S. pombe wtf drivers use dual transcriptional regulation and selective protein exclusion from spores to cause meiotic drive

Meiotic drivers bias gametogenesis to ensure their transmission into more than half the offspring of a heterozygote. In Schizosaccharomyces pombe, wtf meiotic drivers destroy the meiotic products (spores) that do not inherit the driver from a heterozygote, thereby reducing fertility. wtf drivers encode both a Wtfpoison protein and a Wtfantidote protein using alternative transcriptional start sites. Here, we analyze how the expression and localization of the Wtf proteins are regulated to achieve drive. We show that transcriptional timing and selective protein exclusion from developing spores ensure that all spores are exposed to Wtf4poison, but only the spores that inherit wtf4 receive a dose of Wtf4antidote sufficient for survival. In addition, we show that the Mei4 transcription factor, a master regulator of meiosis, controls the expression of the wtf4poison transcript. This transcriptional regulation, which includes the use of a critical meiotic transcription factor, likely complicates the universal suppression of wtf genes without concomitantly disrupting spore viability. We propose that these features contribute to the evolutionary success of the wtf drivers.


Introduction
The transmission of most eukaryotic genes follows Mendel's first law of segregation. This law stipulates that the two alleles of a heterozygous organism (e.g., A/a) segregate randomly into gametes such that each allele is transmitted to 50% of the progeny [1]. There are, however, alleles that can break Mendel's law to force their own transmission into more than half of the offspring. These lawbreaking genes are called meiotic drivers [2,3]. There is a tremendous diversity of meiotic drive genes with distinct evolutionary origins and mechanisms found throughout eukaryotes . However, the molecular details underlying how these systems are expressed and function are limited. Uncovering these details is important for understanding meiotic drive and, more broadly, has the potential to reveal novel insights about gametogenesis. For example, the existence of a sperm-autonomous phenotype, despite cytoplasmic connections between sister sperm, was discovered through study of the t-haplotype driver in mouse [14].
Meiotic drivers can generally be considered selfish or parasitic genes [29]. This is because drivers can persist in genomes due to the transmission advantages of drive, rather than due to fitness benefits they provide to the organisms that carry them. In fact, meiotic drivers often cause decreased fitness through a variety of direct and indirect mechanisms [30][31][32][33]. The fitness costs are especially deleterious amongst the class of drivers known as the killer meiotic drivers (reviewed in [34]). These drivers can achieve up to 100% transmission to viable gametes by destroying the products of meiosis that do not inherit the driver from a heterozygote.
Despite the fitness costs of killer meiotic drivers, many populations harbor these genes (reviewed in [34]). However, the fission yeast S. pombe may represent an extreme case. Different natural isolates of S. pombe contain between 4-14 predicted killer meiotic drivers from the wtf (with transposon fission yeast) gene family [35][36][37][38]. These genes kill the meiotic products (spores) that do not inherit them from a heterozygote using two proteins produced from two transcripts with largely overlapping coding sequences: a Wtf poison protein that kills spores and a Wtf antidote protein that rescues spores that inherit the driver (Fig 1A and 1B; [37,38]). After meiosis in wtf driver heterozygotes, the Wtf poison is found in all spores, while the Wtf antidote is enriched in those that inherit the wtf driver [37,39]. These distribution patterns explain how the driver has the potential to kill all spores, yet specifically rescues those that inherit the driver (Fig 1B). How this dual Wtf protein localization pattern is established and maintained is unclear.
Uncovering the mechanisms of meiotic drivers is important for understanding how these parasites impact the organisms that carry them. The wtf genes, in particular, confer no known fitness benefits and are largely responsible for the infertility observed in heterozygous S. pombe diploids generated by outcrossing between distinct isolates [40][41][42][43][44][45]. Due to the fitness costs of wtf drivers, S. pombe has likely evolved ways to tolerate or suppress drive. There are currently two described mechanisms that can suppress wtf meiotic drivers: suppression by other wtf genes and gamete disomy [39][40][41]. Transcriptional silencing operating through trans-acting regulators or cis-acting chromatin packaging are additional candidates for suppression mechanisms. While this possibility has not been explored during spore formation, histone deacetylation and TOR-mediated RNA processing have been shown to contribute to mitotic silencing of wtf genes [46][47][48][49].
Our goals in this study were to understand how the expression and localization of Wtf driver proteins are coordinated and to find possible routes to transcriptional suppression of wtf drivers. To do this, we use the wtf4 gene from the S. kambucha natural isolate of S. pombe as a model system. We define the promoter sequences that control the wtf4 poison and wtf4 antidote transcripts. We demonstrate that differential expression due to the distinct promoters and differential inclusion of the two Wtf4 proteins in developing spores both contribute to Wtf4 protein localization and thus efficient drive. We also found that the expression of the wtf4 poison transcript is controlled by the transcription factor Mei4, a master regulator of gene expression in meiosis [50,51]. In the context of our results, we discuss why evolving or maintaining transcriptional silencing of wtf genes and drivers in general may be challenging.

