Atypical meiosis can be adaptive in outcrossed S. pombe due to wtf meiotic drivers

Killer meiotic drivers are genetic parasites that destroy ‘sibling’ gametes lacking the driver allele. The fitness costs of drive can lead to selection of unlinked suppressors. This suppression could involve evolutionary tradeoffs that compromise gametogenesis and contribute to infertility. Schizosaccharomyces pombe, an organism containing numerous gamete-killing wtf drivers, offers a tractable system to test this hypothesis. Here, we demonstrate that in scenarios analogous to outcrossing, wtf drivers generate a fitness landscape in which atypical gametes, such as aneuploids and diploids, are advantageous. In this context, wtf drivers can decrease the fitness cost of mutations that disrupt meiotic fidelity and, in some circumstances, can even make such mutations beneficial. Moreover, we find that S. pombe isolates vary greatly in their ability to make haploid gametes, with some isolates generating more than 25% aneuploid or diploid gametes. This work empirically demonstrates the potential for meiotic drivers to shape the evolution of gametogenesis.

(Zanders and Unckless 2019). Due to these costs, variants that suppress meiotic drive can be cell trait, this could lead to evolutionary tradeoffs where variants that are suboptimal for some aspect of gametogenesis may be selected due to their ability to mitigate the costs of meiotic 64 drivers.
necessary to counteract the Wtf poisons ( Figure 2A). Consistent with this model, Sp/Sk heterozygotes do not make more disomic spores per meiosis than Sp or Sk homozygotes 165 (Zanders et al., 2014).

167
To test our model, we engineered an Sk diploid that is heterozygous for two unlinked sets of 168 competing wtf drivers on chromosome 3. We refer to this diploid as the "double driver  Núñez et al., 2018;Bravo Núñez et al., 2020;Nuckolls et al., 2017). Consistent with our 172 hypothesis, we found that 62% of the viable spores generated by the double driver heterozygote 173 inherited both of the parental alleles at ade6, compared to 4% of the viable spores generated by

192
These results are consistent with our hypothesis that diploids carrying multiple sets of heterozygous wtf meiotic drivers generate heterozygous disomic spores due to the destruction 194 of haploid progeny.

198
To further determine the contribution of competing wtf drivers to the high level of disomic 199 spores, we decided to test our model in a strain background with more extensive 200 heterozygosity, like those generated by outcrossing. For these experiments, we started with an 201 Sp/Sk mosaic diploid strain that is heterozygous for eight known or predicted wtf meiotic drivers 202 (Bravo Núñez et al., 2020;Eickbush et al., 2019). This mosaic diploid is homozygous for Sk 203 chromosomes 1 and 2 but is heterozygous for most of chromosome 3 for Sp and Sk-derived 204 sequences ( Figure 3A). These diploids also lack rec12, a gene which encodes the 205 endonuclease that initiates meiotic recombination by generating double-strand DNA breaks 206 (DSBs) (Bergerat et al., 1997;Keeney et al., 1997). The lack of induced recombination in these 207 diploids results in competition between the Sp and Sk wtf drivers on chromosome 3, as haploid 208 spores will generally inherit either every Sp driver or every Sk driver. To determine the

228
However, deleting only one of the six predicted Sk drivers (wtf4) had no effect (Figure 3-figure We also performed analogous experiments in the presence of meiotic recombination by mating the mosaic haploid strain to a rec12+ Sk strain. Meiotic recombination will produce 233 chromosomes with new combinations of Sp and Sk wtf drivers. Our model predicts that 234 heterozygous disomic spores will still have a fitness advantage as they are more likely to inherit were surprised to discover that amongst the G418 R Hyg R progeny, only 15.4% appeared to be 299 aneuploid and none appeared to be diploid. Instead, the majority (84.6%) of the G418 R Hyg R 300 progeny appeared to be haploid ( Figure 4C, diploid 24). We reasoned that an unequal 301 interhomolog crossover event at the ade6 locus could have led to duplication of the wtf driver 302 found on the opposite haplotype ( Figure 4-figure supplement 3). Consistent with this idea, we 303 found that the frequency of G418 R Hyg R progeny that appeared to be haploid fell in the absence 304 of recombination (rec12∆/rec12∆; Figure 4C, diploid 26). Additionally, we directly tested the 305 unequal crossover hypothesis using PCR. We amplified a potential duplication junction in 22 306 haploid G418 R Hyg R spore colonies and found that unequal crossovers moved both of the

311
Therefore, we concluded that this type of atypical meiotic product (duplications) was enriched 312 amongst the progeny of the Sk wtf4/Sk wtf28 heterozygote due to the death of spores that did 313 not inherit both wtf genes.

