• Calum J. Maclean,

    1. Research Department of Genetics, Evolution, and Environment, University College London, Wolfson House, 4 Stephenson Way, London NW1 2HE, United Kingdom
    2. E-mail:
    3. Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, Michigan 48109
    Search for more papers by this author
  • Duncan Greig

    1. Research Department of Genetics, Evolution, and Environment, University College London, Wolfson House, 4 Stephenson Way, London NW1 2HE, United Kingdom
    2. Max Planck Institute for Evolutionary Biology, August Thienemann Straße 2, 24306 Plön, Germany
    3. E-mail:
    Search for more papers by this author


Whole-genome duplication has shaped the genomes of extant lineages ranging from unicellular fungi to vertebrates, and its association with several species-rich taxa has fuelled interest in its potential as a catalyst for speciation. One well-established model for the evolution of reproductive isolation involves the reciprocal loss of redundant genes at different loci in allopatric populations. Whole-genome duplication simultaneously doubles the entire gene content of an organism, resulting in massive levels of genetic redundancy and potential for reciprocal gene loss that may produce postzygotic reproductive isolation. Following whole-genome duplication, different populations can potentially change or lose gene function at different duplicate loci. If such populations come back into contact any F1 hybrids that are formed may suffer reduced fertility as some of the gametes they produce may not carry a full complement of functional genes. This reduction in hybrid fertility will be directly proportional to the number of divergently resolved loci between the populations. In this work, we demonstrate that initially identical populations of allotetraploid yeast subjected to mutagenesis rapidly evolve postzygotic reproductive isolation, consistent with the divergent loss of function of redundant gene copies.

Reproductive isolation can create and maintain species by stopping or reducing gene flow between them. Understanding the mechanisms that underlie reproductive isolation has been a major long-term goal of evolutionary biology (Coyne and Orr 2004). The classic model for the evolution of genetic incompatibilities that cause reproductive isolation was initially developed by Dobzhansky (1934) and later expanded on by Muller (1939). In its simplest form, an ancestral population with, for example, a genotype AABB is split into two populations that evolve new alleles, a and b. These alleles are compatible with the ancestral alleles and become fixed, producing new genotypes aaBB and AAbb. If the populations later come back together, F1 hybrids (AaBb) will form, bringing the two new alleles, which may be incompatible, together for the first time. If an incompatibility is dominant the F1 hybrid may be inviable or sterile, and if it is recessive then some of its gametes (ab), or the F2 hybrids produced from them (aabb), may be inviable or sterile (Turelli and Orr 2000).

Gene duplication plays an important role in evolution (Zhang 2003) and the reciprocal loss of redundant duplicate gene copies between diverging populations is potentially a major cause of reproductive isolation by a form of the Dobszhansky–Muller process (Werth and Windham 1991; Lynch and Force 2000). If an essential gene (A) is duplicated to another locus (B), selection is unlikely to conserve the redundant essential gene function at both loci. If two (or more) isolated populations are formed following gene duplication, the loss or change in function of either copy of a duplicate gene should occur at random, with isolated populations losing the essential function from different copies of a given redundant gene on average half the time. As the likelihood of such reciprocal gene loss between populations increases with the number of duplicate genes carried by the founding genotype, whole-genome duplication (WGD) offers the maximum potential for divergent gene resolution between populations. Figure 1 outlines how reciprocal gene loss (divergent resolution) between initially isogenic allotetraploid (formed by species hybridization and genome duplication) populations can result in postzygotic reproductive isolation. In species whose gametes express their genomes, matings between populations that have undergone divergent resolution will produce viable F1 hybrids with reduced fertility, as one-fourth of their gametes will lack a functional copy of each divergently resolved essential gene and will be inviable. The fertility of such hybrids will be 0.75n, with n denoting the number of essential divergently resolved genes between mated individuals (Werth and Windham 1991). Allotetraploids, unlike autotetraploids that are formed by genome duplication within a species, carry two distinct sets of homeologous chromosomes. If chromosomes successfully pair and segregate only with homologues from the same species, then an allotetraploid can behave in meiosis just like a diploid with double the number of chromosomes and double the number of loci, producing viable balanced gametes genetically equivalent to F1 diploids (Fig. 1). In some taxa, such as Drosophila, genetic incompatibilities already established between species can prevent such chromosome segregation in hybrid diploid and in allotetraploid cells (Dobzhansky 1933). But Saccharomyces yeast hybrids lack such genetic incompatibilities (Greig 2008), and WGD of sterile F1 hybrids gives every chromosome an identical partner to pair with in meiosis, restoring normal meiosis and fertility (Greig et al. 2002).

Figure 1.

Divergent resolution of a duplicated essential gene in an allotetraploid leading to F1 hybrid sterility.
1. Locus A encodes an essential gene. Its homolog in another species is locus B. The two diploid species with AA (black rectangles) and BB (gray rectangles) genotypes produce haploid gametes with genotypes A and B, respectively.
2. Gametes from different species fuse forming a diploid F1 hybrid. Genes A and B become two alleles at a single heterozygous locus with genotype AB.
3. Diploid F1 hybrid undergoes a whole-genome duplication. A and B become two redundant homozygous loci with genotype, AABB.
4. Allotetraploid population is split into two separate populations.
5. Allotetraploids produce diploid gametes, AB.
6. Gene function is lost in redundant loci by new mutations. One population loses the original function of A with new allele a, in the other population the function of B is lost with new allele b (a and b alleles = white rectangles). The new mutations spread to fixation by genetic drift or natural selection, evolving genotypes aaBB and AAbb in the two isolated populations.
7. Evolved tetraploids produce viable diploid gametes with genotypes aB and Ab.
8. Previously isolated populations are reunited. Viable F1 hybrids between the populations have hetrozygous genotype AaBb
9. F1 hybrids produce viable diploid gametes with genotypes aB, Ab, AB, and inviable genotype ab, which lacks any allele encoding the original essential function. F1 hybrid fertility is reduced by 25%.

