Chromosome evolution within the clade

As sequencing was still expensive and time consuming at the time of the first genome being completed, comparative analysis was done via physical- and limited-sequence comparisons. This more limited approach was still valuable in terms of determining phylogenetic relationships and evolutionary processes. Individual representatives of most of the species known at the time were sequenced [S. paradoxus, S. mikatae, S. kudriavzevii, and S. uvarum (Cliften et al. 2003; Kellis et al. 2003)], which allowed a more detailed comparison of genomes. Since then there has been a more complete, high quality assembly of these genomes (Scannell et al. 2011). The Saccharomyces clade shares a great deal of gene content and synteny, having settled down to a similar chromosome karyotype after more rapid chromosomal rearrangements and deletions following the whole-genome duplication (WGD) (Scannell et al. 2006). Other post-WGD species, including Candida glabrata, Naumovozyma castellii, and Vanderwaltozyma polyspora, have very different karyotypes having diverged from the Saccharomyces clade soon after the WGD (Scannell et al. 2006, 2007a).

Early electrophoretic karyotype analysis of some of the species (Naumov et al. 1992) was followed by a more thorough physical analysis of the six known species in 2000 (Fischer et al. 2000). Using probes designed for each chromosome arm and each centromere region, their gross chromosomal structure and breakpoints of large rearrangements between species was described. This analysis was able to map the breakpoints of several translocations relative to S. cerevisiae in the clade. Most of these were coincident with Ty elements or Ty LTRs. Two of the species were collinear with S. cerevisiae and these were not correlated with the phylogenetic relationships. This indicated that the S. cerevisiae chromosome configuration was likely the ancestral configuration. These rearrangements were therefore not the driver of reproductive isolation leading to speciation across the whole clade. The translocations that were mapped were nonrandomly distributed with respect to phylogenetic branch lengths and indeed appeared to have occurred in bursts, perhaps due to Ty activity. In particular, the S. cariocanus karyotype had four reciprocal translocations relative to the very closely related S. paradoxus, which is collinear with S. cerevisiae, the next closest relative. The idea of bursts of Ty activity has been supported by more recent population genomic sequencing (Liti et al. 2009a) where the Ty content of the two S. cariocanus isolates, with four reciprocal translocations mapped, have the highest number of Ty elements, second only to S288C; whereas most S. paradoxus isolates, to which S. cariocanus phylogenetically belongs, have lower numbers of Ty elements than the majority of S. cerevisiae strains. Other examples of large numbers of gross chromosomal rearrangements (GCRs), such as the Malaysian S. cerevisiae lineage (Liti et al. 2009a; Cubillos et al. 2011), are also likely to be due to bursts of Ty activity.

Gene losses and differentiation of function

The individual genome sequences from one strain each of some of the species supported these earlier physical studies but also led to some interesting findings. For example, the entire galactose-utilization (GAL) pathway in S. kudriavzevii (Hittinger et al. 2004) was inactivated due to various mutations in the genes, which could have been due to the independent accumulation in each gene, leading to pseudogenes. This was found in an isolate of the species from the Far East. The more recent finding of a European population supports the nature of this loss as this population has a functioning GAL pathway (Hittinger et al. 2010). Finding this population also solved one of the mysteries of where yeasts originated, given that many European wine strains are S. cerevisiae × S. kudriavzevii hybrids but no European S. kudriavzevii population was known before. Another example where mutation accumulation between species occurs came from the use of genome annotations as a tool for comparative analysis. When the S. uvarum gene annotations were compared to those of S. cerevisiae, as many as 35 apparent breakpoints were discovered (Fischer et al. 2001), yet the physical analysis of the chromosomes described only 8 due to four reciprocal translocations (Fischer et al. 2000). This was rectified by direct sequence comparisons where relics of genes were found between open reading frames (ORFs) in one species compared to the other, resulting in loss of nearest neighbor synteny contacts when only annotated ORFs are used (Lafontaine et al. 2004). There can clearly be rapid accumulation in genes and sets of genes in one lineage compared to another.

As these species are derived from a WGD ancestor, there are many examples of duplicated genes, and these can lead to differentiation of function when one or the other is not lost. Two examples are the GAL1-GAL3 (Hittinger and Carroll 2007) and the SIR2-HST1 (Rusche and Rine 2001) pairs of duplicates. Other examples have been found in comparison of the reference genomes for each species in the clade (Scannell et al. 2011). Most of the differentiation occurred prior to the expansion of the clade but there is still further differentiation among the species (Kuang et al. 2016).

Population genomics

The sequencing of representatives of some of these species was followed after a while by the sequencing of multiple individuals within species, and this has recently been followed by whole-genome sequencing of hundreds to thousands of individuals, mostly of S. cerevisiae. In the days prior to large-scale, second-generation sequencing, sample sequencing of several genes was used to look at diversity and phylogenetic relationships. A few individuals were completely sequenced for specific purposes. EC1118, a wine strain (Novo et al. 2009); YJM789, a clinical isolate (Wei et al. 2007); W303, another laboratory strain (Ralser et al. 2012); Sigma1278b, another laboratory strain (Dowell et al. 2010); and CEN.PK, an industrial strain (Nijkamp et al. 2012); were sequenced. This was in addition to several studies of partial sequences in S. cerevisiae (Fay and Benavides 2005) and S. paradoxus (Tsai et al. 2008), as well as across the clade (Liti et al. 2006). Whole-genome diversity was also assessed using microarrays based on the S288C sequence (Gresham et al. 2006; Schacherer et al. 2007a, 2009).

The early population genomics survey of S. cerevisiae led to the conclusion that there have been two domestication events, one in wine strains and one in sake strains (Fay and Benavides 2005; Schacherer et al. 2009). A comparison of S. cerevisiae variation to that seen in other species revealed much less genetic variation among S. cerevisiae isolates from diverse sources, which was a paradox given the phenotypic diversity that was observed. Populations of S. paradoxus were much more diverse from each other and correlated with geographic boundaries, with no such correlation seen in S. cerevisiae (Liti et al. 2006). The two S. cariocanus isolates were very close in sequence to the North American S. paradoxus isolates, bringing into question their designation as a species as there must have been very recent gene flow between them (Liti et al. 2006). Within S. paradoxus there was determination of the population genetics parameters of the species with a measure of how much sex and outbreeding occurred within a population (Tsai et al. 2008). It is clear that most evolution in the clade is by clonal expansion from vegetative reproduction with a minor contribution of sexual reproduction, however there is some sexual reproduction and recombination and they can have a great deal of influence.

