Mating-Type Switching in Budding Yeasts, from Flip/Flop Inversion to Cassette Mechanisms

SUMMARY

Mating-type switching is a natural but unusual genetic control process that regulates cell identity in ascomycete yeasts. It involves physically replacing one small piece of genomic DNA by another, resulting in replacement of the master regulatory genes in the mating pathway and hence a switch of cell type and mating behavior. In this review, we concentrate on recent progress that has been made on understanding the origins and evolution of mating-type switching systems in budding yeasts (subphylum Saccharomycotina). Because of the unusual nature and the complexity of the mechanism in Saccharomyces cerevisiae, mating-type switching was assumed until recently to have originated only once or twice during yeast evolution. However, comparative genomics analysis now shows that switching mechanisms arose many times independently—at least 11 times in budding yeasts and once in fission yeasts—a dramatic example of convergent evolution. Most of these lineages switch mating types by a flip/flop mechanism that inverts a section of a chromosome and is simpler than the well-characterized 3-locus cassette mechanism (MAT/HML/HMR) used by S. cerevisiae. Mating-type switching (secondary homothallism) is one of the two possible mechanisms by which a yeast species can become self-fertile. The other mechanism (primary homothallism) has also emerged independently in multiple evolutionary lineages of budding yeasts, indicating that homothallism has been favored strongly by natural selection. Recent work shows that HO endonuclease, which makes the double-strand DNA break that initiates switching at the S. cerevisiae MAT locus, evolved from an unusual mobile genetic element that originally targeted a glycolytic gene, FBA1.

INTRODUCTION

The mating behavior of a yeast cell is specified by the set of genes that it contains at its mating-type (MAT) locus. In almost all ascomycete yeast species, typified by Saccharomyces cerevisiae, there are three types of cell, defined by their mating behavior: two types of haploid cell with genotypes MATa and MATα and one type of diploid cell with genotype MATa/α. Each of these cell types can replicate to produce another cell of the same type vegetatively by mitosis and budding (Fig. 1). This asexual cell cycle is the main way that yeast cells proliferate, but many species can also reproduce sexually in a cycle that involves mating followed by meiosis and sporulation. Mating is the process by which two haploid cells of opposite mating types (MATa and MATα) fuse to form a diploid cell (MATa/α). Conversely, meiosis followed by sporulation is the process by which a diploid cell turns into four spores that germinate into haploid cells (two with genotype MATa and two with MATα) (Fig. 1).

FIG 1.

Role of mating-type switching in the life cycle of S. cerevisiae.

About half of the known budding yeast species have a known sexual cycle (1). These are called teleomorphic species. Their predominant vegetative form can be either haploid or diploid, and the species is termed haplontic or diplontic accordingly. For example, most natural isolates of S. cerevisiae are diploid, whereas most natural isolates of Kluyveromyces lactis are haploid. The difference between these two lifestyles depends on how they use a nutrient starvation signal (usually a lack of nitrogen) as a control point in the sexual cycle: some species such as S. cerevisiae use starvation as a trigger for meiosis and are therefore diplontic, whereas other species such as K. lactis use starvation as a trigger for mating and are therefore haplontic (2–4). Thus, the diploid phase is stable in S. cerevisiae (diploid cells will replicate vegetatively for generation after generation until they are starved and sporulate in response), whereas the diploid phase is unstable in K. lactis (zygotes [i.e., newly formed diploid cells] usually sporulate as soon as they are formed by mating, without undergoing diploid vegetative cell cycles). Conversely, the haploid phase is unstable in S. cerevisiae (haploid cells mate as soon as they detect a haploid cell of the opposite mating type, without requiring any other signal), whereas the haploid phase is stable in K. lactis (haploid cells will not mate, even in the presence of haploids of the opposite mating type, unless they are starving).

The other half of the known budding yeast species have no known true sexual cycle and are called anamorphic species. By a true sexual cycle, we mean a complete cycle that includes meiosis as well as mating. The prime example of an anamorphic yeast is Candida albicans. In fact, because yeast taxonomists historically used mating compatibility to define species using Ernst Mayr’s biological species concept, the genus Candida was originally established as a catch-all genus that contained all the ascomycete yeasts that do not have an apparent sexual cycle—regardless of the evolutionary relationship among them. Thus, Candida is a very diverse genus and is not monophyletic (5). C. albicans has a diploid genome and is postulated to be an ancient hybrid (6). Around the year 2000, genome sequencing led to discoveries that C. albicans has orthologs of the MAT genes of S. cerevisiae, which were termed MTL (mating-type-like) genes, and that most strains of C. albicans are MTLa/α heterozygotes (7, 8). Shortly after, C. albicans homozygous diploids with MTLa/a and MATα/α genotypes were found to be able to mate with each other, after transitioning into a mating-competent (“opaque”) state that has no counterpart in the S. cerevisiae life cycle (9, 10). Similar mating ability between MTLa/a and MTLα/α cells has been found in other Candida species too (11). However, meiosis has never been detected in any of the diploid Candida species, even though they have orthologs of many S. cerevisiae meiosis-specific genes (12). Instead, C. albicans has a parasexual cycle consisting of mating followed by “parameiosis,” which is a phase of concerted chromosome loss and increased recombination during which tetraploid cells lose chromosomes in an apparently random manner, forming cells of lower ploidy that can mate if their MTL locus becomes homozygous (13–15). Some other species in the largest Candida clade (which includes C. albicans) have haploid genomes and have been shown to have complete sexual cycles, including meiosis, which resulted in them being given new genus names—for example, Clavispora lusitaniae (formerly called Candida lusitaniae) (16).

STRUCTURE OF THE MAT LOCUS AND ITS FOUR CANONICAL GENES

The yeast MAT locus is not a typical genetic locus in many respects. At most loci in any genome, different alleles have very similar DNA sequences and usually only differ by a few nucleotides, but this is not the case at the MAT locus. The genes contained in the MATa and MATα “alleles” are highly dissimilar from each other. For this reason, the two alternative sequences at the MAT locus are often called idiomorphs rather than alleles.

Each of the MAT idiomorphs typically contains two genes—one coding for a transcription factor (a2 or α1) that acts in haploids and one coding for one subunit of a dimeric transcription repressor (the a1/α2 heterodimer) that acts in diploids (Fig. 2). The nomenclature convention is that MAT is used in the gene names (MATa1, MATa2, MATα1, and MATα2) and the names of the idiomorphs (MATa and MATα), but not in the corresponding protein names (a1, a2, α1, and α2). The a2 protein is a MATa-specific transcription factor that activates expression of the a-specific genes (asgs)—that is, genes whose expression is specifically required to enable MATa cells to mate, such as MFA1, encoding the a-factor pheromone, and STE2, encoding the α-factor receptor. Similarly, the α1 protein is a MATα-specific transcription factor that activates expression of the α-specific genes such as MFα1, encoding the α-factor pheromone, and STE3, encoding the a-factor receptor. In addition to expressing asgs or αsgs, haploid cells also need to express many haploid-specific genes (hsgs) that are shared by the mating pathways in both types of haploid cell, such as GPA1, STE4, and STE18, the genes coding for subunits of the G-protein complex that transmits the signal that a pheromone has been detected on the cell surface (17). Diploid cells need to repress the hsgs, and this is the function of the a1/α2 repressor. Because it is a heterodimer, with one subunit provided by each of the two MAT idiomorphs, the repressor exists only in diploid cells. The MAT locus thereby acts as a sensor of its own heterozygosity, and hence it is a sensor of the cell’s ploidy.

FIG 2.

Functions of the four canonical MAT genes in a typical budding yeast species. In haploids, the a2 and α1 proteins activate the mating programs specific to each cell type. In diploids, the a1/α2 heterodimer represses these programs.

