There is currently limited empirical and theoretical support for the prevailing view that adaptation typically results from the fixation of many mutations, each with small phenotypic effects. Recent theoretical work suggests that, on the contrary, most of the phenotypic change during an episode of adaptation can result from the selection of a few mutations with relatively large effects. I studied the genetics of adaptation by populations of budding yeast to a culture regime of daily hundredfold dilution and transfer in a glucose-limited minimal liquid medium. A single haploid genotype isolated after 2000 generations showed a 76% fitness increase over its ancestor. This evolved haploid was crossed with its ancestor, and tetrad dissections were used to isolate a complete series of six meiotic tetrads. The Castle-Wright estimator of the number of loci at which adaptive mutations had been selected, modified to account for linkage and variation among mutations in the size of their effect, is 4.4. The estimate for a second haploid genotype, isolated from a separate population and with a fitness gain of 60%, was 2.7 loci. Backcrosses to the ancestor with the first evolved genotype support the inference that adaptation resulted primarily from two to five mutations. These backcrosses also indicated that deleterious mutations had hitchhiked with adaptive mutations in this evolved genotype.

---

SOME of the most basic questions of evolutionary genetics concern the way that adaptations are constructed from beneficial mutations. Are adaptive traits typically the product of many mutations, each encoding a small phenotypic effect, or can most adaptation be attributed primarily to a few mutations, each with major effects? The former view was established by FISHER (1930), using a geometric analogy in which a population begins the adaptive process at some hyperdimensional distance from a central point that represents the optimal phenotype. Random mutations whose phenotypic effects are small relative to the distance between the current phenotype and the optimum are more likely to move a population closer to that optimum than are mutations with large effects. Fisher concluded that mutations with very small phenotypic effects and fitness advantages are responsible for most adaptive change. However, upon reviewing both the theoretical basis and the empirical evidence for Fisher's micromutational view, ORR and COYNE (1992) concluded to their surprise that both are weak. ORR (2002)(2003) found that the distribution of selection coefficients of mutations fixed by selection is roughly exponential, regardless of the distribution of phenotypic effects of random mutations. This predicts that adaptive change will typically result primarily from a few mutations with large effects, supplemented by mutations with small effects.

It is rarely possible to sample a sufficiently large number of adaptive mutations to describe their distribution, but some studies have supported the more qualitative prediction that most phenotypic evolution is attributable to genetic changes at very few loci. KEIGHTLEY (1998) analyzed a mouse line selected for increased body weight for 50 generations and found that most of the response to selection resulted from mutations at two loci, with some influence of a third locus. IMHOF and SCHLOTTERER (2001) isolated 66 adaptive mutations in Escherichia coli using frequency changes in plasmids hitchhiking with adaptive mutations. Their observations were consistent with an exponential distribution, although other distributions could not be ruled out. On the other hand, when LANDE (1981) estimated that the numbers of loci responsible for quantitative trait variation in several plant and animal populations ranged from 5 to 22, many of these estimates approached the upper limit that would be detectable given the genetic map of the organism under study. Increasingly precise mapping has led to the identification of 53 loci of QTL affecting bristle number in Drosophila (DILDA and MACKAY 2002). Although the distribution of effects was exponential, the authors noted that this could be an artifact of sampling, linkage, or epistasis.

As a supplement to comparisons among natural populations, experimental evolution offers the advantages of known ancestries and constant, reproducible selective environments. The rarity of spontaneous adaptive mutations can be surmounted using large populations of microbes with short generation times (LENSKI et al. 1991). The simple life histories and freeze tolerance of experimental microbes also permit direct estimates of fitness changes, allowing adaptation to be quantified. The ability to measure fitness in the same environment in which it evolved also avoids the limitation that the relationship of some quantitative traits to fitness, or the nature of potential selection upon them, is unclear (ORR and COYNE 1992). Finally, yeast offer the additional advantage of a readily manipulated sexual cycle that permits the Mendelian analysis of adaptive differences between evolved and ancestral genotypes.

