Programmed Cell Death (PCD) plays a fundamental role in animal development and tissue homeostasis, and abnormal regulation of this process is associated with a wide variety of human diseases (Fuchs and Steller, 2011). Although naturally occurring death was already noticed during amphibian development in the 1849, cell death was long considered a passive phenomenon that accounted for the inevitable end point of biological systems (reviewed in (Steller, 1995; Clarke and Clarke, 2012)). The major breakthrough demonstrating that PCD with the morphological hallmarks of apoptosis was an active, gene-directed process came from genetic studies in the nematode Caenorhabditis elegans. In particular, the identification of mutations with specific effects on programmed cell death and their ordering into a genetic pathway demonstrated that cell death is a developmental fate, with specific genes acting to initiate a program of cell suicide (Ellis and Horvitz, 1986). The characterization of these cell death genes led to identification of a conserved core cell death machinery that centers around a family of cysteine proteases, termed caspases, as key executioners of apoptosis (Steller, 1995; Yuan et al., 1993).

I became interested in PCD in the context of studying the development of neuronal connectivity in the Drosophila visual system. In Drosophila, as in vertebrates, the number of neurons is not genetically predetermined and is controlled by environmental factors (Nordlander and Edwards, 1968; Fischbach and Technau, 1984). Although retinal differentiation can proceed in the absence of connections between photoreceptor neurons and their targets in the optic ganglia, disruption of eye-brain connectivity causes massive cell death of both photoreceptors and brain neurons (Campos et al., 1992; Steller et al., 1987). This suggested that, like in the mammalian nervous system, PCD occurs by default and has to be suppressed by both anterograde and retrograde survival signals (Raff, 1992). In order to understand the mechanism by which survival signals suppress PCD, it was necessary to gain a better understanding of PCD mechanisms. Motivated by the success of the Horvitz lab at MIT in isolating cell death-defective mutations in C. elegans, my laboratory sought to identify Dro sophila cell death mutants. But there were two major technical challenges. First, unlike in C. elegans, PCD in Drosophila does not occur in a predictable, lineage-specific pattern. Second, whereas cell corpses can be detected in C. elegans using Nomarski microscopy, dead and dying cells are not visible in live Drosophila tissues by microscopic inspection. Moreover, the only technically feasible approach for screening a significant fraction of the Drosophila genome for cell death genes at the time was to examine PCD in embryos homozygous for chromosomal deletions. Therefore, it was first necessary to develop a technique to visualize apoptotic cells in live Drosophila embryos, which we achieved using the vital dye acridine orange (AO) (Bonini et al., 1993; Abrams et al., 1993). Using a collection of 129 different chromosomal deletions that covered over 50% of the genome, only three overlapping deletions on the third chromosome produced embryos that lacked all AO staining (White et al., 1994). Further delineation of this region became possible with a smaller deletion, Df(3R)H99, and cloning chromosomal DNA spanning the entire H99 deletion interval by “chromosomal walking” made it possible to isolate the genes responsible for the observed cell-death phenotype (White et al., 1994).

Reaper was the first cell death gene identified in the H99 interval based on the ability of a cosmid transgene to restore cell death in deletion embryos (Figure 1) (White et al., 1994). Strikingly, the mRNA distribution of reaper largely anticipated the pattern of apoptosis in the Drosophila embryo. This suggested that reaper acts as a global and central convergence point of many different signaling pathways to initiate apoptosis. This idea was subsequently confirmed by numerous studies demonstrating that reaper is a direct transcriptional target of many different signaling pathways controlling cell fate, response to hormones, cell stress and damage (Figure 2). Examples includes Hox genes, segmentation genes, steroid hormone signaling, p53, micro-RNAs, JNK-signaling and histone-modifying enzymes (Fuchs and Steller, 2011; Lohmann, 2003; Zhai et al., 2009; Brodsky et al., 2000; Zhang et al., 2008; Stark et al., 2003). In contrast to the vast regulatory region that mediates binding to multiple transcription factors and chromatin-modifying enzymes, reaper encodes a tiny, 65 amino acid protein (White et al., 1994). At its N-terminus, Reaper contains an IAP-binding motif (IBM), which mediated direct binding to and neutralization of DIAP1, the major inhibitor of apoptosis in flies (see below for more details) (Goyal et al., 2000; Shi, 2002). This positioned IAP antagonism as the defining regulatory step controlling cell survival.

