ABSTRACT
The killer phenomenon in yeast (Saccharomyces cerevisiae) not only provides the opportunity to study host-virus interactions in a eukaryotic model but also represents a powerful tool to analyze potential coadaptional events and the role of killer yeast in biological diversity. Although undoubtedly having a crucial impact on the abundance and expression of the killer phenotype in killer-yeast harboring communities, the influence of a particular toxin on its producing host cell has not been addressed sufficiently. In this study, we describe a model system of two K1 killer yeast strains with distinct phenotypical differences pointing to substantial selection pressure in response to the toxin secretion level. Transcriptome and lipidome analyses revealed specific and intrinsic host cell adaptions dependent on the amount of K1 toxin produced. High basal expression of genes coding for osmoprotectants and stress-responsive proteins in a killer yeast strain secreting larger amounts of active K1 toxin implies a generally increased stress tolerance. Moreover, the data suggest that immunity of the host cell against its own toxin is essential for the balanced virus-host interplay providing valuable hints to elucidate the molecular mechanisms underlying K1 immunity and implicating an evolutionarily conserved role for toxin immunity in natural yeast populations.
IMPORTANCE The killer phenotype in Saccharomyces cerevisiae relies on the cytoplasmic persistence of two RNA viruses. In contrast to bacterial toxin producers, killer yeasts necessitate a specific immunity mechanism against their own toxin because they bear the same receptor populations as sensitive cells. Although the killer phenomenon is highly abundant and has a crucial impact on the structure of yeast communities, the influence of a particular toxin on its host cell has been barely addressed. In our study, we used two derivatives secreting different amount of the killer toxin K1 to analyze potential coadaptional events in this particular host/virus system. Our data underline the dependency of the host cell’s ability to cope with extracellular toxin molecules and intracellular K1 molecules provided by the virus. Therefore, this research significantly advances the current understanding of the evolutionarily conserved role of this molecular machinery as an intrinsic selection pressure in yeast populations.
INTRODUCTION
Viruses represent the most ubiquitous entity in the animal kingdom and hold unique evolutionary advantages based on their extremely short generation time. In contrast to lytic viruses, mycoviruses form a stable endosymbiotic relationship with the respective yeast since their lack of an extracellular phase under physiological conditions is a critical feature coupling their reproductive success even closer to the fungal host cell (1). Although most mycoviral infections are asymptomatic or even slightly beneficial concerning host cell fitness, one prominent phenotype, the so-called killer phenomenon, established by two separately encapsulated double-stranded RNA (dsRNA) viruses of the Totiviridae, called LA and M, has been characterized (2, 3). Despite the additional metabolic disadvantages involving killer toxin production, the secretion of toxic compounds sharply increases the competitiveness of killer yeast toward nonkiller strains regarding the availability of natural resources in the environment (4, 5). The genome of LA (4.6 kb) codes for the viral capsid protein (Gag), and an RNA-dependent RNA polymerase (Pol) expressed as a Gag-Pol fusion protein (6, 7). The M virus, in contrast, is a nonautonomous satellite virus exploiting the gene products of LA for replication and maintenance in the host cell cytoplasm. To date, four different killer toxins have been identified in S. cerevisiae (K1, K2, K28, and Klus), which are encoded by the corresponding M dsRNA genome (8–10).
Initially expressed as a precursor protein (preprotoxin [pptox]), each killer toxin traverses the host cell’s secretory pathway and is eventually secreted. The mature K1 toxin resembles a classical A/B toxin composed of one toxic α subunit and one β subunit, the latter being responsible for cell surface binding (11). Although the exact molecular mechanism of toxin action has not been fully elucidated, it is well known that K1 kills sensitive cells in a two-stage receptor-mediated process, initiated by binding to the β-1,6-glucan fraction of the yeast cell wall (12, 13). Subsequently, the toxin is transferred to the plasma membrane, where it interacts with its secondary receptor Kre1p, exerting its lethal effect by forming cation-specific ion channels and thereby disrupting plasma membrane integrity. This ionophoric action of K1 leads to an uncontrolled influx of protons accompanied by a potentially compensating efflux of potassium ions. Eventually, the proton transmembrane gradient collapses, and the resulting energetical and electrochemical drainage cumulates in the death of sensitive yeast cells (14, 15). The current model of K1 action considers both, the possibility of direct insertion of the toxin into plasma membrane structures, as well as interaction with as-yet-unknown primary effectors (11, 15). In a recent study, we were able to characterize the transcriptome kinetics of a sensitive strain in response to K1, revealing insights into both the K1 lethal effects and the possible defense mechanisms of the target cell (16).
