Evolution of pleiotropic costs in experimental populations

Authors

  • J.-N. Jasmin,

    Corresponding author
    1. Department of Biology, Wake Forest University, Winston-Salem, NC, USA
    2. CEFE-UMR 5175, Montpellier, France
    • Correspondence: Jean-Nicolas Jasmin, CEFE-UMR 5175, 1919 route de Mende, Montpellier 34293 Cedex 5, France.

      Tel.: +33/0 4 99 23 48 51; fax: +33/0 4 67 61 33 36;

      e-mail: jnjasmin@gmail.com

    Search for more papers by this author
  • C. Zeyl

    1. Department of Biology, Wake Forest University, Winston-Salem, NC, USA
    Search for more papers by this author

Abstract

The fitness of populations adapting to new environments is expected to decline in different environments, but empirical studies often do not lend support for such adaptation costs. We test the idea that the initial fitness of the selected populations in the environment where the cost is estimated is key for interpreting tests of ecological trade-offs. We isolated single clones of the yeast Saccharomyces cerevisiae every ~250 generations from replicate experimental lineages that had been selected during 5000 generations in a glucose-limited environment. We then selected these clones in a galactose-limited environment for ~120 generations. Finally, we estimated single-clone fitness in both environments, before and after selection on galactose. The pleiotropic effects on glucose of selection on galactose evolved from positive to negative as fitness in glucose increased, providing strong support for the importance of initial fitness for determining the sign and magnitude of pleiotropic effects. This demonstrates that the sign of pleiotropic effects for fitness following adaptation to a new environment can change during long-term adaptation to an original environment. We also found no relationship between the size of the fitness changes in galactose and glucose, such that pleiotropic effects in glucose became relatively smaller as the sizes of direct effects on galactose increased.

Introduction

Pleiotropic fitness costs associated with adaptations are thought to be a common source of genotype-by-environment interactions for fitness, thus explaining some cases of stable polymorphisms (Lenski & Levin, 1985) and ecological specialization (Kassen & Bell, 2000; Dykhuizen & Dean, 2004; Remold, 2012). Adaptation costs can also contain the evolution of undesirable phenotypes such as antibiotic, herbicide or pesticide resistances (Andersson & Hughes, 2010). However, evolutionary experiments in which populations were selected in novel environments have shown that adaptive alleles often do not have negative effects on performance in ancestral or alternative environments (see supporting Table S1 for a survey of examples).

The scarcity of demonstrated adaptation costs – and more generally of genotype-by-environment interactions – in evolutionary experiments can be attributed to any of four non-exclusive causes (omitting low statistical power). First, the environment in which pleiotropic fitness effects are considered may not be sufficiently contrasted from the selective environment. For example, Travisano & Lenski (1996), upon observing a positively correlated response for competitive fitness on other carbon sources to selection in a glucose-limited environment, noted that E. coli use the same uptake mechanism to transport all these sugars across the cell membranes (see also Travisano et al., 1995; Ostrowski et al., 2005; Jasmin & Kassen, 2007). Likewise, it is unlikely that costs of adaptation would evolve in populations selected on distinct but closely neighbouring values of an environmental gradient (Wallace, 1989; Mongold et al., 1996; Kassen & Bell, 2000; Hughes et al., 2007; Moser & Bell, 2011). Adaptive alleles vary in the amount of environmental contrast (and thus genotype-by-environment interaction) that they generate, as indicated by the fact that even when replicate selection lines reveal an overall cost, some lines often do not show a cost (e.g. Lee et al., 2009; Table S1). Second, compensatory mutations may mitigate pleiotropic costs before these costs are assayed (Lenski, 1988; Cowen et al., 2001; Maclean et al., 2004; Cowperthwaite et al., 2005). In this case, even though alleles with pleiotropic costs may spread, their effects are short-lived and epistatic. Third, the new adaptations may be inducible, and this phenotypic plasticity itself may not be costly (Fry, 2001). Fourth, the starting population (prior to its selection in a new environment) may have been maladapted to the environments in which pleiotropic effects are estimated (Joshi & Thompson, 1995; Maclean & Bell, 2002). Whereas compensatory mutations mask costs, high absolute fitness in environments in which pleiotropic effects are estimated may create costs. This fourth cause has often been invoked to explain positively correlated responses to selection in small sexual populations (e.g. fruit flies or mites; Gould, 1979; Service & Rose, 1985; Fry, 1990; Joshi & Thompson, 1995; Chippindale et al., 2003; Fry, 2003), but it is rarely part of the interpretation of results in the microbial evolution literature. Few studies can actually reject low absolute fitness as an explanation for cost-free adaptation (Table S1). Our hope is that this paper demonstrates clearly that this issue applies also to large asexual populations.