Another important finding was that ectopic expression of reaper is sufficient to induce apoptosis during development, and that it is strictly dependent on caspase activation since it is blocked by expression of the baculoviral p35 caspase-inhibitor (White et al., 1996; Grether et al., 1995; Hay et al., 1994; Meier et al., 2000). This supports the model of reaper as a central initiator of a caspase-cascade that is transcriptionally activated by many different death-inducing signals. From a technical perspective, the ability of reaper to induce ectopic cell death served as a powerful tool to define the Drosophila cell death pathway by conduct genetic modifier screens (Goyal et al., 2000; Bergmann et al., 1998).

Shortly after the isolation of reaper, two additional pro-apoptotic genes, head involution defective (hid) and grim were identified in the H99 region (Figure 1) (Grether et al., 1995; Chen et al., 1996). Both genes share with reaper an N-terminal IBM, a role in developmental apoptosis, the potent ability to induce apoptosis, and regulation by multiple developmental and stress pathways. A fourth RHG family member, sickle (skl), was identified in 2002 based on sequence similarity to reaper (Srinivasula et al., 2002; Wing et al., 2002). However, the physiological function of sickle is still not entirely clear. One important difference between hid and reaper is that the expression of hid is significantly broader and not restricted to cells that are fated to die. The hid gene encodes a 410aa protein which is a direct molecular target of the Ras-MAPK pathway (Bergmann et al., 1998; Kurada and White, 1998). Activation of Ras by survival signals causes MAPK to phosphorylate and inhibit Hid protein activity, thereby saving cells that express hid mRNA and are “primed to die” from apoptosis. This mechanism is used, for example, to match the proper number of glial cells in the Drosophila CNS midline to neurons (Bergmann et al., 2002). Neurons secrete the TGFα-like ligand Spitz, and neighboring glial cells, which all express Hid, compete for this survival factor. Spitz activates the Epidermal Growth Factor Receptor (EGFR) only on the surface of midline glial cells that are in contact with neurons. This triggers the Ras/MAPK signaling cascade in “winner cells”, phosphorylates pre-existing HID protein, and inactivates its proapoptotic function. On the other hand, “looser cells” are eliminated by Hid-induced apoptosis (Bergmann et al., 2002). Like for reaper, hid expression is also regulated by micro-RNAs (Brennecke et al., 2003). A combination of transcriptional regulation, post-transcriptional protein modification and micro-RNAs targeting RHG genes provides diverse layers of regulation that allow apoptosis to be precisely patterned in time and space.

Another intriguing layer of regulation is the formation of hetero-multimers amongst RHG proteins. Rpr and Grim contain an α-helical Grim Helix-3 (GH3) domain that promotes their multimerization and localization to the outer mitochondrial membrane (Claveria et al., 2002; Wing et al., 2001; Olson et al., 2003; Sandu et al., 2010). Expression of grim together with reaper or hid induces significantly higher levels of cell death than observed for expressing these genes individually, indicating that each RHG proteins can act synergistically and have distinct cell killing activities (Wing et al., 1998). Moreover, Hid localizes to mitochondria via its C-terminal transmembrane domain and can recruit Rpr to the MOM. This interaction can allow for more effective cell killing by Rpr, and artificially targeting Rpr to the MOM bypasses the requirement for Hid for efficient Rpr-mediated killing (Sandu et al., 2010; Haining et al., 1999). However, some cell types, including embryonic neuroblasts, do not express Hid and are killed by the coordinated expression of reaper, grim and possibly sickle (Tan et al., 2011). Although it is clear that RHG proteins cooperate for the induction of apoptosis, we still do not fully understand the underlying mechanism, and why different cell types have differential requirements for these proteins.

Taken together, the activity of RHG proteins is regulated by many different signaling pathways at both the transcriptional and posttranscriptional level, and these proteins serve as “integrators” to connect signals with the core cell death program. As explained in more detail below, RHG proteins induce apoptosis by directly binding to and neutralizing Inhibitor of Apoptosis (IAP) proteins. The RHG proteins introduced a fundamental principle in cell death regulation: that apoptosis can be triggered by precisely controlled expression of IAP antagonists, rather than activation of caspases alone (Figure 3).

