The idea that cell competition removes aneuploid cells was suggested by studies of p53-independent cell death following chromosome breakage or ionizing irradiation (Titen and Golic, 2008; McNamee and Brodsky, 2009). Importantly, in Drosophila cell competition of Rp^+/- cells does not depend on p53 (Kale et al., 2015). If the model was correct, it would be expected that the p53-independent apoptosis that follows irradiation would depend on the genes recently discovered to be required for cell competition.

We first confirmed the previous findings (Wichmann et al., 2006; McNamee and Brodsky, 2009). Irradiating third-instar larvae resulted in rapid induction of cell death in the wing imaginal disc that was largely p53-dependent (Figure 1A,C,D,E,G,H). The p53-dependent cell death, attributable to the DNA-damage response, was not much affected by the rpS12^G97D mutation that interferes with cell competition (Figure 1B,F,H). While total cell death tailed off with time, p53-independent cell death increased around 18–24 hr post-irradiation, as reported previously (Wichmann et al., 2006; McNamee and Brodsky, 2009; Figure 1I–L). As expected for cell competition, 24 hr after irradiation, p53-independent cell death was reduced by 66% in the rpS12^G97D p53 double mutant compared to the p53 mutant alone (Figure 1K–M). To exclude the possibility that other genetic background differences were responsible, rpS12 function was restored to the rpS12^G97D p53 double mutant strain using a P element transgene encoding the wild-type rpS12 gene, and this restored p53-independent cell death (Figure 1M). Notably, a genomic transgene encoding the rpS12^G97D cell competition-defective allele did not, leading to 86% less p53-independent cell death than the wild-type rpS12 transgene (Figure 1M). Thus, 66–86% of the p53-independent apoptosis was RpS12-dependent and might represent cell competition.

Nuclear Xrp1 protein, which is only at low levels in control wing imaginal discs, was detected in scattered cells throughout p53 mutant wing discs 24 hr after irradiation, similar to the distribution of dying cells (Figure 1N). Strikingly, most (58%) of this Xrp1 expression was rpS12-dependent, similar to the rpS12-dependency of p53-independent cell death itself (Figure 1O–P). This is also as expected for cell competition, which is mediated through the induction of Xrp1 expression (Baillon et al., 2018; Lee et al., 2018; Ji et al., 2019).

Radiation-damaged cells that have reduced Rp gene dose would be expected to differentiate short, thin bristles in adults, as reported previously in studies of DNA repair mutants (Baker et al., 1978), in studies of ionizing radiation (McNamee and Brodsky, 2009), and when aneuploidy is induced by mutation of spindle assembly checkpoint genes (Dekanty et al., 2012). We found Rp^+/--like thoracic bristles at a frequency of ~1/300 following irradiation of either wild type or p53 mutant larvae (Figure 1Q,R). Their frequency increased ~three- to fourfold in rpS12^G97D mutants or p53 rpS12^G97D double mutants (Figure 1Q). In the absence of irradiation, only 2 Rp^+/--like bristles were observed from 500 unirradiated rpS12^G97D mutant flies. Since 16 macrochaetae were examined on each fly thorax, this indicated a frequency of ~1/8000 macrochaetae progenitor cells was Rp^+/--like. We found none in 1000 unirradiated wild type flies (16,000 macrochaetae examined).

