A meiotic driver hijacks an epigenetic reader to disrupt mitosis in noncarrier offspring

Significance

The genome is often viewed as an orchestrated system, with genes working together to ensure the organism’s survival and reproduction. Gene alleles typically follow Mendelian inheritance, with those contributing to higher fitness favored by natural selection. However, some genes can “cheat” during sexual reproduction, selfishly enhancing their own transmission by eliminating offspring that do not inherit them. Here, we report such a gene in fission yeast, previously thought to be essential for survival due to its lethal effect on noncarrier progeny of heterozygous deletion diploids. Through integration of genetics, cell biology, and structural analysis, we show that this gene encodes a protein that interacts with an epigenetic reader and forms supramolecular foci that impede mitotic chromosome segregation in noncarrier progeny.

Abstract

Killer meiotic drivers (KMDs) are selfish genetic elements that distort Mendelian inheritance by selectively killing meiotic products lacking the KMD element, thereby promoting their own propagation. Although KMDs have been found in diverse eukaryotes, only a limited number of them have been characterized at the molecular level, and their killing mechanisms remain largely unknown. In this study, we identify that a gene previously deemed essential for cell survival in the fission yeast Schizosaccharomyces pombe is a single-gene KMD. This gene, tdk1, kills nearly all tdk1Δ progeny in a tdk1+ × tdk1Δ cross. By analyzing polymorphisms of tdk1 among natural strains, we identify a resistant haplotype, HT3. This haplotype lacks killing ability yet confers resistance to killing by the wild-type tdk1. Proximity labeling experiments reveal an interaction between Tdk1, the protein product of tdk1, and the epigenetic reader Bdf1. Interestingly, the nonkilling Tdk1-HT3 variant does not interact with Bdf1. Cryoelectron microscopy further elucidated the binding interface between Tdk1 and Bdf1, pinpointing mutations within Tdk1-HT3 that disrupt this interface. During sexual reproduction, Tdk1 forms stable Bdf1-binding nuclear foci in all spores after meiosis. These foci persist in germinated tdk1Δ progeny and impede chromosome segregation during mitosis by generating aberrant chromosomal adhesions. This study identifies a KMD that masquerades as an essential gene and reveals the molecular mechanism by which this KMD hijacks cellular machinery to execute killing. Additionally, we unveil that losing the hijacking ability is an evolutionary path for this single-gene KMD to evolve into a nonkilling resistant haplotype.

Mendelian inheritance dictates that two gene alleles in a parent are transmitted to gametes with equal chance during meiosis. However, selfish genetic elements known as meiotic drivers subvert this principle by biasing their own inheritance toward future generations (1–3). This advantage in transmission allows them to spread through sexually reproducing populations, even if they reduce the fitness of the organism. There are two main classes of meiotic drivers: true meiotic drivers, which preferentially segregate into eggs instead of polar bodies during female meiosis (4, 5), and killer meiotic drivers (KMDs), which eliminate meiotic products that do not inherit them, in either male or female meiosis, or both (6). KMDs have been found in diverse eukaryotes, including plants, fungi, and animals. Nevertheless, the genes responsible for these drivers have only been identified in a limited number of cases. Even for those that have been molecularly identified, their mechanisms and evolutionary dynamics remain largely unknown (2, 3, 6, 7).

Genetic analyses of KMD systems often reveal the presence of resistant haplotypes, also known as neutral haplotypes, that neither kill nor are susceptible to killing (8–11). For example, the well-studied KMD system Spore killer-2 (Sk-2) in the filamentous fungus Neurospora intermedia exhibits three types of haplotypes with distinct phenotypes: killer haplotype Sk-2, sensitive haplotype Sk-2^S, and resistant haplotype Sk-2^R. In a Sk-2 × Sk-2^S cross, nearly all meiotic products carrying Sk-2^S are killed. However, Sk-2^R is not susceptible to killing by Sk-2 and does not cause killing when crossed with Sk-2^S (10). Sk-2^R haplotype is frequent in areas where Sk-2 strains are present, but absent in regions where Sk-2^S strains dominate (11). Even though the genes responsible for the killing and resistance activities of Sk-2 have been identified (12, 13), the evolutionary origin of Sk-2^R remains unclear.

Currently known KMDs are typically discovered through crosses between genetically divergent natural populations (3, 6). In the fission yeast model organism Schizosaccharomyces pombe, studying hybrid sterility in natural isolates has led to the discovery of the wtf driver family (14, 15). These KMD genes act autonomously to kill spores that do not inherit them. This is achieved through the production of two overlapping transcripts from a single wtf gene. One transcript encodes a toxin that is detrimental to all spores, while the other encodes an antidote that protects carrier spores from the toxin. It has recently been shown that protection by the antidote relies on ubiquitination-mediated Golgi-to-endosome sorting (16). The mechanism by which the toxin exerts its killing effect remains elusive.

S. pombe primarily reproduces asexually as haploid cells but can undergo sexual reproduction upon starvation (17). During sexual reproduction, haploid cells of opposite mating types fuse to form diploids, which then undergo meiosis to produce four haploid spores. In this model organism, whether a gene is essential is determined by first generating a diploid strain with heterozygous deletion and then assessing the viability of the deletion haploid progeny derived from it. This method has been employed in a genome-wide deletion project to determine the essentiality of all protein-coding genes (18, 19). However, in this approach, it is possible for an autonomous KMD gene with high killing efficiency to be erroneously classified as essential. Unlike essential genes, the deletion of a KMD gene should not affect growth in haploids and should not cause progeny inviability when two deletion haploids are crossed (Fig. 1A).

Fig. 1.

tdk1 is a KMD gene in S. pombe. (A) Schematic contrasting a single-gene KMD with an essential gene. (B) Spot assays showing normal growth of tdk1Δ cells of both h+ and h− mating types compared to wild-type controls. Fivefold serial dilutions of each strain were spotted onto YES (rich medium) or EMM (minimal medium). (C and D) Tetrad analyses of tdk1+ × tdk1+, tdk1+ × tdk1Δ, and tdk1Δ × tdk1Δ crosses, showing the inviability of tdk1Δ progeny from heterozygous but not homozygous deletion diploids. Six representative tetrads for each cross are displayed in (C), with progeny labeled A–D within each tetrad. The deletion of tdk1 was tracked using the drug resistance marker kanMX. (D) Statistical analyses of progeny viability. P values were calculated using Fisher’s exact test, comparing the viability of progeny from two crosses (null hypothesis: equal viability between crosses). n, total number of progeny analyzed. (E and F) Tetrad analyses showing that mutation of the start codon (tdk1-M1A) abolishes the killing activity of tdk1. P values were calculated using the exact binomial test, comparing the observed counts of viable progeny with indicated genotypes against the expected Mendelian segregation ratio of 1:1. n, total number of progeny analyzed. The insertion of wild-type tdk1 or tdk1-M1A at the ade6 locus was tracked using a linked drug resistance marker natMX. A schematic representation of the results is shown in (F).

