New Cytoplasmic Virus-Like Elements (VLEs) in the Yeast Debaryomyces hansenii

Yeasts can have additional genetic information in the form of cytoplasmic linear dsDNA molecules called virus-like elements (VLEs). Some of them encode killer toxins. The aim of this work was to investigate the prevalence of such elements in D. hansenii killer yeast deposited in culture collections as well as in strains freshly isolated from blue cheeses. Possible benefits to the host from harboring such VLEs were analyzed. VLEs occurred frequently among fresh D. hansenii isolates (15/60 strains), as opposed to strains obtained from culture collections (0/75 strains). Eight new different systems were identified: four composed of two elements and four of three elements. Full sequences of three new VLE systems obtained by NGS revealed extremely high conservation among the largest molecules in these systems except for one ORF, probably encoding a protein resembling immunity determinant to killer toxins of VLE origin in other yeast species. ORFs that could be potentially involved in killer activity due to similarity to genes encoding proteins with domains of chitin-binding/digesting and deoxyribonuclease NucA/NucB activity, could be distinguished in smaller molecules. However, the discovered VLEs were not involved in the biocontrol of Yarrowia lipolytica and Penicillium roqueforti present in blue cheeses.


Introduction
In addition to the genetic information in the form of DNA existing in the nucleus, yeast may have extra genetic elements located in the cytoplasm in the form of virus-like particles (VLPs) containing dsRNA as well as circular or linear dsDNA plasmids [1,2]. Linear dsDNA plasmids at present called virus-like elements (VLEs) due to their likely viral origin represent the predominating supplementary genetic material in yeasts. However, the number of discovered VLEs carrying strains is not very large. As a result of analysing over 1800 strains belonging to approximately 600 yeast species, molecules of this type have been found only in 40 strains [1,[3][4][5][6][7][8][9][10][11][12][13]. Most of them belong to the Ascomycota phylum (34 strains) and the rest to the Basidiomycota phylum. In the latter, all discovered VLEs containing strains pertain only to one species, Tausonia pullulans, formerly detected two additional strains having such molecules, CBS 1780 (pDH2A/B; 6.5/15 kb) and CBS 4890 (pDH3A/B; 4.7/13 kb). However, the function of all discovered VLEs in D. hansenii cells remains unknown.
Previously, we have been investigating the function of different yeast species including D. hansenii in the cheese ripening process and the possibility to utilize selected strains to shorten the cheese maturation [44][45][46]. We have found that D. hansenii strains originating from the cheese environment express a strong killer activity [39,47,48]. Furthermore, we have developed an industrial technology to produce preparations containing D. hansenii killer toxins for plant protection against fungal phytopathogens [49]. The technology has been implemented by Skotan S.A. (Poland) and the product named HanseFlora®is now available in Poland. In the current study, we have expanded our interests to the dissemination of VLEs in D.
hansenii strains mainly occurring in cheese but also in other ecological niches. We tested strains deposited for years in culture collections as well as freshly isolated from the environment. By analyzing the sequence of three selected new VLE systems, we tried to find/predict a benefit for the host coming from having such elements in the cell.

Frequency of VLEs in D. hansenii Yeast
A set of 135 D. hansenii yeast strains was analyzed for the occurrence of VLEs; 75 of them were obtained from culture collections and 60 strains were freshly isolated from the environment (Tables 1 and 2). Culture collection strains originated mainly from dairy sources but also from ice and seawater, alcoholic and non-alcoholic drinks, meat, insects, and human skin. New isolates were gained from blue-veined cheeses, which are known to be abundant yeast microbiota [40]. To collect a large diversity of samples, eight different types of cheese produced in four European countries were selected. Isolation of yeast strains was performed on a specific medium named YORS and developed particularly for this work. This medium was based on YPD agar (Y) supplemented with oxytetracycline (O), rose bengal (R), and salt (S). Such composition allowed the selective isolation of halophilic yeast such as D. hansenii while inhibiting the growth of bacteria, filamentous fungi, and other yeast species present in blue-veined cheeses.