While the actual nature of radiation-induced genetic changes in cells forming Minute-like bristles is not directly demonstrated, previous studies of DNA-repair mutants including mei-41, the Drosophila ATR homolog, demonstrated using multiply-marked chromosomes that the majority of Minute-like bristles reflect loss of heterozygosity for large, contiguous chromosome regions (Baker et al., 1978). In a small-scale experiment to compare γ-irradiation to the DNA repair defects studied previously (Baker et al., 1978), y^+/- rpS12^G97D larvae were irradiated (1000 Rad) and 2178 adult flies examined for phenotypically y thoracic bristles representing cells where the y⁺ allele had been mutated or deleted. Six times more y bristles were recovered in females than in males. Since the y locus is X-linked, this is most easily explained if y bristles generally result from deletions including essential genes linked to y that could not survive in males. In females, 37.5% of y bristles were also phenotypically Minute, consistent with loss of chromosome regions extending at least from the y locus to the RpL36 gene 0.3 Mb more centromere-proximal that is the nearest Rp locus (Figure 1S). Although this study was small scale, these findings support the conclusion from DNA repair mutant studies that Minute-like bristles seen following irradiation most commonly reflect loss of substantial chromosome segments including Rp genes, (Baker et al., 1978). Accordingly, many Minute-like bristles removed by cell competition genes following irradiation could represent such segmentally aneuploid cells.

Taken together, these bristle results are also consistent with the notion that cell competition removes ~3/4 of the cells with genetic changes that encompass dose-sensitive Rp loci that arise after irradiation.

Having confirmed that cell competition could potentially be important for removing cells following irradiation, we sought to assess the fate of sporadic cells that lose chromosome regions, using an assay where the cell genotypes would be definitively known and the dependence on competition with normal cells could be established. We used the FLP-FRT site-specific recombination system (Golic and Lindquist, 1989) to achieve this, exploiting large collections of transgenic flies that contain FRT sequence insertions at distinct chromosomal locations (Thibault et al., 2004). FLP recombination between pairs of FRT elements linked in cis excises intervening sequences to make defined deletions with a single FRT remaining at the recombination site (Figure 2A,B). Insertion elements of the Exelixis collection exist in several configurations, and FLP-mediated excision from paired FRT strains in the FRT w⁺ … w⁺ FRT configuration removes both the associated w⁺ genes, so that affected cells can be identified in the adult eye by loss of pigmentation (Figure 2A,B). Accordingly, we assembled a collection of genetic strains containing linked pairs of appropriate FRT w⁺ elements, each flanking a distinct genomic region (Supplementary file 1).

It was first necessary to verify FLP recombination between FRT sequences, and investigate any cell-autonomous effects of the resulting segmental-monosomies. We used the eyFlp transgene, which confers continuous FLP expression to the eye and head primordia during larval life, so that FLP-FRT recombination is expected to approach completion (Newsome et al., 2000). Excision should result in white adult eyes, and also reveal any cell-autonomous effect of the resulting heterozygous deletion genotype on growth or differentiation of cells contributing to the adult eye. If the recombined genotype was autonomously cell-lethal, we would expect the developing animal to lack head structures and be unable to emerge from the pupa. If FLP-FRT recombination did not occur (or occurred inefficiently), we would expect adult eyes expressing the parental eye color (or with only scattered white spots that recombine late in development as cell number increases).

We identified 17 paired FRT strains that were efficient FLP targets in this assay. These 17 strains were completely or substantially white-eyed in the presence of eyFlp, indicating excision between FRT sites in most or all eye cells. We also identified paired FRT strains that were poor substrates for Flp (Supplementary file 1; Figure 2C–T). These either retained the parental eye color in the presence of eyFlp or produce a salt and pepper pattern of very small clones (Supplementary file 1; Figure 2U–Z). No genotype tested was inviable in the presence of eyFlp, so there was no evidence that haploinsufficiency of any of the chromosome segments tested was incompatible with cell viability or severely impacted head development. Instead, most of the 17 genotypes that recombined differentiated heads of remarkably normal external appearance and morphology (Figure 2C–T).

