An introgressed gene causes meiotic drive in Neurospora sitophila

Significance In order to survive, most organisms must deal with parasites. Such parasites can be other organisms or, sometimes, selfish genes found within the host genome itself. While much is known about parasitic organisms, the interaction with their hosts, and their ability to spread within and between species, much less is known about selfish genes. We here identify a selfish “spore killer” gene in the fungus Neurospora sitophila. The gene appears to have evolved within the genus but has entered the species through hybridization and introgression. We also show that the host can counteract the gene through RNA interference. These results shed light on the diversity of selfish genes in terms of origin, evolution, and host interactions.

Meiotic drive elements cause their own preferential transmission following meiosis. In fungi, this phenomenon takes the shape of spore killing, and in the filamentous ascomycete Neurospora sitophila, the Sk-1 spore killer element is found in many natural populations. In this study, we identify the gene responsible for spore killing in Sk-1 by generating both long-and short-read genomic data and by using these data to perform a genome-wide association test. We name this gene Spk-1. Through molecular dissection, we show that a single 405-nt-long open reading frame generates a product that both acts as a poison capable of killing sibling spores and as an antidote that rescues spores that produce it. By phylogenetic analysis, we demonstrate that the gene has likely been introgressed from the closely related species Neurospora hispaniola, and we identify three subclades of N. sitophila, one where Sk-1 is fixed, another where Sk-1 is absent, and a third where both killer and sensitive strain are found. Finally, we show that spore killing can be suppressed through an RNA interference-based genome defense pathway known as meiotic silencing by unpaired DNA. Spk-1 is not related to other known meiotic drive genes, and similar sequences are only found within Neurospora. These results shed light on the diversity of genes capable of causing meiotic drive, their origin and evolution, and their interaction with the host genome. meiotic drive | genomic conflict | spore killer | Neurospora S ome genes can spread in a population even though they have negative effects on the organisms that carry them. These are often referred to as "selfish" genes and it is becoming increasingly clear that such genes can be important drivers of a large number of evolutionary patterns (1)(2)(3). One class of selfish genetic elements are killer meiotic drivers (KMDs). When heterozygous at meiosis they can increase their own transmission by destroying or incapacitating meiotic products that do not carry them. KMDs have been found across eukaryotes, and in animals and plants they generally act in male meiosis by killing sperm. In fungi they kill sexual spores, and are therefore known as "spore killers" (4,5).
A KMD must perform two tasks: kill noncarrier (sensitive) meiotic products, and ensure that it does not kill itself. It can do this through either a poison-antidote mechanism, where they produce a poison that kills indiscriminately and an antidote that they keep to themselves, or through a killer-target mechanism, where the killer attacks a target element that can only be found in sensitive meiotic products. In order to successfully drive, the poison and antidote must always be inherited together and a target must never be inherited with a killer. For this reason many KMDs are found in regions of low recombination, such as inversions or on sex chromosomes (6)(7)(8). Some KMDs only require a single gene to function, and so are not associated with regions of low recombination. This is often the case in fungi, where the majority of identified KMDs are single-gene systems. In Podospora, a family of KMD genes named Spok are found at high frequencies (9,10). Spok genes produce a single protein that can act as both poison and antidote simultaneously, probably through the action of different functional domains (10). Podospora anserina also harbors het-s, a single-gene KMD which produces a prion that can cause drive by targeting the protein generated from the sensitive allele (11). In Schizosaccharomyces pombe, the large and highly diverse wtf gene family causes drive by producing poison and antidote products from overlapping transcripts generated from two different start codons (12,13). Strains of S. pombe vary in wtf content and a single strain can carry up to 14 different driving wtf genes, generating extensive sexual incompatibilities between even closely related isolates (14,15).
As exemplified by the wtf genes, meiotic drive can often reduce the fertility of the organism. This puts the KMD in conflict with the rest of the genome, and suppressors are therefore expected to evolve. Suppression of drive has been observed in a number of species, but only in a few cases has the underlying mechanism of suppression been identified. For instance, in Drosophila simulans, RNA interference (RNAi) is necessary for suppression of a sex ratio KMD (16). The reduction of fertility is also hypothesized to drive reproductive isolation between populations, and some KMDs act as hybrid incompatibility loci (17)(18)(19)(20). At the same time, if KMDs can cross reproductive barriers, they may be able to quickly establish themselves in sensitive populations and bring linked genes with them (21,22).