In this study, through a reevaluation of published deletion datasets in S. pombe, we identified a previously annotated essential gene as a single-gene KMD. This gene, dispensable for vegetative growth, efficiently kills noncarrier progeny from heterozygous deletion diploids. We demonstrate that this killing effect manifests through the disruption of mitosis after spore germination. By examining natural polymorphisms within this gene, we identified a nonkilling resistant haplotype. Furthermore, we show that the protein encoded by the wild-type gene, but not the resistant haplotype, possesses the ability to hijack cellular machinery to execute killing.

Results

tdk1 Is a KMD Gene.

In S. pombe, apart from the genome-wide deletion project using heterozygous deletion diploids, two separate deletion analyses on meiotically upregulated genes (mug) have been conducted using haploid deletion strains (20, 21). By comparing the results of these studies, we noticed an apparent discrepancy in the reported deletion phenotype of the uncharacterized gene SPCC330.04c/mug135. While no obvious phenotypes were observed when this gene was deleted in haploids (20, 21), it was classified as essential due to the inviability of haploid deletion progeny derived from heterozygous deletion diploids (18, 19). One possible explanation for this discrepancy is that this gene is a KMD, selectively killing noncarrier offspring during sexual reproduction (Fig. 1A).

Consistent with previous reports, spot assays revealed no growth defects when this gene was deleted in haploid strains (Fig. 1B). Moreover, crosses between deletion haploids resulted in normal progeny viability, exceeding 90% (Fig. 1 C and D). In contrast, when a deletion haploid was crossed with a wild-type haploid, the resulting tetrads predominantly exhibited a 2:2 segregation pattern, with the deletion progeny being inviable (Fig. 1C). Overall, the viability of wild-type progeny was unaffected, whereas the viability of deletion progeny declined to less than 3% (Fig. 1 C and D). These findings reveal that this gene is a selfish KMD. While dispensable for cell growth and sexual reproduction, it favors its own transmission by killing haploid progeny that do not inherit it from heterozygous deletion diploids. We therefore renamed it tdk1 (transmission-distorting killer).

The tdk1 gene exhibits no homology to previously characterized KMDs. It is predicted to encode a protein of 357 residues. This protein, Tdk1, harbors a domain of unknown function called DUF1773 in its C-terminal region (Pfam accession PF08593; InterPro accession IPR013902). According to the InterPro database (22), proteins containing the DUF1773 domain are primarily found in two fungal phyla, Ascomycota and Basidiomycota (SI Appendix, Fig. S1). Within Ascomycota, the distribution of DUF1773 domain-containing proteins is sporadic. They are present in fission yeast species (except Schizosaccharomyces japonicus) of the Taphrinomycotina subphylum and in a limited number of budding yeast families (mainly Debaryomycetaceae, Pichiaceae, and Lipomycetaceae) in the Saccharomycotina subphylum but are absent in the Pezizomycotina subphylum. Our BLAST analysis revealed that, in addition to the DUF1773 domain, amino acids 12 to 118 of Tdk1 exhibit sequence homology with 15 other fission yeast proteins (SI Appendix, Fig. S2).

To determine whether the predicted protein product is responsible for tdk1-mediated killing, we mutated the annotated start codon to alanine. The mutated construct was integrated into the ade6 locus of a tdk1Δ strain and the resulting strain (tdk1Δ ade6::tdk1-M1A) was crossed with tdk1Δ. We observed that ade6::tdk1-M1A exhibited no killing activity (progeny viability ~90%) (Fig. 1 E and F). In contrast, wild-type tdk1 integrated at the same locus (ade6::tdk1) retained its killing activity, reducing the viability of tdk1Δ progeny to ~7% (Fig. 1 E and F). These results demonstrate that tdk1 acts as a KMD in S. pombe, with its encoded protein Tdk1 responsible for the killing activity observed in heterozygous deletion crosses.

tdk1 Kills Noncarrier Progeny by Disrupting Mitosis.

In fungi, KMDs are commonly referred to as “spore killers” due to their ability to disable spores (23). Interestingly, we observed that nearly all of the spores from a cross between tdk1+ and tdk1Δ strains exhibited a normal appearance and successfully germinated on nutrient-rich media (Fig. 2A). No noticeable morphological differences were observed among the four spore progeny from the same tetrad during germination and germ tube elongation. However, out of the four progeny, only two, which were subsequently determined to possess the tdk1+ genotype through replica-plating, were capable of forming colonies (Fig. 2A). The other two progeny (presumably tdk1Δ) usually ceased proliferation after the first cell division (Fig. 2A).

Fig. 2.

tdk1 kills noncarrier offspring by disrupting mitosis. (A) Images of four spores from a tdk1+ × tdk1Δ tetrad showing that tdk1Δ spores undergo normal germination and outgrowth on a rich medium (YES). Genotypes were determined by replica-plating after colony formation. The deletion of tdk1 was tracked using the drug resistance marker kanMX. (Scale bar, 20 μm.) (B and C) Time-lapse fluorescence imaging revealing severe chromosome segregation defects in tdk1Δ but not tdk1+ progeny of heterozygous deletion diploids during the first mitosis after spore germination. Representative time-lapse images of a tdk1+ cell (Top), a tdk1Δ cell with the “anucleate daughter” phenotype (Middle), and a tdk1Δ cell with the “cut” phenotype (Bottom) are displayed in (B). H3-GFP serves as a chromatin marker, and fluorescence images of H3-GFP are merged with differential interference contrast images. (Scale bar, 3 μm.) (C) Quantification of mitotic cells with different phenotypes. n, number of mitotic cells analyzed.

To further investigate the postgermination killing behavior of tdk1, we monitored the cell-cycle progression of germinated tdk1Δ and tdk1+ spores separately. We used the closely linked ura4 auxotrophic marker to selectively permit germination of either tdk1Δ or tdk1+ spores (SI Appendix, Fig. S3). Time-lapse microscopy showed that tdk1Δ spores initially underwent normal germ tube emergence and elongation. However, they displayed abnormality during the first mitotic division after germination (Fig. 2B and Movies S1–S4). Approximately 60% exhibited an “anucleate daughter” phenotype, characterized by a daughter cell lacking chromosomes. Another 20% showed a “cut” phenotype, where a septum bisected the nucleus (Fig. 2 B and C). These findings suggest that Tdk1 kills noncarrier progeny by interfering with chromosome segregation during mitosis. This postgermination killing behavior is distinct from the spore killing behavior of other fungal KMDs.

HT3 Is a Naturally Occurring Resistant Haplotype of tdk1.

To assess the prevalence and diversity of tdk1 among natural S. pombe isolates, we analyzed the genome sequencing data of 56 nonclonal haploid strains (24). Our analysis revealed that tdk1 was present in all 56 strains. Interestingly, we observed that the nonsynonymous nucleotide diversity (π_N) of tdk1 was considerably higher than the genome-wide average (0.0049 vs. 0.0011) (SI Appendix, Fig. S4), similar to patterns observed in certain wtf genes (25). In contrast, the synonymous nucleotide diversity (π_S) of tdk1 was lower than the genome-wide average (0.0041 vs. 0.0066). As a result, the π_N/π_S ratio of tdk1 was much higher than the genome-wide average (1.2 vs. 0.16), suggesting that it may be under positive or balancing selection.