Although we did not measure head size, we noticed three genotypes in which eyes and heads were obviously smaller, consistent with a reduced growth rate of the recombined genotypes. The three small eye regions were 48B2-50C1, 565F16-59B1, and 87B8-89E5 (Figure 2E,H,Q). Since the Drosophila genome is divided cytologically into 102 band intervals, each with lettered and numbered subdivisions, in this paper we refer to the chromosome P{XP}^d09761 pBAC{WH}^f00157, for example, by the cytological locations of the FRT sequences present in the P element and PiggyBac element insertions, which are at 48B2 and 50C1 respectively in this case (Supplementary file 1). This nomenclature quickly communicates the genome location under study, whether it overlaps or is distinct from that affected in another strain, and also indicates that in this case the FRT elements are likely separated by ~2% of the genome. The full description of each insertion strain is given in Supplementary file 1. The three excisions that substantially reduce eye size could delete a copy of one or more haploinsufficient genes important for growth during eye development, but there also could be a dominant effect of the novel junction generated by FLP/FRT recombination. Notably, excision between chromosome bands 87B8-93A2 led to eyes of normal size, although this excises all the sequences between 87B8-89E5 which led to reduced eyes. FLP recombination results in a different junctions in 87B8-89E5 and 87B8-93A2, however (Figure 2R).

Overall, these results showed that 17 segmentally aneuploid genotypes were cell-viable, and able to grow and differentiate in the Drosophila eye, although a minority might have an effect on growth in this tissue. Recombination in these 17 strains each deleted 1.4 Mb – 8.5 Mb of autosomal DNA, representing 1–6% of the sequenced genome each (Supplementary file 1, Figure 3A). Together these deletions encompass 25.3 Mb of DNA, corresponding to 21.1% of the Drosophila euchromatin and 17.7% of the sequenced genome. 11 of these 17 genotypes deleted one or more Rp loci (RpS11, RpS13, RpS16, RpS20, RpS24, RpS30, RpL14, RpL18, RpL28 or RpL36A), whereas six affected no Rp gene (Supplementary file 1, Figure 3A). In this paper, we use the symbol Rp to indicate a mutation affecting any of the 66 ribosomal protein genes that are dominant through haploinsufficiency, in distinction to 13 Rp encoded by Drosophila loci where heterozygous mutations have no phenotype.

The 17 FRT pair strains that were efficient FLP targets were each exposed to a single burst of FLP expression using the heat-shock FLP transgene, intended to stimulate excision in a fraction of cells during early larval life (see Materials and methods). This led to mosaic eyes where segmentally aneuploid cells and diploid cells would be in competition. Clones of excised cells appeared only after heat-shock, confirming strict FLP-dependence.

In contrast to eyFlp recombination, mosaic eyes containing sporadic clones of excised cells were only recovered at high frequencies for four segmental aneuploid genotypes, none of which affected Rp loci (Figure 3). This confirmed that most segmentally aneuploid genotypes were selected against in mosaic eyes where diploid cells were also present, because all had been shown to be intrinsically viable when competing diploid cells were absent (Figure 2). Importantly, no deletion that included an Rp locus showed more than minimal survival of sporadic clones induced with hsFlp, suggesting Rp loci could be the determinants of cell competition between segmental aneuploid and wild type cells (Figure 3B–D). To test this in a specific case, 5.6 kb of genomic DNA encompassing the RpL28 locus was introduced onto the second chromosome using PhiC31-mediated transgenesis. This transgene proved completely sufficient to rescue the survival of cell clones heterozygous for Df(3L)63A3-65A9, a deletion of 3.2 Mb including the RpL28 locus (Figure 4A–C). Clones of Df(3L)63A3-65A9 heterozygous cells barely survived alone, with a median contribution of 2% to the eyes of males and 0% to females (Figure 4H). In the presence of the RpL28⁺ transgene, however, Df(3L)63A3-65A9/+ clones survived in 49 out of 50 eyes, with median contributions of 37% of the eye in males and 53% in females, not statistically different from clones heterozygous for Df(3L)63C1-65A9, an overlapping deletion of 3.0 Mb excluding the RpL28 locus (Figure 4H). Thus in this case, RpL28 gene dose alone determined whether a segmentally aneuploid genotype affecting hundreds of genes would be eliminated in competition with diploid cells.