The first large-scale sampling of genomes from S. cerevisiae as well as S. paradoxus was done using Sanger sequencing and resulted in the start of increased efforts into population genomics surveys (Liti et al. 2009a). At the same time, a microarray survey was done on a larger number of S. cerevisiae strains (Schacherer et al. 2009). The two studies were consistent with each other and there was some overlap in the strains analyzed. The main findings included: there are several distinct populations of S. cerevisiae that are equidistant from each other across their genomes, so called clean lineages; there are several mosaic strains that appear to be the result of recent outbreeding between these clean lineages; and that genome wide there is nothing special about the wine or sake populations, even if they have specific variants associated with adaptation to the fermentation process in which they are usually found (Perez-Ortin et al. 2002). One hypothesis is that the fermentation properties desired already existed in extant populations of yeast and that our activity sampled and selected the strains with the best properties. The mosaic strains clearly had some sequence variation that came from outside the five clean populations described, indicating that there were likely more populations to be discovered. The survey was consistent with the previous sample sequencing and found that the S. paradoxus populations were clean with little or no gene flow between them. The originally characterized S. cariocanus was clearly embedded within the North American population based on shared sequence variation, despite the reproductive isolation.

Clean lineages vs. mosaics

The concept of clean lineages merits some additional discussion, especially in S. cerevisiae. Clean lineages were defined as those populations where the phylogenetic relationship with other populations was the same across the genomes, where the topology of the phylogenetic tree remains the same for any segment of the genome being compared (Liti et al. 2009a). This would be expected for populations that are evolving independently with little or no gene flow between them, as is seen for the S. paradoxus populations in which the genetic diversity is completely correlated with geographic location. The survey of 36 S. cerevisiae genomes revealed that about half fell into five clean lineages by this definition, while the other half were comprised of segments from two or more of the clean lineages with a few segments from unknown populations. As many of these mosaic strains were associated with human activity—such as bread making, food spoilage, fermentation, and even clinical cases—the origins of these recent outbred strains could be due to human activity providing the opportunity for interbreeding. The contrast with S. paradoxus could be a partial explanation for the higher levels of phenotypic diversity seen in S. cerevisiae despite lower overall genetic diversity. The genetic diversity that exists in S. cerevisiae is more mixed up due to outbreeding, which may result in an expansion of the phenotypic space.

Such a global genome analysis has not yet been done on other populations, so it is not known whether the wild populations from China (Wang et al. 2012) or those associated with other human activities (Cromie et al. 2013; Arana-Sanchez et al. 2015; Ludlow et al. 2016) are clean lineages or mosaics, though admixture was detected using genome-wide analysis of variants using restriction site associated DNA sequencing (Cromie et al. 2013; Ludlow et al. 2016). It would be very interesting if some populations associated with human activity are mosaic with genetic combinations that make them useful for particular activities, such as bread making. It is clear that some fermentation-associated populations are clean lineages, such as the sake and the wine populations. A large number of S. cerevisiae strains are now sequenced or about to be sequenced (Strope et al. 2015b; http://1002genomes.u-strasbg.fr/index.html; http://www.ncyc.co.uk/yeast-treasure-trove-goes-live/) and this collection will provide an unparalleled view of diversity within the species, addressing issues of selection and movement by human activity as well as origins of particular adaptations.

Genetic vs. phenotypic diversity

Another result of the survey was confirmation that there was greater genetic diversity in S. paradoxus than in S. cerevisiae. This result was surprising as the strains were all tested for growth under a large number of conditions and it was found that the phenotypic diversity of S. cerevisiae was much greater than that of S. paradoxus. Despite this paradox, there was a significant correlation between genetic diversity and phenotypic diversity in the two species. One hypothesis explaining this paradox is the level of outbreeding seen in S. cerevisiae. Perhaps the outcrossing, likely facilitated by human activity, has led to the increased phenotypic diversity by bringing together novel combinations of variants at different loci.

More diversity has been found in S. cerevisiae with the discovery of at least 10 new populations from China, many from primeval forests (Wang et al. 2012). Whole-genome analysis has yet to be completed but it is clear there is greater diversity in these isolates as seen in Figure 2. Some of the populations from China are close to those previously described (Liti et al. 2009a), but others, particularly the ones from primeval forests (populations CHN I, II, and III in Figure 2) nearly doubles the known diversity. In addition, distinct populations of S. cerevisiae are being found, which are associated with other fermentation activities such as cacao (Cromie et al. 2013; Arana-Sanchez et al. 2015; Ludlow et al. 2016) and coffee (Ludlow et al. 2016) processing.

Open in a separate window

Figure 2

The populations of S. cerevisiae as an unrooted cladogram (not to scale) composite from several studies (Liti et al. 2009a; Wang et al. 2012; Almeida et al. 2015). Populations shown include North American (NA), sake (SA), West African (WA), Malaysian (MA), and wine/European (W) as previously defined (Liti et al. 2009a), with E as the wild European sister clade to the wine group, recently described (Almeida et al. 2015). Eight additional populations from China (CHN I–CHN VIII) were recently described (Wang et al. 2012). In addition a New Zealand (NZ) subpopulation of the wine population has been described (Gayevskiy and Goddard 2012). The original distinction of two domestication events leading to wine and sake fermentation (Fay and Benavides, 2005; Cromie et al. 2013; Arana-Sanchez et al. 2015; Ludlow et al. 2016) has been extended as more populations have been sampled. These include wild populations from oak and related trees, populations associated with fruit trees, and populations associated with fermentation activities. It is clear that many of the populations associated with fermentation are very close to wild populations exemplified by the wine vs. European sister groups, which differ (Almeida et al. 2015) in the presence/absence of the genes acquired by HGT from outside the clade (Novo et al. 2009). The wine population has been spread around the world from its European origins and in some cases, such as New Zealand (Gayevskiy and Goddard 2012), the subpopulation can be placed within the larger population. There are now descriptions of populations associated with ale fermentation not distinguished here (Gallone et al. 2016; Goncalves et al. 2016). Lab, laboratory.

