The initiator caspase Dronc plays a non-apoptotic role in promoting DNA damage signalling in D. melanogaster Free

The presence of damaged DNA evokes a cellular response aimed at sensing the damage in order to initiate signalling leading to cell cycle arrest and, eventually, culminating in DNA repair or, in the case of persistent damage, cell death or senescence. This is collectively referred to as the DNA damage response (DDR) (Sancar et al., 2004; Jackson and Bartek, 2009). Phosphorylation of histone H2Ax (or H2Av in Drosophila) on serine 139 or 137, denoted γH2Ax or γH2Av, serves as one of the early events of DDR (Rogakou et al., 1998; Madigan et al., 2002) and occurs through the action of by PI3K group kinases, namely, ataxia telangiectasia mutated (ATM), ataxia telangiectasia and Rad3-related protein (ATR) and DNA-dependent protein kinase (DNA-PK). Of these kinases, ATM is known to be associated with the response to double-strand breaks (DSBs), whereas ATR plays a more important role in the response to single-strand breaks (SSBs) formed during replication stress and DNA-PK in the response to γH2Ax-associated apoptosis (Podhorecka et al., 2010). It has been shown that γH2Ax/γH2Av marks provide a docking site for many repair factors such as Brca1, NBS1 (also known as NBN) and MDC1 (Kobayashi et al., 2002; Stucki et al., 2005; Lamarche et al., 2010). Recruitment of MDC1 to γH2Ax is the first step through which this mark mediates DSB signalling and repair. MDC1 serves to recruit and tether components of MRN complex, which further facilitate amplification of the γH2Ax signal by a second round of phosphorylation of H2Ax, resulting in the generation of a mega-base level of γH2Ax marks around the damage site (Paull et al., 2000; Stucki and Jackson, 2006). γH2Ax foci have been extensively used as a bonafide marker of the DDR across multiple experimental systems (Rogakou et al., 1999; Madigan et al., 2002; Francisco and Francisco, 2008). Repair of DNA damage is most often accompanied by clearance of γH2Ax marks by phosphatase (PP2A) action, and erasing of γH2Ax marks is essential for attenuating an active DDR, thereby leading to completion of repair (Chowdhury et al., 2005).

Programmed cell death is mediated by a cascade of players, leading to the activation of Cys-Asp proteases (caspases) for execution of cell death (Riedl and Shi, 2004; Fuchs and Steller, 2015). Caspase activation has also been implicated during DNA damage-induced cell death (Norbury and Zhivotovsky, 2004; Roos and Kaina, 2006). In addition to activation of the canonical caspases 8 and 3 downstream of ATM and p53 (also known asTP53) pathway, caspase 2 function has also been implicated in DNA damage-induced cell death (Puccini et al., 2013; Dawar et al., 2016) via the piDDosome complex (PIDD–RAIDD–caspase-2) (Tinel and Tschopp, 2004; Ando et al., 2012; Janssens and Tinel, 2011). Caspase 10 upregulation by p53 following DNA-damaging stimuli, provides another example of a cross-talk between caspases and the DDR (Rikhof et al., 2003). Moreover, caspases are also known to activate kinases such as PAK2 or PKC-δ required for cytoskeletal changes, and MST1 (also known as STK4) for chromatin modification, both leading to efficient apoptosis (Rudel and Bokoch, 1997; Emoto et al., 1995; Wen et al., 2010; Cheung et al., 2003). The requirement for caspases in the activation of caspase-activated DNases (CADs), which are essential for DNA fragmentation, is well established (Yokoyama et al., 2000; Nagata, 2000). γH2Ax has also been observed as a mark of chromatin fragmentation during apoptosis that is artificially induced by, for example, TRAIL ligand or staurosporine treatment (Rogakou et al., 2000; Solier et al., 2009). Cell death-associated γH2Ax marks exhibit ring-type morphology and shows pan-nuclear staining rather than the discrete γH2Ax foci typical of the DDR. They also do not recruit functional MDC1, which distinguishes them from DDR-associated γH2Ax (de Feraudy et al., 2010; Solier et al., 2012; Solier and Pommier, 2011). All the above cited examples underscore the direct involvement of caspases in cell death largely via their protease activity; however, much less is known about their ‘non-apoptotic’ roles, some of which may even be independent of protease activity. Only recently, studies have uncovered multiple other roles for caspases in non-apoptotic process such as dendritic pruning, erythropoiesis, spermatogenesis and even in cell proliferation (Yi and Yuan, 2009). All these newly emerging cases of non-apoptotic roles for caspases raise interesting possibilities of novel cross-talks, perhaps even between caspases and DDR, which we tried to explore here.