The previous paragraph described a “typical” MAT locus, such as the one in Candida albicans, for example (18). It is somewhat unfortunate for us as authors that S. cerevisiae was the first budding yeast species to have its MAT locus characterized, because although the S. cerevisiae MAT locus is very well known (19), it is atypical in two regards. First, S. cerevisiae has no MATa2 gene and no a2 transcription activator. Instead of using a2 to activate transcription of its asgs, S. cerevisiae instead expresses asgs by default in haploid cells, unless they are repressed by a complex consisting of Mcm1 and two molecules of α2; this complex is present in α-cells and in a/α diploids, where asgs need to be repressed (20). The loss of MATa2 in the S. cerevisiae lineage occurred as part of an evolutionary rewiring of the cell-type determination pathway, which also involved the gain of a new interaction between α2 and Mcm1, at approximately the same time as the whole-genome duplication event (WGD) (18, 20, 21). All post-WGD budding yeast species lack MATa2, whereas all non-WGD species have it. Second, S. cerevisiae contains two other MAT-like loci, called HML and HMR, and uses these loci to replace the DNA at its MAT locus, converting MATa cells into MATα cells or vice versa in a process called mating-type switching. However, since (i) most yeast species cannot switch mating types and (ii) most of those that are able to switch do not use the same switching mechanism as S. cerevisiae, we will first introduce nonswitching species and species with simpler switching systems before describing the “prototype,” S. cerevisiae. Presenting the species in this order provides a better evolutionary perspective on how the complex switching system used by S. cerevisiae arose (22).

ANCESTRAL SYNTENY AT THE MAT LOCUS IN ALL BRANCHES OF PHYLUM ASCOMYCOTA

As genome sequencing began to accelerate in the mid-2000s, it soon became apparent that orthologs of the MAT genes are present in almost every budding yeast species, regardless of whether it has a previously described sexual cycle. Moreover, these genes are often at a conserved place in the genome. An early study (21) found that the MAT loci in species from several yeast families, including Lachancea kluyveri (Saccharomycetaceae), Ogataea angusta (Pichiaceae), and Yarrowia lipolytica (Dipodascaceae), were all located beside the gene SLA2 and that linkage of SLA2 to the MAT locus also occurred in Neurospora crassa, which is a filamentous ascomycete (subphylum Pezizomycotina), unlike the budding yeasts (subphylum Saccharomycotina). A decade later, Riley et al. (23) reported evidence that the chromosomal location of the MAT genes has been conserved across the whole phylum Ascomycota: they identified five genes (SLA2, SUI1, NVJ2, APC5, and APN2) that are close neighbors of the MAT genes in at least one species in each of the three subphyla of Ascomycota (Saccharomycotina, Pezizomycotina, and Taphrinomycotina).

This ancestral synteny is surprising and important in view of the changing content of DNA-binding proteins encoded at the MAT locus. In subphylum Saccharomycotina, the transcription repressor subunits a1 and α2 are two very divergent homeodomain proteins, and the transcription activators a2 and α1 are two very divergent HMG domain DNA-binding proteins (α1 was initially described as having a unique “α domain,” but it was later shown that the α domain is a distant member of the HMG domain family [24]). Other ascomycete subphyla also contain homeodomain and HMG domain genes at their MAT loci, but these appear to be from other subfamilies of these domains: for example, Schizosaccharomyces pombe has a homeodomain protein (Pi) that is quite unlike either a1 or α2 from Saccharomycotina, and it has HMG domain proteins (Pc and Mc) that are quite unlike a2 and α1 from Saccharomycotina. In addition, some regulatory genes, such as Sch. pombe Mi and Neurospora crassa mat-A2, seem to lack orthologs outside their own subphyla (23). It is clear that the DNA-binding proteins of the MAT locus are changing rapidly, either by fast sequence evolution or by gene duplication and turnover.

The conservation of the MAT locus’s location in the genome is also a useful feature because it makes the MAT genes easier to find. Because they evolve very quickly, it can be difficult to find the MAT genes in a newly sequenced genome by using the MAT proteins from another species as queries in BLAST (TBLASTN) searches, unless the phylogenetic distance between the species is low. However, we have found that it is often possible to find the MAT genes by first using TBLASTN searches to locate the orthologs of conserved neighboring genes such as SLA2 and then using BLASTX to search the genomic regions near them for traces of homology to known MAT proteins from other species. We have also noticed that the intron/exon structure of the four canonical MAT genes (a1, a2, α1, and α2) is highly variable among species—probably more so than in other genes, though we have not quantified this.

MAT LOCI ARE PRESENT IN VIRTUALLY EVERY YEAST GENOME

A landmark study in yeast comparative genomics was published in 2018 by Shen et al. (25). By sequencing the genomes of more than 200 budding yeast species, including yeasts from several clades that had previously received little attention, they greatly expanded the set of budding yeast genomes available for analysis—to 332 species—and established a solid phylogenomic tree for the subphylum Saccharomycotina. A follow-up study by Krassowski et al. (26) investigated the MAT loci present in the 332 genomes, which provides an overview of mating systems throughout the subphylum. We will make extensive reference to their findings in the rest of this article.

Using BLAST searches as described above, Krassowski et al. (26) were able to identify a MAT locus (containing at least one MAT gene) in 330 of the 332 genome sequences. Even species that have no apparent sexual cycle almost universally contain MAT genes, as was previously found for the main clade of Candida species (12). It therefore seems likely that there are few truly asexual yeasts. For many of the species that are currently considered anamorphic, the presence of a MAT locus in their genome suggests that they must have some sort of sexual cycle (perhaps one involving mating but not meiosis, like C. albicans) and that we have not found the correct conditions, or the correct mating partner, to induce them to mate.

The only species with no apparent MAT genes were Lodderomyces elongisporus and Candida sojae (26). The absence of MAT genes in L. elongisporus remains a puzzle more than a decade after it was first discovered (12). Their absence has been confirmed by analysis of multiple other strains of L. elongisporus as well as the type strain (12, 27). However, the absence of MAT genes reported in C. sojae (26) is probably an artifact. Although no MAT genes are present in the assembly of C. sojae strain GF41 (28), which was the only C. sojae genome sequence available at the time of the 332-species survey (26), a recently published and more contiguous assembly of the type strain of C. sojae contains MATa1 and MATa2 genes in a standard arrangement (29). Comparison of the two assemblies shows that GF41 lacks approximately 100 kb around the MAT locus, which must either be an assembly error or a strain-specific deletion. Thus, based on genome sequence analysis, there is a MAT locus in 331 of the 332 Saccharomycotina species studied by Krassowski et al. (26).

INFERENCE OF THALLISM STATE FROM GENOME SEQUENCES

Taxonomists have traditionally classified each sexual (teleomorphic) yeast species as either homothallic or heterothallic, based on the mating behavior of strains of that species (30). In a homothallic species, any strain can mate with any other strain. In contrast, in a heterothallic species, each strain has a permanent mating type (either MATa or MATα; they cannot switch mating types), and mating occurs only between strains of opposite mating types.

The thallism state of a species can often be inferred from the set of MAT genes present in its genome, even when only one strain of the species has been sequenced, as is usually the case. This approach was first used by Yun et al. (31), who showed in the filamentous ascomycete genus Cochliobolus that heterothallic species had only one idiomorph at the MAT locus (called MAT-1 or MAT-2), but in homothallic species, MAT-1 and MAT-2 genes were both present beside each other, apparently due to recent genomic rearrangements that fused the two idiomorphs and converted heterothallic ancestors into homothallic descendants (these would be called primary homothallics in the terminology used in the next section). The same approach was later used by Riley et al. (23) in a study of 16 budding yeast species of biotechnological importance. They noted a pattern of congruence between the content of MAT genes and the previously known thallism state of each species, as follows. If only MATa or only MATα genes were present, the strain could be inferred to be haploid and the species inferred to be heterothallic. If both MATa and MATα genes were present, but they were located on contigs whose flanking sequences were identical, the MAT genes could be inferred to be allelic, and therefore the strain is diploid and the species is heterothallic. In contrast, if both MATa and MATα sequences were detected, but they were in different (nonallelic) places in the genome, the species could be inferred to be homothallic because only homothallic species contain both types of gene at different locations (for example, in S. cerevisiae the MAT, HML, and HMR loci all contain MAT-like genes and are at three different places).