There are now thousands of S. cerevisiae isolates from a variety of sources. Many cluster with their source even if not with location. Figure 2 is a composite of S. cerevisiae populations sampled over the years and analyzed by different means. The most sampled population is the wine/European, which has subpopulations found around the world wherever wine is produced. One subpopulation analyzed extensively is that from New Zealand which must have arrived with human migration (Goddard et al. 2010; Gayevskiy and Goddard 2012). A sister population was recently found; it is associated with oaks in Europe but not used in fermentation (Almeida et al. 2015). This strain differs by the lack of alien DNA from other species that confers useful properties for wine making (Novo et al. 2009). There are numerous wild populations from around the world as well as distinct populations associated with various human fermentation activities including beer, chocolate, coffee, etc. (Legras et al. 2007; Sicard and Legras 2011; Tapsoba et al. 2015; Ludlow et al. 2016). There are also intermediate populations associated with fruit trees and orchards. As can be seen in Figure 2, almost every population associated with fermentation or fruit trees has a close wild sister population.

As more populations of different species are being found, whole-genome surveys are being expanded, and now there is good data on populations of S. uvarum (Almeida et al. 2014), S. eubayanus (Almeida et al. 2014; Peris et al. 2014, 2016a), and S. kudriavzevii (Hittinger et al. 2010; Peris et al. 2016b), in addition to S. cerevisiae and S. paradoxus. These will make for interesting comparisons with S. cerevisiae, particularly for those species with little human influence.

Domestication or harnessing existing potential

Given the large number of surveys of S. cerevisiae from a wide variety of locations and environments, we can address the issue of domestication more thoroughly. The wine/European population is the most widespread due to human wine production, and the introduction and spread of these yeasts can be observed in some cases (Gayevskiy and Goddard 2012, 2016). The question of domestication, at least with the wine/European population, can be informed by the very recent discovery of a wild population from oak trees that is sister to the wine/European population (Almeida et al. 2015). This finding supports the hypothesis that rather than domestication, our use in fermentation took advantage of the properties of an existing population of yeast with perhaps some adaptations to the wine-making process (see below for discussion of introgressions). However, there are phenotypes and specific genetic variants associated with use in wine fermentation (Gallone et al. 2016). In ale and lager strains, there is stronger evidence of selection on specific genes (Gallone et al. 2016; Goncalves et al. 2016) as well as on the whole genome (Baker et al. 2015) from sequence comparisons. Of the other species used in human fermentation activities, there are probably not enough populations sampled to address the domestication issue. Many strains used in fermentation are hybrids of other Saccharomyces species with S. cerevisiae (Figure 1C), and it may be that the existence of hybrids preceded their use in fermentation.

Subtelomeres and the pan-genome

One thing that became apparent from whole-genome sequence surveys is that you cannot study what you do not know exists. With microarrays and other hybridization techniques, you can only assess variation in known sequences. Several genes and gene families were known from yeast studies that are not found in the S288C genome. Six of these gene families involved in carbon-source utilization are located in the subtelomeres (Liti et al. 2009a). Subtelomere variation and dynamics was already recognized (Louis and Haber 1990, 1992; Louis et al. 1994; Louis 1995) and it was known that the region was a hotbed of gene-family evolution (Brown et al. 2010). The analysis of the Saccharomyces Genome Resequencing Project collection revealed 38 new gene families not previously known to be in S. cerevisiae, in addition to the six known families (Liti et al. 2009a). These varied in presence/absence as well as in copy number across the S. cerevisiae and S. paradoxus isolates. It is clear that as a species, S. cerevisiae is much more than merely represented by S288C.

This pan-genome has become important in the study of genetic variation underlying phenotypic variation among strains. A recent set of tools has been developed to help assess genetic material not found in the reference genome (Song et al. 2015). Quantitative genetic studies have found that much of the causal genetic variation for a given trait maps beyond the ends of assembled contigs into the subtelomeres. In any given strain, there may be ~8% of the genome that is considered subtelomeric. In most quantitative traits analyzed, up to 25% of the causal segregating variation maps to the subtelomeres (Cubillos et al. 2011; Liti and Louis 2012). One explanation of the higher phenotypic diversity on S. cerevisiae compared to S. paradoxus is subtelomeric variation. A total of 50% of the phenotypic variation difference can be explained by copy number and presence/absence variation in subtelomeric genes, which is greater in S. cerevisiae than in S. paradoxus (Bergstrom et al. 2014). This leaves 50% to be explained perhaps by outbreeding, supported by the expansion of the phenotypic variation in every cross between clean lineages as described above.

Horizontal gene transfers

The sequence of a wine strain, EC1118, revealed several genome segments whose origin was from outside S. cerevisiae, but within the Saccharomyces, and some outside the clade (Novo et al. 2009; Marsit et al. 2015). These segments conferred useful properties to the wine niche and therefore may be considered adaptive. One of these appears to have originated from a wine spoilage yeast and so it is possible that somehow DNA from this was incorporated into a wine strain either by some level of cellular interaction or by uptake from the environment. It is certainly possible that the two were occupying the same vessel at the same time during fermentation. The structure of the insertions in various strains indicated a circular intermediate that integrated independently in different configurations (Borneman et al. 2011; Galeote et al. 2011). All wine strains have some permutation of this configuration and the other horizontal gene transfers (HGTs), and so it has spread through the yeast used by the wine community. Another example is in Ty2 elements, which are not thought to be infective though they do generate virus-like particles within the yeast cell. S. cerevisiae and S. mikatae have many copies, yet S. paradoxus in between has none. Either S. paradoxus has lost Ty2 or never had it. The comparison consensus sequences of Ty2 in S. cerevisiae and S. mikatae indicate a time of divergence much more recently than the rest of the genome, and more recently than S. paradoxus and S. cerevisiae diverged (Liti et al. 2006). The most consistent explanation is that Ty2s moved from one to the other in the recent past, postspeciation.