Comparisons between related genomes have shown that WGD has been important for the evolution of many species. The sudden duplication of every gene can potentially be either advantageous or disadvantageous to an organism (for recent reviews see Comai 2005 and Otto 2007). Newly formed polyploids often have lower fitness due to phenotypes intermediate to their parent species, drastic changes in gene expression, and lack of suitable mates of the same ploidy. But they can also benefit from an increased niche range and a larger genome in which redundant genes, freed from selective constraints, can evolve new functions. Although genome duplication is more likely to confer a hindrance than a benefit it is a common occurrence in plants (for review see Soltis et al. 2009) and recent findings suggest that many plants that survived the KT mass extinction were polyploids (Fawcett et al. 2009). WGD has also occurred in yeast (Wolfe and Shields 1997; Dietrich et al. 2004; Kellis et al. 2004) and amphibian lineages (Comber and Smith 2004; Van de Peer 2004) and one or two rounds of WGD are believed to have occurred in the vertebrate lineage (Ohno 1970; see for review Panopoulou and Poustka 2005) along with a further, more recent, additional WGDs in fish (Van de Peer 2004) and desert rat lineages (Gallardo et al. 2004).

Although yet to be proved empirically, WGD is often associated with diverse, species-rich groups, such as the flowering plants (reviewed in Soltis et al. 2009) and the teleost fish (Jaillon et al. 2004), suggesting that it may catalyze speciation (Taylor et al. 2001) The theory of reproductive isolation by reciprocal gene loss is well developed but few concrete natural examples of divergent resolution following WGD have been identified, mainly because it is difficult to detect after rearrangement, sequence divergence and rediploidization of the initially polyploid genome over time (Wolfe 2001; Blanc and Wolfe 2004). Perhaps the best example of reciprocal gene loss following WGD can be found in the yeast lineage. Scannell et al. (2007) found that between Saccharomyces cerevisiae and Kluyveromyces polysporus (two species that diverged shortly after the genome duplication) ∼44% of single copy ancestral loci that could be scored reliably were paralogs rather than orthologs. However, although the authors went on to demonstrate that the rate of reciprocal gene loss was highest shortly after the genome duplication, it is still difficult to ascertain if the gene loss was instrumental in yeast speciation or if some other mechanism was responsible. A similar analysis of the zebrafish (Danio rerio) and pufferfish (Tetraodon nigroviridis) genomes led the authors to estimate that ∼1700 ancestral loci had undergone reciprocal gene loss between the two species and that they likely contribute to their reproductive isolation (Sémon and Wolfe 2007). The primary hurdle to determining whether divergent resolution has been involved in speciation events is the lack of genome sequences that fill the evolutionary time frame between those currently available. As WGD is relatively common in plant and fish lineages, it is likely that more cases of reciprocal gene loss will be detected as more genomic sequences become available. Although it has been difficult to confirm the contribution of divergent resolution to reproductive isolation, the duplication and subsequent divergent resolution of individual genes has been shown to produce genetic incompatibilities. For example, Bikard et al. (2009) recently demonstrated that the divergent resolution of a single gene duplicate results in a “weak root” or “embro lethality” phenotype in F2 Arabidopsis thaliana hybrids, depending on which combination of parental strain alleles they receive. Yamagata et al. (2010) found that the reciprocal loss of a duplicated mitochondrial ribosomal protein is responsible for F1 pollen sterility of hybrids of cultivated (Oryza sativa) and a wild rice (O. glumaepatula). Such studies neatly demonstrate the potential of gene duplication followed by divergent resolution to reduce the fertility of hybrids and strengthen reproductive isolation.

Although detecting and interpreting reciprocal gene loss in lineages is difficult, it is even more difficult to determine whether the differential gene loss caused speciation or whether it simply occurred following reproductive isolation from some other cause. Laboratory experiments using fast-growing, genetically tractable microbes can potentially offer insights into such evolutionary processes. The tetraploid hybrid we used in this work was produced by a WGD in a diploid F1 hybrid between S. cerevisiae and S. bayanus (Greig et al. 2002). The two species show approximately 80% DNA sequence similarity in coding regions, and diverged approximately 20 million years ago. Their genomes are approximately equal in size (∼12 Mb) and gene content (∼5500 genes), and their 16 chromosomes are largely collinear, with the S. bayanus having just five reciprocal translocations and three inversions relative to S. cerevisiae (Kellis et al. 2003). F1 diploid hybrids are sterile because sequence divergence prevents chromosomes from the different species from pairing, recombining, and segregating into gametes in a hybrid meiosis (Greig 2008). But duplicating the hybrid genome gives every chromosome an identical partner to recombine with and the allotetraploid has normal fertility, producing viable gametes (Greig et al. 2002). We find that after repeated rounds of mutagenesis, meiosis, and autofertilization of a single gamete, allotetraploid strains remain fertile. But crosses between replicate allotetraploid lines evolved separately produce many inviable gametes, consistent with the divergent loss of redundant gene function.