The type strain of D. hansenii, CLIB 197 T , forms colonies of characteristic morphology on YORS agar: pink color, smooth surface, circular shape with pulvinate elevation, and entire margin ( Figure S1). Randomly selected strains looking similar to the type strain were chosen for further experiments. All of them were identified as D. hansenii by colony PCR with a species-specific pair of primers-DhPADF and DhPADR developed by Wrent et al. [50], giving the product of 402 bp. In each case, a band of approximately 400 bp was obtained ( Figure S2). The application of this method significantly reduced the time and cost of the analysis compared to amplification and sequencing of the D1/D2 domain of the large ribosomal subunit [51]. The relatively small size of the PCR product allowed us to carry out the PCR with biomass instead of purified DNA, thus, avoiding the time-consuming nucleic acid isolation required by the aforementioned method but also during the identification of D. hansenii by means of intergenic spacer rDNA amplification with subsequent restriction fingerprinting [52,53].
Among yeasts obtained from culture collections, no strain possessing VLEs was detected directly after reviving or after repeated cultivation in a medium containing sodium chloride, which can facilitate the proliferation of cells containing such elements [54]. The procedure of repeated cultures was used for D. hansenii CBS 770 and D. hansenii CBS 7848. It proved to be effective since no VLEs were observed in DNA preparations isolated from cultures directly after recovery from the freeze-dried preparation, while after six passages in YPD with 12% NaCl VLEs were successfully detected ( Figure S3).  Turek niebieski Poland At present, it is difficult to find out whether the lack of the cytoplasmic DNA elements in the studied strains was due to their lack in the primary culture or have been caused by isolating these strains on standard yeast media or the following methods of preservation and storage. Strains of CIRM-Levures are stored in liquid nitrogen or in freeze-dried form, while cultures from DBFM are henceforth stored at −80 • C in a glycerol solution. However, for many years they have been stored under paraffin on YM agar slants at a temperature of 4 • C and have been regularly transferred once a year onto a fresh medium. Such a procedure could have led to the gradual loss of plasmids. No experiments about VLEs content have been conducted before culture conservation in both collections, but it should be implemented from now on. Consideration should, therefore, be given to the development of proper storage or revival conditions for strains deposited in culture collections, which could subsequently enhance chances for maintaining cytoplasmic DNA molecules in cells.
It appeared far more efficient to search for VLEs in fresh isolates of D. hansenii species obtained directly from cheeses with mould overgrowth. In this case, 25% of them contained such molecules (Table 3). There was high variability in the number of VLEs possessing strains between cheeses; VLEs were observed in half of the cheese samples tested. In most cases, low detection rates in the rest of the trials could be correlated with the small number of analyzed strains (four isolates). In cheeses, where yeasts with VLEs were detected, their frequency ranged from 17% to as much as 50% of isolates, which suggested an important contribution of these elements to yeast cell functioning in such an ecosystem. We also observed that in a given cheese both may occur, strains having the same VLE system, as in the case of pDH4A/B/C in the German Edelpilz cheese, as well as coexisting strains with different sets of such elements, as in strains isolated from the Polish Turek niebieski cheese (Table 3, Figure 1). In the latter, four VLEs possessing strains were detected, i.e., 8e, 8g, 8h, and 8i, and each of them had a different set of such molecules. Based on current data, it cannot be distinguished whether they arise from different ancient rearrangements of an ancestral VLEs system or whether there are undergoing current modifications.  Table 3. * Genetic material isolated from strain without VLEs. M. Gene Ruler 1kb Plus DNA Ladder (Thermo Fisher Scientific, Warszawa, Poland).
