The existence of sexual reproduction poses a conundrum because it is more costly than asexual reproduction, and over 100 years of thought has been directed toward understanding why sex is maintained [1–7]. The general idea that sex increases the efficacy of natural selection (the Weismann hypothesis) [1, 4, 8] is supported by an increasing amount of experimental data (reviewed in [9, 10]). However, there are fewer experimental studies that allow us to decipher the mechanisms underlying the advantage of sex.
One long-standing leading theoretical explanation, based on Fisher and Muller's ideas (F-M), suggests that sex allows multiple beneficial mutations to simultaneously permeate populations, resulting in populations achieving combinations of beneficial mutations more rapidly than asexual populations. In contrast, asexual populations must rely on the slower stepwise fixation of individual mutations, which may hinder one another's spread through clonal interference when in different genomes [11, 12]. Studies with a diversity of model systems have shown sexual populations to increase in fitness more rapidly than asexual ones, and the general inference is that this is a result of the F-M effect [9, 13–20]. Some studies have provided a little more understanding of the mechanisms behind sex's advantage. Poon and Chao  demonstrated that recombination has a greater effect on fitness when genetic drift is larger, presumably since here negative linkage disequilibrium between beneficial mutations is greater, and thus sex serves to more effectively concentrate adaptive alleles. Work with Chlamydomonas  also supports the F-M hypothesis as it showed that sex had a greater effect on fitness gain in larger than smaller populations: in larger populations there are more likely to be multiple beneficial alleles segregating and thus sex may more effectively bring these together. Studies directly demonstrating that recombination speeds the fixation of beneficial mutations are rarer, but this has been shown in Drosophila  and bacterial models . In sum, a number of studies, with a diversity of systems, suggest that sex and recombination serve to increase the rate of adaptation under a variety of situations, and this is presumed to be by more effectively bringing together beneficial mutations.
Whilst attractive, the downfall of ideas oriented around beneficial mutations is that any advantage to sex vanishes in the absence of directional selection. There is evidence to suggest that environmental stasis, where purifying rather than directional selection is important, may be the more common state in nature . All populations are subject to detrimental mutations, whether they are adapting or not, and thus detrimental mutation clearance theories are potentially universally applicable. A stochastic based theory concerning the effect of sex on mutation clearance in small populations was originally proposed by Muller , and this idea has some experimental support . The more general mutational deterministic (MD) hypothesis predicts that sex may be maintained since it serves to more effectively purge detrimental mutations in populations of any size . Two conditions are needed for the MD hypothesis to counter the two-fold cost it theoretically imposes: 1, that the per-genome per-generation detrimental mutation rate (U) is above one; and 2, that detrimental mutations interact with negative epistasis . These conditions seem rarely met in nature. Most, but not all, organisms surveyed have detrimental mutation rates below one . The evidence for how detrimental mutations interact is less clear. Work with Chlamydomonas  and insects  suggest negative epistasis, but work with E. coli  and Drosophila  suggest approximately equal frequencies of positive and negative epistatic interactions. Experiments assessing the effect of sex on mutation clearance are fewer. A study by Zeyl and Bell  with Saccharomyces cerevisiae populations suggested that sex served to more effectively clear detrimental mutations than accumulate beneficial ones. In contrast, work with sexual and asexual yeast populations showed no difference in fitness under purifying selection , and Renaut et al.  examined the fate of sexual and asexual Chlamydomonas populations propagated under purifying selection and also found no evidence that sex more effectively cleared detrimental mutations. However, microbial detrimental mutation rates are very low (U ≤ 0.001) [29, 35, 36] and so if the MD idea applies one might not have expected to see a measurable difference in equilibrium fitness between sexuals and asexuals in these experiments.
On balance it seems that beneficial mutation assemblage, rather than detrimental mutation removal, is the stronger evolutionary mechanism underlying the advantage of sex. However, since this is still not clear, it is of interest to attempt specific tests to disentangle the contributions of these processes. Theoretically, the benefit of sexual reproduction may be due to the simultaneous actions of both the F-M and MD mechanisms, but there have been few experimental tests for this. In asexual populations, any detrimental mutations linked to beneficial mutations can be expected to increase in frequency by "hitchhiking" along with them, so long as net genotype fitness positive . This is predicted to increase the rate of molecular evolution, but lower the rate of adaptive evolution as a result of the decrease in effective population size due to the Hill-Robertson effect . A similar argument has also been made that in asexual populations beneficial mutation spread may be suppressed since they likely reside in genomes that become increasingly full of "rubbish"; however, sex may liberate these beneficial mutations . These ideas predict that asexual populations will show a tapering off in fitness gain under directional selection compared with sexual populations as beneficial mutations are not as effectively unlinked from increasingly detrimental backgrounds. This effect will be amplified under greater mutation pressure. Recently Morran et al.  showed that outcrossed nematode populations with increased mutation rates were able to adapt to a novel environment more rapidly than inbred populations, and they inferred that outcrossing both speeds adaptation and impedes the fixation of detrimental mutations compared with inbreeding. One study with S. cerevisiae manipulated mutation rates and compared the effect of sex in static and fluctuating environments , but we are unaware of experiments that have manipulated mutation pressure and compared sexual and asexual populations evolving under differing strengths of purifying and directional selection. Currently an experimenter may manipulate sexual status, mutation rate and environment of selection. However, the ideal experiment would also manipulate the types of mutations present in populations, and compare the effects of sex on the clearance of detrimental mutations and incorporation of beneficial mutations. We are far from having a large enough list of alleles with known differing fitness effects, even for the best characterized of model organisms. Even if there were a comprehensive list, the construction of starting populations with defined suites of characterized alleles, let alone tracking their change, is technically daunting.
