Non-autonomous cell death induced by the Draper phagocytosis receptor requires signaling through the JNK and SRC pathways

Cell clearance is critical for tissue homeostasis and the resolution of inflammation. Failure to remove dead cells from tissue has been associated with autoimmunity, asthma, inflammatory colitis and cancer (Abdolmaleki et al., 2018; Fond and Ravichandran, 2016; Kumar et al., 2017). Phagocytes clear dead cells using orchestrated steps. First, the dying cell releases ‘find-me’ signals, so that phagocytes can migrate to their vicinity. The dying cell also exposes eat-me signals such as phosphatidylserine on its surface. The ‘eat-me’ signals are recognized by phagocyte receptors and this interaction drives phagocyte cytoskeletal rearrangement. The engulfed corpse finally undergoes phagosome maturation, where it is targeted for degradation.

The two main types of phagocytes are professional and non-professional phagocytes. Professional phagocytes have a main function of engulfing dead cells, pathogens and debris, whereas non-professional phagocytes have other tissue-resident functions but can engulf when needed. Understanding the mechanisms by which non-professional phagocytes engulf is important because they are found in almost all tissues of the human body (Arandjelovic and Ravichandran, 2015). Engulfment by non-professional phagocytes might also be critical in tissues that have restricted access to circulating professional phagocytes.

Although much work has been done on understanding the molecular biology that mediates engulfment, a newly emerging theme in the field of cell clearance is the ability of a phagocyte to kill cells. The molecular underpinnings of phagocyte-driven cell death are poorly understood. Phagoptosis is the use of phagocytic machinery to induce cell death (Brown et al., 2015). The defining characteristic of phagoptosis is that the loss of engulfment machinery prevents cell death. The mechanisms that allow for phagocyte-mediated cell death include the loss of ‘don't-eat-me’ signals or the induction of eat-me signals on the target cell. In mammals, macrophages and microglia have been reported to kill erythrocytes, neutrophils, neurons and endothelial cells (Brown and Neher, 2012; Brown et al., 2015; Lobov et al., 2005). Although several in vitro studies have tackled how phagocytes kill, very little in vivo work has been done to understand phagocyte-driven cell death (Chakraborty et al., 2015; Hakim-Mishnaevski et al., 2019; Herzog et al., 2019; Zohar-Fux et al., 2022). The identification of phagocyte-driven cell death mechanisms has clinical implications, such as in cancer therapy for the rapid removal of unwanted malignant cells (Métayer et al., 2017).

The Drosophila ovary is a genetically tractable, in vivo system to study phagocytosis and phagocyte-driven cell death (Serizier and McCall, 2017). Each ovary contains progressively developing egg chambers. Each egg chamber comprises 15 germline-derived nurse cells and an oocyte, which are encompassed by follicular epithelial cells. Egg chamber degeneration, characterized by apoptosis of the germline, can be induced specifically in mid-oogenesis in response to starvation (Drummond-Barbosa and Spradling, 2001; Giorgi and Deri, 1976). Several other influential factors such as developmental defects, chemical treatment and cocaine exposure can also induce apoptosis in mid-oogenesis (Chao and Nagoshi, 1999; de Lorenzo et al., 1999; Panagopoulos et al., 2007; Willard et al., 2006). This stage is thought to be susceptible because it occurs during the pre-vitellogenic stages of oogenesis, before flies require substantial resources to produce yolk (Buszczak and Cooley, 2000; Drummond-Barbosa and Spradling, 2001; Giorgi and Deri, 1976).