Wtf4 poison and Wtf4 antidote proteins have distinct expression profiles
In previous work, we used fluorescent markers to separately tag and visualize the Wtf4 proteins [37,52]. We imaged mostly mature asci (sacks holding spores) and diploid cells present at the same time points that had not completed spore formation. In diploid cells, these analyses revealed signal from a tagged Wtf4 poison -GFP allele, but not an mCherry-Wtf4 antidote allele. In mature asci, we observed Wtf4 poison -GFP in all spores and mCherry-Wtf4 antidote strongly enriched in spores that inherited the locus encoding the allele [37,52]. These analyses, however, were too coarse to reconstruct the full expression dynamics of the two proteins.
In this work, we explored the expression of the two Wtf4 proteins more thoroughly to uncover how the two proteins achieve targeted destruction of spores that do not inherit wtf4 from a heterozygote ( Fig 1B). We again used wtf4 poison -GFP, a separation-of-function allele that does not express Wtf4 antidote and adds a GFP tag to the C-terminus of the Wtf4 poison (S1 Fig allele 1). We have previously shown that this allele acts as a functional poison but is slightly less toxic than the corresponding untagged wtf4 poison allele ( [37], S2A Fig, diploid 4). We also used the mCherry-wtf4 allele, which tags Wtf4 antidote , but still generates an untagged Wtf4 poison (S1 Fig allele 2). This allele functions similarly to the wild type wtf4 in allele transmission and viable spore yield (VSY) assays ( [37], S2A Fig, diploid 3). Viable spore yield is a measure of fertility that assays the number of viable spores generated per cell induced to form spores [53]. In this work, we normalized viable spore yields to those measured in wild-type control cells with an empty vector integrated in the genome instead of a wtf4 allele. Those wild-type cells are therefore considered to have 100% fertility. Complete drive in a heterozygote is expected to reduce fertility by roughly 50% and expression of a Wtf4 poison allele in the absence of Wtf4 antidote is expected to reduce fertility more than 50%.
We integrated the mCherry-wtf4 and wtf4 poison -GFP alleles mentioned above into the ade6 locus of different haploid strains. We crossed these two strains together to generate heterozygous mCherry-wtf4 / wtf4 poison -GFP diploids. We then completed time-lapse microscopy during fission yeast meiosis and spore formation (Fig 1C and 1D). In this text, we depict Wtf4 antidote in magenta and Wtf4 poison in cyan in all the images, regardless of fluorescent protein tag. We consistently observed Wtf4 poison -GFP hours before mCherry-Wtf4 antidote ( Fig  1C and 1D). Given that GFP maturation is only minutes faster than mCherry [54][55][56], the earlier signal from Wtf4 poison -GFP cannot be explained by maturation times of the fluorescent proteins alone. Through the time-course, the GFP signal decreases, but then increases to reach high intensity prior to spore formation. The signal was found in all four spores, not just those that inherited the wtf4 poison -GFP allele (Fig 1C and 1D, S1 Video panel A). These observations suggested either that most of the Wtf4 poison protein is produced prior to spore individualization or that the wtf4 poison transcript and/or protein is freely exchanged between spores. We distinguished between these models by photobleaching two spores or the entire ascus in asci expressing Wtf4 poison -GFP and assaying recovery of fluorescence (i.e., FRAP). We found minimal recovery of Wtf4 poison -GFP signal after photobleaching both spores and asci (S3A- S3D  Fig). These data support that most Wtf4 poison is produced prior to spore formation and then packaged within all spores.
The mCherry-Wtf4 antidote signal was very low prior to spore formation and reached max intensity after spore formation (Fig 1C and 1D, S1 Video panel A). We were curious if the transcription of the wtf4 poison could interfere with the transcription of the wtf4 antidote prior to spore formation because the wtf4 poison transcriptional start site is downstream of the wtf4 antidote transcriptional start site. To test this, we generated an alternate allele of mCherry-wtf4 antidote that lacks introns, and thus lacks the capacity to produce a poison transcript (Figs 1A and S1 allele 3). This allele encodes a functional antidote as it suppresses drive of wild type wtf4, but does not cause drive alone (S2A Fig, diploids 10 and 11). In heterozygotes with this allele integrated at ade6, we again saw minimal signal from this mCherry-Wtf4 antidote protein in diploids and strong enrichment in two spores. This suggests wtf4 poison expression is not significantly affecting the gross localization dynamics of mCherry-Wtf4 antidote (S2B Fig).
We also tested if the difference in signal we observed between the tagged Wtf4 antidote and Wtf4 poison proteins could be due to differences in the N-versus C-terminal tags. To test this, we assayed a C-terminally tagged Wtf4 antidote separation-of-function allele, wtf4 antidote -mCherry. This allele does not encode Wtf4 poison because it lacks introns (Figs 1A and S1 allele 4). We integrated this allele into the ura4 locus and found that it indeed acted as an antidoteonly allele because it does not drive on its own, but it does suppress drive of wtf4 (S2A Fig, diploids 8 and 9). We then crossed a strain carrying the wtf4 antidote -mCherry allele to haploid cells carrying wtf4 poison -GFP to generate wtf4 antidote -mCherry/ura4+, wtf4 poison -GFP/ade6+ diploids. We imaged these diploids over a time course and observed that the Wtf4 antidote localization was different than the pattern we observed with the mCherry-wtf4 allele in that the cells had mCherry signal prior to spore formation ( Fig 1E and 1F, S1 Video panel B). Specifically, we saw Wtf4 antidote -mCherry in diploid cells induced to undergo meiosis and in the ascal cytoplasm (i.e., outside of spores) of immature asci (Figs 1E and 1F and S2C, S1 Video panel B). In mature asci, Wtf4 antidote -mCherry was strongly enriched in two of the four spores, similar to the N-terminally tagged protein, but significant signal remained in the ascal cytoplasm outside spores Figs 1E and 1F and S2C, S1 Video panel B). Another C-terminally tagged Wtf4 antidote separation-of-function allele, wtf4 antidote -GFP (S1 Fig allele 5), localized similarly to wtf4 antidote -mCherry (S2C and S2D Fig) and behaved as a functional antidote allele (S2A Fig, diploids 13 and 15).
We assayed expression of the Wtf4 proteins via western blots. In samples from wtf4 antidote -mCherry/ura4+, wtf4 poison -GFP/ade6+ diploids, we detected free GFP and a band we infer to be full-length Wtf4 poison -GFP when cells are undergoing meiosis and when mature spores are present. We never detect full-length Wtf4 antidote -mCherry, but we detect free mCherry when cells are undergoing meiosis and when mature spores are present (S4 Fig). In diploids expressing the mcherry-wtf4 antidote allele, we also only detect free mCherry, but only when mature spores are present, consistent with our image analyses showing mCherry only in spores (S5 Fig). Our inability to detect full-length tagged Wtf4 antidote on a western was not surprising as we previously showed the protein is quickly trafficked to vacuoles, where it is presumably degraded [52].
We also attempted to detect Wtf4 poison -GFP and Wtf4 antidote -mCherry on Western blots from samples collected from cells induced to undergo synchronous meiosis using an analog sensitive allele of the Pat1 kinase that inhibits meiotic entry (pat1.L95G; [57]). The results were similar to our results in wild-type cells, described above, except we could detect a faint band early in meiosis that may be Wtf4 antidote -mCherry, in addition to the free mCherry band observed at all timepoints (S6 Fig). Overall, the westerns support our imaging as we see signal from the tagged Wtf4 poison locus when cells are undergoing meiosis and in samples containing spores. We observe signal from both tagged Wtf4 antidote alleles in samples containing spores, but only from the C-terminally tagged allele in samples collected prior to spore formation (S5 and S6 Figs).
We conclude that the C-terminally tagged alleles reveal a Wtf4 antidote protein population present prior to spore formation that is not apparent with the N-terminally tagged alleles. The nature of the additional protein is not clear. The early Wtf4 antidote protein could be produced using the second translational start site found in exon 1 (codon 12) via leaky transcriptional scanning [58]. This second start site can be used to encode a fully functional antidote protein (S1 Fig allele 6, S2A Fig diploids 6-7, [37]). In the two N-terminally Wtf4 antidote tagged alleles discussed above, the putative shorter Wtf4 antidote protein would not be tagged, even if it was produced. It is also possible the Wtf4 antidote population revealed with the C-terminal tag is fulllength protein, but its expression prior to spore formation is disrupted by the N-terminal tag. We did not distinguish between these possibilities.
Overall, our results show that signal from the N-terminally tagged Wtf4 antidote used in our previous studies does not reflect the total Wtf4 antidote population. The C-terminally tagged alleles described here reveal additional Wtf4 antidote protein is present prior to spore formation. This interpretation is also consistent with previous long-read RNA sequencing data showing at least some transcription of wtf antidotes prior to spore formation (e.g., 0-6 hours after meiotic induction) [37,38,59]. Both alleles show that additional Wtf4 antidote protein production occurs after spore formation. In contrast, our time course analysis of Wtf4 poison -GFP production revealed that most of the Wtf4 poison protein is present prior to spore formation and the protein cannot move freely between spores.

Distinct promoters contribute to the distinct localization patterns of the Wtf4 proteins
We next wanted to explore if the distinct promoters contribute to the different localization patterns of the Wtf4 poison and Wtf4 antidote proteins in asci. To test this, we generated constructs with fluorescent proteins, mCherry and GFP, under the control of the wtf4 antidote and wtf4 poison promoters (including the 5' untranslated regions (UTRs)), respectively. For the wtf4 antidote promoter (p antidote ), we used the 285 base pairs found upstream of exon 1. This is just downstream of a nearby Tf transposon long terminal repeat (LTR) that is not necessary for Wtf antidote production ( Fig 1A; [38,52]). Previous work assaying cw9 and cw27 (two wtf drivers in the CBS5557 isolate) found that the 288 bp upstream of exon 1 was insufficient to generate a fully functional p antidote promoter [38]. Although the promoter regions of the genes are highly similar, our previous results showed that for wtf4, 285 bp upstream sequence was sufficient to promote production of a functional Wtf4 antidote [52]. This sequence is also well conserved amongst wtf genes that encode for an antidote, which also supports that this region includes the promoter of wtf4 antidote ([36-38,59,60] ; Fig 2A). To characterize the poison promoter (p poison ), we used the 230 bp sequence that makes up intron 1 of the antidote transcript ( Fig 2D). This sequence includes the transcriptional start site of the wtf4 poison transcript and is well conserved amongst wtf drivers (Fig 2D, [36][37][38]59]).
We integrated the promoter reporter constructs at the ade6 locus of different haploid strains. We then generated diploid cells heterozygous for each of the reporters individually (i.e., reporter/ade6+) and imaged them through meiotic induction and spore formation. With the p antidote -mCherry reporter, we observed low signal in diploids undergoing meiosis, but the strongest mCherry signal was observed in two out of the four spores (Fig 2B and 2C), presumably the two that inherited the reporter construct. Importantly, the observed localization of the p antidote -mCherry reporter protein supports that expression occurs both before and after spore individualization.
As mentioned above, a fully functional Wtf4 antidote protein can be made using a second ATG codon at the 12 th codon position [37]. We speculated that additional transcriptional regulatory sequences may be found in those coding sequences upstream of codon 12. To test this, we generated p antidote long -mCherry, a construct with mCherry under a wtf4 antidote promoter that also contains the first 11 codons of the wtf4 antidote coding sequence (S7B Fig). We integrated this allele at lys4 and crossed this strain to wild type to generate heterozygous diploids. In these diploids and in the asci generated via these diploids, we saw mCherry expressed from the p antidote long -mCherry reporter at a similar level to the mCherry expressed from the shorter p antidote (S7A- S7C Fig). The simplest interpretation of our data is that constructs under the p antidote promoter are expressed before and after spore formation. In addition, the pattern of reporter signal enrichment we observe suggests the transcripts and proteins produced in a spore can be retained within that spore. We did not directly explore a role of the 5' UTR in ensuring spore-specific expression, but it is possible the UTR could affect translation or transcript retention in the spore.
For the p poison -GFP reporter, we observed expression in diploid cells induced to undergo meiosis. We also observed similar GFP reporter signal among the four spores (Fig 2E and 2F). These observations are similar to the localization patterns we observe with Wtf4 poison -GFP (e.g., Fig 1C). This roughly equal distribution of p poison -GFP reporter signal in the four spores, even though only two inherited the p poison -GFP reporter, was starkly different from the twospore enrichment observed with p antidote -mCherry reporter. Assuming equal spore to spore mobility between GFP and mCherry, this further supports that the majority of the the p poison transcripts are produced prior to spore individualization. As with the p antidote , it is possible that the 5' UTR affects the translation or mobility of transcripts driven by p poison promoter, but we did not explore these possibilities.
Together, these experiments demonstrate differential gene expression patterns under the p antidote and p poison promoters. Our results also show that these different promoters largely contribute to the localization of Wtf4 poison protein within all spores and the enrichment of the Wtf4 antidote protein within the spores that inherit the wtf4 locus.