In total, eight different VLEs systems have been distinguished: four binary and four ternary ( Figure 1, Table 3). The biggest molecule in each system had a consistent size of 15.1 kb for all tested strains, which is longer than the longest previously known element pSKL from Lachancea kluyveri of 14.2 kb [55]. The size variation was observed for the smallest and medium-sized elements. The length of the smallest one ranged from 4.8 to 9.2 kb and the medium one from 7.5 to 10.0 kb.

The Characteristics of Selected VLEs Systems
Three newly discovered sets of VLEs (pDH4A/B/C, pDH5A/B, and pDH6A/B) were selected for next-generation sequencing. They were isolated from strains acquired from cheeses produced in Germany, Poland, and Denmark (strains 4e, 5c, and 7g). The libraries were made from DNA fractions containing a mixture of linear DNA plasmids separated from mitochondrial and nuclear DNA by ultracentrifugation in CsCl-bisbenzimide gradient ( Figure S4). This approach was not previously used in the preparation of yeast VLEs for sequencing. So far, labour-intensive purification of individual molecules and their subsequent Sanger-based sequencing has been conducted [24,26,[55][56][57][58].
Overlapping reads were assembled into contigs which were compared with GenBank nucleotide database using blastn [59]. It allowed identifying VLE sequences and contaminations with mitochondrial DNA. Such contamination was probably created at the fraction collection stage after separation in the CsCl-bisbenzimide gradient. Due to the small difference in buoyant density between mtDNA and VLEs, both fractions are faintly divided from one another and therefore easy to mix with each other. Nuclear DNA fragments have also been identified among the obtained sequences. Their presence indicates that despite careful operation chromosome fragmentation has not been avoided. Apart from the random fragments of different chromosomes, the sequences of nuclear origin belonged in many cases to the rRNA encoding genes, which are located in many copies in three loci in the genome of D. hansenii CBS 767 T [60]. In addition, fragments of LTR-retrotransposon Tdh5 have also been identified [61]. Frequent detection of the latter may be associated with the presence of more than 20 copies of this element in the D. hansenii genome and subsequent generation of many extrachromosomal DNA copies of this element. While the length of Tdh5 retrotransposons is approximately 5.3 kb, their isolation together with the plasmid fraction was also possible.
Contigs assigned to VLEs constituted almost complete plasmid sequences. In the case of four of them (pDH4C, pDH5B, pDH6A, pDH6B), only the fragment of one terminal inverted repeat (TIR) sequence was missing, while for the other plasmids, sequences at both ends had to be filled by primer walking. Open reading frames were annotated using CLC Sequence Viewer v.8.0.0 (Qiagen Digital Insights). This software was selected based on previously analysed ORF detection accuracy on already known VLEs sequences: pGKL2, pPE1B, pPac1-1, and pSKL. The minimal ORF coding capacity was set to 70 aa as genes encoding 70 aa polypeptides were discovered on pGKL2 and pSKL and proved to be functional [62]. The annotated sequences of pDH4A/B/C, pDH5A/B, and pDH6A/B were deposited in GenBank under the following accession numbers: MF795091/MF795092/MF795093, KX904874/ KX904875, and KX858805/ KX904876, respectively.
All new VLEs turned out to be extremely A+T rich (73−76%) and compact in structure, exactly as for previously known plasmids. Their coding capacity was also very high; 75−83% and 88−90% for non-autonomous and autonomous molecules, respectively. At both ends of all analysed VLEs, terminal inverted repeats (TIR) strictly specific to the element type were found (Figures 2 and 3). Those with a length of 715−716 bp detected on autonomous elements turned out to be about 230 bp longer than the longest TIRs previously reported on pSKL. In contrast, TIRs from the smallest VLEs in the systems also proved to be the smallest known (91 bp) and were identical to the 91-first bases of pDH1A TIRs, which are 96 bp in length [58]. In the case of the middle non-autonomous elements, inverted repeats located at the ends were composed of TIR sequences from the small molecules extended for 33 bp.
