No evidence for different molecular mating types in P. rhodozyma

The existence of a minimum of two MAT alleles or MAT idiomorphs among distinct individuals within a population, defining at least two compatible mating types, can be considered as indicative of the potential for heterothallism, which may co-exist with homothallism in some species [2]. In the tetrapolar (heterothallic) system, the HD genes are multiallelic (more than two alleles are found in the population of a given species) [1,9,16] while the pheromone precursor and pheromone receptor loci can be either multiallelic, as is the case in many mushroom species [9,12,42] or biallelic as in red yeasts [43] and the model smut fungus Ustilago maydis [8] among many others. In each of the three available genomes of P. rhodozyma [39] (Nicolás Bellora and Diego Libkind, personal communication) we identified two clusters encoding each a pheromone precursor and a Ste3 type receptor (henceforth referred to as P/R1 and P/R2) located at approximately 5 kb from each other (Fig 1A). The Ste3 type receptors encoded in the P/R1 and P/R2 clusters are clearly different from each other (S2 Fig). The occurrence in the same genome of multiple pheromone receptor genes is quite common in heterothallic species within the Agaricomycotina. In several well-studied cases, different combinations of P/R clusters define distinct mating type identities [9,42]. To assess whether additional pheromone receptors alleles potentially encoding different mating identities could be retrieved in P. rhodozyma, we obtained partial sequences for the pheromone receptor genes found in the P/R1 and P/R2 clusters of 14 individuals representing all four previously identified populations of P. rhodozyma (S3 Fig) [35]. This survey uncovered 10 variants for both the STE3-1 gene (from P/R cluster 1) and the STE3-2 gene (from P/R cluster 2) that are nevertheless very similar within the cohort formed by genes originating from the same cluster (S3 Fig). A comparison of the three full length amino acid sequences of Ste3-1 and Ste3-2 available (S2 Fig) shows a maximum of seven and three amino acid substitutions between variants of Ste3-1 and Ste3-2, respectively. These differences are found between strains belonging to different populations (CRUB 1149 from population A vs. CBS 7918 and CBS 6838 from population C), while the two strains belonging to the same population (CBS 7918 and CBS 6838) exhibit identical amino acid sequences. We conclude that these variants are very unlikely to encode functionally different receptors. Both the STE3-1 and STE3-2 phylogenies including all the identified variants reproduce well the previously reported relationships between the phylogenetic lineages (populations) within P. rhodozyma (S3 Fig). For the genes encoding the pheromone precursors we did not conduct a broad population survey but a comparison of the genes present in the three genomes available that encompass two distinct P. rhodozyma populations (A and C; S2 Fig) also suggests a high similarity between sequences belonging to the same cluster. Hence, taken together, these results strongly indicate that the P/R locus is not bi- or multiallelic in P. rhodozyma, as opposed to heterothallic basidiomycetes.

Download:

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

Fig 1. Deletion mutants in the P/R locus.

A. Organization of the P/R locus of strain CBS 6938 depicting gene clusters P/R1 (blue) and P/R2 (orange). B. Pheromones and pheromone receptors as are expected to be expressed the wild type cells. C. P/R locus genes expressed and sporulation phenotypes of mutants in which one or more P/R locus genes have been deleted. Plus signs (dark green) indicate that formation of basidia and basidiospores was observed, while minus (red) signs denote complete failure of the mutant to sporulate. D. Light microscopy photographs showing an example of positive and negative sporulation phenotypes, as indicated. Letters b and bs denote basidium and basidiospore, respectively. E. and F. Mean number of basidia observed per plate in sporulation proficient mutants after 10 days of incubation in sporulation medium at 18°C. Error bars represent standard deviations from the mean for the three biological replicates. No significant differences were observed between the WT and the sporulation proficient mutants (Student’s T test).