317
Our results demonstrate that when wtf drivers compete, such as when S. pombe outcrosses, 318 the atypical spores that inherit more drivers are more fit. Disomic spores that inherit two copies 319 of chromosome 3 most likely inherit the maximal number of wtf drivers. Therefore, we 320 hypothesized that the fitness costs of decreasing the fidelity of meiotic chromosome segregation 321 might be offset by the fitness benefits of generating more disomic spores when wtf drivers 322 compete ( Figure 5A and 5B). Consistent with this idea, we previously observed that deleting 323 rec12 imposed no fitness cost on Sp/Sk heterozygotes compared to the Sp/Sp or Sk/Sk 324 homozygotes (Zanders et al., 2014) ( Figure 5C). We wanted to know if this was specific to 325 Sp/Sk heterozygous diploids or if it might apply more generally to outcrossed S. pombe strains.

326
To address this, we compared fertility in the presence and absence of Rec12 in CBS5680/Sp, These results demonstrate that the meiosis fitness optima in inbred strains differs from what is 330 optimal when strains outcross.
We reasoned that competing wtf meiotic drivers were contributing to the dispensability of rec12 333 in the outcrossed diploids. To test that idea, we assayed the fitness costs of deleting rec12 in 334 strains with heterozygous wtf drivers at one or two loci. We found that in a diploid with one set of 335 heterozygous drivers (Sk wtf4/Sk wtf28 at ade6), the cost of deleting rec12 (rec12∆/rec12∆) was 336 similar to that observed in the wild-type background (3-fold decrease in fertility). However, we 337 found that deleting rec12 in a genetic background with wtf drivers competing at both ade6 and 338 ura4 had no cost ( Figure 2C, compare diploid 15 to diploid 13). These results support our model 339 that the costs of disrupting chromosome segregation can be offset by the fitness benefits of 340 disomic gametes in the presence of wtf driver competition.

342
We next tested the fitness costs of deleting other genes that promote accurate meiotic 343 chromosomes segregation (rec10, sgo1, moa1, and rec8) in the presence and absence of  (Krawchuk et al., 1999;Watanabe and Nurse 1999;Yoon et al., 2016).

354
Deleting sgo1 and rec10 had a lower fitness cost in the background with heterozygous wtf 355 drivers than in a background without wtf competition (~3-fold decrease compared to a 6-7-fold

357
deleting moa1 or rec8 had no effect on fitness in diploids with heterozygous wtf drivers, despite 358 the fact that these mutations decrease fertility by 4-and 6-fold, respectively, in the absence of
Deleting one copy of moa1 did not significantly alter the frequency of disomic spores in a wild-the fitness costs of wtf competition ( Figure 6B, diploid 30). Deleting one copy of rec8, however, 367 significantly increased the production of disomic spores in a background without heterozygous 368 wtf drivers ( Figure 6B, diploid 27). This suggests that rec8 exhibits haploinsufficiency and 369 reducing Rec8 protein levels may lead to chromosome segregation errors during meiosis.

370
Consistent with this observation, mutations that reduce the levels of the meiotic cohesin can 371 lead to meiotic defects in mice and flies (Murdoch et al., 2013;Subramanian and Bickel 2008).

372
As hypothesized, rec8 heterozygosity also increased the fertility of diploids with competing 373 drivers ( Figure 6B, diploid 29). Overall, these results suggest that the costs of disrupting 374 chromosome segregation can be partially or totally alleviated by the increased protection 375 against wtf drivers gained by generating more heterozygous disomic spores.

377
Driver competition can facilitate the maintenance or spread of alleles that disrupt meiotic 378 chromosome segregation fidelity in a population

379
Our experiments demonstrate that the effects of meiotic mutants can be quite different in 380 heterozygous S. pombe wherein wtf drivers are competing. To explore this idea further, we 381 turned to population genetic modeling to analyze how drivers affect the evolution of variants that 382 decrease the fidelity of meiotic chromosome segregation.