Analysis of the coding sequence of tdk1 revealed eight distinct haplotypes among the 56 strains, based on single-nucleotide polymorphisms (Fig. 3A and Dataset S1). We excluded an internal repeat region spanning residues 120 to 226 from this analysis due to difficulties in read mapping and assembly. We designated the haplotype present in the reference genome as HT1 and the remaining seven as HT2–HT8. The three most common haplotypes were HT2 (37%), HT1 (28%), and HT3 (19%). HT2 is characterized by three nonsynonymous variants: K50R, D241Y, and G319E, compared to HT1. In contrast, HT3 contains six different nonsynonymous variants: S65A, R102K, E230D, I235M, R255K, and D288E.

Fig. 3.

Naturally occurring HT3 is a nonkilling resistant haplotype of tdk1. (A) tdk1 haplotype network constructed from 56 nonclonal haploid natural isolates of S. pombe (24), with circle size indicating the frequency of each haplotype (detailed information in Dataset S1). A black dot represents a putative ancestral haplotype, and hash marks indicate mutational steps. Nonsynonymous polymorphisms relative to the reference haplotype (HT1) are listed (Right). A multinucleotide polymorphism in HT4 affecting three successive codons (underscored) is considered a single mutational step. The internal repeat region, corresponding to residues 120 to 226, was omitted due to challenges in mapping and assembly. (B) Tetrad analyses performed on S. pombe natural isolates revealing that HT2, but not HT3, is an active KMD. (C) Tetrad analyses showing that the nonsynonymous polymorphisms in HT3 abolish killing activity but do not affect resistance to killing by tdk1. The tdk1-HT3 strain was constructed by introducing the six nonsynonymous polymorphisms of HT3 into the reference strain. (D) Tetrad analyses showing that the nonsynonymous polymorphisms in HT2 affect neither killing nor resistance activity. The tdk1-HT2 strain was constructed by introducing the three nonsynonymous polymorphisms of HT2 into the reference strain. P values in (B–D) were calculated using the exact binomial test on counts of viable progeny with indicated genotypes. n, total number of progeny analyzed.

To determine whether HT2 and HT3 haplotypes act as active KMDs, we examined the viability of progeny resulting from heterozygous deletion crosses. Tetrad analyses revealed that in JB1180 and JB1206, both HT2 strains, progeny inheriting tdk1 exhibited normal viability (>95%) while those lacking it were mostly inviable (JB1180_tdk1Δ: 0%; JB1206_tdk1Δ: ~7%) (Fig. 3B). In contrast, no significant difference in viability was observed between progeny inheriting tdk1 and those lacking it in JB939 and JB953, two HT3 strains (Fig. 3B). These results suggest that the HT2 haplotype, but not the HT3 haplotype, is capable of killing in their respective native strain backgrounds.

To investigate whether the six nonsynonymous variants in HT3 are responsible for the lack of killing, we introduced these variants into the reference strain. The resulting tdk1-HT3 strain displayed no killing when crossed with tdk1Δ (Fig. 3C). Notably, when crossed with a reference strain containing wild-type tdk1, tdk1-HT3 exhibited full resistance to killing. We also constructed a tdk1-HT2 strain, which harbors the three nonsynonymous variants from HT2. This strain exhibited strong killing against tdk1Δ and was fully resistant to wild-type tdk1 (Fig. 3D). These findings suggest that the variants in the HT3 haplotype have resulted in a resistant haplotype that lacks killing activity but retains resistance activity.

Epigenetic Reader Bdf1 Is an Interactor of Tdk1 but Not Tdk1-HT3.

To elucidate the mechanism underlying the observed difference between Tdk1 and Tdk1-HT3, we conducted TurboID-based proximity labeling in S. pombe cells undergoing meiosis and sporulation (26, 27). This experiment revealed that Bdf1, an epigenetic reader, preferentially interacts with Tdk1 compared to Tdk1-HT3 (Fig. 4A and Dataset S2). The spectral count of Bdf1 was more than ten times higher when Tdk1 was used as bait compared to Tdk1-HT3. Interestingly, a previously published proteome-wide binary protein–protein interaction analysis using the yeast two-hybrid (Y2H) method also identified Bdf1 as a Tdk1 interactor in S. pombe (28). Bdf1 is an epigenetic reader belonging to the bromodomain and extraterminal domain (BET) protein family and recognizes acetylated histones through its two bromodomains, BD1 and BD2 (29, 30) (Fig. 4B).

Fig. 4.

Tdk1-mediated killing requires its Bdf1-binding ability. (A) TurboID–mass spectrometry identifies Bdf1 as a highly enriched proximal interactor of Tdk1 but not Tdk1-HT3 (full results in Dataset S2). (B) Domain organization of Bdf1, featuring two bromodomains (BD1 and BD2) and an extraterminal domain (ET). (C) Y2H assays showing that Bdf1 interacts with Tdk1(227-357) (designated as Tdk1C) but not full-length Tdk1, Tdk1(1-226), or Tdk1C harboring mutations from HT3. BIRm refers to I235M and R255K mutations, situated within the BIR. AD and BD denote prey and bait constructs, respectively. −LW denotes the SD/−Leu/−Trp medium; −LWHA denotes the SD/−Leu/−Trp/−His/−Ade medium. (D) Y2H assays narrowing down the Tdk1-binding region of Bdf1 to residues 372 to 554. (E) Tetrad analyses showing that the killing of tdk1Δ progeny from a tdk1+ × tdk1Δ cross is substantially diminished in the absence of the bdf1 gene. (F and G) Cryo-EM density map (F) and structural model (G) of a complex between Tdk1(211-357) and Bdf1(372-554). Residues 211 to 354 of Tdk1 and residues 525 to 547 of Bdf1 are structurally resolved. Six Tdk1 molecules are colored differently, while all Bdf1 molecules are depicted in brick red. (H) Axial view of a Tdk1 trimer unit, related to (G) by a 90° rotation. (I). Ribbon representation of the Tdk1 structure predicted by AlphaFold (31, 32). NTD, stalk domain, and CTD are visually distinguished by distinct colors. (J) Expanded view of the interaction interface between Tdk1 and Bdf1. (K) Y2H assays showing that Bdf1 mutations within the Tdk1-binding interface disrupt interaction between Tdk1C and Bdf1. (L) Tetrad analyses showing that Bdf1 mutations within the Tdk1-binding interface disrupt Tdk1-mediated progeny killing. Wild-type or mutated bdf1 were integrated into a bdf1Δ strain at the ade6 locus. (M) Schematic illustration of Tdk1 and Tdk1-HT3, with amino acid differences in the BIR highlighted in red. (N) Tetrad analyses showing that the BIR mutations abolish the killing activity of tdk1. P values comparing two crosses in (E and L) were calculated using Fisher’s exact test. P values for each cross in (E and N) were calculated using the exact binomial test on counts of viable progeny with indicated genotypes. n, total number of progeny analyzed.

