Our previous results indicated that p53B does not mediate the apoptotic response to radiation in larval imaginal discs and brains (Zhang et al., 2015). One explanation for this lack of function was that p53B protein is expressed at very low levels in those somatic tissues (Zhang et al., 2015). Given the ancestral germline function of the p53 gene family, we considered the possibility that p53B may be expressed and function in the germline. To address this question, we evaluated p53 isoform expression and function in the ovary. During Drosophila oogenesis, egg chambers migrate down a structure called the ovariole as they mature through 14 morphological stages (Figure 1B; King, 1970). Each egg chamber is composed of an oocyte and 15 sister germline nurse cells, all interconnected by intercellular bridges (Figure 1B; Spradling, 1993). The nurse cells become highly polyploid through repeated G / S endocycles during stages 1–10 of oogenesis, which facilitates their biosynthesis of large amounts of maternal RNA and protein that are deposited into the oocyte. The germline cells are surrounded by an epithelial sheet of somatic follicle cells that divide mitotically up until stage 6, and then undergo three endocycles from stages 7 to 10 (Calvi et al., 1998; Deng et al., 2001; Jia et al., 2015). Both germline and somatic follicle cell progenitors are continuously produced by germline and somatic stem cells that reside in a structure at the tip of the ovariole known as the germarium (King, 1970; Drummond-Barbosa, 2019).

To evaluate p53A and p53B expression during oogenesis, we used fly strains transformed with different p53 genomic BAC transgenes in which the p53 isoforms are tagged on their unique N-termini. In one strain, GFP is fused to p53A (GFP-p53A), while in another strain mCherry is fused to p53B (mCh-p53B), with each expressed under control of their normal regulatory regions in these genomic BACs (Zhang et al., 2015). To enhance detection, we immunolabeled these strains with antibodies that recognize GFP and mCherry. Immunofluorescent analysis of the somatic follicle cells revealed that GFP-p53A localized to nuclear bodies, ranging in size from ~0.25 to 1 μm, often in close proximity to the DAPI bright pericentric heterochromatin (Figure 1C–C’). The expression of mCh-p53B, however, was only rarely detected in somatic follicle cells ( < 1 / 50,000 cells) (Figure 1D–D’, Figure 1—figure supplement 1). Thus, similar to our previous results in larval tissues, p53A is expressed at much higher levels than p53B in somatic cells (Zhang et al., 2015). In contrast, both GFP-p53A and mCh-p53B bodies were detected in all germline cells. Early stage nurse cells had one to a few p53 bodies, whereas later stage nurse cells had more bodies that were regionally distributed in the nucleus (Figure 1D–D’). This dynamic pattern suggests that p53 bodies, like some other nuclear bodies, may associate with the polytene chromatin fibers that become dispersed in these nurse cells after stage 4 of oogenesis (Dej and Spradling, 1999; Liu et al., 2006; White et al., 2007; Liu et al., 2009). The oocyte nucleus also had both GFP-p53A and mCh-p53B nuclear bodies, often appearing as one large (~1 μm) and several smaller (~0.25 μm) bodies (Figure 1C–D’). In addition to distinct nuclear bodies, there were low levels of GFP-p53A and mCh-p53B dispersed throughout the nuclei of nurse cells and the oocyte.

We examined females with both GFP-p53A and mCh-p53B to address if they colocalize to the same nuclear bodies. In some cells, co-expression of GFP-p53A reduced the expression of mCh-p53B, perhaps a manifestation of a protein trans-degradation effect that we had described previously (Zhang et al., 2014). Nonetheless, the results indicated that GFP-p53A and mCh-p53B colocalize to the same subnuclear bodies of both nurse cells and oocytes, although the ratio of these two isoforms differed somewhat among bodies (Figure 1E–H). mCh-p53B also co-localized with GFP-p53A in those rare follicle cells that expressed mCh-p53B (Figure 1—figure supplement 1B-C). Examination of the testis indicated that GFP-p53A and mCh-p53B are also expressed and localized to nuclear bodies in the male germline (Figure 1—figure supplement 2A-B'; Mauri et al., 2008; Monk et al., 2012). Altogether, these results indicated that while the p53A isoform is expressed in both somatic and germline cells, the p53B isoform is primarily expressed in the germline.

We next asked which of the p53 isoforms mediate the apoptotic response to DNA damage in the ovary. We had previously addressed this question in larval imaginal discs and brains using mutant BAC rescue transgenes (Zhang et al., 2015). For this study, we used CRISPR/Cas9 to create isoform-specific mutants at the endogenous p53 locus. Since our previous study indicated that the short p53E is a repressor, we focused on making mutants of p53A and p53B isoforms. The resulting p53^A2.3 allele is a 23 bp deletion and 7 bp insertion within the unique p53A 5’ exon (Figure 2A, Figure 2—figure supplement 1A). This deletion extends downstream into the first p53A intron removing both p53A coding sequence and first RNA splice donor site (Figure 2A, Figure 2—figure supplement 1A; Robin et al., 2019). This coding sequence and splice donor site are shared with p53C mRNA, which is predicted to encode a p53 protein isoform that is identical to that encoded by p53A mRNA (Figure 1A). Therefore, p53^A2.3 disrupts both p53A and p53C protein coding. The p53^B41.5 allele is a 14 bp deletion plus 1 bp insertion in the unique second coding exon of p53B, removing p53B coding sequence and creating a frameshift with a stop codon soon afterward (Figure 2B, Figure 2—figure supplement 1A). We had previously shown that p53A mRNA structure is perturbed and p53A protein is undetectable in homozygous p53^A2.3 animals, whereas p53B mRNA is still expressed (Figure 2—figure supplement 1B; Robin et al., 2019). Conversely, RT-PCR indicated that in the p53^B41.5 strain the p53A isoform is still expressed (Figure 2—figure supplement 1B). Thus, p53^A2.3 and p53^B41.5 alleles are specific to each isoform and do not disrupt expression of the other isoform. This is in contrast to the p53^5A-1-4 null allele which deletes the common C-terminus of all the isoforms. To be clear about which of these isoforms are expressed from different p53 alleles, we will annotate wild type p53⁺ as (A+B+), the p53^5A-1-4 null allele as (A-B-), the p53A specific mutant p53^A2.3 as (A-B+) and the p53B specific mutant p53^B41.5 as (A+B-) (Figure 2A and B).

