(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



Predicting population genetic change in an autocorrelated random environment: Insights from a large automated experiment

['Marie Rescan', 'Cefe', 'Cnrs', 'Université De Montpellier', 'Université Paul Valéry Montpellier', 'Ephe', 'Ird', 'Montpellier', 'Université Perpignan Via Domitia', 'Centre De Formation Et De Recherche Sur Les Environnements Méditerranéens']

Date: 2021-08

Most natural environments exhibit a substantial component of random variation, with a degree of temporal autocorrelation that defines the color of environmental noise. Such environmental fluctuations cause random fluctuations in natural selection, affecting the predictability of evolution. But despite long-standing theoretical interest in population genetics in stochastic environments, there is a dearth of empirical estimation of underlying parameters of this theory. More importantly, it is still an open question whether evolution in fluctuating environments can be predicted indirectly using simpler measures, which combine environmental time series with population estimates in constant environments. Here we address these questions by using an automated experimental evolution approach. We used a liquid-handling robot to expose over a hundred lines of the micro-alga Dunaliella salina to randomly fluctuating salinity over a continuous range, with controlled mean, variance, and autocorrelation. We then tracked the frequencies of two competing strains through amplicon sequencing of nuclear and choloroplastic barcode sequences. We show that the magnitude of environmental fluctuations (determined by their variance), but also their predictability (determined by their autocorrelation), had large impacts on the average selection coefficient. The variance in frequency change, which quantifies randomness in population genetics, was substantially higher in a fluctuating environment. The reaction norm of selection coefficients against constant salinity yielded accurate predictions for the mean selection coefficient in a fluctuating environment. This selection reaction norm was in turn well predicted by environmental tolerance curves, with population growth rate against salinity. However, both the selection reaction norm and tolerance curves underestimated the variance in selection caused by random environmental fluctuations. Overall, our results provide exceptional insights into the prospects for understanding and predicting genetic evolution in randomly fluctuating environments.

Being able to predict evolution under natural selection is important for many applied fields of biology, ranging from agriculture to medicine or conservation. However, this endeavor is complicated by factors that inherently limit our ability to predict the future, such as random fluctuations in the environment. Population genetic theory indicates that probabilistic predictions can still be made in this context, but the extent to which this holds empirically, and whether these predictions can be based on simple measurements, are still open questions. Making progress on answering these questions can be achieved by capitalizing on experiments where the environment is precisely controlled over many generations. Here, we used a pipetting robot to generate random time series of salinities with controlled patterns of fluctuations, which we imposed on a microalga, Dunaliella salina. Tracking the frequencies of two genotypes in a mixture by sequencing two short barcode sequences, we were able to show how patterns of fluctuating selection relate to the fluctuating environment. Interestingly, parts of these responses, but not all, could be predicted by simpler measurements in constant environments, allowing precise characterization the limits and prospects for predicting evolution in fluctuating environments.

Introduction

To what extent is evolution predictable? This question has received considerable interest from evolutionary biologists, and has become increasingly quantitative as relevant data have accumulated. Original qualitative arguments about contingency versus necessity [1] or replaying life’s tape [2] have been replaced by more detailed empirical and theoretical investigations on the rate of parallel genetic evolution in replicate lines or populations exposed to similar selective pressures [3–7]. Only recently has predictability in the dynamics (rather than outcome) of evolutionary change gained more prominence [8–11]. This question is indeed of crucial importance for many applications of evolution where the rate of change matters more than the end result, including evolutionary rescue, pest control, antibiotic resistance, and all contexts where strong eco-evolutionary dynamics involve a race between adaptation and population growth or decline [12–16].

A major factor that may alter the predictability of evolution is environmental stochasticity [17,18]. Most natural environments exhibit random fluctuations—also known as stochastic noise—characterized by their variance, which determines their magnitude, and autocorrelation (or color in power spectrum [19,20]), which determines their predictability. Such environmental noise causes randomly fluctuating selection at the genetic and phenotypic levels, which may reduce the predictability of evolution in a number of ways [17,21,22]. First, unaccounted sources of environmental variability (micro-environmental variation) can increase noise in frequency dynamics, thus reducing the precision of selection estimates [23]. And second, even if the environment were perfectly known at a given time, its future would still be uncertain if it fluctuates randomly. Environmental stochasticity thus contributes to chance in evolutionary trajectories, causing allele frequencies to undergo random walks, similarly to genetic drift caused by the finiteness of populations [24–30].

Despite their randomness, stochastic evolutionary dynamics can still be predicted in a probabilistic sense, provided we are able to accurately model them with few parameters. For instance, diffusion approximations (routinely employed to analyze genetic drift [31]) use the magnitude of short-term variance in frequency change conditional on the current frequency (so-called infinitesimal variance [32]) to predict the cumulative influence of stochasticity on allelic frequencies in the long run. For genetic drift, this stochasticity parameter is inversely proportional to the effective population size [31]. Here we wish to measure analog parameters for environmental stochasticity, another major source of chance in evolution [18].