The number of detectable mutations is limited by the genetic map of yeast, which contains 90 crossovers per diploid genome. Since the estimated number of adaptive mutations will be well below this limit, it is unlikely to cause serious bias. A common method of estimating numbers of segregating loci responsible for phenotypic variation is QTL mapping. For the direct comparison of ancestral with evolved genotypes used here, QTL mapping is not applicable because insufficient molecular evolution has occurred to provide the necessary markers. Instead, I use biometrical methods to estimate the effective number of loci (n_e). Studies directly comparing estimates from QTL mapping and biometrics are apparently rare. BUTRUILLE et al. (1999) obtained good agreement between QTL mapping and biometric results for flowering time, plant height, and seed weight in Brassica napus.

Unlike QTL mapping, the approach used here is not compromised by recombination between QTL and marker loci. However, it shares with QTL mapping the risk of interpreting a cluster of tightly linked genes as a single locus. The relatively large number of chromosomes (16) and high recombination rates (see WINZELER et al. 1998) of yeast reduce the risk of linked mutations segregating as a single locus. Nevertheless, a recent combination of QTL mapping with microarray analysis in yeast (STEINMETZ et al. 2002) found not only that three major loci affecting heat tolerance were so tightly linked that they segregated as a single locus, but also that one of the high-tolerance alleles originated from the low-tolerance parent. In yeast populations evolved in glucose-limited chemostat cultures, PAQUIN and ADAMS (1983) counted 64 selective sweeps in haploid and diploid populations in 2612 generations over 11 populations. The average fitness increment for 10 of these mutations was 10%.

Here I analyze the genetic basis of adaptation in a genotype sampled from this population. The results support recent predictions that adaptive change is attributable primarily to mutations at few loci.

Experimental evolution:

Experimental populations were founded by derivatives of strain S288c obtained from the American Type Culture Collection (see Table 1). Haploid MATa and MAT strains 875 and 876 were mixed on an agar plate, and a mating pair of cells was isolated by micromanipulation to yield isogenic diploid strain 56. Single colonies of 876 and 56 picked at random from a thinly spread agar plate were the ancestors of 5 MAT haploid and 5 diploid experimental populations, respectively.

View this table: In this window In a new window   TABLE 1 Strains of S. cerevisiae used in this study

 
Experimental populations were grown in 18 x 150-mm borosilicate tubes containing 10 ml dextrose minimal medium supplemented with leucine (DM + Leu, 0.17% yeast nitrogen base without amino acids, 0.5% ammonium sulfate, 0.25% dextrose, and 60 mg/ml leucine) maintained at a constant 30°. Growth curves (not shown) indicated that at a concentration of 0.25%, dextrose is the limiting nutrient for the ancestral genotypes (ZEYL et al. 2003) and that stationary phase is reached after 21 hr. At 24-hr intervals (±30 min) each tube was vortexed thoroughly and 100 µl of each population, containing 2.7 x 10⁶ cells, was transferred to 10 ml of fresh medium. Each population experienced log₂100 6.6 generations per day and the effective population size, calculated as the harmonic mean of population sizes estimated at hourly intervals from 24-hr growth curves, was 1.3 x 10⁷. Every 15 days (100 generations), a 0.9-ml sample of each population was suspended in 15% glycerol and frozen at –80°. Prior to freezing, all populations were checked for contamination by plating on DM agar lacking leucine, which would permit the growth of wild yeast but not the leu2 experimental populations. No such contamination was detected during the 5000 generations for which these populations were propagated.

Fitness trajectories of evolving populations:

Fitness was estimated by competitions between the ancestral genotype and evolved populations or genotypes under conditions identical to those in which the populations evolved. The common competitor in all fitness assays was a genetically marked version of the diploid ancestor (strain 56G), produced by integrative transformation of strain 56 with a BamHI-EcoRI fragment of plasmid pKanMX4 encoding resistance to the antibiotic G418 (WACH et al. 1994). This marker has been consistently found to have no detectable fitness effect when introduced into yeast strains by transformation (BAGANZ et al. 1997). A diploid ancestral competitor was used because this avoids the complication that a haploid would mate with rather than compete with a haploid of the opposite mating type. There is no detectable fitness difference between haploid and diploid versions of the ancestor (ZEYL et al. 2003).

