Diverse mating phenotypes impact the spread of wtf meiotic drivers in S. pombe

Meiotic drivers are genetic loci that break Mendel’s law of segregation to be transmitted into more than half of the offspring produced by a heterozygote. The success of a driver relies on outcrossing because drivers gain their advantage in heterozygotes. It is, therefore, curious that Schizosaccharomyces pombe, a species reported to rarely outcross, harbors many meiotic drivers. To address this paradox, we measured mating phenotypes in S. pombe natural isolates. We found that the propensity to inbreed varies between natural isolates and can be affected both by cell density and by the available sexual partners. Additionally, we found that the observed level of inbreeding slows, but does not prevent, the spread of a wtf meiotic driver in the absence of additional fitness costs. These analyses reveal parameters critical to understanding the evolution of S. pombe and help explain the success of meiotic drivers in this species.


Introduction 40
Mating behaviours have long been a focus of art, literature, and formal scientific inquiry. This from an estimated 20-60 outcrossing events ). This low outcrossing rate could 81 result from limited opportunities to outcross, preferential inbreeding, or a combination of the two. 82 It is important to note, however, that the outcrossing rate estimates in S. pombe are likely low as The wtf drivers destroy the meiotic products (spores) that do not inherit the driver from a 91 heterozygote. Each wtf drive gene encodes both a Wtf poison and an Wtf antidote protein that, 92 together, execute targeted spore killing of the spores that do not inherit the wtf driver. In the 93 characterized wtf4 driver, the Wtf4 poison protein assembles into toxic protein aggregates that are 94 packaged into all developing spores. The Wtf4 antidote protein co-assembles with Wtf4 poison only in 95 the spores that inherit wtf4 and likely neutralizes the poison by promoting its trafficking to the 96 vacuole (Nuckolls et al. 2020). Spore killing by wtf drivers leads to the loss of about half of the 97 spores and almost exclusive transmission (>90%) of the wtf driver from a heterozygote (Hu et  In this work, we exploited the tractability of S. pombe to better understand how meiotic drivers 103 could succeed. Despite limited genetic diversity amongst isolates, we observed natural variation 104 in inbreeding propensity and other mating phenotypes. Some natural isolates preferentially 105 inbreed in the presence of a potential outcrossing partner, whereas others mate more randomly. 106 Additionally, we found that the level of inbreeding can be altered by cell density and affected by 107 the available sexual partners. To explore the effects of these mating phenotypes on the spread 108 of a wtf driver in a population, we used both mathematical modelling and an experimental 109 evolution approach. We found that while the spread of a wtf driver could be slowed by 110 inbreeding, the driver could still spread in the absence of linked deleterious traits. We

Inbreeding propensity differs amongst S. pombe natural isolates 117
Most laboratory investigations of S. pombe utilize cells derived from 968 (CBS1042), the first 118 haploid isolated from French wine in 1921 by A. Osterwalder (Osterwalder 1924). In this work, 119 we will refer to derivatives of this isolate as "Sp." Homothallic Sp strains switch mating type in a 120 regular pattern such that a clonal population contains an equal number of cells from each 121 mating type (h+ and h-) after relatively few cell divisions (Maki et al. 2018). When starved, 122 haploid S. pombe cells of opposite mating types can mate (fuse) to form diploid zygotes, which 123 then generally proceed directly to undergo meiosis and spore formation (sporulation). Genetics 124 and microscopy experiments have revealed that the homothallic Sp haploids tend to inbreed 125 (Bendezu and Martin 2013;Egel 1977;Leupold 1949;Merlini et al. 2016;Miyata 1981). 126 However, the precise level of inbreeding when S. pombe cells are among nonclonal sexual 127 partners has not, to our knowledge, been formally reported for any isolate. 128 129 To quantify inbreeding in Sp, we first generated fluorescently tagged strains to easily observe 130 mating via microscopy ( Figure 1A). We marked strains with either GFP or mCherry (both 131 constitutively-expressed and integrated at the ura4 locus). We then mixed equal proportions of 132 GFP-expressing and mCherry-expressing haploid cells and plated them on a medium (SPA) 133 that induces mating and meiosis. We imaged the cells immediately after plating to measure the 134 starting frequency of both parent types. We then imaged again 24-48 hours later when many 135 cells in the population had mated to either form zygotes or fully developed spores. We inferred 136 the genotypes (homozygous or heterozygous) of each zygote and ascus (spore sac) based on 137 their fluorescence ( Figure 1A). Homozygotes were produced by mating of two cells carrying the 138 same fluorophore, while heterozygotes were produced by mating between a GFP-labeled and 139 an mCherry-labeled cell ( Figure 1B and C). Finally, we calculated the inbreeding coefficient (F) 140 by comparing the observed frequency of heterozygotes to the frequency expected if mating was 141 random (F=1-observed heterozygotes/expected heterozygotes; Figure 1A). Total inbreeding, 142 random mating, and total outcrossing would yield inbreeding coefficients of 1, 0, and -1, 143 respectively (Hartl and Clark 2007). 144