The follicular epithelial cells function as non-professional phagocytes in the Drosophila ovary and engulf and degrade the apoptotic germline of degenerating egg chambers in mid-oogenesis. Draper is a major engulfment receptor required in the Drosophila follicle cells, epidermis, glia, testis cyst cells and hemocytes (Etchegaray et al., 2012; Han et al., 2014; MacDonald et al., 2006; Manaka et al., 2004; Zohar-Fux et al., 2022). Draper, the Drosophila ortholog of CED-1 in C. elegans and MEGF10 (as well as JEDI; also known as PEAR1) in mammals, is a single pass transmembrane receptor in the Nimrod superfamily that is required for cell clearance (Somogyi et al., 2008). The Draper I isoform has extracellular EMI and NIM domains at the N-terminus followed by 15 epidermal growth factor (EGF) repeats and NPxY and ITAM (YxxI/L-X6-12-YxxI/L) motifs at the cytosolic C-terminus. Previous studies have determined that the ITAM motif of Draper is required for glial phagocytosis (Logan et al., 2012). Phosphorylation at the ITAM motif recruits SH2-containing signaling molecules to activate downstream signaling. The ITAM motif requires phosphorylation via Src42A in glia (Ziegenfuss et al., 2008). Shark kinase interacts with the phosphorylated ITAM motif and is required for Draper-mediated engulfment. The PTB domain of Ced-6, the GULP ortholog, is proposed to interact with the phosphorylated NPxY motif and is required for glial phagocytosis of pruned axons (Awasaki et al., 2006).

Although much work has been done to understand the death and engulfment mechanisms of germline apoptosis and clearance by follicular epithelial cells (Serizier and McCall, 2017; Yalonetskaya et al., 2018), little is known about how the follicular epithelial cells promote death of the nurse cells. Several studies have demonstrated that the cell communication between nurse cells and epithelial follicle cells is required for not only efficient clearance but also germline death progression. For example, during developmental nurse cell death in late oogenesis, draper is required non-autonomously for DNA fragmentation (Timmons et al., 2016). In starvation-induced cell death in mid-oogenesis, draper knockdown in the follicle cells delays oocyte death (Kleinsorge, 2015) and results in a delay in nurse cell chromatin breakdown (Etchegaray et al., 2012). Interestingly, draper overexpression in follicle cells, independently of starvation cues, induces nurse cell death in mid-oogenesis (Etchegaray et al., 2012). Draper overexpression in glia results in dopaminergic and GABAergic neuronal loss (Hakim-Mishnaevski et al., 2019). These findings suggest that Draper can promote cell death non-autonomously in multiple tissues, but how this occurs was not known.

The aims of this study were to characterize and determine how Draper non-autonomously induces nurse cell death in mid-oogenesis. We found that Draper requires its intracellular signaling domain to kill the adjacent nurse cell population. Further investigation revealed that Src, Shark and c-Jun N-terminal kinase signaling and the caspase Dcp-1 (FBgn0010501) were required for Draper-induced nurse cell death. These findings indicate that signaling by Draper activates apoptosis of neighboring nurse cells, providing insight into mechanisms of phagoptosis driven by non-professional phagocytes.

In the Drosophila ovary, the follicle cells function as non-professional phagocytes to clear dying germline material. In response to starvation, nurse cells die and the follicle cells enlarge and engulf the dying germline (Etchegaray et al., 2012). Degenerating egg chambers undergo several nuclear and cytoskeletal changes, which have been used as morphological markers to visualize the progression (‘phases’) of cell death (Etchegaray et al., 2012) (Fig. 1A–F). In healthy egg chambers, the nurse cell DNA is dispersed and intact. In phase 1, nurse cell DNA begins to separate, losing uniformity. In phase 2, the nurse cell chromatin becomes highly condensed, but the fragments of individual nurse cells are distinguishable. By phase 3, the nurse cell chromatin is still highly condensed, but individual nurse cells are no longer distinguishable. In phase 4, the nurse cell nuclei undergo a second round of fragmentation, and by phase 5, the nurse cell chromatin is mostly gone, and some follicle cells begin to die. Concomitant with nurse cell nuclear breakdown, the follicle cells enlarge, as visualized with a plasma membrane marker (Fig. 1A–F).

We had previously found that overexpression of the engulfment receptor Draper in follicle cells could induce cell death in the absence of starvation (Etchegaray et al., 2012), providing a valuable in vivo model to dissect the process of phagoptosis. To determine whether Draper-induced chromatin architectural changes were similar to starvation-induced cell death, we analyzed the nurse cell nuclear changes using DAPI. Egg chambers overexpressing Draper in follicle cells were analyzed under a well-fed diet in parallel to control, starved lacZ-expressing egg chambers. The nurse cell nuclear morphology in egg chambers dying in response to Draper overexpression in follicle cells was similar to starvation-induced nurse cell death; in both cases nurse cells exhibited nuclear separation at phase 1 and highly condensed nurse cell chromatin at phases 2–5 (Fig. 1G–L).

