The Species-specific Acquisition and Diversification of a Novel Family of Killer Toxins in Budding Yeasts of the Saccharomycotina

Killer toxins are extracellular antifungal proteins that are produced by a wide variety of fungi, including Saccharomyces yeasts. Although many Saccharomyces killer toxins have been previously identified, their evolutionary origins remain uncertain given that many of the se genes have been mobilized by double-stranded RNA (dsRNA) viruses. A survey of yeasts from the Saccharomyces genus has identified a novel killer toxin with a unique spectrum of activity produced by Saccharomyces paradoxus. The expression of this novel killer toxin is associated with the presence of a dsRNA totivirus and a satellite dsRNA. Genetic sequencing of the satellite dsRNA confirmed that it encodes a killer toxin with homology to the canonical ionophoric K1 toxin from Saccharomyces cerevisiae and has been named K1-like (K1L). Genomic homologs of K1L were identified in six non-Saccharomyces yeast species of the Saccharomycotina subphylum, predominantly in subtelomeric regions of the yeast genome. The sporadic distribution of these genes supports their acquisition by horizontal gene transfer followed by diversification, with evidence of gene amplification and positive natural selection. When ectopically expressed in S. cerevisiae from cloned cDNAs, both K1L and its homologs can inhibit the growth of competing yeast species, confirming the discovery of a new family of biologically active killer toxins. The phylogenetic relationship between K1L and its homologs suggests gene flow via dsRNAs and DNAs across taxonomic divisions to enable the acquisition of a diverse arsenal of killer toxins for use in niche competition.

toxin from Saccharomyces cerevisiae and has been named K1-like (K1L). Genomic homologs of 23 K1L were identified in six non-Saccharomyces yeast species of the Saccharomycotina 24 subphylum, predominantly in subtelomeric regions of the yeast genome. The sporadic 25 distribution of these genes supports their acquisition by horizontal gene transfer followed by 26 diversification, with evidence of gene amplification and positive natural selection. When 27 ectopically expressed in S. cerevisiae from cloned cDNAs, both K1L and its homologs can 28 inhibit the growth of competing yeast species, confirming the discovery of a new family of 29 biologically active killer toxins. The phylogenetic relationship between K1L and its homologs 30 suggests gene flow via dsRNAs and DNAs across taxonomic divisions to enable the acquisition 31 of a diverse arsenal of killer toxins for use in niche competition. 32 33 Introduction 34 Many different species of fungi have been observed to produce proteinaceous killer toxins that 35 inhibit the growth of competing fungal species [1][2][3][4][5][6][7]. The killer phenotype was reported in the 36 budding yeast Saccharomyces cerevisiae in 1962, when Bevan et al. observed that spent culture 37 medium had antifungal properties [8]. The potential future application of killer toxins as novel 38 fungicides has led to the discovery of many different killer yeasts with varying specificities and 39 toxicities [9,10]. In the Saccharomyces yeasts, including several commonly used laboratory 40 strains, it is estimated that 9-10% are able to produce killer toxins [11,12]. Despite the number of 41 known killer yeasts that have been identified, a complete understanding of the diversity of killer 42 toxins and their evolutionary history is lacking, even within the S. cerevisiae, which has been 43 used as a model organism to study killer toxins for decades. 44 45 In general, killer toxin production by S. cerevisiae is most often enabled by infection with 46 double-stranded RNA (dsRNA) totiviruses of the family Totiviridae [13][14][15]. Totiviruses that 47 infect Saccharomyces yeasts are approximately 4.6 kbp in length and only encode two proteins, 48 Gag and Gag-pol (by a programmed -1 frameshift). These proteins are essential for the assembly 49 of virus particles and the replication of viral RNAs [16,17]. Totiviruses therefore enable killer 50 toxin production by acting as helper viruses for the replication and encapsidation of 'M' satellite 51 dsRNAs, which often encode killer toxins. These satellite dsRNAs are not limited to 52 Saccharomyces yeasts, as they have been identified within other yeasts of the phylum 53 Ascomycota (i.e. Zygosaccharomyces bailii, Torulaspora delbrueckii, and Hanseniaspora 54 uvarum [18][19][20]) and the phylum Basidiomycota (Ustilago maydis [21,22]). In the Ascomycota, 55 the organization of dsRNA satellites is similar, with all sequenced dsRNA satellites encoding a 56 5' terminal sequence motif with the consensus of G(A)4-6, one or more central homopolymeric 57 adenine (poly(A)) tracts, and a 3' UTR containing packaging and replication cis-acting elements 58 [16]. In all known satellite dsRNAs, killer toxin genes are positioned upstream of the central 59 poly ( then raffinose, and finally galactose at 30°C with shaking (250 rpm). The optical density of the 181 final 2 mL culture was normalized to an OD600 of 1.0 and 1 mL was centrifuged at 3,000 ´ g for 182 5 min. The supernatant was removed, and the cell pellet was disrupted by gentle agitation. 2.5 183 µL of the resulting cell slurry was used to inoculate YPD and YPG plates (with 0.003% w/v 184 methylene blue, pH 4.6) seeded with a killer toxin-susceptible yeast strain. Inoculated plates 185 were incubated for 48-72 h at 25°C until killer toxin production was visible (24-72 h). 186 187 Phylogenetic analyses 188 Killer toxin gene sequences were aligned using MUSCLE and manually trimmed to represent the 189 most confident alignment of the a-domain. MEGA (version 7) was used for phylogenetic 190 analysis using neighbor-joining and maximum likelihood methodologies. The optimal model for 191 amino acid substitution was determined as the Whelan and Goldman model with a gamma 192 distribution. 500 bootstrap replicates were used to construct a phylogenetic model with the 193 highest log-likelihood. 194 195 196

