Deregulated G1–S control and energy stress contribute to the synthetic-lethal interactions between inactivation of RB and TSC1 or TSC2

Inactivation of the gene encoding retinoblastoma-associated protein (RB1, hereafter referred to as RB) by deletion, mutation, loss of expression or functional inactivation is considered a general feature of cancer (Sherr and McCormick, 2002). In addition to deregulating cell proliferation, loss of RB function increases apoptosis at least in part through de-repressing activator-type E2F proteins (Chen et al., 2009; Gordon and Du, 2011). Cells that lose tumor suppressors such as RB become dependent on supporting pathways for survival. This reliance on parallel signaling pathways can potentially be exploited by using a synthetic-lethal approach (Kaelin, 2005) – targeting the genes required for the survival of cells that lack RB function.

The Drosophila genetic model provides a unique opportunity to identify RB synthetic-lethal mutations in vivo, as this organism harbors well-conserved yet relatively simple RB–E2F gene families (van den Heuvel and Dyson, 2008). We conducted a forward screen to identify genes that are important for cell survival and differentiation in the absence of the RB ortholog Retinoblastoma family protein rbf (Steele et al., 2009; Sukhanova et al., 2011; Tanaka-Matakatsu et al., 2009) and identified the fly TSC2 ortholog gig (encoding the protein Gigas), which induces synergistic cell death upon rbf loss (Li et al., 2010). This synthetic-lethal interaction between RB and TSC2 is conserved between flies and human cancer cells, as we showed that inactivation of RB and TSC2 leads to the induction of excessive cellular stress, including ROS, which contributes to synergistic cell death in cancer cells in vitro and using a xenograft model (Danos et al., 2012; Li et al., 2010). However, the in vivo mechanisms that mediate the synthetic lethality of RB and TSC2 mutations are unclear.

The products of the tuberous sclerosis genes TSC2 and TSC1 function together to restrict cell growth by inhibiting activation of the rapamycin-sensitive complex TORC1 (Potter et al., 2001; Tapon et al., 2001). Mutation of either TSC2 or TSC1 causes tuberous sclerosis complex (TSC), an autosomal-dominant tumor syndrome (Orlova and Crino, 2010). Studies have shown that mutations in these genes induce endoplasmic reticulum (ER) stress, leading to activation of the unfolded protein response (UPR) and susceptibility to apoptosis (Ozcan et al., 2008). Additionally, deregulation of TORC1 causes glucose addiction (Inoki et al., 2003), as these cells are sensitive to reduced ATP levels and depend on energy stress signaling (Choo et al., 2010). Therefore deregulated TORC1 activity promotes cell growth but also sensitizes cells to nutrient deficiency and/or metabolic-stress-induced cell death.

Here, we show that aberrant S phase entry resulting from inactivation of both rbf and gig/dtsc1 causes increased DNA damage and cell death. Additionally, we show that loss of either rbf or gig/tsc1 induces energy stress and sensitizes cells to ATP depletion, leading to the dependence on signaling by the serine/threonine-protein kinase LKB1 for viability. These results provide new insights into the mechanisms that mediate synergistic cell death when RB and TSC2 are both inactivated and suggest new therapeutic approaches that potentially can be used to target RB-deficient tumors.

We previously showed that loss of gig causes synergistic apoptosis and ablation of rbf-mutant tissue in mosaic eyes and wings (Li et al., 2010). As mutation of either rbf or gig causes G1–S deregulation, we investigated the effect of inactivating both rbf and gig (hereafter indicated ‘rbf,gig’) on G1–S regulation and its impact on synergistic apoptosis. Cells in the morphogenetic furrow (MF) of developing eye discs arrest in G1. While mutation of neither rbf nor gig affects cell cycle arrest in the MF, ectopic S phase cells are observed in rbf,gig double-mutant clones (Fig. 1A–C′), indicating that rbf and gig cooperate to enforce G1 arrest. In addition, overexpressing dE2f1 (together with dDp) together with Rheb, encoding a GTPase that is regulated by gig and activates TORC1 directly (Saucedo et al., 2003; Zhang et al., 2003), leads to synergistic S phase in the MF and posterior (supplementary material Fig. S1). These results, in conjunction with the observations of ectopic S phase in rbf,tsc1 clones (Hsieh et al., 2010), indicate that deregulated E2F and Rheb or TORC1 signaling induce synergistic S phase.