https://doi.org/10.1371/journal.pgen.1006110.g001

Many heterothallic basidiomycete species, both bipolar like Cryptococcus neoformans [44] and Ustilago hordei [45] or tetrapolar, like Leucosporidium scottii [11] or Ustilago maydis [46] harbor two functionally distinct, and thus mating type determining, Ste3 receptors. When phylogenies are constructed using the amino acid sequences of these two pheromone receptors identified in various basidiomycete lineages, trans-specific polymorphism is usually observed, meaning that each of the two Ste3 receptors found in one species is phylogenetically more closely related to the receptor in the corresponding mating type of a different species than to its counterpart in the opposite mating type of the same species [9,47,48]. To find out how the two genes of P. rhodozyma fit in this general picture, a phylogenetic analysis was conducted including the Ste3-1 and Ste3-2 receptors of P. rhodozyma CBS 6938 and their closest known relatives from heterothallic species harboring two mating type specific Ste3 receptors, from the Cryptococcus, Kwoniella and Tremella lineages. While reproducing the expected trans-specific polymorphism for all other species examined, the phylogeny in Fig 2 shows that the two P. rhodozyma receptors share a more recent common ancestor with each other than with the receptors found in the other species and consequently do not exhibit trans-specific polymorphism. In line with this, the protein sequences of Ste3-1 and Ste3-2 receptors have diverged considerably less than the two Ste3 alleles found in the other (heterothallic) species examined (50% amino acid identity, S1 Table and S2 Fig). On the other hand, they seem to be more divergent than some alleles identified in heterothallic species like Coprinopsis cinerea possessing multiple functional receptor paralogs per genome that can be up to 80% identical while exhibiting different pheromone binding properties [42].

Download:

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

Fig 2. Maximum likelihood phylogenetic tree of pheromone receptors (Ste3) from Tremellomycetes.

Phylogeny includes Ste3 protein sequences from the species within Tremellomycetes closest to Phaffia and is rooted with the sequences of both Pucciniomycotina species (L. scottii and S. salmonicolor) and S. cerevisiae Ste3 pheromone receptors.

https://doi.org/10.1371/journal.pgen.1006110.g002

We also identified in P. rhodozyma one pair of divergently transcribed HD-like genes (Fig 3), which are usually hyperpolymorphic in heterothallic, tetrapolar basidiomycetes [8]. The HD-like genes were located in a scaffold different from the one harboring the P/R clusters in the three available draft genome assemblies, the two loci being therefore probably genetically unlinked. A comparison of the complete protein sequences encoded by representatives of all four P. rhodozyma populations shows that the proteins are quite similar among different strains (S2 Fig). Similarly to what was observed for the P/R loci, we found that HD1 and HD2 gene sequences retrieved from strains representing all four P. rhodozyma populations recapitulate previously reported phylogenetic relationships between populations (S3 Fig). Interestingly, the homeodomain region of the HD2 protein is strictly conserved in all the variants identified, while variable amino acid positions seem to be evenly distributed throughout the remainder of the protein (S2 Fig). This is in contrast to what is usually observed for different alleles of HD2 in tetrapolar species, where the N-terminal dimerization domain is much more divergent than the C-terminal domain, as a result of negative frequency dependent selection imposed by the functional constraints on the N-terminus of both the HD1 and HD2 proteins [49]. Notably, HD2 variants representing different P. rhodozyma populations exhibited different C-terminal domain lengths, except for the shortest variant (S2 Fig) that was found in strains representing both populations B and C. A similar analysis conducted for HD1 protein sequences revealed one conservative amino acid substitution within the homeodomain, identical lengths for all the variants and an even distribution of polymorphic sites, as observed for HD2 (S2 Fig).

Download:

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

Fig 3. Deletion mutants in the HD locus.

A. Organization of the HD locus of strain CBS 6938 depicting the HD1 (green) and HD2 (pink) genes. B. Homeodomain transcription factors as are expected to be expressed in wild type cells. C. HD locus genes expressed and sporulation phenotypes of mutants in which one or both HD locus genes have been deleted. Plus signs (dark green) indicate that formation of basidia and basidiospores was observed, while minus signs (red) denote complete failure of the mutant to sporulate. Smaller plus sign was used to denote that only vestigial sporulation was observed in mutants expressing only HD1. D. Number of basidia observed per plate in sporulation proficient mutants after 10 days of incubation in sporulation medium at 18°C. Error bars represent standard deviations from the mean for three biological replicates. Asterisks denote significant difference between the WT and the hd2Δ mutant (Student’s t-test, p = 0,0004).

https://doi.org/10.1371/journal.pgen.1006110.g003

In summary, we conclude that no evidence could be found for the existence of polymorphisms at the MAT loci that might represent cryptic molecular mating types in a set of P. rhodozyma strains that captures the diversity found so far within the species.