Hybrids, introgressions, and reticulate evolution

One of the surprising findings of population genomics surveys is the introgression of sequence (homologous replacement) seen between species. Such introgression requires the ability to hybridize, which all members of the clade have, and some breakdown of the reproductive isolation allowing gene flow from one species into the other. The first examples were between S. paradoxus and S. cerevisiae. When one genome sequence of S. paradoxus was available it became clear that there was a segment with little (0.1%) divergence compared to the S288C sequence, whereas the rest of the genome was 10–15% divergent (Liti et al. 2006). Upon analysis of the region in several isolates, it was shown that the entire European population of S. paradoxus had an introgression from S. cerevisiae not seen in the other populations. Interestingly, this was in a subtelomere. Following this observation, introgressions in both directions have been found in surveys of S. cerevisiae and S. paradoxus (Muller and McCusker 2009). As more populations of different species are surveyed, more examples of such introgressions are found (Almeida et al. 2014). For example, there are numerous introgressions of S. uvarum into S. eubayanus and vice versa (Peris et al. 2014). There are introgressions from S. kudriavzevii into these two species as well. Given that these species are used in human fermentation activities, the potential opportunity for genetic exchange exists.

There are numerous examples of hybrids between yeast species (see What Did We Learn from Comparative Genomics of Other Saccharomycotina?). They are generally sterile and therefore dead ends in terms of long-term evolution, as no sexual recombination is possible anymore. The question then arises as to how segments from one species can move into the genome of another, clearly via homologous recombination, when the hybrids are sterile? The usual thought of how this occurs is through a rare viable spore backcrossing with one of the parents, with subsequent backcrosses becoming easier and easier. The alternative is a single step homologous introgression in a hybrid that does not have to undergo numerous backcrosses, as has been observed in artificial S. cerevisiae × S. uvarum hybrids (Dunn et al. 2013).

The evolution of the Saccharomyces clade appears to be reticulate rather than bifurcating as is usually portrayed in phylogenetic trees. Within populations there is clearly mostly vegetative reproduction. However, there is some outbreeding and sexual recombination so that within a population there is breakdown of linkage disequilibrium and some level of outcrossing (Tsai et al. 2008). Between populations there is less gene flow, however there is admixture in some cases and certainly the movement of yeast through human activity has provided the opportunity for gene flow between populations. This is reticulate evolution and expected for sexually reproducing species. We generally think that once speciation has occurred then gene flow stops and we move from reticulate to bifurcating phylogenies, but the population genomics surveys of Saccharomyces species indicates otherwise. Perhaps gene flow is possible, even if rare, across the whole of the clade, leading to connections between even distant branches. This requires the exploration of reproductive isolation and its role in speciation.

Reproductive isolation and speciation

What is a species and how will we move into the future with species concepts? We have seen above that under one definition, the biological species definition, we define S. cariocanus as a new species distinct from the others in the clade. However under another definition, the phylogenetic species definition, we place these isolates into the American population of S. paradoxus. There are numerous examples of individuals and populations having GCRs relative to closely related other strains, which would make them a species under one definition but not the other. An example is the Malaysian population of S. cerevisiae. It has not been labeled as a new species but as S. cerevisiae based on its sequence and physiological characteristics, yet it has at least 10 breakpoints from GCRs that preclude fertility when crossed to any other S. cerevisiae (Liti et al. 2009a; Cubillos et al. 2011). If S. cariocanus is a new species then this population should also be one. If it remains S. cerevisiae, then S. cariocanus should be relabeled S. paradoxus.

Speciation is generally thought to arise from reproductively isolated populations that continuously diverge from each other. There are three major hypotheses for the underlying mechanism of speciation, two of which are found in other systems and are well established. As discussed above there are numerous examples of introgressions between species, indicating that reproductive isolation is not complete which may bring us to question the concept of species in these yeasts (Louis 2011).

Sequence divergence and mismatch repair

A third mechanism that applies across the entire clade is simple sequence divergence and the action of the mismatch-repair system (Liti et al. 2006; Louis 2011). Here the intermediates of homologous recombination; required for crossing over, chiasmata formation, and proper chromosome segregation; contain numerous sequence differences between diverged parents. These are recognized by the mismatch-repair system and rather than being “repaired” to one or the other parental sequence, as would happen for a single mismatch, the large number of differences lead to abortion of the intermediate. The end result is no recombination and aneuploidy due to random segregation of the chromosomes (Chambers et al. 1996; Hunter et al. 1996). The test of this hypothesis is the restoration of some spore viability associated with increased crossing over when the mismatch repair system is deleted (Chambers et al. 1996; Hunter et al. 1996). This notion holds for the closely related species S. paradoxus and S. cerevisiae (Chambers et al. 1996; Hunter et al. 1996) but also for diverged populations within each of the species (Greig et al. 2003), indicating that it is acting at the right stage for incipient speciation.

In Figure 3 there is a clear linear relationship between sequence divergence and fertility/spore viability. The off-line examples in specific crosses are accounted for by one or more reciprocal translocations and, when corrected for these, the expected fertility fits the linear relationship. In a recent survey of many isolates (Hou et al. 2014), where 60 were crossed to the S288C background to measure spore viability, the conclusion was that sequence divergence was not correlated with the level of spore viability and that GCRs were the major mechanism underlying the observations. There are two problems with this conclusion. One is that they surveyed a narrow range of divergence (0.1–0.5%) where not much change in viability is expected and the divergence in each measure of viability would be greater than any signal of reduced viability with increased divergence. The second and more important issue is that the S288C strain is a mosaic sharing large segments of the genome with low sequence divergence with most other isolates rather than evenly distributed divergence. These low diversity regions will allow normal recombination and chromosome segregation, alleviating any problems elsewhere in the genome.