We previously showed that removing the transcription activation function of de2f1 or inactivation of s6k suppressed the synergistic cell death effect of rbf,gig (Li et al., 2010). To investigate whether the effects of de2f1 or s6k mutation on the synergistic cell death of rbf,gig clones are correlated with their effects on increased S phase, we introduced either de2f1 or s6k mutation into the rbf,gig-mutant background and found that each suppresses ectopic S phase within the MF (Fig. 1D–E′). As caspase activation can induce compensatory proliferation in flies (Ryoo et al., 2004), we tested whether the proliferative phenotype is a consequence of apoptosis by removing the apoptotic gene hid (Tanaka-Matakatsu et al., 2009) from rbf,gig-mutant clones, which significantly reduces the level of apoptosis (Li et al., 2010). As shown in Fig. 1F, inhibition of cell death does not prevent ectopic S phase in the MF (Fig. 1F,F′). These results demonstrate that the transcription-activation function of dE2f1 (Royzman et al., 1999) and TORC1 signaling through S6k (Radimerski et al., 2002) are required for deregulated cell proliferation in the absence of both rbf and gig, which correlates with, but does not result from, apoptosis.

To characterize further the relationship between deregulated proliferation and synergistic cell death in rbf,gig double-mutants, we used fluorescence-activated cell sorting (FACS) to analyze cells dissociated from mosaic wing discs and compared the cell cycle profiles of the rbf and gig mutants alone or in combination. This technique separates mutant from wild-type cells based on GFP expression, thereby providing internal controls for each genotype (Danos et al., 2012). Wild-type cells in late third-instar wing discs arrest in G2. As RBF–E2F regulates both the G1–S as well as G2–M transitions, rbf loss causes a reduction of the G2 population compared with controls (Fig. 2A,A′,2E; P<0.01 between rbf and GFP control cells), consistent with other reported results (Datar et al., 2000). By contrast, mutation of gig, which only promotes the G1–S transition, causes an accumulation of G2–M cells (Fig. 2B,B′,E; P<0.05 between gig and GFP control cells) similar to that reported previously (Gao and Pan, 2001). We also found a modest accumulation of sub-G1 cells among gig mutants that is reminiscent of cell death observed in this tissue (Li et al., 2010). In rbf,gig double-mutants, FACS analysis shows a larger population of sub-G1 cells, reflecting elevated cell death. Interestingly, we also found a corresponding reduction in the G2–M peak (Fig. 2C,C′,E; P<0.01 between rbf,gig and gig), suggesting that cell cycle progression is important for the apoptotic effect. To further test this idea, we determined whether inhibiting cell death affects the rbf,gig cell cycle profile by introducing a mutation in the apical caspase dronc (Steele et al., 2009). We found that rbf,gig,dronc triple-mutants exhibit a decreased sub-G1 population compared with that of rbf,gig double-mutants (Fig. 2D,D′), consistent with the observation that dronc partially suppresses the death of rbf,gig mutants (Li et al., 2010). Importantly, we found that suppressing cell death also partially restores the G2–M peak of the cell cycle profile, whereas mutation of dronc alone did not affect cell cycle phasing (Fig. 2C–E; P<0.05 between rbf,gig and rbf,gig,dronc). Taken together, these results indicate that cell death in the absence of both rbf and gig occurs at the expense of cells that have passed the G1–S checkpoint.

As blocking death alters the G2–M peak of rbf,gig mutants, we further tested whether defective cell cycle regulation contributes to the synergistic cell death phenotype. We thus overexpressed G1 or G2 cell cycle inhibitors specifically in the posterior of mosaic wing discs, thus allowing for the comparison of rbf,gig-mutant clones either with or without transgene overexpression in the same tissue. Expressing the p27 KIP1 ortholog Dacapo (Dap) causes an extended G1 phase, whereas expression of dWee1 prolongs G2 during development (Reis and Edgar, 2004). We found that Dap expression decreases Caspase-3 activation in rbf,gig-mutant clones compared with control clones in the anterior (Fig. 3A,A′,H). This phenotype is evident despite Dap expression on its own causing reduced tissue growth and increased basal levels of apoptosis. By contrast, expressing dWee1 does not significantly alter cell death in rbf,gig-mutant clones (Fig. 3B,B′,H). These results indicate that rbf,gig-mutant wing disc cells that have a reduced ability to enter S phase are protected from apoptosis, whereas inhibiting cell cycle progression through the G2–M checkpoint does not have an observable effect.

