Using the biological species definition, yeasts of the genus Saccharomyces sensu stricto comprise six species and one natural hybrid. Previous work has shown that reproductive isolation between the species is due primarily to sequence divergence acted upon by the mismatch repair system and not due to major gene differences or chromosomal rearrangements. Sequence divergence through mismatch repair has also been shown to cause partial reproductive isolation among populations within a species. We have surveyed sequence variation in populations of Saccharomyces sensu stricto yeasts and measured meiotic sterility in hybrids. This allows us to determine the divergence necessary to produce the reproductive isolation seen among species. Rather than a sharp transition from fertility to sterility, which may have been expected, we find a smooth monotonic relationship between diversity and reproductive isolation, even as far as the well-accepted designations of S. paradoxus and S. cerevisiae as distinct species. Furthermore, we show that one species of Saccharomyces—S. cariocanus—differs from a population of S. paradoxus by four translocations, but not by sequence. There is molecular evidence of recent introgression from S. cerevisiae into the European population of S. paradoxus, supporting the idea that in nature the boundary between these species is fuzzy.
THE concept of a species is central to the biological sciences (Mallet 1995; Coyne and Orr 1998), but the definition of a species is controversial. There are several sometimes conflicting definitions of species. The biological species concept (BSC) is based on patterns of breeding: species are considered to be units reproductively isolated from other such units, but within which interbreeding and genetic recombination reduce the possibility of divergence. Asexual reproduction in many organisms including fungi (Taylor et al. 2000), hybridization in plants (Rieseberg 1997) and fungi (Taylor et al. 2000), and lateral gene transfer between bacteria (Gogarten and Townsend 2005) create problems for the BSC. The genotypic cluster species concept (GCSC) can accommodate some gene flow between species (clusters) as long as their integrity remains such that they can be distinguished (Mallet 1995). Genotypic clusters can be at the subpopulation level as well as at species level. The phylogenetic species concept (PSC) defines the species as a distinct monophyletic group on the basis of evolutionary history and geographical extent. This concept faces the difficulties of incomplete knowledge and a somewhat arbitrary delineation of specific boundaries (Hudson and Coyne 2002).
Sequence and chromosomal evolution can produce genetic divergence and can reduce or prevent interbreeding, as required by these species concepts. Although these concepts do not always agree (Isaac et al. 2004), early analyses of Saccharomyces strains by interbreeding (Naumov 1987) (BSC) and DNA/DNA reassociation (Martini and Martini 1987; Vaughan Martini 1989) (PSC) were concordant in supporting the existence of three distinct species: Saccharomyces cerevisiae, S. paradoxus, and S. bayanus. These studies also confirmed the identification of one natural hybrid (S. pastorianus). Recent studies of reproductive compatibility have defined three new species within the complex on the basis of BSC: S. cariocanus, S. mikatae, and S. kudriavzevii (Naumov et al. 1995a,b, 2000).
S. cerevisiae is a well-known model organism associated with human activity (baking and brewing). Its closest relative is S. paradoxus, which is often isolated from Quercus (oak) trees and has little or no association with human activity. The two species show perfect synteny with no gross chromosomal rearrangements between them (Fischer et al. 2000; Kellis et al. 2003) and a level of sequence divergence of ∼15% (Cliften et al. 2001). The different species within the sensu stricto complex can mate and generate viable hybrids, indicating an absence of prezygotic isolation. Coexistence of these species in similar habitats (Sniegowski et al. 2002) and the isolation of interspecific hybrids (de Barros Lopes et al. 2002; Liti et al. 2005) show that interbreeding not only is possible but also does occur in the wild.