Materials and Methods


The experimental design used in this work is outlined in Figure 2. In detail, a single colony of the S. cerevisiae/S. bayanus allotetraploid strain YDG 624 (Greig et al. 2002) was randomly selected, inoculated into 5-mL minimal media (0.67% yeast nitrogen base without amino acids, 2% glucose), and incubated at 30°C with shaking for 24 h. Twenty isogenic populations were then founded by transferring 50 μL into tubes containing 5-mL fresh minimal media. A sample of the initiating culture was also frozen at −70°C to preserve the ancestral genotype.

Figure 2.

Outline of experimental evolution methods used. The figure shows the experimental design used, specific details of which can be found in the methods section. The design can be broken into two main components “Mutagenesis,” where the individual populations are repeatedly treated with EMS, auto-fertilized and bottlenecked to a single homozygous genotype (Steps 1–6) and “Crossing,” where populations are paired and the spore viability of each line and their F1 hybrid is assessed to determine if they are reproductively isolated (Steps 7–10). Throughout the figures ovals represent vegetative cells whereas circles represent spores.
 1. An initially isogenic allotetraploid population was used to form 20 independent populations, only two of which are shown here for simplicity.
 2. Each population is treated with the mutagen EMS to introduce null mutations in redundant gene duplicates, and allowed to recover for 24 h in rich YPD medium.
 3. The population is transferred to sporulation medium to induce meiosis.
 4. The four meiotic products (gametes) produced by the meiotic division of 10 allotetraploid cells (only three are shown) were dissected onto YPD plates.
 5. Each viable spore germinates and undergoes auto-fertilization (each mother cells mates with its daughter cell following mating-type switching) forming homozygous allotetraploids. Each colony formed was screened for its ability to grow on minimal medium and undergo meiosis, before one was randomly chosen to restart a liquid culture.
 6. Steps 2–5 were repeated 12 times before progressing to “Crossing”.
 7. Each population was sporulated to produce gametes and randomly assigned a partner.
 8. Using a tetrad dissection microscope, gametes from each population were mated both to other gametes from the same population, and also to gametes from their partner population to produce F1 hybrids.
 9. Cells resulting from within-population matings and between-population matings were transferred to sporulation media to induce meiosis.
10. Gametes were dissected and the spore viability of the parental populations and their hybrids was assessed.

All 20 populations were incubated at 30°C for ∼24 h with shaking. The populations were then diluted to an OD 600 nm of 0.6 (∼107 cells of the ancestral genotype/mL), 1 mL of cells were pelleted by micro-centrifugation, the supernatant was removed and the cells were washed in 1-mL 0.01 M potassium phosphate buffer (pH 7.0). Cells were then resuspended in 250 μL of the same buffer. A total of 10 μL (40 μL/mL) of EMS (ethane methyl sulfonate) was then added to each Eppendorf tube, vortexed and incubated at room temperature for 90 min. A total of 1 mL of 5% sodium thiosulphate was then added to each tube and vortexed to neutralize the EMS. The cells were pelleted, washed twice with 1-mL sterile water and finally suspended in 75 μL of sterile water. The populations were then spotted onto YPD (1% yeast extract, 2% peptone, 2% glucose, 2.5% agar) plates and incubated for ∼48 h to allow for cell growth. These plates were then replica-plated to KAc (1% potassium acetate, 0.1% yeast extract, 0.05% glucose, 2% agar) and placed at 25°C for ∼72 h to induce meiosis and sporulation.

Forty gametes from 10 tetrads from each population were dissected onto YPD plates and incubated for at least 72 h to form colonies that were then replica plated to KAc and minimal media. Because the strain contains a functional HO gene, after dividing mitotically, the newly germinated gametes should switch mating-type, and then fuse with a genetically identical cell in the same colony, forming an allotetraploid that is entirely homozygous except at the mating-type locus. To ensure that this auto-fertilization had occurred, we examined the colonies on the KAc medium replica plate to confirm that they contained tetrad ascospores—only cells that had auto-fertilized successfully, making them heterozygous at the mating-type locus, would be able to sporulate. We also examined the minimal medium replica plate to ensure that the colonies were still prototrophic. Because the ability to efficiently enter meioses and sporulate was important to allow future crosses, we selected the best-sporulating colony out of the largest three prototrophic colonies from tetrads containing four viable spores. Large colonies were selected because following mutagenesis some spores formed only very small colonies, perhaps because of mutation of both copies of important, but not essential, genes. Colonies were only taken from tetrads containing four viable spores to ensure that spores were euploid. The selected homozygous tetraploid colony was used to reinitiate the mutagenesis cycle. Thus each population went through a single-gamete bottle-neck every experimental cycle. The experiment was repeated for 12 cycles.


At the end of the experiment, populations were paired at random, and crossed. Crosses were made by spore to spore mating using a Zeiss tetrad dissection microscope. One spore from each population to be crossed was placed in contact with the other on a YPD plate using a tetrad dissection microscope. After ∼4.5 h, plates were checked for zygote formation and the position of the zygote was marked so that the colony formed could be identified later. Once a colony had formed, the hybrid was picked to KAc medium and sporulated. A sample of the strains was also frozen at −70°C in 20% glycerol. Control within-population crosses were also made by crossing two spores from the same population, and backcrosses were made by crossing a spore from each evolved population to a spore from the ancestral strain YDG 624. The spore viability (fertility) of each cross (the within-population control crosses, the “hybrid” between-population crosses, and the backcrosses to the ancestoral strain), was determined by dissecting spores from tetrads onto YPD plates and replica plating the resulting colonies to minimal media plates. Fertility (or spore viability) was defined as the proportion of all dissected spores that formed prototrophic colonies. All statistical analyses were carried out using GraphPad Prism version 5.03 for Windows (2009) or in Microsoft Excel 2007.