In addition to metabolic costs of maintaining the mycoviral genomes and expressing the killer toxin, and in clear contrast to bacterial toxin producers, killer yeasts possess the same receptor population as sensitive cells and are therefore in need of a unique immunity mechanism protecting them from the effect of their own toxin. However, despite decades of research, the exact molecular mechanism of K1 immunity has not been elucidated at a molecular level. Analysis of both the immunity mechanism and the molecular background of K1 toxicity could help us to understand the lethal effects of ionophoric toxins in general and yield essential insights into eukaryotic cell biology and, likewise, adaption processes in yeast communities. In this study, we utilized two K1 killer strains with different toxin secretion levels and sensitivities to externally applied K1 toxin to evaluate the potential cellular adaptations on lipidome and transcriptome levels caused by intrinsic K1-induced selection pressure. Our results provide an overview of K1-induced changes in gene expression after short and prolonged incubation with the killer toxin, as well as differences in basal transcriptome adaptions in both strains, providing insights into an evolutionarily conserved adaption mechanism peculiar to killer yeast.
RESULTS
Characterization of killer strains KIM01 and KIM01s.The killer strain KIM01 was originally constructed by transfecting the sensitive strain GG100-14D with virus particles derived from a standard K1 killer strain (17). The presence of viral dsRNA species in KIM01 and its derivative KIM01s, as well as in the superkiller strain T158c and the wild-type strain BY4742, was analyzed via gel electrophoresis (Fig. 1a). In each strain, LA-dsRNA with a molecular weight of ∼4.6 kb could be detected, whereas the viral M1 genome (∼1.4 kb) coding for the killer toxin precursor was present in each strain except for the nonkiller BY4742. Regarding biological activity, toxin secreted by KIM01s showed very little to no toxicity against intact BY4742 cells (Fig. 1b, upper panel), whereas a small but distinctive zone of growth inhibition could be observed when applied to sensitive spheroplasts generated by enzymatic removal of the cell wall (Fig. 1b, lower panel). Subsequently, secretion of the mature toxin heterodimer was verified via Western analysis of the precipitated supernatant yielding distinctive bands at 18 kDa for all killer strains corresponding to the molecular weight of the mature K1 toxin (Fig. 1c). Quantitative analysis of the corresponding bands showed an circa 40% lower secretion rate for KIM01s cells compared to KIM01.
Characterization of the K1 killer phenomenon. The K1 killer characteristics of S. cerevisiae strains KIM01 and KIM01s were determined at the dsRNA and protein level; the K1 superkiller strain T158c and the sensitive strain BY4742 were used as controls. (a) Visualization of extracted killer viruses. Double-stranded RNA was isolated using TRI Reagent, and separated via gel electrophoresis (1% agarose). LA viruses (upper band, 4.6 kb) were found in all analyzed strains, whereas the smaller M1 virus (1.4 kb, lower band) could only be observed in the strains T158c, KIM01, and KIM01s. DNA-specific molecular weight markers were used to provide approximately sizes of dsRNA genomes. (b) Biological activity of K1 toxin. The toxic activity of secreted K1 against BY4742 was analyzed via MBA assay (pH 4.7). A total of 106 intact cells (upper panel) or spheroplasted cells (lower panel) of the sensitive strain BY4742 were embedded into the agar, respectively. Next, 10 μl of overnight grown cultures (JG medium, 30°C) of the yeast strains T158c, KIM01, KIM01s, and BY4742 were spotted onto the plates, and potential halo zones representing killing activity were documented after incubation for 3 days at 20°C. (c) Analysis of killer toxin secretion. Toxin secretion was quantitatively determined by Western analysis (nonreducing conditions) using a polyclonal antibody detecting the K1 heterodimer (1:1,000). A culture volume at the indicated OD600 was centrifuged, and proteins of cell-free supernatant of indicated strains were precipitated using trichloroacetic acid. Toxin-specific bands could be observed for KIM01, KIM01s, and T158c at approximately 18 kDa (marked by black arrow). For improved visualization, the representative blot images were cropped as indicated by a fine dotted line, and the background brightness was adjusted (unmodified blots are presented in Fig. S3 in the supplemental material).
In addition to the lethal effect of secreted toxin, the establishment of functional immunity of the K1 killer strains was analyzed using a methylene blue agar (MBA) diffusion assay by embedding the respective killer yeast into the agar and applying concentrated K1 toxin. Interestingly, addition of extracellular toxin induced a loss of immunity in KIM01s cells, resulting in the development of killing zones which were approximately 25% smaller compared to sensitive BY4742 cells. In sharp contrast, killer strains T158c and KIM01 were completely immune (Fig. 2a). Most interestingly, functional immunity could be restored via plasmid-driven expression of the wild-type K1 precursor (K1-pptox), pointing to a compensatory effect restoring the intrinsic immunity of KIM01s, concurrently excluding potential general defects in precursor processing and maturation (Fig. 2b). Although the toxic effect of K1 can be attributed to the formation of cation-specific pores in the plasma membrane of a sensitive target cell, it is currently unknown whether this ionophoric effect also affects natural killer yeast or whether these are capable of preventing pore formation (18). Therefore, the occurrence of pores in the K1 killer strains T158c and KIM01, as well as KIM01s, was quantitatively measured via propidium iodide (PI) staining and subsequent fluorescence-activated cell sorting (FACS) analysis. Cells were cultivated in JG medium under toxin stabilizing conditions (20°C; pH 4.7), supplemented with the membrane-impermeable DNA stain, and the percentage of PI-positive cells was determined. Interestingly, 5% (KIM01s), 13% (KIM01), and 8% (T158c) were identified as PI positive when incubated at the optimal pH for K1 toxicity (Fig. 2c).