Master meiotic regulator, Mei4, controls Wtf4 poison expression
We next sought to explore transcriptional regulation of the wtf4 transcripts. Previous work noted that the transcription of multiple wtf genes was controlled by the fork-head transcription factor Mei4, as wtf transcription is decreased if mei4 is deleted and wtf gene expression is increased if Mei4 is overexpressed [61]. Mei4 controls the expression of hundreds of genes and is known as the master regulator of middle meiosis genes [50,51]. However, the prior study was performed before the discovery that wtf drivers can make two transcripts, so it was not clear which transcript Mei4 controls. Depictions and alignments of (A) the wtf4 antidote promoter and (D) the wtf4 poison promoter are shown. For the alignment of antidote promoters, we aligned the 600 base-pairs upstream of 41 predicted antidote-only alleles [36] from three different strains of S. pombe (the reference genome, S. kambucha, and FY29033, [85]). For the alignment of poison promoters, we aligned the intron 1 sequences (flanked by sequences 100 base-pairs upstream and downstream of intron 1) of 28 predicted poison-antidote wtf drivers [36] from three different strains of S. pombe (the reference genome, S. kambucha, and FY29033). The image shows the percent identity at each nucleotide position, excluding gaps. (B) Images of p antidote -mCherry/ade6+ diploid and ascus. (C) Quantification of mCherry fluorescence within p antidote -mCherry/ade6+ asci (n = 24). (E) Images of p poison -GFP/ade6+ diploid and ascus. (F) Quantification of GFP fluorescence within p poison -GFP/ade6+ asci (n = 34). All images were acquired after 3 days on sporulation media. For quantification, we assayed the fluorescence intensity within each spore and then divided that number by the total fluorescence intensity within all 4 spores. TL = transmitted light. All scale bars represent 2 μm. Images of diploids are shown at the same brightness and contrast and were imaged at the same settings as the ascus generated from diploids of the same genotype. https://doi.org/10.1371/journal.pgen.1009847.g002

Transcriptional regulation of wtf meiotic drivers
To understand which wtf transcript Mei4 controls, we first looked for the Mei4-binding motif in the wtf4 promoters. Fork-head transcription factors bind sequences containing FLEX motifs and the complete Mei4 binding motif contains a nine-base pair (GTAAACAAA) core sequence [50,51,62,63]. We found this nine-base pair Mei4 binding motif in the p poison promoter 110 base pairs upstream of the wtf4 poison translational start site ( Fig 3A). Moreover, we found this sequence was conserved amongst known wtf drivers in S. pombe ( Fig 3A) and in wtf drivers found in other Schizosaccharomyces species [64]. In contrast, the p antidote promoter does not contain a FLEX motif.
To test if Mei4 binds the p poison FLEX motif in meiotic cells, we analyzed data from a previous Chromatin-Immunoprecipitation sequencing (ChIP-seq) experiment of Mei4 during meiosis done in the lab isolate of S. pombe [50]. Due to the similarity of the wtf genes, many reads could not be uniquely assigned to wtf4 or any single wtf gene and were thus randomly assigned to matching sites. The sequence of intron 1, however, is distinct between wtf drivers that contain the p poison promoter and antidote-only wtf genes that do not contain the p poison promoter, allowing us to distinguish Mei4 binding between intact wtf drivers and other wtf genes [36,38,39]. Prior to meiosis, we observed that Mei4 binding to wtf drivers (wtf4, wtf13) was low relative to the genome average ( Fig 3B, 0 hours). After 4 hours in meiosis, when most cells are in prophase I, we saw a strong increase in reads in the p poison promoter (intron 1) of the wtf drivers ( Fig 3B, 4 hours). We did not see an increase in Mei4 binding to antidote-only wtf genes (wtf5, wtf9, wtf10, wtf16, wtf18, wtf20, wtf21, and wtf25) during meiosis ( Fig 3B). We also saw a Mei4 peak at the C-terminal region of both subsets of wtf genes. We predict this peak is due to the nearby LTRs, as we saw high Mei4 binding on LTRs independently of their association with wtfs (S8 Fig).
To further test the idea that wtf4 poison expression is controlled by Mei4, we compared two tagged alleles: a fully functional allele (wtf4-GFP, [37]) and an allele lacking the FLEX motif (wtf4 FLEXΔ -GFP; S1 Fig alleles 7 and 8). We integrated the alleles at ade6 to test function. We did not anticipate that deletion of the Mei4-binding motif would affect the expression or function of the Wtf4 antidote protein as we can delete all of intron 1 from wtf4 (as in wtf4 antidote -mCherry), without affecting Wtf4 antidote function (S1 Fig allele 4, S2A Fig diploids 8 and 9). Consistent with our expectations, the wtf4 FLEXΔ -GFP allele encodes a functional antidote as it suppresses drive of wild-type wtf4 in a wtf4 FLEXΔ -GFP /wtf4+ heterozygote ( Fig 3C, diploid  19). In addition, we saw strong GFP signal within two of the four spores in asci generated by wtf4 FLEXΔ -GFP/ade6+ diploids (Fig 3D), similar to tagged Wtf4 antidote alleles (e.g., S2C and S2D Fig). We tested the ability of the wtf4 FLEXΔ -GFP allele to encode a functional Wtf4 poison by assaying if the allele could drive in a wtf4 FLEXΔ -GFP/ade6+ heterozygote. These diploids had Mendelian allele transmission ( Fig 3C, diploid 18). As the wtf4 FLEXΔ -GFP does encode a functional antidote, we conclude that the allele does not drive because the Mei4-binding motif is essential for production of the Wtf4 poison . These data, combined with the previous genetic and biochemical data of others [50,61] support that Mei4 controls expression of the Wtf4 poison and likely the poison proteins produced by other wtf drivers.

