Multiple recent sex chromosome fusions in Drosophila virilis associated with elevated 1 satellite DNA abundance 2 3

Repetitive satellite DNA is highly variable both within and between species, and is often located near centromeres. However, the abundance or array length of satellite DNA may be constrained or have maximum limits. Drosophila virilis contains among the highest relative satellite abundances, with almost half its genome composed of three related 7 bp satellites. We discovered a strain of D. virilis that has 15% more pericentromeric satellite DNA compared to other strains, and also underwent two independent centromere-to-centromere sex chromosome fusion events. These fusions are presumably caused by DNA breakage near the pericentromeric satellites followed by repair using similar repetitive regions of nonhomologous chromosomes. We hypothesized that excess satellite DNA might increase the risk of DNA breaks and genome instability when stressed, which would be consistent with the apparent high rate of fusions we found in this strain. To directly quantify DNA breakage levels between strains with different satellite DNA abundances, we performed the comet assay after feeding flies gemcitabine and administering low-dose gamma radiation. We found a positive correlation between the rate of DNA breakage and satellite DNA abundance. This was further supported by a significant decrease in DNA breakage in an otherwise genetically identical substrain that lost the chromosome fusion and several megabases of satellite DNA. We find that the centromere-to-centromere fusions resulted in up to a 21% nondisjunction rate between the X and Y chromosomes in males, adding a fitness cost. Finally, we propose a model consistent with our data that implicates genome instability as a critical evolutionary constraint to satellite abundance.

strains with different satellite DNA abundances, we performed the comet assay after feeding 23 flies gemcitabine and administering low-dose gamma radiation. We found a positive correlation 24 between the rate of DNA breakage and satellite DNA abundance. This was further supported by 25 a significant decrease in DNA breakage in an otherwise genetically identical substrain that lost 26 the chromosome fusion and several megabases of satellite DNA. We find that the centromere-27 to-centromere fusions resulted in up to a 21% nondisjunction rate between the X and Y 28 chromosomes in males, adding a fitness cost. Finally, we propose a model consistent with our 29 data that implicates genome instability as a critical evolutionary constraint to satellite 30 abundance. 31 32 INTRODUCTION 34 35 Satellite DNA consists of long arrays of tandemly repeated sequences, and is often located near 36 centromeres in heterochromatin (reviewed in Thakur et al. 2021). Satellite DNA varies greatly in 37 sequence and abundance within and between species (Subirana et al. 2015;Wei et al. 2018; 38 Cechova et al. 2019), and this can be partially explained by high rates of copy number mutation 39 (Flynn et al. 2017). Although previously assumed to be inert "junk," recent work has shown that 40 satellite DNA is involved in essential processes in the cell, thus variation in it may be biologically of transposable elements (TEs), the other highly pervasive type of repetitive DNA, seems to be 50 less constrained than satellite DNA with many genomes over 50% and some containing up to 51 85% TE content (Anderson et al. 2019). The nature of satellite DNA with long tandem arrays of 52 the same sequence, may impose instability that prevents it from expanding beyond a threshold, 53 compared to more diverse sequences interspersed in the genome. 54 55 Drosophila virilis is an excellent model for studying satellite DNA variation. D. virilis has the 56 highest relative abundance of simple satellite DNA (defined as satellites with unit length <=20 57 bp) compared to any other studied species. Three 7 bp satellites, AAACTAC, AAACTAT, and 58 AAATTAC take up over 40% of the genome in D. virilis, and they form arrays tens of megabases 59 long in the pericentromeric region (Gall et al. 1971; Gall and Atherton 1974;Flynn et al. 2020). 60 The extremely high relative abundance of satellite DNA in D. virilis makes it an ideal system in 61 which to ask whether there are constraints or maximum limits on satellite abundance. One 62 strain in particular, vir00 (15010-1051.00), contains 15% more pericentromeric satellite DNA 63 compared to other D. virilis strains (Flynn et al. 2020 Figure 4B). In past modeling efforts, 64 satellite DNA arrays have been proposed to be weakly deleterious until they reach a maximum 65 length beyond which they are not tolerated by selection (Charlesworth et al. 1986). Slow DNA 66 replication or development time have been suggested as mechanisms to enforce strong 67 negative selection against long satellite arrays, however empirical evidence for this has been 68 limited (but see