Even though the genetic and molecular hierarchy that regulates cell death (apoptosis) onset following DNA damage in a cell is well understood (Norbury and Hickson, 2001; Borges et al., 2009), how the same pathway might potentially cross-talk with the upstream DDR has not been yet explored, which forms the focus of the current study. Here, using Drosophila wing imaginal disc as a model system, we show that in response to DNA damage, the initiator caspase Dronc (the Drosophila homologue of caspase 9) is required for triggering the γH2Av response during DNA damage signalling in a manner that is independent of its known function in effector caspase activation during cell death. We also show that DNA damage-associated γH2Av ensues in intact nuclei, which is markedly different from cell death-induced γH2Av marks associated with fragmented nuclei that require the activity of effector caspase Drice. Furthermore, we show that the γH2Av activation in the context of DNA damage signalling is mediated by ATM kinase, thus suggesting a genetic crosstalk between Dronc and ATM. Thus, based on this, we propose a novel non-apoptotic role for Dronc in promoting DDR versus cell death post DNA damage, which might be dictated by Dronc activation coupled to the extent of DNA damage experienced by a cell in vivo.

The sequence of events involving DNA damage sensing, repair and cell death has been well described both in terms of hierarchy and molecular events (Petrini and Stracker, 2003; Jackson and Bartek, 2009). Caspase activation is often required for dampening of the DDR, thus driving cells towards death following catalytic cleavage of ATM (Smith et al., 1999; Wang et al., 2006), the apical kinase that is the source of γH2Av signalling. However, the temporal activation of such DDR-associated γH2Av by the caspase cascade, if any, has remained elusive. Here, we studied the DDR associated with either drug-induced (cisplatin) DNA damage (genotoxic stress from external causes) or DNA damage in repair-compromised genetic backgrounds (rad51-null mutant; rad51 is also known as spn-A in flies) (genotoxic stress from endogenous causes). Both these paradigms involving either exogenously or endogenously generated genotoxic stress have now become relevant for uncovering plausible therapeutic targets as exemplified by Dekanty et al. (2015).

We probed for the temporal hierarchy between caspase activation and γH2Av signalling following cisplatin induced DNA damage (DSBs) in Drosophila third-instar larvae wing imaginal disc. We used an antibody specific for γH2Av (Lake et al., 2013) (marking the early onset of the DDR) along with an antibody against cleaved caspase 3 (CC3, active Drice), a marker used for initiator caspase Dronc activity (Fan and Bergmann, 2009), and DAPI staining to mark overall nuclear morphology. Time-course analyses revealed a concomitant rise in the levels of γH2Av and CC3, originating in the middle of the wing pouch (around wingless expression boundary) during the first 5 h and spreading more widely by 10 h after treatment, with both markers showing a similar distribution during these initial time points (up to 10 h), but spatially segregating later (beyond 10 h; i.e. 20 h and 30 h), probably due to tissue remodelling post accumulation of dead cells in the wing pouch (Fig. 1A,C,D,E). We note that the area-normalized positive cells per wing disc as a function of time showed a similar pattern for the γH2Av and CC3 signals (Fig. 1C,D).