A postzygotic barrier exists between the sensu stricto species, resulting in few viable spores (≤1%) (Naumov 1987). Several experiments have shown that this postzygotic barrier is not due to genetic incompatibilities or to chromosomal mechanisms such as translocations, which are the two mechanisms accepted for intrinsic hybrid sterility (Coyne and Orr 2004). Many of the Saccharomyces sensu stricto species are collinear (Fischer et al. 2000; Kellis et al. 2003), ruling out major structural changes as causes of sterility in collinear hybrids. Restoration of collinearity between two of these species did result in the production of viable gametes (Delneri et al. 2003) but these derived fertile hybrids were highly aneuploid. Euploid, 2N hybrids between the collinear species exhibit the low fertility seen between the noncollinear species (Naumov 1987; Naumov et al. 1992, 1995a,b; Hunter et al. 1996). Major dominant genetic incompatibilities are ruled out by converting 2N hybrids into 4N allo-tetraploids (Greig et al. 2002). These allo-tetraploids have high levels of fertility and produce 2N hybrid gametes. The ability to replace individual S. cerevisiae chromosomes with S. paradoxus homologs in the lab reduces the likelihood that there are any recessive genetic incompatibilities maintaining reproductive isolation. Chromosome III has been replaced with no problems of viability (Chambers et al. 1996) and most of the rest of the S. cerevisiae chromosomes have successfully been replaced by S. paradoxus chromosomes (D. Greig, personal communication).
Instead of these two mechanisms, simple sequence divergence and the action of the mismatch repair system have been shown to account for hybrid sterility in Saccharomyces. In an S. paradoxus × S. cerevisiae hybrid only 1% of gametes are viable (Naumov 1987; Naumov et al. 1992; Hunter et al. 1996). These viable gametes are highly aneuploid and exhibit little or no recombination (Hunter et al. 1996). As mismatch repair is known to greatly reduce recombination between diverged genomes in bacteria (Rayssiguier et al. 1989), Hunter et al. (1996) tested whether its action on the sequence divergence between the yeast species could account for the sterility due to prevention of the homologous recombination necessary for proper meiotic segregation of the homologous chromosomes. In mismatch-repair-defective hybrids, fertility increased 10-fold, which correlated with an increase in homologous recombination and reduced aneuploidy. Further experiments on a single S. paradoxus chromosome in an S. cerevisiae background supported the hypothesis that sterility in these hybrids was mostly, if not entirely, due to the normal mismatch repair system acting on the sequence divergence (Chambers et al. 1996). This effect of mismatch repair on fertility was also seen in populations of the same species that have diverged and may play a role in the speciation process itself rather than simply maintaining the postzygotic barrier after speciation by some other mechanism (Greig et al. 2003). The question remains as to how much divergence results in sufficient reproductive isolation to designate a species.
MATERIALS AND METHODS
Saccharomyces strains and crosses:
Strains used in this study are listed in Table 1. Yeast crosses were obtained either by generating haploid versions of strains [by deleting the HO gene using the dominant drug resistance markers KAN and HYG (Goldstein and McCusker 1999)] or by crossing spores from ura3 and lys2 mutants [previously selected in 5-FOA and α-aminoadipate media, respectively (Naumov 1987)] in a minimal medium. The HO deletion was performed by PCR-mediated gene replacement and confirmed using the following primers (5′–3′): HO-DF, (AGAGCTTTTGAAGGTGAACCTGGTAGGTTAGACCCTAGGCGTAGAAC CGTACGCTGCAGGTCGAC); HO-DR, (GGCTCTTTTGTTGTACCGTCACCTAGCCATAGACCAAGCATCCAAGCCATATCGATGAATTCGAGCTCG); HO-A1, (GTTATGTGCGCAGATGGCTC); and HO-A4, (GTCACAGTAGCAGACATTCC).
Sporulation was induced for 3–5 days at room temperature in 1% potassium acetate media and spores were dissected as previously described (Naumov et al. 1994). Some of the crosses involved strains differing in reciprocal and nonreciprocal translocations (Table 2). Each of the translocations studied induces the formation of multivalents during meiosis, causing a 50% loss of gametic viability if reciprocal and a 25% loss if not reciprocal.
Chromosome separation in Saccharomyces strains:
Genomic DNA for CHEF analysis was prepared as previously described (Louis 1998). The program for separation of the chromosomes consisted of two blocks: block 1—20-hr, 60-sec switching time and block 2—12-hr, 90-sec switching time, using a CHEF–DRIII apparatus (Bio-Rad, Hercules, CA). Both blocks were run at a voltage of 6 V/cm, angle 120° in 0.045 m Tris–borate, 0.045 m boric acid, and 0.001 m EDTA (0.5× TBE) at 13°.