To determine which p53 isoforms mediate the apoptotic response to DNA damage, we irradiated adult females from these strains with 40 Gray (Gy) of ionizing radiation (IR) and evaluated cell death 4 hr later by TUNEL. We focused on the follicle cells in the mitotic cycle up until stage six because we had previously shown that endocycling follicle cells in later stage egg chambers repress the p53 apoptotic response to DNA damage (Mehrotra et al., 2008; Hassel et al., 2014; Zhang et al., 2014; Qi and Calvi, 2016). In wild type p53⁺ (A+B+) ovaries that express both isoforms, approximately 30 follicle cells were TUNEL positive in stage six after IR, whereas the p53^A2.3 (A-B+) mutant had very few TUNEL-positive follicle cells (~1 / stage 6), which was not significantly different than irradiated p53^5A-1-4 (A-B-) null or unirradiated controls (Figure 2C–F and H). In contrast, the p53^B41.5 (A+B-) mutant strain had 30 TUNEL-positive follicle cells, a fraction similar to that of wild type (Figure 2D, G and H). These results suggested that p53A, but not p53B, is required for the apoptotic response to DNA damage. A possible caveat, however, is that both the p53^A2.3 and p53^B41.5 alleles also delete part of a non-coding RNA of unknown function (CR46089), which overlaps the 5’ end of p53 and is transcribed in the opposite direction (Roy et al., 2010; Thurmond et al., 2019). Given that this noncoding RNA is mutated in both p53 isoform-specific alleles, its disruption cannot explain the impaired apoptosis specifically in the p53^A2.3 allele. Moreover, similar results were obtained when p53^A2.3 or p53^B41.5 alleles were transheterozygous to the p53^5A-1-4 null allele that does not delete portions of this non-coding RNA. These results strongly suggest that mutation of non-coding RNA CR46089, or possible cryptic mutations on the p53^A2.3 and p53^B41.5 chromosomes, are not contributing to the apoptotic phenotypes. Thus, the p53A protein isoform is both necessary and sufficient for the apoptotic response to IR in somatic ovarian follicle cells.

The low level of expression of p53B in somatic tissues may explain why it does not mediate the apoptotic response. We wondered, therefore, whether p53B participates in the apoptotic response in the germline where it is more highly expressed. Given that endocycling nurse cells and the meiotic oocyte repress p53-mediated apoptosis, we analyzed the apoptotic response of mitotically-dividing germline cells during early oogenesis in the germarium (Mehrotra et al., 2008; Hassel et al., 2014; Zhang et al., 2014; Qi and Calvi, 2016). At the anterior tip of the germarium, the germline stem cells (GSCs) reside in a niche and divide asymmetrically into a GSC and cystoblast (CB) (Figure 3A; Hinnant et al., 2020). This cystoblast and its daughter cells undergo four rounds of divisions with incomplete cytokinesis as they migrate posteriorly through germarium region 1, finally resulting in an interconnected 16-cell germline cyst (Figure 3A; Drummond-Barbosa, 2019; Hinnant et al., 2020). In region 2a, multiple cells in the cyst initiate meiotic breaks and synaptonemal complex formation, but only one cell is eventually specified to be the oocyte, with the 15 other cells of the cyst destined to become nurse cells that enter a polyploid endocycle by germarium region 3 (stage 1 of oogenesis) (Figure 3A). To evaluate which cells in the germarium express GFP-p53A and mCh-p53B, we co-labeled with an antibody against the fly adducin protein ortholog called Hu-li tai shao (Hts), which labels a spherical cytoplasmic spectrosome in GSCs, and a cytoskeletal structure called the fusome that branches through the ring canals that connect the 16 cells of a germline cyst (Figure 3A; Lin et al., 1994). Similar to later stages of oogenesis, both GFP-p53A and mCh-p53B were expressed in GSCs and their daughter germline cells of the germarium where the p53 isoforms colocalized in distinct p53 nuclear bodies (Figure 3B–C).

To determine which p53 isoforms are required for IR-induced germline apoptosis, we irradiated wild type and p53 mutant females with 40 Gy of gamma rays and TUNEL labeled their ovaries 4 hr later. In wild type p53⁺ (A+B+) controls, there were an average of ~13 TUNEL-positive germline cells in region 1 of each germarium (Figure 3D–E,I). Although earlier GSCs and later meiotic cells express both p53 isoforms, they did not label with TUNEL (Figure 3E). In p53^5A-1-4 (A-B-) null ovaries, only ~1 germline cell per germarium was TUNEL-positive in region 1, a number similar to that in unirradiated controls, indicating that most of the germline cell death 4 hr after IR is p53-dependent (Figure 3D, F,I). Similar to p53^5A-1-4 null, the p53^A2.3 (A-B+) mutant also had ~1 TUNEL-positive germline cell per germarium (Figure 3G,I). In contrast, the p53^B41.5 (A+B-) mutant ovaries had ~12 TUNEL positive cells per germarium, a number similar to that in wild type and significantly greater than that in p53 null and p53^A2.3 mutants (Figure 3H–I). These results suggest that, although p53A and p53B are both expressed, only the p53A isoform that is necessary and sufficient for the apoptotic response to IR in the germline.