Fitness assays were performed as described by ZEYL and DE VISSER (2001) and ZEYL et al. (2003). Briefly, frozen samples of the ancestral strain and the populations or genotypes of interest were thawed, reacclimated separately to the conditions of the selection experiment for two 24-hr growth cycles, mixed in replicate tubes of 10 ml DM + Leu, and allowed to compete for one or two 24-hr growth cycles, under exactly the same conditions as in the long-term experiment. Genotype frequencies were estimated by plating and replica plating competing mixtures, using a neutral marker encoding G418 resistance to distinguish the ancestor. Fitness was calculated as the net number of cell divisions by the evolved genotype or population divided by that for the ancestor. Ancestral fitness is therefore defined as 1.0. All fitness assays were replicated 10- to 20-fold. In a typical standard 48-hr competition replicated 10-fold, the smallest fitness difference resolvable with 95% confidence is 2%.

Genetic analysis of two haploid genotypes:

From competitive fitness assays of evolving populations at 500-generation intervals, one of the replicate haploid populations was identified as showing representative fitness increases. The fitness trajectory of this population was analyzed in greater detail by fitness assays performed as described above at 100-generation intervals. A single genotype (henceforth A2000) was isolated from this population as a single colony on DM + Leu agar from a sample frozen after 2000 generations. A second genotype (B2000) was similarly isolated from a different haploid population (Tables 1 and 2).

View this table: In this window In a new window   TABLE 2 Fitness after 2000 generations of five haploid (H1–H5) and five diploid (D1–D5) evolved populations

 
Each genotype was crossed with the haploid strain isogenic to its own ancestor but of the opposite mating type. Mating strains were mixed on DM + Leu agar, and mating pairs of cells were identified visually and isolated by micromanipulation under an inverted microscope. Haploid F₁ progeny were obtained from the resulting diploid genotypes by microdissection of asci following sporulation on standard sporulation agar (2% potassium acetate, 60 mg/liter leucine, 50 mg/liter zinc acetate) at room temperature for 3–5 days. To estimate the number of loci involved in adaptation, fitness estimates were obtained for six complete tetrads obtained by crossing A2000 with its ancestor and for five tetrads from the B2000 x ancestor.

Estimates of numbers of adaptive mutations:

The standard biometrical method of estimating the number of loci responsible for a quantitative trait difference between two genotypes is the Castle-Wright estimator (CASTLE 1921), a ratio of the genetic variance in the parental population to that in a population of offspring. The Castle-Wright estimator as modified for haploids (CHOVNICK and FOX 1953; LYNCH and WALSH 1998) is /4 Var, where _Pi are parental trait means (in the present case, evolved and ancestral fitness), Var(z_Pi) are their sampling errors, and 4 Var(S) is the segregational variance in the F₁ population. The quantity thus estimated is known as the effective number of loci (n_e), reflecting the assumption that all alleles affect the trait equally. It is also assumed that all such loci are unlinked, that all interactions among alleles are additive, and that all alleles increasing the trait are fixed in the same parental population. It is unlikely that all the assumptions are ever met, and the Castle-Wright estimator is biased downward, most seriously by variation among alleles in their effect and by linkage (ZENG et al. 1990). ZENG (1992) corrected for this bias with a modified estimator , where r is an estimator of the mean frequency of recombination among loci, z is a function of variation among alleles in their effects on a trait (and variation among loci in allele frequencies, which in the present case of two haploid parental genotypes can be disregarded), and m is itself derived from the Castle-Wright estimator by averaging allelic effects across loci. An estimator for r in turn is where M is the haploid number of chromosomes, c_i is the genetic length in morgans of the ith chromosome, and The genetic map of our ancestral strain of Saccharomyces cerevisiae is known with great precision, and the expected numbers and chromosomal distribution of crossovers have recently been corroborated at the molecular level by genome-wide analysis of recombination break points (WINZELER et al. 1998). A formula for the sampling variance of n_e was calculated by LANDE (1981) and modified by ZENG (1992; LYNCH and WALSH 1998).