Bilinski et al. 2018 Fachinetti 2018). Robertsonian translocations, one of the most common rearrangements in 82 medical genetics and evolution, occur when there are breaks near the centromere of 83 acrocentric chromosomes and when they are repaired they are fused to each other (Mayrose 84 and Lysak 2021). Robertsonian translocations are associated with multiple miscarriages in 85 humans and aneuploidy disorders like Patau and Down syndromes (Braekeleer and Dao 1990), 86 with increased rates of aneuploidy driven by increased nondisjunction of Robertsonian or fused 87 chromosomes (Schulz et al. 2006). We hypothesize that variation in pericentromeric satellite 88 DNA abundance influences the risk of genome instability events. Specifically, excess satellite 89 DNA might increase the risk of genome instability and genome rearrangements. 90 91 Sex chromosome evolution has been studied for decades, mainly making use of sex 92 chromosomes that have arisen in different time periods (Charlesworth and Charlesworth 2000). 93 In Drosophila, when an autosome fuses to either an X or Y chromosome, a so-called neo-Y 94 chromosome is formed. Because either the fused or unfused version of the chromosome will 95 only be present in males and male Drosophila do not undergo recombination, mutations 96 immediately begin to accumulate through Hill-Robertson interference and other linked-97 selection processes (Charlesworth and Charlesworth 2000 Here, we describe the discovery of two independent and extremely recent sex chromosome-108 autosome fusions in one D. virilis strain, vir00. We hypothesize that this strain has been more 109 prone to DNA instability events, possibly caused by its excessive satellite DNA abundance. After 110 applying DNA replication and physical stressors, we measured DNA damage levels directly and 111 demonstrate that the DNA damage response is associated with satellite abundance. We use 112 two genetically identical strains that differ only by a chromosome fusion and satellite DNA 113 abundance, and demonstrate that the strain with more satellite DNA has significantly increased 114 DNA damage when stressed. Finally, we propose a model that genome instability may impose a 115 constraint on satellite abundance, a model that is entirely consistent with our data. 116

119
Two novel sex chromosome Robertsonian translocations in D. virilis strain vir00 120 121 In summer 2019, we performed DNA fluorescence in-situ hybridization (FISH) on larval 122 neuroblast nuclei of the vir00 strain that we had obtained several months earlier from the 123 National Drosophila Species Stock Center. We discovered it contained a Y-autosome fusion 124 ( Figure 1A). The Y chromosome is recognizable in D. virilis because it has a distinct DAPI staining 125 intensity pattern, and contains a distinct arrangement of satellite DNA (Flynn et al. 2020). 126 However, all other chromosomes are difficult to distinguish in metaphase spreads, so we could 127 not immediately determine the fusion partner. We observed that the fused chromosome 128 Center. We set up single pair crosses and performed larval neuroblast squashes to karyotype 137 multiple male progeny of each cross. Surprisingly, we found that the Y-autosome fusion was not 138 present in any of the larvae karyotyped. However, 3/10 crosses karyotyped contained a 139 different fusion. This new fusion contained the brightly-dapi-staining AAATTAC satellite as the 140 centromere-proximal satellite, indicating that it is completely distinct from the Y fusion and 141 involved different chromosomes. We hypothesized that this new fusion involved the X 142 chromosome for several reasons; 1) based on centromere satellite identity, it had a 67% 143 probability (Flynn et al. 2020); 2) it was found only in single copy in male larvae, but sometimes 144 two copies in female larvae; 3) it was associated with observed X-Y nondisjunction events, such 145 as the presence of an XXYY female ( Figure S1). We performed single-pair crosses and screened 146 the resulting progeny until we isolated a substrain fixed for the X fusion, and called this 147 substrain vir00-Xfus. We maintained one of the cross descendants from the 2019 stock that did 148 not have any evidence of the X fusion, which we designate vir00-Nofus. We inferred that both X We designed two separate two-generation crossing experiments to validate the fusions 164 genetically and identify the autosome each sex chromosome is fused to. Both experiments 165 exploited autosomal markers which we could determine if they were segregating non-166 independently of sex. The first experiment to validate the Y-autosome fusion used crosses 167 between vir00-Yfus and D. novamexicana and scoring of microsatellite loci on each candidate 168 autosome. We scored 82 F2 male progeny for the Chr3 marker, and 79/82 contained both 169 alleles, whereas the null hypothesis was 50% should contain both alleles ( Figure 1C, Figure S2).