384
Our model analyzes the evolutionary fate of a hypothetical mutation that disrupts the 385 segregation of chromosome 3, which houses the majority of wtf drivers. For the sake of 386 simplicity, our model assumes that chromosome 3 exhibits whole-chromosome drive. The 387 model also considers six parameters ( Figure 7A). The first two parameters relate to the wtf 388 drivers. We varied the number of driving alleles in the population (n) and the strength of their 389 drive (t ). Each driving allele was assumed to be at an equal frequency in the population and 390 have the same strength of drive. The next parameters relate to the meiotic mutation. We varied 391 the level of chromosome missegregation caused by the mutation (f ) from 0 (no mutant 392 phenotype) to 1 (50% of the resultant spores are heterozygous disomes and the remaining 50% 393 of the spores lack chromosome 3 and are thus inviable). We considered the dominance of the 394 mutation (h) and any additional fitness costs (s m ) the mutation may incur, such as potential 395 costs relating to the missegregation of other chromosomes. Finally, we considered additional 396 fitness costs (s s ) disomic spores might bear. The full description of the model and additional 397 analyses are presented in Supplementary File 1.
We found that a mutation with no fitness costs (s m and s s =0) that disrupts meiotic segregation 400 could invade a population when: where t critical is the value of drive strength necessary for such invasion in a population of n 403 drivers. Interestingly, the higher the strength of drive (t), the lower the number of drivers 404 required for mutant invasion ( Figure 7B). Importantly, our empirical work demostrates that drive 405 strength is generally high (t >0.9) and that there are ample wtf drivers (n>5) leaving parameter

413
To get a broader perspective on the potential evolutionary trajectories of the segregation fidelity   Figure 8A). We found several natural isolates produce a similar fraction of 436 disomes as the lab strain (<10%), but others produced as many as 32% heterozygous disomic 437 spores ( Figure 8B). This result is consistent with the idea that it is not strongly deleterious for S.

438
pombe strains to generate non-haploid spores at a high frequency.

487
Since wtf drivers are numerous in outcrossed S. pombe, it is very unlikely for a haploid spore to 488 inherit every driver. In this scenario, disomic spores that inherit the two different copies of 489 chromosome 3, which carries nearly every wtf gene, are most likely to inherit every driver and 490 survive. Importantly, an extra copy of chromosome 3 is the only aneuploidy tolerated in S.

491
pombe (Niwa et al., 2006). We have previously speculated that the wtf gene family specifically 492 expanded on chromosome 3 as aneuploid spores provide an avenue to mitigate the fitness amongst the natural isolates assayed in this study ( Figure 8). Strikingly, the strains with the 512 "highest" meiotic fidelity still make ~5% disomic spores, suggesting that chromosome 3 513 missegregates during the first meiotic division in one out of ten meioses.  1942). This type of bias has been hypothesized to drive the widespread rapid evolution of can even generate selective pressure to alter the timing of the first meiotic division (Akera et al., drivers are ubiquitous, and drive represents an incredibly powerful evolutionary force.

538
Appreciating how wtf genes affect S. pombe will likely provide important insights into how 539 genetic parasites can shape the evolution of meiosis in other eukaryotes.

551
To generate a rec12∆::ura4+ deletion in the CBS5680 strain background, we first made a ura4-

552
D18 mutation in SZY2111 (ade6∆::hphMX6 in CBS5680). We amplified the ura4-D18 allele from 553 SZY925 using oligos 35 and 38 and transformed it into SZY2111 to generate SZY3949. We 554 then amplified the rec12∆::ura4+ cassette from SZY580 using oligos 1194 and 1077 and 555 transformed the cassette into SZY3949 to generate SZY3995. We confirmed the rec12 deletion 556 via PCR using oligos (1120 and 1108) that bind 730 bases upstream and 224 bases 557 downstream of the deletion cassette. We generated the rec12∆::ura4+ deletion in the 558 lys1∆::kanMX4 background of the CBS5680 isolate via crosses. We generated the rec12∆ strain 559 in JB844, similarly to how we generated it in the CBS5680 strain.