DIAP1 is an essential survival factor and its inactivation causes spontaneous, widespread apoptosis (Goyal et al., 2000; Wang et al., 1999). IAPs, including Drosophila DIAP1, DIAP2 and dBruce, and the mammalian X-Linked Inhibitor of Apoptosis (XIAP), cIAP1/2, and Bruce/Apollon are all E3-ubiquitin ligases that target specific cell death proteins for regulated degradation by the ubiquitin proteasome system (Bader and Steller, 2009). In cells that live, DIAP1 and its mammalian counterpart XIAP repress caspases by direct binding and ubiquitylation (Goyal et al., 2000; Meier et al., 2000; Wilson et al., 2002). Significantly, deleting the RING domain of either DIAP1 in flies or XIAP in mice completely abrogates the anti-apoptotic activity of these proteins (Goyal et al., 2000; Schile et al., 2008). This demonstrates that the E3-ligase activity of these IAPs is essential for the anti-apoptotic function, and that this mechanism of caspase regulation is conserved in evolution. Importantly, once Rpr accumulates in doomed cells it binds to DIAP1 and promotes self-conjugation and degradation of this protein, thereby removing the major break on death in Drosophila (Ryoo et al., 2002). This represents a radical switch in directing IAP-mediated Ub-tagging from pro-death in living cells to anti-death proteins in doomed cells that is well-suited for a rapid change of cell fate. Interestingly, a similar phenomenon is observed during ARTS-mediated apoptosis in mammalian cells (see below). X-ray crystallography has provided insights into the structural basis for interactions between RHG peptides and different IAP-domains, but information on full-length multimeric protein complexes of IAPs and their antagonists remains very limited (Chai et al., 2003; Yan et al., 2004).

The identification of mammalian Smac/DIABLO, Omi/HtrA2 and ARTS as mitochondrial IAP antagonists validated the evolutionary conservation of the mechanism uncovered in Drosophila (Fuchs and Steller, 2011; Shi, 2002; Larisch et al., 2000; Gottfried et al., 2004; Kornbluth and White, 2005; Verhagen and Vaux, 2002). The IBM is structurally conserved between Drosophila and mammalian IAP-antagonists and is required for IAP-binding and cell killing (Shi, 2002). In contrast to RHG proteins, Smac/DIABLO and Omi/HtrA2 proteins are localized within the mitochondrial intermembrane space and require MOMP to access and bind cytosolic IAPs. ARTS is another mammalian IAP-antagonist that shares no obvious sequence homology with RHG proteins but many functional similarities with Reaper (Larisch et al., 2000; Gottfried et al., 2004). Of particular interest is that ARTS, like Reaper, does not require MOMP and acts upstream of cytoC and Smac/Diablo (Edison et al., 2017; Abbas and Larisch, 2020). ARTS, like Reaper, induces Ub/proteasome-mediated degradation of XIAP (the mammalian DIAP1 homolog) (Lotan et al., 2005). Moreover, transcription of ARTS, like reaper, is induced upon DNA damage by direct binding of p53 to the ARTS promoter (Hao et al., 2021). Finally, inactivation of ARTS in mice causes elevated levels of XIAP protein and attenuates stem cell apoptosis (Kissel et al., 2005; Garcia-Fernandez et al., 2010; Fu et al., 2013; Koren et al., 2018). This suggests that ARTS may be a functional analog of Reaper.

Collectively, this work cemented reaper as a founding concept in metazoan apoptosis regulation. A simplified model to explain the proper spatio-temporal regulation of a caspase cascade culminating in apoptosis is the “gas and brake” model of cell death control (Figure 3) (Song and Steller, 1999). In analogy to a car, coordinated regulation of caspase-activating (“gas”) and inhibitory (“brake”) factors will ensure that apoptosis only proceeds when multiple checkpoints are passed. It is apparent that the complexity of caspase regulation has increased over the course of evolution from worms to flies to mammals. It is plausible that this increased complexity, together with much extended lifespan, required additional layers of control to contain the risks posed by a constitutively expressed cell suicide program.