Furthermore, we calculated the mean intensity of γH2Av in these nuclei and binned them into two categories: intense-discrete and faint γH2Av on the basis of an imposed threshold intensity (Fig. 1B). We focused our analysis only on nuclei displaying high-intensity γH2Av marks, as these γH2Av nuclei correspond to the cells showing high-intensity γH2Av foci in cell culture studies post DNA damage, reflecting true and mature repair foci (Watter et al., 2009). We did not score low intensity or faint (below the imposed threshold) γH2Av as they tend to be close to background staining. Control wing discs showed far fewer γH2Av- and CC3-positive cells, due to low background levels of endogenous DNA damage and sporadic cell death events (Fig. 1F1). The population of γH2Av nuclei in the cisplatin-treated sample were also CC3 positive, with these nuclei displaying different intensities of CC3 staining indicative of differing CC3 levels (marked with coloured boundaries in Fig. 1F2 and quantified as CC3 mean intensity in Fig. 1G): namely, low (Fig. 1F2A, yellow), intermediate (Fig. 1F2b, green) and high CC3 (Fig. 1F2c, red), progressively leading to nuclear fragmentation. Interestingly, a large fraction of the CC3-positive cells that were of the high CC3 category were devoid of γH2Av positivity, perhaps marking the late stage of apoptosis as evidenced by the fragmented morphology of pyknotic bodies (Fig. 1F2d, pink). Quantification in the 10 h cisplatin-treated sample revealed that γH2Av nuclei with high CC3 showed a higher percentage of nuclear fragmentation (on the basis of DAPI stained nuclear morphology) as compared to γH2Av nuclei with low and intermediate CC3 (Fig. 1H). Nuclei with low levels of CC3 were largely intact, and mostly γH2Av positive (Fig. 1F2A,H), perhaps reflecting high DDR. Moreover, the total count of γH2Av-positive nuclei per disc increased in all the three categories of CC3-positive nuclei (low, intermediate and high) of CC3 until after 10–20 h of cisplatin treatment, but dropped somewhat in the 30 h sample (Fig. 1I). More importantly, when we expressed the same data as a relative percentage within each time-point, we found an interesting trend: the relative fraction of γH2Av associated with low CC3 increased dramatically while that with high CC3 decreased as the time of cisplatin treatment increased (Fig. 1J, also reflected in Fig. 1H). This suggested that YH2Av-positive nuclei that were low in CC3 might have protection against damage (as a function of cisplatin treatment time). This could be perhaps attributed to an active DDR, and proficient repair preceding 30 h, which leads to the reduction in high CC3 cells as well as overall cell death levels in the disc at this late time point.

We also probed a rad51-null mutant [spn-A^093A/093A (Staeva-Vieira et al., 2003)] as a genetic model for endogenous DNA damage for another comparison. As expected, control wing discs showed many fewer γH2Av-positive nuclei per disc, whereas mutant wing discs had a higher γH2Av response (Fig. S1A1–A2″). As in the cisplatin-treated set, we observed varying CC3 intensity, with low, intermediate and high levels of CC3 being present in γH2Av-positive nuclei in rad51 mutant wing disc (Fig. S1A2a–A2a″ and A2b1–b4). Quantification revealed a trend that was reminiscent of that observed for cisplatin-induced DNA damage, namely that CC3-positive cells with a lower level of CC3 appeared to have more γH2Av than cells with a high level when expressed as relative percentage levels (Fig. S1C,C′). However, this correlation was not statistically significant, perhaps due to relatively low steady-state level of endogenous damage, which is prone to high stochasticity, in the null mutant tissue as compared to highly time-dependent graded response to damage observed in cisplatin-treated time-course samples (compare Fig. S1A2″ with Fig. 1E4″–E6″).

Furthermore, to distinguish whether the level of γH2Av associated with low to high CC3 stems from a fragmented or intact genome, we performed a TUNEL assay. We analysed the 10 h cisplatin-treated sample for CC3, TUNEL and γH2Av signals. We compared TUNEL positivity vis-à-vis CC3 intensity and found that the majority of cells with low CC3 failed to exhibit any measurable TUNEL signal, whereas cells with intermediate and high CC3 were TUNEL positive (Fig. 2A1–A2″,A2a–c,B). Moreover, high CC3 cells were entirely TUNEL-positive with no TUNEL-negative nuclei (Fig. 2B). Interestingly, most of the γH2Av-positive nuclei (also low CC3) were TUNEL negative, and only a small fraction exhibited TUNEL signal (Fig. 2C1–2″,C2a,b and D). These results, put together, suggested that high CC3 marks fragmented nuclei (Fig. 1H), which are also highly TUNEL positive (Fig. 2B), thus indicating that TUNEL specifically marks late dying cells undergoing genome fragmentation. Since the majority of γH2Av-positive nuclei were also TUNEL negative (Fig. 2D), we tend to conclude that, in the present context, the γH2Av signal reflects a strong DDR mark rather than genome fragmentation associated with cell death. We tried to explore other DDR marks to corroborate the γH2Av results, but unfortunately the lack of Drosophila-specific antibodies precluded these analyses further.