To further evaluate p53 isoform function, we determined whether p53A or p53B protein isoforms induce transcription of proapoptotic genes after IR. Previous studies showed that among the proapoptotic p53 target genes, the gene hid plays a prominent role for inducing germline apoptosis in response to DNA damage (Xing et al., 2015; Park et al., 2019). We therefore used a GFP promoter-reporter for hid (hid-GFP), which contains the hid promoter but not coding region, together with GFP antibody labeling to assay p53 transcription factor activity (Tanaka-Matakatsu et al., 2009). Similar to the results for TUNEL, expression of the hid-GFP reporter in region one cystocytes was significantly induced by IR in p53⁺(A+B+) wild type and p53^B41.5 (A+B-) mutants, but not in p53^5A-1-4 (A-B-) null or p53^A2.3 (A-B+) mutants (Figure 4A–I). Together, these results indicate that, similar to the soma, p53A is necessary and sufficient for induction of proapoptotic gene expression and the apoptotic response to IR in the germline.

We noticed that in the unirradiated wild type controls hid-GFP was constitutively expressed in region 2 of the germarium at levels that were less than 50% that of IR. This hid-GFP fluorescence was reduced to background levels in p53^5A-1-4 null mutants indicating that it is reporting a low level of p53 transcription factor activity (Figure 4C–C’). hid-GFP fluorescence was also reduced to background levels in strains mutant for the fly ortholog of Spo-11, mei-W68, which induces DNA double-strand breaks at the onset of meiosis in late region 1/ early region 2 (Figure 4—figure supplement 1 Mehrotra and McKim, 2006). These results suggest that meiotic DNA breaks induce a low level of p53 transcription factor activity, consistent with a previous report from the Abrams lab who used a reporter for another p53 target gene, reaper (rpr-GFP) (Lu et al., 2010). To determine which isoforms respond to meiotic DNA breaks, we quantified hid-GFP expression in the isoform-specific mutants in the absence of IR. The expression of hid-GFP in the p53^B41.5 (A+B-) mutant was not significantly different than wild type p53⁺ (A+B+), suggesting that the p53A isoform responds to meiotic DNA breaks (Figure 4A–A’ and G–G’). Surprisingly, hid-GFP expression in the p53^A2.3 (A-B+) mutant was significantly higher than in p53⁺ (A+B+), and expression occurred earlier in oogenesis in region 1, including GSCs (Figure 4E–E’). Although hid-GFP reporter expression in p53^A2.3(A-B+) was higher than in p53⁺wild type without IR, it was significantly lower than p53⁺ wild type with IR and did not induce apoptosis at any stage. This hid-GFP expression in the p53^A2.3 (A-B+) mutant is not a response to meiotic DNA breaks because they are not induced until late region 1, nor did we detect evidence of DNA damage before region 2 (see below) (Mehrotra and McKim, 2006). Based on evidence from our previous studies and other systems, a likely explanation is that in the absence of p53A the more active p53B homotetramers have a low-level transcription factor activity in the absence of DNA breaks (see discussion) (Zhang et al., 2014; Zhang et al., 2015). Nevertheless, the level of hid-GFP reporter expression was less than that after IR and was not associated with apoptosis. All together, the comparison of hid-GFP expression in p53⁺ (A+B+), p53^5A-1-4 null (A-B-), p53^A2.3 (A-B+), and mei-W68 suggest that meiotic DNA breaks induce low level activity of the p53A transcription factor.

To investigate the relationship of p53 isoforms to meiosis further, we examined their localization in the early germline. The level of GFP-p53A in p53 bodies was comparable among germline cells in all regions of the germarium, including GSC, dividing cystocytes in region 1, and during early stages of meiosis in regions 2a-2b (Figure 5A–A’). mCh-p53B was also abundant in p53 bodies in GSCs and most region one cystocytes, but then decreased at the onset of meiosis in late region 1/ early region 2, remained low in regions 2a-2b, and then increased again in most cells in late region 2b / early region 3 (Figure 5B–B’). This transient decrease of p53B in bodies coincides with known timing of meiotic break induction in late region 1/ region 2a followed by their subsequent repair by region 3 (Hughes et al., 2018). Quantification of GFP-p53A and mCh-p53B levels within the same p53 bodies of GFP-p53A / mCh-p53B females confirmed that although mCh-p53B and GFP-p53A intensity in the bodies is approximately equal in region 1, p53B levels decrease to ~41% that of p53A in regions 2a-2b, and then increase again to levels comparable to p53A in regions 2b-3 (Figure 5C–G, Figure 5—figure supplement 1). This transient reduction in p53B levels in p53 bodies occurs during ~24 hr of oogenesis. These p53B protein dynamics may reflect its degradation and rapid resynthesis or its relocalization from the p53 body to the nucleoplasm and then back again (King, 1970; Morris and Spradling, 2011). The low magnitude and high variance of the nucleoplasmic fluorescence precluded a determination as to whether the decrease of mCh-p53B in bodies is associated with a commensurate increase in the nucleoplasm (Figure 5E, Figure 5—figure supplement 1C-C''). These results suggest that there may be a functional relationship between p53B protein dynamics and meiotic DNA breaks.