Open in a separate window

Figure 3

Fertility (spore viability) vs. sequence divergence. The biological species definition is based on within-species fertility but between-species sterility after mating. The potential causes of postzygotic reproductive isolation include GCRs, B-D-M incompatibilities, and sequence divergence itself acted upon by the mismatch-repair system during meiotic recombination. Most of the interspecies hybrids exhibit levels of spore viability that are in the range of random segregation of chromosomes leading to an accidental viable combination of <1%. In populations that have diverged yet are not separate species [three populations of S. paradoxus (Sp a, b, and c) and three populations of S. kudriavzevii (Sk a, b, and c)] as well as in the hybrid of the very closely related species S. uvarum and S. eubayanus, there is a linear relationship between sequence divergence and spore viability (Liti et al. 2006; Hittinger 2013). Some close strains by sequence do exhibit spore viability less than expected by the linear relationship; however, when corrected for known translocations, the viability returns to the expected relationship. This is particularly relevant for the S. cariocanus by North American S. paradoxus (1) where the sequence divergence is small and there are four previously described reciprocal translocations (Fischer et al. 2000) and other large GCRs discovered by complete assembly of the genome (D. Delneri, personal communication). Other examples where the correction of GCRs reestablishes the sequence divergence-spore viability relationship include strains of S. paradoxus from the Far East population (2) where one has a translocation, and strains from the European population and Far East populations differing by one translocation (3).

Four major genomic architectures within the Saccharomycotina

From presently available genome sequences, four major subgroups can be recognized within the Saccharomycotina subphylum of yeasts: (i) the Saccharomycetaceae, by far the most extensively studied family; (ii) the “CTG clade,” a diversified subgroup made of yeast species using an alternative genetic code; (iii) the “methylotroph clade,” exemplified by a few recently sequenced yeasts bearing novel signatures; and (iv) several species belonging to distinct and probably distant lineages altogether regarded as “basal” to the Saccharomycotina subphylum but actually very heterogeneous. The four subgroups differ from one another by the presence/absence of specific genomic signatures (Figure 6) and are clearly separable from one another by overall proteome comparisons. Additional species may eventually illustrate novel genomic architectures or interesting evolutionary intermediates between the four subgroups, but data are presently too dispersed for definitive conclusions. The most interesting ones so far are represented by members of the Ascoidaceae and Lipomycetaceae families (see Table 3), which represent early branching lineages at the basis of the Saccharomycetaceae and all Saccharomycotina, respectively (Shen et al. 2016).

Open in a separate window

Figure 6

The four major subgroups of Saccharomycotina as defined from genome architectures. The major species whose genomes served to define the subgroups are indicated (see Table 1, Table 2, and Table 3 for details). Tree topology based on Kurtzman et al. (2011). For more recent global phylogenies, refer to Shen et al. (2016). Most significant properties of each subgroup are summarized in boxes on the right (see text). Red dotted lines symbolize hybridization between members of the ZT and KLE clades of Saccharomycetaceae prior to WGD (Marcet-Houben and Gabaldón 2015; Wolfe 2015).

Genomes of the Saccharomycetaceae family (Table 1) harbor several signatures that distinguish them from all other subgroups: the presence of point centromeres, the triplication of mating loci, and the absence of genes for complex I subunits of the respiratory chain in their mitochondrial DNA (mtDNA). In addition, they generally contain a single rDNA locus either internal to a chromosome arm or, less frequently, located in subtelomeric regions as a result of independent translocation events during the evolution of the family (Proux-Wera et al. 2013). The rDNA locus is made of large tandem arrays of the 35S precursor transcript of ribosomal RNAs (rRNAs), like in all eukaryotes, but also includes the 5S RNA gene, most often in opposite orientation and sometimes duplicated (Bergeron and Drouin 2008). Remarkably, almost all yeasts of this family use the bacterial mode of tRNA decoding instead of the eukaryotic mode for the arginine CGN and the leucine CUN codon families (Marck et al. 2006; Grosjean et al. 2010; see below). Genomes vary in size between ~9 and 14 Mb (haploid equivalent, ignoring the rDNA locus, mitochondria, and plasmids) but are always highly compact (~65–75% coding) and contain ~4500–5900 protein-coding genes with very few spliceosomal introns [3–5% of coding DNA sequences (CDS) are split]. This family is the most extensively studied, all known genera having now at least one representative fully sequenced with the sole exception of Zygotorulaspora and Cyniclomyces (the latter is basal to the family). The 2-μm family of self-replicating plasmids, initially discovered in some isolates of S. cerevisiae (Guerineau et al. 1971), seems to be circumscribed to this family where it was found in several genera (Blaisonneau et al. 1997; Strope et al. 2015a). However, plasmids in general have unfortunately not attracted sufficient attention in recent genomic works to have a precise image of their evolutionary distribution.

By contrast, sequenced members of the CTG clade (Table 2) have regional centromeres (~3–4 kb long), often made of remnants of mobile elements, generally poorer in GC than average genomes and lacking conserved sequence motif. They have typically only one mating-type locus. Their moderately larger genomes (~10–16 Mb, same definition as above) are also highly compact (~55–70% coding) with ~5600–6400 protein-coding genes and few spliceosomal introns (6–7% of CDS are split). Their major distinctive signature, relative to other subgroups, is the usage of the alternative genetic code in which the CUG codon is used to specify serine instead of leucine (or in addition to it), due to an additional tRNA species (Santos et al. 2011). Their mitochondrial genomes contain the seven genes encoding subunits of complex I of the respiratory chain. The different lineages in this large and heterogeneous clade have been unequally studied so far, and numerous genera have not yet been explored at the genomic level.

A smaller number of species have been sequenced so far in the methylotroph subgroup (Table 3). They share common signatures with the CTG clade (regional centromeres, high coding density, and the complete set of genes in mtDNA) except for their usage of the universal genetic code and moderately smaller genomes (~9–13 Mb). They also exhibit a slightly higher number of spliceosomal introns (10–15% of CDS are split). Some species of this subgroup (K. phaffii, Ogataea polymorpha, and Ogataea parapolymorpha) have duplicated mating loci, and haploid cells undergo mating-type switching (Hanson et al. 2014; Maekawa and Kaneko 2014; see below). Methylotroph yeasts appear as evolutionary intermediates between Saccharomycetaceae and basal lineages by their eukaryotic mode of tRNA decoding for the leucine CUN codon family, and bacterial mode for the arginine CGN codon family (Morales et al. 2013). Although associated to this subgroup (Kurtzman et al. 2011), K. phaffii shares common genomic signatures with basal lineages such as the eukaryotic mode of tRNA decoding for both codon families and the dispersion of 5S RNA gene copies in the genome independently of the rDNA locus.