CHEF (Clamped Homogenous Electric Field) analysis was carried out as described by Louis and Haber (1990) and allowed the karyotypes of the evolved strains to be visualized on an agarose gel.



To confirm that each round of EMS mutagenesis was introducing null mutations to essential genes, we treated five replicate populations of the ancestral allotetraploid strain (YDG 624) and a diploid S. cerevisiae strain (YDG 199–Y-55 MATa/MATα ade2/ade2 ura3/ura3—isogenic to the S. cerevisiae portion of the allotetraploid genome—Greig et al. 2002) with the same mutagen treatment used on our experimental lines. Five control populations of each ploidy were also treated in exactly the same manner but without the addition of the mutagen. Ten tetrads from each replicate population were dissected onto YPD plates, replicated to minimal media plates and the spore viability of each line was scored (full spore viability data can be found in Table S1). As expected duplicate loci carried by the allotetraploid largely protected the strains spore viability from the effect of the null mutations introduced by EMS treatment. The spore viability of both the EMS treated (89.5%± 6.937 Standard deviation (SD)) and untreated lines (93.5%± 6.02) remaining high and highly similar (two-tailed unpaired t-test, t0.05, 8= 0.973, P= 0.359). But the EMS mutagenesis was found to greatly reduce the spore viability of the treated diploid lines (8.5%± 2.23 SD) relative to those not exposed to the mutagen (97%± 2.09 SD—Two-tailed unpaired t-test, t0.05, 8= 61.4, P≤ 0.0001). We can therefore be confident that the repeated treatment of our experimental allotetraploid lines with EMS has resulted in the accumulation of null mutations in redundant duplicate loci and that isolated populations could lose the function of different duplicate genes.


Following 12 successive experimental cycles of mutagenesis and meiosis, the spore viability and karyotypes of the 20 initially isogenic populations were assessed by tetrad dissection and CHEF gel analysis (see Methods and Fig. 3). Greig et al. (2002) previously demonstrated that tetraploid, but not diploid S. cerevisiae/S. bayanus hybrids are fertile. The spore viability of 19 of the 20 evolved populations remained high (Mean fertility = 93.02%± 2.189 SD—YDG 624 ancestor = 95%—see Table 1). The remaining population, P7, appears to have rediploidized as it contains the genomes of both species, but has very low spore viability (2.5%, see Table 1 and Fig. 3). Such rediploidization following EMS treatment was previously reported for S. cerevisiae autotetraploids (Mable and Otto 2001). Because of its low spore viability, P7 was excluded from any further analysis along with its randomly assigned partner P6.

Figure 3.

CHEF gel of evolved within-population crosses. The gel shows the electrophoretic karyotypes of S. cerevisiae (C), S. bayanus (B), the Ancestral S. cerevisiae/S. bayanus allotetraploid (A – YDG 624) and the 20 evolved within-population crosses. Chromosomes are labeled as in Fischer et al. (2000). The BOLD labeling identifies the chromosomes of S. cerevisiae whereas the italic labeling identifies the chromosomes of S. bayanus relative to S. cerevisiae. Reciprocal translocations in S. bayanus relative to S. cerevisiae are shown (i.e., IItIV represents a translocation between S. cerevisiae chromosomes II and IV). VIIΔL indicates the nonreciprocal translocation of the left arm of S. cerevisiae chromosome VII to chromosome V of in the S. bayanus lineage.

Table 1.  Spore viability of evolved within-population and between-population crosses.
PopulationSpores viable per tetrad2Total sporesPercent spore viabilityMean within-population spore viability (%)Corrected between-population cross spore viability (%)3Dunn test P-value4
  1. 1Population 7 (P7) showed poor spore viability and is most likely a hybrid diploid. This evolved population, along with P6, was not used in any further analysis.

  2. 2The number of times individual tetrads produced a given number of viable spores.

  3. 3The spore viability of each between-population cross (BPC) as a percentage of the average spore viability of the two within-population crosses (P – see main text for details). The brackets contain the percent change in relative between-population spore viability if the higher or lower spore viability of the two within-population crosses was used rather than their mean.

  4. 4Uncorrected spore viabilities of each between-population cross was compared to that of the relevant within-population cross with the lowest spore viability using a Kruskal–Wallis followed by Dunn multiple comparison tests. The P-values are shown.