Yeast DNA purification, sequencing, and analysis:
Genomic yeast DNA were extracted and purified as previously described (Louis and Borts 1995). DNA sequence was directly obtained from PCR products by Agowa (http://www.agowa.de). We sequenced yKu70, yKu80, NEJ1, EST2, TLC1, and SPT4. Genes were sequenced at two to six times coverage. The region indicated in Figure 4 was sequenced in a subset of strains. Sequences were filtered and assembled using the Staden Package (available at http://staden.sourceforge.net/). Synonymous and nonsynonymous substitutions were calculated using DnaSP software (http://www.ub.es/dnasp/; Rozas and Rozas 1999). Sequence comparison, alignment, and phylogenetic trees were obtained using DNAstar package software (EditSeq and MegAlign, http://www.dnastar.com). Sequence divergence is calculated by comparing sequence pairs in relation to the phylogeny reconstructed by MegAlign. The transfer of S. cerevisiae sequence to S. paradoxus CBS432 was identified when using ad hoc Python scripts to analyze the sequence data (Kellis et al. 2003) and mVISTA (http://genome.lbl.gov/vista/index.shtml).
The time estimation of gross chromosomal rearrangements (GCR) and geographic divergence reported in Table 1 is calculated using t = Ks/(2 × u × g), where Ks is the synonymous substitution, u is the point mutation rate (1.84 × 10−10), and g is the maximum number of generations per year according to Fay and Benavides (2005). The range of years takes into account a range from one (second value) to eight (first value) generations/day.
RESULTS AND DISCUSSION
Genetic variation and gamete viability:
In this report, we test the relationship between sequence divergence and meiotic sterility to determine if there is a level of divergence at which speciation occurs, that is, a level sufficient to result in reproductive isolation. We determined the degree of divergence by sequencing six nuclear genes (∼10 kb) in 41 selected strains within the complex (Table 1). We sequenced yKu70, yKu80, NEJ1, EST2, TLC1, and SPT4. The proteins encoded by yKu70 and yKu80 form a strongly interacting heterodimer and are paradoxically involved in both telomere capping and DNA repair via nonhomologous end joining (NHEJ) (Dudasova et al. 2004). NEJ1 is also involved in regulating NHEJ through ploidy and has a paradoxical role in end protection vs. double-strand break repair as well (Liti and Louis 2003). EST2 and TLC1 are part of the telomerase complex (Smogorzewska and de Lange 2004). EST2 is the catalytic subunit of the reverse transcriptase and TLC1 is the RNA template. Moreover, TLC1 is not translated. Finally, SPT4 is a transcription factor and has been recently shown to be part of the kinetocore complex (Crotti and Basrai 2004). The six genes are on different chromosomes except NEJ1 and EST2, which are ∼100 kbp apart on chromosome XII, and this allows linkage disequilibrium and recombination analyses.
These genes exhibit different levels of divergence and amounts of selective pressure (Figure 1). However, each of the six genes gave a similar phylogenetic topology (Figure 2) and were consistent with the genome-scale approach (Rokas et al. 2003). A phylogenetic tree obtained from a concatenated analysis of four of these genes is shown in Figure 3 and is consistent with ITS1 phylogeny (Liti et al. 2005).
