Predictable genome-wide sorting of standing genetic variation during parallel adaptation to basic versus acidic...

Genomic studies of parallel (or convergent) evolution often compare multiple populations diverged into two ecologically different habitats to search for loci repeatedly involved in adaptation. Because the shared ancestor of these populations is generally unavailable, the source of the alleles at adaptation loci, and the direction in which their frequencies were shifted during evolution, remain elusive. To shed light on these issues, we here use multiple populations of threespine stickleback fish adapted to two different types of derived freshwater habitats-basic and acidic lakes on the island of North Uist, Outer Hebrides, Scotland-and the present-day proxy of their marine ancestor. In a first step, we combine genome-wide pooled sequencing and targeted individual-level sequencing to demonstrate that ecological and phenotypic parallelism in basic-acidic divergence is reflected by genomic parallelism in dozens of genome regions. Exploiting data from the ancestor, we next show that the acidic populations, residing in ecologically more extreme derived habitats, have adapted by accumulating alleles rare in the ancestor, whereas the basic populations have retained alleles common in the ancestor. Genomic responses to selection are thus predictable from the ecological difference of each derived habitat type from the ancestral one. This asymmetric sorting of standing genetic variation at loci important to basic-acidic divergence has further resulted in more numerous selective sweeps in the acidic populations. Finally, our data suggest that the maintenance in marine fish of standing variation important to adaptive basic-acidic differentiation does not require extensive hybridization between the marine and freshwater populations. Overall, our study reveals striking genome-wide determinism in both the loci involved in parallel divergence, and in the direction in which alleles at these loci have been selected.