Coordinated expression of cell death genes regulates neuroblast apoptosis

Programmed cell death is essential for spatial and temporal patterning during development in metazoans, particularly in the central nervous system (CNS), where the majority of cells die during development (Buss et al., 2006). Both mature neurons and dividing progenitors undergo developmentally regulated apoptosis. Mutations in genes important for cell death often result in profound nervous system phenotypes, including increased brain size and defects in morphology (Yuan and Yankner, 2000; De Zio et al., 2005). The appropriate death of proliferating progenitors is particularly crucial to prevent neural hypertrophy (Buss et al., 2006).

We understand relatively little about how particular cells in the CNS are selected to die at a particular time during development. A large number of mature neurons undergo apoptosis in the absence of appropriate targets, and this death is regulated by extrinsic factors, such as the absence of neurotrophins. However, the death of proliferating neural progenitors is likely to be regulated by both extrinsic and intrinsic factors that regulate temporal and spatial identity, as well as interactions with supporting cells (Buss et al., 2006). Examining how neural progenitor cell death is regulated will provide insight into normal neural development. In addition, the viability of these cells might be manipulated to treat disease.

Drosophila neural stem cells or neuroblasts (NBs) provide a well-described, genetically tractable model to examine the regulation of neural stem cell death. In the embryo, NBs are specified and divide to give rise to the neurons and glia of the CNS. At the end of embryogenesis, NBs become quiescent and the vast majority of NBs in the abdominal segments of the developing ventral nerve cord (VNC) undergo apoptosis in late embryogenesis (Fig. 1A, red circle) (Truman and Bate, 1988; White et al., 1994). Of the 30 NBs that are initially present in each hemisegment of the VNC, only three remain in hemisegments A3-A8 at the end of embryogenesis. These occupy characteristic positions and can be identified as the dl, vl and vm NBs in early third instar larvae (Truman and Bate, 1988). During larval life, the dl, vl and vm NBs reinitiate proliferation, giving rise to small neuronal lineages. When NB cell death is inhibited in the embryo, many of the embryonic abdominal NBs survive to become ‘ectopic’ postembryonic NBs (White et al., 1994; Peterson et al., 2002). These continue to divide and produce neuronal progeny, resulting in a massive enlargement of the abdominal segments of the adult CNS (Peterson et al., 2002). In wild-type larvae, the dl, vl and vm abdominal NBs are eliminated by apoptosis at mid third instar (Fig. 1A, purple circle), which limits the number of neuronal progeny they produce (Bello et al., 2003). Our studies focus on the cell death pathways that select NBs to die in the late embryo and in the larvae.

Developmental apoptosis in Drosophila is regulated by the genes head involution defective (hid; Wrinkled, W – FlyBase), grim, reaper (rpr) and possibly sickle (skl), collectively known as the RHG genes (Cashio et al., 2005). These genes are clustered in a 272 kb locus on chromosome III. Developmental apoptosis is completely blocked when hid, grim and rpr are deleted, whereas overexpression of each RHG genes is sufficient to induce caspase-dependent apoptosis (White et al., 1994; Grether et al., 1995; Chen et al., 1996; White et al., 1996; Srinivasula et al., 2002; Wing et al., 2002). A role for skl in developmental apoptosis has yet to be demonstrated. The pro-apoptotic activity of these genes is mediated, at least in part, by the ability to inhibit the anti-apoptotic DIAP1 (Thread – FlyBase) protein (Kornbluth and White, 2005).

Interestingly, the pattern of RHG gene expression largely reflects the pattern of cell death, although hid is expressed both in cells that are fated to die and in cells that survive (Bangs and White, 2000). The dynamic pattern of RHG gene expression must reflect complex transcriptional regulation of these genes. We know some of the upstream regulators of RHG gene expression, such as p53, which activates rpr, skl and hid expression in response to DNA damage, and Ecdysone receptor, which induces transcription of rpr and hid in salivary glands (Brodsky et al., 2000; Jiang et al., 2000; Ollmann et al., 2000; Zhang et al., 2008). The hox protein Deformed directly regulates rpr expression in the developing gnathal segments, and Abd-B regulates rpr in the posterior segmental boundaries (Lohmann et al., 2002). Expression of hid is directly regulated by E2F, modulating sensitivity to DNA damage-induced apoptosis (Moon et al., 2005). Activity of the Ras-MAPK pathway also regulates hid transcription (Kurada and White, 1998). During late embryogenesis, grim, rpr and skl are expressed mainly in the central nervous system whereas hid is expressed in the midline. The similar expression patterns of grim, rpr and skl at this stage of development suggests that these genes might be regulated in a coordinated manner in the embryonic CNS (Bangs and White, 2000; Srinivasula et al., 2002; Wing et al., 2002).