YDG 624 Anc.68921032095
P148660024092.593.125  7.76 (±0.0517)<0.001
BPC 100614703607.22 7.227.22
P1648120002409595.625 33.70 (±0.2188)<0.001
BPC 20630381636032.22   
P848930024093.7593.75 77 (±0)<0.001
BPC 32335154332072.19   
P1348633024091.2593.75 97.19 (±2.524)>0.05
BPC 4602820036091.11   
P548930024092.593.125 48.32 (±0.322)<0.001
BPC 502838740045   
P248660024092.590105.09 (±2.840)>0.5
BPC 650730024094.58   
P348930024093.7593.125 76.30 (±0.510)<0.01
BPC 73317137622471.05   
P1042693024086.2587.03 25.53 (±0.182)<0.001
BPC 80022273132022.19   
P155163002409595  7.57 (±0)<0.001
BPC 90051362320 7.19   
P71000218 80 2.5   


The remaining 18 populations were crossed to their randomly assigned partner and the spore viability (fertility) of the nine between-population crosses was determined by tetrad dissection (Fig. 4 and Table 1). The mean spore viability of the nine between-population crosses (49.19%) is significantly lower than that of the 18 within-population control crosses (92.79%—Mann–Whitney U Statistic = 16.00, P= 0.0006) indicating that postzygotic reproductive isolation has developed between at least some of the populations paired.

Figure 4.

Spore viabilities of within-population and between-population crosses. The figure shows the fertility (spore viability) of each between-population F1 hybrid (BPC1–9, gray bars) as well as the fertility (spore viability) of each parent (least fertile parent = black bar, most fertile parent = white bar). Error bars show the 95% confidence interval for each fertility measurement. The left-hand Y-axis shows the percent fertility whereas the right-hand Y-axis shows the approximate number of viable spores per tetrad dissected giving the populations fertility. The significance of the fertility difference between the least fertile parent (black bar) and the F1 hybrid are also shown above each pair (Kruskal–Wallis test followed by Dunn multiple comparisons—***P < 0.001, **P < 0.01, NS = not significant). Detailed information shown in this figure can be found in Table 1.

To evaluate the significance of spore viability differences between population crosses and the least fertile of the populations used to make them (a more conservative approach than using the mean spore viability of the within-population control crosses), a Kruskal–Wallis analysis (H0.05, 17= 928.33, P= <0.001) followed by Dunn's multiple comparison tests was used. Seven of the nine between-population hybrids produced a significantly smaller proportion of viable spores than the least fertile of the relevant within-population crosses (P= <0.01 in all seven cases, see Fig. 4 and Table 1), providing further evidence that postzygotic reproductive isolation barriers have evolved between the populations.


If the divergent resolution of duplicate genes is responsible for the reduced spore viability of the between-population crosses relative to the within-population control crosses, then their spore viability should match that predicted for a specific number of divergently resolved duplicate gene pairs. The expected fertility of an F1 hybrid following divergent resolution of a specific number of duplicate loci is given by 0.75n, with n being the number of divergently resolved essential gene duplicates when, as in yeast, gametes express their genes (Werth and Windham 1991). As the parental populations (i.e., the within-population control crosses) were not 100% fertile (see Table 1) the spore viability of the between-population crosses were determined as a percentage of the within-population spore viability (see Table 1 and Fig. 5). This allowed the reduction in spore viability of each between-population cross due to divergent resolution to be separated from the other causes of reduced spore viability in the evolved populations. Mean corrected spore viabilities, as well as upper and lower corrected spore viability limits, were calculated using the spore viability of each within-population control and are shown in Figure 5 (upper and lower values were calculated when there was a difference between the with-population fertilities of the two populations that were crossed). Figure 5 also shows the predicted spore viability of crosses when between 0 and 10 essential genes are divergently resolved between the crossed populations. Seven between-population crosses have spore viabilities close to those predicted by the divergent resolution of a specific number of genes (see also Table 1).

Figure 5.

Corrected spore viabilities of between-population crosses. The figure shows the corrected fertility (percent spore viability) of each between-population cross (BPC 1–9). As the evolved populations, mated to form each between-population cross, were not 100% fertile the F1 spore viability was corrected to separate the spore viability decrease caused by incompatibilities between the evolved populations from other factors that reduced the spore viability of the individual evolved populations. Each bar shows the spore viability of the between-population cross corrected for the mean fertility of the evolved parental lines. Error bars show the spore viability of the cross when corrected using the individual parental fertilities (where differences between them existed). The figure also shows the corrected spore viability predicted when between 0 and 10 essential genes have divergently resolved between two populations (predicted spore viability = 0.75n, where n= the number of divergently resolved essential genes between the populations crossed).


Under the Dobzhansky–Muller model for the evolution of postmating reproductive isolation, new alleles that evolve within an isolated population are compatible with the ancestral genotype but incompatible with other new alleles evolving in another isolated population. We therefore tested whether the incompatibilities that had evolved between the populations also caused reproductive isolation from the ancestral population. Each of the 18 evolved populations was crossed back to the ancestral genotype and the spore viability of the resulting backcross strains were recorded. Each of the backcross strains had a high spore viability (Mean = 93.565%± 1.94 SD, Table 2), as expected if the incompatible alleles had evolved by divergent resolution of redundant duplicated genes.

Table 2.  Fertility of backcrosses of evolved populations to ancestor.
Population hybrid1Strains2Viable gametes per tetrad3Total sporesPercent fertility BC vs. BPC Dunn test P-value4 BC vs. Anc. Dunn test P-value4
  1. 1The identity of the between-population cross (BPC) formed by the evolved populations that were backcrossed to the ancestral strain (Anc. – YGD 624). The spore viability data for the between-population hybrids can be found in Table 1.

  2. 2The spore viability of the ancestral (Anc.) strain (YDG 624) and each evolved population backcrossed to this ancestral genotype (BC = Backcrossed).

  3. 3The number of tetrads dissected that produced a given number of viable gametes.