Backcrosses:

To check the accuracy of the estimate for the number of mutations responsible for adaptation, repeated backcrosses to the ancestor were used to isolate adaptive alleles on an ancestral background. From the fraction of backcross progeny with fitnesses that are significantly different from that of the ancestor, the number of segregating factors n_WA can be estimated as

where k is the number of generations of backcrossing and p(r 1) is the fraction of nonparental lines (WEHRHAHN and ALLARD 1965). Such lines are assumed to carry only single segregating factors, which is very likely after only a few generations of backcrossing unless the number of factors is very large (LYNCH and WALSH 1998). The method was developed for inbred diploid lines and is applied here to haploids. The 95% confidence limits (MULITZE and BAKER 1985) were also calculated.

The four genotypes of a tetrad from A2000 were crossed with strain 875 or 876, according to their mating type. From each backcross, one diploid cell was isolated and sporulated as above, and one meiotic tetrad was microdissected. Fitness estimates were obtained for these 16 B1 progeny, and a further round of backcrosses and fitness estimates was performed.

Fitness trajectories:

Fitness estimates for 10 initially isogenic, genetically homogeneous populations after 2000 generations ranged from 1.53 to 1.90 (Table 2). By comparison, fitness estimates for four standard laboratory strains unrelated to the ancestor of the experimental populations (including a pathogenic strain isolated from a human patient) ranged from 0.89 to 1.12. The 53–90% fitness gains by evolved populations therefore represent significant adaptation to the experimental environment and not merely recovery by an inferior genotype. The fitness trajectory of one arbitrarily chosen population at 100-generation intervals is shown in Figure 1. Adaptation slowed down considerably after generation 2000; the fitness of this population after 5000 generations was 1.94 (95% C.I. 1.87–2.01). Fitness trajectories for all 10 populations show significantly greater adaptation by haploid than by diploid populations (ZEYL et al. 2003).

View larger version (11K): In this window In a new window Download PPT slide   FIGURE 1.— Fitness trajectory of a haploid yeast population over 2000 generations in minimal glucose-limited medium. Error bars indicate 95% confidence limits on fitness estimates.

 
Genotype A2000, with an estimated fitness of 1.76 (95% C.I. 1.66–1.85), was randomly isolated from this haploid population, H3 (Table 2). A second genotype (B2000) with a fitness of 1.60 was isolated from population H2, for quantitative genetic analysis of adaptation.

Crosses and backcrosses:

Evolved genotype A2000 was crossed with a strain isogenic to its ancestor but of opposite mating type. From a single colony of this diploid hybrid, six complete meiotic tetrads were obtained and fitness estimates were obtained from these 24 haploid progeny. A Kolmogorov-Smirnov test gives no reason to reject the assumption that the distribution of F₁ fitness (Figure 2) is normal (P = 0.20), although it appears bimodal or possibly trimodal. The distribution of F₁ obtained from B2000 also appears discontinuous and barely passes a Kolmogorov-Smirnov test of normality (P = 0.075). Both distributions suggest that adaptive mutations vary greatly in the magnitude of their fitness effects and that a small number of mutations may be responsible for major fitness gains.

View larger version (20K): In this window In a new window Download PPT slide   FIGURE 2.— Fitness distribution of F1 progeny of crosses between the ancestral genotype and (A) evolved haploid genotype A2000 and (B) evolved haploid genotype B2000. Fitness of the parents is indicated below the x-axis as "anc" and "ev." Open triangles inside solid rectangles above the x-axis represent the mean and 95% confidence limits of F1 fitness. Midparent fitness values are indicated with solid triangles below the x-axis.

 
ZEYL et al. (2003) found that the average dominance coefficient of the adaptive mutations in A2000 is 0.20, indicating partially recessive mutations. Mean F₁ fitness (1.249, 95% confidence limits 1.17–1.33) is less than the midparent fitness (1.38), suggesting synergistic epistasis among adaptive mutations. In contrast, the F₁ distribution from B2000 gives no indication of epistasis: the F₁ mean fitness and the midparent fitness are both 1.31 (Figure 2). Crosses could not be performed with genotypes from additional populations because over the course of adaptation as asexual populations, haploids lost the ability to mate and diploids lost the ability to undergo sporulation and meiosis.