145
In homothallic Sp cells, we measured an average inbreeding coefficient of 0.57 using our 146 microscopy assay ( Figure 1B and 1D). In a mixed heterothallic population (GFP-expressing and 147 mCherry-expressing cells of both mating types), we observed random mating (F = -0.05), 148 consistent with mating-type switching facilitating inbreeding ( Figure 1D). To validate our 149 microscopy results, we also assayed Sp cells using an orthogonal approach employing 150 traditional genetic markers. For this analysis, we mixed haploid cells on supplemented SPA 151 medium (SPAS) to induce them to mate and undergo meiosis. We then manually genotyped the 152 progeny and used the fraction of recombinant progeny to calculate inbreeding coefficients 153 (Supplemental Figure 1A). The average inbreeding coefficients measured using the genetic 154 assay were very similar to the values we measured using the microscopy assay (0.49 for 155 homothallic and 0.05 for heterothallic cells; Supplemental Figure 1B). Together, our results 156 confirm previous observations of self-mating in homothallic Sp cells and quantify the level of 157 inbreeding (Bendezu and Martin 2013;Egel 1977). In addition, we demonstrate that our 158 fluorescence assay provides a powerful tool to observe and measure inbreeding. 159

160
We next extended our inbreeding analyses to other S. pombe natural isolates. We assayed six 161 additional isolates, FY29043, FY29022, FY28981, FY28974, FY29044 and S. kambucha (Sk), 162 using our fluorescence microscopy assay. We chose these isolates because they span the 163 known diversity of S. pombe, are homothallic, sporulated well, were non-clumping, and we were 164 able to transform them with the GFP and mCherry markers described above (Supplemental 165  Figure 2) (Jeffares et al. 2015). We found that the inbreeding propensity 166 varied significantly between the different natural isolates ( Figure 1D). FY29043 inbred similarly 167 to Sp, but other S. pombe isolates, including Sk, mated more randomly ( Figure 1D). We also 168 observed variation in mating efficiency ranging from 10% of cells mating in FY28981 to 50% of 169 cells mating in Sk ( Figure 1E). 170 171 Given that S. pombe cells are immobile, we thought that cell density could affect their propensity 172 to inbreed. To test this, we compared the inbreeding coefficient of both homothallic Sp and Sk 173 isolates at three different starting cell densities: our standard mating density (1X), high density 174 (10X), and low density (0.1X). Because crowding prevented us from assaying high-density cells 175 using our microscopy approach, we used the genetic assay for each condition. We found that 176 inbreeding was increased in both Sp and Sk isolates when cell densities were reduced 177 (Supplemental Figure 1C and 1D). This is likely because cells plated at low density tended to be 178 physically distant from potential sexual partners that were not part of the same clonally growing 179 patch of cells (Supplemental Figure 3). However, heterothallic Sp cells that cannot mate within a 180 clonal patch of cells showed near random mating at all cell densities assayed (Supplemental 181 Figure 1C). Overall, these experiments demonstrate that inbreeding propensity varies within S. 182 pombe homothallic isolates and can be affected by cell density. 183 184 Variation in mating-type switching could contribute to reduced inbreeding in Sk 185 We next used time-lapse imaging to determine the origins of the different inbreeding 186 propensities, focusing on the Sp and Sk isolates. Previous work suggested that Sk cells have 187 reduced mating-type switching efficiency, based on a Southern blot assaying the level of the 188 DNA break (DSB) that initiates switching (Singh 2002). A mutation at the mat-M imprint site was 189 proposed to be responsible for the reduced level of DSBs (Singh and Klar 2003). Since less 190 mating-type switching could lead to less inbreeding, we decided to explore this idea using time-191 lapse assays (Miyata 1981). For these assays, we tracked the fate of individual homothallic 192 founder cells plated on SPA at low density (0.25X to our standard mating density used above) to 193 quantify how many mitotic generations occurred prior to the first mating event. When the first 194 mating event occurred, we recorded the proportion of the cells present that mated (prior to the 195 appearance of cells from next mitotic generation). We also recorded the relationships between 196 the cells that did mate (i.e., sibling or non-sibling cells; Figure 2A). We did not consider cells that 197 were born in mitotic generations past the one in which mating first occurred ( Figure 2A).  (Figure 2A). In Sp cells, we observed mating amongst the clonal 204 descendants of some progenitor cells after a single mitotic division (i.e., at generation 2). By the 205 third generation, we observed mating amongst the descendants of more than half of the 206 progenitor cells. Almost all the observed mating events were between sibling cells ( Figure 2A  207 and 2B). These observations are consistent with published work (Klar 1990;Miyata 1981). 208

