Killer phenotype assays

Killer toxin production by yeasts was measured using killer yeast agar plates (yeast extract, peptone, and dextrose (YPD) agar plates with 0.003% w/v methylene blue buffered at pH 4.6), as described previously [32]. Toxin production was identified by either a zone of growth inhibition or methylene blue-staining of the susceptible lawn yeasts. The pH optima of killer toxins produced by different yeasts was measured on killer yeast agar plates adjusted to pH values of 4.0, 4.5, 5.0 and 5.5.

Killer toxin enrichment

Strains of killer yeast were grown in 2 mL of YPD medium (pH 4.6) overnight at room temperature with vigorous shaking (250 rpm). The culture was centrifuged at 3,100 × g for 5 min followed by filtration of the culture medium through a 0.22 μm filter. Filtered growth medium was added 1:1 with 4°C supersaturated ammonium sulfate solution and mixed by inversion and incubated on ice for 3 h. The precipitated proteins were collected by centrifugation at 20,800 × g for 10 min at 4°C. The supernatant was then removed, and the precipitated proteins suspended in 10 μL of YPD pH 4.6. Killer toxins were incubated either at room temperature, or heat-inactivating at 98°C for 2 min before treating lawns of susceptible yeasts spread onto killer assay agar plates.

RNA extraction and purification

Double-stranded RNAs were extracted as described previously [32], with the following modifications: The extracted dsRNAs were not incubated with oligo d(T)25 magnetic beads and 2× LTE buffer was replaced with 2× STE (500 mM NaCl; 20 mM Tris-HCl, pH 8.0; 30 mM EDTA, pH 8.0). Single-stranded RNAs were extracted from yeasts using the RNeasy RNA extraction kit (Qiagen) with the bead beating method and an on column DNase I digestion. After column elution, an additional incubation with DNase I (Omega Bio-Tek) was performed at 37°C for 10 min followed by 75°C for 10 min when the removal of additional contaminating DNAs was required.

Curing Saccharomyces yeasts of dsRNAs

Cycloheximide, anisomycin, or high temperatures were used to create strains of yeasts that lacked satellite dsRNAs. Yeasts were first grown overnight in 2 mL of liquid YPD medium before 1 μL of cell suspension was transferred to YPD agar with either cycloheximide (0.4–5.0 μM) or anisomycin (0.8 μM). Yeast cultures were incubated for 2–5 days at 23°C to recover surviving cells. Curing satellite dsRNAs using temperature involved incubating yeast cultures on YPD agar for 2–5 days at 30°C, 37°C, or 40°C. Growing cells were streaked onto YPD agar plates from the heat-treated agar plates and were incubated for an additional 2–3 days at 23°C. The colonies resulting from chemical or temperature treatment were analyzed for the loss of killer toxin production by replica plating onto killer assay agar plates seeded with a killer toxin sensitive yeast strain.

Short read sequencing of satellite dsRNAs

The preparation of dsRNAs, cDNAs, Nextera Illumina libraries, and sequence analysis was the same as previously reported, with several amendments detailed below [32]. Reads were cleaned with fastp or HTStream (https://github.com/ibest/HTStream) and de novo contigs were assembled with SPAdes assembler v3.11.1 using default parameters [52,53].

Sanger sequencing of SpV-M1L

Reverse transcriptase PCR (RT-PCR) was used to generate overlapping DNAs that represented the genetic sequence of the satellite dsRNA named SpV-M1L. The approximate molecular weight of these DNAs was determined by agarose gel electrophoresis and capillary electrophoresis (Fragment Analyzer, Agilent Technologies Inc). DNAs were cloned using the pCR-Blunt II-TOPO vector and subjected to Sanger sequencing. 5’ and 3’ RACE were used to determine the terminal ends of the dsRNA molecules using the protocol provided by the manufacturer (Invitrogen) (see S4 File for the full list of primers).