To this end, we studied how pleiotropy for fitness evolves under long-term adaptation in the yeast Saccharomyces cerevisiae. When the strain we used to initiate our experiments was selected either in glucose- or in galactose-limited media, pleiotropic fitness effects were positive in the other medium (this study; note that the two environments differed also in the number of daily population cycle and bottleneck population size). Such reciprocally cost-free adaptation made this pair of environments suitable for testing the idea that long-term adaptation can change pleiotropic effects. Replicate lines were selected in a glucose-limited environment over 5000 generations of selection (Zeyl et al., 2003; Jasmin & Zeyl, 2012), and here we test whether this long-term selection on glucose changed the sign and magnitude of the pleiotropic fitness effect on glucose of subsequent adaptation to galactose. To do this, we isolated 20 clones from each of six glucose-selected lines across 5000 generations of selection on glucose and selected in triplicate these 120 clones in a new, galactose-limited environment for ~120 generations, after which we isolated a single clone from each of these 360 lines to estimate direct and pleiotropic effects of selection (Fig. 1). We found that pleiotropic effects for fitness of new advantageous mutations changed from positive to negative under long-term adaptation.

Figure 1.

Fitness landscape representation of the experiment. Two arbitrary traits x and y underly fitness in the two environments, defined by the distance from the galactose (GAL) or glucose (GLU) adaptive peaks (the isoclines show the distance from each peak). Single clones (empty symbols) are isolated from an experimental population adapting to the glucose-limited environment for 5000 generations (one population and 10 clones are shown, but our experiment involved six replicate selection lines and 20 single-clone isolates per line). Each isolated clone is then selected in the galactose-limited environment for ~120 generations (in triplicate, although only one replicate is shown), and a single clone is then isolated from each galactose-selected population (filled symbols). The pleiotropic effects on glucose of selection on galactose are changes in the distance from the glucose peak following selection on galactose. Our aim is to test whether the effect on competitive fitness in glucose of selection on galactose changed during the long-term adaptation to glucose (the example above shows the evolution of negative pleiotropic effects starting from positive pleiotropic effects).

Materials and methods

Zeyl et al. (2003) selected six asexual populations (A-F) of the budding yeast Saccharomyces cerevisiae for 5000 generations (750 days) in a glucose-limited environment (2.5 g L−1 glucose, 5 g L−1 ammonium sulphate, 1.7 g L−1 yeast nitrogen base, 60 mg L−1 leucine; temperature of 30 °C) at an effective population size of ~6 × 106 individuals. The selection regime consisted of daily 100-fold dilutions of 24-h-old culture into 10 mL of medium. The six selection lines were started from a clone derived from strain S288C. Lines A-E were haploid from the beginning to the end of the selection experiment, whereas line F started as diploid but became haploid through spontaneous mutation and natural selection within the first 300 generations of selection in glucose. During selection, aliquots were frozen bi-monthly (~100 doublings) at −80 °C in 15% glycerol.