Distinct transcriptional regulation of Wtf4 poison and Wtf4 antidote promote efficient meiotic drive
To explore the functional relevance of the differential transcriptional of wtf4 poison and wtf4 antidote , we generated alleles that expressed the antidote protein from the p poison promoter and vice versa. We used the p poison -wtf4 antidote -GFP allele (integrated at ade6), to test if the pattern of p antidote transcription was important for drive (S1 Fig allele 9). We compared the phenotype of this promoter swap allele to the functional wtf4 antidote -GFP allele (integrated at ade6) with the endogenous promoter described above (S1 Fig allele 5). We found that the p poison -wtf4 antidote -GFP was defective in meiotic drive. Specifically, the p poison -wtf4 antidote -GFP allele was unable to suppress the drive of an intact wtf4 driver in a p poison -wtf4 antidote -GFP/wtf4 heterozygote (S9A Fig diploid 22). The control allele (wtf4 antidote -GFP) fully suppressed drive of intact wtf4 (S9A There is a FLEX motif (black box, TGTTTAC in opposite orientation) located 111 to 117 nucleotides upstream of the Wtf4 poison transcriptional start site. Percent consensus nucleotide identity of intron 1 sequences of 28 predicted poison-antidote wtf drivers [36] from three different strains of S. pombe (the reference genome, S. kambucha, and FY29033). The figure shows the percent consensus identity at each nucleotide position, excluding gaps. (B) Mei4 Chip-seq data (data from [50]) showing Mei4 binding on wtf genes at both 0 hours and 4 hours after meiotic induction. On the top panel, the reads are aligned to the predicted wtf meiotic drive genes (wtf4, wtf13) and on the bottom the reads are aligned to the wtf genes that are predicted to only encode antidote proteins (wtf5, wtf9, wtf10, wtf16, wtf18, wtf20, wtf21, wtf25) in the S. pombe genome version ASM294v2. For any reads that mapped to more than one location, only a single location, chosen at random, is reported. (C) Allele transmission and fertility (assayed via viable spore yield, VSY) of diploids with the depicted genotype. Alleles listed are at ade6. The genotype column shows a cartoon depiction of the relevant genotype. The progeny phenotypes are then shown on the right. For diploids heterozygous at one locus (e.g., Diploid 1), two values are shown (top and bottom) that represent the two possible haploid genotypes. Spores exhibiting both parental phenotypes were considered diploid or aneuploid and were excluded from this table but can be found in S1 Data. The expected values assume Mendelian allele transmission. ( � = p < 0.05, NS = not significant; G-test for allele transmission, Wilcoxon test for VSY, in comparison to the empty vector control). We compared all diploids to diploid 1 as the control. The data for diploid 1 is also depicted in S2A and S9A Figs. Diploid 5 is also shown S2A Fig. (D) Images of heterozygous wtf4-GFP/ade6+ (left) and wtf4 FLEXΔ -GFP/ade6+ (right) diploids and asci acquired after 1 day on sporulation media. Wtf4-GFP and Wtf4 FLEXΔ -GFP are shown in green. Scale bar represents 2 μM. TL = transmitted light. All images are shown at the same brightness and contrast for accurate comparison.
https://doi.org/10.1371/journal.pgen.1009847.g003 Fig diploid 15). We also tested the ability of the p poison -wtf4 antidote -GFP to suppress the spore inviability caused by an allele that expresses only Wtf4 poison (mCherry-wtf4 poison ; S1 Fig allele 10). We found that the p poison -wtf4 antidote -GFP provided only minimal rescue of spore death and that the rescue was restricted to the spores that inherited the p poison -wtf4 antidote -GFP (S9A Fig compare diploids 12 and 21).
Our results led us to question why Wtf4 poison expressed from the p poison promoter was able to poison all spores, but the Wtf4 antidote expressed from the p poison promoter was not able to rescue all spores. Such a rescue was expected as we previously observed that Wtf4 poison ectopically induced in mitotic cells is neutralized by expression of Wtf4 antidote from a matching promoter [52]. To explain this discrepancy, we hypothesized that the Wtf4 antidote may be more likely than the Wtf4 poison to be excluded from spores as they individualize within the ascal cytoplasm. This hypothesis was based on our images of Wtf4 antidote proteins with C-terminal tags that tended to show more signal in the ascal cytoplasm than within all newly developed spores. There also tended to be more signal in the ascal cytoplasm than within spores that did not inherit the locus encoding the Wtf4 antidote in mature asci (Figs 1E, S2C and S2D). If the Wtf4 antidote is excluded from spores as they individualize more than Wtf4 poison , Wtf4 antidote would not be expected to protect spores well unless it was expressed in them.
To formally test our hypothesis, we measured the fraction of the Wtf4 antidote -GFP and mCherry-Wtf4 poison protein found within spores in asci generated by diploids carrying the p poison -wtf4 antidote -GFP and/or the mCherry-wtf4 poison allele (Fig 4A-4D). We found that a larger fraction of the mCherry-Wtf4 poison signal was in spores, relative to the fraction of total Wtf4 antidote -GFP signal found in spores (Fig 4D, left panel). This suggests that the Wtf4 antidote protein is excluded from spores more than the Wtf poison as spores individualize. In addition, we found that a smaller fraction of the mCherry-Wtf4 poison signal was in spores produced by diploids carrying the p poison -wtf4 antidote -GFP allele, compared to diploids not expressing a wtf4 antidote allele. This suggests the Wtf4 antidote can promote the exclusion of some Wtf4 poison from spores as they individualize.
To test if the p poison transcriptional timing was important for drive, we generated an allele with Wtf4 poison -GFP expressed by the p antidote promoter (p antidote -wtf4 poison -GFP; S1 Fig allele 11). However, we failed to transform this construct into cells not carrying a wtf4 antidote allele, despite multiple attempts and using multiple transformation protocols (standard lithium acetate and electroporation; [65,66]). We were also unable to generate strains carrying p antidote -wtf4 poison -GFP in the absence of wtf4 antidote via mutation or genetic crosses (S9A Fig diploid  24). These results suggest that the p antidote promoter may cause some Wtf4 poison expression during vegetative growth, killing cells lacking wtf4 antidote . Consistent with this, published longread sequencing data shows some transcription of wtf antidote alleles occurs in cells prior to meiotic induction [59].
We assayed the ability of the p antidote -wtf4 poison -GFP allele (at ade6) to support drive in cells heterozygous for wtf4 antidote -mCherry (at ura4). We compared the p antidote -wtf4 poison -GFP allele to a control mCherry-wtf4 poison allele. We found that the p antidote -wtf4 poison -GFP allele supported minimal drive of the wtf4 antidote -mCherry allele, while the control mCherry-wtf4 poison allele supported strong drive of wtf4 antidote -mCherry (S9A Fig, diploids 23 and 24). Importantly most of the drive of the wtf4 antidote -mCherry allele induced by p antidote -wtf4 poison -GFP stems from the poison allele killing spores that inherit it (100% drive in spores that inherit p antidote -wtf4 poison -GFP), with minimal killing of spores that do not (58% drive in spores that do not inherit p antidote -wtf4 poison -GFP). In the control cross, drive of the wtf4 antidote -mCherry allele was 100% and 91% in the spores that did and did not inherit the mCherry-wtf4 poison allele, respectively (S9A Fig, diploids 23 and 24).
Our genetic results indicate that the Wtf4 poison expressed from the p antidote promoter is defective at killing spores that do not inherit the allele encoding the Wtf4 poison . This is surprising given the high sensitivity of mitotic S. pombe cells to even low levels of Wtf4 poison we observed in previous work [52] and in this study (e.g., S9 Fig Diploid 24). We hypothesized that the Wtf4 antidote is better at preventing Wtf4 poison from entering spores when the proteins were co-expressed from the p antidote promoter than when the Wtf4 poison is expressed from its endogenous promoter. To test this possibility, we compared the amount of Wtf4 poison signal in spores produced by diploids heterozygous for both p antidote -wtf4 poison -GFP allele (at ade6) and wtf4 antidote -mCherry (at ura4) to similar control diploids in which wtf4 poison -GFP was expressed Quantification of the percentage of the total fluorescence intensity found inside of spores of mature asci produced by p poison -wtf4 antidote -GFP/ mCherry-wtf4 poison (left, n = 142), mCherry-wtf4 poison /ade6 + diploids (center, n = 123) and p poison -wtf4 antidote -GFP/ade6 + diploids (right, n = 134). p poison -Wtf4 antidote -GFP fluorescence data is depicted in magenta while mCherry-Wtf4 poison is depicted in cyan. Error bars depict the 95% confidence interval. p poison -Wtf4 antidote -GFP is shown in magenta and mCherry-Wtf4 poison is shown in cyan in merged images. TL = transmitted light. All scale bars represent 2 μm. All images were acquired after 1 day on sporulation media. Images were taken at the same settings and are shown at the same brightness and contrast for accurate comparison.

PLOS GENETICS
Transcriptional regulation of wtf meiotic drivers from its endogenous promoter (Fig 5A and 5B). Consistent with our hypothesis, we found that less of the total Wtf4 poison -GFP signal was inside of spores when the protein was expressed from the p antidote promoter, as compared to when the protein was expressed from the endogenous p poison promoter (59 vs 71%). We also noticed that less Wtf4 poison -GFP (from the Image of a p antidote -wtf4 poison -GFP/ ade6+, wtf4 antidote -mCherry/ura4+ diploid and tetratype ascus. p antidote -Wtf4 poison -GFP is shown in cyan and Wtf4 antidote -mCherry is shown in magenta in merged images. Additional images (parental ditype and nonparental ditype) can be found in S9B Fig. (B) Image of a wtf4 poison -GFP/ ade6+ diploid and ascus. Images of this strain were first presented in [37]. (C) Images of a heterozygous wtf4 antidote -mCherry/ura4+, wtf4 poison -GFP/ade6+ diploid and ascus. Wtf4 antidote -mCherry is shown in magenta and Wtf4 poison -GFP is shown in cyan in merged images. Images of this cross are also depicted in Fig 1E. (D) Quantification of the percent of total ascal Wtf4 poison -GFP fluorescence located outside spores in asci generated by the following diploids: 1. wtf4 poison -GFP/ade6+, wtf4 antidote -mCh/ura4+, 2. wtf4 poison -GFP/ade6+, and 3. p antidote -wtf4 poison -GFP/ade6+, wtf4 antidote -mCh/ura4+ ( � = p < 0.01, t-test, n> 20 asci per sample). TL = transmitted light. All scale bars represent 2 μm. All images were acquired after 2 days on sporulation media. Images were taken using the same settings. Not all images are shown at the same brightness and contrast to avoid over saturation of pixels in the brighter images.