Cloning of genome-encoded killer toxin genes

Genomic DNAs were extracted from K. africana, N. dairenensis, N. castellii, T. phaffii, and P. membranifaciens using the method of Hoffman and Wilson (1987) and were used as templates for PCR (see S4 File for the full list of primers) [54]. Killer toxin genes were cloned into pCR8 by TOPO-TA cloning (Thermo Fisher) and the DNA sequences were confirmed by Sanger sequencing. The K1L gene was commercially synthesized (GeneArt by Thermo Fisher) and used as a PCR template to amplify K1L. The PCR-derived K1L gene was cloned into pCR8 by TOPO-TA cloning and confirmed via Sanger sequencing. All killer toxin genes were sub-cloned using Gateway technology into the destination vector pAG426-Gal-ccdB for ectopic expression in either S. cerevisiae or S. paradoxus (see S5 file for all plasmid sequences) [55,56].

Expression of K1L and homologs from the Ascomycota

For ectopic expression of killer toxins, plasmids encoding toxin genes were used to transform either S. cerevisiae BY4741 or a non-flocculant derivative of S. paradoxus A12 (named A12C) [57]. Transformants were selected on complete medium lacking uracil. To assay toxin expression, a single colony of each transformed strain was used to inoculate a series of consecutive overnight cultures in 2 mL of complete medium lacking uracil, first with dextrose, then raffinose, and finally galactose at 30°C with shaking (250 rpm). The optical density of the final 2 mL culture was normalized to an OD₆₀₀ of 1 and 1 mL was centrifuged at 3,000 × g for 5 min. The supernatant was removed, and the cell pellet was disrupted by gentle agitation. 2.5 μL of the resulting cell slurry was used to inoculate YPD and YPG plates (with 0.003% w/v methylene blue, pH 4.6) seeded with a killer toxin-susceptible yeast strain. Inoculated plates were incubated for 48–72 h at 25°C until killer toxin production was visible.

Phylogenetic analyses

Killer toxin gene sequences were aligned using MUSCLE and manually trimmed to represent the most confident alignment of the α-domain. MEGA (version 7) was used for phylogenetic analysis using neighbor-joining and maximum likelihood methodologies. The optimal model for amino acid substitution was determined as the Whelan and Goldman model with a gamma distribution. 500 bootstrap replicates were used to construct a phylogenetic model with the highest log-likelihood.

The identification of new strains of killer toxin-producing yeasts

A total of 110 strains of Saccharomyces yeasts were obtained from the USDA Agricultural Research Service (ARS) culture collection and screened to identify the production of novel killer toxins. The first screen used eight yeasts from four different species to indicate toxin production on “killer assay medium”, which is a complex YPD medium buffered to pH 4.6, the optimal pH for most Saccharomyces killer toxins. Methylene blue was also added to the medium as a redox indicator of cell death. This assay found that 22% (24/110) of yeasts could inhibit the growth of at least one strain of yeast (S1 File). To identify the types of killer toxins based on their unique spectrum of activities, 13 killer yeasts were further screened against 46 indicator lawn yeasts. Four strains of S. cerevisiae that have been previously described to produce killer toxins of unknown types were also included (NCYC1001, NCYC190, CYC1058, and CYC1113) [58–60]. To facilitate the classification of different toxin types, yeasts that produce canonical K1, K28, and K74 killer toxins, and a non-killer yeast were included for comparison (Fig 1). The degree of growth inhibition by each killer yeast was scored qualitatively based on the appearance of zones of growth inhibition and methylene blue staining of the surrounding indicator strain on agar plates (Fig 1 and S2 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The strain- and species-specificity of killer toxins produced by Saccharomyces yeasts.

A total of 21 killer yeasts were assayed for killer toxin production on agar plates seeded with 46 different lawn yeasts. Yeast from the USDA (ARS) culture collection are designated by either ‘Y’ or ‘YB’. Killer toxin activity was qualitatively assessed based on the presence and size of zones of growth inhibition and/or methylene blue staining around killer yeasts as diagrammed (bottom left). Darker colors on the cluster diagram represent a more prominent killer phenotype with yellow indicating no detectable killer phenotype. The non-killer yeast strain S. cerevisiae 1116 was used as a negative control (bottom right). Results were analyzed using the R package gplots to cluster killer yeasts and lawn yeasts based on killer toxin production and susceptibility, respectively. S. cer—S. cerevisiae, S. par–S. paradoxus, S. bay–S. bayanus, S. kud–S. kudriavzevii, S. mik–S. mikatae, N. dair–N. dairenensis, N. cas–N. castellii, and K. afr–K. africana.