We further tested whether inhibiting G1–S progression affects cell death induced by dE2F1,dDp and Rheb overexpression. Clones of cells overexpressing dE2f1,dDp in eye discs show a pattern of apoptosis near the MF (Fig. 3C,C′) similar to rbf mutation (Moon et al., 2006; Tanaka-Matakatsu et al., 2009). Coexpression of Rheb with dE2f1,dDp significantly increases the level of cell death in these eye discs, particularly in clones anterior to the MF (Fig. 3D–E′,3I). Importantly, overexpression of Dap, while inhibiting ectopic G1–S progression (supplementary material Fig. S1), significantly decreases cell death (Fig. 3F,F′,I). Taken together, these results show that precocious S phase in rbf,gig-mutant clones probably contributes to the synergistic cell death phenotype.

Recent studies show that RB protects cells against chromosome instability and DNA damage (Coschi et al., 2010; Longworth et al., 2008; Manning et al., 2010; van Harn et al., 2010). To test whether rbf mutation during larval development induces DNA damage, we labeled mosaic eye discs with antibodies specific for γ-H2Av (Mehrotra and McKim, 2006a), a phosphorylated form of the Drosophila histone H2A variant that serves as a marker for DNA double-strand breaks (DSBs). Significant levels of γ-H2Av foci are observed in rbf-mutant clones (Fig. 4A,A′,I). While loss of gig also causes a low level of γ-H2Av accumulation proportional to mutant clone area (Fig. 4B,B,I), rbf,gig double-mutant clones show significantly increased γ-H2Av foci per unit clone area (Fig. 4C,C′,I). Therefore synergistic cell death in rbf,gig double-mutant clones correlates with DSB accumulation. Importantly, the increased DSBs in rbf,gig double-mutant clones are dependent on dE2f1 and TORC1 activities, as both de2f1 and s6k mutants suppress the γ-H2Av phenotype of rbf,gig mutants (Fig. 4D–E,I).

To determine whether precocious S phase induced by deregulated RB–E2F and Rheb–TORC1 is important for the accumulation of DNA DSBs, we overexpressed Dap in rbf,dtsc1 double-mutant clones using the mosaic analysis with a repressible cell marker (MARCM) approach (Lee and Luo, 2001). Similar to rbf,gig, inactivation of both rbf and tsc1 leads to synergistic cell death (Hsieh et al., 2010), which we also found to be correlated with high levels of DSBs (Fig. 4F,F′,J). Interestingly, partial inhibition of the G1–S transition by Dap reduces cell death (supplementary material Fig. S2) and decreases the level of γ-H2Av foci (Fig. 4G,G′,J) in rbf,dtsc1 clones. By contrast, expression of the G2–M regulator dWee1 does not reduce the levels of either cell death (supplementary material Fig. S2) or γ-H2Av foci (Fig. 4H,H′,J) in rbf,dtsc1 clones. To characterize further the relationship between increased DNA damage and induction of cell death in rbf,gig double-mutant clones, we determined the effect of impairing DSB repair by mutation of spnA, the Drosophila rad51 homolog required for DNA repair by the homologous-recombination pathway (Sekelsky et al., 2000). As shown in supplementary material Fig. S3, spnA mutation significantly increased both the number of DNA breaks (supplementary material Fig. S3A–C; P<0.001) as well as the level of cell death in rbf,dtsc1 mutant clones (supplementary material Fig. S3D–F; P<0.05).

Taken together, these results indicate that G1–S deregulation in rbf,gig clones induces DNA DSBs, which contribute to synergistic cell death. However as cell death in rbf,gig double-mutants is only partially suppressed by inhibiting S phase, it is possible that other forms of stress beyond DNA damage also affect the survival of these cells.