Dual promoters with distinct regulation are key to wtf meiotic drive
In wtf4 drive, all spores are exposed to the Wtf4 poison , while only those that inherit the wtf4 allele generally receive a sufficient dose of the Wtf4 antidote to survive. Previously, we observed Wtf4 poison in meiotic cells prior to spore formation, but observed Wtf4 antidote enriched only in spores that inherited wtf4 [37]. These observations explained how all spores were poisoned and how only those that inherited the wtf4 locus were rescued by the antidote, but it raised a new puzzle of how meiotic cells survived exposure to the Wtf4 poison in the apparent absence of the Wtf4 antidote . This uncertainty was amplified by subsequent work that showed Wtf4 poison efficiently kills cells, not just spores, that do not express antidote [52]. In this work, our new data offer a solution to the puzzle in that we found there is also Wtf4 antidote expression in cells undergoing meiosis. This result is consistent with previous long-read RNA sequencing data that show a low level of transcription of wtf antidote alleles in early meiosis [59].
Our work also reveals that different transcriptional regulation of the poison and antidote promoters is largely responsible for the distribution of the Wtf4 poison to all spores and the enrichment of the Wtf4 antidote in the spores that inherit the locus, as transcriptional reporters grossly recapitulate localization patterns of tagged proteins (Fig 2). However, different features of the proteins also contribute to their localization patterns. Specifically, we found that the propensity for the two Wtf proteins to be included in developing spores was different. The Wtf4 antidote present in the cytoplasm at the time of spore encapsulation is excluded from the spores more than Wtf4 poison present at the same time (Fig 4D). In addition, our results suggest that Wtf4 antidote protein can promote exclusion of the Wtf4 poison from spores as they individualize, as a smaller fraction of the Wtf4 poison enters spores in the presence of the Wtf4 antidote (Figs 4D and 5D). This change in localization of the Wtf4 poison in the presence of the Wtf4 antidote is consistent with our previous observations that the Wtf4 proteins co-assemble and are trafficked to the vacuole [52].
Our data suggest the following model (Fig 6): the Wtf4 poison is mostly expressed prior to spore formation and then subsequently packaged in spores (e.g., Fig 1E). This likely ensures that each spore gets a lethal dose from a heterozygous zygote. Conversely, Wtf4 antidote is present prior to spore formation, perhaps to prevent the cells undergoing meiosis from succumbing to the poison (e.g., Fig 1E). The antidote that is present at the time of spore formation likely offers little protection to spores that do not inherit the driving locus because it does not prevent all Wtf4 poison from entering spores (Fig 5D) and the Wtf4 antidote is excluded from developing spores (Fig 4). We speculate that the exclusion of Wtf4 antidote from spores is due to its association with vacuoles, which are also excluded from developing spores [52,67]. Because of these factors, spores that inherit wtf4 must produce their own private supply of Wtf4 antidote , while those that do not succumb to the poison.

Difficulties in universal suppression of wtf drivers
This work also adds to our understanding of the evolutionary success of the wtf drivers within Schizosaccharomyces species [64]. In S. pombe, all sequenced isolates contain between 25 and 38 wtf genes, including 4-14 genes that are predicted or shown to be intact meiotic drivers [35][36][37][38]. Each heterozygous driver can cause the destruction of half the spores and antidote proteins generally do not protect against poisons that have distinct sequences [35,38]. Because of these features, diploid zygotes produced by crossing two diverged isolates produce very few viable spores [35,38,[43][44][45]68,69]. The incentive to evolve a suppression mechanism to suppress a wtf driver, therefore seems quite high [70].
There are likely many roads to suppression of wtf drivers. We speculate, however, that none of them have been terribly successful given the wtf gene family has been causing meiotic drive for over 100 million years [64]. The known genic suppressors of wtf drivers are all Wtf antidote proteins. Importantly, Wtf antidote proteins seem highly specific in that they neutralize only Wtf poison proteins with very similar sequences [35,38,39,52]. Because of this, Wtf antidote proteins are not expected to act as general suppressors of wtf drivers, as the genome would require a unique suppressor for each driver. A non-Wtf universal suppressor that interfaces with Wtf poison proteins might also be difficult to envision given the vast diversity of Wtf proteins, which can share as little as 30% amino acid identity within a species [35,36].
Transcriptional repression seems like a more feasible route to universal suppression of wtf drivers given the conservation of their promoters ( [36,38,60] ; Fig 2A and 2D). Previous work has identified some factors that help regulate the expression of wtf genes, however, these studies did not distinguish between wtf poison and wtf antidote transcripts [46,48]. In mitotic cells, a mutation in the histone deacetylase clr6 or inhibition of histone deacetylation with the drug Trichostatin A leads to increased wtf gene transcription [46][47][48]71]. Also in mitotic cells, wtf mRNAs can be eliminated from cells via double-stranded-directed RNA decay [72]. It is possible that these mitotic mechanisms may affect wtf gene expression during meiosis as well, but this has not been determined. In meiotic cells, Cuf2 decreases the expression of wtf genes late in meiosis [51,73]. However, this transcriptional suppression appears to have a minimal effect on drive. For example, Cuf2 decreases expression of wtf13, but even in this suppressed state, wtf13 drives into >90% of the gametes produced by a wtf13+/wtf13Δ heterozygote [39].

PLOS GENETICS
It is important to note that not all wtf genes are equally expressed, and differential expression of the wtf poison transcripts is correlated to drive strength in S. octosporus [37,38,59,64]. This differential expression could reflect an inefficient transcriptional repression system, but that has not been investigated. Still, effective transcriptional repression of all wtf drivers could be challenging for several reasons. The first challenge is the large number of wtf genes and the fact that they are not all found at the same locus. This may make it hard to simply turn them all off en masse by packaging one genomic region in heterochromatin. Even if the genes were found in one location, the genomic region housing the wtf genes would likely be under strong selection to resist such heterochromatization. Specifically, haplotypes that could avoid heterochromatization and thus better maintain drive would be selected by drive. This is because regions linked to meiotic drivers also get to enjoy the benefits of drive and thus profit from enhancing drive, rather than suppressing it [70]. This selective pressure resisting heterochromatization would also apply to the distributed wtf drivers as well. This situation is profoundly different than loci housing transposable elements as those selfish elements generally offer no evolutionary advantages to flanking sequences [2].
Another challenge in establishing suppression of wtf drivers is that partial or imperfect suppression of wtf genes may do more harm than good. For example, if silencing is incomplete in a diploid homozygous for a given driver, only one allele could be expressed, and drive would occur in a cell where no drive would occur in the absence of silencing. In addition, in some instances wtf drivers are suppressed by other wtf genes [35,39]. This is an additional situation in which incomplete silencing of the wrong wtf gene could lead to more, rather than less, drive.
An ideal suppression approach would be to suppress a key transcription factor driving expression from p poison promoters. This approach would allow the cell to stop drive, but it also comes with minimal risk as partial suppression would still be selectively advantageous by decreasing the likelihood of drive. In this scenario, an imperfect system could adapt and improve over time. In this work, we identified Mei4 as the key transcription factor governing the expression of p poison (Fig 3). Suppression of Mei4, however, is not an ideal option for suppressing drive because it is an essential regulator of meiosis. Mei4 controls the expression of over 100 genes and cells lacking Mei4 fail to complete meiosis [50,51,62,63,[74][75][76][77].

Final speculation
It is hard not to admire how beautifully adapted the wtf parasites are at promoting themselves despite the costs. Equally remarkable is how well insulated wtf parasites appear to be against their host genomes acting to stop their expression. That is not to say, however, that Schizosaccharomycetes have not found other ways to mitigate the costs of wtf drivers. Some S. pombe natural isolates preferentially inbreed and some isolates generate disomic (aneuploid or diploid) spores at high frequencies (up to 46% of spores; [41,78]). Although it is hard to say if the wtf genes promoted the evolution of these traits, it is possible, as both traits effectively suppress fitness costs of wtf drivers [41,78]. If these traits were selected for based on their ability to suppress drive, suppression likely came at a steep price to fitness, given that inbreeding and meiotic disruption are generally not good for fitness. Perhaps this was the price that had to be paid, as cheaper options, like universal transcriptional silencing, were out of reach.