Mediation of functional immunity against externally applied K1 toxin. (a) Proper formation of functional immunity was tested via MBA (pH 4.7) by embedding 106 cells of the killer strains T158c (shaded blue), KIM01 (blue), and KIM01s (black), as well as the sensitive strain BY4742 (shaded black), into the agar and applying 1,000 arbitrary units of K1 toxin concentrate into precut holes. The diameters of killing zones were measured after incubation of the plates for 3 days at 20°C and normalized to BY4742 (100%, n = 3). (b) The expression of wild-type K1 pptox restores immunity in KIM01s. A total of 106 cells of the S. cerevisiae strain KIM01s expressing the wild-type K1 precursor (pYES2.1 K1 pptox) were embedded into MBA agar (Ura d/o, galactose [pH 4.7]), and 10-μl portions of an overnight culture of T158c were spotted onto the plates. The potential formation of halo zones was documented after incubation for 3 days at 20°C. Results from one representative experiment are shown (n = 6). (c) Analysis of pore formation in K1 killer strains. Pore formation in the depicted killer strains T158c (shaded blue), KIM01 (blue), and KIM01s (black), as well as the nonkiller strain BY4742 (shaded black), was determined via propidium iodide (PI) staining and subsequent FACS analysis. The applied gate for PI-positive cells was defined by the unstained and heat-treated cells for each strain; samples were measured in triplicates with 100,000 gated events each (n = 3).
Lipidome analysis of KIM01 and KIM01s membranes.Furthermore, a comparative lipidome analysis of the KIM01 and KIM01s was performed in order to identify possible adaptions in membrane composition concerning the different K1 secretion levels. Although the samples of KIM01 and KIM01s cluster differently in the principal component analysis, no significant alterations in the distribution of lipid classes could be determined in terms of the total lipid amount (see Fig. S1a and b in the supplemental material). However, considering the individual lipid species, distinctive variabilities could be observed. In general, membranes of KIM01s displayed a significantly larger amount of ergosterol esters, whereas most of the phospholipid species (phosphatidylcholine [PC], phosphatidylethanolamine [PE], phosphatidylserine [PS], and phosphatidylinositol [PIN]), as well as the triacylglycerol (TAG) species TAG 42:2 and TAG 40:1, were more abundant in membranes of the killer KIM01; only TAG 52:2 was present at higher levels in KIM01s (Fig. 3). These data point to an adaption of the strains to their natural amount of secreted K1 toxin at membrane level.
Comparative lipidomics of the S. cerevisiae killer strains KIM01 and KIM01s. The mean values (± the standard deviations [mol%]) for each triplicate and lipid species differing with high significance between KIM01 (blue) and KIM01s (black) are illustrated (**, P < 0.01; ***, P < 0.001). (a) EE, ergosterol esters; TAG, triacylglycerol. (b) PIN, phosphatidylinositol. (c) PC, phosphatidylcholine; PE, phosphatidylethanolamine; PS, phosphatidylserine; C:D ratio, C atom/double-bond ratio.
Time-dependent transcriptome changes after K1 toxin application.Potential differences in gene regulation of KIM01 and KIM01s induced by additional K1 toxin were analyzed via Illumina sequencing 5 min (t5) and 45 min (t45) after killer toxin application. In a first approach, the transcriptome adaptations were compared to the respective control samples taken before toxin addition. In case of KIM01, 30 and 55 differentially expressed genes (DEGs) were significantly altered 5 and 45 min after toxin addition, respectively; more DEGs were detected in KIM01s cells with 77 (t5 ) and 99 open reading frames (ORFs) (t45). Interestingly, many genes were similarly regulated in both strains, leading to intersections of 18 and 4 DEGs (t5), as well as 26 and 10 DEGs after 45 min (Fig. 4, complete list of DEGs in Table S1). Only the ORF STE3 encoding the receptor for the “a-factor” pheromone was uniquely upregulated in the killer strain KIM01 after 5 min of incubation with K1 toxin. In contrast, after 45 min, two of three genes coding for subunits of the mitochondrial glycine decarboxylase complex were altered in addition to one retrotransposable element (YPL060C-A). Uniquely upregulated DEGs in the killer strain KIM01s mostly comprise genes of cellular carbohydrate and trehalose metabolism (e.g., TPS1, TPS2, TSL1, NTH1, and HXK1). Among others, the stress-related genes CTT1, XBP1, HSP104, and TMC1, as well as the transcription factors TEC1 and CIN5, were significantly upregulated. After 45 min, the upregulation of genes involved in the IMP biosynthetic process, as well as genes associated with carbohydrate metabolism, which were already upregulated before, was sustained. Moreover, ORFs encoding genes essential for cellular processes involved in cell fusion (STE3, FUS2, FUS1, and FIG1) and ion homeostasis (PGM2, FET3, HMX1, FTR1, GLK1, ENB1, and SIT1) were positively regulated. DEGs upregulated early after K1 application for both strains could mainly be classified as genes for glucose transmembrane transport (HXT3, HXT4, HXT7, MTH1, and HXT2) and IMP metabolism (ADE2, ADE1, ADE17, ADE13, MTD1, and SHM2); additionally, transcription factors MSN4 and HAP4 were upregulated in both yeast strains. These positively regulated ORFs were also consistently upregulated in both killer strains after prolonged killer toxin incubation.
