An Evolutionary Advantage of Haploidy in Large Yeast Populations

See allHide authors and affiliations

Science  24 Jan 2003:
Vol. 299, Issue 5606, pp. 555-558
DOI: 10.1126/science.1078417


Although seed plants and multicellular animals are predominantly diploid, the prominence of diploidy varies greatly among eukaryote life cycles, and no general evolutionary advantage of diploidy has been demonstrated. By doubling the copy number of each gene, diploidy may increase the rate at which adaptive mutations are produced. However, models suggest that this does not necessarily accelerate adaptation by diploid populations. We tested model predictions regarding rates of adaptation using asexual yeast populations. Adaptive mutations were on average partially recessive. As predicted, diploidy slowed adaptation by large populations but not by small populations.

The evolutionary origins of diploidy, sex, and recombination, and their subsequent long-term effects, are often intertwined, both in evolutionary theory and in nature (1–3). The masking of recessive deleterious mutations (4–7) and the more frequent production of adaptive mutations (8–10) have both been hypothesized as evolutionary advantages of diploidy. We focused on the rates of adaptation in haploid and diploid yeast populations. In asexual populations, adaptation is typically thought to occur in a series of selective sweeps, each composed of two phases: a waiting period, representing the number of generations before the occurrence of the next adaptive mutation that escapes loss by genetic drift, and a fixation time, which is the number of generations required for selection to fix that mutation in the population. Diploidy may reduce the waiting period by doubling the number of alleles from which the next adaptive mutation could arise (10). However, for any mutation that is not dominant and arises in a heterozygote, diploidy also increases the fixation time, because such mutations confer smaller selective advantages on heterozygotes than on haploids. Which of these two effects is more important depends on population size. If a population is large enough to produce one or more adaptive mutations in an average generation, then the fixation time and not the waiting period limits the rate of adaptation, and haploids should evolve faster (9). More specifically, if the product of population size (N) and the rate at which adaptive mutations arise (v) is greater than 1, then diploids will adapt more slowly by a factor approximately equal to the dominance coefficient h (11).

We tested these predictions using haploid and diploid populations ofSaccharomyces cerevisiae (12). Five populations of each type were propagated through 2000 generations by serial transfer in a minimal liquid medium (12). Dextrose was the limiting nutrient for the founding genotype, and adaptation occurred as mutant genotypes with higher rates of growth and survival in this environment arose and increased in frequency. At intervals of 200 to 300 generations, the relative fitness of each population was estimated from competitions with a neutrally marked version of the diploid ancestor (12). The starting populations were genetically homogeneous, identical except for ploidy, and had the same fitness, population sizes, and generation times (12). Therefore, any consistent evolved differences can be attributed to the effects of ploidy on the fixation of adaptive mutations.

The prediction of faster adaptation by haploids was strongly supported. Major fitness increases were observed in all 10 populations (Fig. 1A) and adaptation was significantly faster in haploids than in diploids (Fig. 2). The average fitness increase in diploids relative to that in haploids after 2000 generations was 0.557/0.802, or 0.69. These 10 populations were subsequently propagated through 5000 generations, after which diploid and haploid populations had converged in fitness despite their earlier divergence (the mean of five replicates ± the standard deviation for haploids and diploids is 1.86 ± 0.073 and 1.81 ± 0.045, respectively; P = 0.25 in a two-tailed heteroscedastict test). Therefore, the difference in rates of adaptation over the first 2000 generations does not appear to result from a difference between ploidies in their potential to adapt to the experimental environment.

Figure 1

Fitness trajectories of (A) large populations (N e = 1.3 × 107) and (B) small populations (N e = 1.4 × 104). Circles and error bars represent means and 95% confidence intervals of fitness for five replicate haploid (open circles and solid lines) and diploid (solid circles and dashed lines) populations. Lines are regression lines of fitness on generation number.

Figure 2

Rates of adaptation in haploid and diploid, large and small populations. Bars represent slopes estimated by linear regression of fitness on the generation number, with the five replicate populations of each size and ploidy pooled for four regressions. Error bars are 95% confidence limits. Regressions were also performed separately for each of the 20 populations. For each population size, the slopes estimated for haploid and diploid populations were compared using two-tailed heteroscedastic t tests, which indicated a highly significant effect of ploidy for large populations (P < 0.001) but not for small populations (P = 0.35).

The relative rates of adaptation in diploids and haploids are expected to depend on population size. In small populations, the advantage of haploidy should be reduced or eliminated because the supply of adaptive mutations, rather than the time taken to fix them, becomes the factor limiting the rate of adaptation. The doubling of adaptive mutation rates presumably conferred by diploidy therefore becomes more important in small populations. When adaptive mutations are rare (Nv≪ 1), the rate of adaptation by diploid populations relative to that in haploid populations approaches 2h (9). Therefore, in small populations, diploidy is an advantage if adaptive mutations are dominant and a disadvantage if they are recessive, and it does not affect rates of adaptation if mutations are additive.