198
The

Identification of New Strains of Killer Toxin-Producing Yeasts. A total of 110 strains of 199
Saccharomyces yeasts were obtained from the USDA Agricultural Research Service (ARS) 200 culture collection and screened to identify the production of novel killer toxins. The first screen 201 used eight yeasts from four different species as indicators of toxin production on "killer assay 202 media" (YPD, pH 4.6 with methylene blue) and found that 22% (24 strains) could inhibit the 203 growth of at least one strain of yeast spread as a lawn (File S1). To identify the types of killer 204 toxins based on their unique spectrum of activities, 13 of the killer yeasts were further screened 205 against 45 indicator lawns of yeasts. Four strains of S. cerevisiae that have been previously 206 described in the literature to produce killer toxins of unknown types were also included 207 (NCYC1001, NCYC190, CYC1058, and CYC1113) [50][51][52]. To facilitate the classification of 208 different toxin types, yeasts that produce K1 (BJH001), K28 (MS300c), and K74 (Y8.5) killer 209 toxins, and a non-killer yeast (S. cerevisiae 1116) were included for comparison. The degree of 210 growth inhibition by each killer yeast was scored qualitatively based on the appearance of zones 211 of growth inhibition and methylene blue staining of the surrounding indicator strain on agar 212 plates ( Fig. 1    indicative of totiviruses (~4.6 kbp) and satellite dsRNAs (<2 kbp) ( Fig. 2A; top). The remaining 245 killer yeasts were found to either contain only totiviruses (16%) or no dsRNAs (16%), which 246 suggests that killer toxin production by these strains is genome encoded (Fig. S1). To confirm 247 that the observed satellite dsRNAs encode killer toxin genes, killer yeasts were treated with 248 either cycloheximide, anisomycin, or incubated at elevated temperatures to select for the loss of 249 dsRNAs [62,63]. The majority of killer yeast (86%) lost their killer phenotype after exposure to 250 chemical or thermal insult ( Fig. 2A; bottom). Analysis of the dsRNAs within yeast strains that 251 had lost killer toxin production showed the loss of satellite dsRNAs, but with the maintenance of 252 totivirus dsRNAs ( Fig. 2A;  The dsRNA-encoded killer toxins identified have a pH optimum between 4.5 and 5, with no 259 inhibitory activity at pH 5.5 (Fig. 2B). To confirm that the identified killer toxins are 260 proteinaceous, each was purified by ammonium sulfate precipitation and used to challenge 261 susceptible yeasts. Zones of inhibition were clearly visible on confluent lawns of yeast cells for 262 all of the killer toxins tested (Fig. 2C). The inhibitory activities of these precipitates were heat-263 labile, and the toxicity was lost after incubation at 98°C for 2 minutes. Together, these data 264 suggest that these killer toxins have similar biochemical characteristics to known proteinaceous 265 killer toxins despite their differing inhibitory effects towards yeasts. 266