Most of the other 10 segmental aneuploid genotypes that deleted one or more Rp loci contributed to adult eyes to a very significantly lower degree that overlapping segmental aneuploidies that spared Rp loci (Figure 3). This reflected the fact that, in contrast to genotypes affecting Rp genes, 4/6 segmental aneuploidies that spared Rp loci survived in eye clones at high frequencies and large sizes (Figure 3A–C). The two exceptions were Df(3R)87B8-89B16/+ and Df(3R)87B8-89E5/+, for which little eye tissue was recovered (Figure 3A–C). Although neither affected any Rp gene, both deleted a locus mapping to 88E5-6 encoding the translation factor eIF2γ. Since independent studies in our laboratory already identified a role for the eIF2α protein in cell competition (Kiparaki, Khan, Chuen and Baker, in preparation), we tested the possible role of eIF2γ by restoring eIF2γ diploidy to Df87B8-89B16/+ or Df87B8-89E5/+ cells using a 11.5 kb genomic transgene including the eIF2γ locus (Tschiersch et al., 1994). This completely rescued the growth and differentiation of these cells to the levels typical for aneuploidies not affecting Rp genes (Figure 4D–G,I). Thus, the locus encoding the translation factor eIF2γ behaved similarly to an Rp gene in triggering competitive elimination of heterozygous cells.

If these studies, which tested a significant fraction of the Drosophila genome, are representative, they indicate that the normal diploid complement of Rp loci is important for sporadic segmentally aneuploid cells to evade cell competition, and that few other genes are comparably important. The one example of such another gene uncovered in our analysis encoded eIF2γ, another protein affecting translation.

If the segmentally aneuploid cells were competed by virtue of their Rp^+/- genotypes, the genetic pathways should be similar. Elimination of Rp^+/- point-mutant cells by competition depends on apoptosis and is suppressed by a genetic deletion, Df(3L)H99, that removes three pro-apoptotic genes reaper (rpr), grim, and head-involution defective (hid)(Moreno et al., 2002; Kale et al., 2015). These genes are also required for the p53-independent cell death that follows irradiation, much of which resembles cell competition (Figure 1; McNamee and Brodsky, 2009). Using the Df(3L)63A3-65A9, where clone loss was demonstrably due to heterozygosity for the RpL28 gene (Figure 4H), we found that recovery of Df(3L)63A3-65A9/+ cell clones was enhanced by genetic suppression of apoptosis in the Df(3L)H99/+ background that experiences loss of heterozygosity for rpr, grim, and hid (Figure 5A,D). A similar rescue was obtained with the Df(3L)63C1-65F5, which deletes the RpL18 locus (Figure 5A,D). The recoveries of Df(3L)63A3-65A9/Df(3L)H99 clones and Df(3L)63C1-65F5/Df(3L)H99 clones approached that of the overlapping genotype Df(3L)63C1-65A5/Df(3L)H99, in which no Rp genes were affected (Figure 5A,D). Recovery was quantitatively inferior to that seen for Df(3L)63A3-65A9/+ p{RpL28⁺} clones (Figure 4H), but it is to be noted that the Df(3L)H99/+ background unexpectedly reduced recovery of the control Df(3L)63C1-65A5/+ cells (Figure 5A). Regardless of whether this reflects the recently described role for basal caspase activity in promoting imaginal disc growth in the wild type (Shinoda et al., 2019), or some other genetic interaction, it complicates assessment of whether H99 heterozygosity and RpL28⁺ transgenesis rescue Df(3L)63A3-65A9/+ clones equally. We attempted to prevent apoptosis more completely using the genetic background hid^WRX1/ Df(3L)H99 in which hid is homozygously affected in addition to heterozygosity for rpr and grim. Although hid^WRX1/ Df(3L)H99 adult animals were recovered in the absence of other mutations, they became exceptionally rare in heat-shocked combinations with the dual FRT chromosomes: as a result, insufficient data could be obtained to address this question. In any case, it is clear from our results that pro-apoptotic genes contribute significantly to eliminating segmentally aneuploid cells (Figure 5A).

