reaper is required for neuroblast apoptosis during Drosophila development

Apoptosis accompanies many developmental processes, from the earliest stages of embryogenesis, to the final sculpting of the mature organism (Jacobson et al., 1997; Vaux and Korsmeyer, 1999). Most physiologic deaths appear to utilize common core effectors, including the caspase proteases. However, less is known about the upstream pathways that initiate programmed cell death. In Drosophila, many of the genes that influence developmental decisions have been identified and characterized. Therefore, the study of developmental cell death in flies has the potential to provide important information about how developmental regulators interact with the apoptotic machinery (Bangs and White, 2000).

All developmental apoptosis in the Drosophila embryo requires the activity of the reaper (rpr), grim and head involution defective (hid) genes. These genes are localized within a small genomic region (White et al., 1994). Embryos homozygous for a deletion (H99) that removes all three of these genes have virtually no developmental apoptosis, and die as embryos with many extra cells. Each of these genes is sufficient to induce apoptosis in a caspase-dependent manner when overexpressed in insect and mammalian systems (Grether et al., 1995; Hay et al., 1995; Chen et al., 1996; Pronk et al., 1996; White et al., 1996; Evans et al., 1997; Claveria et al., 1998; McCarthy and Dixit, 1998; Haining et al., 1999). At least part of the proapoptotic activity of these genes results from their ability to bind and inactivate the Inhibitor of Apoptosis, or IAP proteins (Vucic et al., 1997; Vucic et al., 1998; Wang et al., 1999; Goyal et al., 2000). IAP proteins can bind to caspases and inhibit their function (Kaiser et al., 1998; Hawkins et al., 1999; Meier et al., 2000). Thus, the combined activity of Rpr, Grim and Hid negatively regulates the amount of IAP-mediated caspase inhibition. The recently identified mammalian SMAC/DIABLO protein acts to potentiate apoptosis through a similar mechanism (Du et al., 2000; Verhagen et al., 2000; Wu et al., 2001). Rpr also binds to the antiapoptotic protein Scythe (Thress et al., 1999). This interaction in turn appears to regulate the molecular chaperone activity of hsp70 (Thress et al., 2001). It is not known if Rpr:Scythe interactions regulate apoptosis in Drosophila.

Although rpr, hid and grim may induce apoptosis through similar mechanisms, it is clear that they are not functionally equivalent at the organismal level. The proteins are differentially expressed in dying cells and in response to different signals. Some doomed cells only express a subset of these genes. In both embryonic and adult central nervous systems, rpr and grim are expressed in a large number of dying cells, while hid is only expressed in the midline glia (White et al., 1994; Grether et al., 1995; Chen et al., 1996; Robinow et al., 1997). In contrast, hid and rpr are expressed in non-neural tissues signaled to die during metamorphosis, while grim is not (Jiang et al., 1997). This differential expression reflects the different pathways that regulate these genes. The expression and activity of hid are negatively regulated by the Ras/MAPK pathway (Bergmann et al., 1998; Kurada and White, 1998), while rpr expression is directly regulated by the p53/DNA damage pathway and the ecdysone receptor (EcR) signaling cascade (Brodsky et al., 2000; Jiang et al., 2000; Ollmann et al., 2000). Many other developmentally important pathways are likely to regulate the tissue-specific expression of these genes.

Are there distinct requirements for these genes in developmental apoptosis? This question is best addressed by the analysis of loss-of-function phenotypes. For example, decreased hid function results in extra cells in the developing eye and at the embryonic midline (Zhou et al., 1997; Kurada and White, 1998). These same cells survive when EGF receptor (EGFR) signaling is increased (Dong and Jacobs, 1997; Scholz et al., 1997; Stemerdink and Jacobs, 1997; Miller and Cagan, 1998; Sawamoto et al., 1998). Loss of hid function inhibits apoptosis resulting from decreased EGFR signaling (Bergmann et al., 1998; Kurada and White, 1998). Taken together these data indicate that EGFR activity, by regulating hid function, regulates cell number in the developing fly.

To dissect the role of rpr in apoptosis, we generated a deletion, which removes rpr but not grim or hid. Both developmentally regulated apoptosis as well as DNA damage-induced apoptosis were analyzed in rpr null animals. The majority of developmental apoptosis was unaffected by the absence of rpr. However, the central nervous systems (CNS) of rpr null adults were grossly enlarged. This neural hyperplasia resulted from the failure of some normal nervous system cell death. Among the cells that failed to undergo apoptosis were the neuroblasts of the abdominal neuromeres. When these cells are rescued from apoptosis, they continue to behave like neuroblasts, giving rise to large numbers of neuronal progeny that are incorporated into the adult nervous system. Although both the Drosophila p53 homolog, p53, and EcR regulate rpr expression, we were surprised to find that rpr played only a limited role in apoptosis induced by these regulators.