https://doi.org/10.1371/journal.pgen.1009341.g001

The known strain- and species-specificity of killer toxins have been previously used to biotype different yeasts [61–65] and to identify groups of unknown killer toxins [3,66]. Therefore, cluster analysis was used to group killer yeasts based on their ability to inhibit growth of lawn strains and revealed that no two strains had the exact same spectrum of antifungal activity. S. cerevisiae CYC1113, CYC1058, and NCYC1001 had the broadest antifungal activities but did not cluster with any of the known killer toxins. Clustering found that that the antifungal specificity of S. cerevisiae NCYC190 was most closely correlated with that of a K1 killer yeast, with 85% identical interactions with competing lawn strains. The killer toxin activity of S. paradoxus Y-63717 also clustered with K1 (50% identical interactions), showing both gain and loss of activity. This result was particularly intriguing because evidence of K1 production by S. paradoxus is inconsistent between different research groups with recent studies suggesting that K1 toxins are unique to S. cerevisiae [8,67,68]. The remaining killer yeasts were identified to have weaker inhibitory activities with no clustering with K1, K28, or K74 killer yeasts.

Killer yeasts harbor satellite dsRNAs that encode proteinaceous killer toxins

The production of killer toxins by yeasts is often associated with the presence of satellite dsRNAs that are maintained by totiviruses [25,45]. To determine whether satellite dsRNAs were responsible for the observed killer phenotypes, killer yeasts were treated with either cycloheximide, anisomycin, or incubated at elevated temperatures to select for the loss of killer toxin production [69,70]. The majority of killer yeast (12/19) lost their killer phenotype after exposure to chemical or thermal insult (Fig 2A). The same treatments were unable to select for the loss of the killer phenotype in the remaining seven strains, which suggests that their killer toxins were genome-encoded (S1 Fig). To determine if the killer phenotype was correlated with the presence of dsRNAs, cellulose chromatography was used to selectively purify dsRNAs from the killer yeasts [32]. The analysis of extracted dsRNAs revealed that the loss of the killer phenotype was 100% correlated with the loss of satellite dsRNAs (Fig 2A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Analysis of dsRNAs present within killer yeasts and the properties of their killer toxins.

(A) The extraction and analysis of dsRNAs from totiviruses and associated satellites in Saccharomyces yeasts that produced killer toxins (top panel) or lost the killer phenotype after exposure to cycloheximide, anisomycin, or elevated temperatures (bottom panel). S. cerevisiae Y-27788 is included as a representative killer yeast that lacks a satellite dsRNA. (B) The relative pH optimum of killer toxins against three different indicator lawn yeasts S. cerevisiae YB-4237, DSM70459, and NBRC10991 using YPD buffered to pH 4.0, 4.5, 5.0, and 5.5. (C) Enrichment and concentration of killer toxins from spent growth medium by ammonium sulfate precipitation and their loss of inhibitory activity after incubation at 98°C.

https://doi.org/10.1371/journal.pgen.1009341.g002

The dsRNA-encoded killer toxins identified have a pH optimum mostly between 4.5 and 5, with no inhibitory activity at pH 5.5 (Fig 2B). To confirm that the identified killer toxins are proteinaceous, each was purified by ammonium sulfate precipitation and used to challenge susceptible yeasts. Zones of inhibition were clearly visible on confluent lawns of yeast cells for all of the killer toxins tested but not for a non-killer yeast or a strain cured of its dsRNA satellite (Fig 2C). The inhibitory activities of these precipitates were heat-labile, and the toxicity was lost after incubation at 98°C for 2 minutes. The identified killer toxins have similar biochemical characteristics to known proteinaceous killer toxins despite their differing inhibitory effects towards yeasts.