Previously, we identified a mutant background that results in the ectopic survival of many NBs in the abdominal segments of the larval VNC (Peterson et al., 2002). Overlapping deletions that result in removal of the rpr gene, along with ~68 kb of the surrounding genomic sequence, result in a viable animal with a greatly expanded adult VNC. Although the rpr gene is the only known gene in this deletion overlap, we have found that deletion of rpr alone does not result in ectopic postembryonic NB survival. Here, we show that genomic sequences between rpr and grim control expression of rpr, grim and skl in NBs, allowing the coordinated expression of these cell death regulators. We propose that this genomic region serves as an integrator of multiple developmental signals to select this population of cells to undergo apoptosis at the proper time during development.

Wild-type flies were yw^67c23. grim^A6C and grim^C15E strains were kindly provided in advance of publication by Carrie Brachmann (Wu et al., 2010). The small rpr⁸⁷ deletion was generated by imprecise P excision (Moon et al., 2008). The following mutants were generated by PiggyBac P-P deletion (Parks et al., 2004): (1) MM1 removes 32 kb (pp[d11052-f00984]); (2) MM2 removes 110 kb (pp[d11052-f07860]); (3) MM3 removes 54 kb (pp[f04656-f00984]) between rpr and grim. The grim^C15Erpr⁸⁷ (grim rpr) recombinant was generated in a two-step recombination using flanking markers. The skl^H2 allele was generated by excising the hobo element from the hybrid P(wHy)DG39210 element (Huet et al., 2002). Df(3L)ED224 was obtained from DrosDel, deleting 429 kb on the third chromosome, including hid, grim and rpr. Df(3L)ED225, obtained from DrosDel, deletes 435 kb from 5′ of hid to non-stop. ED225 is a hypomorphic allele of hid. X20 deletes from hid through skl, with the proximal breakpoint ~30 kb proximal to skl. UAS-grimRNAi was obtained from the Vienna Drosophila RNAi Center (VDRC) (Dietzl et al., 2007) and ubiquitous expression was driven by act5C-gal4 (Ito et al., 1997).

Fluorescent in situ hybridization (FISH) or RNA-protein double labeling was carried out as described by Lecuyer et al. (Lecuyer et al., 2008) with the following modifications: the riboprobe concentration was 1 ng/μl, hybridization was done at 57°C and embryos were blocked in PBTB for 30 minutes. In double FISH, the embryos were quenched in 3% H₂O₂ for 30 minutes to completely eliminate residual horseradish peroxidase (HRP) between probes. Riboprobes of biotin-labeled grim and digoxigenin (DIG)-labeled rpr or skl were used in double FISH experiments, and DIG-labeled riboprobes were used to quantify single gene expression. GFP-null embryos were selected from MM3/TM3, Ubi-GFP parents, and rabbit anti-GFP antibody (Invitrogen) was used to detect GFP protein expression.

Fluorescent images were obtained with a Leica TCS 4D or a Nikon A1 SiR scanning confocal microscope. Complete z-series of embryos or larval CNS were taken through the tissue, and z-stack projections were used to quantify NBs, which were identified by their large size, round shape and position. Double staining with an antibody to the neuronal marker ELAV (Robinow and White, 1991) [1:50, Developmental Studies Hybridoma Bank (DSHB)] confirms that these big cells are not neurons and are likely to be NBs. For quantification of gene expression in NBs in the embryo, the CNS of each embryo was divided equally into three sections, and NBs in the posterior third of the VNC were identified and counted. For quantification of early L3 NB number, or late L3 cluster number or cluster size, the NBs in the third to fifth neuromeres of the abdominal VNC were counted and analyzed.