209
Sk cells plated at 0.25X density on SPA divided significantly more than Sp cells prior to the first 210 mating (Wilcox rank sum test; p< 0.005 Figure 2B). Sk progenitor cells most frequently started 211 mating at the fourth mitotic generation ( Figure 2B). This phenotype is consistent with less 212 mating-type switching as more generations would be required on average to produce a cell with 213 the opposite mating type (Singh 2002). In addition, many mating events were between non-214 sibling cells. This phenotype can also be explained indirectly by reduced mating-type switching. To better understand the differences between the number of cell divisions prior to mating 221 between Sp and Sk, we compared the sequence of the mating type locus in the two isolates. 222 Consistent with previous work, we found that the mating-type regions of Sp and Sk are highly 223 similar (Singh 2002). However, using a previously published mate-pair sequencing dataset we 224 discovered a ~5 kb insertion of nested Tf transposon sequences in the Sk mating type region 225 (Eickbush et al. 2019). We confirmed the presence of the insertion using PCR (Supplemental 226 Figure 4A-B). We also found evidence consistent with the same insertion in FY28981, which 227 also mates more randomly than Sp (Supplemental Figure 4B, Figure 1D). We did not, however 228 formally test if the insertion affects mating phenotypes. Even if it does have an effect, it is 229 insufficient to explain all the mating type variation we observed as FY29044 mates randomly, 230 yet it lacks the insertion ( Figure 1D and Supplemental Figure 4B). 231

232
To further explore the hypothesis that decreased mating-type switching efficiency in Sk could 233 contribute to the mating differences we observed ( Figure 2B), we carried out time-lapse 234 analyses of cells at our standard 1X cell mating density. We reasoned that at this density, any 235 given cell is likely to have a cell of opposite mating type nearby, even if mating-type switching is 236 infrequent. We again used mixed populations of GFP and mCherry-expressing cells to facilitate 237 the scoring of mating patterns ( Figure 2C, Supplemental Video 1). We found that the Sp cells 238 predominantly mated in the second and third mitotic generations and most mating events were 239 between sibling cells ( Figure 2C). 240

241
The mating behavior of Sk cells changed more dramatically between 0.25x density and the 242 higher 1X density. Whereas Sk cells tended to first mate in the fourth generation at 0.25X 243 density, at 1X density Sk cells, like Sp, generally mated in the second and third mitotic 244 generations ( Figure 2C, Supplemental video 2). Additionally, we observed significantly reduced 245 levels of mating between Sk sibling cells at 1X density relative to 0.25X (10% and 56%, 246 respectively; Figure 2E and 2B). These phenotypes are consistent with reduced mating-type 247 switching in Sk. Specifically, our data suggest that Sk cells do not need to undergo more 248 divisions before they are competent to mate. Rather, the additional divisions that occurred at 249 0.25X density in Sk could have been necessary to produce a pair of cells with opposite mating 250 types. At 1X density, additional divisions are not expected to be required as additional non-251 sibling compatible partners are available. 252