561
We found it difficult to make gene deletions in many of the natural isolates used in this study.

562
We had more success, however, making mutations using integrating vectors. Because of this, 563 we used integrating vectors to generate the genetic markers used in Figure 8.

585
Deletions of the moa1, rec10, and sgo1 genes in Sp

586
We made moa1, rec10, and sgo1 gene deletions, using standard deletion cassettes and 587 transformation. To make the moa1∆::natMX4 cassette, we amplified the upstream region of 588 moa1 with oligos 1673+1187 and the downstream region with oligos 1190+1191 (or 589 1190+1674) using SZY643 as a template. We also amplified the natMX4 gene (with oligos 590 1675+1189) using pAG25 as a template (Goldstein and McCusker 1999). We then stitched all 591 the PCR fragments together using overlap PCR and transformed this fragment into SZY44 and 592 SZY643 to make strains SZY2479 and SZY2481, respectively. We confirmed the integration of 593 the deletion cassette at the moa1 locus using oligos AO638+1192, AO1112+1191, and 594 1701+1702. We also checked that the moa1 gene was not present somewhere else in the 595 genome by using two oligos (1703+1704) within moa1.

597
To generate a rec10∆::natMX4 strain, we first amplified the upstream region and the 598 downstream region of rec10 from SZY643 using oligos 1723+1724 and oligos 1727+1728, 599 respectively. We also amplified the natMX4 cassette from pAG25 using oligos 1725+1726 together and then transformed the final deletion cassette into SZY643 and SZY44 to make strains SZY2517 and SZY2519, respectively. To confirm the integration of the cassette at the To make the sgo1∆:: hphMX6 allele, we amplified the sequences upstream and downstream of 607 sgo1 from SZY643 using oligos 1224+1225 and 1228+1229. We also amplified the hphMX6 608 cassette from pAG32 using oligos 1226+1227 (

618
These oligos contained the single guide RNA (sgRNA) sequence that targets the Sp wtf4 gene.

619
We then ligated the ends together to generate pSZB570.

621
We also made a deletion cassette to knockout the Sp wtf3 and Sp wtf4 locus. We used oligos 622 574+1138 and 1139+471 to amplify the upstream and downstream sequence of the locus using 623 SZY580 as a template. We then stitched these two fragments together using overlap PCR. We

654
We supplemented the plates with more adenine, but the colonies did not grow faster. This slow 655 growth phenotype was curiously not observed in the ade-parental haploid (SZY2111).

657
Recombination frequency within the ade6 and ura4 interval

662
Determining ploidy of spore colonies

663
In the various tests to assay the ploidy of the spore colonies for Figure 4, we compared the 664 spore colony phenotypes to the following control strains: a homothallic haploid (SZY925), a 665 heterothallic haploid (SZY1180), a diploid (SZY925/SZY1180), and aneuploid (irregular colonies 666 generated by a cross between SZY1994 and SZY1770) controls. The ploidy of the strains was 667 determined by how closely a test strain resembled one of the controls in the following tests:

Spore colony morphology
To determine the morphology of the spore colonies from different diploids, we diluted the spores of each spore colony via replica plating. These images allowed us to correlate the morphology 675 of each spore colony with its genotype. For the spores that had resistance to both G418 and  Some of the G418 R Hyg R progeny we tested appeared to be haploids based on the assays 730 described above. We reasoned they could be the result of an unequal crossover putting both          Figure 1C.

1012
Each column represents the diploid assayed, which matches the diploid number in Figure 1.

1013
The second row shows the diploid number. The third row shows the SZY strain numbers of both 1014 haploid parent strains. We present all the viable spore yield values from independent assays.