274
The Discovery of a New Killer Toxin Produced by S. paradoxus. To identify the unknown killer 275 toxins produced by killer yeasts, dsRNAs were purified and subjected to a short-read sequencing 276 pipeline for dsRNAs [23]. BLASTn analysis of de novo assembled contigs revealed that dsRNAs 277 within strains CYC1058 and NCYC1001 encode canonical K2 toxins and NCYC190 a canonical 278 K1 toxin (Fig. S2). The contigs derived from the dsRNAs of Y-63717 assembled into 125 279 different contigs, with six >750 bp in length and a coverage score >1,000 (Fig. 3A). BLASTn 280 analysis of these high-quality contigs identified the dsRNA genome of the totivirus L-A-45 from 281 S. paradoxus N-45 with 100% coverage and 95.5% nucleotide identity [27]. However, the 282 remaining contigs did not match the nucleotide sequence of any known killer toxin in 283 Saccharomyces yeasts. A combination of 5' and 3' RACE, reverse transcriptase PCR, and 284 capillary electrophoresis was used to assemble the complete sequence of the putative dsRNA 285 satellite from Y-63717 ( Fig. 3B and S3). The novel satellite dsRNA is approximately 2371 bp in 286 length with a single open reading frame (ORF) that encodes a protein of 340 amino acids. The 5' 287 ORF is positioned upstream of a central poly(A) tract of ~220 bp ( Fig. 3B and S3). The 5' 288 terminus has a nucleotide sequence of 5'-GAAAAA that is found in many satellites dsRNAs 289 (Fig. S3) and is predicted to fold into a large stem-loop structure (Fig. S4). The length and positioning of the 5' ORF of the satellite dsRNA in Y-63717 strongly suggests 300 that it encodes a killer toxin (Fig. 2). A PSI-BLAST search of the NCBI database with two 301 iterations found that the putative killer toxin has weak homology to the canonical K1 toxin from 302 S. cerevisiae (99% coverage, 21% amino acid identity, e-value 4 x 10 -17 ) (Fig. 4A). Based on this 303 homology, the putative killer toxin was named K1L (K1-like) and the dsRNA satellite was 304 named Saccharomyces paradoxus virus M1-like (SpV-M1L). The organization of the functional 305 domains of K1L appears to be similar to K1 based on the secondary structure, conserved cysteine 306 residues, and the predicted signal peptidase cleavage sites (Fig. 3C) [69]. K1L contains ten 307 cysteine residues, two of which are likely important for interchain disulfide linkage (Cys91) and 308 killer toxin immunity (Cys257) based on their alignment with cysteines from K1 [32]. To 309 confirm that the K1L is an active killer toxin, it was ectopically expressed by the non-killer strain 310 S. paradoxus A12C using a galactose inducible promoter. A well-defined zone of growth 311 inhibition was visible when the strain was grown on galactose-containing media (Fig. 3D). No 312 K1L toxin expression was observed when cells were plated on dextrose-containing growth 313 media. Together, these data confirm the identification of a new dsRNA satellite in 314 Saccharomyces yeasts and a novel killer toxin related to K1. 315   ORFs was between 153-390 amino acids with 11 of the KKT proteins being similar in length to 334 K1L (~340 amino acids) and an amino acid identity between 25-38% (Fig. 4A). In addition, 335 BLASTn was used to identify 14 additional pseudogenes that, in some species, outnumber intact 336 KKT genes ( Fig. 4B and Table S1). All KKT genes and related pseudogenes are found in multiple 337 copies that vary in frequency between different yeast species and are mostly located within 338 subtelomeric regions (within ~20 kb of the assembled chromosome ends) (Fig. 4B). Of the 38 339 KKT genes and pseudogenes found within six different species, only two are positioned away 340 from the subtelomeric regions in the yeasts N. dairenensis and N. castellii (Fig. 4B). Analysis of 341 the chromosomal position of these two genes revealed that their insertions are unique to each 342 species, are absent from other related species at the syntenic chromosomal location, and are 343 inserted close to tRNAs (Fig. S5). It was noted that six of the KKT genes and pseudogenes from 344 K. africana contained a characteristic "GAAAAA" sequence motif close to the start codon of 345 each ORF (Fig. S5).