Significance
In order to survive, most organisms must deal with parasites. Such parasites can be other organisms or, sometimes, selfish genes found within the host genome itself. While much is known about parasitic organisms, the interaction with their hosts, and their ability to spread within and between species, much less is known about selfish genes. We here identify a selfish "spore killer" gene in the fungus Neurospora sitophila. The gene appears to have evolved within the genus but has entered the species through hybridization and introgression. We also show that the host can counteract the gene through RNA interference. These results shed light on the diversity of selfish genes in terms of origin, evolution, and host interactions.
In the ascomycete fungus Neurospora, three different spore killer KMDs have been identified (23). Sk-2 and Sk-3 are found in Neurospora intermedia and are both poison-antidote systems located in large, highly rearranged, nonrecombining regions (8,24). The rsk gene performs the antidote function for both Sk-2 and Sk-3, but the gene performing the poison function in Sk-2 (rfk-1) is not found in Sk-3 (8,25,26), meaning that an unknown gene must be responsible.
Sk-1 is found in Neurospora sitophila ( Fig. 1 A and B) and was the first spore killer to be identified in Neurospora. In heterozygous Sk-1 x sensitive crosses, nearly all spores carrying the sensitive genotype are killed, generating a distinct pattern of four black, viable spores and four white, dead spores ( Fig. 1C) (23). Little is known about the genetics of Sk-1 and its relationship to Sk-2 and Sk-3, but an extensive sampling effort has identified and isolated over a hundred strains carrying Sk-1. While the Sk-1 element is found globally and in ∼15% of natural N. sitophila isolates, the variation in frequency among locations is large. The element appears to be absent in many populations, fixed in others, and, in two well-sampled locations, Italy and Tahiti, both killers and sensitives are found in roughly equal frequencies (27,28).
In this study, we set out to identify the locus responsible for spore killing in N. sitophila. Using whole-genome sequencing of 56 N. sitophila strains, we show that Sk-1 spore killing is caused by a single gene (Spk-1) and demonstrate that it is responsible for both killing and resistance. Our data suggest that Spk-1 has been introgressed from a different Neurospora species, potentially Neurospora hispaniola. Furthermore, we show that the population structure of N. sitophila is divided into three subclades, one of which is fixed for Sk-1, one where Sk-1 is absent, and a third where killers and sensitives intermix. Finally, we show that spore killing can be suppressed in certain crosses by an RNAi-based genome defense mechanism known as meiotic silencing by unpaired DNA (MSUD). Spk-1 is a class of spore killer genes and we demonstrate that RNAi can protect against meiotic drivers in fungi.

Results
The Locus Responsible for Spore Killing Is Located on Chromosome 6.
We sequenced four N. sitophila strains (two Sk-1 killers and two sensitives) using the PacBio RSII platform, generating high-quality genome assemblies, two of which had most chromosomes assembled from end to end (SI Appendix, Table S1). Aligning these assemblies to the Neurospora crassa OR74 genome assembly (29) reveals that all four strains are largely collinear and that the two Sk-1 strains do not carry any large rearrangements (SI Appendix, Fig. S1), unlike the Sk-2 and Sk-3 spore killer elements in N. intermedia (8).
We also generated short-read, paired-end sequencing data from 56 N. sitophila strains, using the Illumina HiSeq platform. Our sampling strategy aimed to capture the global diversity of N. sitophila and to include a balanced representation of both killer and sensitive strains (SI Appendix, Fig. S2 and Table S2). Special attention was paid to two well-sampled populations in Tahiti (28) and Italy (27), which were both polymorphic for Sk-1.