Neuroblasts were detected with anti-Deadpan (Dpn) staining (Boone and Doe, 2008). Early-mid third instar larvae were collected based on morphology, dissected in PBS (pH 7.4) and the CNS was fixed with 4% formaldehyde for 30 minutes at room temperature. After washing in PBS, the CNS was permeabilized in PBST (PBS with 0.3% Triton X-100) for 30 minutes then blocked with 1% BSA in PBST for 30 minutes. The tissue was incubated with rat anti-Dpn (1:2 or 1:1 in PBST, a gift from C. Doe, University of Oregon) at 4°C overnight. Incubation with donkey anti-rat Alexa 488 secondary antibody (1:200, Invitrogen) was carried out at room temperature for 1.5 hours. To detect newly born neurons, the CNS from late third instar larvae was stained with anti-Neurotactin (Hortsch et al., 1990) (1:50, DSHB) with donkey anti-mouse secondary (1:400, Invitrogen). Anti-phosphohistone H3 antibody (1:400, Millipore) was used to detect mitotic cells.

Embryos of each strain were collected at indicated stages. Total RNA was extracted using RNAqueous-4PCR kit (Ambion) and treated with DNase I to remove genomic DNA. cDNA was synthesized by reverse transcription of total RNA with an Advantage RT-for-PCR Kit (Clontech). Quantitative real-time PCR (qPCR) was performed using ABI Prism 7000 Sequence Detection System (Applied Biosystems). Relative levels of specific mRNAs were determined using the SYBR Green I Detection Chemistry System (Applied Biosystems). Quantification was performed using the comparative C_T method as described in the manufacturer's procedures manual. RP49 (RpL32 – FlyBase) was used for normalization. The following primers were used: grim, 5′-TCGGAGTTTGGATGCTGGGATCTT-3′ and 5′-AGTCACGTCGTCCTCATCGTTGTT-3′; rpr: 5′-AAAGTCCGGCAAATATCGCAAGCC-3′ and 5′-TGTTGTGGCTCTGTGTCCTTGACT-3′; skl: 5′-ACGACAACTCGCCAAGAGTTCAGA-3′ and 5′-TATCGACTTGATCGCCACTGGGTT-3′; RP49: 5′-CTCATGCAGAACCGCGTTTA-3′ and 5′-ACAAATGTGTATTCCGACCA-3′.

Previously, we showed that ectopic neuroblasts were present in the abdominal neuromeres of Df(3L)XR38/Df(3L)H99 (XR38/H99) third instar larvae (Peterson et al., 2002). This overlapping combination of deficiencies removes a 70 kb region that extends from ~15 kb upstream of rpr to 53 kb downstream, between rpr and grim. Because the only identified gene that is completely deleted in XR38/H99 larvae is rpr (hid, grim and skl are all heterozygous), the data suggested that rpr function is required for NB apoptosis.

To study further the role of rpr in the developmental apoptosis of neuroblasts, we generated a small rpr deletion (rpr⁸⁷) by imprecise P-element excision. This deletion removes ~3 kb, including the entire rpr transcription unit (Moon et al., 2008). Surprisingly, rpr⁸⁷ homozygotes did not show a significant increase in abdominal NBs as detected by staining with anti-Deadpan (Dpn) in early-mid third instar larvae, or in late embryos (Bier et al., 1992; Boone and Doe, 2008) (Fig. 1). We considered the possibility that homozygous deletion of rpr along with loss of one copy of hid, grim and skl might result in ectopic NB survival. However, rpr⁸⁷/Df(3L)X20, which recapitulates the hid^−/+, grim^−/+, rpr^−/−, skl^−/+ genotype (Table 1), did not result in ectopic postembryonic NBs (data not shown).