Here, we isolated single clones from samples frozen at ~250-generation intervals from each of the six selection lines in glucose, which we refer as the 20 × 6 ancestors. These clones were cryopreserved at −80 °C in 15% glycerol for assays, and we selected each of them in triplicate in a new environment for ~120 generations (3 × 120 selection lines). The new medium was identical to the glucose-limited medium above, except that glucose was replaced with galactose. The selection regime in galactose consisted of transferring for each line 80 μL of one-day-old culture into 120 μL of fresh media in a single well of 96-well plates (CoStart, Corning). After ~120 generations of selection on galactose (90 daily transfers), we isolated a single clone from each of the 360 selection lines; these are the ‘evolved’ clones (Fig. 1).

The responses to the selection of 120 ancestral and 360 evolved clones were assayed in glucose by head-to-head competition experiments, in which genetically marked strains were mixed and competed. The competitive fitness of the ancestral clones in glucose (Wanc) gives the direct response to selection in glucose. The difference in competitive fitness in glucose between the ancestral clones and the clones selected in galactose is the pleiotropic effect for fitness of selection in galactose (ΔWevoWanc). Competitive fitness assays in glucose used a single common competitor derived from the common ancestor to all the strains and marked with a deletion in the ade2 gene, which causes colonies to turn pink on agar plates containing 15 mg L−1 adenine. The conditions for the assays in glucose were the same as during selection except that 80 mg L−1 of adenine was added to the media. Competitive fitness was estimated by the change in frequency of the focal clone relative to the common competitor in a model with non-overlapping generations,

display math

where p0 and pd are the frequencies of the focal clone before and after two daily cycles of selection, respectively (Chevin, 2010), and d is the number of doubling expected during two cycles of selection [i.e. 2 × (ln100)/ln2 = 13.3].

Growth rate on galactose was measured for the ancestral and evolved clones under the same conditions as during selection. Here, we estimate the direct effect of selection on galactose as the difference in growth rate before and after selection on galactose and test for a relationship between the magnitudes of direct and pleiotropic (ΔW) effects of selection on galactose. To estimate clone growth rate, biomass was estimated every 30 min on 24-h growth curves by the optical density O.D. of the culture at 630 nm in an incubating spectrophotometer (PowerWave XS2; BioTek Instruments, Winooski, Vermont). Clone growth rate was estimated by fitting to the growth curve data (after subtracting from each well, the O.D. of a blank well containing medium only) the Beverton–Holt model (Beverton & Holt, 1957; Jasmin & Zeyl, 2012). For the assays in glucose and in galactose, each evolved clone was paired with its ancestor in adjacent wells (on 96-well plates for galactose) or tubes (on racks of 80 tubes for glucose). The assays were carried out in replicated blocks where each block is one set of 120 evolved clones and their corresponding 120 ancestors.

Results and discussion

Being better adapted to the glucose-limited environment made adaptation to the galactose-limited environment more costly. For all six glucose selection lines, selection on galactose increased the fitness in glucose of genotypes poorly adapted to glucose, but it decreased the fitness of genotypes well adapted to glucose (Fig. 2). These evolutionary changes in the sign of pleiotropic fitness effects in glucose between low- and high-fitness genotypes are indicated by the regression of Wevo on Wanc crossing the equivalency line (Wevo = Wanc) and by the regression slopes being significantly shallower than one for each selection line (Fig. 2). Each regression line was also steeper than zero (all F1,18 > 8, < 0.01). These results demonstrate that pleiotropic effects for fitness following adaptation to a new environment (galactose) can change sign during long-term adaptation to an original environment (glucose). The evolution of ΔW = (Wevo−Wanc) over the 5000 generations of selection on glucose supports the same conclusions (Fig. S1). The most obvious explanation for these main findings is that adaptations of early clones were not specific to the selection regime or carbon source (100-fold daily dilution per day in 10 mL of glucose-limited medium vs. 2.5-fold daily dilution per day in 200 μL of galactose-limited medium), whereas late clones adapted more specifically to either environment. This interpretation is consistent with the first and fourth scenarios outlined in the Introduction section. According to this interpretation, the two environments were not contrasted for early clones simply because they were poorly adapted to both, and aspects common to the two environments were driving early adaptation. This interpretation further suggests that the large-effect mutations available to early clones (as evidenced by the high rate of adaptation of early clones to both the glucose and galactose environments; Jasmin & Zeyl, 2012) may be more ‘generalist’ than the small-effect mutations available to later clones

Figure 2.