253
It is important to note that we were unable to directly measure mating-type switching. Therefore, 254 reduced switching in Sk represents a promising model that remains to be tested. Still, our 255 results demonstrate that the mating phenotypes previously measured in Sp do not apply to all S. 256 pombe isolates. Despite very little genetic diversity, S. pombe isolates maintain significant 257 natural variation in key mating phenotypes (Jeffares et al. 2017). 258

Ascus variation 260
While assaying inbreeding cytologically, we noticed that the Sk natural isolate displayed 261 tremendous diversity in ascus size and shape ( Figure 1C, Supplemental Figure 5A, 262 Supplemental movie 2). This was due to high variability in the size of the mating projections, 263 known as shmoos. Sk produced long shmoos only in response to cells of the opposite mating 264 type and not as a response to nitrogen starvation alone (Supplemental Figure 5B). The long Sk 265 shmoos motivated us to quantify asci length across all the natural isolates described above. We 266 found that most isolates generated zygotes or asci that were ~10-15 μm, similar to Sp. The 267 majority of Sk zygotes and asci also fell within this range, but ~25% of Sk zygotes and asci were 268 longer than 15 μm, with some exceeding 30 μm (Supplemental Figure 5C). We also assayed 269 zygote/ascus length in an additional natural isolate in which we were unable to quantify 270 inbreeding due to a clumping phenotype (FY29033). This isolate also showed populations of 271 long asci, like Sk (Supplemental Figure 5C). 272 273 Additionally, we occasionally noticed a fused asci phenotype in Sk (Supplemental Figure 5D). 274 Time-lapse analyses of mating patterns, described above, revealed these fused asci can result 275 from an occasional disconnect between mitotic cycles and the physical separation of cells 276 (Supplemental Figure 5E). This phenotype is reminiscent of adg1, adg2, adg3 and agn1 277 of mating between sibling cells in the mixed Sk/Sp population (9.2% compared to 9.8% in a 295 homogeneous population; Figure 2E). Amongst the Sp cells, mating between sibling cells 296 decreased significantly from 56.8% to 29.7% in the mixed mating environment ( Figure 2E We were intrigued by the idea that long shmoos (mating projections) of Sk could contribute to its 305 ability to disrupt Sp sibling mating. We were unable to address this idea directly. We did, 306 however, find that Sk/Sp matings produce significantly longer zygotes/asci than either Sk/Sk or 307 Sp/Sp matings (Supplemental Figure 6B). This was true even when we compared Sk/Sp 308 zygote/ascus length to the length of heterozygous Sk/Sk or heterozygous Sp/Sp zygotes/asci. 309 While this result does not prove that long Sk shmoos disrupt Sp sibling mating, it does show that 310 long shmoos tend to be used in these outcrossing events. and thus the frequency at which the wtf driver can act. We varied the inbreeding coefficient (F) 337 from 1 (total inbreeding) to -1 (total outcrossing). 338

339
We used the model to calculate the predicted change in the frequency of a wtf driver after only 340 one sexual generation ( Figure 3A). We also calculated the spread of a wtf driver in a population 341 from a 5% starting frequency over many generations of sexual reproduction ( Figure 3B). Under

Inbreeding and linked deleterious alleles can suppress the spread of wtf drivers 352
We next wanted to test if our predictions reflect the behavior of wtf drive alleles in a laboratory 353 population of Sp cells over many generations. To do this, we constructed an experimental 354 evolution system employing the GFP and mCherry fluorescent markers described above to 355 measure changes in allele frequencies in a population over time using cytometry. To mark drive 356 alleles, we linked the fluorescent markers with the Sk wtf4 driver and integrated the whole 357 construct at the ura4 locus in Sp (Nuckolls et al. 2017). For nondriving alleles, we used GFP or 358 mCherry integrated at the ura4 locus without a linked wtf gene. We call the non-wtf alleles 359 'empty vector. ' We started the experimental evolution populations with a defined ratio of GFP 360 and mCherry-expressing cells. We then induced a subset of the population to mate and 361 sporulate followed by collection and culturing of the progeny (spores). From these cells, we 362 remeasured GFP and mCherry frequencies using flow cytometry, and we initiated the next 363 round of mating and meiosis ( Figure 4A). 364