We also considered the possibility that other cell death genes could be mutated in XR38. We found that the XR38 chromosome carried a point mutation in grim that converts proline 29 to arginine. This proline is conserved in all Drosophila species in the melanogaster group, but not in more distantly related Drosophila species. When expressed in S2 cells pMT-Grim P29A shows slightly reduced killing when compared with wild type (data not shown). We tested whether this grim mutation was significant for NB apoptosis by removing the other copy of grim. Df(3L)X25 deletes hid and grim, and extends ~30 kb proximal to grim (Fig. 2) (White et al., 1994). Larvae of the genotype XR38/ Df(3L)X25 (Table 1) did not contain ectopic abdominal postembryonic NBs (data not shown), indicating that NB cell death occurred normally. These results demonstrated that neither the deletion of rpr, nor the grim P29A mutation in XR38 is solely responsible for the XR38/H99 phenotype. Taken together, these results suggest that deletion of the region around rpr might be responsible for the ectopic abdominal NB phenotype.

To study the role of the genomic region between grim and rpr in NB apoptosis, we generated several genomic deletions (Fig. 2). Using FRT-mediated recombination, we generated a deletion that removes only the intergenic region between rpr and grim, leaving the two transcribed regions intact (MM3). In addition we generated a deletion that extends from grim through rpr (MM2). Deletions around grim were described previously (Wu et al., 2010). Lastly, a small deletion of skl (skl^H2) was generated by transposable element excision.

Early to mid third instar larvae homozygous for the intergenic MM3 deletion were assayed for abdominal NB number, using anti-Dpn staining (Bier et al., 1992; Boone and Doe, 2008). We found a significant number of ectopic abdominal NBs in MM3 homozygous larvae (Fig. 3A), indicating a reduction in embryonic NB cell death (Fig. 1A, red circle). A similar number of ectopic abdominal NBs were found in MM3/XR38 larvae. These overlapping deficiencies result in homozygous deletion of a 22 kb region between rpr and grim.

NBs were also visualized using anti-Neurotactin. Neurotactin marks newly born neurons (Hortsch et al., 1990; Bier et al., 1992; Truman et al., 2004). The neuronal progeny of each NB remain clustered around the NB, and can be seen as distinct clusters (Truman and Bate, 1988; Bello et al., 2003). Thus, visualization of Neurotactin-positive NB clusters can be used as a second method to determine whether NBs persist and continue to proliferate. Ectopic Neurotactin-positive clusters were detected in the abdominal neuromeres in MM3/MM3 and MM3/XR38 larvae (Fig. 3F). This indicates that the 22 kb intergenic region of overlap between MM3 and XR38 must contain a genetic element that is required for abdominal NB apoptosis in the embryo. We call this the NB regulatory region (NBRR).

The NBRR could contain an unidentified transcription unit, either protein coding or a small RNA. Alternatively, this region could contain regulatory elements that control expression of grim or another cell death gene, or multiple cell death genes. No predicted coding genes or non-coding RNAs were identified in this region based on sequence analysis of conserved signatures across the 12 Drosophila genomes (Clark et al., 2007; Lin et al., 2007). In addition, RNA deep sequencing analysis by the ModENCODE project (Celniker et al., 2009), as reported on FlyBase, did not detect any regions of significant transcription. Although we cannot rule out a small RNA transcription unit in the region, we investigated whether a transcriptional enhancer for grim alone, or for multiple cell death genes, could be found within this region.

If MM3 deletes an enhancer that solely regulates grim expression in the NBs, then deletion of grim should reproduce the MM3 NB survival phenotype. Indeed, an increase in abdominal NBs in grim^A6C homozygous larvae was observed, although to a lesser extent than in the MM3/MM3 and MM3/XR38 larvae (Fig. 3A-C). The grim^A6C deletion removes a 25 kb genomic region including grim, CG5103 and CG13700. To exclude the possibility that loss of CG5103 or CG13700 contributed to the NB survival phenotype, grim was ubiquitously knocked down with Act5C-Gal4; UAS-grimRNAi. RNAi knockdown of grim led to a similar increase in abdominal NB number (Fig. 3G), indicating that the grim^A6C phenotype was due to loss of grim. In contrast to grim, deletion of skl or hid did not result in persistent abdominal NBs in early L3 (Fig. 3D,E).