Since the anti-CC3 antibody has been shown to cross-react with additional unknown targets of Dronc that may be involved in non-apoptotic functions (Fan and Bergmann, 2009), we wanted to validate our results by using another antibody specific for Dronc activation. We tested whether another specific marker for Dronc activity, namely the cleaved form of the effector caspase DCP-1 (Drosophila caspase 1), was present in γH2Av-positive nuclei. Even though we observed a high proportion of nuclei with both γH2Av and cleaved DCP-1, unlike for CC3, cleaved DCP-1 showed similar intensities in both γH2Av-positive as well as dying cells (γH2Av-negative nuclei) (perhaps cells undergoing apoptosis) (Fig. 2E,F). We also observed a similar trend in the rad51 mutant, where γH2Av-positive nuclei also showed colocalization with cleaved-DCP-1, thus further confirming our result in another DNA damage regime (Fig. S1D1). Taken together, these results suggest that the γH2Av response coincides with the Dronc activation marks (CC3 and cleaved-DCP-1) which prompted us to probe the epistatic relationship between the two.

We separately tested the genetic requirement of initiator versus effector caspases in triggering the γH2Av response. Firstly, we removed both the effector caspases (Drice and DCP-1) by RNAi followed by DNA damage (10 h cisplatin) treatment, which led to a reduction in CC3 level (although was only partially reduced, if at all, in case of DCP-1 RNAi, as Drice was still active). Interestingly, in both RNAi cases, we observed a high γH2Av response (Fig. 3A1–A2″,B,C; all RNAi controls are shown in Fig. S2A1–A8″), strongly implying that the appearance of γH2Av is not caused by apoptotic nuclear fragmentation or associated activation of CADs, but rather is an output of DDR, where cell death suppression leads to accumulation of high γH2Av due to reduced clearance of DNA-damaged nuclei. The alternative possibility of enhanced cell proliferation in Drice RNAi leading to a high damage response output (γH2Av) was ruled out by analyses of phosphorylated histone H3 (pH3) where we observed that control and RNAi cells showed a similar level this proliferation mark (data not shown). Interestingly, in stark contrast, when we suppressed cell death by means of UAS-DIAP1 (DIAP1 is an inhibitor of both Dronc and Drice), dominant-negative Dronc (UAS-DroncDN) or Dronc RNAi, we observed a reduction in both γH2Av and CC3 (Fig. 3A3–A4″,B,C; Fig. S3A–C). Moreover, the Drice-knockdown compartment following 10 h cisplatin treatment, despite having nuclei with high levels of γH2Av, did not show any TUNEL positivity in comparison with the control compartment (Fig. 3D1). This strengthened our observation that TUNEL specifically labels cells with high levels of CC3, and is more specific to dying pyknotic nuclei, which is a mark of cell death, whereas γH2Av marks DDR-competent DNA damaged cells. The DCP-1 knockdown compartment showed TUNEL-positive cells perhaps indicative of cell death, which was also consistent with the persistence of cell death with DCP-1 RNAi (Fig. 3D2). UAS-DIAP1 and DroncDN showed a reduction in TUNEL-positive cells, most likely due to cell death suppression (Fig. 3D3–D4).

In addition, we did not observe any γH2Av response in Dronc (dronc^I24) mutant wing discs upon 10 h of cisplatin treatment (Fig. S3D1-D2″). Similarly, in the rad51 mutant, suppression of Dronc by either RNAi (Fig. S3E1–E1″,F,G) or with a dominant-negative version (Fig. S3E2–E2″,F,G) led to the suppression of γH2Av along with a reduction in CC3. Together, these results strongly argue in support of the specific requirement for Dronc for triggering or mediating a γH2Av response that is not coupled to cell death. We have already shown above that in the current experimental context, that γH2Av response strongly reflects a DDR rather than cell apoptotic signature. In order to strengthen the hypothesis that it is an increase (rather than the expected decrease if it were a cell death-associated mark) in the γH2Av response that occurs when cells are subjected to DNA damage, such as occurs upon Drice and DCP-1 effector caspase RNAi knockdown, we created a similar situation of ‘undead’ (cells that initiate cell death but fail to die) cells by driving UAS-p35 in the rad51 mutant (p35 is a baculovirus protein; its overexpression is used to artificially block effector caspases, leading to undead cell formation during cell death owing to the presence of active Dronc). This also led to the rise in γH2Av levels (Fig. S3E3–E3″,F,G), which is expressed as normalized to the area of respective compartment (Fig. S3G′). In this context, we believe that Dronc, but not cell death, is perhaps responsible for the rise in γH2Av levels.