We then conducted Y2H assays to verify the interaction between Tdk1 and Bdf1. The results of this analysis revealed that Bdf1 binds specifically to a C-terminal fragment of Tdk1 (residues 227 to 357, referred to as Tdk1C), but not to an N-terminal fragment (residues 1 to 226) (Fig. 4C). Surprisingly, our Y2H assay detected no discernible interaction between Bdf1 and full-length Tdk1 (Fig. 4C). This suggests that in full-length Tdk1, the Bdf1-binding site in the C-terminal region of Tdk1 might be inaccessible under our assay conditions, which may differ from those in the published Y2H study. Consistent with our proximity labeling data, Bdf1 did not interact with the C-terminal fragment of the HT3 variant (Tdk1C-HT3) in Y2H assays (Fig. 4C). Additionally, using a series of truncated Bdf1 constructs in Y2H assays, we identified the Tdk1-binding region of Bdf1 to be within residues 372 to 554 (Fig. 4D).

To determine whether Bdf1 is involved in Tdk1-mediated killing, we analyzed progeny killing in a bdf1 deletion background (tdk1+ bdf1Δ × tdk1Δ bdf1Δ). The absence of Bdf1 substantially impaired killing, as evidenced by an increase in tdk1Δ progeny viability from less than 3 to 64% (Fig. 4E). We further analyzed killing when bdf1 was heterozygously deleted (crosses: tdk1+ bdf1Δ × tdk1Δ bdf1+ and tdk1+ bdf1+ × tdk1Δ bdf1Δ; note that tdk1 and bdf1 genes are not linked). In both crosses, tdk1Δ progeny were effectively eliminated regardless of inheriting the bdf1 gene or not (SI Appendix, Fig. S5), suggesting that zygotically expressed Bdf1 is sufficient for Tdk1-mediated killing. Collectively, these results strongly support the role of Bdf1, a protein that interacts with Tdk1 but not Tdk1-HT3, in Tdk1-mediated progeny killing.

Tdk1-Mediated Killing Relies on its Ability to Interact with Bdf1.

To elucidate the structural basis of the Tdk1–Bdf1 interaction, we used cryoelectron microscopy (cryo-EM) to analyze a fused complex containing residues 211 to 357 of Tdk1 and 372 to 554 of Bdf1 (SI Appendix, Figs. S6 and S7). Within the cryo-EM density map, we resolved residues 211 to 354 of Tdk1 and residues 525 to 547 of Bdf1, achieving an overall resolution of 2.7 Å (Fig. 4F). In this complex structure, Tdk1(211-354) adopts an elongated hexameric structure with D3 symmetry, measuring 195 × 70 × 70 Å (Fig. 4 F and G). Each Tdk1 subunit comprises an extended α-helix (α1, residues 211 to 275) and a C-terminal globular domain (residues 276 to 354) (SI Appendix, Fig. S8A). Three Tdk1 subunits assemble into a mushroom-shaped trimer, with the α1 helices forming a parallel 3-helix coiled coil. This coiled coil is stabilized by extensive hydrophobic interactions within its core, supplemented by peripheral polar interactions (SI Appendix, Fig. S8B). Two of these trimers then arrange in a head-to-head configuration, resulting in the observed hexameric assembly (Fig. 4 G and H).

The predicted structure of full-length Tdk1 in the AlphaFold database can be divided into three domains: an N-terminal domain (NTD; residues 1 to 111), a stalk domain (residues 112 to 275) characterized by a long α-helix, and a globular C-terminal domain (CTD; residues 276 to 357) (31, 32) (Fig. 4I and SI Appendix, Fig. S8 C and D). The cryo-EM structure of Tdk1(211-354) closely matches the predicted structure of full-length Tdk1 (SI Appendix, Fig. S8E). The α1 helix observed in the cryo-EM structure corresponds to the C-terminal portion of the stalk domain. Notably, the first 65 amino acids of the CTD, along with the last eight amino acids of the stalk domain, correspond to the Pfam-annotated DUF1773 domain (residues 268 to 340).

Within the complex structure, residues 529 to 545 of Bdf1 form an α-helix that fits snugly into a groove between the α1 helices of two Tdk1 subunits in a trimer (Fig. 4J). This interface, composed mainly of hydrophobic residues, buries a total solvent-accessible area of 790 Ų (SI Appendix, Fig. S8F). Supporting the structure, Y2H assays demonstrated that Tdk1C interacted with Bdf1(524-554), but not with Bdf1 variants lacking this region or those harboring mutations at the Tdk1-binding interface (Q532R, I536R, L539R, or L543R) (Fig. 4K). To assess the requirement of the Tdk1–Bdf1 interaction in Tdk1-mediated killing, we ectopically expressed these Bdf1 variants in a bdf1Δ background. While Tdk1-mediated killing was restored with ectopically expressed wild-type Bdf1, none of the Tdk1-binding-deficient Bdf1 mutants increased killing (Fig. 4L). Together, these results establish that the interaction with Bdf1 is required for Tdk1’s killing activity.

Our cryo-EM structure pinpointed residues 233 to 255 of Tdk1 as the Bdf1-interacting region (BIR) (Fig. 4J). Notably, the resistant haplotype, HT3, harbors two nonsynonymous polymorphisms (I235M and R255K) within the BIR (Fig. 4M). These two mutations, collectively designated as BIRm, abrogated the ability of Tdk1C to bind Bdf1 in Y2H assays (Fig. 4C). Moreover, tetrad analyses revealed that the BIRm mutations abolished Tdk1’s killing activity (Fig. 4N). In contrast, a Tdk1 variant lacking the BIRm mutations but containing the remaining four polymorphisms (S65A, R102K, E230D, and D288E) present in the HT3 haplotype exhibited strong killing activity (Fig. 4N). These findings conclusively demonstrate that mutations within the BIR that disrupt Bdf1 binding are responsible for the observed loss of killing activity in the naturally occurring resistant haplotype, HT3.

Fission yeast encodes a Bdf1 paralog, Bdf2. Notably, deletion of both the bdf1 and bdf2 genes is synthetically lethal (29). Imaging of spores from a bdf1Δ × bdf2Δ cross revealed that bdf1Δ bdf2Δ spores failed to germinate (SI Appendix, Fig. S9A). Sequence alignment revealed conservation of the Tdk1-interacting region (residues 525 to 547 of Bdf1) in Bdf2 (SI Appendix, Fig. S9B). Proximity labeling and Y2H assays support an interaction between Tdk1 and Bdf2 (SI Appendix, Fig. S9 C–E). Interestingly, while deletion of bdf2 alone did not reduce killing efficiency, increasing the copy number of bdf2 in the bdf1Δ background (bdf1Δ leu1::bdf2) moderately enhanced Tdk1-mediated progeny killing (SI Appendix, Fig. S9F). These findings suggest that Bdf2 likely contributes to the residual weak killing observed in the absence of Bdf1.