Generation of yeast strains
All yeast strains, plasmids, and oligos are described in S1-S3 Tables. We confirmed the plasmids generated in this section via Sanger sequencing. We completed all transformations described in this section using standard lithium acetate protocol [65], first selecting for drug resistance and then screening for the relevant auxotrophy.
Generation of an allele of wtf4 with endogenous promoters and mCherry-wtf4 poison . We ordered a gBlock from Integrated DNA technologies (IDT, Coralville IA) containing the p antidote promoter (-600 bp), exon 1, intron 1, and sequence encoding an mCherry tag and a 5X glycine linker upstream of the wtf4 poison start codon [79]. The gBlock also included 177 bp of exon 2. We amplified this gBlock using oligos 620 and 718. We also amplified the rest of the wtf4 sequence from pSZB189 using oligos 679 and 687 [37]. We then used overlap PCR to stitch the PCR fragments together using oligos 620 and 687. Next, we digested the resulting PCR product with SacI and ligated it into the SacI site of pSZB188 [37], an ade6 integrating vector with a kanMX4 cassette, to create pSZB250. We cut pSZB250 with KpnI and transformed the linearized vector into SZY643 to generate SZY1037.
Generation of a wtf4 allele that encodes only mCherry-Wtf4 poison under the control of the endogenous promoter. This allele has the full-length wtf4 gene (including p antidote promoter) but the two start codons in exon 1 (for the Wtf4 antidote ) are mutated to TAG. We first cloned pSZB259 which contains an untagged wtf4 allele in which both start codons are mutated, similar to pSZB257 [37]. We amplified the p antidote promoter and the mutated version of exon 1 from pSZB259 using oligos 620 and 861. We amplified the mCherry-wtf4 poison sequence from pSZB250 (described above) using oligos 862 and 687. We then used overlap PCR to stitch the two PCR products together using oligos 620 and 687. We digested the resulting PCR product with SacI and ligated the cassette into the SacI site of pSZB188 [37] to make pSZB355. We also digested pSZB355 with SacI to isolate the wtf4 allele and cloned it into SacIdigested pSZB386, a hygromycin B-resistant ade6 integrating vector [37] to generate pSZB824. We digested pSZB824 with KpnI to linearize the construct and transformed into SZY2572 to generate SZY4570.
Generation of a wtf4 allele that encodes only Wtf4 antidote -GFP under the control of the endogenous promoter. We used pSZB203 as a template to amplify the 5' end of wtf4-GFP with oligos 620 and 736 and the rest of the gene with oligos 735 and 634 [37]. The two PCR products were then stitched together using overlap PCR using oligos 620 and 634. The 735 and 736 oligos mutated the start codon for the wtf4 poison to TAC. The resulting PCR product was then digested with SacI and ligated into the SacI site of pSZB188 to generate pSZB260. We cut pSZB260 with KpnI and integrated it into the ade6 locus of SZY643 to create SZY1056.
Generation of a wtf4 allele that encodes only mCherry-Wtf4 antidote under the control of the endogenous promoter. We used site-directed mutagenesis to mutate the translational start site from ATG to TAC of wtf4 poison within pSZB248 [37] to generate pSZB367. We linearized pSZB367 with KpnI and integrated it into SZY643 to make SZY3586.
Generation of a wtf4 allele with the first antidote start site mutated to make the wtf4 short allele. This allele has the full length wtf4 gene, with the first antidote start site (ATG) mutated to TAG. Using pSZB189 as a template, we amplified the 5' end of wtf4 with the mutated start site using oligos 620 and 702. We then amplified the rest of the wtf4 gene using oligos 701 and 686 with pSZB189 as the template. The two PCR products were then stitched together using overlap PCR using oligos 620 and 686 and ligated into the SacI site of pSZB188 to generate pSZB244. We cut pSZB244 with KpnI and integrated it into the ade6 locus of SZY643 to create SZY1026.
Generation of a p antidote -mCherry transcriptional reporter. We first amplified the wtf4 antidote promoter linked to mCherry using pSZB248 [37] as a template with oligos 688 and 1447. Next, we amplified the ADH1 transcriptional terminator from pKT127 [80] using oligos 1448 and 634. We then used overlap PCR to unite these fragments and generate p antidote -mCherry. We then digested the complete p antidote -mCherry cassette with SacI and ligated it into the SacI site of pSZB386 [39] to create pSZB766. We cut pSZB766 with KpnI and integrated it into the ade6 locus of SZY2080 to create SZY2137. We also digested the cassette form pSZB766 and cloned it into SacI-digested pSZB331 [35] to create pSZB744. We cut pSZB744 with KpnI and integrated it into the ura4 locus of SZY2080 to create SZY4534.
Generation of a p antidote long -mCherry transcriptional reporter. We amplified the majority of the p antidote long -mCherry construct (with an ADH1 transcriptional terminator) using oligos 3024 and 634 and pSZB891 [52] as a template. We then added the rest of the upstream to the p antidote long -mCherry construct by using it as a template for PCR with oligos 3025 and 634. This generated the full p antidote long -mCherry construct. We digested this cassette with SacI and ligated it into the SacI site of pSZB322 [39] to create pSZB1361. We cut pSZB1361 with KpnI and integrated into the lys4 locus of SZY2080 [52] to generate SZY4442.
Generation of a strain with wtf4::hphMX6 at the ade6 locus. We digested pAG32 (Goldstein and McCuster, 1999) with NotI to isolate the hphMX6 cassette. We transformed the hphMX6 cassette into SZY887 [37] selecting for HYG resistance and lack of G418 resistance. This created SZY969.
Generation of a p poison -GFP transcriptional reporter. We used pSZB203 as a template to amplify the p poison promoter using oligos 1174 and 1549, and to amplify GFP (with an ADH1 transcriptional terminator) using oligos 1548 and 634 [37]. We used overlap PCR to join these two pieces using oligos 1174 and 634. We then digested the complete p poison -GFP construct with SacI and cloned it into SacI-digested pSZB386 to generate pSZB821. We linearized pSZB821 with KpnI and integrated it into SZY643 to make SZY2279.
Generation of an allele of wtf4 antidote -GFP in which the p antidote promoter is replaced with the p poison promoter. We amplified the poison promoter and exon 1 from pSZB553 (see above) using oligos 1174 and 604. We amplified the rest of the wtf4 coding sequence from pSZB700 [52] using oligos 605 and 997. We amplified GFP (with an ADH1 transcriptional terminator) from pSZB203 [37] using oligos 998 and 634. We used overlap PCR to join the three pieces using oligos 1174 and 634.We digested the complete p poison -wtf4 antidote -GFP construct with SacI and cloned it into SacI-digested pSZB386 to generate pSZB727. We digested pSZB727 with KpnI to linearize the construct and transformed into SZY44 to generate SZY2406.
Generation of the p antidote -wtf4 poison -GFP allele. We amplified p antidote from pSZB203 using oligos 688 and 1383 [37]. We amplified the wtf4 poison coding sequence from pSZB392 [52] using oligos 1384 and 997. We amplified GFP (with an ADH1 transcriptional terminator) from pSZB203 [37] using oligos 998 and 634. We used overlap PCR to join the three pieces using oligos 688 and 634. We digested the complete p antidote -wtf4 poison -GFP construct with SacI and cloned it into SacI-digested pSZB386 to generate pSZB758. We digested pSZB758 with KpnI to linearize the construct and transformed into SZY2572 to generate SZY4525. The p antidote -wtf4 poison -GFP allele lacks exon 1 and thus does not encode Wtf4 antidote . Generation the wtf4 FLEXΔ -GFP allele. We used site-directed mutagenesis to delete the FLEX motif (TTTGTTTAC, [62,63]) within intron 1 of the wtf4-GFP allele in pSZB203 [37]. This generated pSZB848. We digested pSZB848 with KpnI to linearize the construct and transformed into SZY643 to generate SZY1479.
Generation of an ade6+::his5Δ allele. This was completed in the same manner as in [37]. Briefly, we amplified from genomic DNA a region upstream of his5 using oligos 795 and 796, a region downstream of his5 using oligos 797 and 798, and ade6+ using oligos 799 and 800. We stitched these pieces together using overlap PCR and oligos 795 and 798. We then transformed the cassette into SZY631 [39], selecting for ade6+ and then screening for his5-. This generated SZY1285 used in this work.
Generation of a diploid with pat1.L95G, wtf4 antidote -mCherry, wtf4 poison -GFP. We generated h-/h-diploids (SZY5742 and SZY5743) by protoplast fusion of SZY5685 and SZY5638 as previously described [81], with the following changes: We grew 50 mL YEL cultures of each strain to mid-logarithmic phase (~5x10 6 cells/ml) before washing in 10 mL 0.65 M KCl. We then prepared protoplasts by incubating cell pellets in 0.65 M KCl containing 0.1 g/mL Lallzyme MMX (Scott Labs) for 14 minutes. We then plated the protoplasts to YNP -adenine, -histidine, -leucine, -uracil and incubated at 32˚C for 3 days. We then replica plated the colonies and imaged the cells to confirm genotypes and ploidy.

Allele transmission and viable spore yield
All allele transmission and viable spore yield (VSY) assays were completed as previously described [37,53]. Briefly, we generated stable diploids by mixing each haploid parent in a microcentrifuge tube and plating them on SPA (1% glucose, 7.3 mM KH 2 PO 4 , vitamins, agar) for~15 h at room temperature to allow the cells to mate. We scraped the mated cells off of SPA and spread on a medium to select for heterozygous diploids (minimal yeast nitrogen base plates). We grew diploid colonies overnight in 5 mL of rich YEL broth (0.5% yeast extract, 3% glucose, 250 mg/L of adenine, lysine, histidine, and uracil). We then plated~100 mL onto SPA to induce sporulation, as well as diluted samples onto YEA (same as YEL, but with agar). We confirmed the colonies that grew on the YEA plate were truly heterozygous diploid cells by replicating to diagnostic media and counted the colonies for VSY. After 3 days, we scraped the cells from the SPA plates, treated it with glusulase (Sigma (G7017-10ML) and ethanol to isolate spores, and plated dilutions of the spores on YEA [53]. We then counted and phenotyped the spore colonies using standard approaches. At least 2 diploids were assayed per cross and at least 150 spores were genotyped for allele transmission assays. The raw data can be found in the S1 Data file.