Estimation of n_e:

The Castle-Wright estimator for the number of segregating loci affecting fitness, n_e, was calculated from the fitness estimates for ancestral, evolved, and F₁ genotypes. The F₁ segregational variance was estimated as the among-genotype component of variance from ANOVA of the replicate assays of F₁ genotypes. This yields an estimate of n_e = 2.5 for A2000. For B2000 the estimate is n_e = 1.8. Using a gamma distribution of mutation effects (z = 3; see ZENG 1992), our modified estimates of n_e are 4.4 (standard deviation 0.79) for A2000 and 2.7 (SD 2.18) for B2000. These estimates support the inference that a 76% fitness increase is due to six or fewer adaptive mutations. Statistical confidence is tempered by the synergistic epistasis indicated by the F₁ fitness estimates, since the Castle-Wright estimate assumes independent allele effects.

Backcrosses:

One complete tetrad (4 B1 progeny) was chosen from the A2000 crosses for two generations of backcrossing to the ancestor to investigate further the inheritance of fitness. Fitness estimates for the resulting 52 haploid B2 progeny are shown in Table 3. One set of 4 B2s could not be obtained due to the sterility (inability to mate) of a B1 genotype, and for an additional 2 B1 genotypes no further crosses were performed because they had the ancestral fitness of 1.0, indicating that they carried no adaptive mutations. Assuming the latter would have produced 8 B2s with fitnesses indistinguishable from that of the ancestor, 22 of 60 B2s were significantly fitter than the ancestor (whose fitness ± C.I. is 1.004 ± 0.020; Table 3). The Wehrhahn-Allard (WEHRHAHN and ALLARD 1965) estimate is that 3.4 adaptive mutations were segregating in these lines, with 95% confidence limits of 2.10–5.02. If lines with significantly lower fitness are also counted, the estimate is 6.5 mutations, implying that 2–3 deleterious mutations were also segregating. The presence of at least one deleterious mutation in A2000 is indicated by 2:2 segregation of ancestral and below-ancestral fitness in cross BC (Table 3) and the presence of other B2 genotypes with fitness below that of the ancestor. Some tetrads (e.g., DB) suggest that one of the deleterious mutations is less harmful in the presence of one or more of the adaptive mutations than when isolated on the ancestral background.

View this table: In this window In a new window   TABLE 3 Fitness estimates for two backcross generations of a complete F1 tetrad

 
B2 fitness values also support the initial inference of strong positive epistasis, as the 95% confidence limits around mean F₁ fitness (1.002, 1.042) exclude the midparent value (1.062), although the nonnormality of the B2 fitness distribution (Kolmogorov-Smirnov test, P = 0.013) introduces some uncertainty.

A 76% fitness increase in a haploid yeast genotype after 2000 generations occurred by the selection of approximately four adaptive mutations, with one or two deleterious mutations as hitchhikers. Adaptation in a second evolved genotype appears to have occurred by a similarly low number of mutations. These estimates are concordant with the average n_e of 6.3 (range 1–18 loci) in a recent compilation of estimates for 34 traits from 16 different taxa (LYNCH and WALSH 1998, p. 239), although few of the traits in that compilation appear likely to be strongly correlated with fitness.

In contrast to QTL analyses, the Castle-Wright estimate of n_e does not rely on an ability to detect very small effects of individual loci, but rather on the segregational variance in a sample of F₁ progeny. The fitness differences between evolved genotypes and their ancestor are 30–38 x the minimum fitness difference (0.02) detectable by the fitness assay used here, allowing F₁ variance relative to parental variance to be well resolved. Even an insensitive fitness assay would lead to a large estimate of n_e if it showed little variance in F₁ fitness. In contrast, the Wehrhahn-Allard method is downwardly biased by limited power, since it uses estimates of the number of B2 lines differing from the ancestor. This may explain the slightly lower estimate obtained by the latter method.

It should be emphasized that this study is concerned with the genetic factors responsible for a response to selection over thousands of generations, not the factors responsible for the standing variation in natural populations. The implication is not that individuals in a population typically differ by a few, major mutations, but that each differs from an ancestor from thousands of generations ago primarily by a few major factors. It is a major advantage of the yeast system used here that the known ancestor was available for crossing.

Synergistic epistasis is indicated in Figure 2 and Table 3 by the fact that when adaptive mutations were separated by crossing and meiosis, their adaptive effects were reduced. Such epistasis is not surprising, since in this asexual population each mutation was tested by selection only on one particular genetic background. Coadapted combinations are therefore likely to arise. This synergistic epistasis is one factor in the current study that biases downward the estimate of n_e (LYNCH and WALSH 1998).