Eukaryotic microbes present good model systems with which to attempt to test such questions as their mode of reproduction, mutation rates, and environment of growth may be easily manipulated. Yeast divides mitotically when supplied with sufficient nutrients, but starvation induces meiosis (sporulation) in diploids resulting in four haploid recombined spores. Each spore may be one of two mating types (a or α) as defined at a single Mendelian locus, and spores of the opposite mating type may mate once germinated. As meiosis in yeast and other microbial model systems is normally manipulated by starvation, this means that asexual and sexual replicates will experience different selection regimes. Moreover, starvation in both Chlamydomonas and S. cerevisiae is known to increase mutation rates [41, 42], which will tend to increase genetic variation. Thus, these manipulations do not just have effects on the mode of reproduction but also on aspects that may alter the course of evolution independent of sexual reproduction. We employ a S. cerevisiae system that circumvents this issue as sexual and asexual populations experience identical conditions, including starvation, and differ only in their ability to engage in recombination, random assortment and syngamy [10, 15]. Two genes required for normal recombination and meiosis were deleted to create the asexual sporulating strain used here. SPO11 encodes an endonuclease that initiates cross-over events by making double strand breaks in chromosomes: in its absence meiotic recombination does not occur . SPO13 determines whether a cell goes through one or two meiotic divisions by altering the sister chromatid cohesion process ; in its absence only the second non-reductive meiotic division is achieved, and this results in the production of two diploid, as opposed to four haploid, spores. Because chiasmata are required to stabilize chromosome segregation, non-functional mutations of SPO11 would normally lead to aberrant chromosomal segregation, but this phenotype is rescued if SPO13 is non-functional as well, and leaves the asexual diploid double mutant fully fertile, producing diploid spores that are genetically identical to the parent cell [15, 45]. The mitotic fitness effect of deleting these genes appears insignificant, and sporulation rates of sexual and asexual strains are equivalent .
Previous experiments with this yeast system showed that sex conferred a greater rate of adaptation to a stressful environment, but that sex had no effect on fitness in a permissive environment . S. cerevisiae's native U is very small at around 0.001 [36, 46, 47] and thus the genetic load imposed by any detrimental mutations is negligible; consequently this experiment was unable to evaluate the effects of sex on the clearance of detrimental mutations [36, 48]. Here we have additionally elevated the mutation rate by deleting MSH2, a gene involved in DNA mismatch repair , in an attempt to increase the extent of genetic load experience by these populations.
With this yeast system comprising sexual and asexual, wild-type (WT) and mutator populations we attempt a test of the 'strict F-M' hypothesis. This states that sex functions solely to increase the efficacy with which beneficial mutations are incorporated into populations, but that sex has no bearing on the efficacy with which selection operates on detrimental mutations. We are not yet in a position to manipulate the suite of mutations that arise, and thus we instead controlled the environment of selection. We constructed two environments by manipulating osmotic and thermal stresses: a 'permissive' environment at 30°C with 0 M NaCl; and a 'stressful' environment at 37°C and 0.2 M NaCl. We suggest that that purifying selection predominantly operates in the permissive environment, and this serves to remove detrimental mutations as they arise. In the more stressful environment we suggest a greater strength of directional selection will be operating, which will serve to incorporate beneficial mutations. Of course, each environment will not be absolute in the type of selection imposed - there will likely be both types of selection in both environments. In addition, there is evidence from other yeast populations, which shows that the load of detrimental mutations will likely be enhanced in stressful environments [49, 50]. We assume that directional selection is relatively stronger in the stressful environment, but that purifying selection is still important here. Under these assumptions we test the predictions for changes in fitness in environments intended to impose different types and strengths of selection pressures, with populations of varying sexual status and mutation rates, and then infer the actions of sex as related to beneficial or detrimental mutations. The first prediction arising from the strict F-M hypothesis is that under permissive conditions, where purifying selection is likely more important, sex will have no significant effect on fitness regardless of the magnitude of mutation pressure. The second prediction arising from this hypothesis is that under stressful conditions, where directional and purifying selection are important, sexual populations will display greater rates of adaptation compared with asexuals, and that an elevated mutation rate will effect the difference in rate of adaptation between sexuals and asexuals only if the supply of beneficial mutations is limiting. We evolved WT and elevated mutation rate sexual and asexual yeast populations in stressful and permissive environments in order to test the two predictions arising from the strict F-M hypothesis.