Basal lineages of the Saccharomycotina subphylum (Table 3) have long been represented solely by Y. lipolytica, whose genome exhibits a number of important differences with the three previous subgroups, such as its significantly larger size (20.5 Mb), lower compactness (46% coding), multiple rDNA loci located in subtelomeric regions, and dispersed 5S RNA gene (116 copies of this gene are distributed across the genome) like in most Pezizomycotina and Taphrinomycotina (Bergeron and Drouin 2008). With nearly 6600 protein-coding genes, 14% of which are split, it also possesses the largest protein repertoire. Additional species of Yarrowia that are now sequenced (see Table 3) confirm these trends. Historically, the second genome sequenced within the basal lineages was Blastobotrys adeninivorans (Kunze et al. 2014). It shares 49% of conserved microsynteny with Y. lipolytica, indicating a distant but still recognizable common ancestry. But its small size (12 Mb), high compactness (74% coding), and the presence of a single rDNA locus internal to a chromosome arm are more reminiscent of the evolved subgroups of Saccharomycotina. It has, however, a high number (~6150) of protein-coding genes, 11% of which are split by introns. The genome of Geotrichum candidum (Morel et al. 2015) provided another example of the diversity of the basal lineages and the long evolutionary distances separating them. With a size (24.2 Mb), compactness (45% coding), and gene number (~6800) comparable to Yarrowia, it is remarkably intron rich (35% of CDS are split) for a Saccharomycotina. Similar size (16 Mb), compactness (50% coding), and gene number (6820) were also found for Sugiyamaella lignohabitans (Bellasio et al. 2016), but only 5% of its protein-coding genes are split by introns. The recently published genomes of Nadsonia fulvescens and Tortispora caseinolytica (Riley et al. 2016) are comparable to G. candidum for their high numbers of introns, but are much smaller in size (13.7 and 9.2 Mb) and gene number. The genome of Lipomyces starkeyi (Riley et al. 2016), representative of the most basal lineage to all Saccharomycotina, is extremely intron rich with an average of three introns per protein-coding gene. The other sequences of the basal lineages reported in Table 3 (mostly unpublished) have not yet been analyzed in detail. S. lignohabitans, N. fulvescens, T. caseinolytica, and L. starkeyi have dispersed 5S RNA genes, like Y. lipolytica.

Origin of the distinctive features of Saccharomycotina genomes

Centromeres

The presence of point centromeres, originally discovered in S. cerevisiae from their ability to properly partition circular plasmids during cellular divisions (Clark and Carbon 1980), is a specific landmark of the Saccharomycetaceae family (Meraldi et al. 2006) that raises the question of their origin (Figure 7). These short structures, made of two sequence motifs (CDE I and CDE III) separated from each other by a short AT-rich interval (CDEII), contrast with the 3- to 5-kb-long regional centromeres observed in the CTG clade (Sanyal et al. 2004; Padmanabhan et al. 2008; Kapoor et al. 2015; Chatterjee et al. 2016), in Y. lipolytica (Vernis et al. 1997, 2001; Lynch et al. 2010), and in the methylotrophs (Morales et al. 2013; Coughlan et al. 2016), and which are made of variable sequences often associated to clustered traces of mobile elements or flanked by inverted repeats. Regional centromeres of Saccharomycotina yeasts are reminiscent of the epigenetic centromeres of multicellular eukaryotes made of large arrays of DNA lacking sequence specificity (Brown and O’Neill 2014), but are not identical to them. Compared to S. pombe centromeres, whose central region made of CenH3 nucleosomes (the variant form of histone H3 essential for centromere function) is flanked by heterochromatin like in most eukaryotes, the CenH3 nucleosomes of the regional centromeres of C. albicans or C. lusitaniae are not embedded in heterochromatin (Sanyal et al. 2004; Kapoor et al. 2015). This property may facilitate their neoformation (Ketel et al. 2009), although the phenomenon has also been observed for large epigenetic centromeres.

Open in a separate window

Figure 7

Evolution of centromeres. Left: tree topology of major subgroups of Saccharomycotina derived from Figure 6 with major evolutionary events (red lightning bolts) relative to centromeres. Taphrinomycotina used as outgroup. The loss of classical heterochromatin modification occurred very early in Saccharomycotina after the branching (not shown here) of L. starkeyi (Riley et al. 2016). Right: Centromeres (gray shade) are symbolized on chromosomes (red lines) by their characteristic elements (boxes and arrows). Symbolized below are nucleosomes (circles) of corresponding regions. Dark gray background, normal histone H3; void red circles, classical heterochromatin modifications of histone H3; purple background, CENP-A centromeric histone H3 variant (A). Adapted from Ishii (2009). Structures vary between species for regional centromeres (Chatterjee et al. 2016; Coughlan et al. 2016). IR, inverted repeats.

It has been proposed that the loss of RNAi and of classical heterochromatin in Saccharomycotina (see below) was at the origin of the evolution of their centromeres (Baum et al. 2006; Padmanabhan et al. 2008; Ketel et al. 2009; Malik and Henikoff 2009). The emergence of point centromeres in the Sacharomycetaceae, in which the CenH3 protein is conserved, must have taken place by recruiting novel components such as Ndc10p and Ctf13p to bind the evolutionarily conserved kinetochore machinery to the CDEIII motif, forming the CBF3 complex (Meraldi et al. 2006). Malik and Henikoff (2009) have argued that these two genes, which have no homologs in other organisms, might have originated from the capture of the REP1 and REP2 genes of the 2-μm plasmids specifically found in the Saccharomycetaceae yeasts (Blaisonneau et al. 1997). According to this view, the first point centromeres may have derived from the chromosomal integration of the cis-acting STB locus of these plasmids, hence alleviating the need of the ancestral regional centromeres. How this initial structure spread to all chromosomes remains an open question but, once established, point centromeres have been highly conserved in Saccharomycetaceae chromosomes as judged from the conservation of their flanking genes. This evolutionary process, however, has not been unique as revealed by the new type of point centromeres recently discovered at different chromosomal locations in Naumovozyma castelli and Naumovozyma dairenensis (Kobayashi et al. 2015). The fact that the overexpression of CSE4 (encoding the CenH3 protein) reveals new centromere-like regions (CLR) in the genome of S. cerevisiae suggests a possible mechanism at the origin of novel point centromeres (Lefrancois et al. 2013). CLR are similar in size to point centromeres but lack sequence specificity.