To determine whether the phenotype of Draper-induced follicle cell engulfment was similar to that of degenerating egg chambers from wild-type starved flies, we used an antibody against Discs large 1 (Dlg1; hereafter Dlg) to visualize the follicle cell membranes and enlargement. Similar to what was seen with starvation-induced death, Draper-expressing follicle cells induced synchronous follicle cell enlargement once the nurse cell chromatin was highly condensed (Dlg, Fig. 1I–L). Overall, Draper overexpression in follicle cells was found to induce nurse cell death and follicle cell enlargement mimicking starvation-induced death. Moreover, the egg chambers that underwent cell death were between stages 7 and 10 of oogenesis in both the Draper- and starvation-induced pathways.

We have previously shown that follicle cell enlargement coincides with engulfment of germline material using a germline-GFP marker and membrane markers to quantify the amount of engulfed germline (Etchegaray et al., 2012). To more closely examine whether there were subtle differences in the follicle cell engulfment between starvation-induced and Draper-induced nurse cell death, the unengulfed germline was quantified in each phase. When we closely measured unengulfed germline material, nurse cell death was found to precede follicle cell enlargement and engulfment, as shown by statistically significant differences in phases 1 and 2 of engulfment in Draper overexpressing follicle cells (Fig. 1H,I,M). Strikingly, this suggests that Draper signals to induce nurse cell death before engulfment.

To determine the potency by which Draper induced nurse cell death, we scored the number of degenerating egg chambers compared to expression of lacZ control in the follicle cells in flies that were not starved. Flies expressing lacZ in the follicle cells exhibited less than 10 degenerating egg chambers per 100 ovarioles based on nurse cell chromatin morphology, whereas overexpression of Draper led to nearly 70 degenerating egg chambers per 100 ovarioles (Fig. 1N). Thus, Draper killed seven times the number of egg chambers than the lacZ control, suggesting that Draper is sufficient to promote nurse cell death.

To determine whether Draper required intracellular signaling to drive nurse cell death, we expressed full-length Draper I constructs that had mutant NPxY and YxxL domains (Logan et al., 2012) to block signaling via the cytoplasmic tail (Fig. 2A). We examined whether the Draper mutant constructs have similar Draper expression levels by immunostaining (Fig. S1A–D). Quantification of immunostaining indicated that the wild-type Draper overexpression construct produced Draper protein 26 (±5.67, s.e.m.)-fold higher than the lacZ non-starved control (Fig. S1E). The Draper point mutant constructs produced similar levels of Draper protein compared to the overexpression of wild-type Draper.

To determine whether the NPxY and YxxL motifs were required to drive Draper-mediated nurse cell death, the number of degenerating egg chambers per 100 ovarioles was quantified in the mutant Draper-expressing flies. The NPxY signaling motif was found to be dispensable for Draper-induced nurse cell death, since the N855A mutant induced cell death as well as wild-type Draper, suggesting that the Ced-6 adapter molecule (Awasaki et al., 2006) does not contribute to Draper-induced cell death (Fig. 2B). However, cell death was not induced in the Y949F mutant, indicating a requirement for the YxxL motif and Shark kinase interaction in this process.

To identify the molecular signals by which Draper signals downstream of the cytoplasmic tail to kill nurse cells, we next asked which signaling molecules were required for Draper to kill nurse cells. To address this, we knocked down candidate effectors in a draper overexpression background (Fig. 3). To corroborate the intracellular signaling domain findings in Fig. 2, we assessed whether shark knockdown in a draper overexpression background phenocopied the Draper YxxL point mutant phenotype, since Shark has been shown to bind to this intracellular motif to activate downstream signaling. Indeed, shark knockdown in the draper overexpression background blocked draper-induced nurse cell death. Ced-6 knockdown, however, had no effect on Draper-induced nurse cell death, suggesting that it is not required for Draper-mediated downstream effector activation to drive nurse cell death. We next asked whether Draper requires phosphorylation via the canonical Src family kinases. When Src42A was knocked down in a draper overexpression background, death was blocked to the same extent as shark knockdown analysis. Together, these results suggest that Draper relies on Src42A-mediated phosphorylation of the YxxL motif and Shark binding to non-autonomously kill the nurse cells.