One pheromone receptor is necessary and sufficient for sexual reproduction

In heterothallic systems, pheromones and pheromone receptors are usually involved in the process of cell-cell compatibility recognition preceding plasmogamy [1,13]. This process might conceivably be dispensable in self-fertile sexual reproduction but was, on the contrary, found to be relevant for homothallic systems in species of the genera Neurospora [50] and Sordaria [51] (Ascomycota). To address the question of whether pheromone receptor genes were required for sexual reproduction of P. rhodozyma, we produced deletion mutants of each of the two receptor genes STE3-1 and STE3-2 in turn, using homologous recombination to target chromosomal integration of antibiotic resistance markers to the P/R locus so as to delete each of the receptor genes, but leaving the pheromone precursor genes intact (Fig 1 and Table 1). Assessment of the phenotype of the individual pheromone receptor mutants, ste3-1Δ and ste3-2Δ, in sporulation medium showed that their sporulation capabilities were similar to the wild type suggesting that neither receptor is required per se for sporulation (Fig 1E). On the contrary, the double mutant, ste3-1Δ ste3-2Δ, lacking both receptors, failed completely to sporulate indicating that the two receptors are functional and redundant. We subsequently reintroduced the STE3-1 gene in the double mutant ste3-1Δ ste3-2Δ, through integration in the rDNA locus. As expected, functional complementation restored the capability of the resulting strain ste3-1Δ ste3-2Δ +STE3-1 to sporulate (Fig 1C).

Download:

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

Table 1. List of P. rhodozyma strains.

https://doi.org/10.1371/journal.pgen.1006110.t001

Evidence for reciprocal compatibility between receptors and pheromones encoded by the two clusters

Genes encoding distinct pheromone precursors (MFA1 and MFA2) flank each of the two pheromone receptor genes (STE3-1 and STE3-2) that are located at a short distance from each other in the genome (Fig 1A). This arrangement is reminiscent of the P/R locus model proposed for the mushroom Coprinopsis cinerea, a tetrapolar species also possessing multiple P/R clusters per individual [42]. Should the P. rhodozyma P/R clusters functionally resemble those of C. cinerea, the receptor would be expected to be insensitive to the pheromone encoded by its flanking gene. To find out how, if at all, the pheromones encoded in the P. rhodozyma P/R clusters interact with the receptors, we constructed two deletion mutants, ste3-1Δmfa1 and ste3-2Δmfa2. Neither of the resulting strains was able to sporulate, strongly suggesting that each receptor is activated by the pheromone encoded in the other cluster, i.e. Ste3-1 is for example very likely activated by Mfa2 (Fig 1C). In accordance with this, reintroduction of the MFA2 gene in the ste3-2Δ mfa2Δ mutant complemented its sporulation defect (Fig 1C). We subsequently constructed double deletion mutants ste3-1Δmfa2Δ and ste3-2Δmfa1Δ, each expressing a distinct, presumably interacting pheromone receptor/pheromone pair. Both double mutants were able to sporulate at normal levels (Fig 1C and 1F) showing that, as predicted, Mfa1 interacts with Ste3-2 while Mfa2 activates Ste3-1.

Finally, when the ste3-1Δ mfa1Δ and ste3-2Δ mfa2Δ mutants are co-cultured in suitable medium, a low level of sporulation is observed (Fig 1 and S2 Table), two orders of magnitude lower than wild type, which is consistent with the functional receptor remaining in each strain being activated by the pheromone produced and secreted by the other strain.