The discovery of a new killer toxin produced by S. paradoxus

To identify the unknown killer toxins produced by killer yeasts, dsRNAs were purified and subjected to a short-read sequencing pipeline for dsRNAs [32]. BLASTn analysis of de novo assembled contigs revealed that the satellite dsRNAs within strains CYC1058 and NCYC1001 encode canonical K2 toxins and NCYC190 a canonical K1 toxin (S2 Fig and S1 Table). The sequence reads derived from the dsRNAs of Y-63717 assembled into 125 different contigs, with six >750 bp in length and a coverage score >1,000 (Fig 3A). BLASTn analysis of these high-quality contigs identified the dsRNA genome of the totivirus L-A-45 from S. paradoxus N-45 with 100% coverage and 95.5% nucleotide identity [36]. However, the remaining contigs did not match the nucleotide sequence of any known Saccharomyces yeast killer toxin or dsRNA satellite. Short-read sequencing was augmented with a combination of 5’ and 3’ RACE, RT-PCR, Sanger sequencing, and capillary electrophoresis to assemble the complete sequence of the dsRNA satellite from Y-63717 (Figs 3B and S3). The novel satellite dsRNA is approximately 2,371 bp in length with a single open reading frame (ORF) that encodes a protein of 340 amino acids. The 5’ ORF is positioned upstream of a central poly(A) tract of ~220 bp (Figs 3B and S3). The 5’ terminus has a nucleotide sequence of 5’-GAAAAA that is found in many satellite dsRNAs (Figs 3B and S3) and is predicted to fold into a large stem-loop structure (S4 Fig). Downstream of the poly(A) tract in the 3’ UTR there are elements of secondary structure that are indicative of replication (terminal recognition element; TRE) and packaging signals (viral binding site; VBS) that have been well characterized in the canonical M1 satellite dsRNA from S. cerevisiae (ScV-M1) (Figs 3B and S4) [71–73]. In addition to RNA secondary structures, there are also two direct repeats of the sequence motif named “Downstream of Poly(A)” (DPA; 5’-CTCACCYTGAGNHTAACTGG-3’) that is found in different satellite dsRNAs isolated from S. paradoxus (M45, M74, and M62), S. cerevisiae (M1 and Mlus), Zygosaccharomyces bailii (MZb), and Torulaspora delbrueckii (Mbarr) (Figs 3B and S3) [27,74].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Short-read sequencing and analysis of the K1L killer toxin from S. paradoxus Y-63717.

(A) Sequence contigs after de novo assembly of sequence reads represented by contig coverage score and contig length. BLASTx analysis was used to annotate contigs as similar to totiviruses or satellite dsRNAs. (B) Schematic of the organization of the SpV-M1L satellite dsRNA (C) Jpred secondary structure prediction of the K1 and K1L killer toxins with confidence score plotted as a histogram. Predicted domain boundaries are drawn below the secondary structure prediction [76]. Signal peptidase cleavage site were predicted by SignalP and TargetP (S3 File). (D) Ectopic expression of K1L from a plasmid in the non-killer yeast S. paradoxus A12 induced by the presence of galactose in the growth medium.

https://doi.org/10.1371/journal.pgen.1009341.g003

The length and positioning of the 5’ ORF of the satellite dsRNA in S. paradoxus Y-63717 suggested that it encoded a killer toxin (Fig 3B). A PSI-BLAST search of the NCBI database with two iterations found that this putative killer toxin had weak homology to the canonical K1 toxin from S. cerevisiae (99% coverage, 18% amino acid identity, e-value 4 × 10⁻¹⁷) (Fig 4A and S2 Table). Based on this homology, the putative killer toxin was named K1L (K1-like) and the dsRNA satellite was named Saccharomyces paradoxus virus M1-like (SpV-M1L). The organization of the functional domains of K1L appears to be similar to K1 based on the secondary structure, conserved cysteine residues, and the predicted signal peptidase cleavage sites (Fig 3C) [75]. K1L contains ten cysteine residues, two of which are likely important for interchain disulfide linkage (Cys91) and killer toxin immunity (Cys257) based on their alignment with cysteines from K1 [41]. To confirm that the K1L is an active killer toxin, it was ectopically expressed by the non-killer strain S. paradoxus A12C using a galactose inducible promoter. A well-defined zone of growth inhibition was visible when the strain was grown on galactose-containing medium (Fig 3D). No K1L toxin expression was observed when cells were plated on dextrose-containing growth medium. Together, these data confirm the identification of a new dsRNA satellite in S. paradoxus and a novel killer toxin related to K1.