Previously, draper overexpression in follicle cells was found to activate JNK signaling prior to nurse cell death (Etchegaray et al., 2012). Indeed, kayak, which is homologous to mammalian fos and a downstream effector of JNK signaling, was found to be required for full levels of draper-induced nurse cell death (Fig. 3). Expression of a dominant negative version of basket (Drosophila JNK) showed a stronger inhibition of Draper-induced death than kayak knockdown, indicating that JNK signaling is required for Draper-induced death. These results together suggest that Draper kills nurse cells via Src-Shark and JNK signaling pathways.

We next asked what type of cell death is induced by draper overexpression. Because the nurse cells undergo several morphological changes reminiscent of apoptosis, we conducted several assays to measure apoptosis. Effector caspase activation is one hallmark of apoptotic cell death. We first sought to determine whether overexpression of Draper induced activation of Dcp-1 in nurse cells using an antibody that detects cleaved Dcp-1. Cleaved Dcp-1 was expressed in the nurse cells throughout all phases of engulfment (Etchegaray et al., 2012) in draper-expressing follicle cells and the starved control (Fig. 4), similar to what has previously been shown in starvation-induced nurse cell death (Meehan et al., 2015; Sarkissian et al., 2014).

In response to starvation, nurse cell death is dependent on the caspase Dcp-1 and can be prevented by the expression of caspase inhibitors (Baum et al., 2007; Laundrie et al., 2003; Mazzalupo and Cooley, 2006; Peterson et al., 2003). Most apoptotic cell deaths in Drosophila require the initiator caspase Dronc and the effector caspases Drice and/or Dcp-1 (Leulier et al., 2006; Xu et al., 2006); thus, the nurse cell death pathway is unusual for being highly dependent on Dcp-1, with no role for Drice and minor roles for the initiator caspases Dronc and Strica (Baum et al., 2007). In response to starvation, Dcp-1 mutant mid-stage egg chambers do not die but instead become ‘undead’ – defined by characteristic uncondensed nurse cell chromatin and follicle cells that prematurely die (Laundrie et al., 2003). To determine whether Draper-induced cell death was caspase dependent, we expressed the caspase inhibitors p35 (also known as BacA\p35) and Diap1 in the germline, while draper was overexpressed in the follicle cells. Gal4 was expressed in both the germline and the soma simultaneously, but a UASt construct, which is poorly expressed in the germline, was used to express draper, and UASp, which has robust germline expression (Rørth, 1998), was used to express the caspase inhibitors. Similar to what was seen with starvation, the egg chambers became undead, leading to the accumulation of abnormal egg chambers and uncondensed nurse cell debris (Fig. 5A,B). To determine which caspases were required for Draper-induced nurse cell death, draper was overexpressed in the follicle cells and RNAi against individual caspases was expressed using VALIUM20 transgenes (Perkins et al., 2015), which express well in the germline. Most of the caspase knockdown flies showed degenerating egg chambers, similar to what was seen with draper alone (Fig. 5C,E), but RNAi against Dcp-1 gave rise to undead egg chambers (Fig. 5D–F). The number of degenerating and undead egg chambers per 100 ovarioles were quantified, and only Dcp-1 showed a statistically significant difference (Fig. 5E,F). These results indicate that Dcp-1 is required for Draper-induced nurse cell death, and other caspases are not, suggesting that Draper ‘hijacks’ the starvation pathway.

To determine whether draper elicits chromatin changes in nurse cells in a similar manner to starvation in egg chambers, we measured the amount of chromatin breakdown in nurse cells. The diameter of the average largest chromatin fragment (arrows in Fig. 6B,C,E,F) was obtained for phases 4 and 5 of engulfment. There was a higher number of egg chambers with larger condensed chromatin fragments in draper overexpression flies compared to in starved controls, suggesting a delay in chromatin breakdown (Fig. 6H,I).