Alignment of promoters
For the alignment of antidote promoters (Fig 2A), we aligned the 800 base-pairs upstream of 41 predicted antidote-only alleles [36] from three different strains of S. pombe (the reference genome, S. kambucha, and FY29033, Lock et al., 2018). For the alignment of poison promoters (Fig 2D), we aligned the intron 1 sequences (flanked by sequences 100bp upstream and downstream intron 1) of 28 predicted poison-antidote wtf drivers [36] from three different strains of S. pombe (the reference genome, S. kambucha, and FY29033). We utilized Geneious (version 11.0.14.1, https://www.geneious.com) using the Geneious aligner with the "global alignment without free end gaps" setting, taking the percent identity at each nucleotide position.

Protein analysis with western blots
To isolate proteins for pat1+ western blots (S4 and S5 Figs), we grew diploids~20 hours to saturation in 10 ml YEL at 32˚C. We then diluted cells 1:200 in 100 mL PM media and again grew at 32˚C overnight (20-24 hours). We then washed the cells once in PM-N and then diluted to an O.D. 600 of 1.0 (~1x10 7 cells/ml) in 500 ml PM-N incubated at 25˚C. We prepared whole cell extracts at the time-points indicated in figure legends from 100 mL aliquots as described in [82] with the following addition. To block protein degradation, we added 1 mM PMSF to each time-point prior to spinning down cells. The cell lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.5% NP-40, 10% Glycerol) we used was supplemented with 1 mM PMSF and 1 complete Mini, EDTA-free protease inhibitor cocktail tablet per 5 mL. We disrupted the cells using a Mini Bead beater (Biospec Products) and we imaged cells to ensure spores were broken.
To prepare positive controls expressing constitutive GFP or mCherry, we grew strains SZY2636 (GFP) and SZY2638 (mCherry) [78] to saturation in YEL, then pelleted a 1 mL volume, washed the cells in ddH 2 0, resuspended in 1 mL ddH 2 0 and plated 200 μl onto each of three SPA+S plates. We then incubated the plates at 25˚C for three days and then scraped cells into 1 mL ice-cold PBS and prepped whole cell extracts as described above. We imaged the cells to ensure sporulation using a Zeiss Observer.Z1 wide-field microscope with a 63x 1.4 Oil DICII objective and collected the emission onto a Hamamatsu ORCA Flash 4.0 using μManager software. We used BP 440-470 nm to excite GFP and collected BP 525-550 emission using a FT 495 dichroic, and mCherry with BP 530-585 nm excitation and LP 615 emission, using an FT 600 dichroic filter.
To run western blots, we diluted samples 1:1 in LDS sample buffer (Life Technologies) and heated them for 10 min at 75˚C before loading. We then ran the proteins on NuPAGE 4-12% Bis-Tris protein gels (Life Technologies) in 1 x MOPS buffer (Life Technologies) for 50 minutes at 200V and then transferred to PVDF membranes (Bio-RAD #1704156) using the Trans-Blot Turbo system. We probed the membranes with either a rabbit monoclonal, α-GFP antibody (Cell Signaling Technology #2956) at 1:1000 and/or a mouse monoclonal, α-mCherry antibody (Millipore, MAB131873) at 1:1000, overnight at 4˚C with agitation in Odyssey blocking buffer (TBS, from LI-COR biosciences). Secondary, α-rabbit (800 cW) and/or α-mouse antibodies (680 cW) were used for fluorescent visualization of the proteins. We imaged the blots on the Odyssey-CLx (LI-COR biosciences). To show protein loading we either stained membranes with Ponceau S (Cell Signaling Technology #59803) or ran a second gel with the same volume of sample and stained with Imperial protein stain (Thermo Scientific #24615).

Synchronization of pat1.L95G cultures
To generate synchronous meiotic cultures (S6 Fig), we grew h-/h-pat1.L95G/pat1.L95G analog-sensitive diploid strains (SZY5742 and SZY5743) for 20 hours at 30˚C in 10 mL YEL media. We then diluted the cells 1:250 in 50 mL EMM supplemented with 75 μg/ml lysine and grew them for 20-24 hours at 30˚C. We then washed the cells twice in EMM-N and diluted to an O.D. 600 of 1.0 in EMM-N supplemented with 10 μg/ml lysine. After 12 hours at 30˚C, we added 50 μg/ml lysine, 500 μg/ml NH4Cl and 25 μM 3MB-PP1 (Cayman Chemicals) to each culture, and then sampled cells for DAPI staining, imaging and Westerns at the timepoints described in the figures.

DAPI staining
We pelleted a 1 ml volume of each culture at the indicated timepoints and resuspended the cells in 1 ml 70% ethanol. We then incubated the cells at room temperature for one hour and then washed once in 1 ml PBS. We then immediately visualized cells using a Zeiss Observer.Z1 wide-field microscope with a 63x 1.4 Oil DICII objective and collected the emission onto a Hamamatsu ORCA Flash 4.0 using μManager software. We excited DAPI at 365 nm and fluorescence was collected through a 445/50 nm bandpass filter.

Fluorescence microscopy
The raw data for all image quantification is presented in the S2 Data. We generated diploids as previously described [37] and placed them on sporulation agar (SPA, 1% glucose, 7.3 mM KH 2 PO 4 , vitamins, agar) for 1-3 days. We then scraped the cells off the plates and onto slides for imaging. For the pat1.L95G time course experiment shown in S6 Fig, we fixed the cells in 4% paraformaldehyde for one hour at room temperature, washed three times in 1 mL PBS and stored the cells at 4˚C prior to imaging. For all microscopy, except for the experiments listed below, we used an LSM-780 (Zeiss) microscope, with a 40x C-Apochromat water-immersion objective (NA 1.2), in photon-counting channel mode with 488 and 561 nm excitation. We collected GFP fluorescence through a 481-552 bandpass filter and mCherry through a 572 long-pass filter. We also used photon-counting lambda mode, with 488 and 561 nm excitation, collecting fluorescence emission over the entire visible range. We then used these images to linearly unmix the fluorescence spectra using an in-house custom written plugin for ImageJ (https://imagej.nih.gov/ij/) to verify that there was no auto-fluorescence in the cells. Brightness and contrast are not the same for all images. We utilized two independent progenitor diploids and assayed at least 25 asci for each genotype represented. We called an ascus "mature" if all four spores showed distinct, dark outlines via transmitted light, suggesting spore membranes had been formed. For images where fluorescence intensity inside of spores is compared to fluorescence intensity outside of spores, great care was taken in acquiring the data so that only asci where all spores were sharply in focus were used for comparison.
For the wtf4 antidote -mCherryh/ura4+ (S2C Fig) we imaged using a Zeiss Observer.Z1 widefield microscope with a 40x C-Apochromat (1.2 NA) water-immersion objective and collected the emission onto a Hamamatsu ORCA Flash 4.0 using μManager software. We used BP 440-470 nm to excite GFP and collected BP 525-550 emission using a FT 495 dichroic, and mCherry with BP 530-585 nm excitation and LP 615 emission, using an FT 600 dichroic filter, and mCherry with BP 530-585 nm excitation and LP 615 emission, using an FT 600 dichroic filter. Brightness and contrast are not the same for all images.
For the gametogenesis time-lapse imaging (Fig 1C-1F), we crossed a haploid Sp strain carrying mCherry-wtf4 (SZY1142) to one with wtf4 poison -GFP (SZY5047) to generate heterozygous diploids as previously reported [37]. We also crossed a haploid Sp strain carrying wtf4 antidote -mCherry (SZY2572) to one with wtf4 poison -GFP (SZY5047) to generate heterozygous diploids using the same method as above. We grew these diploids to saturation in 5 mLs of rich YEL broth (0.5% yeast extract, 3% glucose, 250 mg/L of adenine, lysine, histidine, leucine, and uracil) overnight at 32˚C. We then diluted 100 μL of these diploid cultures into 5 mLs of PM media (20 mLs of 50x EMM salts, 20 ml 0.4 M Na 2 HPO 4 , 25 mL 20% NH 4 Cl, 1 mL 1000x vitamins, 100 μL 10,000x mineral stock solution, 3 g potassium hydrogen phthalate, 950 mL ddH 2 O, 25 mL of sterile 40% glucose after autoclaving, supplemented with 250 mg/L uracil) and again grew overnight at 32˚C. The next day, we spun to pellet and resuspended the pellet in PM-N media (PM without NH 4 Cl). We shook the PM-N cultures for 4 h at 28˚C. We loaded the plate with PM-N media after washing and flushing the microfluidic plate with PBS. Diploids were trapped in an EMD Millipore CellASIC (Haploid budding yeast) microfluidic plate. PM-N media was flowed through the plate for the duration of the movie. The temperature was maintained through the CellASIC manifold at 25˚C. GFP and mCherry fluorescence were excited with a Spectra III illuminator with a 475 nm and 555 nm dichroic, respectively. The fluorescence emission was then collected with a 515/30 nm emission filter for GFP and a 595/ 31 nm emission filter for mCherry. The fluorescence was recorded by a Prime 95B sCMOS camera (Photometrics). Images from multiple positions were recorded every 10 minutes for 24 hours. Each experiment had two biological replicates of each diploid and was imaged at 4 XY locations analyzed in parallel. We repeated the experiments with both diploids twice. The total number of diploid cells analyzed were n = 25 for mCherry-wtf4/wtf4 poison -GFP diploids and n = 13 for wtf4 antidote -mcherry/wtf4 poison -GFP diploids.
Analysis of these data was performed in Fiji (https://imagej.net/software/fiji/) using a few different plugins written in house (https://research.stowers.org/imagejplugins/). First stage drift was eliminated with "stackregj". The data was then smoothed with a 1-pixel Gaussian blur. Next, the background was subtracted with a 50-pixel rolling ball. After this, ROIs were drawn by hand around each cell of interest. The average intensity over time was plotted using "create spectrum jru v1". Time traces for GFP and mCherry for each cell were recorded separately. Next, each time trace was aligned in time manually by marking the first frame (from transmitted light only) that spores started to appear and setting that frame to time zero with "set multi plot offsets jru v1". Then, all time traces for a particular strain and for GFP and mCherry only were combined into one plot with "combine trajectories jru v1". These plots were then normalized to the minimum and maximum using "normalize trajectories jru v1". The traces were then imported to Prism 9 (https://www.graphpad.com/scientific-software/ prism/) using the XY graph preset, with the mean and 95% confidence interval plotted against time. The time scale for the plots was recalculated using the frame rate of the time lapse.
For the Fluorescence Recovery After Photobleaching experiments, cells were imaged on a Ti2-E (Nikon) microscope coupled to a CSU-W1 Spinning Disc (Yokogawa). GFP and mCherry were laser excited at 488 nm and 561 nm, respectively, through a 60x (spore FRAP NA 1.4) or 100x (whole ascus FRAP NA 1.45) Plan Apochromat objective. The fluorescence emission of GFP was collected through a 525/36 nm filter, while the emission of mCherry was collected through a 605/25 nm filter. Both signals were collected on a Prime 95B camera (Photometrics).
For the spore and whole ascus FRAP experiments, an initial frame of fluorescence was recorded in each channel. Then, many spores or asci were bleached to background in the same large field of view, using 100% power from the 450nm, 550nm and 640nm laser lines. Once bleaching was completed, the recovery was recorded every 10 minutes for~6 hours total time (whole FRAP) or every 5 minutes for 4 hours total time (spore FRAP). To generate the recovery curves, using Fiji, the average fluorescence intensity was quantified inside all spores containing GFP signal and all spores containing mCherry signal per frame, separately. Once the curves were obtained, all curves of the same fluorophore were normalized to the min and max intensity values and averaged to yield the final curves.
To quantify the amount of fluorescence intensity inside and outside spores, asci were imaged on an LSM 780 (Zeiss) in photon counting mode with a 40x LD C-Apochromat Objective (NA 1.1). GFP and mCherry were excited at 488 and 561nm, respectively. GFP fluorescence was collected through a 482-553 nm bandpass filter and mCherry fluorescence was collected through a 572-735 nm bandpass filter. Single z slices were collected with transmitted light. Then for each ascus, an ROI was drawn around the entire ascus, and around each of the 4 spores individually (drawn on the transmitted light without input from the fluorescence). This occurred either manually or via Cellpose (https://www.cellpose.org/) with manual editing of the auto generated ROIs. Then the sum-total intensity of the ascus in each channel was measured. Then the sum total intensity of each spore in each channel was also measured. For each channel, the 4 spores' intensities were added together and divided by the total intensity to yield the fractional spore intensity. This was done for each channel individually. The results of many fractional spore intensities were plotted together.