In addition to regulating cell growth and proliferation, tumor suppressors such as RB and TSC2 also regulate the energy-intensive metabolic processes of DNA replication and protein synthesis. Therefore, inactivation of rbf and gig/dtsc1 could cause metabolic and energetic imbalance. To test this, we induced mutant clones in a Minute background to generate eye-antenna discs that comprised mostly mutant tissue and determined the ratio of intracellular ATP:ADP within these discs. As a control, we determined the effect of mutant cytochrome c oxidase Va (cova), which encodes a mitochondrial protein that was shown to abrogate ATP synthesis and induce cell cycle arrest (Mandal et al., 2005; Owusu-Ansah et al., 2008). A modest but statistically significant decrease in the ATP:ADP ratio is observed in cova-mutant eye-antenna discs compared with controls (Fig. 5A; P = 0.003). A similar decrease in the ATP:ADP ratio is also observed in rbf- or tsc1-mutant discs compared with controls (Fig. 5A; P≤0.02); however, an even more pronounced decrease in the ATP:ADP ratio is observed in rbf,tsc1 double-mutant discs (P<0.03 compared with rbf- or tsc1-mutant discs). These results indicate that inactivation of rbf or tsc1 induces energy stress, which is more severe in rbf,tsc1 double-mutants. To characterize further the effect of deregulating RB–E2F and TORC1 signaling on cellular energy levels, we determined the effect of overexpressing Rheb and dE2F1–dDp. Decreased ATP:ADP ratios were observed in dE2f1,dDp,Rheb-overexpressing eye discs, similar to rbf,tsc1 double-mutant discs (supplementary material Fig. S4), suggesting an important role of RB–E2F and TSC–TORC1 signaling pathways in regulating energy metabolism, in addition to their roles in the cell cycle and cell growth.

Signaling by LKB1 is known to be required for balancing cellular energetic needs and supply through AMPK-dependent and -independent pathways (Shackelford and Shaw, 2009). The altered metabolism and increased cellular stress we observed in rbf- and tsc1-mutant clones suggests the likelihood that the viability of rbf- or dtsc1-mutant cells depends on LKB1 function. Thus, we used an lkb1 mutation (Lee et al., 2006) to test this possibility directly. Although mutation of lkb1 alone does not show a significant apoptotic effect (Fig. 5B), lkb1 loss significantly enhances cell death in rbf-mutant clones in both wing (supplementary material Fig. S5) and eye imaginal discs (Fig. 5E,G) and significantly decreases the contribution of mutant clones to mosaic eye tissue in adults (Fig. 5I,L,M). As loss of lkb1 does not significantly affect ATP levels either on its own or in combination with rbf mutation (Fig. 5A), it is likely that the cell death effect of lkb1 mutation is mediated by the inability of mutant cells to respond properly to energy stress. Similarly, inactivation of lkb1 also increases cell death in dtsc1- and rbf,dtsc1-mutant clones during larval development (Fig. 5C–H) and leads to the development of adult eyes that are largely devoid of mutant tissue (Fig. 5I–O).

rbf,tsc1 double-mutants show significantly decreased ATP levels compared with those of either single mutant (Fig. 5A), raising the possibility that excessive metabolic stress and ATP depletion contribute to the synergistic cell death. To test this, we impaired ATP production using the aforementioned cova mutant. rbf,cova double-mutant discs have significantly lower ATP levels than either single-mutant (Fig. 5A; P<0.001). Importantly, this decrease in ATP levels correlates with increased apoptosis in rbf,cova double-mutants, particularly anterior to the MF (Fig. 6G–I; Fig. 6M). Furthermore adult eyes are largely devoid of rbf,cova double-mutant tissue, even when induced in a Minute background, whereas both single mutants are visible (Fig. 6A–C). Similarly loss of cova significantly increases cell death in dtsc1 and rbf,dtsc1 mutant clones (Fig. 6J–M) and significantly decreases the amount of dtsc1-mutant adult tissue while eliminating the rbf,dtsc1-mutant tissue from adult eyes (Fig. 6B,D–F). These results indicate that loss of rbf or gig/dtsc1 alters metabolism and sensitizes cells to additional metabolic stress or energy depletion and that combining these mutations results in excessive metabolic stress that contributes to synergistic cell death, resulting in the elimination of mutant tissue from adult structures.