To test the prediction that diploids fare relatively better with smaller population sizes, we repeated the evolution experiment with much smaller populations [effective population sizeN e ≈ 1.4 × 104(12)]. As expected, small populations adapted much more slowly than large populations (Figs. 1B and 2). The average fitness increase in diploids relative to that in haploids after 2000 generations was 0.973, and haploids and diploids adapted at nearly identical rates (Fig. 2). An analysis of variance (ANOVA) of rates of adaptation in large and small populations confirmed significant effects of ploidy (F 1, 16 = 41.93,P < 0.001), population size (F 1, 16 = 1113.29,P < 0.001), and their interaction (F 1, 16 = 24.26, P < 0.001). This is not simply because slower adaptation precluded significant fitness divergence among the small populations: An F ratio test of variance in fitness indicated that fitness after 2000 generations was not significantly more variable among large populations than among small populations (F 9, 9 = 3.06,P = 0.09).

Because asexually propagated yeast tend to change ploidy or become aneuploid, the DNA content of the ancestral and evolved populations were compared by flow cytometry. One large population and four small populations switched ploidies (12). This is almost certainly due to the origin and increase in frequency of new ploidy mutants within these populations rather than to contamination among populations (12). Although ploidy changes complicate quantitative comparisons between theoretical predictions and results from the small populations, their occurrence is consistent with the predicted effects of ploidy on rates of adaptation. Among large populations, where haploids adapted faster, the only ploidy change involved one initially diploid population becoming haploid. In contrast, two switches from haploidy to diploidy and two switches from diploidy to haploidy occurred among small populations, where ploidy had no detectable effect on rates of adaptation. Repeating the ANOVA of rates of adaptation with populations classified by their final rather than initial ploidy did not alter the conclusion that ploidy, population size, and their interaction all had significant effects (P = 0.002,P < 0.001, and P = 0.006, respectively).

Our results appear to contradict a previous report that adaptive mutations are fixed more frequently in diploid than in haploid yeast populations (10). One difference between the experiments is that we used serially transferred cultures in which selection is largely density independent, whereas Paquin and Adams (10) observed density-dependent adaptation in chemostats, which may select for a different set of mutations. Also, rather than directly estimating fitness gains by evolved populations, Paquin and Adams obtained estimates of the numbers of adaptive mutations that arose in each population, which may not have indicated the number that were fixed or the cumulative fitness increases (13).

Our results are in qualitative agreement with the main prediction of Orr and Otto's (9) model: With large populations, adaptation occurs more rapidly in haploid than in diploid populations, and this haploid advantage is lost at much smaller population sizes. A quantitative comparison between theory and our observations requires estimates of h and Nv and acceptance of the model assumptions that neither h nor the selection coefficients varies among adaptive mutations. In reality, adaptive mutations must vary in both parameters, and their distributions and covariance probably affect relative rates of adaptation. We estimated the average value of h for the suite of adaptive mutations present in one haploid and one diploid genotype randomly chosen from large 2000-generation populations. One potential complication is that mutations that are adaptive in haploids may not be adaptive in diploids, or vice versa, because the differences in gene regulation between haploid and diploid yeast may select for different types of mutation. To test for this possibility, we performed fitness estimates on haploids, homozygous diploids, and heterozygous diploids derived from each of the two evolved genotypes (12). There was no indication that the adaptive effects of mutations were restricted to the ploidy background on which they were selected; in fact, at least some mutations selected in the diploid were maladaptive in homozygotes but adaptive in both heterozygotes and haploids (Fig. 3). Mean dominance coefficients for the mutations in the haploid and diploid genotypes were 0.20 and 0.75, respectively. One or more of the mutations selected in the diploid evolved genotype appear to be overdominant (h > 1).

Figure 3

Fitness estimates for haploid, homozygous diploid, and heterozygous diploid strains derived from (A) a haploid genotype and (B) a diploid genotype, sampled from populations after 2000 generations of adaptation. Bars indicate means of 10 replicate estimates, and error bars are 95% confidence limits.

The greater dominance of mutations selected in diploids is an intuitively logical outcome of selection acting on mutations that vary in dominance, but it is difficult to interpret this comparison without a known pattern of covariance between s and h for adaptive mutations. For example, s and h may be negatively correlated (mutations conferring greater fitness advantage may be more recessive). The current lack of theory that incorporates variation in s and h in modeling the effect of ploidy on adaptation and our empirical ignorance of these parameters for adaptive mutations point to the need for more work in both areas. Finally, in predicting relative rates of adaptation (11), we assume that adaptive mutations remained heterozygous in diploids, because all populations were asexual. However, loss of heterozygosity due to processes such as gene conversion and mitotic crossing-over occurs at appreciable rates in asexual diploid yeast (14), and it is possible that some diploids became homozygous for adaptive mutations or that some mutations were lost by gene conversion to the wild-type allele.

The results presented here support a simple theoretical prediction: Haploid populations adapt faster than diploid populations when both are large, but not when both are small. A greater rate of adaptation is not a general consequence of diploidy and cannot by itself explain the prominence of the diploid phase in plant and animal life cycles.

Supporting Online Material

Materials and Methods

References and Notes

  • * To whom correspondence should be addressed. E-mail: zeylcw{at}

  • Present address: Population Biology, Ecology and Evolution Program, Emory University, Yerkes Primate Research Center, 954 Gatewood Drive, Atlanta, GA 30329, USA.

  • Present address: Immunology and Infectious Diseases, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115–6018, USA.


View Abstract

Stay Connected to Science

Navigate This Article