HD1 and HD2 are required for sexual development in P. rhodozyma

Divergently transcribed candidate HD1 and HD2 genes were also identified in the P. rhodozyma genomes, resembling the genomic arrangement of homologous genes found in most tetrapolar species (Fig 3). Hence, we proceeded to investigate whether the putative HD1 and HD2 genes uncovered in the genomes were involved in sexual development. To that end, single mutants, hd1Δ and hd2Δ, in which each of the two genes was deleted in turn, as well as a double mutant hd1Δhd2Δ, were constructed. Sporulation was abolished in the hd1Δ single mutant and in the hd1Δ hd2Δ double mutant, but vestigial sporulation was observed for the hd2Δ mutant (Fig 3 and S3 Table). Reintroduction of the HD1 gene in the rDNA of the hd1Δ strain restored the wild type sporulation phenotype, confirming that loss of sporulation was truly a consequence of HD1 deletion (Fig 3C and 3D). Similarly, ectopic expression of HD1 in the hd1Δ hd2Δ mutant restored a limited ability to sporulate to levels similar to those observed for the hd2Δ mutant (S3 Table). Taken together, these results suggest that sporulation requires both proteins but that the absence of HD2 does not completely block completion of the sexual cycle. The most likely explanation for these observations is that the divergently transcribed HD1 and HD2 genes in P. rhodozyma work together to regulate genes required for sexual development, unlike similarly arranged genes found in tetrapolar species across the Basidiomycota. To assess whether the HD1 and HD2 proteins might be capable of forming a heterodimer, we first used a bacterial two-hybrid assay [52] to try to detect an interaction between the HD1 and HD2 proteins of P. rhodozyma. The results, shown in S4 Fig, fail to consubstantiate the occurrence of an interaction between the two proteins sufficiently strong to be detected by this assay. To consubstantiate this apparent absence of interaction, we subsequently performed a second assay using the yeast two-hybrid system, which was previously used to detect interactions between HD1 and HD2 proteins of U. maydis [14] and C. neoformans [53]. To this end, fusions were constructed mimicking those successfully employed to detect interactions between the U. maydis proteins, including fusion proteins comprehending the complete HD proteins as well as shorter versions including solely the N-terminal domains normally involved in dimerization and the homeodomain region (S4 Table). In line with results obtained for the bacterial two-hybrid system, we did not detect a clear interaction using the two possible combinations of the short dimerization domains of HD1 and HD2 (S5 Fig) with only one of the four transformants expressing the HD1 and HD2 N-terminal domains showing some activation of the MEL1 reporter gene and no discernible activation of the remaining two reporter genes. However, a weak interaction signal denoted only by the MEL1 reporter gene was consistently detected in all combinations involving fusion proteins that comprehended one complete coding region of either HD1 or HD2 (S5 Fig). Taken together, these results lead to the tentative conclusion that the two P. rhodozyma proteins may interact, albeit weakly. Interestingly, in the yeast assay, we also observed interactions that might support the formation of homodimers (S5 Fig). Homodimers of homeodomain proteins were observed in S. cerevisiae where they have a well-defined function [54], but no biological roles have been ascribed so far to HD protein homodimers in the basidiomycetes.

Genetic evidence for the involvement of meiosis in the sexual cycle of P. rhodozyma

Close inspection of the P. rhodozyma genomes available revealed the presence of all but one (MLH2) of the core genes required to complete meiosis [39] (Nicolás Bellora and Diego Libkind, personal communication). However, some available evidence also argued against the occurrence of a typical meiosis: the ploidy of different P. rhodozyma strains is uncertain and may vary [40], extensive intraspecies chromosome length polymorphisms were noted [36,55,56] and segregation of markers was observed to deviate from Mendelian distribution [36]. This prompted us to examine the dependence of P. rhodozyma sporulation on SPO11, a core meiosis gene encoding an endonuclease [57] shown to be required for meiotic recombination in C. neoformans [58]. To achieve that, we compared the ability of a spo11Δ mutant to complete the sexual cycle and the viability of the spores produced with that of the wild type strain. The results, shown in Fig 4, indicate that sporulation is less efficient in the spo11Δ mutant than in the wild type, although this difference did not reach statistical significance. In addition, the viability of F1 spores isolated from the spo11Δ mutant was significantly lower than that of the wild type. These results suggest that meiosis is indeed part of the sexual life cycle of P. rhodozyma.

Download:

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

Fig 4. SPO11 deletion mutant.