  4. 4The significance of fertility differences between the population hybrids (BPC) and their backcrossed parental lines (BC) and between the ancestral genotype and each backcrossed population was assessed by using a Kruskal–Wallis analysis (H0.05, 27= 1274, P= <0.001) followed by Dunn multiple comparison tests. The P-value of each comparison is shown.

Anc.68 921032095 – –
BPC 1BC 1471030024093.33<0.001>0.05
 BC 948 903024092.5<0.001>0.05
BPC 2BC 16461130024092.92<0.001>0.05
 BC 2054 330024096.25<0.001>0.05
BPC 3BC 847 940024092.92<0.001>0.05
 BC 1150 730024094.58<0.001>0.05
BPC 4BC 1352 440024095>0.05>0.05
 BC 1850 460024093.33>0.05>0.05
BPC 5BC 551 810024095.83<0.001>0.05
 BC 17421620024091.67<0.001>0.05
BPC 6BC 251 450024094.17>0.05>0.05
 BC 451 540024094.58>0.05>0.05
BPC 7BC 3312900024087.92>0.05>0.05
 BC 1448 390024091.25<0.01>0.05
BPC 8BC 10481110024094.58<0.001>0.05
 BC 1248 930024093.75<0.001>0.05
BPC 9BC 1550 730024094.58<0.001>0.05
 BC 1951 630024095<0.001>0.05



Seven of the nine between-population crosses produced significantly fewer viable spores (were less fertile) than the within-population control crosses. In addition, each of the between-population crosses has a spore viability close to that predicted by the divergent resolution of a specific number of essential genes. Each individual in the evolved populations carries a complete complement of essential genes and therefore within-population crosses produce gametes that carry functional copies of all genes essential for growth. But crosses between two individuals from populations that have lost different duplicate copies produce some gametes that lack a functional copy of an essential gene, and are therefore inviable.

If the loss of a particular duplicate gene copy occurs at random between populations, it would be expected that there be variation in the spore viability of the between-population crosses and this was found to be the case. The spore viability of the hybrids ranges from the levels expected for zero (∼100% corrected fertility) to nine (∼7.5% corrected fertility) divergently resolved genes. Our repeatedly mutagenized lines accumulated more and more null mutations in redundant loci encoding essential genes, allowing for different copies to lose function in the isolated populations. The control mutagenesis of the S. cerevisiae diploid (YDG199) reduced spore viability to an average of 8.5%, indicating that on an average three to four essential genes lost their function in each diploid exposed to EMS (0.53= 12.5%; 0.54= 6.25%;). Giaever et al. (2002) previously determined that 1106 genes are essential for S. cerevisiae growth in rich YPD media whereas a further 98 are known to be essential for growth on minimal media (i.e., those involved in amino-acid biosynthesis—Saccharomyces genome database—, a total of 1199 “minimal media essential” genes. Because the genome of our allotetraploid is doubled, we would expect the 2398 duplicated essential genes to experience between six and eight loss-of-function mutations each experimental cycle, accumulating between 72 and 96 such mutations (eight (or six) mutations × 12 cycles = 96 (or 72)) by the end. We would therefore expect the number of divergently resolved genes between any pair of populations to be between approximately 3.84 (96/2398 × 96) and 2.16 (72/2398 × 72). The mean fertility of the nine between-population crosses was 49.19%, so on average two (0.752= 56%) or three (0.753= 42%) essential loci were divergently resolved, fitting our expectation. However two crosses (see Fig. 4) had spore viabilities of less than 8%, corresponding to nine (0.759= 7.5%) divergently resolved genes, far higher than expected. One possibility is that mutagenesis can sometimes cause loss of DNA repair functions, producing mutator phenotypes that greatly increase the mutation rate. Another explanation is that the mutagenesis and autofertilization cycle occasionally fixed small chromosomal deletions containing many essential genes, a possibility supported by the observation that essential genes tend to be clustered in groups (Pal and Hurst 2003). Alternatively the mutagenesis we applied to our lines may not be completely random. EMS mutagenesis is known to produce GC:AT mutations by reacting with guanine residues. This reaction often causes DNA polymerases to incorrectly place a thiamine residue opposite to the affected guanine and subsequent rounds of replication can result in the GC:AT mutation (Mable and Otto 2001). It follows that genes with higher GC content will be preferentially targeted during the course of our experiment, effectively increasing the likelihood of null mutations in certain genes. With certain genes more likely to accumulate null mutations than others, the likelihood of two populations undergoing divergent resolution is increased above the level expected if mutations can arise anywhere within the genome.

Following successive rounds of mutagenesis, 19 of the 20 within-population crosses remained highly fertile (93.02%± 2.189 SD), and still contained the genomes of both species (See Fig. 3). They are likely, therefore, to have remained tetraploid, because diploid S. cerevisiae/S. bayanus hybrids are sterile, producing only ∼1% viable, often aneuploid, gametes (Hunter et al. 1996; Greig et al. 2002). Population 7 appears to have rediploidized (it has low spore viability but chromosomes from both species are present) indicating that the loss of chromosomes can take place following repeated EMS treatment. Mable and Otto (2001) witnessed similar rediploidization following EMS mutagenesis of an autotetraploid S. cerevisiae strain and prolonged maintenance of the same autotetraploid strain without application of mutagen also showed rediploidization (Gerstein et al. 2006) although their later discovery that the founding strain was in fact aneuploid may have been an important contributing factor (Gerstein et al. 2008).


