
Activation of caspases has been observed in most of these processes; however, it is unclear why caspases do not induce apoptosis under these conditions. Compartmentalization of active caspases might contribute to this control. Furthermore, the Drosophila IKKε-related kinase (DmIKKε), which regulates Diap1 turnover, might contribute to control caspase activity at the threshold required for non-apoptotic functions.162 The recently reported mutual suppression between Ark and Dronc might also account for the control of caspase activity in non-apoptotic processes.114
In addition to being death-executioners, caspases play essential roles in promoting compensatory proliferation, a mechanism that replaces dying cells through stimulation of proliferation, which contributes to maintaining tissue homeostasis (reviewed in ref. 163). The investigation of the underlying mechanisms of compensatory proliferation was initially thwarted by the quick removal of dying cells following apoptosis. However, rendering dying cells alive through expression of the effector caspase inhibitor p35 (thus producing ‘undead’ cells) allowed the analysis of signaling involved in compensatory proliferation.109,159,164-167
Remarkably, the initiator caspase Dronc was found to be required for compensatory proliferation in Drosophila larval wing tissue.109,159,166 Recently, the requirement of Dronc for compensatory proliferation in proliferating eye tissue was validated.167 Expression of dpp and wg is upregulated in this context.164,165 In contrast, phospho-MAD and Vestigial, two downstream targets of Dpp and Wg, seem to be downregulated.166 Thus, the roles of Dpp and Wg in compensatory proliferation need further validation. In addition to Dronc and potentially Dpp and Wg, JNK (Jun N-terminal kinase) and Dp53 (Drosophila p53), which form a regulatory loop with hid, rpr and dronc, appear to be involved in this type of compensatory proliferation.166
Recently, a second type of compensatory proliferation has been described which occurs in GMR-hid eye imaginal discs.88 This type of compensatory proliferation requires the effector caspases DrICE and Dcp-1, and the Hedgehog (Hh) signaling pathway.167 The primary difference between these two types of apoptosis-induced compensatory proliferation lies in the developmental potential of the affected tissue. The Dronc-dependent Wg/Dpp pathway is engaged in apoptotic proliferating tissue such as the third instar wing and anterior eye imaginal discs, whereas the Dcp-1/DrICE-dependent Hh pathway occurs in differentiating tissue such as the posterior compartment of the eye imaginal disc. Consistently, the Hh signal is released by dying photoreceptor neurons to induce compensatory proliferation in undifferentiated cells.167
Thus, the initiator caspase Dronc and effector caspases DrICE and Dcp-1 stimulate distinct mechanisms to trigger apoptosis-induced compensatory proliferation in tissues of different developmental origin. The mechanisms by which caspases trigger the downstream signaling pathways involved in compensatory proliferation are currently the subject of intensive research.
Although most of the core components of the cell death pathways are known and although we have a good understanding of their biochemical interactions, several important questions remain. How is cell death integrated in the complex organization of a developing multi-cellular organism? In particular, how do cell/cell interactions influence the life-or-death decision? Why do activated caspases in some cases trigger apoptosis, but in non-apoptotic processes do not? What are the mechanisms of apoptosis-induced compensatory proliferation? These are the challenges of the years to come. Drosophila will continue to play an important factor in elucidating these questions. No doubt, the answers to these questions will be of importance for human health and the understanding of the pathology of apoptosis-related diseases.
We would like to thank Zhihong Chen and Clare Bolduc for excellent assistance. This work was supported by the NIH (R01GM068016, R01GM074977, R01GM081543) and The Robert A. Welch Foundation (G-1496).