In order to identify the locus responsible for spore killing, we conducted a genome-wide association test. Illumina reads were mapped to the PacBio assembly of the killer strain W1434, singlenucleotide polymorphisms (SNPs) were called, and the association of each biallelic site to the killer phenotype was calculated using Fisher's exact test ( Fig. 2A). Only 25 SNPs covering a 2-kbp region on chromosome 6 showed a perfect association with the killer phenotype ( Fig. 2 B and C). We call this region sk1c1, for Sk-1 candidate 1. A closer inspection revealed that in both killer and sensitive strains, the region contains a homolog to the N. crassa gene NCU09865. Sensitive strains carry the whole NCU09865 sequence, but in the killer strains the allele is truncated and is missing the first half of the open reading frame (ORF); in its stead there is ∼1 kbp of new sequence (Fig. 2D). NCU09865 is of unknown function in N. crassa, but it is predicted to contain a methyltransferase domain and is most highly expressed during sexual development (30). The methyltransferase domain is present in the sensitive N. sitophila allele, but is partially missing in the Sk-1 allele.
Confirmation of the Killing and Resistance Activity of sk1c1 k . We tested if the killer sk1c1 allele (sk1c1 k ) is responsible for spore killing using three different strategies. First, we tested if it showed distorted segregation. A cross between a killer strain (W1446) and a sensitive strain (W1426) was performed and 46 sexual spores were picked and germinated separately. Segregation of the killer allele was assessed by PCR, with primers detecting a size polymorphism at sk1c1. All 46 spores carried the killer allele, showing complete segregation distortion.
We then generated a deletion mutant of the locus by deleting 2.8 kbp surrounding sk1c1 k in strain W1434 (SI Appendix, Fig.  S3). When the deletion mutant was crossed to sensitive strains no killing was observed, and when it was crossed to killer strains killing was detected (Fig. 2G). These results show that sk1c1 k is necessary both for killing sensitive strains and for resistance to killing from other Sk-1 strains, and presumably itself.
Finally, we inserted sk1c1 k at the same locus in a sensitive N. sitophila strain (W1426), by replacing NCU09865. When we crossed this knockin strain to a sensitive strain (5941), spore killing was again detected (Fig. 2G), and when we crossed it to an Sk-1 strain (W1434), killing was not observed, indicating that sk1c1 k is sufficient When crossing a sensitive strain to an Sk-1 strain, four viable black and four dead white spores are produced, indicating spore killing (1). In a cross between two sensitive N. sitophila strains, eight viable spores are formed in an ascus (2). When deleting the sk1c1 k locus (Sk-1Δ) and crossing to a sensitive strain, killing is lost (3). When crossing Sk-1Δ to another Sk-1 strain, spore killing is observed, indicating loss of resistance to spore killing (4). When inserting sk1c1 k at the same locus in a sensitive strain (9865Δ:::sk1c1) and crossing it to a sensitive strain, killing is observed (5) and when crossed to Sk-1, killing is lost, indicating gain of resistance (6). 5941 and W1432 were used as Sk-1 tester strains, and 4739 and W1446 were used as sensitive tester strains.
for both spore killing and for resistance. These results are consistent with a poison-antidote mechanism for Sk-1, where the presence of a resistance factor at sk1c1 k can rescue any strain from killing, independent of genetic background.
An ORF with Homology to NCU01957 Is Generating the Causative Agent for Spore Killing. We generated transcriptomic data from W1434 (Sk-1), 5940 (sensitive), and a cross between the two strains. We identified a 1,450-bp transcript containing two introns at sk1c1 k in the data from W1434 and from the cross ( Fig. 2E and SI Appendix, Fig. S4). This result indicates expression of this transcript both during sexual development and vegetative growth. The two longest predicted ORFs are just over 400 bp, one of which corresponds to the NCU09865 fragment (ORF2) and the other which shows homology to the N. crassa gene NCU01957 (ORF1). Furthermore, in Neurospora, A-to-I RNA editing is active during sexual development and ascospore formation (30). The transcript contains a single edited site, which converts a tyrosine in ORF1 to a cysteine. In order to determine which ORF causes spore killing, or if both are necessary, we molecularly dissected the transcript. First, we performed RT-PCR using several primer pairs within the transcript, results of which support the existence of a single 1,450-bp locus-spanning transcript with two introns (SI Appendix, Fig. S5). Second, we generated deletion mutants of each ORF (SI Appendix, Figs. S5 and S6). The deletion mutant of ORF1 matched the expected phenotype of a sensitive strain: It showed loss of killing when crossed to a sensitive strain and loss of resistance when crossed to a killer strain. In contrast, the deletion mutant of ORF2 showed loss of killing but no loss of resistance.