Each of the newly sequenced VLEs had its own DNA polymerase-encoding gene that enables its replication in the cytoplasm independently of the nucleus using a previously disclosed protein-primed mechanism [reviewed by 30]. The sequence of this gene was closely related to the type of element, highly conserved for autonomous (99−100% identity) and non-autonomous (95−98% id.) molecules, while much more diverse between genes from these two types of VLEs (~29% id.).
The confirmation of distinctiveness of these two kinds of DNApol can also be found in the phylogenetic tree illustrating the relationships between various groups of proteins that are representatives of the type B superfamily of DNA polymerases ( Figure 4). Similar phylogenetic analyses have been performed before but on a smaller number of sequences [57,58,63,64]. This time, the analysis was expanded due to an increase in the number of available data. Nevertheless, proteins encoded by autonomous and non-autonomous plasmids formed separated groups within the clade of VLEs-origin DNApols. It can be hypothesized that both groups derived from a common ancestor in which the DNApol gene was probably duplicated, then single copies were divided into two linear molecules of different functions and further differentiated. Moreover, our results indicated that VLEs-origin DNA polymerases are closely related to DNApol of adenoviruses and bacteriophages of both Gram-positive and -negative bacteria, which suggests a common origin. The structure of the phylogenetic tree obtained in this work showed that mitochondrial DNApols differentiated earlier from the common progenitor than those of yeast cytoplasmic VLEs, Adenoviridae, and phages. Furthermore, the evolution of plasmid DNApol of mitochondria apparently followed more complicated paths. The basis for this presumption is that the DNApol of mitochondrial linear plasmids of fungi (Ascomycota and Basidiomycota) are arranged in three different clades distant from cytoplasmic and viral B-type polymerases. Therefore, our results confirm the conclusions already reached by Klassen and Meinhardt [64] from the analysis of a smaller pool of sequences.  Tables 4 and 5. The arrows of light blue color correspond to ORFs encoding enzymes involved in replication, dark blue in transcription, and yellow of unknown function. The products of ORFs colored in purple show some similarities with immunity determinants to VLE killer toxins. The direction of arrows indicates the transcription direction; black triangles represent terminal inverted repeats (TIRs) at the ends of plasmids.  It seems that genes of DNA polymerases have been pulled into VLEs that were sourced from a different origin. The research of Sýkora et al. [65] and Vopálenský et al. [66], who analyzed the transcriptional mechanism of pGKL from K. lactis, showed that RNA polymerase encoded on the autonomous molecules as well as the promoter sequences of all genes lying on both autonomous and non-autonomous molecules probably derived from the protoplast of poxviruses. Furthermore, the DNA polymerases found in viruses of the Poxviridae family do not show homology to the polymerases of VLEs origin.
Sequence analysis of pDH4C, pDH5B, and pDH6B showed their extraordinary conservation, with 99% identity. Therefore, in those molecules, all open reading frames were located precisely in the same orientation and order, and 10/11 of them had the same length, start, and stop codons ( Figure 2, Table 4). Individual plasmids differed only in the length of the ORF9 located on the right end. In pDH6B, this ORF was the longest (1104 bp), while in pDH4C and pDH5B, two and one nucleotide insertions, respectively, resulted in the shift of the reading frame and the early appearance of a stop codon, which shortened the ORF length to 861 and 786 bp. A similar genetic organization of core 10 ORFs has already been reported for pGKL2, pPE1B, pPac1-1, and pSKL [22,24,25,55]. However, no homology to DhORF9 was found in these plasmids, although, the additional ORF (ORF1) was found on the left side of pSKL and pGKL2.
The predicted products of all extra ORFs located on VLEs of different species show some similarities to proteins providing immunity to plasmid-encoded killer toxins ( Table 5).