Krassowski et al. (26) applied this approach to the 332-genome data set of Shen et al. (25) to infer the thallism state of each species. Genomes were categorized as containing a MATa locus if either of the genes MATa1 or MATa2 was detected by BLAST and as containing a MATα locus if either MATα1 or MATα2 was detected; this rule was applied both because there are some well-documented evolutionary losses of individual genes from MAT loci (32) and because some MAT genes might not be detected in BLAST searches because of their rapid evolution.

Among the species in which MAT genes were detected (331 species, after correcting C. sojae), 193 (58%) were inferred to be heterothallic from their MAT locus content, and the remaining 138 (42%) were homothallic (26). Homothallism and heterothallism were not restricted to particular clades but were distributed all over the phylogenetic tree: of the 12 major clades of Saccharomycotina in the tree, there were heterothallic species in all 12 and homothallic species in 9 (26). The homothallic species can be further subdivided into primary and secondary types of homothallism, as described in the next section.

TWO VERY DIFFERENT TYPES OF HOMOTHALLISM IN YEASTS: PRIMARY AND SECONDARY

In classical mycology, the term homothallism was used to indicate that a teleomorphic species did not have separate mating types (33–35). In other words, any strain of the species could mate with any other strain to produce fertile offspring. The concepts of homothallism and heterothallism were originally used in the context of filamentous fungi and literally meant whether the thalli (mycelia) of different strains of a species were all one homogeneous type or two heterogeneous types, in terms of their mating ability. The same concepts were also applied to unicellular fungi (yeasts): in a homothallic yeast species, mating and sporulation could be observed in a monosporic culture (i.e., a culture derived from a single germinated spore), whereas in a heterothallic yeast species, mating and sporulation only occur if two monosporic cultures of opposite mating types are mixed together. In 1949, Öjvind Winge and Catherine Roberts (36) identified a single gene that differentiated homothallic from heterothallic S. cerevisiae strains: the allele for homothallism is dominant, and the gene was originally called D for diploidization, but is now called HO for homothallism. However, subsequent investigation showed that there are two very different forms of homothallism in different yeast species, called primary and secondary homothallism. Although these terms were first used in 1949 (34), the molecular basis of the distinction between primary and secondary homothallism as they apply to yeasts was not understood for a long time (30, 35). In short, in a secondary homothallic species, mating occurs only between cells with opposite MAT genotypes (MATa and MATα), but cells can change their mating type by mating-type switching, so the species is phenotypically homothallic (any strain can mate with any other strain). In a primary homothallic species, all cells contain the same MAT genes (usually all four canonical genes—MATa1, MATa2, MATα1, and MATα2) in the same places, and mating can apparently occur between any cell and any other cell, although many aspects of primary homothallism in budding yeasts remain uninvestigated. Two other forms of homothallism—pseudohomothallism and unisexual reproduction—are known but are rare in Saccharomycotina and will not be discussed here (30, 35, 37).

SECONDARY HOMOTHALLISM BY MATING-TYPE SWITCHING

Flip/Flop Inversion Systems

Secondary homothallism occurs by means of mating-type switching. Although mating-type switching was first characterized in S. cerevisiae (38) and Sch. pombe (39), we will begin by describing the simpler flip/flop switching systems that were discovered much more recently in species such as Ogataea polymorpha, because (i) they are the most common type of switching system found in yeasts, and (ii) we hypothesize that the more complex S. cerevisiae and Sch. pombe systems evolved from progenitors of this kind.

O. polymorpha is a member of the methylotrophic yeast family Pichiaceae. It has two MAT-like loci, about 19 kb apart on chromosome 3 (Fig. 3A). One contains genes MATa1 and MATa2, and the other contains genes MATα1 and MATα2 (40, 41). The 19-kb region is flanked by two identical 2-kb sequences that form an inverted repeat (IR). Recombination between the two copies of the IR leads to inversion of the 19-kb region, which has the effect of swapping the positions of the MATa and MATα genes relative to the rest of the chromosome and, most importantly, relative to the centromere of chromosome 3. The centromere lies close to, but outside, the invertible region—slightly to the left of the left copy of the IR—and it represses transcription of the pair of MAT genes that lie close to it (Fig. 3A). The details of how this repression occurs are not known, but centromeric histone Cse4 (CenH3) extends onto the pair of MAT genes located close to the centromere (41), and only the pair of MAT genes located further away from the centromere are expressed. When the 19-kb region becomes inverted, it therefore causes a switch of mating types. In one orientation, the MATα genes are close to the centromere and repressed, whereas the MATa genes are 19 kb away and expressed, so the cell is an a-cell. In the other orientation, the situation is reversed and the cell is an α-cell. Diploids are heterozygous for the two orientations of the 19-kb region.

FIG 3.

Organization of the MAT loci in three budding yeast species. Hatched lines indicate regions that are transcriptionally silenced. CEN and TEL indicate centromeres and telomeres, respectively. Asterisks in gene names indicate truncated duplicated genes located in the repeat regions. (A) Ogataea polymorpha, a species with a flip/flop system involving one IR (FF1). The regions highlighted in blue form the IR. In the orientation shown, only the MATa genes are expressed. (B) Komagataella phaffii, a species with a flip/flop system involving two IRs (FF2). The regions highlighted in blue form the outer IR, and the regions highlighted in purple form the inner IR. In the orientation shown, only the MATα genes are expressed. (C) Saccharomyces cerevisiae, a species with a 3-locus cassette system. The three identical X regions are highlighted in blue, and the three identical Z regions are highlighted in purple; they form direct repeats on S. cerevisiae chromosome III. In the arrangement shown, only the MATα genes are expressed.

Inversion of the 19-kb region in O. polymorpha is induced by nitrogen starvation, which is also a signal for mating (40, 41). Inversion is first detectable 4 to 6 h after nitrogen starvation begins and occurs on most carbon sources, except methanol (42). It is unknown whether inversion of the region involves a site-specific recombinase, but the homologous recombination (HR) DNA repair system is required (43). Induction of switching by nitrogen starvation in O. polymorpha also requires the EFG1, RME1, and STE12 transcription factors and is repressed by the a1/α2 heterodimer present in diploids (43, 44). Similar flip/flop inversion systems for mating-type switching have been characterized in two other Ogataea species: Ogataea minuta (45) and Ogataea thermomethanolica (46).

A similar but slightly more complex flip/flop inversion system controls mating-type switching in another methylotrophic yeast, Komagataella phaffii (formerly called Pichia pastoris) (Fig. 3B) (41, 47). In this species, repression of transcription of one MAT locus occurs because it is beside a telomere, rather than beside a centromere as in O. polymorpha. K. phaffii has two MAT-like loci situated more than 100 kb apart on chromosome 4. They are flanked by two distinct IRs—an inner IR and an outer IR (Fig. 3B). Recombination between the two copies of the outer IR causes inversion of the whole region and swaps the positions of the two sets of MAT genes relative to the telomere, resulting in a change of the MAT genes expressed (41). Diploid K. phaffii strains are heterozygotes for the two orientations of the 138-kb region between the two MAT loci, which also includes a centromere at its approximate center (Fig. 3B). The function of the inner IRs is unknown, but in natural isolates of K. phaffii, all possible combinations of MAT gene locations and orientation of the central 123-kb region are found, which indicates that recombination between the inner IRs must occur in nature (41).

Following the initial discoveries of flip/flop mating-type switching systems in O. polymorpha and K. phaffii, genome analysis uncovered two other examples of species that appear to switch mating types by flip/flop inversion (23): Ascoidea rubescens (family Ascoidaceae) and Pachysolen tannophilus (family Pichiaceae). Neither of these systems has been characterized in detail, but they both involve an invertible genomic region with MATa genes at one end and MATα genes at the other end, flanked by a single IR about 2 kb long. In A. rubescens, isolates with both orientations of the invertible region have been identified, and the invertible region is immediately adjacent to a telomere. In P. tannophilus, it was shown that nitrogen starvation induces inversion of the region between the IRs (23). Transcription of the MAT genes has not been investigated in detail in these species, but it has been suggested that one set of MAT genes is repressed by a telomere in A. rubescens and by a centromere in P. tannophilus (48).