Mutations in the rpS12 and Xrp1 genes are more specific for cell competition than mutations in cell death genes. The rpS12 and Xrp1 mutations prevent the elimination of Rp^+/- point mutant cells from mosaics, but otherwise lead to seemingly normal flies, and do not affect other cell death processes (Kale et al., 2018; Lee et al., 2018). These genes should be required if segmental aneuploid cells are competed due to reduced Rp gene dose. Our results strongly support this conclusion in nearly all cases. Because the results are too extensive to present together in a single figure, they are presented in groups according to chromosome region (Figure 5, Figure 6, Figure 7, Figure 8).

Beginning with the Df(3L)63A3-65A9/+ genotype where clone loss was demonstrably due to heterozygosity for the RpL28 gene (Figure 4H), we found that heterozygosity for an Xrp1 mutation greatly restored contribution of Df(3L)63A3-65A9/+ clones (RpL28^+/-) to the eye (Figure 5B,D). Similar results were seen for the Df(3L)63C1-65F5/+ genotype that is heterozygous for RpL18, but Xrp1 heterozgosity did not affect recovery of clones of the overlapping Df(3L)63C1-65A9/+ that is Rp^+/+ (Figure 5B,D). The Xrp1 mutation even improved the survival of larger segmental-aneuploidies where combinations of the RpL18, RpL28, and RpL14 genes were affected (Figure 6A–C).

Because the rpS12^G97D mutation that affects cell competition recessively maps to the third chromosome, it was simpler to examine in combination with segmental aneuploidies affecting chromosome 2. These aneuploid eye clones, which like those discussed above were also recovered in the presence of an Xrp1 mutation, included Df(2L)26A1-29F8/+, heterozygous for RpL36A and RpS13, and Df(2R)56F16-59B1/+, heterozygous for RpS16 and RpS24 (Figure 7A,B,D). Clones of the Df(2L)26A1-28C3/+ or Df(2R)56F16-58E2/+ cells that did not affect any Rp loci were recovered at high rates, independently of Xrp1 genotype (Figure 7A,B,D). As expected, rpS12^G97D homozygosity also led to significant recovery of Df(2R)56F16-59B1/+ clones that were heterozygous for RpS16 and RpS24, although to a quantitatively lesser degree than Xrp1 (Figure 7C,D) The contribution of Df(2R)56F16-58E2/+ cells, where no Rp gene is affected, was unaltered by the rpS12^G97D mutation (Figure 7C,D).

Xrp1 also affected other segmentally aneuploid regions. Xrp1 mutations had a minor but statistically significant effect on clones of Df(2R)48B2-50C1/+ cells, heterozygous for RpS11 (Figure 8A,D), a genotype that also had an autonomous effect on eye growth (Figure 2E). Xrp1 mutations enhanced the contributions of Df(3R)87B8-89B16/+ and Df(3R)87B8-89E5/+ cells that were heterozygous for the eIF2γ gene (Figure 8B–D). Crossing to Xrp1 only slightly improved survival of Df(3R)87B8-93A2/+ clones that were heterozygous for the eIF2γ, RpS20 and RpS30 genes, and did little to enhance recovery of Df(3R)89B13-93A2/+ genotypes that were heterozygous for RpS20 and RpS30 alone. As these deficiencies already delete the Xrp1 locus itself (within 91D3-5 region), introducing an Xrp1 mutation in trans leads to Xrp1^-/- genotypes (Figure 8B–D). Heterozygous mutation of Xrp1 is already sufficient to suppress competition of Rp^+/- point mutant cells (Lee et al., 2018), probably explaining why Xrp1 homozygosity had little further effect.