Venn diagrams of the time-dependent transcriptome alterations after K1 toxin application. Significantly altered DEGs (compared respective untreated control samples) for each yeast strain and time point (t5 and t45) were clustered into upregulated (green) and downregulated (red) ORFs (Padj. < 0.05). The depicted intersecting region of the Venn diagrams illustrates identical DEGs altered in both strains for the particular time point.
Compared to upregulated genes, fewer ORFs showed a significant negative regulation. Regarding early adaptations, seven DEGs were uniquely downregulated for the killer strain KIM01, including ALD6, GPP1, RPL26, and STL1, whereas only the gene coding for a protein of the large ribosomal subunit was affected in KIM01s. In addition, the genes CIT1, GPD1, HXT1, and SPO24 were negatively regulated in both yeast strains. After 45 min of incubation with K1 toxin, 15 ORFs were significantly downregulated in KIM01 that can mostly be summarized in the GO term “small molecule biosynthetic processes” (ARG1, ENO2, HOR2, PDC1, IMD3, RIB4, ADH1, and LYS20). For the K1-sensitive strain KIM01s, 19 DEGs were found to be negatively altered in their expression, including HXT3, CPA2, IMD4, and HIS5. After 45 min, GPD1 and HXT1 could still be found downregulated like GPP1, IMD2, and LYS1 (Fig. 4).
Pairwise comparison of transcriptome data.In addition to the time-dependent analysis, transcriptome sequencing data were used to compare the analyzed states of the cells in a pairwise comparison (see the complete list of DEGs in Table S2 in the supplemental material). Interestingly, even without killer toxin application, both strains remarkably differed in their gene expression, as illustrated in the corresponding MA plot (Fig. S2a). In total, 153 genes (excluding dubious ORFs) were expressed at higher levels in the strain KIM01, whereas 134 DEGs were significantly less expressed on a basal level (Fig. S2b). A gene ontology (GO) enrichment for “biological process” was calculated for both strains, and subsequent clustering and visualization of the data sets were carried out using REVIGO. In case of upregulated DEGs (Fig. 5, left panel), the GO term “IMP metabolism” showed the most enrichment, including the DEGs ADE4, ADE2, ADE1, ADE17, ADE5,7, ADE13, ADE6, and ADE12. Furthermore, many transposable elements summarized in “DNA integration” and “transposition” were found to be significantly upregulated in KIM01. Genes involved in amino acid metabolism (e.g., ARO10, GCV1, ARG1, CHA1, and SER1) and cell adhesion processes (SAG1, FLO9, AGA1, and FIG2) showed greater expression levels in the killer strain. Interestingly, genes of the trehalose metabolism like TPS1, TSL1, and UGP1 (“carbohydrate biosynthesis”) showed also a higher basal expression level. In less-expressed DEGs of the strain KIM01 (Fig. 5, right panel), 62 ORFs could be summarized in the GO term “cellular biosynthetic process,” including LYS4, URA5, CYT1, ERG28, ERG2, ERG11, and IMD4. In addition, genes encoding both ribosomal subunits were found to be basally less expressed in KIM01 (“rRNA processing”).
Basal differences in the transcriptome of KIM01 and KIM01s independent of K1 toxin application. Visualization of upregulated (left panel) and downregulated (right panel) GO terms in KIM01 in comparison to KIM01s. DEGs (Padj. < 0.05) were clustered according to their respective log2-fold change in upregulated (left) or downregulated (right) genes. GO enrichment was performed with GOrilla for each data set, and subsequent visualization of the significantly enriched GO terms (FDR-corrected P value of <0.01) was performed using REVIGO. x axis, log10 P value; y axis, semantic space Y (implicating similarity between GO terms). The colors and sizes of the spheres correlate with the log10 P value (blue = highest significance).