Although we have shown that the populations have probably remained allotetraploid, we cannot exclude the possibility that populations have lost some chromosomes. Between-population crosses could have reduced spore viability if one population had lost the S. cerevisiae copy of a chromosome whereas the population with which it was crossed had independently lost the S. bayanus copy of the same chromosome. Individually, the populations would be viable but during hybrid meiosis the chromosomes from the different species would be unable to pair and segregate correctly, producing inviable unbalanced gametes just as in diploid F1 hybrids of the two species (Hunter et al. 1996). Each chromosome reciprocally lost like this would reduce spore viability by 25%. This potential explanation for the reduced spore viability can be ruled out for most chromosomes, as both species’ chromosomes are clearly visible in CHEF gels of the individual within-population crosses (Fig. 3). But because a few (particularly the larger) chromosomes from the different species migrate to the same position on the gel they are difficult to distinguish, so this explanation cannot be completely excluded. However, the reduction in spore viability of many of the between-population crosses is so great that it would require the differential loss of several different chromosomes, which could not go undetected.

Second, instead of the complete loss of both copies of a chromosome from a particular species described above, it is possible that only one of the copies is lost from the evolved strains genome, producing trisomy for certain chromosomes. For example, if an evolved strain carried two copies of S. cerevisiae chromosome I and a single S. bayanus chromosome I (trisomy for chromosome I), this would not be visible on the CHEF gel because both species’ copies of chromosome I are present. Could trisomy explain the reproductive isolation between evolved populations? A single copy of a chromosome within an evolved tetraploid would have no partner at meiosis whereas the remaining diploid chromosome pair from the other species would pair and segregate correctly into the gametes. Thus meiosis would produce gametes that are viable, as they have all received a copy from the correctly segregating pair, but only two will also carry the other species’ chromosome. If another evolved line lost a single copy of the other species’ chromosome (e.g., if another strain carried two copies of S. bayanus chromosome I and but a single S. cerevisiae chromosome I) then half the hybrid zygotes formed between these two populations would contain only a single pair of heterologous chromosomes I, one from each species. This hybrid pair of chromosomes would not segregate properly at meiosis, and would reduce fertility of the between-population cross. However, we think that such trisomy is unlikely to explain the reproductive isolation between populations that we observe. The repeated auto-fertilization of gametes used following each EMS cycle to produce homozygous strains would either repeatedly reestablish a tetraploid chromosome complement (if the auto-fertilized gamete contained a copy of the chromosome from both species) or a diploid chromosome pair (if the auto-fertilized gamete contained only one species’ copy of a given chromosome). Because of this, trisomy could occur only transiently in the parental populations, and the presence of both species’ chromosome versions in the evolved populations’ genomes (Fig. 3) shows that they have remained tetraploid rather than diploidized. The likelihood of two paired populations being transiently trisomic for the same chromosome, but with a different species’ copy lost is likely to be low (one could estimate it as the trisomy rate multiplied by 1/16 multiplied by 1/2). Furthermore, if trisomy did occur within parental lines in addition to the accumulation of null mutations in duplicate essential genes, the spore viability of the within-population crosses would also be reduced, which was not observed. This is because, if the species’ chromosome of which there is only one copy carried the only functional copy of an essential gene, half of the gametes produced by the within-population crosses would be inviable as they lack the chromosome and therefore a copy of the essential gene. Such reduced fertility of the evolved parental lines was not observed despite the fact that EMS-induced null mutations were known to accumulate. For all these reasons, therefore, we think that whole chromosome loss cannot explain the evolution of reproductive isolation we observe.

Chromosomal rearrangements could potentially cause reproductive isolation between our populations. Translocations can be excluded, as they would be visible in our CHEF gel analysis. But inversions could not be detected because the inversion of a chromosomal segment does not change its length. But we think that inversions are unlikely to have occurred widely, because we saw no translocations at all, and EMS induces G:C to A:T mutations, not rearrangements (Mable and Otto 2001). Also, heterozygozity for an inversion can only result in reduced spore viability if a very unlikely crossover occurs within the inverted region during meiosis, producing unbalanced gametes that may lack a copy of an essential gene. Furthermore, in an allotetraploid (such as our between-population crosses) the homeologous chromosome pair, which is unlikely to carry a heterozygous inversion in same region, will stop such gene loss from affecting spore viability. For example, if a between-population cross was heterozygous for an inversion on S. cerevisiae chromosome III, a rare crossover within the S. cerevisiae inverted region could still result in loss of S. cerevisiae copies of essential genes in the inverted region. But spore viability would remain high as the noninverted S. bayanus version of chromosome III carried by the allotetraploid would behave normally during meiosis, ensuring all gametes receive at least one copy of any essential genes lost from the S. cerevisiae chromosome. If rare inversions had fixed within a population during the course of our experiment, the homeologous chromosome pair masks their potential to affect the spore viability of between-population crosses. Although we cannot be certain that gross chromosomal changes have not occurred, we do not think they are likely or sufficient to explain the strong reproductive isolation that evolved.


The conditions used in these experiments were designed to promote loss of genetic redundancy by loss-of-function mutations. The rate of gene loss and the fixation of null mutations within populations was greatly increased by random mutagenesis, repeated bottlenecking of the population size down to a single individual, and restoration of full homozygosity each cycle by auto-fertilization. Selection was minimized relative to drift, but by selecting a single homozygous individual in each cycle based on colony size and sporulation ability, artificial selection was applied which may have improved these traits. It is therefore possible that the genetic incompatibilities causing reproductive isolation in our experiment are not caused by genetic drift of divergently resolved, selectively neutral, loss-of-function mutations, but by artificial selection of mutations changing the function of redundant genes. Even simple deletions eliminating redundant genes may actually confer small increases in fitness by reducing the metabolic cost of unnecessary genome expression and maintenance. It is noteworthy that yeast has a very compact genome, presumably the result of this type of selection for metabolic efficiency.

