Fig 1.
Evolutionary relationships and patterns of genome-wide variation across the radiation.
(A) The black tree was constructed from a concatenated alignment of genome-wide SNPs and is rooted using M. clevelandii. The 387 gray trees were constructed from 500-kb genomic windows. The first number associated with each node or taxon is the bootstrap support for that clade in the whole-genome tree, and the second number is the percentage of window-based trees in which that clade is present. Flower photos were taken by the authors. (B) Levels of differentiation (FST), (C) divergence (dxy), and (D) diversity (π) within and among taxa based on the same 500-kb windows. For simplicity, FST and dxy are shown only for comparisons with the red ecotype of subspecies puniceus. See S7 Fig for the distributions of FST and dxy across all pairs of taxa.
Fig 2.
Common genomic landscapes mirror variation in the local properties of the genome.
(A) Tree concordance scores for 500-kb nonoverlapping genomic windows plotted across the 10 bush monkeyflower chromosomes. (B–D) Plots of PC1 for FST, dxy, and π in overlapping 500-kb windows (step size = 50 kb). PC1 explains 66%, 70%, and 85% of the variation in FST, dxy, and π, respectively, and is Z-transformed such that above average values have positive values and below average values have negative values. (E–F) Gene count and recombination rate (cM/Mbp) in overlapping 500-kb windows. (G) Mean fd (admixture proportion) in 500-kb nonoverlapping genomic windows. See S4 Fig for the same plot made for 100-kb windows. LG, linkage group; PC1, first principal component.
Fig 3.
Correlations between measures of diversity and intrinsic features reveal the impact of heterogeneous indirect selection.
Matrix of pairwise correlations between PC1 FST, PC1 dxy, PC1 π, tree concordance, mean fd, gene density, and recombination rate, all estimated in 500-kb nonoverlapping windows. The heat map indicates the strength of the correlation and its sign. All correlations are statistically significant at p < 0.001. (See S8 Fig for a more detailed correlation matrix and S9 Fig for the correlation matrix from 100 kb windows.) PC1, first principal component.
Fig 4.
A large-effect adaptive locus shows a lineage-specific signature of positive selection.
Plots of Z-transformed FST across the genome, estimated in 500-kb sliding windows (step size 50 kb). The red line shows values between the red and yellow ecotypes of subspecies puniceus, whereas the gray lines show the values of all other comparisons. The dashed blue line shows PC1 calculated across all of the comparisons. The triangle marks the position of the gene MaMyb2. A cis-regulatory mutation that is tightly linked to this gene is responsible for the shift from yellow to red flowers. See S10 Fig for the same plot made for 100 kb windows. LG, linkage group; PC1, first principal component.
Fig 5.
The range of divergence times reveals static and dynamic signatures of recurrent indirect selection.
Correlations between variables (500-kb windows) for all 36 taxonomic comparisons (gray dots) plotted against the average da as a measure of divergence time. The left panels show how the relationships between π (each window averaged across a pair of taxa) and (A) gene count and (B) recombination rate vary with increasing divergence time. The middle panels (C and D) show the same relationships but with FST. The right panels show the relationships between (E) dxy and π and (F) FST and π. The regressions (dashed lines) in each plot are fitted to the 8 independent contrasts (colored points) obtained using a phylogenetic correction. The color gradient shows the strength of the correlation.
Fig 6.
Emergence of a heterogeneous differentiation landscape across 1 million years of divergence.
(A) Plots of FST (500-kb windows) across the genome for pairs of taxa at early (red versus yellow), intermediate (red versus parviflorus), and late stages (red versus M. clevelandii) of divergence. (B) Average nucleotide diversity (for the red ecotype of subspecies puniceus, yellow ecotype of subspecies puniceus, subspecies parviflorus, and M. clevelandii) across the genome in 500-kb windows. The geographic distribution (parapatric or allopatric), sequence divergence (da × 10−2), and correlation between FST and mean π are provided next to each taxon pair. LG, linkage group.
Fig 7.
Genomic landscapes simulated under different divergence histories.
Each row of plots shows patterns of within- and between-population variation (π, dxy, and FST) across the chromosome (500-kb windows) at 5 time points (N generations, where N = 10,000) during one of the scenarios The selection parameter (Ns, where s = Ns/N), proportion of deleterious (−) and positive mutations (+), and number of migrants per generation (Nm; 0 unless stated) for these simulations are as follows: (i) neutral divergence (no selection), (ii) BGS (−Ns = 100; −prop = 0.1), (iii) BDMI (−Ns = 100, −prop = 0.05, Nm = 0.1), (iv) positive selection (+Ns = 100, +prop = 0.001), (v) BGS and positive selection (−Ns = 100, −prop = 0.1; +Ns = 100, +prop = 0.005), and (vi) local adaptation (+Ns = 100, +prop = 0.001, Nm = 0.1). The gray boxes in the first column show the areas of the chromosome that are experiencing selection, while the white central area evolves neutrally. Note that π (in populations a and b) and dxy have been mean centered so they can be viewed on the same scale. Uncentered values and additional simulations with different parameter combinations and more time points can be found in S13 Fig. BDMI, Bateson-Dobzhansky-Muller incompatibility; BGS, background selection.