We then generated mutants where the start codon (ATG) of each ORF was changed to a leucine (TTG) (SI Appendix, Fig. S7).
Here we observed loss of killing and resistance in the ORF1 mutant (SI Appendix, Fig. S7 B and E) but loss of neither in the ORF2 mutant (SI Appendix, Fig. S7 C and F). We also generated an ORF2 frameshift mutant by inserting a guanine nucleotide immediately after the predicted start codon. Like the ORF2 ATG-to-TTG mutation, the ATG-to-ATGG mutation altered neither spore killing nor resistance to spore killing (SI Appendix, Fig. S7 H and K).
Finally, we created a mutant version of ORF2 called ORF2 [mut5]. This sequence contains five single-nucleotide variants, nine single-nucleotide insertions, seven single-nucleotide deletions, and a 3-nt-long deletion (SI Appendix, Fig. S7M). The goal of ORF2[mut5] was to disrupt any and/or all coding sequences within ORF2. Interestingly, while ORF2[mut5] kills spores, it does so with reduced efficiency (SI Appendix, Fig. S7I and Table  S3). In contrast, ORF2[mut5] appears to be fully efficient with respect to resistance to spore killing (SI Appendix, Fig. S7L).
Our results (summarized in SI Appendix, Table S4) suggest that a single gene exists at the sk1c1 k locus, herein referred to as Spore killer-1 (Spk-1, the gene name starting with an uppercase letter because of its dominant phenotype). A single transcript is produced from Spk-1 and a protein with 134 amino acids is translated from the Spk-1 messenger RNA.
Spk-1 Has Many Different Homologs in Neurospora. The SPK-1 protein contains no predicted functional domains, but an Inter-Pro scan identifies two transmembrane domains, a central cytosolic region and two noncytosolic flanks (Fig. 2F). Searching the National Center for Biotechnology Information (NCBI) nonredundant protein database and FungiDB identifies no similar sequences outside of Neurospora but several hits in N. crassa and Neurospora tetrasperma. Two N. crassa genes show significant sequence identity (SI Appendix, Fig. S8), NCU01957 (35% amino acid identity) and NCU16600 (32% amino acid identity). NCU01957 contains a plasma membrane ATPase domain not found in SPK-1 and mutants are sterile (31). No information on NCU16600 function is available, but it is highly up-regulated late in perithecial development, when ascospores start to be delimited (32). Spk-1 also does not share any detectable homology with any of the other known spore killer genes in Neurospora, Podospora, or Schizosaccharomyces.
We searched genome assemblies for 31 Neurospora strains from seven different species (N. crassa, Neurospora discreta, N. hispaniola, N. intermedia, Neurospora metzenbergii, N. sitophila, and N. tetrasperma; SI Appendix, Table S5) for homologous sequences of Spk-1 using tblastn; 115 hits over 100 amino acids were identified (including NCU01957 and NCU16600) with amino acid sequence identities ranging from 24 to 98.5% (SI Appendix, Fig.  S9). It is not clear to what extent these hits represent actual genes, but in most cases they correspond to potential ORFs of approximately equal or greater length than Spk-1. sk1c1 k Shows Signals of Introgression from N. hispaniola. The homologous sequence that shows the highest sequence identity to Spk-1 (98.5% amino acid identity) is found in N. hispaniola strain 10403 (33). This sequence is not only highly similar to Spk-1 but it also truncates NCU09865 at the same site ( Fig. 3A and SI  Appendix, Fig. S10). A phylogenetic analysis of the NCU09865 fragment present in 10403 and the Sk-1 killers, including alleles from several different species of Neurospora, reveals that all Sk-1 killers group with 10403 but are clearly divergent from all sensitive N. sitophila alleles (Fig. 3B).
The two genes flanking Spk-1 are located in the larger region that is fixed in 21 out of 25 killer strains and a phylogenetic analysis of these reveals that the Sk-1 alleles are also most similar to alleles found in N. hispaniola (Fig. 3C). These sequences cluster not only with sequences from strain 10403 but also with N. hispaniola strain 8817, which does not carry Spk-1 or the truncated NCU09865 sequence. The N. sitophila killers and the two N. hispaniola strains form a monophyletic group clearly separated from the sensitive N. sitophila strains. A phylogenetic analysis based on SNP data in sliding windows across the genome also identifies the sk1c1 k allele as a candidate for introgression from N. hispaniola (Fig. 3D), but on a genome-wide level there is only evidence for very limited gene flow between the two species (SI Appendix, Figs. S11 and S12).