To investigate whether p53 isoforms regulate germline DNA breaks, we labeled ovaries with antibodies against the phosphorylated form of the histone 2A variant (γ-H2Av), which marks sites of DNA damage and repair, evident as distinct nuclear DNA repair foci (Madigan et al., 2002; Lake et al., 2013). It has been shown that labeling for γ-H2Av detects repair foci at meiotic DNA breaks beginning in late region 1/ early region 2 a of the germarium (Jang et al., 2003; Mehrotra and McKim, 2006; Lake et al., 2013). Mei-W68 induces breaks in most cells of the 16-cell cyst, but γ-H2Av repair foci are most abundant in four cells, one of which will become the oocyte while the others are destined to become nurse cells (Carpenter, 1975; Jang et al., 2003; Mehrotra and McKim, 2006; Lake et al., 2013; Hughes et al., 2018). Consistent with these previous reports, we observed that ovaries from wild type females had four cells per cyst with prominent γ-H2Av labeling, first appearing at the onset of meiotic recombination in germarium region 2a, decreasing in region 2b, and then undetectable in 97% of oocytes by germarium region 3 (oogenesis stage 1), a time when most meiotic DNA breaks have been repaired (Figure 6A–A”; Jang et al., 2003; Mehrotra and McKim, 2006; Lake et al., 2013).

In contrast, females homozygous mutant for the p53^5A-1-4 (A-B-) null allele had more than four germline cells per cyst with strong γ-H2Av labeling (Figure 6B–B”). This phenotype is similar to that previously reported for mutants required for meiotic DNA break repair, which increase the steady state number of unrepaired meiotic DNA breaks, and thereby the number of cells per cyst that label strongly for γ-H2Av in germarium stage 2 (Mehrotra and McKim, 2006; Wei et al., 2019). Also similar to known DNA repair mutants, repair foci in p53^5A-1-4 persisted into later stages, with 56% of ovarioles having γ-H2Av labeling in both nurse cells and oocytes in stage 1, and 8% of ovarioles having γ-H2Av labeling as late as stage 4 (Figure 6B–B”, Supplementary file 1 for p values). Females homozygous for the p53^A2.3 (A-B+) allele also had more than four cells per cyst that labeled intensely for γ-H2Av, as well as γ-H2Av labeling in oocytes up to stage 1 in 25% of ovarioles, not as frequent as that in the p53^5A-1-4 (A-B-) null (56%) (Figure 6C–C”). Another difference with p53^5A-1-4 (A-B-) null was that the in p53^A2.3 (A-B+) the frequency of γ-H2Av labeling in oocytes was higher than in nurse cells (Figure 6C”, see Supplementary file 1 for p values). Females homozygous for p53^B41.5 (A+B-) also had more than four γ-H2Av-positive cyst cells and repair foci that persisted up to stage 2, later in oogenesis than in p53^A2.3 (A-B+) (stage 1), but not as late as in the p53^5A-1-4 (A-B-) null (stage 4) (Figure 6D–D”). To confirm scoring of oocyte and nurse cells in stage 1, we marked oocytes with anti-Orb antibody, which yielded similar results (Figure 6—figure supplement 1A-G). These experiments also revealed that the nurse cell adjacent to the Orb-positive oocyte was often more intensely labeled for γ-H2Av than other nurse cells (Figure 6—figure supplement 1H). This cell may be the descendant of one of the two ‘pro-oocytes’ in the germarium that are known to have the most numerous meiotic DNA breaks, and which then became a nurse cell while its sister cell adopted the oocyte fate. To test whether the increased DNA damage in p53 mutants is the result of a defect in repair of meiotic breaks, we asked whether the failure to form these breaks in a mei-W68 mutant would suppress the persistent DNA break phenotype of p53 mutants. Labeling of mei-W68; p53^A2.3 (A-B+) and mei-W68; p53^B41.5 (A+B-) double mutants with anti-γH2Av indicated that the persistent γ-H2Av labeling in the p53 mutants is dependent on the creation of DNA breaks by Mei-W68 (Figure 6—figure supplement 3). We also quantified the intensity of γ-H2Av labeling in stage one oocytes and nurse cells, which showed that the p53 mutants had significantly more unrepaired DNA breaks per cell than wild type (Figure 6—figure supplement 2, Supplementary file 2 for p values). γ-H2Av foci were not observed in egg chambers after stage six in any genotype, suggesting that DNA breaks are eventually repaired in the p53 mutants. All together these data reveal that p53A and p53B protein isoforms are required for the timely repair of meiotic DNA breaks.

To further explore the role of p53 isoforms in DNA break dynamics, we tested whether p53 plays a prominent role in the germline when there are defects in meiotic DNA recombination. We examined females doubly mutant for p53 and okra (okr), the fly ortholog of Rad54L, which is required for homologous recombination (HR) DNA repair in meiotic germline and somatic cells (Ghabrial et al., 1998; Sekelsky, 2017; Hughes et al., 2018). It was previously reported that females doubly mutant for okra and a p53 null allele result in egg chambers with extra nurse cells and shorter eggs, which was partially suppressed by mutation of mei-W68, but the relationship of this genetic interaction to DNA break repair, and the possible role of the different p53 isoforms, have not been explored (Lu et al., 2010).