8
The three progeny that did not contain both alleles were determined to have nondisjunction 171 events and did not contain the Y chromosome ( Figure S3). We concluded that the Y 172 chromosome is fused to Chr3, and that the fusion was fixed in this subline. The other markers 173 on Chr2, Chr4, and Chr5 segregated independently of sex and acted as negative controls (Table  174 S1). We also did a negative control with the same crossing scheme except with vir08 instead of 175 vir00 (Chr3 chi-square p = 0.39, N=22, Table S1). 176

177
The second experiment to validate the X-autosome fusion used crosses to D. virilis transgenic 178 lines containing GFP markers on one of each of the candidate autosomes. For the crosses to 179 Chr4-GFP (vir95), we phenotyped 304 F2 female progeny, and found that progeny containing a 180 GFP signal were significantly depleted compared to the Mendelian expectation of 50% ( Figure  181 1D). This indicated that the X chromosome was fused to Chr4. This crossing scheme with two 182 other candidate autosomes did not show association of GFP signal with sex (Chi-square p-value 183 > 0.1, Table S2). We also performed a negative control with the D. virilis genome strain instead 184 of vir00-Xfus crossed to the Chr4-GFP line and found the GFP signal was independent of sex 185 (Chi-square p-value = 0.92, Table S2). 186

187
We validated that vir00-Yfus and vir00-Xfus are the same genetic line and not the result of 188 contamination from other lines either in our lab or the stock center. We made use of medium-189 coverage whole genome sequencing from Flynn et al. (2020) to design primers to amplify 190 singleton insertion/deletion variants present only in vir00 (the version that was sequenced was, 191 in hindsight, vir00-Yfus) and not in any other wildtype D. virilis strain present in the stock center 192 (Table S3). We designed primers to amplify four loci on chromosomes 2, 3, 5, and 6 which 193 contain a homozygous 12-13 bp deletion in vir00 compared to the reference and the other 194 strains (Table S4). We found that both versions of vir00 contained the deletion at each of these 195 loci, confirming that these substrains are indeed the same strain and not a contamination 196 ( Figure S4). Although we did the above indel experiment first, the whole genome resequencing 197 SNP analysis (below) was also concordant with these substrains being the same genetic line. 198 199

Satellite DNA decreased in vir00-Nofus compared to vir00-Yfus 200 201
We used Illumina to resequence the three vir00 substrains: vir00-Yfus, vir00-Xfus, and vir00-202 Nofus. We then used k-Seek to quantify the satellite abundance in each substrain (Wei et al. 203 2014). Assuming the most ancestral strain was vir00-Yfus, we found that satellite abundances 204 decreased in the two derived lines vir00-Xfus and vir00-Nofus (Table S5). In vir00-Xfus, there 205 was an 8% decrease only in the centromere proximal satellite of ChrX and Chr4 (AAATTAC). 206 vir00-Nofus had an overall 12% loss of satellite DNA compared to vir00-Yfus ( Figure 3b). vir00-207 Nofus had a similar 8% decrease in AAATTAC, in addition to a 10% loss in the pericentromeric 208 satellite (AAACTAC) and a 13% loss in the centromere-proximal satellite of ChrY and Chr3 209 (AAACTAT). These data are consistent with our interpretation that vir00-Nofus is the result of 210 breaking apart of the fusion chromosomes, and we conclude that significant satellite DNA was 211 lost from all three pericentromeric and centromeric satellites in vir00-Nofus. Since spontaneous DNA breakage events are rare and we did not want to confound our data 238 with breaks that occur as part of meiotic crossing over, we used stressors to elevate the rate of 239 DNA damage. This would allow us to potentially detect a difference in the phenotype of interest 240 between strains and also amplify possible types of stress imposed by excess satellite DNA 241 content. We fed 0-1 day old adult flies the nucleoside analog gemcitabine for 8 days, which 242 stalls replication forks and acts as a sensitizer for radiation via the Rad51 pathway 243 (Kobashigawa et al. 2015). We then irradiated these sensitized flies with gamma rays at low 244 level radiation (10 Gy). For each of seven strains tested, we included a control which was fed 245 with the same liquid food with no gemcitabine and did not receive radiation treatment. We 246 used the comet assay or single-cell gel electrophoresis to measure DNA damage in the male 247  (Table  249 S6). To detect differences in DNA damage in response to stress between strains, we took the 250 mean difference in olive moments between the control and stress treatments (see Methods). 251