1015
We calculated the p-value using the Wilcoxon test by comparing the heterozygous diploid to the 1016 homozygous parent 1 (p1) and parent 2 (p2) strains. Diploid 1 was compared to control diploids compared to control diploids 9 (p1) and 11 (p2); and diploid 7 was compared to control diploids

1032
Column 11 indicates the p-values calculated when comparing the frequency of Ade+ Hyg R 1033 progeny produced by heterozygous diploids to the frequency of Ade+ Hyg R progeny produced 1034 by both homozygous diploid parent strains. Diploid 1 was compared to control diploids 8 and 9; 1035 diploid 2 was compared to control diploids 8 and 10; diploid 3 was compared to control diploids 1036 8 and 11; diploid 4 was compared to control diploids 8 and 12; diploid 5 was compared to 1037 control diploids 9 and 10; diploid 6 was compared to control diploids 9 and 11; and diploid 7 was compared to control diploids 9 and 12. Column 12 shows the percentage of the progeny that were Ade+ Hyg S (excluding Ade+ Hyg R progeny). Column 13 indicates the p-value calculated when comparing diploids 2-4 to diploid 8 and diploids 5-7 to diploid 9. The last column shows the total number of independent diploids assayed for each cross.

1226
The raw data can be found in Figure

1295
The raw data for diploid 23 are also presented in Figure

1303
Haploid spores that only inherit one wtf allele will be killed by the poison of the wtf they did not 1304 inherit. Spores that inherit both wtf drivers due to a wtf duplication or disomy (aneuploidy or         (C) Narrow parameter space allows for stable equilibrium. Solid line represents t critical , below 1410 which the segregation mutant cannot invade. Dashed line represents t fixation , above which the mutant fixes. In the space between the lines, the equilibrium ranges from zero (solid line) to one  loci, all heterozygotes are likely to produce gametes that will be killed by the wtf poison .

1457
The strength of drive (t) ranges from zero (wtf poison does not kill any gametes) to one (wtf poison 1458 kills all gametes that do not produce the specific wtf antidote ). Therefore, the relative number of 1459 gametes produced by a heterozygous parent is 1-t and the average relative fitness of a 1460 population with n drivers all at equal frequency (P i ) and strength of drive (t) is: without drivers (Figure 7-figure supplement 1A). This "drive load" is quite severe with large t and large n. Based on our findings, natural populations show drive strength around t=0.9 and segregate for n≈5 drivers (this number varies considerably), suggesting the population fitness is 1466 0.28 relative to a population without drive (Bravo Núñez et al., 2020;Eickbush et al., 2018;Hu et al., 2017). Such a burden would select for resistance mechanisms that might include positive chromosomes with both poisons but also both antidotes and therefore would be protected.
defect in meiosis I. We also assume that this mutation only influences segregation on and are therefore inviable. The recursion for the frequency of the infidelity mutant is: 1490

1491
Invasion: For the mutant to invade, we must have ′ > (the frequency of the mutation in the 1492 next generation must be higher than in the current generation). Solving for ′ > in terms of t, 1493 we find: Surprisingly, the critical value of t allowing for invasion is independent of the degree of In an attempt to get a more intuitive feel for the path a segregation infidelity mutant might take, we plotted trajectories of such mutants using variable parameters. Figure 7C shows that (h=1) spread more rapidly than recessive ones (h=0); c) more segregating drivers (n=5) leads to and Figure 7B. Note that in this no-cost model, there is no stable internal equilibrium. All

1569
Limitations and caveats: Our model is simplistic in several ways but does suggest that a wtf-like 1570 poison-antidote system might select for mutations that disrupt "normal" chromosome 1571 segregation fidelity. The first caveat is that we assume the population is randomly mating.

1572
However, due to spatial proximity of cells, we expect that inbreeding may be common. Such

1573
inbreeding would reduce the proportion of matings between haploids with different drive 1574 chromosomes and therefore decrease the benefit of segregation infidelity. A second caveat is 1575 that we assume all drive strengths are equal and all drivers are at equal frequency in the 1576 population. This is a logical starting point but is unlikely to be the case in real populations.

1577
Increasing the variance of the frequencies of individual drive chromosomes within a single 1578 population will also reduce the proportion of matings between haploids with different drive 1579 chromosomes and will decrease the benefit of segregation infidelity. Finally, we do not allow for 1580 non-driving third chromosomes. Diploids with a wild-type and a driving chromosome would still 1581 benefit from segregation infidelity. This benefit would be due to the disomic spores generated, in 1582 which both poison and antidote would be produced, and thus the disomic spore would be saved.
However, selection for the infidelity mutation would be weaker because the drive load in the population would be less since fewer gametes are killed each generation due to drive.