Like mammals, Drosophila contains both initiator and effector caspases and this topic has been extensively reviewed elsewhere (Figure 3) (Fuchs and Steller, 2011; Fuchs and Steller, 2015; Kietz and Meinander, 2023; Denton et al., 2013). The initiator caspase Dronc is homologous to mammalian caspase-9 and requires binding to Dark/Hac-1, the Apaf-1 homolog, to form the apoptosome (Meier et al., 2000; Rodriguez et al., 1999; Zhou et al., 1999). Drice and Dcp-1 are caspase-3-like effector caspases that carry out cellular demolition (Song et al., 1997; Fraser and Evan, 1997; Fraser et al., 1997). Although cytochrome c appears not strictly required for apoptosome activation in flies, mutations can delay developmental apoptosis, consistent with the idea that cytoC accelerates rather than initiates a caspase cascade (Mendes et al., 2006). Moreover, mitochondrial dynamics and ROS critically influence cell death outcomes (Clavier et al., 2016). In general, the Drosophila apoptosome functions analogously to its mammalian counterpart. This conservation makes Drosophila a powerful model for caspase biology. While Drosophila research cannot reveal mammalian-specific mechanistic details, it has and will undoubtedly continue to identify new principles for cell death regulation to guide mammalian work.

Apoptosis has long been recognized as an important morphogenetic process that removes unwanted cells during development (Fuchs and Steller, 2011). However, complete ablation of apoptotic cell death in C. elegans, the system central to identifying the core apoptotic machinery, produces viable animals that appear grossly normal and have only relatively minor neurological and behavioral defects (Conradt et al., 2016; Kochersberger, 2023). At the other end of the spectrum, results from studies aimed at blocking apoptosis during mouse development are consistent with an important role of apoptosis but have encountered numerous technical difficulties, including enumerating apoptotic cells during development and possible compensatory mechanisms (Voss and Strasser, 2020). Here is where Drosophila shines again: inhibition of apoptosis in Drosophila unequivocally revealed the essential role of this process during both embryonic and post-embryonic development. The many essential developmental processes that rely on apoptosis include head involution (a critical morphogenetic process shaping the embryo), determining the size and pattern of the embryonic nervous system, forming proper segment bound, shaping leg joints, wing margins and the developing eye during post-embryonic development (Grether et al., 1995; Bergmann et al., 2002; Suzanne and Steller, 2013; Peterson et al., 2002; Lohmann et al., 2002; Rusconi et al., 2000). Moreover, as discussed in more detail below, apoptotic effector caspases are also required for pruning of synaptic connections (Yu and Schuldiner, 2014; Williams et al., 2006; Rumpf et al., 2011). Furthermore, apoptosis is essential for metamorphic remodeling of larval tissues, including salivary glands, midgut, and larval neurons that are all removed by a combination of coordinated apoptosis and autophagy (Suzanne and Steller, 2013; Xu et al., 2020). Apoptosis also contributes mechanically to shaping tissues because local caspase activation drives epithelial folding by altering junctional tension and contractility (Suzanne and Steller, 2013). Apoptosis also determines germ cell survival and shapes Drosophila gonads, ensuring correct organ size, composition and architecture (Baum et al., 2007; Monier and Suzanne, 2015; Suzanne et al., 2010). Finally, as a general principle, cells that fail to adopt correct identities during development are removed through RHG induction (Rusconi et al., 2000; Xu et al., 2009; Zhai et al., 2012). These diverse inputs relay spatial developmental cues, such as cell stress and damage, mis-specification of cells, competition for limiting amount of survival factors, lineage-specific apoptosis, to the induction of RHG genes and apoptosis. Thus, Drosophila is arguably the best system at this time for investigating the mechanisms by which apoptosis drives tissue morphogenesis.

Over the last 2 decades, Drosophila has emerged as a leading model demonstrating that caspases have critical non-lethal roles in development, tissue remodeling, cell differentiation, proliferation, neuronal remodeling, and stress responses. These functions rely on sublethal, spatially restricted, or transient caspase activity, controlled by Diap1 and RHG proteins. This is a very exciting and active field of research that has been extensively reviewed elsewhere. Therefore, only a brief overview is provided here.