We blocked the activity of caspases by using the chemical inhibitor Z-VAD (a tripeptide specific for caspase active site, Z-Val-Ala-Asp), which is a highly specific pan-caspase inhibitor and is widely used in cell culture studies to inhibit apoptosis (Solier et al., 2012; Wen et al., 2010). As expected, wing discs in culture (Fig. 2SB) exhibited cisplatin-dependent DNA damage revealed by γH2Av staining (Fig. 3E2–E2′,G). However, prior treatment of wing disc with Z-VAD not only led to suppression of cell death but also a concomitant reduction in γH2Av levels (Fig. 3E4–E4′,F,G), a result consistent with the genetic results (Fig. 3A3,A4,B,C) pointing to the requirement of caspases in triggering the γH2Av response. Since Z-VAD is a broad-specificity caspase inhibitor, its effects cannot be ascribed to any specific caspase. We also found that overexpressing full-length Dronc led only to a marginal increase in cell death as well as a γH2Av response (Fig. S4A1–A2b′), suggesting that wild-type Dronc on its own is not sufficient for promoting the γH2Av response but requires DNA damage-dependent activation. Furthermore, suppression of pro-apoptotic gene hid by RNAi led to a reduction in γH2Av levels (Fig. S4B1–B2″), suggesting that DIAP1 (which is negatively regulated by hid) might be exerting its negative effect on γH2Av levels by lowering Dronc activity in hid RNAi cells. A similar mechanism has been previously implicated for the non-apoptotic role of Dronc during the compensatory proliferation response (Kondo et al., 2006; Ryoo et al., 2004).

Furthermore, to attest to the lack of nuclear fragmentation, as reflected by the maintenance of nuclear integrity in γH2Av-positive nuclei, we assayed for the intactness of nuclear lamina (Rao et al., 1996; Kihlmark et al., 2001), a bonafide marker of the nuclear membrane. As expected, wild-type controls showed intact lamina except in few sporadic cell death events (Fig. 3H1′,H1a′). Cells in wild-type discs, following a 10 h cisplatin treatment, while undergoing death (high CC3) (Fig. 3Hb″), exhibited disintegration of nuclear lamina (Fig. 3H2b′); however, cells low in CC3 intensity (Fig. 3H2A″) appeared to be intact and did not show any apparent nuclear lamina disintegration (compare Fig. 3H2A′ with H2b′). Furthermore, similar analysis for Drice RNAi (10 h cisplatin) discs showed a stark difference between the anterior (control) and posterior (expressing RNAi) compartments. In the control compartment, we observed CC3-positive apoptotic pyknotic bodies as evidenced by both DAPI staining (Fig. 3H3a) and lamina disintegration [Fig. 3H3a′ (marked with red arrow) and H3a‴]. However, the Drice RNAi compartment did not show any sign of cell death (Fig. 3H3b″), but exhibited an intact nuclear lamina (Fig. 3H3b′ and H3b′″), reflecting intact nuclear morphology, as also evident from DAPI staining (Fig. 3H3b; Fig. S5). This further supported our hypothesis and strongly suggested that the increase in γH2Av-positive nuclei is not associated with nuclear disintegration nor cell death, but arises as a mark of the DDR after DNA damage.

In the following experiments, as a control for comparison, we created genetically induced cell death and monitored the γH2Av response in wing disc system when it was not linked to any DNA damage. The purpose of this approach was solely to define the genetic requirement for the cell death-induced γH2Av response and contrast the same with that of DDR-induced response. Since we relied on γH2Av response as a sole marker for DDR, this internal comparison with the cell death-induced γH2Av response became mandatory.