Males carrying a ry⁺ marked P element (P(ry⁺)l(3)02069) were X-rayed and progeny scored for loss of the marked P element. Out of 120,000 scored flies, 50 showed loss of the P element as judged by the loss of ry⁺. One candidate, XR38, showed loss of rpr DNA in 27% of embryos as assessed by single embryo PCR (see below). Both grim and hid DNA were present (data not shown). Subsequent in situ hybridization experiments confirmed that grim and hid expression was normal, while rpr was not expressed. GFP-null and GFP-positive embryos (controls) were selected from XR38/TM3,(Kr-GFP) (Casso et al., 1999) parents, and whole-mount in situ hybridizations to detect hid and grim were carried out according to the methods of Grether et al. and Chen et al. (Grether et al., 1995; Chen et al., 1996).

Single embryo PCR was performed as described previously (Franc et al., 1999). Single embryos collected from a cross of XR38/TM6B parents were assayed for the presence of rpr, grim or hid DNA. In each reaction, primers for an irrelevant control sequence (doom) were included to control for DNA quality. rpr-specific primers were: 5′GGCAGTGGCATTCTACATACCCG and 5′CCCGTATTTGTTAGTTACTCGAATCC. hid primers: 5′TCGACGGGCGAGGATGAGCGCGAG and 5′GACTGATGTGGCCATGGATGGCAC. grim primers: 5′CAACAACAGCAGCAGCAGCAGCGAC and 5′CCGTCGGTTGACGCTGGCTCGAACT. doom-specific control primers: 5′AGGGTAAACGGCCACAGAATGT and 5′GATATCGTTGTAGTTGGCCCG. Out of 48 embryos with a doom product, 13 showed no rpr PCR product. No loss of hid or grim DNA was detected. Further mapping by single embryo PCR demonstrated that the XR38 distal breakpoint was at least 20 kb proximal of grim and at least 30 kb distal of rpr.

Acridine Orange (AO) staining (Abrams et al., 1993) was performed on the rpr null embryos from XR38/TM3,(Kr-GFP) (Casso et al., 1999) parents. To test for suppression of grim or hid killing by the rpr mutation, GMR-grim/+; XR38/H99 was compared to GMR-grim/+; +/+ and SM1-GMR-hid/+; XR38/H99 was compared to SM1-GMR-hid/+; +/+. Apoptosis of midline glia was assayed in P(1.0 slit-lacZ); XR38/XR38 embryos according to the method of Zhou et al. (Zhou et al., 1997). X-ray-induced apoptosis was assessed in third instar larval wing discs. XR38/H99 (the Tb⁺ progeny of XR38/TM6B,Tb X H99/TM6B,Tb) and y w control larvae were mock treated or irradiated with 4000 rads and subsequently aged for 4 hours at 25°C. Wing discs were dissected and apoptotic cells were visualized by AO (Brodsky et al., 2000). Steroid hormone-induced death was assessed in salivary glands and larval midgut. XR38/H99 and y w white prepupae were collected and aged for 4 hours at 25°C. Midguts were dissected and the presence of gastric caecae was assessed (Jiang et al., 1997). In both rpr null and wild-type pupae, the caecae were almost completely histolyzed at this time. To assess salivary gland histolysis, animals were dissected at 16 hours post-puparium formation. Both wild-type and rpr mutant salivary glands are undetectable at this time.

Larval stainings were done on the Tb⁺ progeny of XR38/TM6B,Tb X H99/TM6B,Tb, and on y w controls. Larvae were fed continuously on Kankel/White medium (White and Kankel, 1978) containing 0.1 mg/ml of BrdU (Sigma). Central nervous systems from wandering third instar larvae or adults were dissected in calcium-free Ringer’s solution and subsequently fixed for 30 minutes in Carnoy’s fixative (60% ethanol, 30% chloroform, 10% glacial acetic acid), and rehydrated for 2× 20 minutes in PBT (PBS with 0.1% Triton X-100). The nervous systems were then treated with 2 N HCl in PBT for 90 minutes, washed for 2× 10 minutes in PBT and blocked for 2 hours in PBT with 2% BSA and 1% normal goat serum. Mouse anti-BrdU (Roche) was added at 1:50 in PBT with 1% BSA, incubated overnight at 4°C, washed 3× 20 minutes in PBT at room temperature and incubated with FITC-conjugated anti-mouse (Jackson Immunoresearch) 1:200 for 4 hours at room temperature. The tissue was washed 4× 20 minutes in PBT and mounted in Fluoromount-G (Southern Biotech) for confocal microscopy. For double labeling, rat anti-Elav (1:200) (Developmental Studies Hybridoma Bank, University of Iowa) was added with the anti-BrdU, and an anti-rat secondary (Jackson Immunoresearch) was used at 1:200.