K1L homologs are found in yeasts of the Saccharomycotina

The PSI-BLAST search that identified K1L as a homolog of K1 also identified 24 hypothetical “K1-like Killer Toxin” (KKT) genes in six diverse species of non-Saccharomyces yeasts from the subphylum Saccharomycotina (Figs 4A and S5). The species Kazachstania africana, Naumovozyma castellii, Naumovozyma dairenensis, Tetrapisispora phaffii, and Pichia membranifaciens encode KKT genes that are homologous to K1L (aligned >300 amino acids, e-value <10⁻⁸⁰) and represent yeasts from the families Saccharomycetaceae and Pichiaceae (Fig 4A and S3 Table) [77]. Importantly, genomic KKT genes appear to be unique to these particular species and absent from other related yeasts (i.e. Kazachstania naganishii, Tetrapisispora blattae, and Pichia kudriavzevii). The length of all Kkt proteins are between 153–390 amino acids with 11 being similar in length (340 amino acids) (Fig 4A) and amino acid identity (16–28%) to K1L (S2 Table). In addition, BLASTn was used to identify 14 additional pseudogenes that, in some species, outnumber intact KKT genes (Fig 4B and S3 Table). All KKT genes and related pseudogenes are found in multiple copies that vary in frequency between different yeast species and are mostly located within subtelomeric regions (within ~20 kb of the assembled chromosome ends) (Fig 4B). Of the 38 KKT genes and pseudogenes found within six different species, only two are positioned away from the subtelomeric regions in the yeasts N. dairenensis and N. castellii (Fig 4B). Analysis of the chromosomal position of these two genes revealed that their insertions are unique to each species, absent from other related species at the syntenic chromosomal location, and inserted close to tRNAs (S6 Fig). It was noted that six of the KKT genes and pseudogenes from K. africana contained a characteristic “GAAAAA” sequence motif close to the start codon of each ORF, similar to K1L and other killer toxins (S6 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. KKT genes are distributed across the subphylum Saccharomycotina and are primarily subtelomeric in their genomic location.

(A) PSI-BLAST analysis of the killer toxin encoded upon the satellite dsRNA from S. paradoxus Y-63717. (B) Location of KKT genes and pseudogenes on the linear representation of the chromosomes of four yeasts of the Saccharomycetaceae. Chromosome I of N. castellii did not contain any KKT genes and was truncated in the diagram due to its large size. (C) Unrooted maximum likelihood phylogeny of the aligned α-domain of 21 killer toxins from six species of yeast and one dsRNA satellite. Numerical values represent the bootstrap support for the placement of each node. The domain organization of each protein is illustrated and annotated based on the four-domain structure of K1. Signal peptidase cleavage site predicted by SignalP and TargetP (S3 File). Underlined killer toxin names represent those that were cloned and functionally tested for antifungal activity. *Killer toxins with confirmed antifungal activities.

https://doi.org/10.1371/journal.pgen.1009341.g004

To ascertain the evolutionary history and relatedness of K1L and its homologs, a multiple amino acid sequence alignment was constructed. The most confident alignment was achieved between the putative α-domain of each protein and included eight truncated proteins with premature stop codons. Phylogenetic analysis was performed using maximum likelihood (Fig 4C) and neighbor-joining (S7 Fig) methodologies with 500 bootstrap iterations as implemented by MEGA [78]. The tree topologies of the two phylogenetic models are in general agreement and show the distinct clustering of killer toxins from each species (Figs 4C and S7). Importantly, K1L is most closely related to the Kkt1 proteins from T. delbrueckii and P. membranifaciens. The Kkt proteins from K. africana form a monophyletic clade that contrasts those from T. phaffii and N. dairenensis that appear to be polyphyletic. There also appears to be some support for a common ancestry of killer toxins within the Naumovozyma yeasts. BLASTn analysis and nucleotide alignment of the 5’ and 3’ UTRs of the KKT genes from K. africana indicates that flanking nucleotide sequence is between 83–94% identical over ~2,000 bp, suggesting gene duplication. There is also evidence of gene duplication of KKT genes within N. dairenensis based on their close phylogenetic relationships and similar untranslated regions. To ascertain the evolutionary trajectory of KKT paralogs, the rate of accumulation of nonsynonymous (dN) and synonymous (dS) mutations were calculated for KKT paralogs that could be confidently aligned by their nucleotide sequences. When dN/dS was calculated for all codons over the evolutionary history of the sequence pairs, all three pairs of paralogs have evolved under purifying selection since their duplication (dN/dS = 0.72, 0.41, 0.23) [79]. A domain-resolution approach using a sliding window to calculate dN/dS supports the predominance of purifying selection with a few localized peaks of possible weak positive selection (dN/dS >1) in the α- and/or γ-domains (S8 Fig).