Since the γH2Ax response can also arise as a result of apoptotic chromatin fragmentation (Rogakou et al., 2000; Solier et al., 2009), we genetically induced cell death using heat-shock-driven Hid (hs-hid) and UAS-rpr driven by vg-Gal4, and we found that CC3-positive cells also exhibited a γH2Av response (Fig. 4A1–A2″; Fig. S5B1–B1′) even in the absence of any DNA damage-inducing treatment. Inducing cell death by 30 min of heat-shock in hs-hid third-instar larvae, clearly showed CC3-positive cells associated with a γH2Av staining in the wing discs (Fig. 4A2). However, controls without heat-shock showed only basal levels of cell death and γH2Av staining (Fig. 4A1–A1″). Furthermore, to distinguish the DDR-associated γH2Av response from that arising out of nuclear fragmentation during apoptosis, we performed a genetic analysis for both initiator and effector caspases after hs-hid apoptosis induction. Suppression of both initiator and effector caspases with UAS-DIAP1 in the background of hs-hid, as expected, led to a reduction of CC3 as well as associated γH2Av positivity (Fig. 4A3–A3″). Interestingly, suppression of Drice using an RNAi in the background of hs-hid led to a marked reduction in both CC3- and γH2Av-positive cells (Fig. 4A4-4″). This was in contrast to what was observed for the DNA damage-induced γH2Av response, where Drice RNAi led only to the suppression of nuclear fragmentation but not the suppression of DDR-associated γH2Av response (compare Fig. 3A1-1″,B,C with Fig. 4A4–A4″). In support of this, we also observed that almost all of the γH2Av-positive nuclei arising as a result of cell death during hid or rpr overexpression was associated with high CC3 and cell death (Fig. 4B1–B1a″ and Fig. S5C1,C2). A high magnification analysis of γH2Av-positive nuclei revealed high CC3 levels and fragmented nuclei, but the cells lacked the graded pattern of CC3 intensities (low, intermediate and high) that was observed in the DDR-induced system above, suggesting that such γH2Av was associated with apoptotic nuclei rather than repairing nuclei (Fig. 4B1a-a″,C1 and Fig. S5C1–C2). In support of this argument, we could show that CC3-positive nuclei exhibit complete nuclear lamina disintegration within 30 min of heat-shock-mediated expression of Hid (Fig. 4C2). One can note in the colocalization image that CC3-positive nuclei have entirely lost lamina staining, while the CC3-negative nuclei in their vicinity fully retain intact lamina staining.

Thus, these experiments clearly indicated that γH2Av associated with apoptosis is markedly different from that in the DDR-induced response, which does not arise from nuclear fragmentation. Furthermore, our results also suggest that the γH2Av-positive nuclei observed here in response to the DDR appear to be different from the pan-nuclear ring forms reported earlier for nuclear apoptosis, which are devoid of active repair due to MDC1 cleavage (Paull et al., 2000; Solier et al., 2009, 2012; Solier and Pommier, 2011). We reiterate that such chromatin fragmentation-related γH2Av staining in apoptotic cells was independent of any DNA damage treatment, thereby precluding the need of any active repair there.

Next, we assessed the relative contribution of apical kinases ATM and ATR in mediating the γH2Av response (Bartek et al., 2007). Suppression of ATM in the rad51 mutant or during cisplatin treatment (10 h) led to the abrogation of γH2Av response, but surprisingly also led to slight increases in CC3 levels, indicating the requirement of ATM for H2Av phosphorylation but not for cell death, and suggesting that ATM function is downstream of caspase activation (Fig. 5A1–A1″,B,C,D1–D1″,E ,F). However, suppression of ATR led to different results: while ATR suppression had almost no effect in the rad51 mutant (Fig. 5A2–A2″,B,C), it led to a significant increase in both γH2Av-positive nuclei (as a result of increased genomic instability) and cell death during cisplatin-induced damage (Fig. 5D2–D2″,E,F). This can be rationalized by the fact that ATR, unlike ATM, is known to play a relatively more important role in later steps of DSB repair (LaRocque et al., 2007), but may not be required for early sensing of DSBs via γH2Av. No difference was observed with ATR RNAi in the rad51 mutant, perhaps due to the low steady state level of DNA damage in the mutant as compared to relatively high damages induced in cisplatin-treated wing discs where loss of ATR function leads to enhanced genome instability. It has already been shown that ATM rather than ATR function is more important during the DDR to low levels of DNA damage in Drosophila (LaRocque et al., 2007; Bi and Gong, 2005). This interpretation is also consistent with ATM being predominantly associated with telomere maintenance (Bi et al., 2005), while ATR performs more wider functions in the DDR in Drosophila. We also confirmed these results by blocking ATM and ATR function independently, using specific inhibitors for ATM (KU55933) and ATR (NU6027), which recapitulated the results obtained in the genetic knockdown experiments (Fig. 5G,H). Moreover, our results showing that loss of ATM leads to no significant increase in cell death is also consistent with a published report where only about an twofold increase in cell death and a marginal rise in lethality was observed in the γH2Av mutant (Madigan et al., 2002).

Taken together, our results raise an interesting possibility of a novel role for caspases (specifically Dronc) acting upstream of the ATM-mediated γH2Av response, as summarized in Fig. 5I. Our model is in contrast with the known functions of caspases, largely developed on mammalian cell line studies where DDR is attenuated by catalytic cleavage of ATM leading to cell death (Smith et al., 1999; Wang et al., 2006). However, the molecular basis of Dronc acting upstream to ATM to promote DDR requires further investigation. Molecular analyses of Drosophila ATM, its possible cleavage or the cleavage of a putative negative regulator leading to ATM activation are all open possibilities that await future studies.