We interpret these results as suggesting that the killer locus has been introgressed into N. sitophila from N. hispaniola. However, we have a limited sample of N. hispaniola strains in our dataset and thus we cannot exclude the possibility that the source is another unknown species, or an unsampled and diverged population of N. sitophila.
N. sitophila Is Split into Three Major Clades. A phylogenetic analysis of whole-genome SNP data reveals clear population structure within N. sitophila and the investigated strains were split into three major clades (Fig. 4A), from now on referred to as clades 1, 2, and 3. Clade 1 has a global distribution with strains from Asia, Oceania, Africa, the Caribbean, and Central and South America. Clade 2 contains strains from Asia, Africa, and Europe, and clade 3 contains strains from Europe, Australia, and the continental United States. All strains from Tahiti, both Sk-1 and sensitives, are placed in clade 1, whereas all sensitive strains from Italy are placed in clade 2 and all Sk-1 strains from Italy are placed in clade 3. In fact, clade 2 is completely fixed for the sensitive genotype and clade 3 for Sk-1. We only find evidence for limited gene flow between the three clades (Fig. 4B), even though strains from clades 2 and 3 were sampled at the same site and at the same time in Italy (27). These strains show no reduced fertility in laboratory crosses, but nevertheless reproduction between strains of these clades appears to be rare in nature. Additionally, one strain of N. sitophila (6850) has been classified as resistant to Sk-1 (28), but based on our analysis it shows comparatively strong differentiation from all other N. sitophila strains, and it may be classified as a different species. Spore Killing Can Be Suppressed through RNAi. We wanted to investigate if the differences in frequency of Sk-1 between the clades could be influenced by the genetic background. We therefore set out to introgress Sk-1 alleles from Italy and Tahiti into sensitive backgrounds from both regions and create isogenic lines by multiple rounds of backcrossing (Fig. 5A). In the first round of crosses, we observed normal levels of spore killing, but when backcrossing first generation offspring (F1) individuals with a Tahitian sensitive parent to that same sensitive parent, we saw a strong reduction in the number of four-spored asci (Fig. 5B). Furthermore, after three generations of backcrossing, all lines generated from Tahitian sensitive parents had either completely lost spore killing or had a large reduction in spore killing (SI Appendix, Fig. S13). For the Italian sensitive lines, all lines displayed normal spore-killing phenotypes. This pattern could be explained by Spk-1 becoming unlinked from a modifying locus that improves killing efficiency, but when the Tahitian F1 individuals were outcrossed to another sensitive strain from either location, the number of four-spored asci was close to normal levels. This pattern suggests that the loss of spore killing is not caused by unlinking a modifier locus but instead that it is an effect of increased relatedness between the killer and sensitive strains. MSUD is a genome defense mechanism identified in Neurospora that operates through the RNAi machinery to posttranscriptionally silence genes that are unpaired during meiosis (34). Theoretically, MSUD should be capable of targeting and inactivating Spk-1 as it is partially unpaired in a killer-sensitive cross; however, MSUD activity has not been evaluated in N. sitophila. MSUD has been shown to have higher efficiency in inbred lines of N. crassa (35), and we hypothesized that loss of killing could be driven by an increased MSUD activity caused by increased relatedness of the killer and sensitive strains. To test this hypothesis, we generated a strain that lacks functional MSUD by deleting sad-2 (36) in the sensitive Tahitian strain 4746. When Sk-1-proficient offspring of 4746 were backcrossed to the sad-2-deficient parent, close to 100% four-spored asci was observed (Fig. 5C and SI Appendix, Figs. S13 and S14), indicating that MSUD is indeed responsible for the loss of killing.