A. Number of basidia observed per plate in sporulation proficient mutants after 10 days of incubation in sporulation medium at 18°C. Error bars represent standard deviations from the mean for the three biological replicates. No significant differences were observed between the sporulation ability of the wt and spo11Δ mutant. B. Total number of basidiospores recovered from wt and spo11Δ mutant and fraction of viable spores (asterisks denote significant difference, P value = 0.000302, Chi-square statistic).

https://doi.org/10.1371/journal.pgen.1006110.g004

Construction of heterothallic strains

Interestingly, the results described in the previous sections show that the genetic determinants and mechanisms involved in the homothallic life cycle of P. rhodozyma are similar to those of heterothallic, tetrapolar basidiomycetes. On the other hand, from the point of view of strain improvement of P. rhodozyma for biotechnological applications, it would be very useful to be able to generate strains in which outcrossing is strongly favored, because it facilitates the selection of strains harboring desirable combinations of characteristics using classical genetic approaches. This prompted us to use our recently acquired knowledge of the sexual reproduction system to try to generate obligate outcrossing P. rhodozyma strains. To do that, we generated artificial “mating types” by constructing two triple mutants harboring complementary components of the two mating recognition systems, ste3-1Δ mfa1Δ hd1Δ and ste3-2Δ mfa2Δ hd2Δ. Our model (Fig 5) predicts that a cross between these two triple mutants should almost exclusively yield spores resulting from conjugation between independent cells belonging to complementary “mating types”, while formation of basidia originating from single cells or from pedogamy is almost completely prevented by the absence of complete HD1/HD2 gene pair in each of the “mating types”. Hence, while in a cross between ste3-1Δ mfa1Δ and ste3-2Δ mfa2Δ mutants, extracellular diffusion of the pheromones would permit any of the three possible modes for formation of basidia (S1 Fig) because each mutant possessed a complete HD1/HD2 pair, in the triple mutant cross, sporulation is expected to occur almost exclusively after cell fusion brings together the HD1 and HD2 partners originating from different “mating types”. Indeed, in a cross between the artificially created “mating types”, basidia with basidiospores were formed, albeit at levels three orders of magnitude lower than wild type and two orders of magnitude lower than the cross between ste3-1Δ mfa1Δ and ste3-2Δ mfa2Δ (S2 Table). The occurrence of sporulation in the triple mutant cross, even if at very low levels, paves the way for strain improvement based on outcrossing.

Download:

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

Fig 5. Model describing molecular interactions required for sexual reproduction.

Pheromones and receptors encoded in the P/R locus were both shown to be reciprocally compatible, forming two functionally redundant pairs. The HD1 and HD2 proteins encoded at the HD locus are proposed to contribute synergistically for sexual development, possibly through a weak interaction allowing them to form a stable heterodimer upon DNA binding.

https://doi.org/10.1371/journal.pgen.1006110.g005

Strains and culture conditions

Escherichia coli DH5α (S6 Table) strain was used for all cloning steps and was grown in LB medium (10 g/L Tryptone, 5 g/L Yeast Extract and 5 g/L NaCl) with 100 μg/ml ampicillin at 37°C. Wild type Phaffia rhodozyma strain CBS 6938 was grown in YPD medium (10 g/L Yeast Extract, 20 g/L Peptone and 20 g/L glucose) at 20°C, while mutants strains were grown in YPD medium supplemented with appropriate antifungal drugs (50 μg/ml geneticin, 50 μg/ml hygromycin or 100 μg/ml zeocin). All wild type P. rhodozyma and mutant strains used and generated in this work are listed in Table 1.

Construction of the gene deletion fragments and complementation plasmids

Different gene deletion fragments (GDF) were generated in order to construct P. rhodozyma deletion mutants, using a common strategy, consisting in cloning the upstream and downstream flanking regions of the selected gene upstream and downstream of an antifungal resistance cassette, to promote integration of the GDF in the targeted genomic locus [34]. Standard molecular biology methods were employed [72,73] and three distinct plasmids were used as backbone, namely pPR2TN [74], pBS-HYG [75] and pJET1.2+ZEO (S7 and S8 Tables), encoding geneticin, hygromycin and zeocin resistance genes, respectively (S6 Fig). Specific primers for the STE3-1, HD1, HD2, SPO11, regions STE3-1/MFA1 and STE3-2/MFA2 flanking regions were designed to include restriction sites that allowed cloning of the amplified regions onto the chosen plasmids (S7 Table). The GDF for the STE3-2 gene was constructed by overlap extension PCR and then cloned into the pJET1.2 vector using CloneJET PCR Cloning Kit (Thermo Scientific) (S8 Table). Nested primers were used to amplify the complete GDFs by PCR using Phusion High-Fidelity DNA Polymerase (Thermo Scientific) (S7 and S8 Tables). PCR products were purified using GeneJET Gel Extraction Kit (Thermo Scientific) or Illustra GFX PCR DNA and Gel Band Purification Kit (GE Healthcare) and finally used to transform P. rhodozyma. Complementation of selected deletion mutants was accomplished using plasmid pUC18+rDNA+ZEO that was constructed by inserting the constructed 1.8 Kb zeocin resistance cassette (S9 Table) and a 3 Kb fragment of the rDNA from plasmid pPR2TN in plasmid pUC18 at the SmaI and SacI restriction sites respectively. DNA fragments containing the ORFs’ of the genes pertaining to the complementation with approximately 300 bp flanking regions, were subsequently cloned into the PstI and BamHI restriction sites of pUC18+rDNA+ZEO. Each of the four complementation plasmids were then linearized with ClaI within the rDNA sequence to promote integration of the plasmid into the ribosomal DNA of the mutant strains to be complemented (S10 Table) [76].