365
Because our experiments rely on comparing the frequency of GFP and mCherry-expressing 366 cells over time, we needed to test the fitness of the markers. We found that both fluorescent 367 markers were lost from all our experimental populations over time (Supplemental figure 8A-B). these experiments, we also assayed heterothallic populations in parallel. As described above, 403 heterothallic cells cannot switch mating type and therefore cannot inbreed. In the heterothallic 404 populations, the wtf4 driver spread significantly faster than in homothallic cells. In generations 1-405 6, the spread of wtf4 was very similar to that predicted by our model if we assumed random 406 mating. In later generations, our observations did not fit the model well. We suspect extensive 407 loss of fluorescent cells, especially those with mCherry, and the resulting decrease in population 408 size could contribute to this effect ( Figure 4D  We next wanted to further explore the effects of linked deleterious traits on the spread of wtf 414 meiotic drivers. We used the population genetic model to calculate the ability of a driver to 415 spread when tightly linked to alleles with fitness costs ranging from 0 to 0.4. We also varied the 416 inbreeding coefficient from 0 (random mating) to 1 (complete inbreeding). We found that in the 417 absence of costs, wtf drivers are predicted to spread in a population at all initial frequencies 418 greater than 0 ( Figure 5A). As described above, this spread is slowed, but not stopped, by 419 incomplete inbreeding (coefficients less than 1). When the driver is burdened by additional 420 fitness costs, it can still spread in a population. Importantly, the driver must start at a higher 421 initial frequency to spread when linked to deleterious alleles ( Figure 5A). For example, when the 422 driver is linked to a locus with a fitness cost of 0.11, like the GFP allele described above, it is 423 expected to spread in a randomly mating population if its initial frequency is 0.125 or higher. In 424 an inbreeding population, the minimal initial frequency required for the driver to spread 425 increases as the degree of inbreeding increases ( Figure 5A). 426

427
We next tested these predictions experimentally using the Sk wtf4 allele linked to GFP in 428 homothallic cells. As described above, the GFP allele is linked to an unknown deleterious trait 429 (cost 0.11). We varied the inbreeding of the population by assaying cells mated at 1X and 0.1X 430 density. As reported above, homothallic cells mated at 1X density exhibit an inbreeding 431 coefficient of 0.5 to 0.57, but that is increased to 0.8 to 0.95 by mating the cells at low (0.1X) 432 density ( Figure 1D, Supplemental Figure 1B-C). Consistent with the predictions of our model, 433 we observed in the experimental populations that the driver failed to spread when the initial 434 frequency was less than 0.25 ( Figure 5B). When the wtf4::GFP allele was found in roughly half 435 of the population, it could spread under low levels of inbreeding but decreased in frequency 436 when inbreeding was increased ( Figure 5B). Similar, but less dramatic, effects were observed at propose that Sk switched less frequently than Sp when they noticed less of the DNA break that 464 initiates switching (Singh 2002). We discovered a large, nested insertion of transposon 465 sequences in the mating type locus of Sk, and we posit that this insertion could contribute to 466 reduced DNA break formation and, potentially, decreased mating-type switching. The long 467 shmoos we observed in Sk may also contribute to more random mating in this isolate, as the 468 long shmoo may increase the available number of partners within range. 469 470 Additional previously described natural variation that we did not functionally explore may also 471 contribute to differences in inbreeding propensity in S. pombe. For example, heterothallic 472 natural isolates are predicted to exclusively outcross to isolates with the opposite mating type 473  The effect of mating-type phenotypes on the spread of wtf meiotic drivers 492 To understand the evolution of the wtf drive genes, it is not necessarily essential to understand 493 how frequently significantly diverged natural isolates mate. Instead, it is important to understand 494 how often a driver is found in a heterozygous state. This is likely a significantly higher frequency 495 than the frequency of mating between diverged isolates because of the rapid evolution of the wtf 496  Those data, along with the inbreeding coefficients measured in this study, allowed us to 513 mathematically model the spread of a wtf meiotic driver in a S. pombe population. Our modeling 514 showed that the incomplete inbreeding we observed in S. pombe could slow the spread of a wtf 515 driver. Importantly, the incomplete inbreeding observed in S. pombe does not halt the spread 516 except in cases where the driver is found in low frequencies and linked to a deleterious allele. 517 Given the tractability of S. pombe, we were also able to test the predictions of the model 518 experimentally. Overall, our experimental results were quite similar to the model's predictions 519 discussed above. This suggests that our model encompasses all critical parameters. In addition,