To determine the extent to which Bdf2 contributes to Bdf1-independent Tdk1-mediated progeny killing, we generated a bdf1Δ bdf2Δ strain, rendered viable by expressing the Tdk1-binding-deficient variant Bdf1-Q532R. Tetrad analysis in this mutant background (bdf1Δ bdf2Δ bdf1-Q532R) revealed no significant viability difference between tdk1+ and tdk1Δ progeny, indicating a complete loss of Tdk1-mediated killing (SI Appendix, Fig. S9F). Collectively, these results demonstrate that Tdk1’s killing activity strictly relies on its interactions with the paralogous histone readers, Bdf1 and Bdf2, with Bdf1 exerting a far more pronounced impact than Bdf2.

Disrupting the α-Helical Continuity of the Stalk Domain Converts Nontoxic Tdk1 to Toxic in Vegetative Cells.

Coimmunoprecipitation (co-IP) experiments corroborated our Y2H results, showing that Bdf1 interacted with Tdk1C, but not with full-length Tdk1 in vegetative S. pombe cells (Fig. 5A). This suggests that full-length Tdk1 may adopt an autoinhibited configuration in vegetative cells, preventing its interaction with Bdf1. Consistent with this lack of interaction, vegetative cells expressing Tdk1 displayed normal growth in spot assays (SI Appendix, Fig. S10A). We hypothesized that mutations disrupting Tdk1’s autoinhibition might lead to toxicity in vegetative cells. To test this, we introduced random mutations using error-prone rolling circle amplification (RCA) to screen for toxic variants of Tdk1 (33, 34). This approach yielded Tdk1-L229P, a variant with a single leucine-to-proline mutation in the stalk domain upstream of the BIR. Tdk1-L229P elicited toxicity when expressed from the inducible Pnmt41 promoter (35) in vegetative cells and interacted with Bdf1 in Y2H assays (SI Appendix, Fig. S10 A and B). We then investigated whether the toxicity of Tdk1-L229P depends on Bdf1. Compared to the bdf1Δ strain, which displayed a moderate growth defect, the bdf1Δ strain expressing Tdk1-L229P showed no more severe growth defect (SI Appendix, Fig. S10A). These findings suggest that altering the stalk domain can generate a Bdf1-dependent toxic Tdk1 variant.

Fig. 5.

Linker-containing but not wild-type Tdk1 is toxic in vegetative cells. (A) Coimmunoprecipitation showing that in vegetative S. pombe cells, Tdk1C but not full-length Tdk1 interacts with Bdf1. IP, immunoprecipitation. IB, immunoblotting. (B) Schematic representation of Tdk1 variants containing linkers. Tdk1-1L, Tdk1-2L, and Tdk1-3L harbor a flexible linker (SGGGSSG) inserted at distinct positions (after residue 119, 184, or 275, respectively) in the stalk domain upstream of the BIR. (C) Y2H assays showing that linker-containing Tdk1 variants (Tdk1-1L, Tdk1-2L, and Tdk1-3L), but not full-length Tdk1, interact with Bdf1. (D) Spot assays showing that expression of linker-containing Tdk1 variants (Tdk1-1L, Tdk1-2L, and Tdk1-3L), but not wild-type Tdk1, induces toxicity in vegetative cells. Mutations in the BIR (BIRm) or deletion of the bdf1 gene abrogate this toxic effect. (E and F) Time-lapse fluorescence imaging showing mitotic defects in cells expressing Tdk1-1L, in contrast to those expressing wild-type Tdk1. Representative time-lapse images of a cell expressing Tdk1 (Left), a cell expressing Tdk1-1L with the “anucleate daughter” phenotype (Middle), and a cell expressing Tdk1-1L with the “cut” phenotype (Right) are displayed in (E). Ish1 serves as a nuclear envelope marker. BF, bright field. (Scale bar, 3 μm.) (F) Quantification of mitotic cells with different phenotypes. n, number of analyzed mitotic cells. Only cells with a detectable level of Tdk1 or Tdk1-1L were analyzed. In (A and D–F), Tdk1 variants were expressed from the inducible P41nmt1 promoter (35).

Given that proline is an α-helix disruptor (36, 37), we hypothesized that the toxicity of the Tdk1-L229P variant could be due to a disruption of the continuous α-helical configuration of the stalk domain. To test this, we introduced a flexible linker (SGGGSSG) at three different positions in the stalk domain upstream of the BIR (Fig. 5B). These linker-containing variants (Tdk1-1L, Tdk1-2L, and Tdk1-3L) all gained the ability to interact with Bdf1 in Y2H assays (Fig. 5C). Consistently, spot assays demonstrated strong toxicity of these Tdk1 variants when expressed from the Pnmt41 promoter in vegetative cells (Fig. 5D). As expected, the toxic effect of Tdk1-1L was abolished either by mutating the BIR (Tdk1-1L-BIRm) or by deleting bdf1 (Fig. 5D).

We then used live-cell imaging to examine the cellular phenotype caused by the induced expression of Tdk1-1L in comparison with the induced expression of wild-type Tdk1. In this experiment, we tagged Tdk1-L and wild-type Tdk1 with a fluorescent protein. While the expression of wild-type Tdk1 did not lead to mitotic defects in vegetative cells, the expression of Tdk1-1L resulted in aberrant chromosome segregation in approximately 80% of dividing cells during the first mitosis after the appearance of the Tdk1-1L signal (Fig. 5 E and F). Of these, half displayed the “anucleate daughter” phenotype, and the other half exhibited the “cut” phenotype. Notably, both phenotypes resembled those observed in tdk1Δ progeny from a tdk1+ × tdk1Δ cross. Interestingly, while wild-type Tdk1 exhibited a diffuse nuclear distribution, Tdk1-1L formed distinct nuclear foci (Fig. 5E). To assess the Bdf1-binding ability of Tdk1-1L foci, we employed a Tdk1-binding fragment of Bdf1, Bdf1(524-554), as a probe. Unlike full-length Bdf1, this fragment lacks histone-binding domains and is therefore freely diffusible. We observed that this fragment colocalized with Tdk1-1L foci (SI Appendix, Fig. S10C), indicating that Tdk1-1L molecules within these foci are capable of Bdf1 binding. Collectively, these findings demonstrate that the disruption of the α-helical continuity within the stalk domain of Tdk1 results in its release from the autoinhibited state. This enables Tdk1 to interact with Bdf1, leading to a killing effect in vegetative cells that mimics the Tdk1-mediated killing observed in noncarrier progeny during sexual reproduction.

Tdk1-Mediated Killing Requires Its Attachment to Chromosomes.