We next tested whether DNA fragmentation occurs in the Draper overexpression mutants by conducting a TUNEL assay to label DNA fragments. Nurse cell DNA was not fragmented in healthy egg chambers of control and Draper-overexpressing egg chambers (Fig. 7A′,C′) and nurse cell DNA fragmentation occurred similarly between starved controls and Draper-overexpressing egg chambers (Fig. 7E). DNA fragmentation was detected in phase 1, when egg chambers began to degenerate and continued to the end of engulfment in phase 5 (Fig. 7B,D,F). All morphological assays conducted suggest that Draper induces Dcp-1-dependent cell death in nurse cells, with a mild delay in chromatin breakdown and follicle cell enlargement compared to starvation-induced apoptosis.

To investigate candidate pathways that might activate Dcp-1, we examined markers of oxidative stress, including the reporter GSTD–GFP (Sykiotis and Bohmann, 2008). Interestingly, GSTD–GFP was activated in the anterior follicle cells of egg chambers dying in response to Draper expression, but it was not activated in egg chambers dying in response to starvation (Fig. 7G–J″). These findings suggest that Draper might activate an oxidative stress pathway in follicle cells, which, in turn, triggers Dcp-1-dependent cell death in nurse cells.

Cell clearance is critical for tissue homeostasis and the resolution of inflammation. Although the mechanisms of internalization and phagosome maturation are well characterized, the molecular underpinnings of phagocyte-dependent cell death are understudied. When draper is overexpressed in the follicle cells, the nurse cells die, which is apparent by the nuclear changes that occur. When signaling at the cytosolic tail of Draper was blocked, the YxxL motif was found to be required, but the NPxY motif was dispensable. These results suggest that the YxxL motif, a site of phosphorylation by Src42A and binding of Shark, is critical for draper-induced germline death progression. Investigation of candidate signaling molecules using RNAi corroborated these results; shark and Src42A attenuation blocked draper-induced germline death, but ced-6 attenuation had no effect. Overall, these results support that the Draper/Src/Shark axis, but not the Draper/Ced-6 axis functions in promoting germline death in the ovary. The Draper/Shark interplay aligns with what has previously been shown in engulfment by Drosophila glia, where Src42A is required for Shark binding and Draper-mediated cell clearance (Ziegenfuss et al., 2008). However, there are other contexts that require a Draper/Ced-6 axis, such as larval axonal clearance during metamorphosis (Awasaki et al., 2006). Perhaps the Draper/Ced-6 axis is critical for generating a more fine-tuned response that helps glia determine which axons to prune, whereas the Draper/Shark axis induces diverse downstream effectors for bulk engulfment and signaling to dying cells. Draper has been shown to form microclusters in response to phosphorylation, and additional phosphorylated residues drive engulfment in hemocyte-derived S2 cells (Williamson and Vale, 2018). Given these results, Shark might be required to prime the cytosolic tail of Draper for additional phosphorylation events to take place for oligomerization to exclude phosphatase binding and promote downstream effector activity.

The c-Jun N-terminal kinase (JNK) pathway is activated in response to stress. Knockdown of the fos ortholog kayak in starvation-induced nurse cell death strongly blocks engulfment (Etchegaray et al., 2012). Knockdown of kayak also results in a strong persisting nurse cell nuclei phenotype in late oogenesis, where nurse cells fail to die during developmental nurse cell death (Timmons et al., 2016). In the draper overexpression context, however, we found that kayak knockdown in the draper-overexpressing follicle cells only partially blocked nurse cell death. Expression of a dominant-negative version of basket (JNK) had a stronger effect, suggesting that JNK signaling might have other targets beyond Kayak that promote nurse cell death.

Draper induces several hallmarks of nurse cell apoptosis, namely nurse cell condensation and fragmentation and a requirement for the Dcp-1 caspase. Dcp-1 is the key caspase required for starvation-induced nurse cell death (Baum et al., 2007; Laundrie et al., 2003), indicating that Draper activates a pathway that is related to the starvation-induced pathway. Similar to what is seen in the starvation pathway, the initiator caspase Dronc was not required, and it remains to be determined how Dcp-1 is activated in a Dronc-independent manner.