252
The strain with the lowest satellite abundance, the genome strain, contained the lowest DNA 253 damage response and this was significantly lower than all other strains tested (Figure 2A contributing to the variation we found. Ideally, to demonstrate that satellite DNA is a causal 258 factor, we would manipulate satellite DNA abundance and test the DNA damage phenotypes, 259 but multi-megabase long arrays of satellite DNA cannot be manipulated with traditional 260 genome editing. However, vir00-Yfus and vir00-Nofus differ only by a chromosome fusion and 261 12% satellite abundance. Thus, if there is a difference in DNA damage response between these 262 substrains, it may be caused by differing satellite abundance. vir00-Nofus, which contained 12% 263 less satellite DNA than vir00-Yfus, had a significantly reduced DNA damage response, which is 264 concordant with our expectations that satellite DNA plays a causal role ( Figure 3B Flies treated with gemcitabine and radiation are suffixed with "rad" and control flies are 272 suffixed with "con". The lower panel shows the unpaired mean difference between "rad" and 273 "con" for each fly strain (black dot) and 95% confidence intervals (black line) produced from 274 dabestr (5000 bootstrap method  Elevated rates of X-Y nondisjunction represent a fitness cost because zygotes with infertile or 314 lethal karyotypes will form at increased frequency. We found some evidence of nondisjunction 315 in the vir00-Yfus genetic validation ( Figure S3), and also common Y chromosome aneuploidy in 316 the stock of vir00-Xfus ( Figure S1). Since both fusions involve the sex chromosomes, we tested 317 the rate of primary X-Y nondisjunction in males in all three versions of vir00, along with the D. 318 virilis genome strain as a control. We first fully isolated the vir00-Nofus strain to ensure no 319 fusion chromosomes were segregating, which even at low frequency could increase the rate of 320 nondisjunction in the line. 321

322
We crossed individual males of each strain to genome strain females and genotyped for the 323 presence or absence of the Y chromosome in progeny. Female progeny containing a Y 324 chromosome indicate XY sperm from the father, and male progeny lacking a Y chromosome 325 indicate nullisomic sperm from the father. Since nondisjunction was extremely high in vir00-326 Xfus, 2/7 fathers tested were of XYY karyotype, which we could infer if more than half of his 327 female progeny contained a Y chromosome (Maggert 2014). We eliminated these fathers' 328 progeny from the primary nondisjunction rate calculation. No fathers tested from other 329 sublines were determined to be XYY. Males without a Y chromosome would not produce 330 progeny. We found that vir00-Yfus had a slightly elevated nondisjunction rate of 4.5%, 331 compared to the genome strain control of 1.2%, but it was not statistically significant with the 332 sample sizes we used (Table 1). vir00-Xfus had an extremely high primary nondisjunction rate of 333 21% (e.g. Figure S5), which was significantly higher than that of all other substrains (p < 0.004, 334 pairwise proportion test, Table 1). Surprisingly, vir00-Nofus had a significantly elevated 335 nondisjunction rate compared to the genome strain control, at 5.7%. This was not statistically 336 different from vir00-Yfus. 337 338  We identified elevated nondisjunction rates as a cost of chromosome fusions. Surprisingly, the 389 rate of nondisjunction was 5-fold higher in vir00-Xfus compared to vir00-Yfus. With a 390 nondisjunction rate of 21%, a significant proportion of abnormal karyotypes will be produced, 391 such as XO (sterile), XYY (viable and fertile), and XXY or XXYY (viable and fertile), all of which we 392 found in cytological samples. Further