Global adaptation complicates the interpretation of genome scans for local adaptation

Abstract

Spatially varying selection promotes variance in allele frequencies, increasing genetic differentiation between the demes of a metapopulation. For that reason, outliers in the genome-wide distribution of summary statistics measuring genetic differentiation, such as F_ST , are often interpreted as evidence for alleles that contribute to local adaptation. However, theoretical studies have shown that in spatially structured populations the spread of beneficial mutations with spatially uniform fitness effects can also induce transient genetic differentiation. In recent years, numerous empirical studies have suggested that such species-wide, or global, adaptation makes a substantial contribution to molecular evolution. In this perspective, we discuss how commonly such global adaptation may influence the genome-wide distribution of F_ST and generate genetic differentiation patterns, which could be mistaken for local adaptation. To illustrate this, we use forward-in-time population genetic simulations assuming parameters for the rate and strength of beneficial mutations consistent with estimates from natural populations. We demonstrate that the spread of globally beneficial mutations in parapatric populations may frequently generate F_ST outliers, which could be misinterpreted as evidence for local adaptation. The spread of beneficial mutations causes selective sweeps at flanking sites, so in some cases, the effects of global versus local adaptation may be distinguished by examining patterns of nucleotide diversity within and between populations in addition to F_ST . However, when local adaptation has been only recently established, it may be much more difficult to distinguish from global adaptation, due to less accumulation of linkage disequilibrium at flanking sites. Through our discussion, we conclude that a large fraction of F_ST outliers that are presumed to arise from local adaptation may instead be due to global adaptation.

Keywords: FST outlier; genetics of adaptation; genome scans; global adaptation; local adaptation.

Figure 1

Selective sweeps of globally beneficial alleles can generate ephemeral F_ST peaks. (A) Manhattan plot of F_ST calculated between parapatric populations subject only to global adaptation. F_ST was calculated in sliding windows of 10,000 bp with a step size of 500 bp. (B) The allele frequency of beneficial mutations in deme 1 (solid lines) and deme 2 (dashed lines). (C) F_ST over time for 10,000 bp analysis windows containing beneficial alleles. The vertical red line indicates the time point for which the Manhattan plot in A) was constructed. The dashed gray horizontal line indicates the 99.999th percentile of F_ST from neutral simulations. Simulation parameters, N_e = 2000 diploid individuals, ${\overline{S}}_a$ = 0.1, p_a = 0.0001, Nm = 1.

Figure 2

The number F_ST outliers per 10,000 windows for parapatric populations subject to global adaptation. The text on the horizontal axis details the proportion of adaptive substitutions (α) observed in simulated populations as well as parameters of the distribution of fitness effect assumed. The mean effect of beneficial mutations is given by ${\overline{\mathbf{S}}}_{\mathit{a}}$ and the proportion of new mutations that were beneficial is given by p_a. F_ST was calculated for 10,000 bp analysis windows centered on simulated “gene‐like regions” using the method of Weir and Cockerham (1984). Outliers were determined using the 99.99th percentile of the distribution of F_ST from neutral simulations. Plusses indicate the point estimate and violins indicate the distribution of 1000 bootstraps samples from 2000 simulated datasets. Populations were simulated with N = 5000 diploid individuals per deme.

Figure 3

Summaries of population genomic data in regions surrounding F_ST outliers generated by global adaptation, local adaptation, and genetic drift. Weir and Cockerham's F_ST and nucleotide diversity between (d_XY) and within ($\overline{{\pi }_W}$) populations are shown. We identified 100 F_ST outliers from simulations of each of the four processes shown in the plot. Summary statistics were calculated in 10,000 bp sliding windows with a 500 bp step. In the cases of global and local adaptation, alleles with s_a = 0.05 were simulated. In all simulations shown Nm = 1.