The putative DhORF9p appears to be most closely related to the immunity determinant of W. robertsiae encoded by pWR1A ORF5 (35% identical amino acids on 90% of protein length) and of M. acaciae pPac1-2 ORF4 (36% id. aa of 83% of p.l.), which are also closely related to each other (49% id. aa of 97% p.l.). No significant similarity of DhORF9p was found to the zymocin immunity protein encoded by pGKL1 ORF4. The deduced immunitylike proteins of the pSKL and pGKL2 probably belong to a different group of proteins resembling immunity as they have 48% common amino acids on the length of 80% of the protein sequence and they are homologous to the mentioned zymocin resistance protein (25% id./52% p.l.).  The function of the immunity-like genes is as yet unknown. However, Schaffrath et al. [70] demonstrated using pGKL2ORF1p as an example that this protein does not guarantee resistance to zymocin. In addition, they also showed that this protein does not participate in the replication processes and maintenance of the VLE system.
DehaORF9 appears to be intact with a putative upstream control region (UCR) acting as yeast plasmid gene's promoter [71][72][73]. This sequence contains the ATA TGA key fragment located 38 bp upstream from the start codon. Such sequences (consensus: ATNTGA), called upstream conserved sequences (UCS), usually precede the start codon about 100 bp for autonomous and around 30 bp for non-autonomous plasmids [25,30,70,74,75]. UCS in the promoter sequence is necessary for the correct binding of RNA polymerase of VLEs origin and handling of the transcription of plasmid genes in the cytoplasm.
The sequences of new non-autonomous VLEs were searched for genes, which could be involved in killer toxin production. In all new systems, ORFs containing either a full domain of chitinase activity (GH18) or their fragments were detected (Figure 3, Tables 6 and 7). Polypeptides with such function are an inseparable part of VLE-encoded killer toxins [27,29]. However, they are also present on cryptic elements which function is still unknown [9,58]. The most interesting ORF containing GH18 is located on pDH5A. The predicted protein encoded by ORF2 contains the full domain of chitinase along with the domain responsible for chitin binding (ChBP1). There is also an incomplete autolysin domain of 1,4-beta-N-acetylmuramyl-hydrolase activity. The autolysin contains elements of the LysM domain that is responsible for binding to peptidoglycan in bacterial cell walls as well as fungal chitin which is structurally similar [76]. LysM domains were also detected in proteins encoded on plasmids pDH1A, pGKL1, pPin1-3, pPac1-2, pWR1A, and pPE1A ( Figure 3).
Additionally, in pDH5AORF2p, a fragment of the deoxyribonuclease domain (DNase NucA/NucB) was also found. Parts of DNase NucA/NucB were also present in other pDH systems, i.e., in ORF2 of pDH4B and in ORF3 of pDH6A. The prediction of protein localization with DeepLoc-1,0 [77] revealed that pDH5AORF2p and pDH6AORF3p can be extracellular proteins. However, only in pDH5AORF2p was the signal peptide of 18 aa detected confirming the possible secretion of this protein outside the cell. Interestingly, in each of the pDH systems, a protein that probably functions in the nucleoplasm (pDH4AORF3p, pDH5AORF4p, and pDHORF7p) was detected. These proteins are almost identical and highly homologous to the hypothetically cytoplasmic pDH4AORF2p, pDH5AORF3p, and pDH6AORF6p proteins. It could be assumed that these proteins may be an element of the host's genetic material defense against the killer activity of DNases.  However, these assumptions need to be confirmed in laboratory experiments. DNase activity has not yet been identified in any protein derived from linear plasmids, including killer ones. However, PaT and zymocin killer toxins indirectly affect DNA integrity in M. acaciae and K. lactis. These toxins cleave selected tRNAs, which leads to damages in the DNA repair systems and accumulation of untidy mutations [29,[78][79][80].
The pDH5AORF2p protein may undergo modifications with KEX2 protease because five motifs recognizable by this enzyme can be found in its sequence: one in the autolysin domain, two in the chitinolytic domain, and two before the fragment of the putative DNase. The post-translational modifications by KEX1/KEX2 proteases are necessary for the secretion of active killer toxins encoded both chromosomally and by VLEs (for a review, see [81]). Further, ORF2p secondary structure analysis with Phobius [82] showed that probably the fragment involved in chitin-binding and its cleavage is localized on the outer side of the cell membrane of the target yeast or mould cell, then, there is one transmembrane region followed by the element of the DNA digesting activity located on the cytoplasmic side.