Stainings were performed on XR38/H99, NT1B1/+; XR38/NT1B1 H99 (White et al., 1994), hid⁰⁵⁰¹⁴/H99 (Grether et al., 1995), dark^CD4/dark^CD4 (Rodriguez et al., 1999) larvae and y w controls, as in Uv et al. (Uv et al., 1997). Staining was detected by confocal microscopy. Grh-expressing cells were counted in y w, XR38/H99 and NT1B1/+; XR38/NT1B1 H99 larvae. The number of Grh-expressing cells was counted in A2-A8 on each confocal image, with care not to count the same cell twice. An average of 143±8 Grh-expressing neuroblasts could be counted in the abdominal neuromeres of XR38/H99 larvae (n=3), while 117±2 Grh-expressing neuroblasts could be counted in the same region of NT1B1/+; XR38/NT1B1 H99 larvae (n=3). 9.5±3.5 Grh-expressing cells were counted in y w control larvae (n=2).

Central nervous systems from wild-type, XR38/TM6B, H99/TM6B and XR38/H99 pharate or 2- to 6-day old adults were dissected in PEM (100 mM Pipes, 2 mM EGTA, 1 mM MgSO₄). Tissues were fixed for 20 minutes in 3.7% formaldehyde in PEM, rinsed 2× in PEM, dehydrated in a methanol series, and rinsed 5× in PBT. Nervous systems were incubated overnight at room temperature with 1:1000 anti-CCAP (Ewer and Truman, 1996) or 1:200 anti-EcR-A (Talbot et al., 1993) in PBT and 10% normal goat serum, rinsed 4× in PBT and incubated with biotinylated anti-mouse or anti-rabbit secondary (1:200 in PBT, 10% normal goat serum) for a minimum of 2 hours at room temperature. Nervous systems were rinsed 4× in PBT, incubated with ABC (Vector) and developed in the presence of diaminobenzidine.

Canton-S, XR38/TM6B, H99/TM6B or XR38/H99 males were collected within 12 hours of eclosion and aged in individual food vials for 3-5 days. Each male was placed in a courtship chamber with a wet filter paper and an aged Canton S female for observation and videotaping. Courtship indices (CIs) reflect the percentage of time each male directed any courtship step toward his female during a 10 minute observation period. In some cases a male mated before the 10 minute observation period was completed. CIs for these males were based upon percentage of time spent courting until the point of successful copulation.

A specific loss-of-function rpr mutation was essential to dissect the role of rpr in developmental apoptosis. The isolation of such a mutation has proved challenging; previous attempts to use chemical mutagens to create lethal or visible point mutations in the H99 region only resulted in the isolation of hid alleles (White et al., 1994), prompting us to use an alternative strategy. Males carrying a P element located in the non-stop gene, 225 kb proximal to rpr (Fig. 1A) (Berkeley Drosophila Genome Project) were irradiated and candidate genomic deletions were identified. Loss of rpr genomic sequence was assayed by single embryo PCR (Fig. 1B,C). A single rpr deletion, XR38, was isolated. As assessed by in situ hybridization, homozygous XR38 embryos showed no rpr expression, while no quantitative or qualitative changes in grim or hid mRNA expression were detected (Fig. 1D-G). The XR38 deletion is large, removing several genes (Fig. 1A), and XR38 homozygotes are lethal. However, flies of the genotype XR38/H99 are likely to be homozygously deleted for the rpr gene alone, as the proximal breakpoint of H99 lies only 15 kb from rpr (White et al., 1994), and no other predicted genes lie between rpr and this breakpoint (Berkeley Drosophila Genome Project). The distal breakpoint of the XR38 deletion lies between rpr and grim and was found to map more than 30 kb distal to rpr and more than 20 kb proximal to grim. There are no predicted genes between rpr and grim.

XR38/H99 flies are viable, and emerge at the expected frequency. They have a shortened lifespan, but do not show any obvious visible defects. Acridine Orange (AO) staining (Fig. 2A,B) did not reveal detectable decreases in the overall levels of embryonic apoptosis. Apoptosis was also assessed in embryos laid by rpr null (XR38/H99) mothers, to eliminate any potential maternal contribution of Rpr protein or mRNA to embryonic apoptosis. No changes in overall apoptosis were detected in embryos that lacked both maternal and zygotic rpr (data not shown).