mating between these abnormal karyotypes will produce 393 significant proportions of sterile or inviable karyotypes like XYYY or XXX, which will further 394 decrease the fitness of this line. Furthermore, karyotypes with extra Y chromosomes such as 395 XYY and XXY have been found to have decreased lifespan (Brown et al. 2020b). Although we did 396 not assay the nondisjunction rate of the fused autosome, it is possible the nondisjunction rate 397 of Chr4 is also elevated in vir00-Xfus, which would further decrease the fitness of the line 398 because Chr4 aneuploidy is expected to be lethal (Lindsley et al. 1972). The extreme 399 nondisjunction rate indicates X-Y pairing is severely disrupted due to the X-4 fusion, but only 400 slightly if at all due to the Y-3 fusion. When the Y fusion presumably broke apart and the X 401 fusion formed, it is possible other rearrangements on the X and/or Y occurred that disrupted 402 pairing. Furthermore, the nondisjunction rate did not decrease to a level similar to the genome 403 strain in vir00-Nofus. We believe this indicates a remaining rearrangement in vir00-Nofus 404 affecting the pairing and disjunction of the X-Y. We only found a modest difference in estimated 405 rDNA copy number between the substrains (Table S7), which has been found to mediate pairing 406 of the X and Y (McKee and Karpen 1990). Detailed analysis of structural rearrangements in the 407 heterochromatin will be required to determine the mechanism of the elevated nondisjunction 408 rates. 409 410 Our study has several limitations. First, the vir00-Nofus flies we used for resequencing and for 411 DNA damage assays had the X-4 fusion segregating at low frequency (<15%), which was 412 unknown to us until we were able to correct it for the nondisjunction assays. However, we 413 believe our results hold firmly because the main comparison was with vir00-Yfus, the substrain 414 with the most satellite DNA. For testing our hypothesis, the fusion status in the other substrains 415 matters less than the satellite DNA abundance, which was markedly lower in the vir00-Nofus 416 substrain we used. Furthermore, the increase in damage level between the control and stressed 417 flies may not be directly applicable to the risk of DNA breaks and genome instability in natural 418 conditions. Although we find a difference in the DNA damage in response to stress between 419 vir00 substrains with different abundances of satellite DNA, their fusion status is also different. 420 We cannot eliminate the possibility that the presence of the Y fusion itself increased the rate of 421 DNA damage instead of the abundance of satellite DNA. Finally, we cannot eliminate the 422 possibility that satellite DNA increased in vir00-Yfus after the fusion occurred and not prior to as 423 we suggest in our model. 424 425 We believe the system we discovered will be useful for a variety of future studies. The vir00 426 fusion substrains will be useful for studying centromere identity. In both the Y-3 and X-4 427 fusions, two spherical regions of satellite DNA are present at the centromere of these fusions, 428 representing one from each acrocentric chromosome (Figure 1). We note that the X-4 fusion in 429 We designed an experiment that would both validate the Y chromosome fusion and to 459 distinguish which autosome is fused. We first designed primers flanking microsatellite loci on all 460 four autosomes that met the following criteria: 1) had 100% conserved non-repetitive and 461 unique priming sites between D. virilis and D. novamexicana; 2) amplicon length differed 462 between the species by at least 15 bp as to be distinguished on an agarose gel; 3) locus 463 contained a mono or tri nucleotide repeat; 4) locus length ~200 bp. We next set up a two-464 generation crossing scheme ( Figure 1A). We crossed D. novamexicana virgin females with vir00-465 Yfus males and selected the male progeny, which we backcrossed to D. novamexicana virgin 466 females. We then genotyped the male F2 progeny from this cross at the 4 sets of primers 467 corresponding to the four non-dot autosomes (Chr2, 3, 4, 5) (Table S4). We performed single-fly 468 DNA extraction in strip tubes with Tris-EDTA buffer and 0.2 mg/mL