Discussion
We have identified the gene responsible for the Sk-1 spore killer phenotype in N. sitophila. A 1,450-bp transcript generates a single protein product of no more than 134 amino acids which is capable of both killing sensitive sibling spores and of protecting the spores producing the protein from self-killing. We have found no obvious mechanism behind either the killing or resistance phenotypes, but the gene carries predicted transmembrane domains, presenting the possibility that it could disrupt membrane integrity, as has been observed in some bacterial toxin-antitoxin systems (37,38). On the other hand, the wtf genes in Schizosaccharomyces were initially also predicted to contain transmembrane domains (13), but recent work has shown that these are hydrophobic regions which are important for forming large toxic protein aggregates essential for spore killing (39), and a similar mechanism might be at play here as well.
The Spk-1 transcript is a chimera of the 3′ end of NCU09865 and a sequence similar to NCU01957. While the remaining fragment of NCU09865 including ORF2 appears to be wellpreserved, our results indicate that only ORF1 with homology to NCU01957 is translated. Even so, the remains of NCU09865 still play a role, as deleting or heavily mutating ORF2 affects the killing function. The mechanistic basis for loss of killing in ORF2Δ but not ORF2[ATG-to-TTG] or ORF2[ATG-to-ATGG] start-codon mutants is unclear. It could be that the Spk-1 3′ untranslated region (UTR) contains a regulatory element that is required for killing but not resistance. This idea is consistent with our findings for ORF2[mut5], which kills spores with decreased efficiency. The decreased efficiency suggests that one or more of the mutations in ORF2[mut5] lies within an important regulatory sequence of the 3′ UTR. Furthermore, a single site in the Spk-1 transcript experiences A-to-I RNA editing, which converts a tyrosine to a cytosine in 26% of all transcripts. Further work is necessary to determine if it is important for either killing or resistance, but one intriguing hypothesis is that it is RNA editing that makes it possible for the single Spk-1 gene to perform both functions by generating two different proteins. This question could be addressed by for instance constitutively expressing edited and unedited products in vegetative and sexual tissues.
Spk-1 shows no homology to any known spore killer gene and, compared with the wtf genes in Schizosaccharomyces and the Spok genes in Podospora, Spk-1 is significantly smaller (9,10,12,13). On the other hand, ORF1 appears to belong to a family of   highly diverse Neurospora-specific genomic sequences. It is still unclear to what extent these sequences are protein-coding, but among them we found two annotated genes in the N. crassa reference genome. NCU01957 is located very close to the mating-type locus of N. crassa, but is not found at this location or elsewhere in most other investigated Neurospora genomes. Exceptions are N. hispaniola, where it is found at the same locus, and N. sitophila, where a highly similar gene is found at a distance of ∼20 kbp. NCU01957 is of unknown function, but a strain carrying a single-amino acid mutation displayed both stunted vegetative growth and total sterility (31). NCU16600 is also of unknown function, but it is up-regulated during ascosporogenesis (32). The highly diverse set of homologous sequences is reminiscent of the wtf genes in Schizosaccharomyces, with complex phylogenetic patterns and extensive presence-absence polymorphisms (14). Further work is necessary to determine if these are genes with a regular function for normal sexual development, or if they could be selfish genetic elements that have either gone to fixation or are controlled by some suppressing mechanism. The sk1c1 k allele found in N. sitophila Sk-1 strains is highly diverged from sensitive N. sitophila strains and together with the flanking genes shows greater similarity to the closely related species N. hispaniola. Strain 10403 carries a copy of Spk-1 that shows both high sequence and structural similarity to the one found in Sk-1 strains, suggesting that it may also be able to cause spore killing. Only four strains of N. hispaniola have ever been collected (33), and we have not been able to observe spore killing when crossing these to each other or in hybrid crosses to N. tetrasperma. This could mean that Spk-1 in 10403 is no longer functional, but it can also indicate that the strains available carry an unknown resistance factor and further sampling would be necessary to assess whether Spk-1 is actively driving within the species.
While sk1c1 k appears to be introgressed from N. hispaniola, there is little sign of introgression elsewhere in the genome. N. sitophila and N. hispaniola show high levels of reproductive isolation with each other and we have so far not observed viable offspring in any crosses between the species. While selfish genetic elements such as meiotic drivers have often been argued to contribute to restricting gene flow between species (17,18), here we see the opposite result, where meiotic drive instead may have facilitated a limited exchange of genetic material. Similar results have been observed in D. simulans and Drosophila mauritiania where the Winters sex ratio element has crossed between species despite strong reproductive barriers (21). At this time, we cannot exclude that Sk-1 entered N. sitophila through a yet-undiscovered species or population of Neurospora that shows greater sexual compatibility with both N. sitophila and N. hispaniola. We also do not know if Spk-1 evolved in N. hispaniola or if it originated in another species and has spread within Neurospora through multiple introgressive steps.