To more closely examine embryonic apoptosis, the death of glial cells at the embryonic midline was assessed. The death of these cells can be easily quantitated in vivo using the midline marker P(1.0 slit/lacZ) (Zhou et al., 1995). During embryonic development the number of lacZ-expressing midline glia decreases from 9 per segment (stage 12) to 3 (stage 17) due to apoptosis. Previous work has shown that deletion of hid, grim and rpr (Df(3L)H99) blocks the death of these cells completely (Zhou et al., 1997). Deletion of hid alone doubles the number of surviving cells from 3 to 6 per segment (Zhou et al., 1997), while deletion of hid and grim (Df(3L)X25) increases the number of surviving cells to 8. We found only 3 surviving P(1.0 slit/lacZ)-expressing cells in XR38/XR38 embryos (data not shown). These data support the idea that rpr and grim function cooperatively with hid to induce apoptosis. The killing activity of grim and rpr appears to be approximately additive in the midline glia, while hid activity is more important. In sum, we could detect no essential role for rpr in embryonic apoptosis.

Previous results suggested that rpr might be required for the loss of some larval tissues during metamorphosis. Early in pupal development both the larval salivary gland and much of the larval midgut undergo programmed cell death (Jiang et al., 1997). Four hours after puparium formation the midgut has shrunk in size and no longer contains gastric caecae, while salivary gland histolysis occurs during a 30 minute span 14.5 hours after puparium formation. This process is regulated by pulses of the steroid hormone ecdysone (Baehrecke, 2000). rpr and hid expression immediately precede salivary gland and midgut histolysis. rpr is induced directly by the EcR hormone receptor complex, while hid is induced as a secondary response to ecdysone during salivary gland histolysis (Jiang et al., 1997). Based on these findings, the loss of rpr was predicted to delay or inhibit salivary gland or midgut histolysis. Surprisingly, these deaths were not detectably altered in rpr mutant pupae (data not shown). Thus, other apoptotic regulators must compensate for the loss of rpr in these hormone-regulated deaths.

X-irradiation leads to increased levels of apoptosis in the Drosophila embryo and larva. This apoptosis is preceded by a rapid induction of rpr expression (Nordstrom et al., 1996). In mammalian systems p53 has been shown to be required for apoptosis induced by DNA damaging agents such as X-irradiation (Lowe et al., 1993). Recently a Drosophila homolog of p53 has been identified (p53) (Brodsky et al., 2000; Ollmann et al., 2000). p53 induces apoptosis when ectopically expressed, and a dominant negative p53 inhibits apoptosis induced by X-rays. Importantly, p53 was shown to activate a radiation responsive element in the rpr promoter (Brodsky et al., 2000). This suggests that p53 induces apoptosis in response to DNA damage by activating rpr transcription. Our rpr-specific deletion allowed us to test whether rpr expression is required for p53 and radiation-induced apoptosis.

When p53 is overexpressed in the wild-type eye using an eye-specific driver (GMR-p53), ectopic apoptosis is induced, resulting in small adult eyes (Ollmann et al., 2000) (Fig. 3A). When GMR-p53 was expressed in rpr mutants, the eye size was not modified by the absence of rpr (Fig. 3B). As the phenotype of GMR-p53 might not only result from p53-induced apoptosis, but might reflect additional effects of p53 on eye development, we assayed apoptosis directly in developing eyes, and found that the loss of rpr did not suppress or delay this death (Fig. 3C,D). These data indicate that p53 can induce apoptosis in a rpr-independent manner.

We directly examined the role of rpr in X-ray-induced apoptosis in the developing wing disc. In wild-type larvae, apoptosis is induced in the developing wing within 4 hours of treatment with a high dose of X-rays (Fig. 3F). This death is dependent on the activity of p53 (Brodsky et al., 2000; Ollmann et al., 2000). In the absence of rpr, X-ray-induced apoptosis was significantly inhibited, although some ectopic death is clearly visible (Fig. 3G). Thus, rpr function is required for high levels of apoptosis in response to DNA damage. The residual apoptosis seen in Fig. 3G indicates that p53 can activate apoptosis through other, rpr-independent mechanisms. It is likely that strong overexpression in GMR-p53 obscures a requirement for rpr in p53-induced death in the eye.