We previously showed that antagonizing JNK signaling during larval development reduces apoptosis in rbf,gig wing disc clones (Li et al., 2010), indicating that activated JNK signaling also contributes to rbf,gig cell death. JNK activation can be monitored by expression of puckered (puc) (Martín-Blanco et al., 1998; Riesgo-Escovar et al., 1996) or directly using a phospho-specific antibody against JNK (Ohsawa et al., 2011). We observed elevated JNK signaling in eye discs mosaic for gig loss (Fig. 7B,B′; supplementary material Fig. S6F,F′) or overexpressing a strong Rheb construct (Fig. 7E,E′), whereas neither loss of rbf (Fig. 7A,A′) nor overexpression of dE2f1,dDp (Fig. 7D,D′) causes observable JNK activation. Furthermore, loss of rbf does not obviously affect JNK activation due to loss of gig (Fig. 7C,C′), nor does dE2f1,Dp coexpression affect JNK activity induced by Rheb (Fig. 7F,F′). Notably, mutation of s6k blocked elevated levels of phospho-JNK induced by gig mutation (supplementary material Fig. S6G,G′), indicating that deregulated TORC1 signaling through S6k is required for the activation of JNK signaling. While inhibiting JNK signaling by overexpressing Puc reduces cell death in rbf,gig double-mutants (Li et al., 2010), we also found that Puc reduces the apoptotic effect of dE2f1,Dp,Rheb (Fig. 3G,G′) as well as that induced by overexpression of a strong Rheb construct on its own (supplementary material Fig. S6A–E). Therefore, deregulated TORC1 activity activates JNK-mediated stress signaling, which also contributes to the rbf,gig cell death phenotype.

We have shown that cell death induced by simultaneous inactivation of rbf and gig/dtsc1 correlates with synergistic deregulation of G1–S control and increased energy stress. While it is not surprising that deregulated RB–E2F and TSC–TORC1 pathways leads to uncontrolled S phase progression, it is not expected that the observed G1–S deregulation contributes to cell death. We found that inactivation of rbf and gig/dtsc1 also leads to the increased accumulation of DSBs. Manipulations that suppress deregulated G1–S progression, by de2f1 and s6k mutations or Dap overexpression, reduce the accumulation of DSBs and suppress synergistic cell death of double-mutants. By contrast, inhibition of G2–M progression by expression of dWee1 failed to affect double-mutant DSB accumulation or synergistic cell death. These observations suggest that inactivation of rbf and gig/dtsc1 causes deregulated G1–S progression, leading to DNA DSB accumulation and contributing to synergistic cell death (Fig. 8).

In addition to inducing DNA DSBs, our results show that inactivation of rbf and gig/dtsc1 also induces energy stress. Evidence of the ability of oncogenes and tumor suppressors to modulate metabolism in addition to their roles in regulating cell proliferation and/or differentiation is emerging. TORC1 signaling, for example, is known to promote both G1–S progression as well as anabolic processes such as protein and lipid synthesis, and TORC1 deregulation sensitizes cells to various forms of stress, including energy stress. Indeed, studies have shown that mammalian cells lacking TSC1 or TSC2 are particularly sensitive to glucose deprivation (Inoki et al., 2003). Similarly recent studies show that RB–E2F regulates genes involved in oxidative metabolism (Bateman and McNeill, 2004) and that RB is required for the upregulation of mitochondrial biogenesis during erythropoiesis (Sankaran et al., 2008). Interestingly, cells with either rbf or gig/dtsc1 mutations crucially depend on cellular mechanisms that balance metabolic or energy supply and demand. While we found that inactivation of the metabolic regulator LKB1 significantly increases the death of rbf or dtsc1 mutants and largely eliminates mutant tissue from adult eyes, others showed LKB1 to be required for hematopoietic stem cell survival in mice (Gan et al., 2010; Gurumurthy et al., 2010; Nakada et al., 2010). Furthermore rbf and gig/dtsc1 mutants are sensitive to ATP depletion, as inactivation of cova, which impairs ATP production, also increases the death of either rbf or dtsc1 mutants and largely eliminates mutant tissue from adults. Therefore the enhanced energy stress we observed in rbf,gig double-mutants also probably contributes to the synergistic cell death phenotype (Fig. 8).