We noted that in late L3 grim mutant larvae, Dpn-expressing NBs were detected in the locations of the normal dl, vl and vm clusters described by Truman and Bate (Truman and Bate, 1988), in addition to some ectopic NBs (data not shown). This suggests that the grim mutant might prolong the lifespan of the dl, vl and vm abdominal NBs that normally die in mid-L3 (purple circle in Fig. 1A). In the absence of cell death at mid-L3, the normal dl, vl and vm abdominal NBs continue to survive and proliferate, resulting in an enlarged cluster of neuroblast progeny (Bello et al., 2003). We investigated whether continued proliferation of abdominal NBs could be detected in grim^A6C mutant larvae using BrdU, anti-phosphohistone H3 (pH3) and Neurotactin staining. BrdU feeding labels proliferating postembryonic NBs and their progeny (Truman and Bate, 1988; Peterson et al., 2002). Additional BrdU staining was observed in grim^A6C mutant larvae. By contrast, no ectopic proliferation of abdominal NBs was found in rpr⁸⁷ mutant larvae (data not shown). Staining with anti-pH3 antibody also demonstrated prolonged proliferation in grim^A6C mutant larvae (Fig. 4A).

Anti-Neurotactin was used to quantify the number of progeny of abdominal NBs. Clusters of Neurotactin-positive cells surround each dividing NB and the number of these cells corresponds well with the NB progeny visualized by BrdU staining [our unpublished data and Bello et al. (Bello et al., 2003)]. Therefore, we examined the number of Neurotactin-positive cells in dorsolateral clusters in grim^A6C mutant larvae (Fig. 4B,C). The dl cluster is the largest of the normal clusters in the wild type, with about six neurons in late third instar larvae. Deletion of grim (grim^A6C/MM2) resulted in a fourfold increase in Neurotactin-positive cells in clusters in this position. We conclude that grim is required for NB death in mid-L3, and that in the absence of grim, NBs continue to divide and produce neuronal progeny. Notably, the deletion of both grim and rpr (MM2/ED225) (Fig. 4B,C) or MARCM clones for H99 (Bello et al., 2003) did not increase cluster size further, suggesting that grim is the sole apoptosis gene required to regulate the normal loss of the dl, vl and vm NBs in mid L3. Cluster size in MM3/MM2 was similar to that observed in the grim^A6C mutant (Fig. 4C). This suggests that the regulatory sequences deleted in MM3 control grim expression in larval NBs.

We considered the possibility that other cell death genes might contribute to abdominal NB loss. It is unlikely that hid is involved, as we can not detect expression of hid in the embryonic CNS outside of the midline (Grether et al., 1995). In addition, hid mutations did not result in ectopic NB survival (Fig. 3E). However, we cannot rule out the contribution of a low level of Hid expression.

We examined the expression of grim, rpr and skl in the embryonic CNS using fluorescent in situ techniques (Lecuyer et al., 2008). Expression of all three genes was almost completely restricted to the CNS from stage 15 (data not shown) (White et al., 1994; Chen et al., 1996; Srinivasula et al., 2002; Wing et al., 2002). Interestingly, grim and skl mRNAs were almost completely colocalized (Fig. 5A). At stage 15, this included co-expression in many large cells in the abdominal segments of the CNS, identified as NBs by their location, size and shape (Fig. 5A). Double labeling with the neuronal marker anti-Elav confirmed that these cells were not neurons (Fig. 5C,D). Surprisingly, grim colocalization with rpr was much less extensive at stage 15 (Fig. 5B). Some NB co-expression of rpr and grim was detected. However, rpr was also expressed in many smaller cells in the CNS, most likely to be neurons. Given that the rpr gene lies between grim and skl, this suggests that regulation of gene expression in this region is quite complex.

To test the idea that the MM3 deletion removes a common transcriptional regulator of multiple cell death genes in embryonic NBs, we examined overall levels of expression of grim, rpr and skl at late stages of embryogenesis in wild-type and MM3 homozygous embryos. As abdominal NB loss occurs at mid-stage 16 (Fig. 1D), we examined whole embryo expression by quantitative RT-PCR (qPCR) at stages 15 and 17. All three genes showed similar temporal patterns of expression, with high expression at stage 15, and a significant drop in expression levels by stage 17 (data not shown). Homozygous MM3 embryos did not show a significant alteration in overall grim, rpr or skl expression at stages 15 and 17. As NBs represent only a small proportion of the cells expressing grim, rpr or skl at this time, the lack of differences in overall expression suggests that any potential regulatory element deleted in MM3 might be specific for a small number of cells, including NBs.