The first mechanistic proof that apoptotic caspases can drive cell differentiation without inducing apoptosis was during mouse lens fiber cell differentiation, overturning the dogma that caspases are inherently lethal (Ishizaki et al., 1998). This was followed by a report that caspase-3 activity is required for skeletal muscle differentiation in mice (Fernando et al., 2002). The first demonstration of an essential non-lethal role of apoptotic effector caspases in Drosophila was for spermatid differentiation (Arama et al., 2003). Inhibition of caspase activity prevents cytoplasmic clearance during spermatid individualization and results in male sterility. Interestingly, in this context caspase activation depends on the testis-specific cytochrome c paralog cyt-c-d (Arama et al., 2006). Loss-of-function mutations in cyt-c-d impair Dronc activation and consequently block individualization. Additional components of the apoptosome, including Ark and Dronc, are likewise essential for caspase activation during spermatogenesis (Arama et al., 2006; Huh et al., 2004). Conversely, mutations in the giant IAP-like protein dBruce lead to nuclear degeneration and spermatid death, underscoring its critical role in safeguarding the nucleus from inappropriate caspase activity (Arama et al., 2003).

Another fascinating example for the critical non-lethal role of apoptotic proteins is the remodeling of neurons during Drosophila metamorphosis (Yu and Schuldiner, 2014; Mukherjee and Williams, 2017; Colon-Plaza and Su, 2022; Maor-Nof and Yaron, 2013; Schuldiner and Yaron, 2015). During post-embryonic development, many larval neurons are extensively remodeled to generate adult-specific morphologies. Classic examples are Drosophila sensory neurons and mushroom body γ-neurons, which prune their larval axons and dendrites in response to the steroid hormone ecdysone (Williams et al., 2006; Watts et al., 2003; Kuo et al., 2005). Strikingly, this remodeling requires activation of apoptotic caspases, yet the neurons survive. Localized activation of the initiator caspase Dronc occurs selectively within neurites destined for removal (Yu and Schuldiner, 2014). This spatial restriction results from a combination of DIAP1 downregulation, compartment-specific assembly of Dark apoptosomes, and the sensitivity of cytoskeletal substrates to low-level caspase cleavage (Williams et al., 2006; Rumpf et al., 2011; Lee et al., 2009). Effector caspases such as Drice and Dcp-1 are also activated locally, where they dismantle cytoskeletal structures, degrade adhesion molecules, and promote fragmentation of targeted processes. At the morphological level, phagocytosis of cellular material during neuronal pruning is morphologically and mechanistically similar to the engulfment of apoptotic corpses (Ji and Han, 2024). Importantly, apoptotic thresholds are never reached in the soma during pruning due to robust maintenance of DIAP1 and survival pathways that prevent global cell death. The pruning of neuron thus leverages a “micro-apoptotic” program that is powerful enough to remodel defined domains but tightly contained to ensure viability. Taken together, work in Drosophila revealed that neuronal pruning is achieved through exquisitely tuned, sublethal caspase activation. These findings highlight a broader principle: apoptotic proteins, long viewed as executioners, can also serve as precise sculptors of neural architecture. This paradigm has since influenced studies in vertebrates, where caspase-dependent pruning is now recognized as a conserved mechanism for the refinement of neural circuit (Simon et al., 2012; Faust et al., 2021; Nguyen et al., 2021).

Another example is epithelial morphogenesis in Drosophila, which requires coordinated changes in cell shape, adhesion, and cytoskeletal architecture (Tepass, 2012). During this process, caspases fulfill essential lethal and non-lethal roles (Suzanne and Steller, 2013; Nakajima and Kuranaga, 2017). Non-apoptotic caspase functions depend again on precisely controlled, low-level caspase activity that remodels specific cellular structures while preserving epithelial integrity (Nakajima and Kuranaga, 2017). Non-apoptotic effector caspase activity is also required to maintain tissue integrity by suppressing cell migration and invasion, and caspases participate in epithelial cell delamination and controlled cell extrusion (Gorelick-Ashkenazi et al., 2018; Villars et al., 2022). Collectively, studies in Drosophila have revealed that epithelial morphogenesis relies on caspases not simply as executioners but as precision remodeling enzymes. Sublethal, spatially confined caspase activity provides a versatile mechanism for sculpting epithelial tissues, coordinating mechanical forces with patterning cues, and preserving epithelial organization during dynamic developmental transitions. Interestingly, many caspase substrates are shared between apoptosis and non-lethal use of effector caspases, including NF-κB subunits (p65/RelA, p50), PARP, lamins, caspases, RIPK1, TRAF1, PKC, and STAT1 (Julien and Wells, 2017; Abdelghany et al., 2024). Collectively, this work supports the notion that the non-lethal use of effector caspases amounts to “apoptosis without death”.