The relationship between competitive fitness in glucose before (Wanc) and after (Wevo) short-term selection on galactose for clones isolated from six glucose long-term selection lines (a–f). The slope (±SE), R2 and P-value of a t-test against an expected slope of 1 are also indicated for each regression line.

Two hypotheses are usually evoked to explain costs of adaptation: the ‘mutation accumulation’ hypothesis and the ‘antagonistic pleiotropy’ hypothesis (Kassen, 2002). The mutation accumulation hypothesis, according to which costs of adaptation are caused by conditionally neutral mutations (Maclean & Bell, 2002), is unlikely to explain our results because (1) the six replicate selection lines in glucose evolved adaptation costs in parallel (Figs 2, S1), supporting antagonistic pleiotropy rather than a stochastic process (Cooper & Lenski, 2000), and (2) selection on galactose was too brief (~120 generations) for the isolated clones to have accumulated conditionally neutral mutations (only one or two beneficial mutations are expected to fix during that period (Adams, 2004; Zeyl, 2004), and given a mutation rate affecting fitness of one per 10 000 genome replications in yeast (Zeyl & De Visser, 2001; Lynch et al., 2008), the probability of an isolated clone carrying a conditionally neutral mutation is on the order of 0.0001). Thus, the effects on glucose of selection on galactose are more likely to result from pleiotropic gene action. However, we do not know whether mutations adaptive in the glucose or galactose environment affected the same loci or metabolic pathways.

Another variable that can impact the size of pleiotropic fitness effects in glucose (the absolute value of ΔW) is the size of direct effects (Fisher, 1930; Haldane, 1932; Lande, 1983; Otto, 2004), here the amount by which clones improved on galactose. We estimated the direct response to selection in galactose as the change in growth rate on galactose following selection in that environment (growth rate can be used as a proxy for fitness in the galactose environment because selection generally increases that life-history trait in that environment, Fig. 3a; Vasi et al., 1994; Jasmin & Zeyl, 2012). We found no correlation between direct and pleiotropic effect sizes (Fig. 3a; F1,115 = 0.53, = 0.60), indicating that mutations with large direct effects on galactose had relatively small pleiotropic effects in glucose in comparison with mutations with small direct effects.

Figure 3.

Relationship between the size of pleiotropic effect on glucose following selection on galactose (absolute value of ΔW) and the direct response on galactose for all 120 clones (a). Large direct responses to selection on galactose tended to occur in genotypes poorly adapted to glucose (b).

Variation in ancestral fitness in glucose may also explain this latter result. Adaptive changes of large effect on galactose tended to occur in clones poorly adapted to glucose (small Wanc) and, conversely, small adaptive changes in galactose occurred in clones well adapted to glucose (there is a negative correlation between the magnitude of direct effects in galactose and ancestral competitive fitness in glucose; Fig. 3b, regression slope ± SE, −6.5 ± 1.3, F1,117 = 24, P < 0.0001, R2 = 0.17; the regression is still statistically significant if negative direct effects are excluded, −4.3 ± 1.2, F1,101 = 13, P = 0.0005, R2 = 0.11). This suggests that the fitness in glucose of clones well adapted to glucose was relatively more sensitive to genetic perturbations (generated by selection on galactose) than was the fitness of clones poorly adapted to glucose. If this interpretation is correct, our results provide compelling evidence for epistasis in the pleiotropic effects of beneficial mutations. Evidence for such epistatic effects is accumulating in studies of antibiotic resistance (e.g. Maclean et al., 2010; de Visser et al., 2011), but it remains limited for other traits. What may be the most surprising here is that the interaction between Wanc and the size of direct effects in galactose almost exactly cancelled out the possible association between direct effects in galactose and correlated effects in glucose (ΔW).