Transcriptome adaptations in both strains after 5 and 45 min of incubation with K1 toxin were also compared. Similar to the control samples without toxin, significant differences in gene expression levels could be detected for each incubation time illustrated in the corresponding MA plot (Fig. S2a). In detail, 53 (t5) and 57 (t45) ORFs were significantly upregulated in the killer strain KIM01, whereas 65 and 41 DEGs were distinctively less expressed after 5 and 45 min of K1 intoxication, respectively. For each cluster (up- and downregulated DEGs), a substantial intersection of genes deregulated at both time points could be observed comprising of 33 and 23 ORFs, respectively (Fig. 6a). In order to get a general overview of cellular activities altered by K1 toxin, data sets of significantly up- and downregulated DEGs were subjected to a GO term enrichment, and significantly enriched terms were illustrated using the calculated fold enrichment (FE). In comparison to the sensitive strain KIM01s, processes involving IMP and one-carbon metabolism were highly upregulated in KIM01 at both time points with fold enrichments ranging from 53.4 to 68.5 (“IMP MP”) and from 25.7 to 30.9 (“one-carbon MP”), respectively. GO terms referring to DNA- or RNA-dependent processes such as “DNA integration” (18.9 and 20.4 FE), “RNA-dependent DNA biosynthetic process” (13.2 and 14.3 FE) or “transposition” (10.5 and 12.4 FE) were significantly enriched for both time points. Furthermore, the biological processes “DNA duplex unwinding,” “neurotransmitter CP,” “regulation of neurotransmitter levels,” and “pigment metabolic process” were solely upregulated after 45 min of incubation with the killer toxin with respective fold enrichments of 10.8, 41.7, 23.4, and 14.8. The GO term “serine family amino acid catabolic process” was also found to be highly enriched in upregulated DEGs (61.8). In the case of significantly downregulated ORFs, the GO term analysis revealed several considerably enriched terms including “DNA integration” (19.0 and 20.7 FE), “RNA-dependent DNA biosynthetic process” (13.3 and 14.5 FE), “transposition” (10.4 and 13.2 FE), and “transposition RNA-dependent” (11.0 and 14.0 FE) for both analyzed time points (Fig. 6b and c). Moreover, terms referring to biological processes involving genes of the carbohydrate metabolism were downregulated after a short incubation with the killer toxin with fold enrichments of 21.1 (“carbohydrate transmembrane transport”), 26.7 (“glucose transmembrane transport”), and 12.0 (“glucose metabolic process”), respectively. In addition, the GO terms “acetate metabolic process” and “antibiotic biosynthetic process” were significantly enriched in downregulated DEGs after 5 min of K1 toxin application (for a complete list of the GO terms, see Table S3 in the supplemental material).
Difference between KIM01 and KIM01s in transcriptome changes after K1 toxin application. (a) Pairwise comparison of significantly deregulated ORFs in KIM01 and KIM01s after K1 application. Transcriptome data sets of both yeast strains were compared for each time point (t5 and t45), and significantly altered DEGs (Padj. < 0.05) were further clustered into upregulated (green) and downregulated (red) ORFs. The depicted intersection of the Venn diagrams illustrates identical DEGs deregulated at both analyzed time points for the respective gene cluster. GO term enrichment of upregulated (b) and downregulated (c) DEGs in KIM01 cells associated with biological processes. DEGs were classified as up- and downregulated genes according to the corresponding log2-fold change (Padj. < 0.05). GO term enrichment was calculated via the GOrilla software. For simplification reasons, only GO terms with an enrichment of >10 are depicted (for a complete list, see Table S3 in the supplemental material). AA, amino acid; BP, biological process; CP, catabolic process; MP, metabolic process.
DISCUSSION
Chronic infection of yeast cells with viruses is a common characteristic of various strains and is often accompanied by a symptomless phenotype (reviewed in reference 5). Their ubiquity and abundance in multiple fungal taxonomic groups underline their importance in shaping yeast communities and their essential role in biological diversity. One unique form of such a mutualistic endosymbiosis is the establishment of the killer phenomenon in yeasts based on the cytoplasmic persistence of two distinct mycoviruses, which enables the killer strain to kill sensitive cells. We characterized a model system for analyzing toxin-specific effects on the host cell itself by comparing the K1 killer strain KIM01 and its derivative KIM01s at various levels; the latter strain exhibits a distinct and aphenotypic sensitivity when additional K1 toxin is applied.
Secretion of biologically active K1 could be verified for both strains, whereas a distinctively reduced toxin concentration was determined in the supernatant of KIM01s compared to KIM01. Considering the logarithmic correlation between the amount of secreted toxin and the generation of killing zones that is limited by the diffusion properties of the toxin into the agar, this decrease in secretion levels has an observable impact on the toxicity. Accordingly, a pronounced toxic effect of K1 toxin secreted by KIM01s could only be observed against spheroplasted cells of the sensitive strain. In addition to the reduced secretion rate, disturbed binding properties of K1 toxin secreted by KIM01s cannot entirely be excluded and need further validation. Most interestingly, KIM01s cells showed a distinct sensitivity against additionally applied K1, although producing the killer toxin themselves and without being a self-killer. This loss of immunity, generally granted by the precursor toxin within the secretory pathway, implicates a fine-tuned system of mechanisms and molecules, which are essential for the general mediation of immunity. Taken together with the lower K1 secretion, KIM01s can protect itself against the amount of its own secreted toxin; however, as soon as additional toxin molecules are present, intrinsic mechanisms involved in the mediation of functional immunity are not able to compensate for the emerging damage (Fig. 7). This hypothesis is further strengthened by the reestablishment of full immunity by the intracellular expression of wild-type K1 toxin.
