Report

Adaptation and Evolutionary Rescue in Metapopulations Experiencing Environmental Deterioration

See allHide authors and affiliations

Science  10 Jun 2011:
Vol. 332, Issue 6035, pp. 1327-1330
DOI: 10.1126/science.1203105

Abstract

It is not known whether evolution will usually be rapid enough to allow a species to adapt and persist in a deteriorating environment. We tracked the eco-evolutionary dynamics of metapopulations with a laboratory model system of yeast exposed to salt stress. Metapopulations experienced environmental deterioration at three different rates and their component populations were either unconnected or connected by local dispersal or by global dispersal. We found that adaptation was favored by gradual deterioration and local dispersal. After further abrupt deterioration, the frequency of evolutionary rescue depended on both the prior rate of deterioration and the rate of dispersal. Adaptation was surprisingly frequent and rapid in small peripheral populations. Thus, evolutionary dynamics affect both the persistence and the range of a species after environmental deterioration.

Global environmental change is causing elevated rates of population extirpation (1, 2), giving rise to concern that the rate of environmental change may exceed the capacity of populations to persist and maintain their range (3). A population that is exposed to a stress may adapt through natural selection and thereby avoid extinction (47). This process of evolutionary rescue has been documented for single populations in the laboratory (8) but has not been studied in metapopulations representing the range of a species. Hence, it is not known whether evolutionary rescue can occur across metapopulations to prevent range collapse and species extinction. Patterns of range collapse (9, 10) show that species can persist in the periphery of their historical range. Peripheral populations tend to show lower levels of genetic diversity than central populations, although this difference is not large (11). Theory indicates how species will respond to stress, as their range collapses or expands or shifts in space (12), but these predictions remain untested experimentally. Here we describe an experiment showing that adaptation to environmental deterioration is most likely to evolve when change is slow and individuals can disperse. Consequently, the history and spatial structure of metapopulations affect the likelihood of evolutionary rescue and thereby the maintenance of the species range.

We set up model ranges of baker’s yeast on standard 96-well plates where each well constituted a population (13). Yeast is increasingly being developed as a model organism in ecology and evolution (8, 1416). We maintained the populations in continual growth by transferring a small quantity from each well at regular intervals to the corresponding well on a new plate supplied with fresh growth medium. We stressed the populations by adding different amounts of salt to the growth medium to create a gradient across the plate, with more favorable conditions (low salt concentration) to the “south” and more stressful conditions (high salt concentration) to the “north” (fig. S1). This created a corresponding gradient of population density, higher in the south than in the north. We then intensified this stress by shifting the gradient to the south between transfers, compressing the range and eventually creating conditions in the far north of the plate that would be lethal to the ancestral populations. This enabled us to track shifts in population size and study evolutionary rescue across the metapopulation.

The likelihood of evolutionary rescue in a metapopulation will depend on the quantity of genetic variation, the rate of environmental deterioration, and the rate of dispersal between individual populations. Our yeast populations were derived from a single colony and grew asexually, so that genetic variation was at first minimal and was thereafter generated by mutation and changes in gene expression (17). Almost 500 loci contribute to tolerance of high concentrations of salt in yeast (18), and about 60 have severely reduced competitive ability in saline medium (19), so the loci that potentially contribute to evolutionary rescue at high salt concentration make up at least 1% of the yeast genome. Candidates for genes involved in evolutionary rescue at high salt concentration were previously identified by functional profiling (8, 19). We imposed three rates of environmental deterioration: zero (“constant” treatment, gradient unchanged between transfers), low (“slow” treatment, gradient shifted southward every four transfers), and high (“fast” treatment, shifted every two transfers). In each case, there were three modes of dispersal from well to well during each transfer: no dispersal (each well is inoculated exclusively from the corresponding well on the old plate), local dispersal (each well receives a small contribution from wells with the next lower salt concentration on the old plate), and global dispersal (each well receives a small contribution from all the wells on the old plate). Hence, the experiment comprised nine combinations of rate of deterioration and mode of dispersal. It was conducted in two phases. In the first phase, we followed the dynamics of adaptation by recording the yield (cell density at the end of the growth cycle) (13) of each population as the environment deteriorated at the prescribed rate. In the second phase, we transferred each experimental plate regardless of its history to the same severe gradient to measure the frequency of evolutionary rescue across all the populations in the range.