Females that were transheterozygous for two mutant okra alleles (okra^RU/AA) had more than four cells per cyst with strong γ-H2Av labeling in the germarium, which abnormally persisted into germarium region 3 (stage 1), consistent with previous reports that meiotic HR repair is delayed in okra mutants (Figure 7A–B”; Ghabrial et al., 1998; Jang et al., 2003; Mehrotra and McKim, 2006; Lake et al., 2013). Analysis of the timing and intensity of γH2Av labeling in the different okra; p53 double mutants indicated that they all had severe DNA break repair defects (Figure 7C–E”, Figure 6—figure supplement 1D-H, Supplementary files 1 and 2). In okra^RU/AA; p53^5A-1-4 (A-B-) double mutants, γ-H2Av foci persisted up to stage 2 in 100% of ovarioles, with some ovarioles having repair foci in nurse cells and oocyte up to stage 5–6, ~ 30 hr later in oogenesis than wild type (Figure 7C–C”, Figure 6—figure supplement 1D-D'; Lin and Spradling, 1993). The okra^RU/AA; p53^A2.3 (A-B+) double mutants also had repair foci up to stage six in almost all ovarioles (Figure 7D–D”, Figure 6—figure supplement 1E-E', G). Unlike okra^RU/AA; p53^5A-1-4 (A-B-), however, the okra^RU/AA; p53^A2.3 (A-B+) ovaries had much less damage in the nurse cells than in the oocyte, which was not significantly different than γH2Av labeling in wild type nurse cells in either frequency or intensity (Figure 7D–D”, Figure 6—figure supplement 1E-E', G, Figure 6—figure supplement 2, Supplementary files 1 and 2). Given that p53^A2.3 expresses p53B but not p53A, these results suggest that in okra mutants the p53A isoform is required to protect genome integrity in the oocyte, while the p53B isoform plays a prominent role in DNA break repair within nurse cells. okra^RU/AA; p53^B41.5 (A+B-) also had repair foci up to stage 6, but in these ovaries lacking p53B there were numerous repair foci in both the nurse cells and oocyte (Figure 7E–E”, Figure 6—figure supplement 1F-G, Figure 6—figure supplement 2, Supplementary files 1 and 2). Thus, p53B is required for DNA repair in nurse cells and oocytes. These data suggest that p53 isoforms are crucial when HR is compromised, and that they have both overlapping and distinct functions in nurse cells and oocytes to protect genome integrity.

Defects in meiotic DNA recombination and repair are known to activate a meiotic pachytene checkpoint arrest in multiple organisms (Bähler et al., 1994; Gebel et al., 2017). During oogenesis in mice and humans, the pachytene arrest is mediated by p63 and p53 (Suh et al., 2006; Bolcun-Filas et al., 2014; Coutandin et al., 2016; Gebel et al., 2017; Marcet-Ortega et al., 2017; Rinaldi et al., 2020). In Drosophila, it is known that defects in the repair of meiotic DNA breaks activate the pachytene checkpoint, but it is not known whether this checkpoint response requires p53 (Ghabrial et al., 1998; Joyce and McKim, 2011; Hughes et al., 2018). To address this question, we used an established assay that depends on a visible manifestation of the Drosophila pachytene checkpoint arrest, which is a failure of the oocyte nucleus to form a compact spherical structure known as the karyosome, which normally occurs during stage 3 of oogenesis (Ghabrial et al., 1998). It has been shown that activation of the pachytene checkpoint results in an oocyte nucleus with either a diffuse or ellipsoidal morphology (Ghabrial et al., 1998). We examined karyosome formation by labeling with antibodies against the synaptonemal complex (SC) protein C(3)G and DAPI (Page and Hawley, 2001). In wild type females, 95% of ovarioles had a spherical compact karyosome beginning in stage 3 (Figure 7F and K). In okra^RU/AA mutants, compaction was normal in only 5% of ovarioles, with the oocytes in 95% of ovarioles appearing either diffuse or ellipsoidal, consistent with previous reports that DNA repair defects in these okra^RU/AA mutants activate the pachytene checkpoint (Figure 7G and K; Ghabrial et al., 1998). In okra^RU/AA; p53^5A-1-4 (A-B-) null double mutants, however, karyosome compaction was normal in 68% of ovarioles, a fraction that is significantly different than the okra single mutant, suggesting that p53 is required for normal activation of the pachytene checkpoint (Figure 7H and K). Similarly, in okra^RU/AA; p53^A2.3 (A-B+) double mutants karyosome compaction was normal in 68% of ovarioles (Figure 7I and K). In contrast, okra^RU/AA; p53^B41.5 (A+B-) mutants had normal karyosome compaction in only 9% of ovarioles, a fraction that was not significantly different than okra single mutants, indicating that the pachytene checkpoint was activated in most of these ovarioles that expressed p53A but not p53B (Figure 7J and K). Altogether these data suggest that the p53A isoform is required for normal pachytene checkpoint activation when meiotic DNA recombination is impaired, analogous to the functions of mammalian p53 and p63.

The DNA repair defects in the p53 mutants prompted an inquiry into what consequences this genome damage has on oogenesis and female fertility. It is known that mutation of okra and other genes required for meiotic DNA break repair disrupt patterning signals from the oocyte, resulting in ventralized eggs and eggshells and embryos that fail to hatch (Ghabrial et al., 1998; Ghabrial and Schüpbach, 1999; Hughes et al., 2018). Analysis of okra^AA/RU females confirmed that they have a variably expressive, maternal-effect eggshell phenotype, producing eggs that ranged from wild type to different degrees of ventralization, evident as closely spaced or fused eggshell dorsal appendages, while other eggs were small and misshapen with thin shells (Figure 7—figure supplements 1–2). None of the eggs produced by the okra^AA/RU mothers hatched, consistent with previous reports that they are completely female sterile (Figure 7—figure supplement 3; Ghabrial et al., 1998). The p53^5A-1-4 (A-B-) null and p53^B41.5 (A+B-) isoform-specific mutant females also produced eggs with abnormal eggshells, but this phenotype was less severe than in the okra^AA/RU mutants, and only the p53^B41.5 (A+B-) mothers produced clearly ventralized eggs (Figure 7—figure supplements 1–2). The phenotype of p53^A2.3 (A-B+) was much less severe, with most eggs appearing normal (Figure 7—figure supplements 1–2). Egg hatch rates for all of the p53 mutants were significantly lower than wild type controls, indicating that all the p53 mutant alleles have a partial maternal-effect embryonic lethal phenotype (Figure 7—figure supplement 3). The eggshell phenotypes of okra; p53 double mutants were much more severe than either single mutant. The okra^AA/RU; p53^A2.3 and okr^AA/RU; p53^B41.5 mothers produced eggs that were shorter and severely ventralized, with dorsal appendages often fused, shortened, or missing (Figure 7—figure supplements 1–2). The okra^AA/RU; p53^5A-1-4 null mothers produced eggs with the most severe phenotypes, which ranged from completely fused dorsal appendages to very small eggs with extremely thin shells (Figure 7—figure supplements 1–2). Similar to okra single mutants, none of the eggs produced by okra; p53 double mutant mothers hatched (Figure 7—figure supplement 3). These results indicate that mutation of p53 isoforms result in defects in egg patterning and embryo survival that are enhanced by mutation in other genes required for meiotic DNA break repair.