S. pombe as a tool to experimentally model complex drive dynamics 534
To conclude, we would like to highlight the potential usefulness of the S. pombe experimental 535 evolution approach developed for this study. With this system, we were able to observe the 536 effects of altering allele frequencies, inbreeding rate, and fitness of a driving haplotype. In the 537 future, this system could be used to experimentally explore additional questions about drive 538 systems. For example, one could experimentally model meiotic drivers that bias sex ratios by 539 linking the driver to the mating type locus in a heterothallic population. In addition, one could 540 explore the evolution of complex multi-locus drive systems employing combinations of multiple 541 wtf meiotic drivers or drivers and suppressors. This tool could lead to novel insights about 542 natural drivers, but it may also be particularly useful for exploring potential evolutionary 2020b)) to generate pSZB437 and pSZB882, respectively. We then linearized the plasmids with 556 KpnI and transformed them into S. pombe using the standard lithium acetate protocol (Schiestl 557 and Gietz 1989). We independently transformed the isolates GP50 (S. pombe), S. kambucha, 558 FY28974, FY28981, FY29022, FY29033, FY29043, and FY29044. We were unsuccessful in 559 transforming FY28969, FY29048, and FY29068. FY29033 was not included in the inbreeding 560 analyses due to its proclivity to clump. The homothallic and heterothallic strains carrying 561 mCherry or GFP were transformed using the same method. 562

563
To add Sk wtf4 to the Sp genome, we again used a ura4-integrating plasmid. To generate this 564 plasmid, we amplified Sk wtf4 from SZY13 using the oligos 688 and 686. We digested the 565 amplicon with SacI and ligated into the SacI site of pSZB332 to generate pSZB716 (Bravo 566 Nunez et al. 2020a; Bravo Nunez et al. 2020b). We then separately introduced the GFP and 567 mCherry gBlocks into the SpeI site of pSZB716 to generate pSZB904 and pSZB909, 568 respectively. We introduced the resulting plasmids into yeast as described above. 569 570

Crosses 571
We performed crosses using standard approaches (Smith 2009). We cultured each haploid 572 parent to saturation in 3 ml YEL (0.5% yeast extract, 3% dextrose, and 250 mg/L adenine, 573 histidine, leucine, lysine, and uracil) for 24 hours at 32°C. We then mixed an equal volume of

Iodine staining 584
We grew haploid isolates to saturation in 3 ml YEL overnight at 32°C. We washed the cells once 585 with ddH2O then resuspended them in an equal volume ddH2O. We then spotted 10 μL of each 586 strain onto an SPA+S plate, which we then incubated at 25°C for 4 days prior to staining with 587 iodine (VWR iodine crystals) vapor (Forsburg and Rhind 2006). 588 589

Mating-type locus assembly and PCR 590
We used mate-pair Illumina sequencing reads to assemble the mating-type locus of S.