Yield at first declined linearly from south to north across the plate in response to increasing salt concentration. The rate of decline and the spatial distribution of yield changed over time as the populations evolved (Fig. 1). There was no statistical interaction between the rate of environmental deterioration and the mode of dispersal [analysis of variance (ANOVA), F4, 4 = 1.69, P = 0.18)] The rate of change in mean yield over the whole range varied with the rate of environmental deterioration (ANOVA, F2,2 = 625.45, P < 0.001). It increased slightly in the constant treatment, and declined substantially under slow deterioration and much more rapidly under rapid deterioration. This was modulated by dispersal (ANOVA, F2,2 = 8.74, P = 0.0012) because local dispersal extended the habitable zone northward and slowed the decline of yield relative to both no dispersal and global dispersal (fig. S2).

Fig. 1

The effect of environmental change and dispersal on average (±95% CI) rate of change in yield across all 60 populations per metapopulation over 12 transfers.

Within the metapopulation there is a pronounced U-shaped relationship between the rate of change in yield and the level of stress, such that populations exposed to intermediate salt concentration declined more rapidly than those at very low or very high concentration. This was apparent in the constant treatment (fig. S3) and was exaggerated by environmental deterioration (Fig. 2 and table S1). The effect is modest at low concentration, amounting to a few percent of overall yield, but is much more substantial, amounting to 30% or more of overall yield, at high concentration. This U-shaped relationship arises because the rate of adaptation to stress is necessarily constrained by the concomitant reduction in population size. Smaller populations are expected to have a lower overall mutation supply rate and therefore to adapt more slowly to a stressor such as salt (20). The fraction of these mutations that are beneficial with respect to a specific stress depends on the state of adaptedness of the population. In benign conditions to which a population is well adapted, there may be no mutations available that would further increase growth. In mildly stressful conditions a few beneficial mutations may exist, and the fraction of mutations that are beneficial will increase with the level of stress, up to the point where the breakdown of fundamental biological processes hinders or entirely prevents any alleviation. The overall rate of beneficial mutation is the product of the mutation supply rate, which falls with increasing stress, and the fraction of mutations that are beneficial, which rises with increasing stress up to some limit. Consequently, the rate of adaptation may be a U-shaped function of stress, with populations subjected to mild stress adapting slowly, or failing to adapt at all, whereas those subjected to severe stress adapt more rapidly (fig. S4). The effect of environmental deterioration is to expose a new range of beneficial mutations in populations that had become adapted to worsening conditions (21, 22). This shifts the curve relating the fraction of beneficial mutations to the level of stress to the right, thus deepening the U-shaped pattern of adaptation (Fig. 2 and fig. S5). This was so pronounced that we observed a positive response to selection only under nearly lethal levels of stress at the edge of the range, a surprising effect that was observed even in isolated populations experiencing rapid environmental deterioration.

Fig. 2

The mean rate of change in yield across the gradient of stress (NaCl, g/liter). Salt concentration is the initial value for the treatment: in deteriorating environments, this concentration rises over time. The panels show the patterns of response in relation to treatment combinations [(13), and main text for treatment descriptions] of environmental change (top to bottom: constant, slow, and fast) and dispersal mode (left to right: no dispersal, local, and global). The gray lines show the locally weighted polynomial regression (lowess) fit, and the horizontal black lines provide a guide for no change in yield. The panel at top right is reproduced in more detail as fig. S3.

In the second phase of the experiment a further episode of deterioration, of the same magnitude as those of previous episodes, was imposed on all the experimental ranges and resulted in lethal conditions appearing in the southern section of the plates (13). For the slow and fast treatments, this was a continuation of the previous tempo of deterioration, whereas for the constant treatment it represented an abrupt severe stress. Subsequently, we often observed the abrupt decline in yield followed by an increase typical of population recovery by evolutionary rescue. We expected that the frequency of rescue would be increased by prior selection in a deteriorating environment because beneficial mutations conferring tolerance to lethal conditions would have spread by virtue of their correlated advantage at sublethal levels of stress. We further expected that rescue would be facilitated by dispersal, because beneficial mutations that had spread in other populations could be introduced into poorly adapted populations as immigrants.

