Extensive linkage disequilibrium and parallel adaptive divergence across threespine stickleback genomes

Abstract

Population genomic studies are beginning to provide a more comprehensive view of dynamic genome-scale processes in evolution. Patterns of genomic architecture, such as genomic islands of increased divergence, may be important for adaptive population differentiation and speciation. We used next-generation sequencing data to examine the patterns of local and long-distance linkage disequilibrium (LD) across oceanic and freshwater populations of threespine stickleback, a useful model for studies of evolution and speciation. We looked for associations between LD and signatures of divergent selection, and assessed the role of recombination rate variation in generating LD patterns. As predicted under the traditional biogeographic model of unidirectional gene flow from ancestral oceanic to derived freshwater stickleback populations, we found extensive local and long-distance LD in fresh water. Surprisingly, oceanic populations showed similar patterns of elevated LD, notably between large genomic regions previously implicated in adaptation to fresh water. These results support an alternative biogeographic model for the stickleback radiation, one of a metapopulation with appreciable bi-directional gene flow combined with strong divergent selection between oceanic and freshwater populations. As predicted by theory, these processes can maintain LD within and among genomic islands of divergence. These findings suggest that the genomic architecture in oceanic stickleback populations may provide a mechanism for the rapid re-assembly and evolution of multi-locus genotypes in newly colonized freshwater habitats, and may help explain genetic mapping of parallel phenotypic variation to similar loci across independent freshwater populations.

Figure 1.

Genomic distributions of haplotype structure and recombination rates. Vertical grey shading and Roman numerals at the top indicate the 21 linkage groups (LGs; unassembled scaffolds not shown). (a) Integrated haplotype homozygosity (iHH), a measure of the decay of LD from a locus, within the combined oceanic population. Units of iHH are megabases (integral under the EHH curve; see text for details). (b) iHH in the freshwater population Boot Lake (BL). (c) Cross-population extended haplotype homozygosity (XP-EHH) comparing the oceanic with BL populations. Positive values indicate larger values of LD in BL, and negative values indicate larger values in the ocean. (d) Recombination rate estimated from an oceanic by freshwater F₂ cross using a hidden Markov model. Grey lines indicate upper and lower boundaries of the Bayesian 95 per cent credible interval. Three apparent peaks of recombination greater than 40 cM Mb⁻¹, most likely the result of chromosomal rearrangements relative to the reference genome sequence, have been truncated in this plot; the height of these peaks in cM Mb⁻¹ is given above each one.

Figure 2.

Haplotype structure, recombination and differentiation along LG IV. (a) iHH in the oceanic (blue) and BL (red) populations. (b) Recombination rate in an F₂ cross. Black dots represent the location of markers used. Grey lines indicate the upper and lower boundaries of a Bayesian 95 per cent credible interval about the estimate. Two extreme peaks in apparent recombination rate are truncated (see figure 1). (c) Population differentiation between BL and the combined oceanic population (black), and between all three freshwater populations and the oceanic population (green). Bars above the plot indicate regions of p < 10⁻⁵ bootstrap significance (see Hohenlohe et al. [90] for details).

Figure 3.

Haplotype structure, recombination and differentiation along LG VII. All details as in figure 2.

Figure 4.

Long-distance LD and population differentiation on (a) LG IV and (b) LG VII. Pair-wise LD among loci, measured as D′, estimated in the combined oceanic (above the diagonal) and freshwater Boot Lake (below the diagonal) populations. The yellow rectangle highlights a region of potential long-distance LD between two genomic regions previously implicated in adaptation to fresh water. Differentiation between oceanic and freshwater populations, adapted from Hohenlohe et al. [90], is shown along the bottom and side. The black line compares the oceanic populations with BL, and the green line compares the oceanic populations with three independently derived freshwater populations including BL. Bars above the plot indicate regions of p < 10⁻⁵ bootstrap significance (see Hohenlohe et al. [90] for details). A key to the colour scheme for D′ is shown at the bottom.

Figure 5.

(a) Long-distance LD between LGs IV and VII in the combined oceanic population. Plots of population differentiation (F_ST) on the left and the bottom are as in figure 4 (adapted from Hohenlohe et al. [90]). Three areas of pair-wise LD between previously identified adaptive genomic regions are highlighted by yellow rectangles. (b) Histogram of D′ values for all inter-chromosomal pairs of SNPs in the combined oceanic population. Labels on the horizontal axis represent the upper limit of bins of width 0.05. The three areas highlighted in (a) occupy the far right bin at D′ = 1.0.