There were subtle morphological differences between Draper-induced cell death and starvation-induced cell death. Engulfment was slower to start in the draper-overexpressing egg chambers (Fig. 1), and chromatin breakdown was less efficient (Fig. 6). This delay in Draper-overexpressing egg chambers suggests a requirement for downstream effector signaling activation to initiate follicle cell enlargement and engulfment events. Alternatively, starved flies might have signaling machinery that is poised in previtellogenic egg chambers allowing for an immediate engulfment program, whereas Draper-overexpressing egg chambers need more time to recruit effectors for cell clearance. Interestingly, we found a striking increase in GSTD–GFP in the anterior follicle cells of egg chambers dying in response to draper expression, which was not observed in egg chambers dying in response to starvation. This suggests that overexpression of Draper induces oxidative stress. Taken together, our findings indicate that active signaling by Draper is required to kill nurse cells via apoptosis, providing new insights into mechanisms of phagoptosis driven by non-professional phagocytes.

Most of the UAS-based RNAi stocks were generated by TRiP (Perkins et al., 2015) and obtained from the Bloomington Drosophila Stock Center: UAS-luciferase RNAi (TRIP.JF01355), UAS-egfp RNAi (Bloomington #41560),UAS-kayak RNAi (TRiP.JF02804), UAS-shark RNAi (TRiP.JF01794), UAS-Dronc RNAi (TRiP.HMS00758), UAS-Drice RNAi (TRiP.HMS00398), UAS-Damm RNAi (TRiP.HMJ30189), UAS-Dcp-1 RNAi (TRiP.HMS01779) and UAS-Strica RNAi (TRiP.HMJ21268). Additional lines obtained from the Bloomington Stock Center were UAS-lacZ, UASp-dronc-CARD and UAS-bsk.DN (dominant-negatives). UAS-Src42A RNAi (KK108017) was provided by the Vienna Drosophila Resource Center. GR1-GAL4 was a gift from Trudi Schüpbach (Department of Molecular Biology, Princeton University, USA) and was used to drive expression of UAS lines in follicle cells (Goentoro et al., 2006). nos-GAL4, a gift of Pernille Rørth (1998), was recombined with GR1-GAL4 for simultaneous expression in the germline and follicle cells. UASp-p35 (Baum et al., 2007) was received from Hermann Steller (Strang Laboratory of Apoptosis and Cancer Biology, The Rockefeller University, USA) and UASp-Diap1 was generated in our laboratory (Peterson et al., 2003). UAS-Ced6 RNAi (Awasaki et al., 2006), UAS-draper I and UAS-draper I NPxY and YxxL point mutants (Logan et al., 2012) were gifts from Marc Freeman and Mary Logan (Vollum Institute and Jungers Center, Oregon Health & Science University, USA). The point mutant strains were confirmed to express the correct transgenes using PCR and DNA sequencing. GSTD–GFP (Sykiotis and Bohmann, 2008) was kindly provided by Richard Binari and Norbert Perrimon (Department of Genetics, Harvard Medical School, USA).

Flies were maintained on cornmeal molasses food. Flies for dissection were 3–10 days old and were supplemented with yeast paste for 2–3 days for well-fed conditions. For protein starvation, flies were shifted to apple juice agar lacking yeast for 16–20 h prior to dissection. Flies were grown at 25°C and shifted to 29°C to increase GAL4 activity for several days before dissection.