p53 isoforms colocalized to subnuclear bodies in the Drosophila male and female germline (Figure 8A). This finding is consistent with a previous study that reported p53 bodies in the Drosophila male germline, although that study did not examine individual isoforms (Monk et al., 2012). We deem it likely that these p53 bodies form by phase separation, an hypothesis that remains to be formally tested (Mitrea and Kriwacki, 2016; Alberti, 2017). Drosophila p53 subnuclear bodies are reminiscent of human p53 protein localization to subnuclear PML bodies (Mauri et al., 2008). Evidence suggests that trafficking of human p53 protein through PML bodies mediates p53 post-translational modification and function, although the relationship between nuclear trafficking and the functions of different p53 isoforms has not been fully evaluated (Fogal et al., 2000; Chang et al., 2018). Similarly, we observed a decline in abundance of p53B within p53 bodies in germarium region 2a, followed by a restoration of p53B within bodies in region 3. This fluctuation of p53B in bodies temporally correlates with the onset of meiotic DNA breaks in region 2a and their repair in regions 2b - 3. These observations are consistent with the idea that nuclear trafficking of p53B out of bodies may mediate its response to meiotic breaks, although it is also possible that p53B is degraded and rapidly resynthesized during this 24 hr period (Figure 8A). Future analysis of Drosophila p53 bodies will help to define how p53 isoform trafficking mediates the response to genotoxic and other stresses.

TUNEL labeling indicated that p53A is necessary and sufficient for apoptosis in both the germline and soma. IR induced apoptosis to a similar frequency in p53⁺ (A+B+) wild type and p53^B41.5 (A+B-) mutants, whereas the frequency of apoptosis in p53^A2.3 (A-B+) mutants was equivalent to that of p53^5A-1-4 (A-B-) null and unirradiated controls. Consistent with this, hid-GFP reporter expression was not induced by IR in the p53^5A-1-4 (A-B-) null mutant, whereas IRinduced hid-GFP expression in the p53^B41.5 (A+B-) mutant was equivalent to p53⁺ (A+B+) wild type, indicating that the p53A isoform is required for the transcriptional response to IR-induced DNA breaks. It is interesting to note that while germline cystocytes in germarium region one apoptosed after IR, their ancestor GSCs and descendent meiotic cells did not (Figure 8B). The observed IR-induced expression of the hid-GFP promoter reporter in GSCs is consistent with previous evidence that apoptosis is repressed in these stem cells downstream of hid transcription by the miRNA bantam (Wylie et al., 2014; Xing et al., 2015; Ma et al., 2016). How meiotic cells repress apoptosis is not known, although it is crucial that they do so because they have programmed DNA breaks. Together, these data suggest that p53A is necessary and sufficient for induction of proapoptotic gene expression and apoptosis in response to IR-induced DNA breaks in the soma and germline.

While our manuscript was in preparation, it was reported that p53A and p53B both participate in the apoptotic response to IR in the ovary (Park et al., 2019). That study used the GAL4/ UAS system to express either p53A or p53B rescue transgenes in a p53 null background. In contrast, we created and analyzed loss-of-function, isoform-specific alleles at the endogenous p53 locus, which we believe more accurately reflect the physiological function of p53 isoforms. We favor the conclusion, therefore, that it is the p53A isoform that has the primary function of mediating the apoptotic response to IR in the soma and germline.

In the absence of IR, there was a lower but detectable hid-GFP expression at the onset of meiosis in germarium region 2. This region 2 expression was dependent on p53 and formation of meiotic breaks by Mei-W68, consistent with previous reports that used a rpr-GFP reporter to show that p53 responds to meiotic DNA breaks (Lu et al., 2010). This low level of hid-GFP expression in region two without IR was similar between p53⁺ (A+B+) wild type and p53^B41.5 (A+B-) mutants, suggesting that the p53A transcription factor activity responds to meiotic DNA breaks. The results for the p53^A2.3 (A-B+) mutant were not informative, however, because in that mutant hid-GFP expression was constitutively higher than wild type beginning in early region 1 of the germarium. We did not observe γ-H2Av labeling before late region 1/ region 2 a indicating that this low-level activity of p53B is not a response to DNA breaks. While further experiments are required to define the mechanism, a cogent hypothesis is that in the absence of the p53A subunit p53B homotetramers have somewhat higher basal activity. This hypothesis is consistent with our previous evidence that the p53B isoform with a longer transactivation domain is a much stronger transcription factor than p53A, and that p53A and p53B can form heterocomplexes (Zhang et al., 2015). It is also consistent with evidence that the shorter p53 isoforms in humans and other organisms repress the transcriptional activity of longer isoforms in heterotetramers (Anbarasan and Bourdon, 2019). It is important to note, however, that while hid expression was higher in the p53A mutants than in wild type, it was not associated with apoptosis. Overall, while the hid-GFP reporter evidence suggests that p53A responds to meiotic DNA breaks, it is unclear whether this low-level activation of p53A transcription factor activity is related to its role in meiotic DNA break repair or checkpoint activation, which we discuss further below.

