Adaptation and the Covariance Among SNPs and Population Differentiation in Puget Sound Chinook

Chinook salmon spawning in rivers draining into Puget Sound and Strait of Juan de Fuca, Washington are currently considered parts of 22 populations, with spring, summer, and fall adult run times, and sub-yearling and yearling juvenile life histories.  Many populations are supplemented by hatchery production, and most populations with fall run times have been genetically introgressed by individuals with Soos Creek Hatchery (Green River) ancestry.  The WDFW Molecular Genetics Laboratory genotyped 3700 individual Chinook salmon from 24 hatchery and wild populations using 96 SNPs, 75 of which are part of the GAPS laboratory consortium standardized SNP Taqman assays.  SNPs were genotyped using either the Fluidigm EP1 or Applied Biosystems 7900 systems.  We used Principal Component Analysis to examine the covariance structure among 77 of the 96 SNPs.  The first two principal components accounted for 65% of the among-population genetic variance and separated most of the populations into three clusters: (1) populations from north Puget Sound with spring and summer adult run times, (2) populations from south Puget Sound and Hood Canal, with fall adult run times and populations dominated by Soos Creek Hatchery lineage, and (3) populations in the Elwha and Dungeness River, both of which drain into the Strait of Juan de Fuca.  Clusters 1 and 2 separated mostly along the first component, and Cluster 3, while trending towards Cluster 1 along the first component, was well differentiated from the other clusters along the second component.  By using the correlation between the vectors that defined these clusters and the first two eigenvectors, plotted using polar coordinates, we associated specific SNP loci with population differentiation.  Twenty seven of the 77 SNPs were significantly correlated with the three clusters of populations, with 13 SNPs differentiating the Puget Sound clusters, and 14 different SNPs differentiating the Strait of Juan de Fuca cluster from the two Puget Sound clusters.  Only one of the 27 SNPs showed evidence of directional selection, using an FST outlier test.  We projected onto the principal component plane defined by these Puget Sound populations Chinook samples from four outgroup populations.  The ordination of all populations along the first two principal components suggests that the overall covariance structure among these SNPs is derived in Cluster 2 populations.