Fitness reaction norms and experimental evolution design.

(A) Heritable genotype-by-environment (GxE) fitness variance implies that genotypes (colored lines) have different growth rates along the value of the environmental factor(s) considered [3, 13]. The function of each genotype’s growth rate with environmental value can be denominated its “fitness reaction norm”. Under density- and frequency-independent conditions, the relative magnitude of genotypic growth rates to the average population growth rate (thick line) will determine its deterministic frequency trajectory during evolution [25]. In general, fitness reaction norms of genotypes present in standing genetic variation will cross each other somewhere along the environmental value. A history of evolution in variable environments, with balancing selection, or population structure, for example, will determine the extent of GxE fitness variance in the ancestral population above that expected under a mutation-drift equilibrium. Crossing of fitness reaction norms means that genotypes will be favored at some environmental values while disfavored in others. Assuming no input from mutation or recombination during evolution, with a sudden change to an environmental value of 1 (vertical dotted line), from an ancestral environment 0, selection will favor the grey genotype, while under a gradual change selection will initially favor the red genotype, then the blue one, and only at a later period the grey genotype. Vertical lines broadly define the three population genetic stages expected under a gradual environmental change. In a first stage (I), the best genotype at the most extreme environment (i.e., the grey genotype) will be selected against and kept at such low frequency that its stochastic loss by genetic drift is likely under small population sizes. In a second stage (II), around where the fitness reaction norms cross, reduced fitness variance will slow down adaptive changes in allele frequency [16], and depending on the amount of standing variation [11], the best genotypes can be kept at such low frequencies that again their stochastic loss is likely. Under slower environmental change, therefore, a population will spend more time in stages I and II, reducing the chance that the best genotypes will be present at a third stage (III), when they can be selectively favored. Assuming no de novo mutation or recombination during gradual evolution the loss of the best genotypes will restrict adaptation to the most extreme environment [11]. At stage III, it is also possible, particularly in large populations with limited recombination between extant genotypes, that interference between the best genotypes (the grey and the blue genotypes for example) will transiently reduce selection efficacy on the best one [28, 29]. This interference process in turn will favor stochastic loss of the best genotypes if at the start of stage III they are at low frequencies. Note that without prior knowledge of fitness reaction norms, genotypes that increase in frequency and then decrease in frequency (for example the red genotype) would suggest the presence of negative-frequency dependence. (B) Experimental evolution design reported in [18] (S1 Table), used here to address whether slower environmental change hinders adaptation. A single 140-generation lab-adapted C. elegans population with abundant genetic diversity [19, 21], reproducing only by self-fertilization, was the ancestor for experimental evolution [18]. In the sudden regime, 4 replicate populations were faced from the first generation onwards to 305 mM NaCl in their growth media (high salt, black line). In the gradual regime, 7 replicate populations were faced with an 8 mM NaCl increase each generation until generation 35, being then kept at 305 mM until generation 50 (dark grey). In the control regime, 3 replicate populations were kept at 25 mM NaCl, the conditions to which the ancestor was adapted to (light grey). Here we genotype individual hermaphrodites at several time points during experimental evolution, at single nucleotide polymorphisms (SNPs) across pairs of chromosomes (vertical dashed lines, Fig 2 and Fig 3). Analysis of this data allow us to infer the fitness reaction norms of extant genotypes and predict with a deterministic model their frequency trajectories (Fig 4 and Fig 5), which are then empirically validated (Fig 6). To address the role of genetic drift and founder effects in the loss of the best genotypes during slower rates of environmental change we perform a second set of evolution experiments, where 7 of the gradual populations at generation 35 are kept in constant high salt for an extra 30 generations (Fig 7).