The sequences of S. paradoxus fall into three clades correlated with geography on a continental scale. We found three major subpopulations: European, Far Eastern (East Russia and Japan), and North American. Within these subpopulations, the levels of sequence divergence range from 0.08 to 0.21%. Between them, the divergence ranges from 1.5% (between European and Far Eastern populations) to 4.6% (between European or Far Eastern and North American populations; see Figure 3). The values of sequence divergence between S. paradoxus strains raise a question: using the PSC, at what level of divergence should monophyletic taxa be considered as subgroups, subspecies, or distinct species? With enough population data, the application of the GCSC may help with this distinction. A single lineage seems to occupy the whole North American continent. Different sets of strains, isolated in different places and at different times (Table 1) have similar sequences. The recently described species S. cariocanus, which has so far been found only in Brazil, appears in our analysis within the North American lineage of S. paradoxus. Strain NBRC1803 of S. kudriavzevii is highly divergent (4.6%) from the other three strains, despite being found geographically with the other isolates (Figure 3). Variation among S. cerevisiae strains, on the other hand, shows little in the way of geographical structure. Isolates from neighboring areas do not necessarily group together. Moreover, sequence divergence between locations was significantly lower in S. cerevisiae than in the other species, ranging from 0 to 0.6%. This is not really different from the phylogenetic studies seen in the analysis by Aa et al. (2006) where they report differences between North American isolates (some of these are included in this study) and vineyard isolates, for example. The S. cerevisiae structures reported here and in Aa et al. (2006) are not as distinct as those seen in S. paradoxus. The relative lack of diversity is probably a consequence of domestication (Fay and Benavides 2005), could reflect movement through human activity, and is consistent with a more recent common ancestor, or founding population, for all extant S. cerevisiae strains (Sniegowski et al. 2002; Aa et al. 2006).
Our data were used to assess the correlation between reproductive isolation and sequence divergence. The degree of divergence partially fills the gap between the extremes of ∼1 and ∼15% found within and between the species previously analyzed (Naumov 1987; Greig et al. 2003). When the percentage of viable spores is plotted against sequence divergence (Figure 4A), several points lie outside a general linear trend. When chromosomal translocations are taken into account (see below), these data points move into the general mass of data (Figure 4B). The degree of sequence variation strongly correlates with meiotic sterility and several smooth, monotonic relationships can fit the data (including linear, exponential, etc.). The best fit is an exponential decay curve (y = −12.75 + 104.52 × e(−0.149x); r = 0.98, where x is sequence divergence and y is gamete viability). The linear and exponential curves are also very good fits (see Figure 4B). All combinations of crosses between the three major subpopulations in S. paradoxus exhibit partial reproductive isolation (Table 2 and Figure 4), in agreement with previous reports (Sniegowski et al. 2002; Greig et al. 2003). Crosses within the same subpopulation exhibit a high frequency of viable spores (>∼80%).
Translocations and gamete viability:
The new species S. cariocanus is interesting because, when the reductions in viability due to the four reciprocal translocations (16-fold; see materials and methods) are accounted for, we find that crosses between S. paradoxus and S. cariocanus strains can produce a high corrected frequency of viable spores (77–81%; Table 2). Although these strains differ in chromosomal organization, they differ little in their sequences. Support for S. cariocanus as a member of the S. paradoxus species and in particular the American subpopulation comes from analysis of the viable spores. In hybrids between species, aneuploidy occurs at high frequencies due to the lack of homologous recombination (Hunter et al. 1996). In the S. cariocanus × S. paradoxus hybrids we do not see any aneuploidy in the viable spores (Figure 5). Instead, we see the balanced chromosome sets expected for heterozygous translocations.
We analyzed the viability of spores in crosses between S. paradoxus strains with chromosomal rearrangements and varying levels of sequence divergence. We used a Far Eastern isolate, N-43, carrying one reciprocal translocation (VIII t XV and XV t VIII, using the nomenclature of Fischer et al. 2000) and a European strain, CBS5829, carrying one nonreciprocal translocation (XI t V). We crossed each of these with strains from all three geographic subpopulations of S. paradoxus (Figure 4). Reciprocal and nonreciprocal chromosomal translocations produce, respectively, 50 and 25% drops in gametic viability due to unbalanced segregation of essential genes on the translocated arms. The translocations are all large enough to contain at least one essential gene. The strains with chromosomal translocations do not show as large a sequence divergence compared to others within the same population, as might have been expected, thereby reinforcing the idea that chromosomal evolution has a minor impact on speciation in the Saccharomyces complex (sensu stricto) (Fischer et al. 2000; Delneri et al. 2003). A rapid accumulation of translocations, as happened in S. cariocanus, may influence the speciation process over a longer period as they clearly have an effect on fertility. At this point in time, however, S. cariocanus has not diverged much in sequence from the rest of the American S. paradoxus.