Most studies ascribe caspase activation leading to cell death as a response either downstream to or parallel to DDR associated with excessive or unrepaired DNA damage. We postulated that given the possibility of intense cross-talk between these two cellular pathways, upstream regulation of DDR by caspases cannot be completely ruled out. Recent findings uncovering novel non-apoptotic functions of caspases also makes this an exciting possibility. In the current study, by performing genetic epistasis experiments between cell death and DDR, we uncovered an interesting dependence of DDR on initiator caspase Dronc in the Drosophila wing disc, as discussed below.

We observed that suppression of Dronc led to a reduction in the γH2Av response, whereas, in contrast, removing the effector caspases Drice and DCP-1 led rather to a significant increase in γH2Av-positive nuclei (Fig. 3; Fig. S3). These results, where the γH2Av response was apparently not dependant on effector caspases, and therefore perhaps not directly linked to the execution of cell death, implied that the response was not related to nuclear fragmentation. This was supported by the result showing that a larger fraction of γH2Av-positive nuclei were low in CC3 (a marker used for initiator caspase Dronc activity) and also failed to show nuclear fragmentation as evidenced by the absence of TUNEL signal (DDR proficient), whereas, in contrast, a smaller fraction of γH2Av-positive nuclei high in CC3 were also TUNEL positive (apoptotic) (Fig. 2). Nuclear lamina staining provided a direct readout of intact nuclear morphology in the above described DDR proficient versus fragmented nuclear morphology in apoptotic nuclei (Fig. 3). While all these observations collectively pointed towards the exciting possibility of DDR (via its γH2Av response) being regulated upstream by the initiator caspase Dronc, we were curious to assess how the γH2Av response might be different should the cells be induced to undergo apoptosis even in the absence of DNA damage.

Therefore, next, we probed γH2Av response in cells where apoptosis was artificially induced by genetic manipulation (hs-hid and UAS-rpr driven by vg-Gal4) where such death induced γH2Av-positive nuclei that uniformly colocalized with high CC3 and were also positive for nuclear fragmentation despite of no DNA damage-inducing treatment (cisplatin). Interestingly, suppression of both initiator and effector caspase with UAS-DIAP1 or Drice RNAi in the background of hs-hid led to a drop in γH2Av positivity (Fig. 4), thus showing that the γH2Av response during cell death had a strong dependence on effector caspase activation unlike that of DDR-associated response.

Integration of these results involving genetic epistasis in an exogenously induced (cisplatin treatment) DNA damage system in wild-type wing discs or in a system with physiologically generated endogenous damage (rad51 mutant wing discs) where the DDR-centric γH2Av response predominates prompted us to speculate about how might Dronc act differently to in a cell death response. Here, we speculate that there might exist a distinct signalling axis involving Dronc leading to either DDR or a cell death response where a DNA damage-based threshold mechanism might operate. Specifically, we propose that Dronc plays a dual role that can either initiate DDR or apoptosis depending upon the level and the required threshold of its activation in damaged cells. Activating Dronc below a certain threshold required for cell-death might well promote DDR and cellular repair, while the same above the set threshold leads to Drice activation sufficient for cell death. We note that similar threshold-based mechanism for apoptotic versus non-apoptotic roles of caspases have already been proposed in Drosophila wing and eye imaginal discs (Florentin and Arama, 2012).

In conclusion, we posit that caspase acting upstream of γH2Av activation means there is a possible non-apoptotic role for Dronc in the DDR, which is consistent with several recently described examples of non-apoptotic functions ascribed to caspases. The mechanistic basis of non-apoptotic activation of caspases both during differentiation and cellular proliferation is not yet well addressed. It is however, suggested that restricting caspase activity to a localized sub-area within the cell might facilitate a novel specific outcome mediated by the caspase, as observed during dendritic pruning and spermatogenesis (Wang and Sheng, 2014; Feinstein-Rotkopf and Arama, 2009; Nakajima and Kuranaga, 2017). Radically, non-protease-dependent new functions of caspases cannot be ruled out, although they are yet to be discovered. Our current experimental paradigm involving DDR in wing disc might offer an interesting opportunity to explore this new area of research for discovering such novel functions of caspases. In the current experimental context, we speculate that Dronc might execute its regulatory function on ATM kinase perhaps by cleaving a putative inhibitor of ATM (a novel protease function), thereby enhancing the γH2Av response or alternately that it might delay the attenuation of the DDR signal by directly interacting with chromatin via some mediator proteins (non-protease function), and thus providing a cell enough time to repair, or that indeed both of these processes might function. However, to understand and explain these possibilities better, it is important to perform detailed biochemical and proteomic studies on the Dronc interactome. Such future studies might unveil novel exciting therapeutic targets directed against the DDR via the caspase interactome.