It is important to note that in the flip/flop systems (Fig. 3A and B), there are only two MAT-like loci in the genome (a MATa locus and a MATα locus), and mating-type switching swaps their positions. Switching simultaneously moves the outgoing MAT genes from an expression site to a repression site and moves the incoming MAT genes from a repression site to an expression site. Mating-type switching by flip/flop does not involve the synthesis of new DNA, just recombination. These features are different from the 3-locus (3LOC) system of S. cerevisiae described in the next section.

The 3-Locus Mating-Type Switching System of Saccharomyces cerevisiae

The mechanism of mating-type switching in S. cerevisiae has been reviewed in detail by Haber and colleagues (49, 50), so we will present only a brief summary here. S. cerevisiae has three MAT-like loci called MAT, HML, and HMR (Fig. 3C). The MAT locus is transcribed and specifies the cell type (MATa haploid, MATα haploid, or MATa/α diploid), whereas HML and HMR are “silent” loci that are never transcribed due to silencing by SIR proteins. During mating-type switching in a haploid S. cerevisiae cell, the MAT locus is physically cleaved by an endonuclease (HO), making a double-strand DNA break in the chromosome. The cleaved MAT locus is then repaired by using either HML or HMR as a template for DNA synthesis. HML normally contains the DNA sequence necessary to repair the MAT locus to a MATα genotype, so it is designated HMLα. HMR normally contains the DNA sequence necessary to repair the MAT locus to a MATa genotype, so it is designated HMRa. The MAT, HMLα, and HMRa loci are at three different places in the genome: in S. cerevisiae they are all on chromosome III, but MAT is in the middle of the chromosome, HMLα is near the left telomere, and HMRa is near the right telomere. Transcription at HMLα and HMRa is silenced by local cis-acting silencer elements, but silencing is probably also reinforced by their location near the ends of chromosomes, which are subject to telomeric silencing (49). Each of the three loci contains an a-specific region or an α-specific region (designated Ya or Yα), which in S. cerevisiae are both approximately 700 bp long. The Y regions are flanked on each side by two regions (X and Z) that are common to all three loci and also a few hundred base pairs long. It is these X and Z regions that enable HMLα and HMRa to interact with the MAT locus and act as the templates for DNA repair during mating-type switching (Fig. 3C). Herskowitz and colleagues used the analogy of a cassette tape player to describe the S. cerevisiae 3-locus system: HML and HMR are alternative cassettes that can be inserted into the MAT locus for playback (51).

Mating-type switching in S. cerevisiae is a highly regulated process (49). Switching begins when HO endonuclease cuts the MAT locus, at the junction between the Y and Z regions (Fig. 3C). The ends of the double-strand DNA break at the MAT locus are resected (trimmed back), leaving long single-stranded tails that can interact with HMLα or HMRa in the X and Z regions that they share. The gap at the MAT locus is then repaired by DNA polymerase, which first synthesizes one DNA strand in the Z-to-X direction using HMLα or HMRa as a template and then uses the new strand as a template for second-strand synthesis in the opposite direction to complete the repair (52).

There is a bias in the choice of repair templates such that a cleaved MATa idiomorph preferentially chooses HMLα as a donor for DNA repair, and a cleaved MATα idiomorph preferentially chooses HMRa, so that the process of cleavage and repair succeeds in changing the mating type and is not futile. This bias is generated by a locus called the recombination enhancer (RE), which is located near HMLα (53). The RE contains a binding site for an α2-Mcm1 corepressor, which is present only in MATα cells. If the RE is bound by α2-Mcm1, it represses the use of HMLα as a donor for DNA repair, thereby favoring HMRa, whereas the use of HMLα as a donor is preferred if α2-Mcm1 is absent (54).

The process of switching takes about 1 h from initiation to completion (55), which is a long time relative to the cell cycle. It occurs during the G₁-to-S interval of the cell cycle, at which point the cell is already committed to DNA replication and mitotic cell division. Switching also occurs only in mother cells (in daughter cells, transcription of HO is repressed by Ash1), and as a result, switched cells appear in pairs (50). For example, after a MATa spore germinates, it will bud to form a MATa daughter. The mother cell will switch mating types during the G₁-to-S interval and then enter S phase and produce a second daughter. The mother cell and the second daughter will both be MATα, while the first daughter (and in turn, that cell’s first daughter) will remain MATa (50, 56). This process retains a 1:1 ratio between MATa and MATα cells in the population, which maximizes the chances of mating among them.

It is likely that most other species in family Saccharomycetaceae have switching systems that are similar to that in S. cerevisiae, although this conclusion comes mostly from inferences from genomics. The process of switching has been studied experimentally in only a few other budding yeast species with 3-locus systems (57–62). From comparative genomics, it seems to be universally true that HML is always on the same chromosome as the MAT locus—typically 100 to 200 kb away—and that the genotype of HML is always HMLα (63). This conservation of synteny between MAT and HML is probably due to the presence of the RE near HML, with natural selection to maintain the mechanism of donor bias. However, HMRa can be on any chromosome, even though all three loci are on the same chromosome in S. cerevisiae. In several species, there are multiple copies of HMRa on different chromosomes (63, 64).

Flip/Flop Switching Systems Originated Many Times Independently

In an article by Riley et al. (23) and in two review articles published in 2017 (22, 48), we interpreted the presence of flip/flop inversion systems in four genera (Ogataea, Komagataella, Pachysolen, and Ascoidea) that are quite distantly related to each other as evidence that mating-type switching by flip/flop inversion was an ancient process that was already present in their common ancestor, near the base of the subphylum Saccharomycotina. However, we now think that this interpretation is wrong. It assumed that flip/flop switching had a single origin and therefore that the many heterothallic yeast lineages that do not switch must have lost the ability to switch. The newer data from the 332-species survey (26) make this view untenable. The new data show that a single origin of switching in Saccharomycotina is highly unparsimonious, because it would require switching to have been lost again in a very large number of lineages. Instead, the most parsimonious explanation is that there were at least 10 independent, and recent, origins of flip/flop switching in Saccharomycotina, as described below.

In their study of 332 budding yeast genome sequences, Krassowski et al. (26) were able to identify 138 probable homothallic species because their genomes contained both MATa and MATα genes at nonallelic locations (Fig. 4). They were further able to distinguish between primary and secondary homothallism (in 41 and 97 species, respectively) based on the presence of DNA sequence repeats beside the MAT genes, which is a hallmark of species that can switch mating types (i.e., secondary homothallics). In species with flip/flop systems, the repeats are the IRs that enable inversion of the chromosomal region containing MAT genes. In S. cerevisiae and other species with a 3-locus MAT/HML/HMR system, the repeats are the X and Z regions that facilitate the use of the HM loci as DNA repair templates. In all these known examples, the repeat sequences are almost 100% identical to each other in DNA sequence and at least several hundred base pairs long. They are usually formed from parts of protein-coding genes (often parts of the MAT genes themselves), and their copy number in the genome is only 2 or 3, so they can be distinguished easily from other types of repeat sequences, such as mobile genetic elements. Therefore, if a yeast genome contains both MATa and MATα genes at different (nonallelic) locations, and these genes are flanked by DNA sequence repeats that could act as IRs or X/Z regions, the strain can be inferred to be able to switch mating types. Moreover, we expect a genomic feature that can repress transcription, such as a centromere or a telomere, to be located near the MAT genes that are silenced.

FIG 4.