A new family of K1-like antifungal killer toxins

The relatedness of the KKT genes to the killer toxin K1L suggests that they encode antifungal killer toxins. To assay for killer toxin production by the yeasts that encode KKT genes, P. membranifaciens, N. dairenensis, N. castellii, T. delbrueckii, T. phaffii, and K. africana were used to challenge 19 different yeast on killer assay plates. With the exception of N. castellii, all of the KKT-encoding yeast species produced killer toxins that caused growth inhibition of at least one other yeast (Fig 5A and 5B). Each of these killer yeasts was also immune to its own killer toxin, but susceptible to those produced by other KKT-encoding yeasts. The production of killer toxins by these species is consistent with the previously reported killer activity of P. membranifaciens, T. delbrueckii, and T. phaffii [27,80,81]. There was no evidence of satellite dsRNAs in any of the KKT-encoding yeasts, except for the detection of an unknown high molecular weight dsRNA within P. membranifaciens NCYC333 (S9 Fig). The differences in killer toxin production by five strains of P. membranifaciens suggested that there could be strain-specific polymorphisms in KKT genes (S10 Fig). The published genome sequence of P. membranifaciens Y-2026 revealed a large central deletion in the γ-domain of its KKT gene (S10 Fig). Sanger sequencing of the same KKT gene from P. membranifaciens Y-2026 acquired directly from the NRRL culture collection failed to identify the same deletion, instead there was an indel within the γ-domain that caused the truncation of the killer toxin gene (S10 Fig). Sequencing of the KKT gene from P. membranifaciens NCYC333 confirmed a full-length gene that correlated with robust killer toxin production by the strain. However, P. membranifaciens Y-2026 was still able to express killer toxins, suggesting the production of other alternative antifungal molecules.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The antifungal activity of Kkt toxins from K. africana, N. dairenensis, and P. membranifaciens.

(A) KKT-encoding yeasts were assayed for killer toxin production on killer assay agar plates seeded with indicator strains. Killer toxin production was judged by the presence of zones of growth inhibition or methylene blue staining (see the key in Fig 1). (B) Representative pictures of killer toxin production by different species of yeast from the Saccharomycotina. (C) PCR detection of KKT gene DNA and RT-PCR detection of RNA transcripts under laboratory conditions. S. paradoxus Y-63717 served as a dsRNA control. (D) Galactose-dependent ectopic expression of KKT genes from K. africana, P. membranifaciens and N. dairenensis can inhibit the growth of K. africana. N. dai–N. dairenensis, N. cas–N. castellii, K. afr–K. africana, P. mem–P. membranifaciens, T. pha–T. phaffii, S. cer–S. cerevisiae, S. mik–S. mikatae, S. par–S. paradoxus, and T. del–T. delbrueckii.