In the case of the remaining proteins potentially associated with killer activity (pDH4BORF2p, pDH6AORF2-5p), none of them have a signal peptide, the presence of which would suggest the possibility of extracellular secretion. Only pDH4BORF2p has a 31-amino acid transmembrane region located between the fragments identified by the Phobius software [82] as cytoplasmic and non-cytoplasmic. The transmembrane region (beta subunit) is present, e.g., in zymocin, and is probably responsible for the transport of the active gamma subunit to the cytoplasm of the attacked cell [30].
Some ORFs on the non-autonomous VLEs are not conserved in the other D. hansenii systems. An example is ORF3 from pDH4B, which has not been detected in other D. hansenii VLEs. However, ORF3p showed homology to the immunity-like protein of ORF1 from the autonomous plasmids pSKL and pGKL2. Their function is still unknown but blastp search showed some similarities with alpha-mannosyltransferases, which may suggest that they could be involved in protein glycosylation processes.

The Relationship between New VLEs and Killer Activity
Considering the above data, we decided to check the relationship between new VLEs and killer activity in practice. For that purpose, selected yeasts possessing VLEs were cultured for six weeks in media without selection pressure (without NaCl) and at an elevated temperature (30 • C). Similar conditions caused the loss of the pDHL1/2/3 system in cells of D. hansenii CBS 7848 [8]. The lack of VLEs was first confirmed by electrophoretic separation of total DNA isolated from yeast cells, and then, by polymerase chain reaction ( Figure S5). Gunge et al. [8] have shown that cells can gradually lose their VLEs from a system, eventually losing their autonomous molecules. Therefore, primers enabling the amplification of a 963 bp DNA polymerase fragment from the autonomous plasmids were designed for PCR reactions. As it is shown in Figure S5, the applied procedure of treatment enabled the complete removal of VLEs from D. hansenii cells.
After confirmation of the VLEs' lack, a cross test was performed with Y. lipolytica PII6a yeast and P. roqueforti PR1 mould used as sensitive organisms (Figures S6-S9). Both species coexist with D. hansenii in blue-veined cheeses and could be the target of their VLE-encoded toxins due to the presence of chitin in their cell wall [83,84]. No killer activity was observed at neutral pH, which is optimal for VLE-origin killer toxins, zymocin, and PaT [30,85]. The test was performed at 14 • C, which is used for blue-veined cheese ripening as well as at 28 • C which is the standard temperature for yeast cultivation. In addition, the loss of immunity of D. hansenii strains to its own toxin due to the loss of linear plasmids was also not observed, neither at pH 4.6 nor 7 ( Figure S10). The killer effect of the studied strains on Y. lipolytica and P. roqueforti was observed at pH 4.6, which is optimal for killer toxin activity of D. hansenii strains isolated from cheese [39,42,86]. However, no difference in lethal activity between VLEs possessing and cured of them isolates could be noted (Figures S11 and S12). Thus, the antifungal activity of the yeast D. hansenii is not encoded within VLEs. It is in accordance with results obtained previously by Gunge et al. [8] for the pDHL1/2/3 system.

The Relationship between Plasmids and Halotolerance
Bearing in mind that the function of linear cytoplasmic genomic elements of D. hansenii remains cryptic, it was also examined if their presence increases the yeast resistance to a high concentration of sodium chloride. Although sustaining VLEs in D. hansenii cells may be dependent on the presence of salt in the environment [54], there was no change in growth profiles of yeast deprived of such elements in comparison with strains possessing them in YPD medium containing NaCl in the range of 0 to 24% (for an example see Figure S13). Another possibility for D. hansenii to maintain VLEs is its putative toxicity against lactic acid bacteria present in cheese. This hypothesis remains to be tested.