Evolutionary transitions between different mating systems in 332 budding yeast species (modified from reference 26). The three gray boxes show MAT locus arrangements consistent with heterothallism (center), primary homothallism (left), or secondary homothallism (right). Within secondary homothallism, species were classified into three groups (FF1, FF2, or 3LOC) depending on the arrangements of repeat sequences (shown in blue and purple) flanking the MAT genes. Within primary homothallism, species were classified into two groups (PHC and PHN) depending on whether the MATa and MATα genes in a genome were contiguous or not. Arrows show the inferred numbers of evolutionary transitions between systems (Fig. 5); black arrows mark transitions from heterothallism to homothallism. Modified from reference 26 (with reassignment of Candida sojae from NOMAT to HET) published under the terms of the Creative Commons Attribution License (CC BY 4.0).

Applying this approach to the 332 genomes, Krassowski et al. (26) identified 31 species predicted to be able to switch mating types by a flip/flop mechanism. These comprised 23 species whose MAT loci were flanked by a single IR, as in O. polymorpha, and 8 species whose MAT loci were flanked by two IRs, as in K. phaffii (Fig. 4). These flip/flop groups were designated FF1 and FF2, depending on whether they have one or two IRs. What was most surprising about this result is that the 31 species were widely distributed over the phylogenetic tree of Saccharomycotina (Fig. 5). The flip/flopper species did not fall into a single clade, but instead there were 8 separate clades of FF1 species and 3 separate clades of FF2 species.

FIG 5.

Transitions to homothallism on the phylogenomic tree of Saccharomycotina. Thickened branches represent homothallic species, colored as in the key. Thin branches represent heterothallic species. The tree topology and clade labels are from Shen et al. (25). Genera in which transitions to homothallism occurred are named. (This does not necessarily mean that all species in the genus are homothallic.) Modified from reference 26 published under the terms of the Creative Commons Attribution License (CC BY 4.0).

Most of these clades appeared to be very young. For example, in the genus Cyberlindnera, most species were classified as heterothallic with only MATa or only MATα genes, but a small clade of three species (C. saturnus, C. suaveolens and C. mrakii) had an FF1 arrangement with MATa and MATα genes 49 kb apart and flanked by an IR (Fig. 6). Comparison between the FF1 and the heterothallic (HET) Cyberlindnera species showed that one of the two MAT loci of the FF1 species was in the same genomic location as the MAT locus of the HET species (between the genes SLA2 and VPS75), whereas the second set of MAT genes and the repeat that formed the IR were insertions relative to the ancestral chromosomal organization. This arrangement indicates that the common ancestor of the clade of three Cyberlindnera species underwent an evolutionary transition from ancestral heterothallism to FF1-type secondary homothallism. The genomic data are consistent with the previously known homothallism of the three species in the C. saturnus clade and heterothallism of most other Cyberlindnera species (1).

FIG 6.

Recent transition from heterothallism to flip/flop mating-type switching (secondary homothallism) within the genus Cyberlindnera. The tree shows an expanded view of part of Fig. 5. Blue branches indicate species inferred to switch mating types by an FF1 mechanism, and black branches indicate species inferred to be heterothallic, based on their genome sequences (26). The cartoons show gene organization around the MAT loci in two species. C. jadinii is inferred to be heterothallic diploid, with allelic MATα and MATa loci on different contigs in the genome assembly. C. saturnus has MATα genes at a position syntenic with the C. jadinii MAT locus and MATa genes 49 kb away on the same contig. The two sets of C. saturnus MAT genes are flanked by an IR, so it is inferred to switch mating types by an FF1 mechanism similar to Ogataea polymorpha. Modified from reference 26 published under the terms of the Creative Commons Attribution License (CC BY 4.0).

In contrast to the multiple independent origins inferred for FF1 and FF2 secondary homothallism, the three-locus MAT/HML/HMR system as used by S. cerevisiae appears to have only a single evolutionary origin. In Krassowski et al.’s study, 66 species had this type of system (designated 3LOC), but they were all in the family Saccharomycetaceae (Fig. 5). Only one other 3-locus mating-type switching system is known, in Schizosaccharomyces pombe and other species in the genus Schizosaccharomyces (65–69). Schizosaccharomyces is a member of a different ascomycete subphylum, Taphrinomycotina, so it is an outgroup to the tree in Fig. 5. Even though these two 3-locus switching systems have extensive similarities, many of their molecular components are different, and the 3-locus systems of S. cerevisiae and Sch. pombe are analogous rather than homologous (22, 70, 71).

Mating-type switching equivalent to that in budding and fission yeasts does not seem to occur at all in the third ascomycete subphylum, Pezizomycotina, nor in basidiomycetes or any other fungi outside the phylum Ascomycota. Several Pezizomycotina species perform so-called unidirectional mating-type switching, predictably changing from self-fertile to self-sterile in approximately half of their meiotic offspring by undergoing a programmed deletion of MAT1-2 genes from within a fused MAT1-1/MAT1-2 locus by recombination between direct repeats (these species include Ceratocystis fimbriata [72], Sclerotinia trifolium [73], Chromocrea spinulosa [74], and Thielaviopsis cerberus [75]). However, we do not consider unidirectional switching in Pezizomycotina to be fully equivalent to mating-type switching as seen in budding and fission yeasts because (i) it is not reversible, and (ii) it is a switch from a homothallic genotype (genes from both idiomorphs are present at the MAT locus, on the same chromosome) to a heterothallic genotype (genes from only one idiomorph are present), rather than a switch that replaces one idiomorph by the other. An inversion that occurs recurrently in the MAT locus during meiosis in Sclerotinia scleriotiorum and is inherited by half the offspring is more similar to yeast mating-type switching because it is reversible, but both orientations of this inversion are self-fertile (76).

PRIMARY HOMOTHALLISM BY GENE RELOCATION

In contrast to the extensive research that has gone into mating type switching, there has been much less investigation of primary homothallic yeast species, and we still do not have a good understanding of the molecular basis of primary homothallism. Several species in the CUG-Ser1 clade (families Debaryomycetaceae and Metschnikowiaceae) (Fig. 5) have MAT loci with MATa and MATα genes in close proximity, with no DNA repeat sequences that could indicate a switching mechanism, so they appear to be primary homothallics (48). These species include Debaryomyces hansenii, which has a haplontic homothallic life cycle with a full sexual cycle (57, 77), and Scheffersomyces stipitis, which has been reported to show normal meiotic segregation of markers in genetic crosses (78, 79), as well as Millerozyma farinosa and Spathaspora passalidarum (48). Beyond this, we do not know much. Since it seems that in these primary homothallic species any cell can mate with any other cell, we must ask the question of how cells signal to each other (22). Genes coding for both types of pheromone (a-factor and α-factor) and both types of pheromone receptor (STE2 and STE3) are present in these species. Does each cell produce both a-factor and α-factor? If so, how does a cell avoid responding to its own pheromones?

As well as the species described in the previous paragraph, the 332-species survey uncovered many other Saccharomycotina species whose genome organization indicates that they are primary homothallics. Krassowski et al. (26) used the term PHC (primary homothallic, contiguous) for species such as D. hansenii and S. stipitis that have both MATa and MATα genes close to each other (within 20 kb) and that do not have any DNA repeats that could allow mating-type switching (Fig. 4). There were 25 species of this kind, from 11 different phylogenetic clades (Fig. 5). One of these PHC species is Lipomyces starkeyi, a member of the family Lipomycetaceae, which is an outgroup to most other Saccharomycotina species (Fig. 5). Recently, Takayama (80) showed that mating in wild-type L. starkeyi is homothallic, but by UV mutagenesis, she was able to obtain two heterothallic derivatives that were self-sterile but could mate with each other and sporulate—suggesting that the two mutants acted as complementary mating types. Further characterization of these mutants could provide a route to dissect how primary homothallism works in budding yeasts.