Further evidence that segmentally aneuploid cells are eliminated by cell competition due to their Rp/+ genotypes came from studies in homotypic Rp mutant backgrounds (Figure 9). It is known from previous work that cells heterozygous at two Rp loci do not suffer more severe competition than cells heterozygous for only one Rp mutation, and therefore that cells heterozygous for two Rp loci generally cannot be eliminated by cells heterozygous at only one Rp locus (Simpson and Morata, 1981). Accordingly, clones of segmental aneuploid cells affecting the RpS11, RpS13, RpS16, RpS20, RpS24, RpS30, RpL14, RpL18, RpL28, or RpL36A genes were all recovered significantly better in an RpS3 point mutant background, ie RpS3^+/- Df(Rp)/+ clones were not eliminated from RpS3^+/- tissues (Figure 9A,B,D). This applied to segmental aneuploid clones heterozygous for eIF2γ as well (Figure 9C,D). On the other hand, the RpS3 point mutant background usually had no effect on the survival of clones of genotypes that did not delete other Rp loci (Figure 9C). One exception was Df(2L)26A1-28C3/+, for which the RpS3^+/- background generated significantly larger clones (Figure 9C,D). Notably, Df(2L)26A1-28C3/+ cells had previously been recovered less than the other non-Rp segmental aneuploidies (Figure 3B,C; Figure 9C). Although this could also have reflected a lower rate of FLP-recombination between the 26A1 and 28C3 FRT sites, in the RpS3^+/- background the recovery of Df2(2L)6A1-28C3/+ clones was similar to that of Df(2R)48F6-50C1/+, Df(2R)56F16-58E2/+ or Df(3L)63C1-65A9/+, suggesting instead that Df(2L)26A1-28C3/+ might be subject to a mild cell competition that can be rescued in the RpS3^+/- background (Figure 9C).

Although suppression of apoptosis, mutation of cell competition genes, or a germline-inherited RpS3 background all restored the growth and survival of cells hemizygous for Rp loci, they may not have done so equally. Suppressing apoptosis was least effective at expanding the contribution of aneuploid cells in the rescued eyes (Figure 5A). Xrp1, rpS12^G97D, and the RpS3^+/- background suppressed cell competition to similar extents, although the general RpS3 background often had the greatest effect, comparable to those of RpL28⁺ and eIF2γ⁺ transgenes (see statistical comparisons for the 63A3-65F5 and 56F16-59B1 regions in Figures 5 and 6 legends). Although several explanations could justify these differences, it is worth noting that the results correlate with the effects of these genetic backgrounds on translation and growth. Thus Df(3L)H99, which suppresses apoptosis with no known increase translation or cellular growth, is expected to suppress the competition of Rp^+/- cells but not restore their translation. As a consequence, clones of Rp^+/- Df(H99)/+ cells, although surviving, are not expected to grow as rapidly or contribute as much to the eye as clones of Rp^+/+ cells. By contrast the rpS12^G97D and Xrp1 mutations restore the general translation rate of Rp^+/- cells, with Xrp1 mutation also restoring more normal rates of cellular and organismal growth (Lee et al., 2018; Ji et al., 2019). A background mutation in RpS3 is not expected to restore translation or growth to segmentally aneuploid cells, but by equally impairing the unrecombined cells, and systemically delaying the growth and developmental rate of the organism as a whole, it equalizes the contributions of aneuploid and control genotypes.

In summary, our results strongly support the conclusion that the growth and survival of most segmentally aneuploid regions is determined by cell competition according to Rp gene copy number, and show that the RpS12/Xrp1-dependent process that eliminates Rp^+/- point mutated cells also acts on cells with large losses of genetic material that include Rp genes.

Experiments using the hsFLP transgene should stimulate recombination and segmental aneuploidy in all tissues, not only in the eye where excision causes loss of pigmentation. To test this, we looked for cells with deletions encompassing Rp loci in the thorax, where Rp haploinsufficiency leads to small, thin thoracic bristles (Marygold et al., 2007). This was explored using Df(2R)56F16-59B1, which deletes the RpS16 and RpS24 loci. Minute-like bristles were not observed on the thoraces of heat-shocked flies carrying Df(2R)56F16-58E2 heterozygous clones, which affect no Rp locus, or on the thoraces of heat-shocked 56F16-59B1 flies lacking rpS12 or Xrp1 mutations, but they represented 0.5% of the thoracic bristles in the 56F16-59B1/+ rpS12^G97D flies and 0.25% of the thoracic bristles in the 56F16-59B1/+ Xrp1^m2-73/+ background (Figure 10A,B). These findings indicate that Df(2R)56F16-59B1/+ cells also survive to adulthood in the thorax if cell competition is suppressed, albeit at lower frequency than observed in the eye.