We examined NB expression of grim, rpr and skl in MM3 homozygous embryos using fluorescent in situ analysis. The overall tissue distribution of expression of all three genes was not affected by the deletion of the intergenic region (Fig. 6A,A′,B,B′,C,C′), in late embryogenesis. To specifically examine NB expression, z-stacks were taken through the ventral nerve cord, and the number of NBs expressing each gene was counted (Fig. 6A″,B″,C″). NBs were recognized by size and location.

Analysis of the in situ data showed that the number of abdominal NBs expressing grim and skl was similar (compare Fig. 6A″ with C″). In wild type, NB expression of grim was visible at stage 10, whereas skl expression was observed at stage 11. The number of NBs expressing grim and skl peaked at stage 14. In MM3 homozygotes, the number of grim- and skl-expressing NBs at earlier stages was similar to wild type, but the stage 14 peak was blunted. In very late embryogenesis, the number of abdominal NBs expressing grim and skl in MM3 was similar to that of wild type.

Expression of rpr in abdominal NBs was visible at stage 12. Interestingly, the number of embryonic abdominal NBs expressing rpr was about half of the number expressing grim at all stages (compare Fig. 6A″ with B″). Similar to grim and skl, the peak at stage 14 of rpr-expressing NBs was essentially missing in MM3 homozygous embryos. However, the number of rpr-expressing NBs in MM3 stage 14/15 embryos was ~75% of that seen in wild type, whereas the decrease in grim-expressing NBs at the same stages was ~50%, and skl ~60%. These data indicate that the MM3deletion has a larger effect on grim and skl expression in NBs than it does on rpr expression. We conclude that MM3 deletes a regulatory element that is responsible for the rise in the number of NBs expressing grim, rpr and skl at stage 14, and in the absence of this peak of expression, NBs survive into larval stages.

If grim and rpr act together to regulate NB apoptosis, deletion of the region encompassing both genes should result in a strong NB survival phenotype. In MM2/ED224, we observed a fourfold increase in abdominal NBs in early-mid third instar larvae (Fig. 7). The additional deletion of one copy of skl in MM2/ED225 slightly increased the number of NBs. The morphology of the VNC in these mutants was abnormal, with thickened caudal projections that were not seen in wild type. This genotype was semi-lethal; the flies that emerged lived for only one or two days, suggesting that this greatly enlarged CNS is detrimental. However, defects in other tissues cannot be ruled out as the cause of decreased viability.

MM2 deletes the intergenic region as well as both the grim and rpr coding regions. To confirm that the ectopic NB survival phenotype is due to the loss of both grim and rpr, we generated a recombinant of the grim^C15E and rpr⁸⁷ deletions. Many ectopic abdominal NBs were seen in grim^C15Erpr⁸⁷ recombinant larvae (grim rpr) (Fig. 7). The number of abdominal NBs in grim rpr larvae (10.4 NBs/abdominal hemisegment) was comparable to that seen in MM2/ED224 (11.8 NBs/abdominal hemisegment), but was much greater than that observed in either rpr⁸⁷ or grim^A6C larva (3-4 NBs/abdominal hemisegment). These data suggest that grim and rpr, expressed in the embryonic abdominal NBs, act together to induce the apoptosis of these cells.

If the loss of intergenic sequences in MM3 results in ectopic NB survival due to decreased rpr and grim expression, then MM3/grim rpr should result in ectopic NB survival. Indeed, the number of ectopic NBs in MM3/grim rpr (5 NBs/abdominal hemisegment) was comparable to the number in MM3/MM3 (6.2 NBs/abdominal hemisegment) (compare Fig. 7 with Fig. 3). This demonstrated that loss of the intergenic regulatory sequences on the MM3 chromosome, coupled with loss of grim and rpr on the recombinant chromosome, leads to a phenotype that approximates that of two copies of the intergenic deletion (MM3/MM3). This finding supports our model of gene regulation in NB cell death.