Krassowski et al. (26) also identified a second category of primary homothallic species, called PHN (primary homothallic, noncontiguous) (Fig. 4). The genomes of these species contain both MATa and MATα genes, but they are far away from each other, usually on different chromosomes. In these species, one of the two MAT loci shows conserved synteny with the MAT locus position in other closely related species (often beside the gene SLA2), and so is at the ancestral MAT locus site. In contrast, the second MAT loci in these PHN species were at species-specific sites in regions of the genome that lacked interspecies synteny (Fig. 4). Several of these second MAT loci were near telomeres, and one was beside the ribosomal DNA (rDNA) array. Based on these observations, Krassowski et al. (26) proposed that the second MAT loci of PHN species were recent introgressions into their genomes, so that a heterothallic ancestor (with a single MAT locus at the ancestral position) gained a second MAT locus of the opposite mating type and was converted into a primary homothallic species. This arrangement of the two MAT loci in PHN budding yeasts is very similar to the arrangement seen in the primary homothallic filamentous ascomycete Aspergillus (Neosartorya) fischeri (81). In Aspergillus fischeri, the MAT1 locus is found at an ancestral location between homologs of S. cerevisiae APN2 and SLA2, whereas the MAT2 locus containing the opposite idiomorphs is at an unlinked location and flanked by pseudogenes of APN2 and SLA2 (whether this location is telomeric is unclear in current genome assembly data), so MAT2 seems to have been gained recently by either a DNA duplication and transposition (81, 82) or a DNA introgression that converted a heterothallic ancestor to a homothallic descendant.

The telomeric and rDNA-linked locations of second MAT loci in PHN species suggest that their transcription could be regulated epigenetically (26, 83, 84). It seems plausible that in primary homothallic yeasts (both PHN and PHC species), each haploid cell stochastically expresses only one of its two types of MAT genes, but this idea has not yet been tested. It would avoid the problems of self-signaling and formation of the a1/α2 repressor in haploids.

NATURAL SELECTION FAVORS HOMOTHALLISM

Shen et al. (25) established the phylogenetic relationships among the 332 Saccharomycotina species, using state-of-the-art phylogenomic methods. Krassowski et al. (26) then used this tree to infer the transitions between different thallism states that have occurred during Saccharomycotina evolution. Remarkably, almost all the evolutionary transitions were away from heterothallism and toward homothallism. There were 31 transitions from heterothallism to homothallism and only 3 in the opposite direction (Fig. 4). This imbalance was unexpected, particularly because in theory it should be easy to convert a homothallic species to a heterothallic one by simply deleting one set of MAT genes, whereas to transition from heterothallism to homothallism requires more complex genomic rearrangements. The fact that 90% of the evolutionary transitions were from heterothallism to homothallism indicates that transitions in this direction must have been strongly favored by natural selection (26, 85).

Most of the evolutionary transitions to homothallism occurred near the tips of the phylogenetic tree and affected relatively small clades, indicating that they are recent (Fig. 5). There were more transitions to primary homothallism (19 in total: 11 to PHC and 8 to PHN) than to secondary homothallism (12 in total: 8 to FF1, 3 to FF2, and 1 to 3LOC) (Fig. 4). The relative magnitudes of these numbers seem consistent with the complexity of the genomic changes required to make the transition: a three-locus (3LOC) system requires the most complex changes, whereas primary homothallism requires the least.

The conclusion that homothallism emerged repeatedly concurs with the results of laboratory evolution experiments in Sch. pombe by Nieuwenhuis and colleagues (86). After imposing repeated cycles of strong selection for successful mating and spore formation (by killing vegetative cells), they found that homothallic (switcher) strains of Sch. pombe arose spontaneously from parental mixtures of two heterothallic (nonswitcher) strains of opposite mating types, at a high frequency and especially at low culture densities. The parental nonswitcher strains each contained only one silent cassette of the same idiomorph as their MAT locus. The evolved switcher strains acquired spontaneous genomic rearrangements in the MAT region, which created new 3LOC structures that contained silent cassettes of both idiomorphs and enabled switching (86).

EVOLUTION OF 3-LOCUS SYSTEMS FROM 2-LOCUS SYSTEMS

In the analysis of 332 Saccharomycotina genomes, a 3-locus system (MAT/HML/HMR) was seen to have originated only once, on the branch that was the common ancestor of family Saccharomycetaceae (Fig. 5) (26). The closest outgroup clade is family Saccharomycodaceae (containing the genus Hanseniaspora), which is heterothallic. Although this could be taken as evidence that a direct transition from heterothallism to 3-locus secondary homothallism occurred, in view of the complexity of the 3-locus system, we think that it is more likely that it arose via a series of intermediate steps—HET→FF1→FF2→3LOC—even though there is no direct evidence for the intermediates (26). In other words, we hypothesize that a flip/flop system with one IR initially arose in an ancestor of Saccharomycetaceae, and then it gained a second IR to become an FF2 system. The gain of a second set of repeats in FF2 species may be favored because it allows the distance between the two sets of MAT genes to increase, because recombination in these repeats can overcome the problem that the central region between the two sets of MAT genes is in opposite orientations on the two chromosomes in a MATa/α diploid, which could otherwise lead to unbalanced chromosome rearrangements if this central region underwent one crossover (or any odd number) during meiosis (26, 41). The arrangement of the two IR sequences on each side of the MAT genes in FF2 species is identical to the arrangement of the X and Z repeats around the MAT genes in 3LOC species (Fig. 4), so the major step required to convert an FF2 system into a 3LOC system is simply to duplicate one of the two existing sets of MAT genes to make a third locus. The third locus also needs to be silenced, but this could be achieved by retaining the silencing mechanism used for the silent locus in the FF2 system. For example, an FF2 species like K. phaffii could be converted to a 3LOC species by a genomic rearrangement that duplicated the silenced MAT genes and their adjacent telomere, attaching them to a different chromosome.

CLEAVING THE MAT LOCUS IN 3-LOCUS SYSTEMS

Mating-type switching in a 3-locus system such as that of S. cerevisiae involves degrading the DNA at the outgoing MAT locus and replacing it by newly synthesized DNA that is made by using HML or HMR as a template. In contrast, in species that switch mating types by a flip/flop mechanism (FF1 or FF2), the outgoing MAT locus is not degraded and replaced. Instead, recombination between the IRs moves the DNA of the outgoing MAT locus from a site where it is expressed to a site where it is repressed, and this DNA will be moved back again the next time the cell switches its mating type.

This difference means that 3-locus systems need a way to cleave the outgoing MAT locus so that it can be degraded, whereas 2-locus systems do not. Interestingly, even though the 3-locus system in Saccharomycetaceae has a single origin, different clades of species within this family use different mechanisms to cleave the MAT locus, as described below. Sch. pombe uses a different mechanism again, involving a single-strand DNA nick that makes an imprint at the MAT locus, leading to its breakage during DNA replication (87).

In S. cerevisiae, the double-strand DNA break that initiates degradation of the outgoing MAT locus is made by HO endonuclease. However, not all species in family Saccharomycetaceae contain an HO gene, and its origin can be pinpointed to a branch of the tree that is significantly younger than the family. HO is present only in the Saccharomycetaceae species that underwent whole-genome duplication (WGD) and in one non-WGD clade, the clade containing the genera Zygosaccharomyces and Torulaspora (Fig. 7). The WGD was probably an interspecies hybridization between one parent from the Zygosaccharomyces/Torulaspora (ZT) clade and one parent from the Kluyveromyces/Lachancea/Eremothecium (KLE) clade (88). If so, the ZT parent contributed an HO gene to the hybrid, but the KLE parent did not. Some models of the WGD postulate that it occurred by interspecies mating (forming a diploid hybrid), followed by deletion of one of the two MAT loci (forming a cell that behaved as a haploid), followed by mating-type switching and autodiploidization (forming a cell that has a MATa/α genotype and two identical copies of every gene from both parents, so it is fertile and genetically diploid) (88–90). For such a model to work, the single HO gene of the hybrid must have been able to cleave the single retained MAT locus: i.e., the MAT locus derived from the ZT parent must have been the one retained. The origin of HO is discussed in more detail in the next section.

FIG 7.