Mating cassettes and mating-type switching

Mating-type switching by the conversion of one idiomorph to its opposite at the MAT locus during mitotic cycles of yeasts has retained considerable attention during the last decades (Heitman et al. 2013). It is a characteristic feature of yeasts that differentiates them from their multicellular fungal ancestors. In S. cerevisiae and S. pombe, the molecular mechanisms at the basis of the phenomenon have been precisely elucidated and they extensively differ (Haber 2012; Klar et al. 2014). Despite the fact that these mechanisms are inherently destructive for genomes (Gordon et al. 2011), mating-type switching has emerged independently several times during yeast evolution, suggesting some selective advantage (Figure 8). One of them might be that switching facilitates the finding of a mating partner of the same species in harsh environmental conditions (Hanson et al. 2014). Another one might be that, by forming homozygous diploids, switching alleviates the risk of dissociation of favorable allelic combinations at the next meiosis, such combinations resulting from insidious selection during the long phases of clonal propagation that are so characteristic of yeasts.

Open in a separate window

Figure 8

Evolution of mating-type switching. Left: same as Figure 7. Red lightning bolts refer to capture of site-specific nucleases. Right: MAT cassettes (colored boxes) are symbolized on chromosomes (gray bars) with MATa (horizontal black stripes) or MATα (vertical black stripes) idiomorphs. Silenced cassettes are shaded. Shown are the topologies observed in S. cerevisiae (for the Saccharomycetaceae) and K. phaffii (for the methylotrophs), but significant variations are found in both subgroups (see text). The α3 and Kat1 transposases of K. lactis are shown acting at the same place as the HO endonuclease of S. cerevisiae, despite important mechanistic differences.

In S. cerevisiae, mating-type switching involves two extra copies of the MAT locus (designated HML and HMR cassettes, respectively), located close to telomeres on the same chromosome and silenced by a combination of the SIR1–4 proteins. Note that the common histone modification (H3K9me), characteristic of silent chromatin in most eukaryotes, is absent in Saccharomycotina due to their original loss of the CLR4 and SWI6/HP1 genes (Hickman et al. 2011). Such a triplication is another intrinsic landmark of genomes of the Saccharomycetaceae family not found in other Saccharomycotina (Figure 8). Members of the CTG clade as well as Y. lipolytica have only one MAT locus per haploid genome, designated MTL, and do not switch mating type. MTL bears either the MATa or MATα idiomorphs with some rare cases of mixing of both (Dujon et al. 2004; Butler et al. 2009). In the heterothallic members of the methylotrophic subgroup capable of switching, the MAT locus is duplicated (see below). Triplication of the MAT locus also occurred in S. pombe and other Schizosaccharomyces species, but this represents an evolutionary convergence (Klar et al. 2014), not an ancestral feature. During evolution of the Saccharomycetaceae family, the triplication has received secondary modifications such as the loss of one or two cassettes in some species with the concomitant loss of switching or even sex (Fabre et al. 2005; Müller et al. 2007; Wolfe et al. 2015), or the expansion to a fourth locus in members of the Eremothecium clade (Dietrich et al. 2013).

In S. cerevisiae, the HML cassette bears copies of the MATα1 and MATα2 genes, and the HMR cassette has a copy of the MATa1 gene (encoding a homeodomain protein that drives meiosis and sporulation). The MATa2 gene, encoding a transcriptional activator belonging to the HMG family conserved in most yeast species, has been lost in the post-WGD species of the Saccharomycetaceae family but is present in members of the ZT and KLE clades (see below and Table 1). In S. cerevisiae, the switching mechanism requires (i) a specific double-strand endonuclease (HO), cleaving DNA within the active MAT locus while leaving intact the silenced copies; (ii) a double-strand break repair machinery using the sequence homology of the flanking regions (designated X and Z); and (iii) a donor preference enhancer (RE) to ensure repair using the appropriate silent cassette (Haber 2012). The HO gene is also found in Z. rouxii, Torulaspora delbrucki (ZT clade), and almost all post-WGD species of Saccharomycetaceae (lost in Kazachstania africana), but it is not found in species of the KLE clade of Saccharomycetaceae. It encodes an endonuclease belonging to the LAGLIDADG family of intein-related proteins that has probably been recruited for mating-type switching in the ancestor of the ZT clade and transmitted to the post-WGD species, or, perhaps, recruited in the ancestor of the entire family but then lost in members of the KLE clade (no species has a complete HO gene). The presence of sequences similar to HO in some of these last species (Fabre et al. 2005) corresponds to relics of an ancestral gene of the same LAGLIDADG family that may or may not have been involved in mating-type switching.

The entirely different switching mechanism discovered in K. lactis (Barsoum et al. 2010; Rajaei et al. 2014) indicates the independent recruitment of a completely different machinery in this lineage. In this case, the MATα to MATa switching events and the reverse MATa to MATα events involve two domesticated transposases, α3 and Kat1, respectively. α3 is encoded by the MATα3 gene located next to the MATα locus and shares homology to the MULE family of transposable elements. Kat1 is the product of the KAT1 gene containing a programmed −1 frameshift and shares homology to the hAT family of transposable elements. Both are under the transcriptional control of the MTS1/RME1 gene product, induced by nutrient limitation. In MATα strains, binding of Mts1 to sites close to the MATα3 gene stimulates the excision of the MULE, inducing MATα to MATa switching to repair the chromosome break. In MATa strains, Kat1 generates two hairpin-capped, double-strand breaks at the fossil imprints of an ancient transposon located within the MATa locus; inducing MATa to MATα switching to repair the chromosome break. Within the KLE clade, the KAT1 gene is specific to the Kluyveromyces species (absent from L. thermotolerans and E. gossypii), suggesting a domestication event of the transposon at the origin of this genus (Rajaei et al. 2014).