Analysis of published Mei4 Chip-Seq data
FASTQ data from [50] was retrieved from the European Nucleotide Archive (ENA), with accession number: ERP001894. Data was trimmed for quality using Trimmomatic [83] and was aligned to S. pombe genome version ASM294v2 with bowtie2 [84] using default parameters. Peaks were called using MACS2 with callpeak -gsize 1.38 e7 and -q 0.05. Metagene plots were constructed by averaging the RPM (Reads Per Million) signal for individual genes (either the predicted wtf meiotic drive genes (wtf4, wtf13, top) or the wtf genes that are predicted to only encode antidote proteins (wtf5, wtf9, wtf10, wtf16, wtf18, wtf20, wtf21, wtf25, bottom)) across 400 bins and then averaging multiple genes to create a single profile followed by loess smoothing with a span of 0.05. For any reads that mapped to more than one location, only a single location, chosen at random, is reported.
Supporting information S1 Fig. Representation of alleles used in this study. Each allele is represented as a cartoon with the allele numbers referenced through the text. The depictions describe the promoters (orange for p antidote and yellow for p poison ), the translational start site of the proteins (black arrows), fluorescent tags (mCherry or GFP), translational start site mutations (red stars) and each protein produced. Wtf4 antidote is represented as "A" and Wtf4 poison is represented as "P". We also depict the FLEX motif (black box) found within p poison . These depictions are also present in figures where each of these alleles is used. Spores exhibiting both parental phenotypes were considered diploid or aneuploid and were excluded from this table but can be found in S1 Data. The expected values assume Mendelian allele transmission. We used the viable spore yield assay (VSY) to quantify fertility with values normalized to the relevant empty vector control ( � = p < 0.05, NS = not significant; G-test for allele transmission, Wilcoxon test for VSY, in comparison to the empty vector control). We compared diploids 3, 4, 5, 6, 8, 10, 12, 13 to control diploid 1 and diploids 7,9,11,14,15 to control diploid 2. The data for diploids 1, 2, and 12-15 are also depicted in S10A Fig and the data for diploids 1 and 5 are also depicted in Fig 3C. The data from diploids 3, 4, and 14 were previously published in [37]. (B) Images of a heterozygous mCherry-wtf4 antidote /ade6+ diploid cell and mature ascus. mCherry-Wtf4 antidote is shown in magenta in merged images. (C) Images of a heterozygous wtf4 antidote -mCherry/ura4+ diploid cell and mature ascus. Wtf4 antidote -mCherry is shown in magenta. (D) Images of a heterozygous wtf4 antidote -GFP/ade6+ diploid cell and mature ascus. Wtf4 antidote -GFP is shown in magenta. All Images were taken after 3 days on sporulation media. TL = transmitted light. All scale bars represent 2 μm. Images were taken with the same settings. Not all images are shown at the same brightness and contrast to avoid over saturation of pixels in the brighter images.  (B). The images of the two different p antidote -mCherry reporters / + were from heterozygous diploids and asci. (C) Quantification of mCherry fluorescence within heterozygous (p antidote -mCherry reporter/ +) diploids and asci. At least 25 diploids and 25 asci were quantified per reporter. All images were acquired after 3 days on sporulation media. TL = transmitted light. All scale bars represent 2 μm. Images were taken at the same settings and are shown at the same brightness and contrast for accurate comparison. (TIF) S8 Fig. Mei4 associates with LTRs in meiosis. Peaks (called by MACS2) from Mei4 ChIP-seq (data from [50]) 4 hours past meiotic induction are compared to LTR regions (identified by BLAST). If the regions overlap, they are shown in the shared region of the Venn diagram. For any ChIP-seq reads that mapped to more than one location, only a single location, chosen at random, was selected.  Fig 3C. (B) Images of p antidote -wtf4 poison -GFP/ade6+, wtf4 antidote -mCherry/ura4+ parental and non-parental ditype asci. p antidote -Wtf4 poison -GFP is shown in cyan and Wtf4 antidote -mCherry is shown in magenta in merged images. TL = transmitted light. All scale bars represent 2 μm. All images acquired after 2 days on sporulation media. All images are shown at the same brightness and contrast for accurate comparison. (TIF) S1 Video. Time-lapse videos of representative cells shown in Fig 1C and 1E. Panel A in the video shows a representative mCherry-wtf4/wtf4 poison -GFP diploid undergoing meiosis, captured by the time lapse microscopy we conducted in Fig 1C. mCherry-Wtf4 antidote is in magenta and Wtf4 poison -GFP is in cyan in the composite. Panel B in the video shows a representative wtf4 antidote -mcherry/wtf4 poison -GFP diploid undergoing meiosis, captured by the time lapse microscopy we conducted in Fig 1E. Wtf4 antidote -mCherry is in magenta and Wtf4 poison -GFP is in cyan in the composite. Scale bar represents 4 μm. The videos were captured using the same settings. The videos are not shown at the same brightness and contrast to avoid over saturation of pixels in the brighter frames. Each video is comprised of 115 frames captured at 10 minute intervals. (MOV) S1 Table. Yeast strains used. Column 1 is the name of strain used, while column 2 refers to the genotype. Columns 3 lists the reference for the yeast strain. If it was made in this study, we also detail how the strain was made in column 4. Column 5 lists the figure(s) in which the