Adaptations of a K1 killer yeast in response to different toxin secretion levels. After secretion, the mature K1 heterodimer exerts its lethal effect on a sensitive target cell via a two-staged receptor-mediated process involving β-1,6 glucans in the cell wall, as well as its secondary receptor Kre1p. This toxic effect could be observed for KIM01 killing intact cells and spheroplasted cells, whereas a significant killing zone using KIM01s could only be evoked against sensitive cells lacking the cell wall fraction. This effect corresponds to the distinct lower secretion levels detected by Western blotting (toxicity, left). Although both strains possess the necessary M1 viruses and are immune to their own secreted toxin (black), KIM01s showed a substantial loss of immunity when additional K1 toxin (gray) was extracellularly applied. Basal transcriptome analysis showed higher expression of various stress-related genes, as well as osmoprotectants, in KIM01 compared to KIM01s. These intrinsic cellular adaptions to a high toxin secretion are sufficient to provide KIM01 with the ability to cope with additionally added toxin. In contrast, based on the low secretion of toxin, the intrinsic mechanisms of KIM01s are incapable of providing immunity against increased external K1 levels culminating in the death of the killer cells (immunity, right).
In addition to the exact molecular mechanism underlying this self-defense mechanism, it is also not known whether the pore formation induced by rebinding toxin molecules is entirely repressed in K1 killer yeast. In line with this, PI staining was performed to observe potential toxin-induced disruption of the plasma membrane in killer strains, and PI-positive cells for all analyzed killer yeast were observed. The reduced pore formation in strain KIM01s can probably be explained by the strongly decreased secretion of the toxin, whereas the molecular machinery conferring immunity in the superkiller T158c is likely to be optimally adjusted to reinternalized toxin molecules. This leaky immunity correlates with findings displaying the same phenotype when K1 toxin is expressed via a centromeric plasmid (19). However, we previously were able to demonstrate that the formation of toxin-induced pores takes place considerably before a significant decrease in cell viability of sensitive cells can be observed (16). Based on this observation, a corresponding analysis correlating the percentage of PI-positive and living cells using the killer strains could help to determine whether K1 toxin is able to merely insert into the plasma membrane of its producing host cell or whether the killer yeast is effectively killed during the process.
Due to the ionophoric effect of the killer toxin and the observed incomplete immunity of the various killer yeast, we raised the question whether significant differences in the membrane composition, especially the plasma membrane, of KIM01 and KIM01s cells could be detected. Although the lipid arrangement of yeast cells is rather simple compared to higher eukaryotes, it is highly flexible and can be actively adapted to changes in environmental conditions (20). The most abundant membrane lipid species in yeast are ergosterol esters (EE), phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PIN), phosphatidylserine (PS), and diacylglycerol (DAG), whereas triacylglycerols (TAGs) are mostly used as storage lipids. The general composition of the fungal plasma membrane tends to be more negatively charged due to differences in the ratio of anionic and neutral lipids, especially larger amounts of PIN and PA. These electrostatic properties make the fungal plasma membrane a prominent target for cationic antimycotic peptides and ionophoric killer toxins such as Zygocin (21). Significant differences in the membrane composition could be detected when comparing KIM01 and KIM01s, indicating lipidome adaptations to the particular secreted killer toxin species. Interestingly, membranes of KIM01s appeared to possess a significantly larger amount of ergosterol than KIM01 cells, which correlates with transcriptome data indicating a basally lower expression level of genes essential for the ergosterol biosynthetic pathway (ERG2, ERG11, and ERG28) in cells of the killer strain KIM01. In a genome-wide screening analyzing K1 toxin sensitivity, different deletion mutants defective in sterol synthesis showed a significantly reduced toxin sensitivity, also indicating the importance of ergosterol in K1 toxicity (22). This sterol is an essential component of lipid rafts and drastically contributes to the permeability and fluidity of the yeast plasma membrane. In addition to modulating the physicochemical properties of the membrane per se, the activity of membrane-associated proteins is indirectly affected by changes in the lipid raft abundance and composition (23, 24). In addition, a reduction in the ergosterol content has been associated with a severely decreased frequency of endocytotic events and could also affect the general function of the secretory pathway, leading to partial toxin resistance, as shown for several mutants with defects in protein secretion and trafficking (22, 25). Taken together, these findings suggest the smaller amount of ergosterol in KIM01 may hinder the insertion of the toxic α subunit into the lipid bilayer, potentially due to alterations of the cellular membrane potential. The reduction of the sterol content could also impair the general toxin binding to membrane proteins such as Kre1p or other potential primary effectors, disturbing or reducing the endocytosis of K1 toxin-effector complexes.