Flies were reared on standard medium at 25°C unless otherwise noted. The genetic strains used are described in the Key Resources Table.

Strains carrying pairs of FRT transgenic elements in cis were obtained after meiotic recombination using appropriate genetic crosses, monitoring recombination frequency to confirm the expected transgene locations. FLP expression was induced by 37°C heat shock for 30 or 60 min at 36 ± 12 hr after egg laying. Adult flies were aged ~1 week to allow eye color to darken fully, then stored at −20°C for later measurement and photography. The fraction of each adult eye populated by unpigmented cells was estimated manually under a dissecting microscope. Samples were blinded for genotype before scoring by an independent investigator. We estimate the clonal composition of the eye by conceptually dividing each eye into segments so as to focus on the composition of the mosaic subregions. For example, an eye that is 56% white might be half white with an additional quarter of the eye that was one quarter white. Estimates are no doubt approximate although we do not think the errors are large. Importantly, the Mann Whitney procedure used to compare results statistically ranks relative clone size between genotypes rather than using the absolute values of the estimates.

Many of the genetic backgrounds in which mosaics were generated carried other, distant FRT sites as part of the FRT82B Xrp1^m2-73, FRT82B RpS3, rpS12^G97D FRT80B, Df(3L)H99 FRT80B chromosomes. Accordingly, the control backgrounds in these experiments always included FRT82B or FRT80B, as appropriate, and as described in the figure legends.

The RpL28 rescue transgene was obtained by inserting genomic sequences 3L: 3220152–3225729 (Drosophila genome Release 6) into pTL780, which uses DsRed expression as a transgenic marker (Blanco-Redondo and Langenhan, 2018). The genomic DNA was amplified from the Drosophila genomic reference strain (Adams et al., 2000). The resulting pTL780(RpL28+) plasmid was used for integration at the VK37 landing site on chromosome 2 (Venken et al., 2009).

For irradiation, food vials containing larvae were exposed to 500, 1000 or 4000 rad from a γ-ray source 84 ± 12 hr after egg laying. Dissection, fixation, and immuno-labeling of wing imaginal discs with anti-active Dcp1 and anti-Xrp1 was performed as described previously (Baker et al., 2014; Lee et al., 2018).

Frequencies of cell death and of Xrp1 expression were compared pairwise by t-tests (Figure 1). For multiple comparisons, one-way ANOVA was used with the Holm correction for multiple testing (Figure 1, Figure 10). Previous studies indicated that significant results could be obtained from five biological replicates, where a biological replicate is an imaginal disc preparation labeled, imaged, and quantified (McNamee and Brodsky, 2009). N for each experiment is reported in the figure legends. The extent of white tissue in mosaic eyes was compared using pairwise Mann-Whitney tests with the Benjamini-Hochberg (BH) correction for multiple testing, using FDR ≤ 0.05. There are 109 pairwise Mann-Whitney comparisons made in the main text of this paper, their P-values and the BH corrections are summarized in Supplementary file 2. Where the extent of white tissue in mosaic eyes was compared between multiple genotypes simultaneously, the Kruskal-Wallis test was used with post-hoc follow-up tests using the method of Conover with BH correction using FDR ≤ 0.05. No explicit power analysis was used. All flies obtained were scored in initial experiments, sometimes leading to unequal sample sizes, subsequently we considered 20 eyes of each sex generally sufficient for significant results (while the number of flies that can be obtained is rarely limiting, blinding and scoring clone sizes is time-consuming). N is given in the figures for each experiment. All the figures show experimental and control data obtained from simultaneous parallel experiments in each case, for which all the data scored were included. Some of the genotypes have been generated on multiple occasions with similar results, not all included in the figures.