358
To ascertain the evolutionary history and relatedness of K1L and its homologs, a multiple amino 359 acid sequence alignment was constructed. The most confident alignment was achieved between 360 the putative a-domain of each protein and included eight truncated proteins with premature stop 361 codons. Phylogenetic analysis was performed using maximum likelihood (Fig. 4C) and 362 neighbor-joining (Fig. S6)  toxins that caused growth inhibition of at least one other yeast ( Fig. 5A and B). Each of these 386 killer yeasts was also immune to its own killer toxin, but susceptible to those produced by other 387 KKT-encoding yeasts. The production of killer toxins by these species is consistent with the 388 previously reported killer activity of P. membranifaciens, T. delbrueckii, and T. phaffii 389 [19,73,74]. There was no evidence of satellite dsRNAs in any of the KKT-encoding yeasts, 390 except for the detection of an unknown high molecular weight dsRNA within P. 391 membranifaciens NCYC333 (Fig. S8). The differences in killer toxin production by different 392 strains of P. membranifaciens indicated that there could be strain-specific differences in KKT  393 genes. The published genome sequence of P. membranifaciens Y-2026 revealed a large central 394 deletion in the g-domain of its KKT gene (Fig. S9). Sanger sequencing of the same KKT gene 395 from P. membranifaciens Y-2026 acquired directly from the NRRL culture collection failed to 396 identify the same deletion, instead there was an indel within the g-domain that caused the 397 truncation of the killer toxin gene (Fig. S9). Sequencing of the K1L gene from P. 398 membranifaciens NCYC333 confirmed a full-length K1L gene that correlated with robust killer 399 toxin production by the strain. However, Y-2026 was still able to produce killer toxins, 400 suggesting the production of other antifungal molecules by P. membranifaciens. 401 402 Although P. membranifaciens, N. dairenensis, T. delbrueckii, T. phaffii, and K. africana are 403 killer yeasts (Fig. 5A and B), it was unclear whether KKT genes were directly responsible for the 404 observed production of killer toxins. Indeed, T. phaffii has been reported to express an antifungal 405 glucanase and K2 killer toxin-related genes have been found in the genome of K. africana 406 [68,73]. To demonstrate that KKT genes are active killer toxins with antifungal activities, 10 full-407 length KKT genes were cloned into galactose inducible expression vectors (labelled in Fig. 4C). 408 Active killer toxin production from a non-killer strain of S. cerevisiae transformed with KKT 409 genes was assayed against 13 lawns of yeasts using galactose containing agar plates. The 410 majority of KKT genes did not cause any noticeable growth inhibition or methylene blue staining 411 of competing yeasts (Fig. S10). However, killer toxins from P. membranifaciens NCYC333, K. 412 africana, and N. dairenensis were able to create visible zones of growth inhibition (Fig. 5C). No 413 growth inhibition was observed when the genes were not expressed by plating cells on dextrose 414 (Fig. S10), or when S. cerevisiae was transformed with an empty vector control (Fig. 5C). 415 Altogether, these data show that KKT genes encode active killer toxins and confirm the 416 discovery of a new family of genome and dsRNA-encoded antifungal proteins in the 417 Saccharomycotina. 418 419 420 421  The most significant finding of this study is the discovery of a novel satellite dsRNA that 432 encodes a killer toxin related to K1 and a larger family of DNA-encoded homologs in yeasts. 433 The relatedness of killer toxins encoded on dsRNAs and DNAs suggests that the origins of K1L 434 are outside of the Saccharomyces genus, with killer toxin gene mobilization and interspecific 435 transfer by dsRNAs. Many of these killer toxins have been shown to be biologically active and 436 are diverse in their amino acid sequences with evidence of their rapid evolution by gene 437 duplication and elevated rates of non-synonymous mutations. This demonstrates the likely 438 benefits of killer toxin acquisition and the ongoing mobilization of these genes between 439 divergent species of yeasts. The more specific implications of our findings are discussed below. 440 441

Horizontal Acquisition and Copy Number Expansion of Killer Toxins in Fungi. KKT genes 442
have most likely been acquired by horizontal gene transfer because of their sporadic distribution 443 and lack of common ancestry in closely related yeast species. Moreover, the lack of relatedness 444 of KKT genes in these species suggests that they have independent origins.