Our evidence suggests that both p53 isoforms are required for the timely repair of meiotic DNA breaks in the Drosophila female germline. p53 null and isoform-specific mutants had a persistent germline DNA break phenotype that was dependent on the creation of double-strand DNA breaks by Mei-W68 (Figure 8B). Further consistent with a role in meiotic DNA break repair, p53 mutants had an increased number of cells with γ-H2Av foci beginning in germarium stage 2a, the time when Mei-W68 induces programmed meiotic DNA breaks. Moreover, the number of persistent breaks per cell was higher in oocyte and adjacent nurse cell, the presumptive pro-oocyte, which are known to have more meiotic breaks. This p53 DNA break mutant phenotype is similar to that of okra (RAD54L) and other genes required for meiotic break repair and was enhanced in okra; p53 double mutants. It was previously shown using p53 null alleles that p53 also protects the germline genome by restraining mobile element activity, but we did not evaluate whether one or both of the p53 isoforms are required for this function (Wylie et al., 2016; Wei et al., 2019). Overall, our data strongly suggest that both p53 isoforms have an important role in the repair of meiotic DNA breaks.

Our analysis also revealed that p53 isoforms have overlapping and distinct requirements for meiotic break repair in different cell types (Figure 8B). Both p53A and p53B were required in the oocyte, whereas p53B played the more prominent role in nurse cells, even though nurse cells express both p53A and p53B isoforms. This differential requirement for p53 isoforms may reflect differences in how meiotic breaks are repaired in nurse cells versus oocytes. While it is not known whether DNA repair pathways differ between nurse cells and oocytes, evidence suggests that the creation of meiotic breaks does, with breaks in pro-oocytes but not pro-nurse cells depending on previous SC formation (Mehrotra and McKim, 2006). Important questions motivated by our results are how distinct responses to DNA damage in different cells are determined by different types of DNA lesions, checkpoint signaling and repair pathways, and p53 isoform structure.

The consequences of p53 null and isoform-specific alleles for oogenesis were also similar to okra mutants in that they caused reduced female fertility and defects in eggshell patterning and synthesis. Previous evidence suggested that defective meiotic DNA break repair causes these maternal effect phenotypes in part through disrupting patterning signals from the oocyte to somatic follicle cells (Ghabrial and Schüpbach, 1999). The maternal effect on egg hatch rates, however, was much more severe in the okra mutants, which were completely female sterile, consistent with previous studies (Ghabrial et al., 1998). Thus, although the p53 and okra null mutants had similar levels of germline DNA damage, the severity of their maternal-effect on egg patterning and embryo viability differ, suggesting that some of their pleiotropic effects on oogenesis are distinct. Together, the results indicate that defects in repair of meiotic DNA breaks in both p53 and okra mutant females negatively impact embryo patterning and female fertility.

The requirement for Drosophila p53 in the repair of meiotic DNA breaks is consistent with evidence from other organisms that p53 has both indirect and direct roles in DNA repair. It is known that Drosophila p53 and specific isoforms of human p53 induce the expression of genes that are required for different types of DNA repair (Brodsky et al., 2004; Gong et al., 2015; Williams and Schumacher, 2016). p53 also acts locally at DNA breaks in a variety of organisms, including humans, where it can mediate the choice between HR versus non-homologous end joining (NHEJ) repair (Moureau et al., 2016; Williams and Schumacher, 2016). In fact, it has been shown that human p53 directly associates with RAD54 at DNA breaks to regulate HR repair, consistent with our finding that p53; okra (RAD54L) double mutants have severe DNA repair defects (Linke et al., 2003). Moreover, the C. elegans p53 ortholog CED-4 localizes to DNA breaks to promote HR and inhibit NHEJ repair in the germline (Mateo et al., 2016). Although the hid-GFP reporter indicated that meiotic DNA breaks induce a low level of p53A transcription factor activity, Hid has no known role in DNA repair, and it remains unknown whether p53-regulated expression of DNA repair genes is required for the timely repair of meiotic DNA breaks. We deem it likely that the persistent DNA damage that we observe in the germline of Drosophila p53 mutants may, in part, reflect a local requirement for p53 protein isoforms to regulate meiotic DNA repair (Figure 8A–B). Important remaining questions include whether different p53 isoforms participate indirectly in DNA repair by inducing transcription and directly at DNA breaks to influence the choice among different DNA repair pathways.

Our study has also uncovered a requirement for Drosophila p53 in the meiotic pachytene checkpoint. This function was isoform-specific, with p53A, but not p53B, being required for full checkpoint activation in oocytes with persistent DNA breaks. The failure to engage the pachytene checkpoint in the majority of okra; p53^A2.3 double mutant oocytes is more striking given that these cells had more severe DNA repair defects than the okra single mutants that strongly engaged the checkpoint. While the pachytene arrest was compromised to similar extents in okra; p53 null and okra; p53^A2.3 mutants, some egg chambers in both genotypes did engage a pachytene arrest. This observation suggests that there are p53-independent mechanisms that also activate the checkpoint, perhaps in response to secondary defects in chromosome structure, which are known to independently trigger the pachytene checkpoint in flies and mammals (San-Segundo and Roeder, 1999; Wu and Burgess, 2006; Li et al., 2007; Joyce and McKim, 2009). Moreover, although the pachytene checkpoint was strongly compromised in the p53 null and p53A mutant alleles, it did not suppress okra female sterility, suggesting that other mechanisms ensure that oocytes with excess DNA damage do not contribute to future generations. Altogether, the results indicate that p53A is required for both DNA repair and full pachytene checkpoint activation in the oocytes.