Evidence of introgression from S. cerevisiae into the European population of S. paradoxus:
The smooth monotonic relationship between sequence divergence and gamete viability, rather than the appearance of a discontinuity or rapid decrease, makes the species boundary hard to define. Indeed, the lower, but not negligible, probability of successful breeding between diverged lineages may have strong influences on the population and genomic structures of the species concerned. Traces of interbreeding may still be present in the genomes of Saccharomyces. We searched for evidence of past interbreeding by comparing the available whole-genome sequences of S. paradoxus and S. cerevisiae. Pairwise comparisons of the two genomes show an average identity of 85% across chromosomes, a finding consistent with earlier analyses (Kellis et al. 2003), except for one region that is aberrantly high. A subtelomeric segment on the left arm of chromosome XIV, 23 kb long, exhibits an anomalously high identity between S. cerevisiae S288C and S. paradoxus CBS432. The region extends between 15 and 38 kbp from the left telomere (in S. cerevisiae) and encompasses 12 open reading frames (Figure 6A). The average identity in this region (>99%) is significantly higher than in the rest of the genome. This could have been error in cloning or sequencing within the genome project, an error in handling the strains concerned, or a true introgression from one species into the other. To exclude a mistake in sequencing. we resequenced a small region within this segment in S. paradoxus and confirmed the original results (Figure 6, A and B). We then sequenced the same region in several isolates of S. cerevisiae and S. paradoxus. We found similar sequences in all strains of S. cerevisiae and all European strains of S. paradoxus, but not in the Far Eastern and American isolates (Figure 6B), which exhibit an average divergence from S. cerevisiae of 15%, in line with the rest of the genome. These results indicate that S. cerevisiae was the donor and S. paradoxus the recipient, either in a direct lateral transfer or through introgression after hybridization. They also indicate that the transfer took place after the split of the European and Far Eastern lineages. There is the possibility of a partial hybridization, where a single chromosome moves from one genome to the other, as in kar1 mutants (Nilsson-Tillgren et al. 1980). However, we do not know how often such events may occur.
We favor introgression via hybridization for several reasons. Horizontal gene movements are quite rare in yeast (Dujon 2005; Liti and Louis 2005) and a large DNA segment is unlikely to be transferred. Furthermore, the segment involved is a homologous replacement and does not extend up to the telomeric repeats. S. cerevisiae and S. paradoxus can be found in the same localities in nature (Sniegowski et al. 2002) and therefore have the opportunity to hybridize. Naturally occurring hybrids between these species have been found (Liti et al. 2005). We propose a scenario where a spontaneous hybridization between S. cerevisiae and S. paradoxus resulted in a rare viable gamete, perhaps with a majority of S. paradoxus sequences. Viable gametes could be composed of a range of chromosomes from either parent due to random segregation and therefore could have a majority of chromosomes from one parent or the other (G. Liti, unpublished results). The viable gametes, although aneuploid, will be nearly euploid as most of multiple aneuploidies are inviable (Parry and Cox 1970; Louis and Haber 1989).This could be followed by many backcrosses within the S. paradoxus parent population, leaving a single DNA segment from S. cerevisiae. To become fixed in the European population, this segment would probably require a selective advantage. A bottleneck in the ancestral European populations could also have fixed this transferred segment. The KRE1 gene, which can confer resistance to killer toxins, seems at first sight to be a good candidate in this regard, but the different European isolates proved to have different resistances to killer toxins (results not shown). No cause for a selective advantage has so far been identified.