A standard assumption of quantitative genetic studies is that all the alleles that increase a trait value come from one parent. The results of STEINMETZ et al. (2002) illustrate that using parental lines with the most extreme trait values provides no guarantee of meeting this assumption. In the present study, given the known recent history of the parental genotypes, the evolved parent could reasonably have been assumed to carry the adaptive allele at all loci where it differed from the ancestor, but in fact A2000 carried at least one deleterious mutation, probably two or three. The hitchhiking of these deleterious mutations illustrates one of the disadvantages of asexuality in the experimental populations, since in a sexual population recombination could have separated adaptive from deleterious mutations, allowing selection to fix only the former.

Genotypes A2000 and B2000 and several of the backcrossed genotypes showed varying degrees of sterility and inability to undergo meiosis and sporulation. The other evolved populations suffered much greater loss of sexual competence, limiting the potential to replicate estimates of n_e. A subject of current investigation is whether there is a trade-off between sexual and asexual fitness in the evolved yeast populations (as observed in Chlamydomonas reinhardtii by DASILVA and BELL 1992), so that some of the mutations selected for their beneficial effect on fitness during this asexual experiment have the pleiotropic effect of reducing sexual competence. Alternatively, genes encoding sexual functions may have gradually decayed due to the accumulation of mutations that were neutral in the asexual selection regime.

The experimental population differed from those modeled by ORR (1998)(2003) in being asexual. Other things being equal, the fitness effects of mutations selected in clonal populations are expected to exceed those in sexual populations due to clonal interference: competition between simultaneously segregating adaptive mutations. Assuming that selective sweeps do occur, when two or more adaptive mutations are present at the same time in an asexual population, only the one with the greatest fitness effect will be selected, and the others will be lost. This will inflate the average effect of adaptive mutations fixed in a clonal population. ORR's (1998, 2002, 2003) conclusions that mutations with large effects will be responsible for a substantial fraction of adaptive change will therefore be amplified by asexuality.

I thank M. Carter and A. Pearce for their diligent technical assistance. This work was supported by National Science Foundation grant DEB-0075594.

BAGANZ, F., A. HAYES, D. MARREN, D. C. J. GARDNER and S. J. OLIVER, 1997 Suitability of replacement markers for functional analysis studies in Saccharomyces cerevisiae. Yeast 13: 1563–1573.[CrossRef][Medline]

BUTRUILLE, D. V., R. P. GURIES and T. C. OSBORN, 1999 Linkage analysis of molecular markers and quantitative trait loci in populations of inbred backcross lines of Brassica napus L. Genetics 153: 949–964.[Abstract/Free Full Text]

CASTLE, W. E., 1921 An improved method of estimating the number of genetic factors concerned in cases of blending inheritance. Science 54: 223.[Free Full Text]

CHOVNICK, A., and A. S. FOX, 1953 The problem of estimating the number of loci determining quantitative variation in haploid organisms. Am. Nat. 87: 263–267.[CrossRef]

DASILVA, J., and G. BELL, 1992 The ecology and genetics of fitness in Chlamydomonas VI. Antagonism between natural selection and sexual selection. Proc. R. Soc. Lond. Ser. B 249: 227–233.[CrossRef]

DILDA, C. L., and T. F. C. MACKAY, 2002 The genetic architecture of Drosophila sensory bristle number. Genetics 162: 1655–1674.[Abstract/Free Full Text]

FISHER, R. A., 1930 The Genetical Theory of Natural Selection. Oxford University Press, Oxford.

IMHOF, M., and C. SCHLOTTERER, 2001 Fitness effects of advantageous mutations in evolving Escherichia coli populations. Proc. Natl. Acad. Sci. USA 98: 1113–1117.[Abstract/Free Full Text]

KEIGHTLEY, P. D., 1998 Genetic basis of response to 50 generations of selection on body weight in inbred mice. Genetics 148: 1931–1939.[Abstract/Free Full Text]

LANDE, R., 1981 The minimum number of genes contributing to quantitative variation between and within populations. Genetics 99: 541–553.[Abstract/Free Full Text]

LENSKI, R. E., M. R. ROSE, S. C. SIMPSON and S. C. TADLER, 1991 Long-term experimental evolution in E. coli. Am. Nat. 138: 1315–1341.[CrossRef]

LYNCH, M., and B. WALSH, 1998 Genetics and Analysis of Quantitative Traits. Sinauer, Sunderland, MA.