proteinase K. We did 12 uL 469 standard PCR reactions (3 min at 95, 30 cycles of 30 sec 95, 30 sec 55, 50 sec 72, final extension 470 5 min). Each primer on each PCR plate had a homozygous (D. novamexicana) and heterozygous 471 (D. novamexicana-D.virilis F1 hydrid) control. We then ran the PCR product on 2.5% agarose 472 gels. If there was indeed an autosome fused to the Y chromosome, we would expect to see 473 100% of the male progeny being heterozygous for the virilis and novamexicana alleles (except 474 for rare cases of non-disjunction). For the autosomes that are not fused, we would expect to 475 see 50% of the progeny being homozygous for the novamexicana allele, and half heterozygous, 476 due to Mendel's law of random segregation. We successfully validated the existence of the Y 477 fusion, and found that it is fused to chromosome 3 (Muller D) (Table S1, Figure 1). There were 3 478 male progeny that were homozygous for the Chr3 novamexicana allele. We verified that these 479 were cases of nondisjunction (opposed to the Y fusion not being fixed in this subline) by finding 480 that the Y chromosome was absent in controlled Y chromosome PCR assays ( Figure S3). 481 482 Isolation of vir00-Xfus and vir00-Nofus 483 484 The X fusion was found to be segregating with a no fusion substrain in the 2019 stock of vir00. 485 We wanted to isolate these into two separate substrains where the karyotype is fixed. From the 486 progeny of the three original crosses in which we found the X fusion, we made 10 single pair 487 crosses and did neuroblast squashes of 6-8 larval progeny per cross, including both sexes. By 488 chance, we should be able to find a cross in which the mother had two copies of the fusion and 489 the father had a single copy -in which the derived line would be fixed for the fusion. If all 490 progeny imaged contained the fusion (and females contained two copies of the fusion), then it 491 is likely that this was the case. We created this line, and call it vir00-Xfus. We maintained a line 492 isolated from the 2019 stock that had no evidence of the X fusion and called it vir00-Nofus. We 493 later found that vir00-Nofus still had the X fusion segregating at low frequency. We isolated a 494 fixed Nofus version in the same way as above for nondisjunction assays, because a low 495 frequency fusion could increase the nondisjunction rate greatly. GFP signal in their brain. We chose to phenotype at the larval stage since we would be crossing 507 GFP strains to wildtype red-eyed flies and the visibility of GFP in the adult eye would be low. We 508 then designed a crossing scheme which would allow us to both validate that the X chromosome 509 was fused and distinguish which autosome it was fused to ( Figure 1B). We crossed GFP-line 510 males to vir00-Xfus virgin females. We then selected the male F1 progeny and crossed them to 511 virilis genome strain virgin females. We then phenotyped F2 larvae, classifying each as either 512 GFP positive or negative. When the phenotyped flies emerged, we sexed and counted them. If 513 the candidate autosome is fused to the X, we would expect sex to segregate with the GFP 514 marker: all female progeny will be GFP negative, and all male progeny will be GFP positive 515 (except for phenotyping errors or rare nondisjunction events). For all other lines, sex should not 516 segregate with GFP status. We performed negative control crosses in which the parental cross 517 was replaced by genome strain virgin females, to ensure the crossing scheme produced the 518 expected results (Table S2). We found that the X chromosome is fused to Chr4 (Muller B). 519 520

Validation of the three versions of vir00 with private fixed indels 521 522