https://doi.org/10.1371/journal.pgen.1009341.g005

Although P. membranifaciens, N. dairenensis, T. delbrueckii, T. phaffii, and K. africana are killer yeasts (Fig 5A and 5B), it was unclear whether KKT genes were directly responsible for the observed production of killer toxins. Indeed, T. phaffii has been reported to express an antifungal glucanase and K2 killer toxin-related genes have been found in the genome of K. africana [74,80]. To confirm the expression of KKT genes in killer yeasts, total genomic DNA and RNA were extracted from six species of killer yeasts and used as templates for PCR and RT-PCR with primers targeting ten different KKT genes (S11 Fig). Due to the similarity of the KKT genes it was not possible to unambiguously distinguish between transcripts from TpKKT1 and TpKKT2, or KaKKT1 and KaKKT2. Of the six species assayed, KKT RNA transcripts were detected in T. phaffii, N. dairenensis, and K. africana indicating active gene transcription under laboratory conditions (Fig 5C). To demonstrate that KKT genes are active killer toxins with antifungal activities, a subset was cloned into a galactose inducible expression vector (labelled in Fig 4C). Active killer toxin production from a non-killer strain of S. cerevisiae transformed with KKT genes was assayed against eight lawns of yeasts using galactose containing agar plates. The majority of KKT genes did not cause any noticeable growth inhibition or methylene blue staining of competing yeasts when expressed from S. cerevisiae (S12 Fig). However, killer toxins from P. membranifaciens NCYC333 (PmKKT1), K. africana (KaKKT2), and N. dairenensis (NdKKT1) were able to create visible zones of growth inhibition (Fig 5D). No growth inhibition was observed when gene expression was repressed by plating cells on dextrose. Altogether, these data show that KKT genes are transcriptionally active and encode active killer toxins. This confirms the discovery of a new family of K1-like genome-encoded antifungal proteins in the Saccharomycotina.

Horizontal acquisition and copy number expansion of killer toxins in fungi

KKT genes have most likely been acquired by horizontal gene transfer because of their sporadic distribution and lack of common ancestry in closely related yeast species. Moreover, the lack of relatedness of KKT genes in these species suggests that they were acquired independently. Fungi are known to acquire foreign DNAs from other species of fungi [14–18,82–84] as well as bacteria [14,85,86]. The interspecific capture of DNAs derived from retrotransposons, viruses, and plasmids has also been observed [19,20,87,88]. Specifically, genome integrated copies of dsRNA-encoded killer toxins homologous to KP4, K1, K2, Klus, and Kbarr have been found within bacteria and fungi, indicating gene flow between dsRNAs and DNAs across taxonomic divisions [74,89]. However, the vast majority of these putative killer toxins are uncharacterized, and it remains unclear as to whether they are biologically active.

Phylogenetic evidence suggests that cross-species transmission of viruses between fungi has occurred on multiple independent occasions [90,91]. Laboratory experiments have also successfully demonstrated extracellular [92,93] and interspecific virus transfer [36]. In particular for Saccharomyces yeasts, mating and hybridization between different species has been observed, and is a mechanism for gene introgression, as well as for the acquisition of retrotransposons and plasmids. The close association of many yeast species in natural and anthropic habitats may increase the likelihood of horizontal gene transfer or invasion by dsRNA viruses and satellites [94–96]. Specifically, the satellite dsRNA that was identified in this study (named SpV-M1L) and an unrelated satellite dsRNA (SpV-M45) are both found within sympatric Far Eastern yeast strains of S. paradoxus [36,97]. The parasitism of L-A-45 by both of these satellite dsRNAs in different strains of yeast strongly implies that they were acquired by horizontal gene transfer. The apparent structural similarities of the satellite dsRNA-encoded VBS elements suggest that, despite the evolutionary distance between SpV-M1L and SpV-M45, each has evolved a similar mechanism to enable the hijack of the replication and packaging machinery from the L-A-45 totivirus. Unlike Saccharomyces yeasts, the presence of active RNAi within the KKT-encoding yeasts would hinder the horizontal acquisition of dsRNA viruses from other yeasts, which could explain the abundance of genome-encoded killer toxins and paucity of dsRNA viruses and satellites (S5 Fig) [98,99]. However, T. delbrueckii can support the replication of the putative totivirus TdV-LAbarr and an associated satellite [27], which suggests that this strain has lost its RNAi machinery or the totivirus has the ability to evade or disrupt RNAi similar to other mycoviruses [100–102]. The apparent invasion of these yeasts by viruses could enable the acquisition of killer toxins encoded upon satellite dsRNAs.

