Local adaptation is often facilitated by loci clustered in relatively few regions of the genome, termed genomic islands of divergence. However, the mechanisms that create, mold, and maintain these islands are poorly understood. Here, we use sockeye salmon as a model species to investigate the mechanisms responsible for creating islands of divergence linked to adaptive variation. Previous research suggests that multiple islands are involved in adaptive radiation of sockeye salmon. However, these studies were based on low-density genomic methods that genotyped tens to thousands of loci, making it difficult to elucidate the mechanisms responsible for islands. We used whole genome resequencing to genotype millions of loci to investigate these mechanisms. We discovered 64 islands, 16 of which were shared between two isolated populations; these 16 islands were clustered in four genomic regions. Characterization of the shared regions suggested that three of four were likely created by chromosomal inversions, while the other was created by processes not involving structural variation. Additionally, all four regions were relatively small (< 600 kb), suggesting inversions and other low recombination regions do not have to span megabases to be important for adaptive divergence. In sum, our study demonstrates that heterogeneous selection can lead to a mosaic of islands created by different mechanisms within the same genome. Future studies should continue to investigate how gene flow, selection, and the architecture of genetic traits interact to influence the genomic landscape of adaptive divergence.

Introduced and invasive species make excellent natural experiments for investigating rapid evolution. Here, we describe the effects of genetic drift and rapid genetic adaptation in pink salmon (Oncorhynchus gorbuscha) that were accidentally introduced to the Great Lakes via a single introduction event 31-generations ago. Using whole-genome resequencing for 134 fish spanning five sample groups across the native and introduced range, we estimate that the progenitor population’s effective population size was 146,886 at the time of introduction, whereas the founding population’s effective population size was just 72—a 2040-fold decrease. As expected with a severe bottleneck, we show reductions in genome-wide measures of genetic diversity, specifically a 37.7% reduction in the number of SNPs and an 8.2% reduction in observed heterozygosity. Despite this decline in genetic diversity, we provide evidence for putative selection at 47 loci across multiple chromosomes in the introduced populations, including missense variants in genes associated with circadian rhythm, immunological response, and maturation, which match expected or known phenotypic changes in the Great Lakes. For one of these genes, we use a species-specific agent-based model to rule out genetic drift and conclude that a strong response to selection occurred in a period gene (per2) that plays a predominant role in determining an organism’s daily clock, matching large day length differences experienced by introduced salmon during important phenological periods. Together, these results inform how populations might evolve rapidly to new environments, even with a small pool of standing genetic variation.