Although rpr mutant flies survive to adulthood, the males are completely sterile. Spermatogenesis appeared normal in rpr null males, and large numbers of motile sperm were present in the testes. However, when wild-type females were placed with rpr mutant males for several days, no sperm was transferred to the females. Closer analysis revealed that rpr mutant males showed almost wild-type amounts of overall courtship behaviors, but only 1 male out of 16 mated with the female. Courtship indices (CIs) were calculated for wild-type, XR38 and H99 heterozygous and XR38/H99 males. The CIs represent the mean amount of time males spent courting individual Canton S females during a given observation period. Wild-type males had a courtship index (CI) of 58±10% (n=13), rpr heterozygotes had a CI of 49±9% (n=8) and rpr mutant males had a CI of 47±9% (n=16). These numbers are not statistically different by ANOVA analysis (F_{(2, 0.28)}=0.63, P>0.5). The major block in the mating process appears to be an inability of the males to bend their abdomens sufficiently for copulation.

These defects in adult behavior prompted us to examine the morphology of the adult CNS. Surprisingly, the nervous systems of rpr mutant flies were much larger than those of wild-type flies. The thoracic and abdominal ganglia of the ventral nerve cord (VNC) were enlarged (Fig. 4B,D,F), with the most extensive hyperplasia found in the abdominal ganglion.

To determine if this enlargement was due to decreased apoptosis of larval neurons, we looked at the survival of two subsets of neurons. Adult nervous systems were stained with an antibody directed against the crustacean cardioactive peptide (CCAP). Anti-CCAP labels a set of 34-36 neurons in the VNC (Draizen et al., 1999). Six of these neurons are located in the ventral thoracic portion of the VNC, while the remaining neurons are in the abdominal ganglion. Most of these neurons die within 24 hours of eclosion in the wild type, leaving a population of only 4 CCAP-positive cells in the abdominal ganglion (Draizen et al., 1999). Similarly, we observed an average of 3±0.7 CCAP-expressing neurons (n=5) in the abdominal ganglia of 2- to 6-day old wild-type flies (Fig. 4A,C). In 2- to 6-day old XR38/TM6B flies we found an average of 5.3±0.9 CCAP-positive neurons in the abdominal ganglion (n=4), a number that is not significantly different from wild type. As previously reported, the 6 ventral thoracic CCAP-expressing cells survive in the absence of 1 copy of rpr, grim and hid (H99/TM6B) (Draizen et al., 1999). In XR38/H99 adults, many more of these cells survived, with 26.6±1.2 CCAP-immunoreactive neurons present 2-6 days after eclosion (n=16) (Fig. 4B,D). These cells include the 6 ventral thoracic cells that survive in H99/+, as well as 20 cells in the abdominal ganglion.

To examine the survival of another class of neurons, an antibody to the A isoform of the ecdysone receptor (EcR-A) was used to identify a group of around 300 neurons that normally undergo apoptosis within a day of adult eclosion (Robinow et al., 1993). The death of these cells requires falling levels of ecdysone, and is accompanied by increased transcription of rpr and grim (Robinow et al., 1997). In 1- to 2-day old XR38/TM6B adults 2.8±1.4 (n=4) EcR-A expressing cells were present (Fig. 4E). In H99/TM6B adults 6.9±1.4 (n=9) EcRA positive cells were found, while approximately 65±3.8 (n=5) of these cells persisted in the XR38/H99 adults (Fig. 4F). These data reveal two populations of neurons that fail to undergo normal programmed cell death in XR38/H99 rpr mutants, suggesting that rpr is required for at least some post-eclosion neuronal cell death.

The persistence of neurons that should die during or after metamorphosis likely accounts for some of the increased size of the rpr null CNS. However, the significant enlargement of the abdominal ganglia of the VNC led us to speculate that this increase might additionally reflect the survival and proliferation of neuroblasts. This speculation was based on patterns of neuroblast apoptosis. In the embryo, neural development begins with the specification of about 50 neuroblasts in each neuromere (repeating segment) of the VNC (Goodman and Doe, 1993). Neuroblast stem cell divisions produce the neurons of the larval CNS, and then the neuroblasts become quiescent. In early larval life, almost all neuroblasts in the brain and the thoracic neuromeres of the VNC reinitiate division, giving rise to the neurons of the adult CNS (Truman and Bate, 1988; Prokop and Technau, 1991). However, very few neuroblasts are reactivated in the abdominal neuromeres (A2-A8) of the larval VNC (Truman and Bate, 1988), as the majority of neuroblasts in the abdominal neuromeres undergo apoptosis by the end of embryogenesis in the wild type (Bray et al., 1989; White et al., 1994).

The apoptosis of the abdominal neuroblasts is prevented in H99 homozygous embryos (White et al., 1994). However it is not known whether these rescued cells are competent to function as neuroblasts. Can they proliferate? If so, do the progeny of these divisions differentiate as neurons? The embryonic lethal stage of H99 homozygotes prevented us from assessing whether neuroblasts rescued from death are competent to resume proliferation in the larvae.