Using whole-genome SNP data, we can show that N. sitophila is split into three clades which show remarkably different patterns of spore killing. While 15% of all N. sitophila strains carry the Sk-1 element, these are not evenly distributed within the clades. Clade 3 is fixed for Sk-1, clade 2 is fixed for sensitives, and clade 1 contains both genotypes. This population structure is especially puzzling as strains from clades 2 and 3 were sampled at the same location at the same time and show no signs of reduced fertility when crossed under laboratory conditions but still show only very limited signatures of gene flow. Some mechanism must prevent individuals from these two clades from crossing in nature, but unfortunately we currently know too little about the ecology and life cycle of N. sitophila to have a good understanding of what this might be.
In clade 1, we find both Sk-1 and sensitive strains. It is unclear if Sk-1 has recently been introduced into the clade and is on its way to fixation or if it is maintained as a stable polymorphism due to unknown fitness costs of carrying the spore killer element. One tantalizing suggestion is that the Sk-1 polymorphism can be explained by the observed suppression through MSUD. While we only observe strong suppression in highly isogenic crosses, we do not know how common such crosses are in nature. If inbreeding deletion on the ability of backcrossed strains to kill. In a cross between a sensitive strain from Tahiti (4746) and a killer F1 from a cross between that strain and a killer strain (4738), killing is strongly reduced. Deleting the sad-2 gene inactivates MSUD, and in a cross between 4746 Δsad-2 and the killer F1, spore killing is restored to normal levels. This indicates that MSUD is responsible for suppressing spore killing in inbred crosses with a genetic background from Tahiti. This pattern is not seen if the sensitive strain comes from Italy.
is a common pattern in N. sitophila and MSUD is more efficient between inbred strains, MSUD may be sufficient enough to keep Sk-1 at a low frequency. It is also possible that MSUD is more efficient under natural growing conditions where environmental factors may promote a higher activity. RNAi has been shown to be essential for suppressing sexlinked meiotic drive in D. simulans and is suspected to be involved in other Drosophila species as well (16,40,41). In D. simulans the sex ratio driver Dox is suppressed by two genes named Nmy and Tmy (16,42,43). Both genes are partial duplicates of Dox and generate small RNAs which silence Dox expression and prevent inactivation of Y-bearing sperm. The production of small RNAs has also been observed from duplicated genes that are suspected to cause sex ratio distortion in Drosophila miranda (41).
Here we show that the RNAi-based MSUD system can act as a suppressor of spore killing in highly isogenic crosses. Transcriptional silencing through MSUD is not dependent on gene duplications but can act directly on genes that are unpaired at meiosis. Silencing through gene duplication may actually not be possible in Neurospora, as duplicated sequences are targeted and mutated by the RIP (repeat induced point mutations) system, which will introduce high levels of C-to-T mutations in larger sequences found at least twice in a genome (44). Both MSUD and RIP are thought to defend against selfish genetic elements such as transposable elements, and MSUD should at least in theory be capable of defending against invading meiotic drivers that are sufficiently divergent to not be able to pair at meiosis (45). It is interesting to note that both Spk-1 in N. sitophila and rfk-1, which is the causative factor behind the Sk-2 spore killer element in N. intermedia (26), manage to evade MSUD in most crosses. In the case of rfk-1, the small size together with proximity to pairing DNA appears to protect it from being silenced (26), and in the case of Spk-1 a potentially higher MSUD efficiency in isogenic crosses is enough to tip the balance against the spore killer gene. We can hypothesize that MSUD is often a sufficient defense against most KMDs which try to invade a population of Neurospora, and the only ones that are able to persist or go to fixation are the rare cases that have some feature that allows them to evade genome defense mechanisms such as MSUD.