Measuring inbreeding coefficients by microscopy 620
We mixed haploid parents (a GFP and an mCherry-expressing strain) in equal proportions on 621 SPA as described above. We then left the plate to dry for 30 minutes and then took a punch of 622 agar from the plate using a 1271E Arch Punch (General Tools, Amazon). We then inverted the 623 punch of agar into a 35 mm glass bottomed dish (No. 1.5 MatTek Corporation). We used this 624 sample to count the initial frequency of the two parental types. We then imaged a second punch 625 of agar taken from the same SPA plate after 24 hours incubation at 25°C for homothallic cells 626 and 48 hours for heterothallic cells.  with width 50 pixels, we then measured the average intensity for the GFP and mCherry 641 channels. When measuring the log10 ratio of GFP over mCherry, the mCherry homozygotes 642 have the lowest ratio, homozygotes for GFP the highest ratio, and heterozygotes intermediate.  was not included in the analysis. We recorded two videos for each isolate. 680 681

Calculation of mating efficiency 682
We calculated mating efficiency from microscopic images using the following formula: 683 (%) = 2 + 2 + 2 7 + 2 + 2 + 2 7 * 100 684 Where represents the number of zygotes, represents the number of asci, represents the 685 number of free spores and represents the number of vegetative cells (Seike and Niki 2017). 686 687

Measuring inbreeding coefficients using genetics 688
We used a high-throughput system to genotype the progeny from each cross. First, we crossed 689 two parental populations to generate spore progeny as described above. In addition to placing 690 the mixed haploid cells on SPA+S, we also diluted a subset of the mix and plated it onto 691 YEA+S. We genotyped the colonies that grew on that YEA+S plate to measure the starting 692 frequency of each parental strain in the cross. We plated the spores produced by the cross on 693 YEA+S and grew them at 32°C for 4 days. We picked the colonies using a Qpix 420 Colony 694 Picking System and cultured them in YEL for 24 hours at 30°C in 96 well round-bottom plates 695 (Axygen). We then used a Singer RoTor robot to spot the cultures to YNP dropout and YEA+S 696 drug plates and incubated them at 32°C for three days. We then imaged the plates using an 697 S&P robotics SPImager with a Canon EOS Rebel T3i camera. We analyzed each picture using 698 the subtract background feature in Fiji and assigned regions of interest (ROIs) to the 384 spots 699 where cells were pinned. We then measured the average intensity of each spot and classified 700 cells as grown or not by a heuristic threshold. We genotyped some cross progeny manually 701 using standard techniques due to robot unavailability, with indistinguishable results. 702

703
We then inferred the frequency of outcrossing based on the frequency of recombinant progeny 704 using a combination of either two or three unlinked genetic markers. If mating was random, we 705 expect the progeny to reflect Hardy-Weinberg expectations ( ! + 2 (1 − ) + (1 − ) ! = 1), 706 where ! + (1 − ) ! reflect the expected frequency of homozygotes and 2p(1-p) reflects the 707 expected frequency of outcrossing. If the parental strains inbreed to make homozygotes, they 708 can only produce offspring with the parental genotypes. If the strains outcross to generate 709 heterozygotes, they will make the parental genotypes and recombinant genotypes all in equal 710 frequencies (2 n total genotypes where n is the number of segregating markers). For our crosses 711 with three markers, we therefore expected the true 'observed' frequency of progeny produced 712 by outcrossing to be equal to the number of observed recombinants divided by 6/8. For our 713 crosses with two unlinked markers, we divided by 2/4. We then calculated the inbreeding 714 coefficient using the formula, 715 . 716 717

Zygote frequency expectation under inbreeding and different mating efficiencies 718
We calculated the expected zygote frequencies when isolates were outcrossed with different 719 isolates on SPA (see Measuring inbreeding coefficients by microscopy) using an additive model 720 that incorporated mating efficiencies and inbreeding coefficients measured from the isogenic 721 crosses. The model assumed that each strain contributes equally to inbreeding, and that they 722 do not change their own mating in response to the mating partner. 723

724
We calculated the expected frequency of homozygotes for parental strain 1 as: 725 where the inbreeding coefficient is "# and the mating efficiency is (1 − $# ) for parental strain 1 727 ( 1), considering its initial frequency . The expected heterozygote frequency is: 728 The expected fraction of homozygotes for parental strain 2 is: 730