In an earlier study, we showed that another cellular stress, oxidative stress, is synergistically induced by inactivation of RB and TSC2 and contributes to the synergistic death of human cancer cells (Li et al., 2010). Notably, different types of stress in mammalian cells are often linked. For example, increased energy stress can induce accumulation of ROS as cells try to balance energy needs with supply by promoting mitochondrial oxidative phosphorylation, which is a principle intracellular source of oxidative stress. Although cancer cells often exhibit aerobic glycolysis, an observation known as the Warburg effect (Vander Heiden et al., 2009), they usually still depend on mitochondrial oxidative phosphorylation for survival (Fogal et al., 2010; Funes et al., 2007; Weinberg et al., 2010). Accumulation of ROS can also contribute to oncogene-induced DNA damage and cell death (Vafa et al., 2002), while increased DSBs increase ROS levels (Kang et al., 2011). Therefore increased energy stress and DSBs can induce accumulation of ROS, which contributes to the death of RB-mutant human cancer cells in which TSC2 is inactivated. By contrast, ROS do not appear to play a sufficient role in Drosophila apoptosis. For example, mutation of Pdsw, which encodes a component of mitochondrial complex I, causes accumulation of very high ROS levels and G1 cell cycle arrest but not cell death (Owusu-Ansah et al., 2008). Therefore, mechanisms in addition to ROS probably contribute to metabolic- or energy-stress-induced cell death in Drosophila. Interestingly, a recent study showed that caspase activation and cell death can be directly modulated by metabolic byproducts in flies (Kim et al., 2008).

A variety of cellular stresses can activate JNK signaling (Johnson and Nakamura, 2007). We found that JNK activity is involved in the sensitivity of rbf,gig cells to apoptosis. As Drosophila JNK signaling has been shown to promote expression of hid (Luo et al., 2007; Ryoo et al., 2004), an apoptotic gene similar to Diablo (Smac) that is involved in dE2f1-mediated cell death, it is possible JNK activation by deregulation of TORC1 can synergize with the loss of rbf to induce apoptosis owing to their effects on apoptotic gene expression (Fig. 8). The involvement of multiple mechanisms in the synergistic cell death of these double-mutants is consistent with our observations that hid and dronc are both important, but neither is absolutely required for the cell death phenotype of rbf,gig-mutant clones (Li et al., 2010). It should be noted that the ability of gig, tsc1, lkb1 or cova to induce synergistic cell death with rbf varies depending on the different regions of the eye disc. This is likely due to the fact that additional cell-intrinsic and extrinsic factors also influence the level of synergistic cell death. For example, it has been shown that Hid protein is preferentially accumulated near the MF in rbf-mutant eye discs (Tanaka-Matakatsu et al., 2009) and that epidermal growth factor receptor (EGFR) signaling, which is an important survival signal in the posterior of the eye disc (Yang and Baker, 2003), negatively regulates rbf-mediated cell death (Moon et al., 2006).

The mechanisms revealed by this and our previous research suggest there could be multiple benefits of transient activation of TORC1 signaling in therapies that target cancers with inactivated RB. In addition to inducing a significant death phenotype in RB-deficient cells, activation of TORC1 signaling in the absence of RB is expected to deregulate G1–S control synergistically and induce metabolic and energy stress. These effects could further sensitize RB-deficient cancer cells to conventional cancer treatments such as radiation and chemotherapy, as well as to new strategies that target cancer metabolism (Vander Heiden, 2011). As resistance to cancer therapy is believed to arise from the persistence of subsets of cells, such as cancer stem cells that survive the cytotoxic effects of chemotherapy or radiation by exiting the cell cycle and remaining dormant until growth conditions are more favorable (van Gent et al., 2001), transient activation of TORC1 activity might lead to better cure rates for RB-deficient cancers.

The fly stocks used in this study were: rbf^15aΔ (Tanaka-Matakatsu et al., 2009), de2f1ⁱ² (Royzman et al., 1999), de2f1^rm729 (Duronio et al., 1995), gig⁶⁴ (Li et al., 2010), dtsc1²⁹ (Gao and Pan, 2001), s6k^l1 (Radimerski et al., 2002), lkb1^X5 (Lee et al., 2006), cova^tend (Mandal et al., 2005), hid¹³⁸ (Tanaka-Matakatsu et al., 2009), dronc⁰¹ (Tanaka-Matakatsu et al., 2009), puc^E69 (Riesgo-Escovar et al., 1996), UAS-E2f1,UAS-dDp (Bloomington stock no. 4774), UAS-Dap (Reis and Edgar, 2004), UAS-dWee1 (Price et al., 2002), UAS-Puc (Martín-Blanco et al., 1998), and UAS-Rheb(III) or strong UAS-Rheb(II) (Saucedo et al., 2003).