The mechanisms of mating-type switching in some methylotrophic yeasts may help us understand the origin of the complex three-locus system that evolved in the Saccharomycetaceae (Hanson et al. 2014; Maekawa and Kaneko 2014). In O. polymorpha and K. phaffii (see Table 3), two MAT loci bearing opposite idiomorphs lie in opposite orientation on the same chromosome. They are flanked by inverted repeats. One is the active locus that determines cell type, the other is silenced by its proximity to the centromere (O. polymorpha) or the telomere (K. phaffii). In both cases, switching occurs by a flip-flop mechanism that inverts the chromosomal segment between the two loci at the same time as it exchanges idiomorphs between active and silent locations. The phenomenon is induced by nutrient starvation. The same situation was recently found in Pachysolen tannophilus and Ascoidea rubescens (Riley et al. 2016). Whether or not this two-locus system has a common origin with the three-locus system of the Saccharomycetaceae remains an open question given the important differences between the molecular mechanisms involved. But its presence in A. rubescens, a member of the Ascoideaceae family, is consistent with an evolutionary intermediate between the methylotrophs and the Saccharomycetaceae.

Tentative reconstruction of the evolution of Saccharomycotina genomes

Considering the above elements, the important evolutionary changes separating the four major subgroups of Saccharomycotina genomes can be summarized as illustrated by Figure 9 using parsimony as a guiding principle, but keeping in mind that more complex scenarios, including multiple and reversible changes, are not excluded. As discussed above, some changes concern RNA molecules and others concern chromosome structure. Among the latter, some events correspond to the intrinsic dynamics of chromosomes in vertical lineages and others are related to horizontal exchanges such as the capture of extrachromosomal elements or interspecific hybridizations (see below). The loss of ancestral elements and functions also played a critical role in subsequent evolutionary events of the Saccharomycotina subphylum. Among those, the early losses of HP1-mediated formation of heterochromatin (Hickman et al. 2011; Riley et al. 2016) and of canonical eukaryotic RNAi (Bernstein et al. 2012a) were probably the two most important events. The decreasing number of spliceosomal introns and the increasing compactness of genomes are general trends that seem to have persisted throughout the evolution of the Saccharomycotina subphylum, but were already significant at its origin.

Open in a separate window

Figure 9

Overview of Saccharomycotina genome evolution. Evolutionary innovations (blue arrows) or losses (red crosses) attributed from parsimony to major separations between subgroups of Saccharomycotina (secondary losses of preceding innovations, not shown for simplicity). Left: RNA-related events. Note that, in some lineages, noncanonical enzymes replaced the ancestral Dicer and that the phylogenetic distribution of Dicer and Argonaute proteins in Saccharomycotina is discontinuous (see text). Right: chromosome-related events. Note that the recruitment of α3 and Kat1 transposases concerns the Kluyveromyces, not the entire KLE clade, and that the loss of H3K9me chromatin modification occurred after the separation of L. starkeyi (data not shown) from other basal lineages. Purple cross: massive post-WGD gene loss. Note that the loss of complex-I genes in mtDNA is not a chromosomal event, but its occurrence is indicated because of its association with the duplication of nuclear genes encoding alternative dehydrogenases (see text).

Among the RNA-related events, a fixation of the 5S RNA gene into the rDNA locus must have occurred very early, as only some members of the basal lineages have dispersed 5S RNA genes (a normal situation for eukaryotes) and no 5S gene in the rDNA locus. Most other yeasts have one (or rarely two) 5S gene(s) in the rDNA locus and no (or few) dispersed copies. But this fixation was probably reversible since some members of the CTG and methylotroph clades, such as M. bicuspicata, Hyphophichia burtonii, and K. phaffii, only have dispersed copies in multiple numbers (Riley et al. 2016). Note, however, that these species represent lateral branches to these clades. The switch to the bacterial decoding mode of the Arg-CGN codon family also occurred very early, whereas the situation appears more complex for the Leu-CUN codon family (with some reversions and multiple reassignments of the CUG codon).

Most chromosome-related events concern the Saccharomycetaceae family, but this is also where they have been most extensively studied, particularly in S. cerevisiae. The WGD was recognized very early (Wolfe and Shield 1997) and its origin and consequences have been extensively studied to the point that post-WGD species are overrepresented among sequenced yeasts (see Table 1). The event, suspected to have occurred from an ancestor with eight chromosomes (Gordon et al. 2009), was followed by extensive gene loss (Scannell et al. 2007a,b), while neofunctionalization or subfunctionalization of the remaining paralogous copies (ohnologs) have played an important functional role in the evolution of these yeasts. The existence in post-WGD genomes of numerous paralogs predating the WGD event was recently interpreted as an indication that the duplication event followed an interspecies hybridization event that probably occurred between a member of the ZT clade (which appears phylogenetically closer to post-WGD) and a member of the KLE clade of Saccharomycetaceae (Marcet-Houben and Gabaldón 2015; Wolfe 2015).

The triplication of the MAT cassettes exists in both clades (with variations in some species) but the recruitment of the HO endonuclease for mating-type switching (probably from a LAGLIDADG intein) is specific to the ZT clade and, consequently, post-WGD species. Less data are unfortunately available about the recruitment of the α3 and Kat1 transposases discovered in K. lactis (Barsoum et al. 2010; Rajaei et al. 2014) and about the mechanism of mating-type switching in other members of the KLE clade. How the duplication of the MAT cassettes in A. rubescens and some methylotroph yeasts (Hanson et al. 2014; Maekawa and Kaneko 2014; Riley et al. 2016) relate to the triplication in Saccharomycetaceae is not yet entirely solved, but it was inferred that mating-type switching evolved soon after the loss of the classical H3K9 chromatin modification, a very early event at the origin of all Saccharomycotina after the separation of L. starkeyi (Riley et al. 2016).

Another highly spectacular evolutionary transition at the chromosomal level is the replacement of the regional centromeres found in all Saccharomycotina, except Saccharomycetaceae, by the point centromeres specific to the latter family; a phenomenon proposed to result from the capture of genes and cis-acting elements from the circular 2-μm plasmids into chromosomes (Malik and Henikoff 2009). Other captures have been reported which, together with other mechanisms (below), have contributed at various levels during the evolution of Saccharomycotina.

While writing this review article, the authors’ laboratories were supported in part by the Centre National de la Recherche Scientifique (UMR3525), Université Pierre et Marie Curie (UFR927) and the L’Agence Nationale de la Recherche (contract DYGEVO from program blanc 2011 SVSE6) for B.D.; and by the Biotechnology and Biological Sciences Research Council and the University of Leicester for E.J.L. The authors declare no conflict of interest.