Schematic phylogeny of budding yeast genera, showing the inferred points of origin of the 3-locus MAT/HML/HMR cassette system (3LOC), HO endonuclease, WHO endonucleases, α3, and KAT1 genes. Ψ symbols indicate WHO pseudogenes, and numbers of plus symbols indicate abundance. Most of the species shown are in family Saccharomycetaceae; Hanseniaspora, Wickerhamomyces, and Candida albicans are outgroups from other families.

Kluyveromyces lactis is a member of the KLE clade, a group of genera within family Saccharomycetaceae that do not have HO endonuclease genes (Fig. 7). Like S. cerevisiae, K. lactis has a 3-locus system with MAT, HMLα, and HMRa loci. Åstrom and colleagues have characterized in detail how K. lactis makes the initial double-strand DNA break in its MAT locus during switching. Remarkably, it uses two different mechanisms, depending on whether the outgoing MAT locus is MATα or MATa. Both of these mechanisms involve making a double-strand break by a process that resembles the first step of mobilization of DNA transposons. During switching from MATα to MATa, the break is made by α3 protein, which is related to the DNA transposase of mutator-like elements (MULEs), a type of DNA transposon (58). The α3 gene is located in the Yα region in K. lactis (in addition to the usual MATα genes, MATα1 and MATα2), so it is always present at the HMLα locus and is also present at the MAT locus in MATα haploid cells. During switching in the MATα-to-MATa direction, α3 protein is expressed from HMLα and makes a break in the MATα locus (58). This step is similar to how transposases cut at the ends of MULE elements to initiate transposition. During switching in the opposite direction, from MATa to MATα, the double-strand break is made by Kat1 protein, which is related to the hAT family of DNA transposases (60). Kat1 cleaves at a site between the MATa1 and MATa2 genes, which is only present in the MATa idiomorph. After the initial cut is made, switching in both directions proceeds similarly to switching in S. cerevisiae, with homology in the X and Z regions flanking the MAT locus allowing recombination and repair using HML or HMR as a template for new DNA synthesis. Switching in both directions in K. lactis also requires the ortholog of the S. cerevisiae transcription factor Rme1 (called Mts1 in K. lactis, for “mating type switching 1”) (58). Rme1 plays no role in switching in S. cerevisiae, but it is required for switching in the flip/flop system of O. polymorpha (43, 44).

The α3 and KAT1 genes are specific to the genus Kluyveromyces (Fig. 7). They are transposase genes that originally came from mobile genetic elements but became “domesticated” during evolution. They are no longer mobile, and their DNA-cleaving ability has been harnessed by the host cell for a different purpose—cleaving its MAT locus to initiate mating-type switching. The two other genera in the KLE clade, Lachancea and Eremothecium (also called Ashbya), have not been investigated in depth, but most species in these genera have MAT, HML, and HMR loci, so they are presumed to use a 3-locus system (91). Switching has been detected in one species, Lachancea waltii (59). However, these species do not have α3, KAT1, or HO genes, so it is unknown how Lachancea and Eremothecium cleave the MAT locus.

ORIGIN OF HO ENDONUCLEASE FROM THE WHO FAMILY OF MOBILE GENETIC ELEMENTS

Recent research has shown that, like α3 and KAT1, the HO endonuclease gene also originated when yeasts domesticated a selfish genetic element. Coughlan et al. (92) discovered a large and diverse family of HO-like genes, which they called WHO genes (for “weird HO”), coding for endonucleases related to HO. The WHO endonuclease genes are contained on homing genetic elements called WHO elements, and HO is proposed to have originated by domestication of a WHO gene (92). WHO and HO endonucleases both also have sequence similarity to inteins, which are homing genetic elements that interrupt a host protein and excise from it by peptide splicing (93). The only intein present in Saccharomycetaceae species is the VDE intein, which interrupts the VMA1 gene coding for a subunit of vacuolar ATPase (94). However, the similarity between HO and WHO endonucleases is more extensive, and they both have a C-terminal zinc finger domain that is not present in VDE or other inteins.

WHO genes have a very limited phylogenetic distribution, being mostly found only in the yeast genera Torulaspora and Lachancea (Fig. 7), and they have characteristics of homing genetic elements. They occur in clusters downstream of the gene FBA1, which codes for the glycolytic enzyme aldolase. Focusing on Torulaspora delbrueckii, Coughlan et al. (92) found that different natural isolates of this species have WHO gene clusters that are highly variable in both structure and sequence (Fig. 8). The clusters frequently contain WHO pseudogenes as well as intact WHO genes, and they often contain duplicated fragments of the 3′ end of FBA1. The WHO genes in the T. delbrueckii clusters code for a very diverse family of HO-like proteins, with only 24 to 55% amino acid sequence identity among them and 22 to 25% identity to HO itself.

FIG 8.

Organization of WHO genes and pseudogenes and FBA1 fragments downstream of the full-length FBA1 gene in Torulaspora and related species. Six different alleles are shown for T. delbrueckii, three for T. pretoriensis, and two for T. globosa. Numbers indicate different families of WHO genes. Reproduced from reference 92 published under the terms of the Creative Commons Attribution License.

WHO genes code for endonucleases in the LAGLIDADG family (95). Most endonucleases in this family are encoded by homing genetic elements, which are a form of selfish DNA. The most familiar examples of homing genetic elements are inteins and mobile introns (93, 96). Homing elements code for an endonuclease that cuts the genome at a specific site in a host gene. Alleles of the host gene that do not contain the element are sensitive to cleavage by the endonuclease, whereas alleles containing an element are resistant to cleavage. The cleaved DNA is then repaired by using the uncleaved element-containing allele as a template.

Coughlan et al. (92) found that the host gene for WHO elements is the aldolase gene FBA1. They proposed that WHO element homing occurs during meiosis in T. delbrueckii, similar to the VDE intein (93, 94). If a strain whose FBA1 locus contains a WHO element mates with a strain with an “empty” FBA1 locus, expression of the WHO endonuclease gene in the resulting heterozygous diploid cell can lead to cleavage of the allele that lacks the element. The double-strand DNA break will then be repaired by copying the intact FBA1 allele, including the WHO element, making the cell homozygous for the element (Fig. 9A). Therefore, when the heterozygous diploid cell completes meiosis, all the spores it produces will contain the WHO element, and the element spreads selfishly through the population by this non-Mendelian mechanism, called homing.

FIG 9.

Similarity of (A) the proposed mechanism of WHO element homing into the FBA1 locus, (B) the known mechanism of VDE intein homing into the VMA1 locus, and (C) the known mechanism of mating-type switching in S. cerevisiae. R and S indicate alleles resistant or sensitive to cleavage by endonuclease, respectively. Green shading of the 3′ end of FBA1 indicates sequence variants resistant to cleavage by this WHO endonuclease. Modified from reference 92 published under the terms of the Creative Commons Attribution License.

WHO genes occur in clusters downstream of their host gene, FBA1 (Fig. 8), whereas in all other known homing elements (i.e., inteins and mobile introns), a single copy of the element interrupts the host gene and clusters do not form. This difference led Coughlan et al. (92) to propose that the mechanism by which an inserted WHO element makes its host gene FBA1 resistant to cleavage is different from the mechanism in inteins and mobile introns and involves replacing the 3′ end of FBA1 (Fig. 9A). Integration of a WHO mobile element replaces the 3′ end of the intact FBA1 gene, copying sequence variants from the donor locus to the recipient locus. These variants make FBA1 resistant to cleavage by this particular WHO element’s endonuclease. However, the same sequence variants that make the 3′ end of FBA1 resistant to one WHO endonuclease may also make it susceptible to a different WHO endonuclease, allowing clusters to form.