Abdominal neuroblast survival in rpr null XR38/H99 larvae was directly assessed by staining for the Grainyhead (Grh) protein (Bray et al., 1989). Grh is expressed in neuroblasts throughout embryogenesis and larval life (Bray et al., 1989; Uv et al., 1997; Prokop et al., 1998). The nervous systems of XR38/H99 larvae contain many cells expressing Grh in the abdominal neuromeres, whereas only a few are present in wild-type nervous systems (Fig. 5A,B).

The death of the abdominal neuroblasts specifically requires rpr function. In larvae that completely lack hid function, we found no ectopic survival of abdominal neuroblasts (data not shown). Even in larvae that lack hid function, and have reduced rpr and grim function (hid⁰⁵⁰¹⁴/H99) (Grether et al., 1995), there was no increase in the survival of these cells (Fig. 5C). Thus, the reduction in grim and hid in XR38/H99 is unlikely to contribute to this phenotype. In addition, we were able to partially rescue the rpr phenotype with a cosmid transgene containing 30 kb of genomic sequence around rpr (NT1B1) (data not shown) (White et al., 1994). The incomplete nature of this rescue may reflect poor rpr expression from this transgene in this tissue.

Interestingly, we also found a few ectopic abdominal neuroblasts in dark mutant larvae (Fig. 5D). dark is a Drosophila homolog of Apaf-1, an important component of the apoptotic effector machinery (Kanuka et al., 1999; Rodriguez et al., 1999; Zhou et al., 1999). Apaf-1 and Dark act to facilitate caspase activation in response to the release of cytochrome C from the mitochondria. Mutations in dark suppress killing by ectopically expressed rpr, suggesting that Dark activates caspases in response to rpr expression. The small number of ectopically surviving neuroblasts in dark mutants demonstrates that dark is essential for a limited amount of rpr-dependent apoptosis in the developing animal.

To determine if these surviving neuroblasts retained their proliferative capacity, larvae were fed on BrdU-containing food. In the larval VNC, virtually the only cycling cells are the neuroblasts and their progeny, the ganglion mother cells (Truman and Bate, 1988). The dividing cells are localized to the ventral surface of the VNC, and the progeny of each neuroblast can be visualized as a cluster of BrdU-positive cells. In wild-type larvae this proliferation is confined mainly to thoracic neuromeres, with the exception of a few divisions in the terminal abdominal neuromere (Truman and Bate, 1988; Taylor and Truman, 1992). In rpr mutants, there is a substantial increase in the number of BrdU-labeled cells in the abdominal neuromeres (Fig. 5E,F). These cells are ventrally localized and clustered, similar to the progeny of wild-type neuroblasts. We conclude that neuroblasts rescued from apoptosis in the rpr mutant remain competent to proliferate.

The progeny of these rescued neuroblasts are responsible for the enlarged abdominal neuromeres in the adult CNS. Cells that have been BrdU labeled during the larval stages are abundant in the abdominal neuromeres of the rpr mutant adult VNC, and much less abundant in the small abdominal neuromeres of the wild type (Fig. 5G,H).

The progeny of these ectopic neuroblast divisions differentiate into neurons. In both wild-type and rpr mutant larval VNCs most BrdU-labeled cells expressed Elav, a pan-neural marker (Robinow and White, 1991) (Fig. 6). These newly born neurons are confined mainly to the thoracic neuromeres of the wild-type VNC, but are present throughout the entire VNC in the rpr mutant. Taken together, these data indicate that rpr function is required to eliminate neuroblasts in the embryo. Neuroblasts that survive this developmental death are competent to divide, giving rise to ectopic neurons in the larval and adult CNS.

The combined activity of the genes rpr, hid and grim is required for the initiation of all apoptotic death in the developing Drosophila embryo. Both genetic and molecular data suggest that these proteins act in an additive manner in the cell to induce apoptosis. However, the activities of these apoptotic regulators are governed by diverse developmental and environmental signals. To assess their individual roles in developmental apoptosis we have generated a mutation that removed only rpr. The phenotypes of this rpr mutant reveal both overlapping and unique functions for this gene. Although rpr is expressed in many cells that are fated to die during development, loss of the gene results in very specific defects. In the absence of rpr, neuroblasts in the developing CNS survive and proliferate, resulting in neural hyperplasia. Substantial numbers of neurons normally fated to die after eclosion survive in these mutants, suggesting that rpr is required for many, but not all, post-eclosion neuronal cell deaths. In contrast, overall levels of apoptosis are not detectably altered in rpr mutant embryos, and the ecdysone-induced death of the larval salivary gland and midgut, as well as some DNA damage-induced death can occur normally.