MULITZE, D. K., and R. J. BAKER, 1985 Evaluation of biometrical methods for estimating the number of genes. 1. Effect of sample size. Theor. Appl. Genet. 69: 553–558.[CrossRef]

ORR, H. A., 1998 The population genetics of adaptation: the distribution of factors fixed during adaptive evolution. Evolution 52: 935–949.[CrossRef]

ORR, H. A., 2002 The population genetics of adaptation: the adaptation of DNA sequences. Evolution 56: 1317–1330.[CrossRef][Medline]

ORR, H. A., 2003 The distribution of fitness effects among beneficial mutations. Genetics 163: 1519–1526.[Abstract/Free Full Text]

ORR, H. A., and J. A. COYNE, 1992 The genetics of adaptation revisited. Am. Nat. 140: 725–742.[CrossRef]

PAQUIN, C., and J. ADAMS, 1983 Frequency of fixation of adaptive mutations is higher in evolving diploid than haploid yeast populations. Nature 302: 495–500.[CrossRef][Medline]

STEINMETZ, L. M., H. SINHA, D. R. RICHARDS, J. I. SPIEGELMAN, P. J. OEFNER et al., 2002 Dissecting the architecture of a quantitative trait locus in yeast. Nature 416: 326–330.[CrossRef][Medline]

WACH, A., A. BRACHAT, R. POHLMANN and P. PHILIPPSEN, 1994 New heterologous modules for classical or PCR-based gene disruptions in Saccharomyces cerevisiae. Yeast 10: 1793–1808.[CrossRef][Medline]

WEHRHAHN, C., and R. W. ALLARD, 1965 The detection and measurement of the effects of individual genes involved in the inheritance of a quantitative character. Genetics 51: 109–119.[Free Full Text]

WINZELER, E. A., D. R. RICHARDS, A. R. CONWAY, A. L. GOLDSTEIN, S. KALMAN et al., 1998 Direct allelic variation scanning of the yeast genome. Science 281: 1194–1197.[Abstract/Free Full Text]

ZENG, Z-B., 1992 Correcting the bias of Wright's estimates of the number of genes affecting a quantitative character: a further improved method. Genetics 131: 987–1001.[Abstract]

ZENG, Z-B., D. HOULE and C. C. COCKERHAM, 1990 How informative is Wright's estimator of the number of genes affecting a quantitative character? Genetics 126: 235–247.[Abstract]

ZEYL, C., and J. A. G. M. DE VISSER, 2001 Estimates of the rate and distribution of fitness effects of spontaneous mutation in Saccharomyces cerevisiae. Genetics 157: 53–61.[Abstract/Free Full Text]

ZEYL, C., T. VANDERFORD and M. CARTER, 2003 An evolutionary advantage of haploidy in large yeast populations. Science 299: 555–558.[Abstract/Free Full Text]

Communicating editor: P. OEFNER

This article has been cited by other articles:

				L. Perfeito, L. Fernandes, C. Mota, and I. Gordo Adaptive Mutations in Bacteria: High Rate and Small Effects Science, August 10, 2007; 317(5839): 813 - 815. [Abstract] [Full Text] [PDF]

				J. A. Hill and S. P. Otto The Role of Pleiotropy in the Maintenance of Sex in Yeast Genetics, March 1, 2007; 175(3): 1419 - 1427. [Abstract] [Full Text] [PDF]

				P. Sliwa and R. Korona Loss of dispensable genes is not adaptive in yeast PNAS, December 6, 2005; 102(49): 17670 - 17674. [Abstract] [Full Text] [PDF]

International Access Link

Online ISS 1091-6490
Copyright 2008 by the Genetics Society of America
phone: 412-268-1812   fax: 412-268-1813   email: genetics-gsa{at}andrew.cmu.edu