Most of the killer meiotic drive elements that were initially discovered and which have been studied in greatest detail are large multigene systems found in regions of suppressed recombination (5)(6)(7)(8). It has sometimes even been stated that meiotic drive can only be caused by the interaction of multiple loci and that the association with regions of low recombination is a defining characteristic of the phenomenon. Spk-1 joins a growing number of recently discovered single-gene KMDs which contradict this view. In fungi, four out of five known meiotic drive gene families are caused by single genes which can cause drive on their own (9,11,13,26). All of these gene families appear to have evolved independently, suggesting that there is ample potential for de novo evolution of single-gene KMDs. But if this is the case, why have we not found more of these systems? Single genes are less likely to be tightly linked to deleterious mutations, which could make them more likely to go to fixation and become invisible in crosses. They are also less likely to be tightly linked to visible phenotypes that would allow easy detection. Although spore-killing phenotypes can be relatively easy to detect, the genomic techniques used here can also be used to detect segregation distortion in the absence of visible phenotypes. Indeed, we expect that future work combining genomic techniques, controlled crosses, and sampling of more nonmodel organisms will unmask a vast diversity of cryptic meiotic drivers and shed new light on meiotic drive as a major factor in the evolution of species.

Methods
All natural isolates used in this study are listed in SI Appendix, Table S2. Strains were ordered from the Fungal Genetics Stock Center (46) or provided by D.J.J. (27). We verified the phenotype of all strains by crossing them to both Sk-1 and sensitive tester strains and observing patterns of dead and living spores at 500× magnification. This method was also used to determine the efficiency of spore killing among the introgressed lines generated between Tahitian and Italian strains for the MSUD analysis.
Whole-genome paired-end Illumina HiSeq sequencing data were collected from 56 N. sitophila strains (SI Appendix, Table S2), and long-read PacBio data were collected from 4 strains (SI Appendix, Table S1). Transcriptomic data were also collected by paired-end Illumina sequencing of RNA from vegetative tissue of the Sk-1 strain W1434 and the sensitive strain 5940, and sexual tissue was obtained from a cross between the two. We assembled the PacBio data de novo using HGAP 3.0 (47), and checked completeness and synteny by aligning the assemblies to the N. crassa reference genome. We then called variants by mapping the Illumina reads to the PacBio assembly of strain W1434, and calculated the association of each SNP with the killing phenotype using Fisher's exact test. A transcriptome was also assembled for W1434 by mapping RNA-sequencing (RNA-seq) reads to the PacBio assembly with STAR (48) and calling transcripts using Cufflinks (49).
We tested whether sk1c1 showed segregation distortion by evaluating 46 progeny from a cross between strains W1426 and W1446 for the abundance of each parental allele. We developed a PCR protocol that produces an ∼900-bp product in W1446 (killer locus) and an ∼600-bp product in W1426 (sensitive locus) and assessed the genotype of the progeny with gel electrophoresis. All genetic transformations were performed by electroporation and gene-deletion and transgene-insertion vectors were constructed by double-joint PCR.
Homologs of Spk-1 were identified with BLASTp against the NCBI nonredundant protein database, FungiDB, and 31 additional high-quality Neurospora assemblies. Hits longer than 100 bp were extracted and a phylogeny was inferred using IQ-TREE (50). We tested the hypothesis of introgression by extracting the fragment of NCU09865 found in Spk-1 from 177 Neurospora assemblies, as well as the two neighboring genes NCU09864 and NCU09866, and generating phylogenies using RAxML (51). A whole-genome phylogeny of all 56 N. sitophila genomes was also generated from SNP data with RAxML. We searched for a genome-wide signal of introgression using a sliding window phylogenetic approach with Twisst (52), and evaluated introgression and between-clade gene flow with ADMIXTURE (53).
Full methods are included in SI Appendix, Methods.
Data Availability. Raw reads reported in this article have been deposited in the Sequence Read Archive at the NCBI (under PRJNA649678). Genome assemblies, genome annotations, SNP data in variant call format, and alignments are available at FigShare (https://doi.org/10.6084/m9.figshare.c.5093585.v1) (54). All natural N. sitophila isolates used in this study that were previously not publicly available, together with key mutants generated for this study, have been deposited at the Fungal Genetics Stock Center. Scripts used for data analysis are available at https://github.com/johannessonlab/sitophila_spore_killer/.