Flies were maintained at 25°C on standard media containing yeast, cornmeal, sugar and agar. The genotypes used were as follows:

yw, hsFLP/Y; Ubi-GFP, FRT80B/gig, FRT80B

rbf^15aΔ,w, hsFLP/Y;RBF-G3, Ubi-GFP, FRT80B/FRT80B

rbf^15aΔ,w, hsFLP/Y;RBF-G3, Ubi-GFP, FRT80B/gig, FRT80B

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT80B/FRT80B

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT80B/gig, FRT80B

yw, eyFLP/Y; Ubi-GFP, FRT80B/gig, FRT80B, puc^E69

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT80B/gig, FRT80B, puc^E69

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT40A/FRT40A;puc^E69/+

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT80B/s6k, gig, FRT80B

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT80B/hid, gig, FRT80B

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT80B/dronc, gig, FRT80B

rbf^15aΔ,w, eyFLP/Y;RBF-G3, Ubi-GFP, FRT80B, de2f1^rm729/gig, FRT80B, de2f1ⁱ²

rbf^15aΔ,w, hsFLP/Y;en-Gal4/UAS-Dap;RBF-G3, Ubi-GFP, FRT80B/gig, FRT80B

rbf^15aΔ,w, hsFLP/Y;en-Gal4/UAS-dWee1;RBF-G3, Ubi-GFP, FRT80B/gig, FRT80B

rbf^15aΔ,w, eyFLP/Y;Act >y >Gal4, UAS-GFP/+;FRT82B, RBF-G3, tub-Gal80/FRT82B dtsc1

rbf^15aΔ,w, eyFLP/Y;Act >y >Gal4, UAS-GFP/UAS-Dap;FRT82B, RBF-G3, tub-Gal80/FRT82B

dtsc1

rbf^15aΔ,w, eyFLP/Y;Act >y >Gal4, UAS-GFP/UAS-dWee1;FRT82B, RBF-G3, tub-Gal80/FRT82B

dtsc1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP/FRT82B

yw, eyFLP/Y;FRT82B, Ubi-GFP/FRT82B, dtsc1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP/FRT82B, dtsc1

yw, eyFLP/Y;FRT82B, Ubi-GFP/FRT82B, lkb1

yw, eyFLP/Y;FRT82B, Ubi-GFP/FRT82B, lkb1, dtsc1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP/FRT82B, lkb1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP/FRT82B, lkb1, dtsc1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B

rbf^15aΔ,w, eyFLP/+;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B, dtsc1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B, dtsc1

rbf^15aΔ,w, eyFLP/+;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B, cova

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B, cova

rbf^15aΔ,w, eyFLP/+;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B, cova, dtsc1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B, cova, dtsc1