Transformation of Phaffia strains and mutant confirmation

Linearized plasmids and GDF were used to transform P. rhodozyma by electroporation as previously described (S7–S10 Tables) [76]. Transformants were selected in YPD medium with the appropriate antifungal drugs. Mutants with multiple deletions were obtained by transforming a confirmed deletion mutant with a second or third GDF. Correct integration of the disruption cassettes was verified by PCR (S7–S10 Tables) as previously described [34]. Briefly, a primer inside the resistance cassette and a primer outside of the flanking region present in the GDF at both the 5’ and 3’ extremities were used in diagnostic PCR reactions to identify the desired mutants. Absence of the gene targeted for deletion was also assessed by PCR for each mutant. Key deletion mutants were also confirmed by Southern blot in order to ensure that integration of the GDF occurred only once and in the correct locus. For Southern blot, 5μg of genomic DNA was digested with ClaI and run in a 0.8% agarose gel. Southern blot was performed using standard methods. Primers MP091-MP092 and MP062-MP063 were employed to amplify fragments of the resistance genes present in the geneticin and hygromycin cassettes to be used as probes. Labeling of the probes was performed with (α-³²P) dATP using the Prime-a-Gene Labelling system (Promega). Signals were detected on X-ray films (Hyperfilm MP, GE Heathcare Life Sciences) (S7 Fig).

Sporulation assays

In order to test the ability of the deletion mutants to sporulate, the various mutants were inoculated on DWR (2.5% agar and 0.5% ribitol) solid medium, incubated at 18°C and observed regularly for up to two months. Sporulation efficiency assays were performed as described previously by Kucsera [36]. Briefly, cells were grown on YPD medium overnight (180 rpms, 20°C, in 10% of the volume of the flask), collected by centrifugation and washed thoroughly with sterile distilled water to remove culture medium. Cells were subsequently distributed in 10 μl drops over the surface of DWR plates that were subsequently incubated at 18°C for 10 days. The number of basidia on each plate was determined by direct observation of the perimeter of the colonies using an optical microscope. Three independent assays were performed with CBS 6938 wild type (WT) strain and with all deletion mutants (in triplicate). Student’s t-test was performed (with a significance of 99%) to ascertain the statistical significance of the differences observed between the WT and each of the sporulating deletion mutants. Additional assays were conducted to determine the viability of F1 progeny of the spo11Δ mutant and wild type strain CBS 6938. Basidiospores were recovered by micromanipulation and were transferred to YPD solid medium to determine viability. Chi-square statistic was performed to verify if the difference between WT and spo11 deletion mutant was statistically significant.

Crosses of mutant strains

Mutant strains to be crossed were cultivated and washed as previously described, were mixed 1:1 and were finally distributed in 10 μl drops on the surface of DWR plates. Additionally, direct mixture of strains on DWR plates was also performed. Plates were incubated at 18°C and observed daily.