Conclusions
Although it is known that VLEs are the most widespread type of additional genetic information in yeast, we still do not know everything about them. This is due, among other things, to the fact that until now sequences of only three complete VLE systems and four individual linear dsDNA molecules have been deposited and analyzed. In this work, the available data were expanded to include sequences of the full three new systems from one yeast species-D. hansenii, including, for the first time, a system consisting of three elements. These plasmids, apart from the genes associated with their own replication and transcription, have ORFs resembling genes of killer activity and resistance. However, our experimental studies showed no relationship between killer activity and the presence of VLEs in host cells and, therefore, the function of VLE-encoded proteins is still not deciphered. Further analyses are needed to check the conditions under which VLE-derived genes are expressed and correlate them with organisms that may be targeted for potential VLE-encoded toxins. It would be also interesting to isolate Debaryomyces strains harboring VLEs from different environments than dairy products. The linear plasmid purification procedure followed by NGS used in this work can be utilized for rapid sequencing of similar systems in other organisms and can contribute to a better understanding of their functions.

New Yeast Strain Isolation
Yeast isolation was performed on YORS agar (yeast extract 10 g L −1 , peptone 20 g L −1 , glucose 20 g L −1 , oxytetracycline 0.1 g L −1 , rose bengal 0.033 g L −1 , NaCl 120 g L −1 , agar 15 g L −1 ). The medium without oxytetracycline was sterilized at 121 • C for 20 min. After cooling to approximately 90 • C, sterilely weighed oxytetracycline was added to the medium and quickly mixed until dissolved. Cheese samples of 5 g were homogenized in 45 mL of physiological salt solution (Stomacher Lab Blender, Seward, Worthing, UK) and diluted in the same liquid. Decimal dilutions were spread onto the surface of YORS agar. Colonies were taken from plates after 5 days incubation at 25 • C and streaked onto YPD agar plates with 3% NaCl.

Yeast Isolate Identification
As we wanted to analyze only D. hansenii strains, new yeast isolates were initially examined by colony PCR with D. hansenii species-specific pair of primers DhPadF/DhPadR developed by Wrent et al. [50] under their conditions except for the initial denaturation step, which was prolonged from 5 to 10 min. The reaction was performed with a small portion of 24 hour-old biomass taken from a plate with a 10 µL tip and mixed with 10 µL of 2 × Taq Nova Red Master Mix (Blirt-DNA-Gdansk, Gdansk, Poland), 0.4 µL of 10 µM each primer, and 9.2 µL of distilled water. The product was visualized with ethidium bromide after electrophoresis in 1.2% agarose gel at 120 V for 40 min.

Plasmid Detection
The VLEs presence in cells of new yeast strains was studied immediately after isolation, while for culture collection strains, it was carried out twice: first directly after reviving and then after four propagations in YPD medium with gradually increased NaCl concentration in the range of 0−9%. Each culture was shaken at 25 • C at 160 rpm (G10 Gyrotory Shaker, New Brunswick Scientific Co., Edison, NJ, USA) and after 7 days 1 mL was transferred to a new medium. The last culture was made in YPD containing 12% of NaCl and took 24 h. Then, the biomass was centrifuged at 6000 rpm (Sigma 16−32K), washed with distilled water, and used for DNA isolation.
The total DNA was obtained with GeneMATRIX Bacterial and Yeast Genomic DNA Purification Kit (Eur x , Gdansk, Poland) according to the manufacturer's protocol. Plasmids were separated from nuclear and mitochondrial DNA in 0.8% agarose gel with 1×TAE buffer at 90 V for 1.5 h and visualized with ethidium bromide.