rbf^15aΔ,w, eyFLP/Y;FRT82B, RBF-G3, Ubi-GFP, rps3/FRT82B, cova, dtsc1

yw,hsFLP/+;UAS-E2f1, UAS-dDp/+; Act >y >Gal4, UAS-GFP/+

yw,hsFLP/+;UAS-E2f1, UAS-dDp/UAS-Rheb; Act >y >Gal4, UAS-GFP/+

yw,hsFLP/+;UAS-E2f1, UAS-dDp/Act >y >Gal4, UAS-GFP;UAS-Rheb/+

yw,hsFLP/+;UAS-E2f1, UAS-dDp/UAS-Dap;Act >y >Gal4, UAS-GFP/UAS-Rheb

yw,hsFLP/+;UAS-E2f1, UAS-dDp/Act >y >Gal4, UAS-GFP;UAS-Rheb/UAS-Puc

yw,hsFLP/+;UAS-E2f1,dDp/UAS-Rheb, UAS-Dap; Act >y >Gal4, UAS-GFP/+

yw,hsFLP;UAS-E2f1, UAS-dDp/UAS-Rheb; Act >y >Gal4, UAS-GFP/UAS-Puc

yw,hsFLP;UAS-Rheb/Act >y >Gal4, UAS-GFP;puc^E69/+

yw,hsFLP;UAS-E2f1, UAS-dDp/+;Act >y >Gal4, UAS-GFP/puc^E69

yw,hsFLP;Act >y >Gal4/UAS-Rheb;puc^E69

yw,hsFLP;UAS-Rheb/Act >y >Gal4, UAS-GFP;UAS-Puc/puc^E69

yw,hsFLP;UAS-E2f1, UAS-dDp/UAS-Rheb;Act >y >Gal4, UAS-GFP/puc^E69

All steps were performed at room temperature. Larval imaginal discs were dissected in 1× PBS, fixed in 4% formaldehyde in 1×PBS for 30 minutes and incubated with primary antibody diluted in 1× PBS plus 0.3% Triton-X100 (PBSTx) with 10% normal goat serum overnight at 4°C. The primary antibodies used were: rabbit anti-activated caspase-3 (C3, 1:300 from Cell Signaling), anti-P-H2Av (Mehrotra and McKim, 2006b) at 1:500; mouse anti-β-galactosidase (1:500, DSHB), anti-BrdU (1:100, DSHB) and anti-P-JNK (1:50 from Cell Signaling). Following incubation with primary antibody, samples were washed three times (10 minutes each) in PBSTx and incubated with secondary antibodies from Jackson ImmunoResearch (1:200 to 1:400). Sample mounting was performed using 70% glycerol with 1,4-diazabicyclo[2.2.2]octane (DABCO) at 12.5 mg ml^–1. The β-galactosidase and BrdU antibodies were obtained from the Developmental Studies Hybridoma Bank (DSHB), developed under the auspices of the NICHD and maintained by The University of Iowa (Department of Biology, Iowa City, IA, USA). Imaging was achieved with the Zeiss Axioscope–ApoTome microscope using the AxioCam CCD camera controlled by Zeiss software.

The percentage of pixels that had an above-background level of Caspase 3 (C3) staining was determined in mutant clones using the histogram function in Adobe Photoshop. The background level of C3 staining was determined as the level that was ≥95% of the C3 signal in adjacent wild-type tissues that had no apoptosis. The levels of γ-H2Av were determined as the number of positive cells per unit clone area, as quantified by Adobe Photoshop.

Incorporation of BrdU (Sigma) was performed at 75 ug ml^–1 for 60 minutes in 1× PBS, followed by standard fixation. After two washes in PBST, samples were treated with 0.02 U µl^–1 DNase I (Promega) for 1 hour at room temperature and then labeled using antibodies against BrdU following standard procedures. The EdU (Invitrogen) incorporation assay was performed according to the manufacturer's instructions; the EdU labeling reaction was achieved at 10 µM for 60 minutes and detected using the AlexaFluor 594 azide. To minimize GFP quenching, the fluorophore reaction was performed for 5 minutes.

G.M.G., T.Z. and W.D. conceived the experiments, analyzed the data and wrote the manuscript. G.M.G. and T.Z. were involved in all aspects of the experiments and data generation in this manuscript. J.Z. contributes to the Facs analysis data generation and manuscript preparation.

The method used to analyze the dissociated wing disc cells using FACS was adapted from de la Cruz and Edgar (de la Cruz and Edgar, 2008). Briefly, at least 20 larvae were used to procure 40 wing discs for each genotype. Discs were cleanly dissected in 1× PBS and then transferred into 4 ml of 5× trypsin-EDTA solution for 2–4 hours, with gentle agitation. Cells were then labeled for 30 minutes before FACS analysis with 1 µl of Vybrant DyeCycle Violet stain per 4 ml of trypsin solution. Dissociated cells were then sorted using the LSR-II Flow Cytometer (BD Biosciences) at the University of Chicago Flow Cytometry Facility. Data were analyzed using FlowJo software version 6.0 (Tree Star).

Relative ATP levels were determined after treating dissected eye imaginal discs with 120 µl 1× passive lysis buffer (Promega) by pipetting 15 times in a 1.5 ml tube on ice, boiling for 5 minutes, then incubating on ice for 2 minutes. After centrifugation at 18,000 g for 2 minutes, 20 µl of each sample was used to assay the ADP:ATP ratio by using the Enzylight kit according to the manufacturer's protocol (BioAssay Systems).

We thank Kim McKim, Duojia Pan, Shelagh Campbell, Utpal Banerjee, Jongkyeong Chung and Bruce Edgar for generously supplying antibodies, fly stocks and reagents. We thank the Bloomington Drosophila Stock Center and the Developmental Studies Hybridoma bank at the University of Iowa for providing fly stocks and antibodies. We also thank Binghui Li for assistance with FACS analysis and members of the Du laboratory, past and present, for helpful discussions.