Bacterial two-hybrid assays

In order to ascertain the possible interaction between HD1 and HD2 protein bacterial two-hybrid assays were performed using the Bacterial adenylate cyclase two-hybrid system kit from Euromedex. The cDNA of each of the HD genes (S1 File) was synthesized at Eurofins and delivered as an insert in vector (pEX-K4). Synthetic HD genes were amplified by PCR (S11 Table) and sub-cloned into plasmids pKNT25 and pUT18 using the Hind III and Pst I restriction sites present in the MCS of both of plasmids, according to standard molecular biology techniques and the Euromedex manual [52]. Genes were cloned in frame at the N-terminus end of the T25 and T18 peptides. After transformation, integrity of the constructs was assessed by PCR amplification and sequencing of the cloned fragments (primers MP191/MP192 and MP193/MP194) (S11 Table). Different combinations of the recombinant plasmids (S4C Fig) were co-transformed into BTH101 E. coli cells (S6 Table). After successful transformation, the phenotype of 8 clones (identified as 1.1–10.8) of each of the different co-transformations was assessed in X-gal and MacConkey/maltose media upon incubation at 30°C for 48h. Results were scored after 24h and 48h incubation times. The presence of both fragments and correct plasmids in each of the clones was assessed by PCR (S11 Table).

Yeast two-hybrid assays

Matchmaker Gold Yeast Two-Hybrid System from Clontech was also used to assess a possible interaction between the homeodomain proteins of P. rhodozyma. Synthetic HD genes (S1 File) were cloned into pGBKT7 and pGADT7 plasmids by transforming each of the PCR fragments and the digested plasmid into the pertinent yeast strain (S4 Table and S5C Fig). All S. cerevisiae transformations were performed according to Yeastmaker Yeast Transformation System 2 User Manual from Clontech. Primers used to amplify the HD genes carried 40 bp 5’ tails homologous to the ends of the linearized plasmids to promote recombination. Plasmids pGBKT7 and pGADT7 were linearized with Pst I and Cla I, respectively. Two distinct versions of the HD genes were used, a shorter one comprising the complete N-terminal and homeodomain region of each of the genes (the first 183 amino acids of HD1 protein and the first 196 amino acids of HD2 protein respectively) and another, comprising the complete proteins. Each of the haploid S. cerevisiae strains generated was tested for the ability of the fusion protein expressed to activate any of the reporter genes on their own (S5E Fig). Haploid S. cerevisiae strains (S5D Fig) were them mated and diploid strains selected as described in the Matchmaker Gold Yeast Two-Hybrid System user manual. Three diploid strains (named X, Y and Z), selected from each mating experiment were tested for their ability to activate the reporter genes. Plates were incubated at 30°C for 72h and photographed daily (S5 Fig).

Construction of Ste3 phylogeny

Previously published pheromone receptor sequences were used to reconstruct the phylogenetic tree depicted in Fig 2. The accession numbers of sequences used are listed in S12 Table. Available draft genome sequences of Kwoniella mangrovensis CBS 10435, K. heveanensis BCC 8398 and Tremella fuciformis tr26 were searched for the presence of pheromone receptor homologues by TBLASTN using P. rhodozyma Ste3-1 as query. Genomic regions corresponding to positive hits were retrieved (GenBank accession numbers ASQD01000019.1, ASQB01000005 and LBGW01000351, respectively) and protein sequences were deduced after removal of likely introns, either manually or using AUGUSTUS [77]. The final protein dataset was aligned using an iterative refinement method (L-INS-i) in MAFFT v.7.221 [78]. Poorly aligned regions were removed with trimAl v.1.2 [79] using the "gappyout" option. The resulting alignment containing 338 positions was analyzed in ProtTest v.3.2 using the corrected Akaike information criterion (AICc) to determine the model of sequence evolution that best fitted our data. A maximum likelihood-based phylogenetic tree was built in RAxML v.8.1.24 using PROTGAMMAILGF model of amino acid substitutions and branch support was determined using 1000 rapid bootstraps. The Saccharomyces cerevisiae Ste3 pheromone receptor was used to root the tree.

Construction of MAT phylogenies

MAT genes STE3-1, STE3-2, HD1 and HD2 were partially amplified and sequenced using primers listed on S7 and S8 Tables. All sequences obtained were deposited in Genbank and are listed in S12 Table. Nucleotide unrooted maximum likelihood phylogenies were inferred with General Time Reversible model and 1000 bootstrap replications on MEGA5.1 software [80]. The trees with the highest log likelihood are shown with branch lengths measured in the number of substitutions per site.