Fly stocks and their sources are included in the Table S1.

Imaginal discs were fixed using 4% formaldehyde in 1× PBS, followed by 2 h blocking with 0.5% bovine serum albumin (BSA) dissolved in 0.1% 0.1% PBST (1×PBS with 0.1% Triton X-100). Primary antibody incubations were performed at 4°C overnight, followed by 2 h secondary incubation at room temperature. All washes following antibody incubation were performed in 0.1% PBST. Primary antibodies were against CC3 (Sigma 8487, 1:200), active DCP-1 (Cell Signalling 9578, 1:200), γH2Av (DSHB, 1:50), engrailed (DSHB, 1:50) and GFP (DSHB, 1:50). Alexa-Fluor-conjugated secondary antibodies were used at 1:200 dilution. TUNEL labelling was performed by using the S7111 ApopTag (R) Plus in situ apoptosis fluorescein detection kit as per the protocol already described for Drosophila larval imaginal discs (Chakraborty et al., 2015). Image acquisition was performed on Olympus FV 1000 microscope, and intensity adjustments were made using Photoshop PS5 and Image J during image processing. For whole disc imaging, the entire set of stacks were z-projected, whereas for single nuclei depiction, three or four stacks were maximum intensity projected.

Cisplatin treatment was performed by feeding early third-instar stages with cisplatin (dissolved in DMSO to 40 mM and added to the fly medium to the final concentration of 1 mM) for 2, 5, 10, 20 and 30 h (method of feeding was adapted from Kooistra et al., 1999). Larvae, feeding on cisplatin-containing medium, were collected at regular intervals of 2, 5, 10, 20 and 30 h. For controls, a similar amount of DMSO was added to the medium, followed by sample collection.

In our analysis, γH2Av nuclei were analysed for their mean intensity with Image J across samples from 10 h cisplatin treatment and were binned in two categories: intense-discrete and the other being faint, on the basis of threshold of intensity (Fig. 1B). Moreover, we also observed that γH2Av intensity across nuclei in the same set was not highly variable. We considered only the intense-discrete γH2Av nuclei for further analysis as γH2Av-positive nuclei and for counting of numbers (using the Image J cell counter) from whole discs. Similar, intensity analysis was performed for CC3-positive cells by calculating mean intensities from the 10 h cisplatin treated sample. CC3-positive cells were placed into three different intensity-categories (low, intermediate and high) on the basis of mean intensity threshold (Fig. 1G). Apoptosis was quantified by drawing a boundary manually around the CC3-positive (high CC3) area using Image J, and normalizing to the total disc area as a percentage. For all the Gal4 perturbation experiments, quantification was done separately in Gal4-expressing and non-expressing regions. For independent and dependent samples, the Student's t-test was used to perform statistical analysis using socscistatistics tool (http://www.socscistatistics.com).

In vitro conditions were standardized to culture wing imaginal discs by incubating larvae heads in Schneider's medium with 2% fetal bovine serum (FBS). Wing discs from each culture condition were scored for cell death and proliferation for up to 8 h (Fig. S2B). For cell death suppression, Z-VAD (OMe)-FMK (Abcam) was used at 200 μM concentration and for DNA damage signalling ATM inhibitor (KU55933) (Calbiochem) and ATR inhibitor (NU6027) (Calbiochem) were used at 20 μM. Discs were pre-incubated with inhibitors for 1 h prior to the treatment in 5 h cisplatin.

We thank Trudi Schupbach (Princeton University) for the rad51 stocks; Prof. L.A. Baena-Lopez (University of Oxford) for the dronc allele and valuable input; Profs. Maithreyi Narasimha and Ullas Kolthur from TIFR for critical input, and the Bloomington Stock Center, Developmental Studies Hybridoma Bank and Vienna Drosophila Resource Center for reagents.