The apparently normal level of death in rpr null embryos indicates that grim and hid functions are sufficient for the majority of embryonic apoptosis. As grim is expressed in a very similar pattern to rpr, it is likely that the activity of grim is able to compensate for the absence of rpr in many tissues, with the notable exception of the neuroblasts in the abdominal neuromeres.

In contrast to what is seen in rpr mutants, embryos lacking hid function show a slightly decreased level of overall apoptosis (Grether et al., 1995; Zhou et al., 1997). In the absence of hid, rpr and grim activity are not sufficient to initiate apoptosis in a significant number of cells. These results suggest that hid activity is quantitatively or qualitatively different from rpr and grim activity.

Both in vitro binding studies and genetic data support the model that rpr, grim and hid kill by binding to a Drosophila IAP, DIAP1, inactivating its caspase inhibitory activity. In this model, all three proteins are functionally equivalent. The combined activity of the three proteins determines the likelihood that a cell will undergo apoptosis. However there are some findings that suggest functional differences between rpr, hid and grim. Overexpression of hid and rpr together in the midline glia has been shown to induce apoptosis much more potently than equivalent expression of either protein alone (Zhou et al., 1997). In addition, some caspases appear to be more important for killing by rpr than by hid, suggesting that there are molecular distinctions between rpr and hid within the cell (Song et al., 2000). Our preliminary data reveal that deletion of rpr and hid together has a strong synergistic effect on overall levels of embryonic apoptosis (C. P. and K. W., unpublished). This observation is consistent with distinct mechanisms of action of Rpr and Hid.

The steroid hormone ecdysone regulates programmed cell death at metamorphosis and in the adult central nervous system (Baehrecke, 2000). It is interesting to note that rising levels of ecdysone initiate degeneration in the larval midgut and salivary glands, while falling levels of the hormone are required for the death of the type II neurons in the newly eclosed adult. These different responses to ecdysone may be mediated by different isoforms of the receptor, as the doomed larval midgut and salivary gland cells express primarily the B1 isoform of the receptor, while the doomed neurons express the A isoform (Robinow et al., 1993; Talbot et al., 1993). Although these receptor isoforms share both ligand binding and DNA binding domains, they show functional differences (Bender et al., 1997; Schubiger et al., 1998).

Expression of rpr is rapidly induced in the salivary glands after the prepupal pulse of ecdysone. A binding site for the ecdysone receptor complex is present in the rpr promoter, which is essential for rpr expression in the doomed salivary gland (Jiang et al., 2000). The type II neurons also express rpr before they undergo apoptosis (Robinow et al., 1997). Thus rpr is a strong candidate to be important in both of these deaths. We found that salivary gland death was not affected in rpr mutant pupae, while the death of type II neurons was significantly inhibited. This disparity may be explained by the differences in the other genes expressed in these tissues. In the salivary glands, the induction of rpr expression is rapidly followed by hid expression (Jiang et al., 2000). In this tissue, as in the embryo, it is likely that hid activity compensates for the absence of rpr. Expression of the caspase Dronc is also increased in response to ecdysone in these tissues (Dorstyn et al., 1999). High levels of Dronc can induce apoptosis (Quinn et al., 2000) and may contribute to the histolysis of these tissues.

In contrast to the findings in salivary gland and midgut, the ecdysone-regulated death of EcR-A-expressing neurons in the adult nervous system was inhibited in the absence of rpr. These cells express rpr and grim but not hid prior to their death (Robinow et al., 1997). This expression pattern may be a common feature of neuronal tissue, as hid expression is not detectable in the embryonic central nervous system outside of the midline glia (Grether et al., 1995). In the adult nervous system, grim function is apparently not sufficient to induce apoptosis in many of the type II neurons. However, in the embryonic nervous system there is significant apoptosis in the absence of rpr. At this stage grim activity must be sufficient for most neural apoptosis, with the important exception of the death of the neuroblasts.