We used GATK recommended practices to do genotyping of our low-coverage whole genome 523 sequencing data from Flynn et al. (2020). We used vcftools to subset singletons present only in 524 vir00, which was, in hindsight, vir00-Yfus. We then used GATK's SelectVariants to select only 525 non-reference homozygous indels 12 bp or more with a depth of at least 10 in vir00 and at least 526 2 in the other strains. We then manually inspected each potential candidate in IGV to ensure: 527 no reads in other strains supported the indel, all reads in vir00 supported the indel, and there 528 were no nearby indels in other strains. We then designed primers for the four loci (on Chr 2, 3, 529 5, 6) that met these criteria and also had enough SNP-free sequence flanking the indel in order 530 to design primers that would amplify a locus 100-200 bp equally in all strains. We performed 531 PCR and gel electrophoresis (2.5% gel, 98 V, 90 min). 532 533 DNA damage assays 534 535 We chose two stressors that would moderately increase the rate of DNA breaks and allow us to 536 potentially detect differences between strains. Gemcitabine is a nucleoside analog that induces 537 replication stress by stalling polymerases, and also sensitizes cells to radiation via the RAD51 538 pathway (Kobashigawa et al. 2015). We selected dosage and a fly-feeding regime based on 539 (Kislukhin et al. 2012). Ionizing radiation has long been used to increase the rates of DNA breaks 540 in flies for mutagenesis. We chose a dose ¼ -½ of what has been typically used in mutagenesis 541 (Carlson and Southin 1962). 542 543 We collected male flies 0-1 days old and fed them gemcitabine (0.718 mM) mixed with liquid 544 food in vials with 8-12 adult flies as in (Kislukhin et al. 2012). Liquid food consisted of 12.5g 545 sucrose, 17.5 g dry yeast, 5 mL corn syrup, and 95 mL PBS (autoclaved for 30 min immediately 546 after adding the yeast). Flies were fed the drug for 7-9 days before radiation. Controls were fed 547 with the same liquid food for 7-9 days, except no gemcitabine was added. Flies were moved to 548 fresh vials every 3-4 days. For radiation treatment, we transferred flies into 50 mL conical tubes 549 with 5 mL agar because these tubes were compatible with the radiation source. Control flies 550 were also transferred to new tubes. We used a J.L. Shepherd & Associates Mark I Irradiator with 551 1,100 Ci of Cs-137, and flies were irradiated at approximately 400 rad/min for a total of 10 Gy. 552 In one case, for the GDvir stress treatment, the radiation was not stopped on time so 4 extra Gy 553 were applied. We believe this did not affect our results, especially because the GDvir strain had 554 the lowest DNA damage increase with gemcitabine and radiation stress. We did comet assays 555 to measure DNA damage (Angelis et al. 1999) over three different dates (Table S6), but ensured 556 experimental conditions were practically identical each time. For some samples we had to 557 combine results from two different dates to have enough nuclei for statistical analysis (Table  558 S6). We dissected testes from approximately 8 flies from each treatment within one hour of 559 radiation treatment to minimize the opportunity for breakage repair (Shetty et al. 2017). We 560 then homogenized the testes tissue using a dounce, filtered the homogenate through a 40 561 micron filter to remove debris, and centrifuged and resuspended the cells to approximately 10 5 562 cells/mL. 563 564 We next performed the alkaline comet assay as directed by the Enzo comet kit (ADI-900-166), 565 which provides higher sensitivity than the neutral comet assay (Angelis et al. 1999). We imaged 566 slides on a metamorph imaging system at 10x magnification using a fluorescent green filter to 567 detect the CyGreen dye included in the comet kit. We quantified damage levels using the 568 software OpenComet as a plugin in ImageJ (Gyori et al. 2014). We filtered called nuclei that 569 were not comet shapes or contained background interference. We used the measure of "olive 570 moment," which is the product of the percent of DNA in the tail and distance between 571 intensity-weighted centroids of head and tail (Gyori et al. 2014) as the statistic to compare 572 between strains and treatments. 573 574 Resequencing sublines to determine differences in their satellite abundance. 575 576 Pools of 6 male flies were DNA extracted with Qiagen DNeasy blood and tissue kit. PCR-free 577 libraries were then prepared with Illumina TruSeq PCR-free library prep. Libraries were 578 sequenced on a NextSeq 500 single end 150 bp. We removed adapters and poly-G signal with 579 fastp and then ran k-Seek to count satellite abundances (Wei et al. 2014). We used average 580 read depth to normalize the kmer counts. We also mapped the reads to the D. virilis rDNA 581 consensus sequence (http://blogs.rochester.edu/EickbushLab/?page_id=602) to estimate the 582 rDNA copy number in the three vir00 substrains as well as a vir08 as a control (Table S3, S7). 