We then investigated the roles of Bdf1’s bromodomains and ET domain in Tdk1-mediated killing. Bromodomains bind acetylated lysines on histone tails via a hydrophobic pocket, while the function of ET domains is not fully understood (38–42). To disrupt the histone-binding activity of each bromodomain, we mutated the conserved tyrosine within the acetyl-lysine binding pocket to alanine (40). In crosses using strains with a single mutated bromodomain (bdf1-Y123A or bdf1-Y293A), Tdk1-mediated progeny killing was mildly affected, with the viability of tdk1Δ progeny increased to around 20% (Fig. 6A). Simultaneously mutating both bromodomains (bdf1-Y123A, Y293A) resulted in a complete abrogation of killing (Fig. 6A). This complete loss of killing contrasts with the substantial yet incomplete reduction of killing observed in the bdf1Δ background. This can be attributed to a dominant-negative effect, where the Bdf1-Y123A, Y293A mutant competes with Bdf2 for Tdk1 binding. In a cross where the ET domain was deleted (bdf1ΔET), tdk1Δ progeny were efficiently killed, with viability below 2% (Fig. 6A). These findings suggest that the bromodomains, but not the ET domain, of Bdf1 are essential for Tdk1-mediated progeny killing.

Fig. 6.

Tdk1-mediated killing requires its attachment to chromosomes and its formation of foci. (A) Tetrad analyses showing that the histone binding activity of Bdf1 is required for Tdk1-mediated progeny killing. Mutations Y123A and Y293A disrupt the histone-binding pockets of BD1 and BD2, respectively. For the ET domain deletion, residues 430 to 510 of Bdf1 were deleted. Wild-type or mutated bdf1 were integrated into a bdf1Δ strain at the ade6 locus. P values were calculated using the exact binomial test on counts of viable progeny with indicated genotypes. n, total number of progeny analyzed. (B) Spot assays showing that Tdk1-1L-BIRm, but not Tdk1 or Tdk1-BIRm, exhibits toxicity when artificially tethered to BD1. Artificial tethering was achieved through the interaction between GBP and the GFP variant mECitrine. (C) Spot assays showing that Tdk1-1L-BIRm exhibits toxicity when tethered to two other histone reader domains: the PWWP domain, recognizing Set9-catalyzed H4K20 methylation (43, 44), and the PHD domain, recognizing Set1-catalyzed H3K4 methylation (45, 46). Toxicity is abolished in the absence of Set9 or Set1, respectively. (D) Spot assays showing that tethering Tdk1-1L-BIRm with histone H3 elicits toxicity that increases with the level of H3-GBP. Padf1, Prps901, and Pcyc1 are promoters of increasing strength. (E) Immunoblotting showing H3-GBP levels in (D). (F) Fluorescence micrographs showing localization of Tdk1, Tdk1-1L, Tdk1-BIRm, and Tdk1-1L-BIRm in vegetative cells. DIC, differential interference contrast. (Scale bar, 3 μm.) (G–I) Time-lapse fluorescence imaging showing that attaching a Tdk1 focus to a specific chromosomal site impedes mitotic chromosome segregation at that site. A schematic of the site-specific attachment assay is presented in (G). The tetO array was integrated adjacent to the lacO array. LacIw is a LacI variant with reduced affinity for lacO (47). Representative time-lapse images of cells undergoing mitosis with repressed (Top) or induced (Bottom) expression of Tdk1-1L-BIRm are shown in (H). (Scale bar, 3 μm.) (I) Statistical analysis of (H), with P value calculated using Fisher’s exact test. A dividing cell is defined as a cell containing two separating nuclei, marked by the nuclear envelope marker Ish1. n, number of mitotic cells analyzed. In (B–I), Tdk1 variants were expressed from the inducible P41nmt1 promoter (35).

To further examine the role of Bdf1 in Tdk1-mediated killing, we artificially tethered Tdk1-1L-BIRm, which cannot bind Bdf1, to BD1, the first bromodomain of Bdf1. The tethering was achieved through the interaction between mECitrine, a fluorescent protein derived from GFP, and the GFP-binding protein (GBP) (48). Coexpression of Tdk1-1L-BIRm-mECitrine with BD1-GBP resulted in a toxic effect, whereas coexpression with either BD1 alone or the BD1-Y123A-GBP mutant did not elicit toxicity (Fig. 6B). These results suggest that the only role of Bdf1 in Tdk1-mediated killing is to bridge an association between Tdk1 and acetylated histones, leading to the attachment of Tdk1 to chromosomes.

We then investigated whether attaching Tdk1-1L-BIRm to chromosomes through other means could lead to toxicity. Coexpression of Tdk1-1L-BIRm-mECitrine with either a GBP-fused PWWP domain of Pdp1 (a H4K20 methylation reader) or a GBP-fused PHD domain of Png1 (a H3K4 methylation reader) resulted in toxicity (43, 45) (Fig. 6C). Moreover, loss of the respective methyltransferases (Set9 for H4K20 and Set1 for H3K4) (44, 46) abolished the toxicity of each combination (Fig. 6C). Similarly, coexpressing Tdk1-1L-BIRm-mECitrine with GBP-fused histone H3 resulted in a dose-dependent toxic effect, intensifying with H3-GBP levels (Fig. 6 D and E). Collectively, these results demonstrate that chromosomal attachment of Tdk1 is essential for Tdk1-mediated killing.

Tdk1-Mediated Killing Requires the Formation of Tdk1 Foci.

While tethering Tdk1-1L-BIRm to BD1 exhibited toxicity, tethering either Tdk1 or Tdk1-BIRm to BD1 did not exhibit toxicity (Fig. 6B). This contrast suggests that chromosomal attachment alone is insufficient for Tdk1 to manifest its killing effect. The observation that Tdk1-1L but not wild-type Tdk1 formed foci in vegetative cells led us to hypothesize that foci formation might be another requirement for Tdk1-mediated killing. We reasoned that these foci, when attached to chromosomes, could potentially create aberrant chromosomal adhesions, impeding their mitotic segregation. Supporting this hypothesis, we found that all Tdk1 variants exhibiting toxicity, including Tdk1-L229P and those with linker insertions, displayed nuclear foci in vegetative cells (SI Appendix, Fig. S10D). To determine whether Bdf1 binding is necessary for Tdk1 foci formation, we introduced the BIRm mutations or deleted the bdf1 gene. Neither manipulation affected Tdk1 foci formation (Fig. 6F and SI Appendix, Fig. S10E), indicating that Bdf1 binding is not essential for Tdk1 to form foci.

To examine whether chromosome-attached Tdk1 foci could locally impede chromosome segregation at sites of attachment, we employed a simplified model system. We introduced a fusion protein of GBP and LacI (GBP-LacI) into a strain harboring a 256× lacO array integrated at chromosome I (49). The specific interaction between LacI and lacO enabled the tethering of Tdk1-1L-BIRm-mECitrine, under the control of an inducible promoter, to the lacO array (Fig. 6G). The position of this locus was visualized using a fluorescently tagged TetR protein binding to a tetO array integrated adjacent to the lacO array (50). When the expression of Tdk1-1L-BIRm-mECitrine was repressed, a single TetR focus was observed to split into two distinct foci during mitosis (Fig. 6H), suggesting proper sister chromatid segregation. Conversely, inducing Tdk1-1L-BIRm-mECitrine expression resulted in a failure of TetR focus separation in approximately 90% of mitotic cells (Fig. 6 H and I). Notably, the unseparated TetR focus was invariably located near a Tdk1-1L-BIRm focus (Fig. 6I), suggesting a local hindrance effect of Tdk1 foci on chromosome segregation. Collectively, our data demonstrate that both Tdk1’s chromosomal attachment and its foci formation are essential for killing, likely by generating aberrant chromosomal adhesions that impede segregation.