In flies, as in mammalian tissues, cells undergo apoptosis in response to DNA damage, and this apoptosis requires the activity of the transcription factor p53 (Ko and Prives, 1996; Brodsky et al., 2000; Ollmann et al., 2000). In flies, the expression of a dominant negative form of p53 largely inhibits X-ray-induced apoptosis (Brodsky et al., 2000; Ollmann et al., 2000). Drosophila p53 (p53), can directly bind to a radiation-inducible enhancer in the rpr promoter (Brodsky et al., 2000). These data strongly suggest that p53 induces apoptosis in response to DNA damage by activating rpr expression. Unexpectedly, no suppression of p53-induced apoptosis was detected in rpr null animals. However, X-ray-induced apoptosis is reduced in the absence of rpr. These data indicate that rpr is an important regulator of apoptosis induced by DNA damage, and that other apoptotic regulators are also involved. When p53 is strongly overexpressed in the eye, these other targets must be sufficient to overcome the absence of rpr. The functions of hid and/or grim are doubtless also involved in DNA damage-induced death, as X-ray-induced apoptosis is very strongly inhibited in H99 embryos (White et al., 1994).

Two striking phenotypes are found in rpr mutants: the adult central nervous system in both males and females is enlarged, especially the abdominal part of the ventral nerve cord, and males are sterile. The hyperplasia of the CNS results in part from the abnormal persistence of some larval neurons in the adult ventral ganglia. More importantly, neuroblasts also survive inappropriately in rpr mutants. In the wild-type animal, most of the neuroblasts in the abdominal neuromeres die at the end of embryogenesis, while in the rpr mutant many of these neuroblasts survive and proliferate. The progeny of these ectopic neuroblast divisions differentiate into neurons that are integrated into the adult nervous system. Why are the neuroblasts particularly sensitive to the loss of rpr? One possibility is that rpr is the only apoptosis regulator expressed in these cells. hid is not expressed in the embryonic nervous system (Grether et al., 1995). Although widespread expression of grim is detected in the embryonic CNS, it is not known if neuroblasts express grim. A distinct expression of other apoptotic factors could also account for the specific requirement for rpr in neuroblast apoptosis.

Mutations in the Drosophila Apaf1 homolog dark also result in enlargement of the larval CNS (Kanuka et al., 1999; Rodriguez et al., 1999; Zhou et al., 1999). This increased size results at least in part from the survival and proliferation (data not shown) of a few neuroblasts in the abdominal neuromeres. This implicates dark as being required for some rpr-dependent apoptosis. It is interesting to note that dark mutations, like rpr mutations, cause significant male sterility (Rodriguez et al., 1999).

The sterility of rpr mutant males appears to be behavioral, as they are unable to copulate, although other courtship behaviors appear normal. The cause of the male copulation defect is unknown, but it is interesting to speculate that the reduction in normal cell death in the abdominal neuromeres is in some way responsible for this behavioral deficit. Indeed, the focus of male copulatory behavior has been mapped to the abdominal nervous system by mosaic mapping techniques (Ferveur and Greenspan, 1998). The presence of additional neurons in the nervous system of rpr mutants might interfere with the organization of the appropriate neurons into a functional neural circuit required for copulation. Alternatively, the neural circuit in the CNS might be properly constructed but the presence of additional motorneurons might prevent coordinated movement of the abdomen during copulation.

In C. elegans, the majority of developmental apoptosis occurs in the nervous system (Metzstein et al., 1998). In worms that are mutant for the genes ced-3 or ced-4, and thus lacking all apoptosis, there are extra neurons (Ellis and Horvitz, 1986). However, ectopic cell proliferation has not been reported in these mutant animals. Neural hyperplasia is also seen in mice carrying engineered mutations in caspases 3 and 9 and in the Apaf1 caspase activator (Kuida et al., 1996; Cecconi et al., 1998; Hakem et al., 1998; Kuida et al., 1998; Woo et al., 1998; Yoshida et al., 1998). A detailed analysis of brain development in caspase 3 knockout mice showed a marked increase in proliferating neuroblasts, similar to what is seen in rpr mutants (Pompeiano et al., 2000). These mutants provide a graphic example of how normal development can be particularly disrupted when apoptosis of a stem cell population is inhibited, and these cells continue to proliferate. In the future, the rpr mutant flies will provide a unique model to explore the fate of ectopic neural stem cells and their progeny in the context of the nervous system.

We would like to thank Elvan Tahaoglu for excellent technical assistance with the excision mutagenesis, Jizeng Qiao and Hui Su for help with confocal microscopy, and Nathalie Franc for comments on the manuscript. Important fly stocks, reagents, and protocols were supplied by Mike Brodsky, John Nambu, Steve Robinow, Barbara Wakimoto, John Abrams, David Hogness, Jim Truman, Margrit Schubiger, John Ewer and Sarah Bray. K. W. is supported by NIH grant GM55568 and a grant from the Shiseido Company of Japan to Massachusetts General Hospital/Harvard Medical School. B. T. is supported by NIH grant GM56920 and G. C. by NIH NRSA GM20030.