The mechanism by which WHO elements are proposed to home into “empty” alleles of FBA1 is very similar to the mechanism by which inteins home into their host genes (Fig. 9B) and is also very similar to the steps that occur during mating-type switching in S. cerevisiae (Fig. 9C). In each case, a LAGLIDADG endonuclease cleaves the target locus at a specific site, after which the double-strand break is repaired by using a template that has homology to the cleaved target on both sides but contains a nonhomologous region in the center (Fig. 9). While the mechanisms of WHO and intein homing are essentially identical, there are important differences between them and how HO operates. First, mating-type switching occurs only in haploids (49), whereas homing of inteins and WHO elements occurs (or is proposed to occur) only in diploids during meiosis (92, 94), so the endonucleases are regulated in opposite ways. Second, HO does not insert its own gene. When a WHO gene was domesticated to become HO, its cleavage target changed from a site in FBA1 to a site in the MAT locus (specifically, a site in the MATα1 gene [22]). This change of cleavage site led to a change of the donor locus that was used as a DNA repair template, because the sequences flanking the double-strand break were different. Instead of using a repair template containing the gene coding for the endonuclease itself (as occurs after cleavage by WHO endonucleases and VDE), cleavage by HO leads to the use of a repair template (HML or HMR) that inserts a gene unrelated to the endonuclease. The change of the endonuclease’s target from FBA1 to MATα1 during the domestication process that produced HO from WHO may have been caused by sequence changes in the zinc finger domain of the endonuclease (92).

WHY SWITCH?

The results of Krassowski et al. (26) indicate that natural selection strongly favors homothallism in budding yeasts, whether it is primary or secondary homothallism (Fig. 4). Heterothallic ancestors have given rise to homothallic descendants on about 31 separate occasions. Yeast clades that became homothallic have only rarely lost that ability again (Fig. 5), indicating the existence of natural selection to maintain homothallism. However, switching may incur a cost during asexual growth, and a minority of natural isolate strains are nonswitchers in both S. cerevisiae and Sch. pombe (86, 97). Furthermore, mating-type switching has been maintained even though it is error-prone and mutagenic. Synthesis of the new MAT genes during switching has a high point mutation rate (98). In 3LOC species, natural isolates frequently show evidence of illegitimate exchanges between MAT, HML, and HMR loci, causing interchromosomal translocations and sometimes resulting in strains that have only α-genes or only a-genes at all three loci (64, 99, 100). A comparative genomics study of Saccharomycetaceae (3LOC) species found that genes situated beside the MAT locus frequently become deleted or relocated to other places in the genome, which was attributed to errors that occurred during mating-type switching (63). Deletions were particularly prominent just after the WGD, which introduced a high level of redundancy into the genome and allowed large sections of DNA beside the MAT locus to be deleted—in one case, a region spanning 60 consecutive genes. So, the continued existence of mating-type switching, despite the damage it inflicts on the genome, implies that homothallism confers a strong evolutionary benefit to a yeast species.

What is the nature of this benefit? Since evolutionary transitions to primary and secondary homothallism are about equally frequent (Fig. 4), we can conclude that the benefit comes from homothallism of any kind, and not specifically from mating-type switching (secondary homothallism). Ira Herskowitz (101) suggested that by enabling haploid cells to leave diploid descendants, even in the absence of a mating partner, a species that can switch mating types gains the benefits that diploidy confers in terms of DNA repair—specifically, the ability to repair a damaged gene by using the homologous chromosome as a repair template—an idea we call the “diploid DNA repair” hypothesis. Alternatively, the “genome renewal” hypothesis proposed by Robert Mortimer for S. cerevisiae (102, 103) points out that mating-type switching enables a heterozygous diploid parental lineage to produce large numbers of homozygous diploid descendants, each with a different combination of alleles (some of which may be recessive) at different loci. Not only does this process purge the population of recessive deleterious alleles, it also enables the descendants to sample a large portion of the possible genotypic and phenotypic landscape, resulting in survival of the fittest and rapid evolutionary adaptation (104).

However, neither the genome renewal nor the diploid DNA repair hypothesis can explain the existence of mating-type switching in haplontic yeast species such as Kluyveromyces, Ogataea, or Schizosaccharomyces, in which diploids normally sporulate as soon as they are formed by mating. Haplontic species do not have the benefit of diploid DNA repair. Recessive mutations in their genomes are not masked. In these species, if a heterozygous zygote is formed by outcrossing, its descendants will sample only a small fraction of the possible genotypic and phenotypic landscape because there will only be four such descendants (in contrast to diplontic S. cerevisiae, where the diploid lineage can amplify mitotically and give rise to thousands of genetically varied descendants after sporulation and autodiploidization). Thus, although genome renewal may confer a real evolutionary advantage to diplontic species such as S. cerevisiae, it cannot occur in haplontic species and cannot account for the existence of mating-type switching in them.

Instead, we (22, 63) and others (86, 105) have argued that the evolutionary benefit of homothallism to budding yeasts comes from the dependence of spore formation on meiosis, which can occur only in diploids. Spores (ascospores) are life-support structures that enable cells to survive exposure to harsh environments (106). Meiosis and sporulation are essentially a single continuous pathway (107), and only diploid cells can form spores because the subcellular structure that eventually becomes the spore wall develops directly from the spindle pole body of meiotic cells. At the end of meiosis, proteins forming the outer plaque of each spindle pole body are replaced by proteins from which the prospore membranes begin to grow and envelop each nucleus (4). Even in Ashbya gossypii, an unusual Saccharomycetaceae species that can sporulate without mating, spore formation requires the ortholog of S. cerevisiae Spo74 (a meiotic outer plaque component), as well as regulators of meiosis such as Ime1 and Ndt80, and it has been suggested that sporulation in Ashbya involves a meiosis-like step (108–111).

This dependence of sporulation on meiosis means that when a spore germinates and turns into a haploid cell, the cell will be unable to form a spore again until after it has managed to mate. As a result, a spore’s “decision” about when it should germinate takes on a great evolutionary importance. Commitment to germination is an irreversible step (112). It occurs when a spore senses that its environmental conditions are improving (113, 114), and even isogenic strains of S. cerevisiae germinate asynchronously, suggestive of a bet-hedging strategy (115). If a spore germinates early relative to its peers, and if the environmental conditions are sufficient to support its growth and mitotic division, the spore’s descendants will have an immediate numerical advantage over the descendants of other spores that germinated later (Fig. 10A). This creates selective pressure in favor of early germination. But if the spore germinates too early and finds that the environment is insufficient to support life, or if the environment deteriorates after it germinates (Fig. 10B), the cell will die unless it can find a partner to mate with and resporulate. Under these conditions, there is a clear advantage to being homothallic and able to mate with one’s clonal relatives (Fig. 10C), as opposed to being heterothallic and depending on spores of the opposite mating type to germinate and be close enough to mate with. Even though mating often occurs between sister spores in the same ascus, it cannot occur until both of them have germinated, and if one spore fails to germinate another must be left with no mating partner (101, 116). This “resporulation” hypothesis is similar to the concept of reproductive assurance at low population density (86, 105) and the “lonely spore” hypothesis (101), but it proposes that the need to ensure sexual reproduction arises specifically from a need to ensure resporulation when germinating in a fluctuating environment. Although it remains unproven, and relatively little is known about how germination is controlled (112, 117), we believe that the resporulation hypothesis remains the most plausible explanation for why homothallism—both primary and secondary—has emerged repeatedly during budding yeast evolution.

FIG 10.

The “resporulation” hypothesis postulates that homothallism enables spores to germinate early without risking extinction. The black curves represent two scenarios: one in which the environment improves continuously over time (A) and one in which the environment fluctuates over time (B and C). The cartoons show two spores (red and blue) germinating and expanding clonally as environmental conditions improve. Circles represent spores, and ovals represent vegetative cells. Spore 1 (red) has a lower threshold for the quality of environment required for germination than spore 2 (blue). (A) If the environment improves smoothly, spore 1 leaves more descendants than spore 2 because it germinates earlier. (B) If the environment fluctuates and returns to conditions in which vegetative cells cannot survive, spore 1 germinates too early. If it is heterothallic and cannot find a mating partner, it will leave no descendants and go extinct. (C) In the same fluctuating environment, if spore 1’s haploid descendants are homothallic and able to mate with each other, the lineage can resporulate and survive the uninhabitable period. When the environment improves again, the spore 1 lineage will germinate earlier and outnumber the spore 2 lineage.