Several studies suggest an effect of starting fitness on pleiotropic costs. Evolutionary experiments with phytophagous insects have repeatedly shown that prior adaptation to a focal host is required before adaptation to an alternative host can generate a fitness cost on the focal host (Gould, 1979; Joshi & Thompson, 1995; Agrawal, 2000), although costs are not inevitable (Magalhaes et al., 2009; Table S1). Bacterial populations initiated from a single clone adapted at pH 7.0 paid a cost at pH 7.8 when selected at pH 5.3, but not the reverse (Hughes et al., 2007). The process demonstrated in our study predicts that, had the E. coli ancestor been adapted to pH 6.0 rather than 7.0, the sign of the two pleiotropic effects may have been reversed. Pleiotropic effects may depend on initial absolute fitness as much as genetic correlations for fitness across environments. Similarly, differences in the cost of rifampicin resistance alleles in two Mycobacterium tuberculosis clinical isolates (Gagneux et al., 2006) may be explained by the initial absolute fitness of these isolates to the assay conditions. The fact that the fitness of well-adapted populations declines faster than that of less-well-adapted populations under strong random genetic drift (e.g. single-individual bottleneck in asexuals, full-sib mating in sexuals or relaxed selection in both) provides further evidence that the fitness of starting populations determines the vulnerability of genotypes to new mutations (e.g. Burch & Chao, 2004; Zeyl et al., 2005; Silander et al., 2007). Lastly, positive epistasis among deleterious mutations (de Visser et al., 2011) and the cost of resistance to different drugs (Trindade et al., 2009; Ward et al., 2009) also suggests that well-adapted genotypes are more vulnerable to genetic changes than less-fit genotypes.

Initial absolute fitness may provide insights about two other patterns regarding pleiotropy. First, there is often much more variation among replicate lines in the pleiotropic effect sizes of advantageous mutations than in their direct effect sizes (Travisano & Lenski, 1996; Ostrowski et al., 2005, 2008; Bennett & Lenski, 2007). Second, reciprocal pleiotropic effects, in which direct and pleiotropic effects are each estimated in a pair of environments, tend to be asymmetrical, in both sign and magnitude (Falconer, 1953; Bohren et al., 1966; Bradford, 1968; Gromko, 1995; Kassen, 2002; Nidelet & Kaltz, 2007). The two patterns could be explained by the starting populations being relatively maladapted to the correlated (pleiotropic) environments, causing average pleiotropic effects to be around zero.

We have shown that the pleiotropic effects of adaptive mutations in a new environment evolved from positive to negative during long-term adaptation in an original environment (i.e. there is epistasis for pleiotropic effects on fitness; de Visser et al., 2011). Therefore, pleiotropic effects of adaptive mutations can only be interpreted when the absolute fitness of the starting population in the environments in which pleiotropic effects are measured is known. Positive pleiotropic fitness effects may reflect the fact that the starting genotypes are maladapted to all environments (Joshi & Thompson, 1995). Of course, these positive pleiotropic effects are not uninteresting; but the role of initial absolute fitness should be included in inferences arising from them. Experimental evolutionists are starting to estimate the distribution of pleiotropic effects on fitness for selected mutations (e.g. Bataillon et al., 2012). Our results suggest that the sign and magnitude of these effects are expected to be genotype and environment specific.

Acknowledgments

We thank Josh B. Kelley and the work-study students for their help in the laboratory. Comments from C. Devaux, I. Olivieri and two anonymous reviewers improved a previous version of the manuscript. J.N.J was supported in part by the Fonds de Recherche du Québec – Nature et Technologies (FQRNT), and the research was supported by US National Science Foundation grant # 0820969 to C.Z.

Ancillary