We observed that the larval CNS of grim rpr had an abnormal morphology. Interestingly, homozygous MM3 larvae did not have this abnormal CNS morphology, and were viable at expected frequencies with a normal lifespan. As MM3 embryos still had significant NB expression of grim and rpr (Fig. 4), additional regulatory elements clearly exist for the transcription of these genes and their expression is important for normal CNS development.

When we examined the CNS in one- to two-day-old adults, we saw that the VNC of grim rpr/MM2 was more than twice as long as wild type (Fig. 7C). The abdominal segments extended far into the abdomen, unlike in the wild type where they extended just to the caudal thorax. The thoracic segments of the CNS were also greatly expanded. When we compared the adult CNS from XR38/H99 to XR38/X20, we found that the VNC of XR38/X20 was more extended. As XR38/X20 deletes skl in addition to rpr and the NBRR deleted in XR38/H99, this finding suggests that skl might contribute to cell death in the VNC.

The patterns of developmental cell death are extremely complex and dynamic, and the fate of an individual cell probably represents the integration of multiple signaling pathways. Intrinsic pathways that define the identity and health of a cell, as well as extrinsic growth and survival pathways, are all likely to contribute to the life or death decision. In Drosophila, the transcriptional regulation of the RHG genes is an important output of the pathways that regulate cell death. The clustering of these genes to a single large genomic locus suggests that cell death signals can be integrated by regulatory sequences that control the coordinated expression of these genes.

In this work we show that rpr, grim and skl are co-regulated to eliminate NBs during embryonic development. We identified a genomic region important for this regulation, located between rpr and grim. Identified genes or transcripts are surprisingly lacking in the 93 kb region between these genes, which is highly conserved between Drosophila species (Flybase, 2003). This suggests that crucial cis-regulatory elements might be localized to this region.

We narrowed the genomic region important for NB apoptosis down to 22 kb. The number of NBs expressing grim and skl is significantly decreased in stage 14/15 embryos homozygous for a deletion of this region. The number of NBs expressing rpr is also affected, but to a smaller extent. Interestingly, there is still expression of grim, rpr and skl in NBs from stage 10 through stage 16. However, in the wild type, the number of NBs expressing all three genes increases at stage 14, presaging the loss of these cells by several hours. This increase in NBs expressing rpr, grim and skl is not seen in MM3 homozygous embryos. As overall grim, rpr and skl expression levels are not significantly altered in the mutants, the deleted regulatory sequences must control grim, rpr and skl expression in both a temporal- and tissue-specific manner, and baseline NB expression must be regulated by sequences outside of this region.

It is also interesting to note that the MM3 deletion has a more significant effect on grim and skl expression than on rpr expression. grim and skl are co-expressed in most cells of the stage 15 embryonic CNS, whereas rpr expression is less overlapping. As grim and skl lie 134 kb apart, on either side of rpr, long range enhancer interactions might regulate subsets of genes in this cell death locus. This type of long-range interaction has been implicated previously in the response to irradiation (Zhang et al., 2008). Here again, a single regulatory region regulates multiple cell death genes at a great distance. The radiation responsive element activates hid and rpr, which are 200 kb apart, without having a major effect on the intervening gene grim. It will be important to examine higher order chromatin interactions to understand these long-range enhancer interactions.

The structure of the RHG cell death locus is conserved in other Drosophila species, but not obviously in more distantly related insect species, even though IAP inhibitory proteins are found in other species (Flybase, 2003; Zhou et al., 2005). This suggests that a coordinately regulated RHG gene cluster might be particularly important for Drosophila development.

As demonstrated by these studies, the individual roles for the cell death genes in developmental apoptosis are not well understood. Strong genetic and molecular data support a role for hid in regulating cell death in the embryonic head and in the developing eye (Grether et al., 1995; Bergmann et al., 1998; Kurada and White, 1998). A role for grim in the death of microchaete glial cells has been described recently (Wu et al., 2010). Here, we demonstrate a specific role for grim in the elimination of the normal abdominal neuroblasts in the third instar larvae. As described in this work, rpr alone is not required for abdominal NB death in the embryo. However, our double mutant analysis demonstrates that rpr and grim must act together to eliminate these cells.

