Genomic time-series data show that gene flow maintains high genetic diversity despite substantial genetic drift in a butterfly species

Abstract

Effective population size affects the efficacy of selection, rate of evolution by drift, and neutral diversity levels. When species are subdivided into multiple populations connected by gene flow, evolutionary processes can depend on global or local effective population sizes. Theory predicts that high levels of diversity might be maintained by gene flow, even very low levels of gene flow, consistent with species long-term effective population size, but tests of this idea are mostly lacking. Here, we show thatLycaeidesbutterfly populations maintain low contemporary (variance) effective population sizes (e.g., ∼200 individuals) and thus evolve rapidly by genetic drift. Contemporary effective sizes were consistent with local census populations sizes. In contrast, populations harbored high levels of genetic diversity consistent with an effective population size several orders of magnitude larger. We hypothesized that the differences in the magnitude and variability of contemporary versus long-term effective population sizes were caused by gene flow of sufficient magnitude to maintain diversity but only subtly affect evolution on generational time scales. Consistent with this hypothesis, we detected low but non-trivial gene flow among populations. Furthermore, using population-genomic time-series data, we documented patterns consistent with predictions from this hypothesis, including a weak but detectable excess of evolutionary change in the direction of the mean (migrant gene pool) allele frequencies across populations, and consistency in the direction of allele frequency change over time. The documented decoupling of diversity levels and short-term change by drift inLycaeideshas implications for our understanding of contemporary evolution and the maintenance of genetic variation in the wild.

Spatial variation in population genomic responses to over a century of anthropogenic change within a tidal marsh songbird

“…Morro Bay and Newport Bay in southern California were not significantly different from 0 (no correlation). These patterns were consistent with the expectation that greater gene flow leads to greater correlation between the direction and magnitude of allele frequency change (Gompert et al 2021).…”

“…Further, if outlier loci associated with local adaptation to tidal marshes exhibit increased resilience to gene flow then we would expect less homogenization of allele frequencies at these loci relative to non-outlier loci. We tested these predictions by estimating the degree of correlation between historical deviation in allele frequency and change in allele frequency through time for each SNP (Gompert et al 2021). In this analysis, stronger correlations will reflect greater levels of gene flow.…”

Monitoring genetic diversity with new indicators applied to an alpine freshwater top predator

“…We observe a general trend of small local N e , but with higher metapopulation N e , and migration rates between subpopulations are crudely estimated as ~0.5–2.5 individuals per generation (assuming an island model of migration). These findings suggest that the metapopulation structure buffers the lake systems against loss of genetic diversity and that protected areas need to be large enough to support a large meta‐ N e (compare with Gompert et al, 2021; Jorde & Ryman, 1996).…”

Laplace's demon in biology: Models of evolutionary prediction

“…Phenotypic effects were assigned to the loci by sampling from a standard normal distribution, and allele frequencies for each bi-allelic locus were drawn from a beta distribution, beta(0.6,0.6) (this gives a U-shaped distribution of allele frequencies). Then, for each simulation of evolution, the subset of causal variants was determined by sampling loci according to their probabilities of association (analogous to posterior inclusion probabilities from Bayesian polygenic models for genomic prediction, see, e.g., Zhou et al, 2013;Gompert, 2021). Sampled loci were assigned their respective effect sizes, and other loci were assigned an effect size of 0.…”

“…, where b i is the average excess of locus i, σ 2 z denotes the phenotypic variance and was set to 2 to give a heritability of about 0.5 (the exact value varied based on the specific causal loci), and S σ 2 z equals the selection gradient β (Kimura & Crow, 1978;Walsh & Lynch, 2018;Gompert, 2021). This approximation assumes the trait remains normally distributed, effect sizes are small, and causal loci are unlinked.…”