VLEs for NGS sequencing were isolated according to an adapted protocol of Querol et al. [87]. Briefly, cells were harvested, lysed with zymolyase 100T (ICN Biochemicals, Aurora Ohio USA), and treated with SDS. Total DNA was precipitated within ethanol and resuspended in TE buffer. CsCl (1 g mL −1 ) and 10 µL mL −1 of 10 mg mL −1 stock solution of bisbenzimide (Hoechst; Thermo Fisher Scientific Poland, Warszawa, Poland) were added to total DNA. Then, the plasmid fraction along with mitochondrial DNA was separated by ultracentrifugation (Beckman L7-55 Ultracentrifuge, Beckman Coulter, Indianapolis, IND, USA) after 60 h at 50,000 rpm at 20 • C. The fraction with plasmid DNA was purified four times with isoamyl alcohol buffer made of 100 mL of isoamyl alcohol, 1 mL Tris-HCl 1 M pH 8, 20 g CsCl, and 20 mL water. DNA was precipitated with Na-acetate and isopropanol and finally resuspended in 50 µL of TE buffer. Gene annotation on VLEs was performed with CLC Sequence Viewer (QIAGEN Digital Insights). The function of open reading frames was at first predicted by sequence similarity on the basis of blastx [67]. Then, in case of no hits, the Conserved Domains Database [68], as well as the Pfam database [69], were used to search for conserved domains.

Phylogenetic Analysis
Amino acid sequences of DNA polymerases were aligned in CLC sequence viewer v.8.0.0 (QIAGEN Digital Insights). The same software was used for the construction of the phylogenetic tree based on the neighbor-joining method with Jukes-Cantor protein distance measure after 1000 bootstrap replications.

VLE Curing
To force the yeast cells to lose linear cytoplasmic DNA, strains were cultivated on YPD medium without NaCl at 30 • C, with shaking at 160 rpm for 7 days and transferring 1 mL into a fresh medium. Such procedure was repeated 8 times. Then, the total DNA was isolated and inspected on agarose electrophoresis as described previously. Furthermore, the isolated DNA was used as a template for PCR reaction with primers designed with Primer3Plus [88] specific for the DNA polymerase gene from autonomous D. hansenii VLEs: pCpolDNAF1 (5 TTG GTG CGA TAT AGA TGG AAA A 3 ) and pCpolDNAR1 (5 TGC TTT TTA AAC CAG CTC CA 3 ). The reaction mixture was: 0.4 µL DNA mixed with 10 µL of 2 × Taq Nova Red Master Mix (Blirt-DNA-Gdansk), 0.4 µL of 10 µM each primer, and 8.8 µL of distilled water. The PCR reaction was performed in a Personal Thermocycler (Biometra, Göttingen, Germany) under the following conditions: initial denaturation at 95 • C for 5 min; 35 cycles of denaturation at 94 • C for 1 min, primers annealing at 49 • C for 30s, extension at 72 • C for 1 min; and final elongation at 72 • C for 10 min. The product was visualized with ethidium bromide after electrophoresis in 1.2% agarose gel at 120 V for 40 min.

Antifungal Activity Test
The connection between antifungal activity and the presence of VLEs was tested according to the method described by Woods and Bevan [89]. Strains with such elements and cured of them were streaked in a thick line on a lawn of a sensitive Y. lipolytica PII6a strain (5 × 10 5 cells mL −1 of YPD-MB agar pH4.6 or pH 7.0 containing 2.5% NaCl) or P. roqueforti PR1 (1 × 10 4 cells per plate of PDA + 2.5% NaCl pH4.6 and 7.0) and cultivated at 14 • C and 28 • C for 72 h. An analogous experiment was performed with D. hansenii strains possessing VLEs streaked on a lawn of D. hansenii cured of them. The experiment was performed on YMB pH 7.0 with 2.5% NaCl. The streaked strain was considered a killer yeast if the line of the biomass was surrounded by a clear growth inhibition zone of the sensitive strain. In the case of P. roqueforti, changes in sporulation were also inspected. All experiments were run in four replicates.

Acknowledgments:
The authors would like to thank Noémie Jacques for providing strains from the CIRM-Levures collection.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.