A possible role for hid in NB death cannot be eliminated, although we do not detect hid mRNA in the VNC outside of the midline (Grether et al., 1995). A recent study suggests that rpr requires hid for efficient induction of apoptosis (Sandu et al., 2010). However, loss of hid does not result in increased NB survival, as would be expected if hid is required for rpr activity. Furthermore, loss of one copy of hid along with rpr and grim does not increase NB survival (MM2/ED224 versus grim rpr in Fig. 7B). Knockdown of rpr, grim and hid in NBs with an RHG miRNA (Siegrist et al., 2010) also does not increase the number of surviving NBs over loss of rpr and grim (6.2±0.7 NB/abdominal hemisegment in UAS-dcr2; wor-gal4; UAS RHGi versus 10.4±1.8 NBs/abdominal hemisegment in grim rpr). Further studies are needed to examine how hid and grim, rpr and skl differentially contribute to developmental apoptosis.

Although skl has similar pro-apoptotic activity to hid, grim and rpr, a role for skl in developmental apoptosis has not previously been demonstrated, owing to the lack of a skl mutation. Here, we show that skl deletion alone does not alter NB death. In addition, we found that deletion of one copy of skl along with grim and rpr slightly increase NB survival over deletion of grim and rpr alone (MM2/ED225 versus MM2/ED224). Deletion of skl and rpr, along with the NBRR in XR38/X20 also results in a larger VNC than deletion of rpr and the NBRR in XR38/H88, again suggesting a role for skl in cell death in this tissue. Additional double mutant analysis will be needed to test whether deletion of skl along with grim or rpr supports a role for skl in developmental apoptosis. Given the strongly overlapping expression of grim and skl in the embryonic VNC, we might find that skl is redundant with grim in the CNS.

What are the developmental advantages of multiple cell death regulators? This work demonstrates that multiple RHG genes are activated within a single NB to activate apoptosis. In the salivary gland, hid and rpr, but not grim, are transcribed in the dying tissue, whereas the radiation response appears to involve hid, rpr and skl (Jiang et al., 1997; Zhang et al., 2008). These data suggest that the complex upstream regulation of cell death is likely to impact on different subsets of the RHG genes. If each gene is controlled by different combinations of enhancers this could provide greater flexibility in controlling the activation of cell death. In addition, a requirement for multiple cell death activators could ensure that cells are not killed inappropriately by the accidental activation of a single RHG gene.

The requirement for multiple RHG genes to kill cells might also reflect differences in the pro-apoptotic activities of these genes. For example, the RHG proteins show differential binding preferences to specific domains of the DIAP1 protein, and inhibit DIAP1 anti-apoptotic functions through different mechanisms (Wu et al., 2001; Yoo et al., 2002; Zachariou et al., 2003; Yokokura et al., 2004). Additional activities, such as mitochondrial permeabilization, translational repression or interaction with other proteins such as Bruce or Scythe, also differ between RHG genes (Thress et al., 1998; Vernooy et al., 2002; Thomenius and Kornbluth, 2006; Abdelwahid et al., 2007; Sandu et al., 2010). Furthermore, post-translational regulation of particular RHG proteins might regulate their activity. This has been demonstrated for Hid; phosphorylation by MAPK downregulates the killing activity of Hid (Bergmann et al., 1998). It remains to be demonstrated whether different cell types have differential sensitivities to particular combinations of cell death genes when expressed at normal levels. This might provide yet another layer of control of the cell death process.

In sum, this work demonstrates that the regulation of NB apoptosis during development involves the coordinated temporal and spatial regulation of multiple cell death genes, controlled by a common regulatory region. By dissecting the RHG gene locus, specific genomic elements that regulate these genes in other developmentally important cell deaths are likely to be identified.

Thanks to Chris Doe for antibodies and advice, Ginger Carney and Barbara Taylor for important observations early in the project and the Bloomington, VDRC and Harvard Exelixis stock centers for fly stocks. Reena Patel provided technical assistance on the project at early stages. The work was funded by NIH grant GM55568 (K.W.), a Career Award in the Biomedical Sciences from the Burroughs Wellcome Fund (C.B.), and an MGH Fund for Medical Discovery award (Y.T.). Deposited in PMC for release after 12 months.