Furthermore, we were interested in comparing the transcriptome adaptations of both strains after the addition of external toxin, as well as the basal differences in expression levels. In contrast to a previously conducted study analyzing transcriptome adaptations of a sensitive strain, only a few genes were found to be significantly deregulated in both K1 killer strains (16). Although the expression of some genes was altered in both strains, the K1-sensitive strain KIM01s showed a generally stronger response when incubated with additional K1. Interestingly, the majority of DEGs in both strains showed the same pattern as the sensitive strain BY4742, including the upregulation of stress-responsive genes and genes encoding carbohydrate metabolism proteins. The moderate response of both strains in comparison to BY4742 may reflect the constant exposure to K1 toxin. Nevertheless, the application of higher toxin concentrations led to some transcriptome adaptions potentially coping with higher intracellular stress. Noteworthy also is the upregulation of the transcription factors CIN5, HAP4, MSN4, and TEC1; this was also observed in the transcriptome of BY4742 after K1 exposure, pointing to a common stress response. Early transcriptome adaptations were mostly similar to BY4742, and only prolonged incubation with K1 killer toxin provoked the regulation of other genes, especially in the K1-sensitive strain KIM01s.
In contrast to the time-dependent analysis, where DEGs were determined with respect to the untreated control, in a pairwise comparison, gene expression in KIM01 cells was compared to the individual expression data of KIM01s cells. We found significant differences in both strain derivatives regarding the basal transcriptome without interference of additional K1 toxin. Intriguingly, the K1-resistant strain KIM01 showed significantly higher expression levels of several genes essential for energy metabolism, as well as cell adhesion, IMP metabolism, and DNA integration. Taken together with the basally less-expressed genes, the transcriptome of KIM01 appears to be preadapted to secretion and reinternalization of moderate to high K1 toxin levels. In addition, we compared the transcriptome adaptations of both strains after toxin application. Our data support the positive transcriptional regulation of many genes related to purine de novo biosynthesis and tetrahydrofolate (THF) metabolism in KIM01, which already showed a higher basal transcription. As with BY4742, this could indicate a shift of the purine biosynthesis pathway to the generation of ATP since genes such as SER1, SHM2, GCV1, GCV2, and GCV3 have a crucial role in THF metabolism by generating and recycling THF derivatives (26). We already demonstrated that sensitive cells are able to initially restrain the ionophoric action of K1 toxin to a certain degree and maintain relatively stable intracellular ATP levels despite the leakage of this metabolite into the extracellular space. ATP is not only the most essential energy source of the cell, but it can also act in a hydrotropic fashion, maintaining protein solubility independent of other chaperones (27). Moreover, higher ATP levels would also support potential cellular repair mechanisms against the pores formed by K1. In addition, genes essential for the trehalose and glycogen synthesis were more severely regulated in KIM01 (e.g., TSL1, TPS1, TPS2, and GSY1). Trehalose and glycogen are osmolytes that are part of a critical cellular defense mechanism against different stress situations, helping in the detoxification and stabilization of cellular proteins (28). Trehalose not only serves as a reserve energy supply but is also directly, acting as an osmoprotectant, is involved in ROS detoxification and probably the reduction of lipid peroxidation (29–32). Previous studies additionally linked trehalose to the stabilization of membranes under stress conditions (33).
The killer phenomenon in yeast serves not only as a eukaryotic model for host-virus interaction but has great potential to investigate coevolutionary and coadaptational events. Based on the data generated in this study, we propose a model of K1 immunity based on titratable mechanisms involving the number of precursor proteins intracellularly available in comparison to external toxin molecules attacking the cell (Fig. 7). Furthermore, coadaption of the host cell to the amount and biological activity of the secreted toxin seems to play an essential role in surviving physiological and elevated extracellular toxin concentrations. Transcriptome analysis of both strains further revealed significantly higher expression level of various stress-related genes in the K1-resistant strain KIM01, pointing to generally higher stress tolerance compared to its derivative KIM01s. In addition, lipidome analysis showed distinct alterations in the membrane composition, which could impact toxin insertion into the membrane or K1 binding to membrane-associated proteins. Based on preliminary data of a whole-genome sequencing of both killer strains, we could also exclude chromosomal mutations in genes essential for the cellular lipid metabolism which could potentially be responsible for the observed phenotype (unpublished data). Moreover, we were also able to show that killer yeast cells are not entirely resistant to the pore-forming capacity of the toxin, which could help to understand the molecular mechanism involved in K1 immunity. It is possible that killer yeast that are not able to cope with the high levels of applied cellular stress are killed by their own toxin or at least more susceptible to external environmental stress. This would not only provide extra resources to the residual killer yeast population but would add additional selective pressure on the killer yeast community. Thus, the system used in this study consisting of a natural K1 killer and an aphenotypic isogenic strain helps to elucidate the importance of cellular adaptations of a yeast cell expressing active toxin variants (Fig. 7). Although KIM01s shows no features of a self-killer phenotype, its distinct sensitivity against additionally applied K1 toxin could provide an interesting basis for further studies on K1 toxin biology and the sensitive interaction between viruses and host cells from an evolutionary point of view.
MATERIALS AND METHODS
Strains and culture media.S. cerevisiae strain BY4742 (MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0) was used as a sensitive strain. K1 toxin was produced by S. cerevisiae superkiller strain T158c (MATα his4C-864). RNA extraction was performed on the K1 killer strain KIM01 and its derivative KIM01s [MATα ura3-52 his3 trp1 pho3 pho5 M(1) L-A(1)] (17). Cultivation was conducted in standard yeast extract-peptone-dextrose medium at 30°C.