We do not think that the reproductive isolation that emerged in our experiment was due to preexisting genetic incompatibilities between S. cerevisiae and S. bayanus. Dominant incompatibilities do not cause F1 hybrid sterility between the sensu stricto yeasts (Greig et al. 2002), and to date only three recessive cytonuclear incompatibilities have been identified between S. cerevisiae and S. bayanus (Lee et al. 2008; Chou et al. 2010). These incompatibilities, between the S. cerevisiae mitochondria and the S. bayanus nuclear gene AEP2 and between the S. cerevisiae nuclear genes MRS1 and AIM22 and the S. bayanus mitochondria, do not cause F1 hybrid sterility, but instead make F2 hybrids unable to respire, so they can only grow by fermentation and are therefore unable to undergo meiosis. The apparent absence of any nuclear–nuclear incompatibilities (Greig 2007; Lee et al. 2008) between the Saccharomyces sensu stricto species means we can rule out the novel expression of preexisting genetic incompatibilities as a likely cause of the reproductive isolation between our experimental strains, which was caused by hybrid spore inviability, not respiration defects.

But although the founding allotetraploid used in this work is viable and fertile, without any sterilizing incompatibilities, it is possible that weaker incompatibilities may have affected the outcome of our experiment. Recent work has demonstrated that the replacement of individual S. cerevisiae chromosomes with their S. bayanus homolog results not in complete inviability, but in reduced fitness (Lee et al. 2008), and we have found similar reductions in fitness when chromosomes from the more closely related species S. paradoxus are used to replace S. cerevisiae chromosomes (C. J. Maclean and D. Greig, unpubl. data). Allotetraploids carry two complete sets of such coevolved chromosomes, whose genes may work well with genes of the same species, but not so well with those from the other species. The loss of function at one or other redundant locus may not, therefore, occur at random, because the retention (or loss) of a particular gene copy may be favored as it provides a greater fitness advantage to the evolving allotetraploid. This would cause gene loss to occur in parallel between the populations, rather than at random, reducing the chance of divergent resolution

On the other hand, it is also possible that selection for coadapted gene complexes could strengthen the effect of divergent resolution. If one gene in a complex is divergently resolved by chance, then selection may also select for the maintenance of other genes from the same species that make up the co-adapted complex. This will result in all such genes being divergently resolved, strengthening the reproductive isolation due to reciprocal loss of function. In our experiment, identical selective conditions were applied to all evolving lines, but if different selection was applied to different lines, it might directly favor the maintenance of coadapted gene complexes from one species over another, which would increase the likelihood of divergent resolution. A natural progression from the experiments presented here is to determine how increased or decreased selection, relative to drift, affects the evolution of reproductive isolation.


Although the relative importance of divergent resolution as a general speciation process has been questioned (Coyne and Orr 2004) the widespread occurrence of WGD makes it difficult dismiss completely, even if it acts only to increase the reproductive isolation between two populations already partially isolated by some other barrier to gene flow. Because of this, divergent resolution may be more important in establishing reproductive isolation between initially isogenic allotetraploids that are adapting to different conditions, particularly when the interacting genes from one species are better suited for one set of conditions than the other. In this sense, the divergent resolution of duplicate genes may be important in accelerating the accumulation of reproductive isolation.

A criticism of the importance of divergent resolution as a general mode of speciation is that it is most effective in organisms whose gametes express their genomes, and therefore require a full complement of functional genes (Lynch and Force 2000; Coyne and Orr 2004). This same criticism also applies to other forms of recessive Dobzhansky–Muller incompatibility, which can sterilize heterozygous F1 hybrids only if the recessive incompatibility is expressed in the hemizygous gametes they produce. In organisms without gametic genome expression, recessive incompatibilities such as divergently resolved genes cannot affect gamete survival because they are not expressed, but they will affect the resulting homozygous F2 zygotes. Therefore 1/16th of the F2 generation will be inviable or sterile for each divergently resolved essential gene (0.9375n F2s will not be affected, with n denoting the number of essential divergently resolved genes). However, the need for more than one functional copy of a gene (haploinsufficiency) could result in a higher rate of F2 zygote death or sterility, as a further one-fourth of zygotes receive only one functional gene copy (Lynch and Force 2000). Divergent resolution is likely to be most important as a mechanism for reproductive isolation in organisms that are commonly polyploid and have gametic gene expression, such as plants, fungi, and algae (Werth and Windham 1991; Xu et al. 1999; Lynch and Force 2000).

The increasing availability of whole-genome data for closely related species, including nontraditional model species, is allowing the mechanisms that have shaped modern genomes to be slowly revealed. WGD and the genome reorganization that follows is likely to have significantly contributed to the species richness seen today. In this work, we have demonstrated the feasibility of a mechanism that can produce reproductive isolation, a process that in natural population would likely take a great deal of time. Further laboratory-based studies, in combination with more traditional genomic analyses of a broader range of species, will likely increase our understanding of the mechanisms that underlie speciation.

Associate Editor: P. Turner