Minimum age of common ancestor:
The sequence divergence between populations and species can be used to estimate the minimum age since the most recent common ancestor. Using known mutation rates and estimates of generation times from studies in S. cerevisiae (see materials and methods and Fay and Benavides 2005), absolute dates can be estimated (Table 3). By assuming that the generation times and mutation rates are similar in all of the Saccharomyces sensu stricto isolates over time, we can also estimate the perhaps more informative relative age of populations. Using a range of potential generation rates, from one to eight/day on average, we can set the time of the most recent common ancestor of S. cerevisiae and S. paradoxus to between 0.4 and 3.4 MYA. The relative age of the common ancestor of the most diverse S. cerevisiae isolates is 2.7% of this, which is similar to the diversity seen within subpopulations of S. paradoxus. The relative age of the different subpopulations of S. paradoxus is much higher, ranging from 6.3% between Far Eastern and European isolates to 28.6% of the age of the species when comparing North American to Far Eastern or European populations (Table 3). This analysis can be applied to isolates with gross chromosomal rearrangements, which would be expected to have increased diversity if they were old and prevented gene flow with neighbors not harboring the same rearrangements. In every case, the relative ages of the GCRs containing isolates was on the order of 0.2–0.3% of the age of the species. The South American species S. cariocanus has a relative age, compared to the North American species S. paradoxus, of only 1.4% of the time to the most recent common ancestor, or between 5000 and 40,000 years ago. The four reciprocal translocations seen in this population must have occurred in a burst on an evolutionary time scale as predicted by Fischer et al. (2000). However, we cannot infer if the translocations occurred simultaneously or sequentially.
Fuzzy boundaries in Saccharomyces species:
Earlier results have shown that spontaneous hybrids can be found at reasonable frequencies (de Barros Lopes et al. 2002; Liti et al. 2005) and that the hybrid species S. pastorianus originated from several hybridizations rather than from one (Liti et al. 2005). Although lateral gene transfer has been rarely detected in Saccharomyces yeast, we have recently suggested that one of the yeast LTR–transposons, Ty2, has been transferred between S. cerevisiae and S. mikatae (Liti et al. 2005).
Overall, our data suggest a continuum of genetic differentiation: extant species may simply be the result of an extensive decay in sequence homology. A gradient of gamete viability is correlated with sequence divergence. These analyses are consistent with studies in Drosophila (Coyne and Orr 2004) and Neurospora (Dettman et al. 2003) where reproductive isolation (both pre- and postzygotic) increases over time (genetic distance), but in those studies the correlation appears weaker. The findings in Drosophila are consistent with the model of Bateson–Dobzhansky–Muller (BDM) (Coyne and Orr 2004) where two or more genetic incompatibilities account for the reproductive isolation. A special case of the BDM model acted in speciation of the hemi-ascomycete yeasts soon after the whole-genome duplication event (Scannell et al. 2006). Indeed, “speciation” genes have been identified (Orr et al. 2004). One might assume that the findings presented here are consistent with the BDM model of intrinsic postzygotic isolation with a large number of genes involved (Coyne and Orr 2004); however, there is no evidence of genetic incompatibilities here—only divergence acted upon by the normal process of mismatch repair. Any genetic incompatibilities between pairs of strains used in this study should result in fertilities inconsistent with the trend seen for sequence divergence. After accounting for the chromosomal incompatibilities of some crosses due to translocations, all points fit the correlation with sequence divergence. In Neurospora, there was a great deal more variation in reproductive success over genetic distance than seen here with phylogenetically diverged strains still capable of successful reproduction (Dettman et al. 2003).
Evidence of frequent spontaneous hybridization, lateral gene transfer, and introgression suggests that species boundaries are fuzzy and that networks may be present in the phylogeny. Classical phylogenetic trees can give only a glimpse of true relationships. We suggest that a clear picture of the population structure of Saccharomyces requires modified phylogenies that can account for ongoing gene flow due to hybridization, introgression, or lateral transfer with evolving reproductive barriers. The GCSC concept may be the best for describing yeast species and its application at the population level up to the species and genus level may yield a better understanding of species and speciation in Saccharomyces. This view may be relevant to many other species.
We thank Bryan Clarke, Paul Sharp, John Brookfield, Jim Mallet, and Daniel Hartl for helpful discussions and The Wellcome Trust for funding.
- Received June 16, 2006.
- Accepted August 4, 2006.
- Copyright © 2006 by the Genetics Society of America