To capture killer toxin genes from dsRNAs, the erroneous reverse transcription of mRNAs by endogenous retrotransposons would allow their insertion into a yeast genome by the retrotransposon integrase protein [103]. The ancestral KKT gene in K. africana could have been captured directly from dsRNAs as all extant paralogs encode a 5’UTR “GAAAAA” motif, which is characteristic of satellite dsRNAs. The 5’ UTR is often conserved during mRNA capture by retrotransposons [103]. While the majority of KKT genes are subtelomeric in their location, we have observed that non-subtelomeric KKT genes from N. dairenensis and N. castellii have been uniquely inserted into genomic loci near tRNA genes. The genomic integration of retrotransposon cDNAs is selectively targeted to tRNA genes and many extrachromosomal nucleic acids are identified at loci adjacent to tRNAs [19]. This means the potential mobilization of these genes by retrotransposons and direct insertion by integrase or cellular DNA repair mechanisms [104]

The potential benefits of killer toxin acquisition

Acquisition of foreign nucleotide sequences can be associated with adaptation to a specific environmental niche, including genes associated with nutrient acquisition, virulence, the stress response, and interference competition (e.g. allelopathy) [18,19,84]. Killer toxin-like genes have been identified within the genomes of many different species of fungi and could represent selection for functional killer toxins to improve competitive fitness. Indeed, the production of killer toxins by different species of yeasts has been consistently shown to provide a competitive advantage, particularly in a spatially structured environment at an optimal pH for killer toxin activity [105–110]. However, competition between different killer yeasts likely selects for locally adapted populations that are immune to the predominant killer toxins in a specific environmental niche [6]. The laboratory evolution of killer yeast populations has also shown that killer toxin exposure increases the prevalence of killer toxin resistance [111]. High levels of killer toxin resistance perhaps account for the low prevalence of killer yeasts within natural yeast populations and that killer toxin production might not always be advantageous. However, the rise of killer toxin resistance within a population could perhaps drive the acquisition of new killer toxins or the subfunctionalization of existing killer toxins to maintain a selective advantage. In the context of the multicopy dsRNA satellites and expanded killer toxin gene families in yeasts, it would seem logical to assume that natural variation would provide a pool of genetic diversity on which natural selection could act.

The majority of KKT genes are found to have been acquired within the subtelomeric regions of chromosomes that would facilitate gene expansion due to elevated rates of homologous recombination between telomeric repeat sequences. As has been noted for the subtelomeric MAL gene family, gene duplication enables evolutionary innovation that is also evident in KKT paralogs by possible weak signatures of positive selection [112]. Other genome-encoded fungal killer toxin families have also undergone copy number expansion and are experiencing elevated rates of non-synonymous substitutions [89,113]. Both KP4-like and Zymocin-like killer toxin families appear to have roles during antagonistic interactions with plants and fungi, respectively, which could drive the continued evolution of novel killer toxins. In addition to the expansion of KKT genes in yeasts, there is also gene loss and pseudogenization. KKT gene inactivation is biased towards truncations that leave the α-domain and a small portion of the γ-domain. This same α/γ region of K1 is the minimal sequence required for functional killer toxin immunity that could provide a selective advantage by protecting yeasts from exogenous killer toxins related to KKT [50,51]. KKT yeasts all encode C-terminally truncated KKT genes and are mostly resistant to other KKT killer yeasts. However, despite encoding several truncated KKT genes, K. africana appears to be naturally susceptible to many of the killer toxins produced by other KKT encoding yeasts, including its own killer toxin when ectopically expressed by S. cerevisiae. Subtelomeric Sir-dependent gene silencing could account for this apparent susceptibility under laboratory conditions [114,115]. This would prevent the expression of KKT genes and associated immunity by K. africana, although transcripts from full-length KKT genes were detected in K. africana.

K1L and its homologs represent a unique example of the mobilization and subtelomeric expansion of DNA- and dsRNA-encoded killer toxin genes in different species of yeasts. The diversity of dsRNAs and the known prevalence of killer yeasts suggests that more killer toxins await future discovery and characterization. This will enable a better understanding of their role in fungal biology and provide greater insights into the mechanisms of horizontal gene transfer in eukaryotes.