The rate of environmental deterioration and mode of dispersal interact to affect the frequency of rescue (Fig. 3 and table S2). We scored a rescue event when the yield of a population after three transfers exceeded its yield just before the abrupt stress. The fewest events occurred in unconnected ranges that had previously experienced no environmental change [mean ± 95% confidence interval (CI) for constant, zero-dispersal treatments: 25.25 ± 4.9]. Both local and global dispersal significantly increased the mean number of rescue events in the constant environment (local dispersal: 35.75 ± 5.85; global dispersal: 49.75 ± 6.91). Environmental deterioration before abrupt environmental change markedly increased the mean number of evolutionary rescue events. This was most apparent in unconnected metapopulations (no dispersal: constant = 25.25 ± 4.9, slow = 40.75 ± 6.25, and fast = 45.75 ± 6.62); dispersal did not significantly improve the number of rescue events in the presence of historical environmental change. Hence, the likelihood of evolutionary rescue depends on both the history of environmental change and the connectivity of the metapopulations.

Fig. 3

The effect of dispersal (isolated, local, and global) and historical environmental change (constant, slow, and fast) on the mean (±95% CI) number of evolutionary rescue events (maximum = 60) across the metapopulations after abrupt environmental change.

A history of environmental deterioration also affected the degree of adaptation that evolved after abrupt change (Fig. 4 and table S3). In the constant treatment, there was a U-shaped relationship between population growth rate and salt concentration, as in the first phase of the experiment, indicating evolutionary rescue in populations at the extreme ends of the stress gradient. In metapopulations that had previously experienced environmental deterioration, whether slow or fast, the mean rate of change in yield was positive at all points along the concentration gradient, showing the overwhelming effect of prior adaptation to severe stress in enhancing adaptation at lethal levels of stress. The curvilinear relation between adaptation and stress is no longer apparent; instead, adaptation increases steadily from lower to higher salt concentrations. This is expected because adaptation increases population size, making the relation between stress and the mutation supply rate more shallow, and thereby tending to establish a monotonic relation between stress and adaptation (fig. S6).

Fig. 4

The effect of past environmental change and dispersal mode on the mean rate of change in yield over four transfers after abrupt environmental change. The panels show the patterns of response to salt (NaCl, g/liter) stress in relation to treatment combinations [(13), and main text for treatment descriptions] of environmental change (rows top to bottom: constant, slow, and fast) and dispersal mode (columns left to right: no dispersal, local, and global). The gray lines show the locally weighted polynomial regression (lowess) fit, and the horizontal black lines provide a guide for no change in yield.

Here we have shown that slow prior deterioration and modest levels of dispersal can foster the ability of a species to evolve tolerance to environmental stress sufficient to extirpate its ancestor. The mechanism responsible appears to be the spread of beneficial mutations conferring adaptation to intermediate levels of stress, which have the correlated effect of also conferring some degree of tolerance to severe stress before this is actually experienced, as previously documented for the yeast-salt system in single populations (15). Local dispersal transfers these alleles up the stress gradient and thereby prevents the collapse of the metapopulation; global dispersal is sometimes less effective because both the mean fitness of immigrants and the probability of introducing a well-adapted mutant are much lower. Populations embedded within connected metapopulations that have experienced historical environmental change have a greater probability of recovering and persisting by evolutionary rescue after a severe perturbation.

Our microcosms are incomplete models of metapopulations occupying natural landscapes, where other processes such as interactions between species or between combinations of stressors also influence range dynamics (23). They are an important first step toward developing more realistic laboratory models using automated transfer systems with high levels of replication to analyze the factors responsible for adaptation and evolutionary rescue in deteriorating environments (24). We believe that such experiments can provide the basis for the synthesis of spatial ecology with evolution that is required to understand the impacts of anthropogenic stress on the Earth’s biodiversity.

Supporting Online Material

www.sciencemag.org/cgi/content/full/332/6035/1327/DC1

Materials and Methods

Figs. S1 to S6

Tables S1 to S3

Additional Data tables S1 to S3

References (2529)

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. R Development Core Team, R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, 2010).
  3. Acknowledgments: The robot was programmed and operated by Z. M. Wang, assisted by A.-M. LeHeureux. G.B. and A.G. are supported by funding from the Natural Sciences and Engineering Research Council of Canada and by the Canadian Foundation for Innovation. A.G. is supported by the Canada Research Chair Program. We thank two anonymous reviewers for comments that improved the manuscript.
View Abstract

Navigate This Article