Evidence suggests that the ancient function of the p53 family was of a p63-like protein in the germline (Levine, 2020). Consistent with this, our findings in Drosophila have parallels to mammals where the TAp63α isoform and p53 mediate a meiotic pachytene checkpoint arrest, and the apoptosis of millions of oocytes that have persistent defects (Di Giacomo et al., 2005; Suh et al., 2006; Bolcun-Filas et al., 2014; Gebel et al., 2017; Rinaldi et al., 2017; Rinaldi et al., 2020). Our evidence suggests that the different isoforms of the sole p53 gene in Drosophila may subsume the functions of vertebrate p53 and p63 paralogs to protect genome integrity and mediate the pachytene arrest. Unlike p53 and p63 in mammals, however, Drosophila p53 does not trigger apoptosis of defective oocytes. Instead, the activation of the pachytene checkpoint disrupts egg patterning, resulting in inviable embryos that do not contribute to future generations (Hughes et al., 2018). Thus, in both Drosophila and mammals, the p53 gene family participates in an oocyte quality control system that protects the integrity of the transmitted genome.

Fly strains were reared at 25°C and were obtained from the Bloomington Drosophila Stock Center (BDSC, Bloomington, IN, USA) unless otherwise noted. The hid-GFP fly strain was a generous gift from W. Du (Tanaka-Matakatsu et al., 2009). y w^67c23 and w¹¹¹⁸ strains were used as wild type. The okra mutant strains were obtained from T. Schupbach. mei-W68 alleles were obtained from K. McKim and the BDSC. The mei-W68 null genotype for Figure 6—figure supplement 3 was mei-W68¹/ Df(2R)BSC782. See the Key Resources Table for a complete list of fly strains used in this study.

Isoform-specific alleles of p53 were generated by injection of gRNA encoding plasmid into embryos of a nanos-Cas9 strain using standard methods (Gratz et al., 2014; Ren et al., 2013). Candidate lines for p53B-specific alleles were initially identified by screening for co-knockout of white (Ge et al., 2016). Injections were performed by Rainbow Transgenics (USA). Alleles were identified and characterized by PCR genotyping / sequencing. The p53^A2.3 allele is a 23 bp deletion (coordinates 23053346–23053368 in the D. melanogaster genome version 6.32), and has a seven base-pair insertion (Robin et al., 2019). The p53^B41.5 allele is a 14 bp deletion (coordinates 23053726–23053739) with an insertion of a single Adenine. RT-PCR analysis of p53 isoform-specific mutants was performed on mRNA from adult flies using standard methods. See Figure 2—figure supplement 1 and Key Resources Table for further information about alleles, gRNAs and primers.

Adult females were mated and conditioned on wet yeast for three days. They were then irradiated with a total of 4000 rad (40 Gy) from a Cesium source and were allowed to recover at 25 °C for 4 hr before TUNEL labeling. TUNEL labeling (In Situ cell death detection kit, Fluorescein, Roche) was performed according to manufacturer’s instructions. Follicle cell death in Figure 2 was quantified by counting TUNEL-positive cells in oogenesis stage 6.

Dissection, fixation, antibody labeling, and immunofluorescent microscopy of testes and ovaries were as previously described (Thomer et al., 2004). Primary antibodies, sources, and concentrations used were rabbit anti-GFP (Invitrogen) 1:500, rabbit anti-dsRed (Clontech) 1:200, mouse anti dsRed (Clontech) 1:200, mouse anti Hts1B1 (DSHB) 1:20 (Zaccai and Lipshitz, 1996), mouse anti-γH2Av (DSHB) 1:1,000 (Lake et al., 2013), and mouse anti-Orb 4H8 isotype IgG1 (DHSB) 1:500. The anti-γH2Av antibody was preabsorbed against fixed wild type ovaries before use. Secondary antibodies were Alexa 488 anti-rabbit, Alexa 488 anti-mouse, Alexa 568 anti-rabbit, and Alexa 568 anti-mouse (Invitrogen) all used at 1:500-1-750. For Figure 6—figure supplement 1, specialty isotype-specific Alexa 488 (IgG1) and Alexa 568 (IgG2b) were used to distinguish between γ-H2Av and orb-4H8 as previously described (Collins et al., 2014). Samples were counterstained with DNA dye 4′,6-diamidino-2-phenylindole (DAPI) at 1 μg/ml. Confocal micrographs were captured on a Leica SP8 confocal using a 63 X multi-immersion lens.

Fluorescence intensities were quantified using the LASX software of the Leica Sp8 confocal microscope. The hid-GFP fluorescence intensity of Figure 3 was measured across z-stacks of each germarium. The intensities of GFP-p53A and mCh-p53B in p53 bodies in Figures 1 and 5, Figure 1—figure supplement 1, and Figure 5—figure supplement 1 were quantified along a line. For strains expressing both GFP-p53A and mCh-p53B, the ratio of GFP-p53A: mCh-p53B was quantified within each body. For those expressing a single tagged isoform, the ratios of intensities in germarium regions 1:2 or 2:3 among different bodies was calculated, all within the same germarium to control for technical variation. For all experiments, germline cells were identified by labeling with either anti-Hts, anti-Vasa, or anti-Orb. In Figures 6 and 7, and Figure 6—figure supplement 1, the persistence of γ-H2Av into later stages was quantified by scoring oocyte and nurse cells whose total fluorescent intensity was significantly above that of wild type controls. Similar methods were used in Figure 6—figure supplement 2 to quantify total anti-γH2Av intensity in individual oocytes or nurse cells in stage 1. Statistical analyses were performed using GraphPad Prism. See figure legends and Supplementary files 1 and 2 for sample sizes and p values.