Calculating expected allele frequencies after sexual reproduction 733
To model the expected changes in allele frequencies in a randomly mating population over time, 734 we used the equations described by Crow (Crow 1991  Simulations for the spread of a driver in Figure 3 only considered drive and inbreeding. To 739 simulate drive in fluorescent populations, the starting frequencies of each allele (i.e. ' ') were 740 determined empirically for each experiment using either traditional genetic approaches or 741 cytometry. For relative fitness values ## , #! , !! , we assigned mCherry/mCherry 742 homozygotes a fitness of ## = 1, regardless of whether they were EV/EV or wtf4/wtf4 743 homozygotes. In all but one cross (see below), we observed a fitness cost linked to the GFP 744 alleles relative to mCherry alleles during sexual reproduction, regardless of wtf4. We therefore 745 used our data (see below) to calculate the 0.11 as the fitness cost of the GFP-linked variant. 746 Because of that, we assigned a fitness value of 0.78 to GFP homozygotes, !! . We assumed 747 the fitness cost linked to GFP was codominant and thus assigned a fitness value of 0.89 for 748 GFP/mCherry homozygous for wtf4 or Empty vector, #! . For GFP/mCherry heterozygotes that 749 were also heterozygous for wtf4, we assigned a fitness of 0.46, #! . This accounts for the cost The calculation for allele frequency for a wtf meiotic driver in consecutive sexual cycles from 753 haploid populations is, 754 Where the mean fitness of the population at each generation ( ), To calculate the minimum initial frequency of driver linked to alleles with varying additional 769 fitness costs, we assumed codominance for the additional alleles ( ). Then the relative fitness of 770 heterozygotes for the allele is #! = 1 − and the relative fitness of homozygotes for the allele 771 is ## = 1 − 2 . For simplicity we also assumed complete transmission bias, = 1. The relative 772 fitness of the alternative allele was assumed to be !! = 1. A wtf meiotic driver linked to 773 deleterious allele spread under the condition that 774 Where is the initial frequency of a wtf driver linked to allele c and was the inbreeding 776 coefficient. 777 778

Measuring allele frequencies for experimental evolution analyses 779
We performed the crosses and collected spores as described above. We then started the next 780 generation by culturing 60 µL of spores from each cross in each of three different wells with 600 781 µL fresh YEL media in 96 deep-well-round-bottom plates (Axygen) and cultured for 24 hours at 782 1200 rpm at 32°C. We then transferred 60 µl from each culture to a new plate with 600 µL YEL 783 and cultured for 12-14 hours at 1200 rpm at 32°C. We then pooled the culture replicates in an 784 Eppendorf tube, spun down, and resuspended them in an equal volume of ddH20. We then took 785 100 µL of this sample to assay via cytometry (described below). We also plated 200 µL of each 786 sample on SPA+S plates and incubated at 25°C for 5 days to allow the cells to mate and 787 sporulate.   Experimental strategy to quantify inbreeding. GFP (cyan)-and mCherry (magenta)-expressing 857 cells were mixed and placed on SPA medium that induces mating and meiosis. An agar punch 858 from this plate was imaged to assess the initial frequencies of each haploid strain.  suggests the insertion is also in the FY28981 isolate. C) We aligned long Nanopore sequencing 953 from the indicated strains to the Sp reference genome mat locus, which is h-. We then 954 quantified the frequency of h+ alleles from the aligned reads. We used Fisher's exact tests to 955 compare our observations to the indicated ratios. Sk cells with a septum (black arrow) that remained uncleaved after two rounds of fission. 966 Additional septa (white arrows) are also shown that cleaved prior to mating. Scale bar 967 represents 10 µm.  Experimental strategy to monitor allele frequency through multiple generations of sexual 1006 reproduction. GFP (green) and mCherry (magenta) markers are used to follow empty vector 1007 (EV) alleles or the wtf4 meiotic driver. Starting allele frequencies and allele frequencies after 1008 each round of sexual reproduction were monitored using cytometry. B) Homothallic population 1009 with mCherry marking wtf4 and GFP marking an EV allele. Allele dynamics were predicted 1010 using drive and fitness parameters described in the text assuming inbreeding (black lines) and 1011 Homothallic Sk cells plated mating inducing media (SPA) were recorded for 24 hours. Cells 1115 labelled with constitutively -expressing fluorophores mCherry (magenta) or (GFP) were mixed in 1116 equal proportion at 1X standard cell density. 1117