Fly ovaries were dissected in phosphate-buffered saline (PBS) or Grace's insect medium (Thermo Fisher Scientific) for no more than 20 min. Ovaries were incubated in a fixative comprised of 300 µl of PBS or Grace's insect medium, 200 µl of heptane and 100 µl of fresh 16% EM grade paraformaldehyde for 20 min. Samples were rinsed quickly two times and washed three times for 20 min in PBS with 0.1% Triton X-100 (PBT). Samples for DAPI staining alone were rinsed with 1× PBS and incubated in Vectashield with DAPI (Vector Labs) overnight and mounted on slides. Samples for antibody staining were incubated in PBT with 0.5% BSA and 5% normal goat serum (PBANG) for 1 h. Samples were incubated in primary antibody diluted in PBANG at 4°C at least overnight. Samples were washed two times with PBT and incubated in PBT with 0.5% BSA two times for 1 h each. Samples were washed two times with PBT and incubated in secondary antibody diluted in PBANG for 1 h at room temperature. Samples were washed two times with PBT and incubated in PBT with 0.5% BSA two times for 1 h each. Samples were rinsed with 1× PBS and incubated in Vectashield with DAPI (Vector Labs) overnight and mounted on slides. Primary antibodies were against: Dlg (1:100, Developmental Studies Hybridoma Bank, 4F3; Parnas et al., 2001), Drpr (1:50, Developmental Studies Hybridoma Bank, 8A1; Musashe et al., 2016), and cleaved Dcp-1 (1:100, Cell Signaling, 9578; Meehan et al., 2015; Sarkissian et al., 2014). Secondary antibodies (Jackson Labs) were used at 1:200, and were goat anti-mouse-IgG conjugated to Cy3 and goat anti-rabbit conjugated to Alexa Fluor 488. Egg chambers were imaged on a Fluoview 10i confocal microscope or Nikon C2 si. Images were processed in Fiji software and Adobe Photoshop. For quantification of Draper immunostaining, Image J was used to determine the integrated density of probe for a given area that spanned germline and oocyte. All measurements were normalized to a background signal in the same position that did not contain Draper antibody staining. Corrected total cell fluorescence was defined as total integrated signal density minus the background signal.

Any mid-stage egg chamber that contained highly condensed nurse cell chromatin as visualized with DAPI was scored as a degenerating egg chamber. The total number of germaria was also quantified, to have a reference point for total number of ovarioles examined. The number of degenerating egg chambers per 100 germaria was determined.

Samples for measurement were stained with DAPI and antibodies against Dlg, and imaged on a Fluoview 10i confocal microscope. To score chromatin breakdown, Fiji software was used to measure the diameter of the largest chromatin fragment detectable with DAPI staining. The average largest nurse cell chromatin diameter was obtained for phase 4 and 5 egg chambers. 20 egg chambers were analyzed for each phase and genotype. To score engulfment, Fiji software was used to measure the germline area and total egg chamber area of the largest central area of each egg chamber determined by staining of the membrane marker Dlg. As follicle cells enlarge, they engulf dying material, leading to a decrease in the germline area. The area of unengulfed germline was obtained by dividing the germline area by the egg chamber area.

Each experiment was performed at least three times unless indicated in the figure legend. Statistical analysis was performed using Excel and Graphpad Prism using either unpaired two-tailed t-tests or one-way ANOVA with Dunnett's post-hoc test.

In situ cell death detection kit in TMR red (Roche) was used for TUNEL labeling. Ovaries were dissected and teased apart in 2% paraformaldehyde with 0.1% PBT. Ovaries were fixed in fresh 2% paraformaldehyde with 0.1% PBT for 45 min with rotation. Tissue was rinsed twice with 1% PBT and washed three times in 1% PBT for 30 min. Tissue was further permeabilized in fresh 0.1% sodium citrate in PBT, and incubated at 65°C for 30 min, inverting every 15 min. Tissue was washed three times with 1% PBT over 20 min. 30 µl of TUNEL reaction mixture (27 µl label solution, 3 µl enzyme solution) was added to each sample and incubated at 37°C for 3 h in the dark, shaking every hour. Samples were washed four times with PBS and two drops of Vectashield with DAPI were added. Samples were imaged on the Nikon C2 si and processed in Fiji software.

We would like to thank members of our lab for all of their ongoing support. We would also like to thank Marc Freeman, Mary Logan, Trudi Schüpbach, Pernille Rorth, Richard Binari, Norbert Perrimon, the Developmental Studies Hybridoma Bank, Harvard TRiP, and the Bloomington Stock Center for reagents.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.250134.reviewer-comments.pdf.