Tdk1 in Spores Forms Bdf1-Binding Foci That Disrupt Mitosis in Noncarrier Progeny.

We next investigated whether during sexual reproduction, wild-type Tdk1 exhibits Bdf1 binding and foci formation, two abilities required for the killing effect of toxic Tdk1 variants in vegetative cells. Microscopic analysis of tetrads expressing Tdk1-mECitrine revealed distinct Tdk1 nuclear foci in spores (Fig. 7A and SI Appendix, Fig. S11A). Consistent with our proximity labeling data obtained from cells undergoing meiosis and sporulation, these nuclear foci colocalized with the aforementioned Bdf1(524-554) fragment (Fig. 7A), supporting Tdk1’s ability to bind Bdf1 in spores. Notably, in the absence of Bdf1, Tdk1 still formed foci in spores (SI Appendix, Fig. S11 A–C), similar to the behavior of linker-containing Tdk1 in vegetative cells. These findings indicate that, in spores, wild-type Tdk1 acquires the two abilities essential for its killing effect: Bdf1 binding and foci formation.

Fig. 7.

Tdk1 in spores forms toxic Bdf1-binding foci that impede chromosome segregation in noncarrier offspring. (A) Fluorescence micrographs showing colocalization of Bdf1(524-554) with Tdk1 foci in spores of tdk1Δ/tdk1Δ ade6::tdk1-mECitrine/ade6::tdk1-mECitrine lys3::Pbdf1-CCHex-mCherry-bdf1(524-554)/lys3::Pbdf1-CCHex-mCherry-bdf1(524-554). (Scale bar, 3 μm.) (B) Fluorescence micrographs showing persistence of Tdk1 foci in tdk1Δ progeny of ura4-D18::tdk1-mTurquoise2/tdk1Δ::ura4 after spore germination. (Scale bar, 3 μm.) (C) Box plot showing the quantity of Tdk1-mECitrine molecules within individual foci in spores of ura4-D18::tdk1-mECitrine/tdk1Δ::ura4. Cnp1-mECitrine (endogenously tagged) was used as a calibration standard for fluorescence intensity comparison (51). Boxes represent median and interquartile range (25th-75th percentiles). Whiskers extend to minimum and maximum values. (D) Fluorescence micrographs showing resistance of Tdk1 foci to 5% 1,6-hexanediol (1,6-HD) treatment. Germinating tdk1Δ spores of ura4-D18::tdk1-mECitrine/tdk1Δ::ura4 were subjected to a 10-min exposure. (Scale bar, 3 μm.) (E) A diagrammatic model depicting how Tdk1 impedes chromosome segregation by forming foci and interacting with the epigenetic reader Bdf1.

We then examined whether Tdk1 foci formed in spores persisted after spore germination. We generated a strain with a fluorescently tagged tdk1 integrated at the ura4 locus (ura4-D18::tdk1-mTurquoise2), which is closely linked to the tdk1 locus (SI Appendix, Fig. S12A). The ura4 auxotrophic marker enabled selective germination of only the tdk1Δ progeny from a cross between ura4-D18::tdk1-mTurquoise2 and tdk1Δ::ura4 strains. Following spore germination, the tdk1Δ progeny displayed “anucleate daughter” and “cut” phenotypes at rates comparable to those seen in tdk1Δ progeny from a tdk1+ × tdk1Δ cross (SI Appendix, Fig. S12B). Microscopic analysis revealed that Tdk1-mTurquoise2 nuclear foci persisted in germinated tdk1Δ progeny and were still observable during the first mitosis, a stage when chromosome segregation failures occurred (Fig. 7B).

To estimate the number of Tdk1 molecules within each Tdk1 focus in spores, we employed a quantitative fluorescence imaging approach, using well-characterized Cnp1 foci as a calibration standard (52, 53). A previous study has determined that a single anaphase Cnp1 focus contains around 680 molecules (51). By measuring the fluorescence intensity of Cnp1-mECitrine and Tdk1-mECitrine foci under identical imaging conditions, we found that the median fluorescence intensity of a Tdk1 focus in spores from a ura4-D18::tdk1-mECitrine × tdk1Δ::ura4 cross was approximately 8.7 times that of a Cnp1 focus (Fig. 7C; see Materials and Methods). Based on these data, we estimate that each Tdk1 supramolecular foci within a spore contains an average of approximately 6,000 molecules.

We then assessed the stability of Tdk1 foci by treating germinating tdk1Δ spores from the ura4-D18::tdk1-mECitrine × tdk1Δ::ura4 cross and vegetative cells expressing Tdk1-1L with 5% 1,6-hexanediol (1,6-HD). This compound is known to disrupt weak hydrophobic interactions (54). A 10-min treatment with 1,6-HD efficiently dissolved liquid droplets of Cdr2, as reported (55) (SI Appendix, Fig. S13A). However, this treatment had no effect on Tdk1 foci in germinating tdk1Δ spores or Tdk1-1L foci in vegetative cells (Fig. 7D and SI Appendix, Fig. S13B). This resistance to 1,6-HD suggests that these foci are stabilized by strong intermolecular interactions. In summary, these findings support that Tdk1 adopts a toxic conformation in spores, similar to that of linker-containing Tdk1 variants in vegetative cells. This conformation enables Tdk1 to form stable nuclear foci and interact with the epigenetic reader Bdf1. These chromosome-attached Tdk1 foci persist after spore germination in tdk1Δ progeny and hinder subsequent mitosis, likely by generating aberrant chromosomal adhesions (Fig. 7E).

Discussion

Our reevaluation of S. pombe deletion datasets has revealed a KMD gene, previously thought to be essential, which we have named tdk1. While dispensable for cell growth, tdk1 kills noncarrier haploid progeny from heterozygous deletion diploids by disrupting mitosis. Additionally, we have identified a tdk1 haplotype in natural isolates that is nonkilling but resistant to killing by the wild-type tdk1. The protein encoded by wild-type tdk1 interacts with the epigenetic reader Bdf1 during sexual reproduction, but a version harboring polymorphisms from the resistant haplotype HT3 lacks this interaction. Cryo-EM analysis has revealed that two polymorphisms in HT3 affect the Tdk1–Bdf1 binding interface. In spores, Tdk1 forms Bdf1-binding foci, which persist in tdk1Δ progeny and disrupt the segregation of mitotic chromosomes after spore germination. A companion study demonstrated that Tdk1 expressed during germination in carriers adopts a nontoxic conformation (56). This nontoxic form does not bind Bdf1 and can actively promote the dissolution of the preformed toxic Tdk1 foci, ensuring the survival of progeny inheriting the tdk1 gene (56). This remarkable structural duality enables Tdk1 to constitute a toxin–antidote pair for meiotic drive.