Toxin production and in vivo toxicity assay.Toxin production and concentration were conducted as described before (16). Activity of K1 toxin was analyzed via agar diffusion assay using SC-glucose MBA supplemented with appropriate amino acids and nucleotide supplementation (0.5% ammonium sulfate, 1.92% citrate, 2% glucose, 0.17% yeast nitrogen base, 1.5% agar, and methylene blue [pH 4.7]) as described previously (16). Likewise, the mediation of intrinsic immunity of KIM01 and KIM01s against externally applied K1 toxin was analyzed.
Analysis of toxin secretion and biological activity.Secretion of K1 toxin was analyzed via SDS-PAGE under nonreducing conditions, followed by Western analysis. Yeast strains were cultivated in JG medium for 3 days at 20°C, aliquots of the equivalent optical density at 600 nm (OD600) were centrifuged, and proteins of the cell-free supernatant were precipitated via trichloroacetic acid supplementation. Toxin heterodimers were detected by using a polyclonal primary antibody (α-K1, 1:1,000, rabbit) and a secondary antibody coupled with horseradish peroxidase (α-rabbit, 1:5,000, goat; Roche, Basel, Switzerland). The biological activity of potential secreted K1 toxin was checked by an MBA diffusion assay embedding BY4742 and spotting 10 μl of a fresh overnight-grown culture of the respective yeast; the killer activity was documented after incubation of the plates for 3 days at 20°C. Spheroplast preparation of BY4742 was conducted as described previously using Zymolyase T20 treatment (15).
Yeast transformation and selection.Transformation of KIM01s was performed by the standard lithium acetate method (34), and transformants were grown on uracyl-deficient (Ura d/o) synthetic medium supplemented with the appropriate amino acids and nucleotides (0.17% yeast nitrogen base, 0.5% ammonium sulfate, and 2% glucose). Immunity against wild-type K1 toxin was assayed using MBA (Ura d/o, 3% galactose [pH 4.7]) by embedding 106 cells of the transformant and spotting 10 μl of an overnight culture of the superkiller T158c.
Isolation of dsRNA.S. cerevisiae dsRNA was isolated by using TRI Reagent (Merck, Darmstadt, Germany), followed by purification on a silica filter spin column (35). Extracted dsRNA was visualized by agarose gel electrophoresis.
Flow cytometry.Pore formation was determined via propidium iodide (PI) staining and subsequent flow cytometry (FACS). Yeast cells were grown in JG medium (pH 4.7) overnight at 20°C (110 rpm) and washed with 10 mM McIlvaine buffer (pH 4.7). PI in a final concentration of 1 μg/ml (stock solution, 1 mg/ml in phosphate-buffered saline; Merck) was added. Flow cytometry was performed using a BD LSRFortessa Cell analyzer, and data were further evaluated using BD FACSDiv software (BD Bioscience, Heidelberg, Germany). Unstained and heat-treated cells were used for instrument settings; samples were measured in triplicates with 100,000 gated events each.
Lipidome analysis.Lipidome analysis of S. cerevisiae killer strains KIM01 and KIM01s was performed by Lipotype GmbH Services (Dresden, Germany). Sample preparation was conducted according to the manufacturer’s instructions.
Experimental design, RNA isolation, and Illumina sequencing.In analogy to similar conducted transcriptome analysis using the sensitive yeast S. cerevisiae BY4742, the KIM01 and KIM01s cells were metabolically preadapted to experimental setting conditions (20°C, 110 rpm). RNA isolation and sample and cDNA library preparation, as well as Illumina sequencing, demultiplexing, and quality and adaptor trimming, were performed as described previously (16). Raw data were deposited at the European Nucleotide Archive (ENA) under accession number PRJEB34586.
Analysis and visualization of transcriptome data.Data analysis and visualization of each replicate data set was conducted as previously described using Salmon (v0.8.2) for gene expression quantification (16, 36). DEseq2 (v1.18.1) was used to determine the DEGs using read counts obtained with Salmon (37). The resulting P values were false discovery rate (FDR) adjusted for multiple testing using the Benjamini-Hochberg approach (38). DEG analysis of both strains was conducted by comparing transcriptome changes after incubation for 5 and 45 min in the presence of K1 toxin with the corresponding untreated control samples. Also, pairwise comparison of each condition in the respective strains (control, t5, and t45) was performed. The DEG results were further interpreted by separating them into up- and downregulated genes and performing GO enrichment analysis using the GOrilla tool (39). Clustering and visualization of significantly enriched GO terms were conducted as previously described using REVIGO (16, 40).
ACKNOWLEDGMENTS
We thank Robert Ernst for helpful discussions during the preparation of the manuscript.
We declare there are no competing financial interests.
FOOTNOTES
- Received 22 October 2019.
- Accepted 26 November 2019.
- Accepted manuscript posted online 6 December 2019.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.