Another major contribution of Drosophila to cell death biology is the discovery that apoptotic cells can release mitogenic signals to stimulate proliferation of neighboring cells, a phenomenon now generally referred to as apoptosis-induced proliferation (AiP) (Ryoo et al., 2004; Perez-Garijo et al., 2004; Bergmann and Steller, 2010; Mollereau et al., 2013; Bergantinos et al., 2010; Ryoo and Bergmann, 2012; Mollereau and Ma, 2014; Fan and Bergmann, 2008; Fogarty and Bergmann, 2017). A series of experiments in Drosophila first demonstrated that cells undergoing apoptosis in response to DNA damage can stimulate the proliferation of neighboring cells and thereby aid in tissue regeneration (Ryoo et al., 2004; Perez-Garijo et al., 2004). This response is mediated by p53 that causes direct transcriptional activation of reaper and hid, leading to inhibition of Diap1 and de-repression of caspases. JNK signaling stimulates a positive feedback loop between p53 and RHG genes that amplifies the initial stimulus and is critical for stress-induced apoptosis (Shlevkov and Morata, 2012). Depending on the cellular context, both initiator and executioner caspase influence the release of mitogens from stressed or injured cells, thereby promoting regeneration. The mitogens secreted by apoptotic cells during AiP include Wingless (Wg, Wnt orthologue), Decapentaplegic (Dpp, TGF-β orthologue) and Hehdgehog (Fogarty and Bergmann, 2017). These findings are significant because AiP is an evolutionary conserved process with obvious implications for tissue regeneration and cancer (Fuchs and Steller, 2011; Ryoo and Bergmann, 2012). For example, chemotherapy using DNA-damaging agents may cause AiP, which in turn may contribute to the formation of metastases (Moreno-Celis et al., 2022; Hagan et al., 2025; Morata and Calleja, 2020; Morana et al., 2022). Therefore, studies on AiP are likely to have considerable implications for the clinic.

In 1994, PCD was basically synonymous with apoptosis. Perhaps ironically the one classic exception was always the death of intersegmental muscles during insect metamorphosis, based on which the term “programmed cell death” was coined by Lokshin and Williams (Lokshin and Williams, 1965; Schwartz, 1992). Since then many different forms of non-apoptotic PCD have been described (Fuchs and Steller, 2015; Arama et al., 2025; Tait et al., 2014). These include autophagy-dependent developmental cell death, regulated necrosis, pyroptosis-like inflammatory death and immune-induced death (Dziedziech and Theopold, 2022; Denton and Kumar, 2019; Napoletano et al., 2017; Li et al., 2019; Liu et al., 2014; McCall, 2010). Besides a well-established connection between apoptosis and autophagy, a direct and physiological role of reaper in any of these processes remains to demonstrated. However, reaper may contribute to other forms of cell death, especially under pathological conditions. For example, when apoptosis is blocked by caspase inhibitors, Eiger (the fly TNF homolog) can induce necroptosis in a process that requires the initiator caspase Dronc (Li et al., 2019). Since Dronc is inhibited by Diap1, and this repression is relieved by RHG proteins, it is possible that reaper and friends can affect other forms of cell death when core apoptotic pathways are genetically or pharmacologically inhibited.

The discovery of reaper in 1994 established the foundation for modern apoptosis research in Drosophila. RHG–DIAP1–caspase interactions revealed that apoptosis is a transcriptionally regulated, developmentally patterned process. Over the past 32 years, Drosophila has provided fundamental insights into apoptosis, the non-lethal use of apoptotic proteins, alternative death pathways, and proliferation–death coupling. Many processes that are controlled by RHG proteins in Drosophila mirror phenomena in vertebrates. RHG genes will remain essential tools to drive conceptual advances with relevance to cancer, neurodegeneration, immunity, and regeneration.