In theoretical models, population genetics in stochastic environments are usually parameterized by the distribution (mean, variance, autocorrelation) of selection coefficients over time, which drives the evolutionary dynamics in this context [24,28,30,33,34]. For instance, the stochastic variance in selection coefficients over one generation (or infinitesimal time step in continuous time) can be used to predict evolutionary outcomes over multiple generations, such as probabilities of fixation [27,29] or expected heterozygosities [28], analogously to the influence of effective population size for genetic drift. However, while the demographic consequences of the magnitude and autocorrelation of environmental variations have been experimentally explored [35–38], and evolutionary experiments have been performed under randomly changing environments [39–43], we are not aware of attempts to measure the stochastic variance of population genetic change under conditions where patterns of random environmental fluctuations have been experimentally manipulated.

Furthermore, even though these parameters of fluctuating selection are the most directly relevant for evolutionary predictions, it would also be extremely useful to be able to project evolutionary change based on how the environment itself fluctuates. Indeed, measurements of selection are complex, time consuming, and costly (often involving substantial sequencing effort), while massive environmental time series can readily be obtained from e.g. the Intergovernmental Panel on Climate Change [44] or collected anew using automated devices such as temperature loggers (Thermochrons [45]). Whether or not these abundant environmental data can be used to project evolutionary change depends on our ability to predict fluctuating selection from a fluctuating environment. The answer to this question is however not straightforward, and several degrees of simplification can be envisioned. First, it may be possible to measure selection coefficients at a few constant values of the environment, producing a form of “selection reaction norm”, which could then be combined with the pattern of environmental fluctuations to project population genetic change. Going one step further, one may simply measure the population growth rates in isolation of all genotypes across environments, to estimate their environmental tolerance curves [37,46,47]. Since selection arises from the differential growths rates of genotypes in competition, changes in selection across environments could then be inferred from genetic differences in tolerance curves (Fig 1), without requiring any sequencing effort to identify genotypes in mixtures. However, the usefulness and limits of such mechanistic links between the environment and selection first need to be evaluated under controlled conditions. For instance, the selection reaction norm approach would be compromised if selection at a given time in a given environment depends not only on the current environment, but also on the sequence of environments a population was exposed to, because of a memory of past environments mediated by phenotypic plasticity [37,40,48]. Similarly, fitness in competition may not be predictable from growth rates in monoculture (as assumed by the tolerance curve approach), if specific interactions between genotypes cause selection to be frequency- or density-dependent [49,50].

PPT PowerPoint slide

PowerPoint slide PNG larger image

larger image TIFF original image Download: Fig 1. Relationship between tolerance curves and selection across environments. Panels a-d represent tolerance curves, with absolute fitness (population growth) in isolation plotted against the environment (here, salinity). In panels a, b and c, the tolerance curve of a reference strain (in gray) is contrasted to that of an alternative genotype or mutant (in color), which varies in only one parameter of the tolerance curve: maximal growth (A), salinity optimum (B), or niche width/breadth (C). Selection for the mutant equals the difference in growth rates between strains (shown as arrows) if fitness is density- and frequency-independent. In addition, the past environment may influence the current tolerance curve of each strain, as illustrated in panel d, where the tolerance curve of the mutant (in black) and reference strain (in gray) vary depending on whether they were transferred from low (plain line, 0.5M) or high (dashed, 4M) salinity. Panel e shows the selection reaction norm for the mutant based on these tolerance curves (hence assuming density- and frequency-independent selection), using the same line types and colors as in panels a-d. The gray normal curve materializes the salinity distribution used in our fluctuating salinity experiment (rescaled vertically for graphical convenience). Panel f plots the resulting distribution of selection coefficients in the fluctuating environment, with line types and colors as in previous panels. Note that the distribution of selection coefficient is Gaussian if both strains have the same tolerance breadth (yellow), but can otherwise be highly skewed. https://doi.org/10.1371/journal.pgen.1009611.g001

To investigate how random environmental fluctuations translate into randomly fluctuating selection, and elucidate the predictability of these responses, we here used experimental evolution in the laboratory, a powerful approach for quantifying the influence of environmental drivers on evolutionary dynamics [51–53], and assess precision limits in the measurement of key evolutionary parameters [23]. We worked with the halotolerant micro-alga Dunaliella salina, which we exposed to randomly fluctuating salinities over a realistic continuous range (instead of the more common practice of switching the environment between low and high constant levels [39–41]). We used a liquid-handling robot to control the mean, variance, and autocorrelation of salinity over many replicate lines. We tracked the frequencies of standing genetic variants through time by Illumina amplicon sequencing of two natural DNA barcodes, as done with engineered barcodes in other studies (BarSeq, [54,55]). We have previously shown that the stochastic demography of these populations was well predicted by combining patterns of environmental variation with short-term salinity tolerance curves [37]. Here, we ask whether, and how, key parameters of stochastic fluctuating selection are influenced by parameters of environmental variation: How much variance in selection is caused by variance in the environment? Does environmental variation also affect the mean selection coefficient? And is there an influence of environmental autocorrelation − which determines the predictability of environmental fluctuations − on patterns of fluctuating selection? We finally ask whether population genetics in a stochastic environment can be predicted by combining environmental time series with simpler population measurements, such as selection reaction norms and tolerance curves.

[END]

[1] Url: https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1009611

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.