583 584 Using sequencing data to estimate the age of the Y-3 fusion 585 586 Scripts for this section are available here: https://github.com/jmf422/D-virilis-fusion-587 chromosomes/tree/main/simulate_degradation. We used the sequencing data from Flynn et al. 588 (2020) in addition to data produced here for vir00-Yfus, vir00-Xfus, and vir00-Nofus and 10 589 other D. virilis strains. We mapped the data to the RS2 genome assembly using bowtie2. We 590 then genotyped with GATK following standard procedures (McKenna et al. 2010). We extracted 591 heterozygous singleton sites for vir00, and counted how many occurred on each autosome. We 592 calculated the enrichment on Chr3 in vir00-Yfus based on the difference from the average SNP 593 density on the other autosomes (excluding the dot chromosome Chr6). To determine whether of mutations, we randomly permuted the total number of heterozygous singleton SNPs on all 596 autosomes and calculated the proportion falling on Chr3, and repeated this 1000 times. 597 598 We then performed simple simulations to determine approximately how many generations of 599 mutation accumulation without recombination or selection would result in the enrichment we 600 observed. Since heterozygous singletons are challenging for the genotyper to call with 601 moderate coverage sequencing data, we incorporated this into our simulation. First, we made 602 the genome assembly diploid then used mutation-simulator (Kühl et al. 2020) to simulate 603 random mutations (transition/transversion ratio 2.0) at a rate of 2 x 10 -9 per bp per generation 604 for 500, 1000, 2000, and 5000 generations on one copy of Chr3 only. We then simulated 605 Illumina reads with ART (Huang et al. 2012) at the same depth as we have for vir00-Yfus in our 606 real data (23 x haploid or 11.5 x diploid). We next used standard GATK genotyping and selected 607 out heterozygous singletons on Chr3. We repeated the simulation 10 times for each number of We crossed a single male from vir00-Yfus, vir00-Xfus, vir00-Nofus (fixed), and GDvir (control) to 614 one or two GDvir (genome strain vir87) females. We collected the virgin progeny from each 615 cross, extracted DNA with a squish-proteinase K prep, and genotyped with PCR and gel 616 electrophoresis for the presence or absence of the Y chromosome in up to 16 female and 16 617 male progeny. We amplified a locus unique to the Y chromosome (primers designed by Yasir 618 Ahmed-Braimah for a different project, Table S4). For a subset of individuals, we also 619 performed multiplex controls with an autosomal locus. Otherwise, we performed DNA 620 extractions in large batches with the same proteinase K mixture to minimize the chance of DNA 621 extraction failure. A very small quantity of DNA is required for a standard PCR with robust 622 primers. To control for the completeness of the PCR mastermix, we included male samples in 623 the same batch as female samples. If a male lacked a Y chromosome, we inferred the father's 624 sperm was missing the Y chromosome (nullisomic), and if a female contained a Y chromosome, 625 we inferred the father's sperm contained both X and Y. We used R prop.test to evaluate 626 whether there were any differences between nondisjunction proportions for the different 627 strains. After finding this highly significant, we used pairwise.prop.test in R with Holm-628 Bonferroni multiple test correction to determine which pairs of substrain nondisjunction rates 629 were significantly different from each other. 630 631 Acknowledgements 632 We thank Yasir Ahmed-Braimah for discussions and use of primers. We greatly appreciate the 633 use of the Cs-137 irradiator from the Robert Weiss lab and to Amanda Loehr for operating the 634 device. Asha Jain prepared sequencing libraries for this project, and Yassi Hafezi provided 635 advice on the nondisjunction assays. We also thank members of the Clark lab for discussions 636 and encouragement on this project. 637 638