While our understanding of how KMDs execute killing is limited, research in fungi has provided valuable insights (23, 57). In Podospora anserina, the Spok driver family encodes proteins containing both kinase and nuclease domains (58). A recent study suggests these proteins likely kill by disrupting DNA metabolism, potentially through nuclease-mediated cleavage of DNA (59). Another KMD in P. anserina, het-s, exhibits its drive activity when crossed with strains carrying the het-S haplotype (60). Maternally inherited HET-s protein converts Het-S into a toxic protein, which perforates the cell membrane in spores carrying the het-S haplotype (61, 62). Our study reveals that tdk1 executes killing by disrupting mitotic chromosome segregation, a killing mechanism not known among fungal KMDs. To our knowledge, the only previously known case of KMD killing through disrupting chromosome segregation is the Paris sex-ratio KMD system in Drosophila simulans. In this case, the driver encodes a dysfunctional heterochromatin protein that fails to facilitate the segregation of sister chromatids of the Y chromosome during meiosis II (63). Further research is needed to determine whether different KMDs tend to converge on common cellular processes to execute their killing effects.

Unlike other fungal KMDs, which kill spores, tdk1 exhibits a late killing effect and executes killing only after spore germination. This delayed action is due to its specific targeting of mitosis, a cellular process absent in gametes. This late killing effect, analogous to the behaviors of embryo-killing elements found in nematodes (64–67), indicates a broader range of killing modes for KMDs than previously realized. It is possible that there are additional KMDs yet to be identified that target processes that are absent in gametes but occur during postgametic development.

The transmission advantage of KMDs allows them to spread through sexually reproducing populations, even if they reduce organismal fitness (1–3). This ability positions KMDs as promising candidates for biocontrol applications in the wild (68–70). However, a major challenge to harnessing KMDs lies in the potential evolution of suppressor mutations that can counteract the drive (71–73). Our findings suggest that mutations disrupting the Tdk1-interacting ability of Bdf1, or completely inactivating Bdf1, could act as potential suppressors of the tdk1 KMD. Nevertheless, sequence analysis of natural S. pombe isolates revealed no such mutations (SI Appendix, Fig. S14 and Dataset S3) (24). This absence is likely due to two reasons. First, the disruption of bdf1 moderately reduces the growth of haploid S. pombe cells (Fig. 5C and SI Appendix, Fig. S10A), imposing a selective disadvantage. Second, the suppression of killing requires mutations in both parental bdf1 copies (Fig. 4E and SI Appendix, Fig. S5). This “two-hit” requirement significantly reduces the chance of suppression. These findings offer potential insights for the design of synthetic gene drive systems with a lower risk of suppressor emergence (68).

Natural resistant haplotypes of KMDs that exhibit no killing activity but can confer resistance to killing are frequently observed. Molecularly characterized examples include the “antidote-only” wtf genes that can express an antidote isoform but not a toxin isoform (14, 15, 25, 73–75). The identification of the HT3 haplotype of tdk1 reveals a distinct mechanism, demonstrating that resistance can arise through the loss of the ability to hijack a host factor required for killing. Such resistant haplotypes likely represent an intermediate stage in KMD degeneration, where the killing activity is lost but the protecting activity remains unaffected. In KMDs with separate toxin and antidote genes, this stage can be achieved through loss-of-function mutations in the toxin gene (74). However, for KMDs that use a single protein, loss-of-function mutations would result in sensitive haplotypes. Therefore, separation-of-function mutations, like those observed in HT3, which specifically disrupt the killing activity, are necessary to generate resistant haplotypes.

The presence of close Tdk1 homologs across fission yeast species that have diverged about 100 Mya provides an opportunity to explore the evolutionary trajectory of a KMD. Investigating the potential drive activity of the Tdk1 homologs in other species is of great interest. If they demonstrate drive activity, it would suggest the presence of a likely ancestral KMD, implying an extraordinarily long persistence (over 100 My) for this KMD family. Deciphering the evolutionary strategies that underlie their remarkable success and escape from extinction would be valuable. Alternatively, if these homologs lack drive activity, studying how tdk1 acquired its drive activity compared to its homologs would provide insights into the evolutionary origins of KMDs. Interestingly, many non-fission-yeast fungal species possess remote Tdk1 homologs that contain the DUF1773 domain (SI Appendix, Fig. S1A). Determining whether these remote Tdk1 homologs act as KMDs or have other biological functions would offer further insights into the evolutionary origin of Tdk1.

Despite their evolutionary importance, the number of KMDs characterized at the molecular level remains limited. The identification of KMDs based on genetic incompatibility between diverged populations often requires laborious genetic analyses. Our identification of tdk1 as a KMD underscores an alternative and complementary strategy in organisms capable of proliferating as haploids. This strategy involves reevaluating genes previously classified as “essential” but with unclear functions, particularly when the lethal phenotype only occurs in loss-of-function mutant haploid progeny derived from a heterozygous diploid, but not in vegetative haploid mutant cells. It is important to note that active KMDs may not exhibit as complete progeny killing as tdk1 and thus may not have been classified as essential genes. A quantitative comparison of mutant phenotypes between haploid progeny from heterozygotes and vegetative haploid cells could potentially lead to the identification of KMDs with moderate killing efficiency.

Materials and Methods

S. pombe strains, plasmids, and oligonucleotides used in this study are listed in Dataset S4 A–C. Detailed methods are available in SI Appendix, SI Materials and Methods. Details of the measurements and quantifications in this study are provided in Dataset S5.

Cryo-EM data collection and refinement statistics are summarized in SI Appendix, Table S1.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Movie S1.

A germinated tdk1+ spore from heterozygous deletion diploids exhibiting normal mitosis, corresponding to the cell shown at the top of Fig. 2B. H3-GFP serves as a chromatin marker.

Movie S2.

A germinated tdk1Δ spore from heterozygous deletion diploids exhibiting the “anucleate daughter” phenotype, corresponding to the cell shown in the middle of Fig. 2B. H3-GFP serves as a chromatin marker.

Movie S3.

A germinated tdk1Δ spore from heterozygous deletion diploids exhibiting the “cut” phenotype, corresponding to the cell shown at the bottom of Fig. 2B. H3-GFP serves as a chromatin marker.

Movie S4.

Long-time imaging of a germinated tdk1Δ spore exhibiting the “anucleate daughter” phenotype. The germinated spore displayed the “anucleate daughter” phenotype during the first mitosis. The nucleated cell then progressed to a second round of mitotic division, during which it again experienced chromosome segregation defects. H3-GFP serves as a chromatin marker.

Data, Materials, and Software Availability

The cryo-EM density map and coordinates data have been deposited in Electron Microscopy Data Bank [EMDB, EMD-61290 (76)] and Protein Data Bank [PDB, 9JA5 (77)]. All other data are included in the manuscript and/or supporting information.