Transcriptomic variation among lakes
In order to visualize similarities in transcriptomes among lake trout populations, hierarchical clustering and PCA was conducted on transcriptome-wide count data. Results of sample-to-sample Euclidian distance clustering indicated an approximate clustering of L223 and L373 populations and clustering of L260 and L223 populations within separate clades (Figure S1). Transcriptomic differences among the lakes were further elucidated by the PCA of transcriptome-wide count data, which clearly showed that samples clustered by lake from which they were collected (Figure 2A). Again, L224 and L373 lake trout clustered more closely together, whereas L260 and L223 lake trout clustered apart from all other lakes. The eigenvalues of the two first principal components represented 53% of the total variance (PC1 33%; PC2 20%). There were no patterns based on sex of the lake trout in the PCA results, suggesting transcriptomic patterns were most likely driven by lake (Figure S2).
Hierarchical clustering of the top 500 most-variable transcripts revealed similar results, with individuals clustering into clearly defined clades by lake (Figure 2B). As with the PCA results, L224 and L373 lake trout clustered together, whereas L260 and L223 each clustered apart from the other lakes. These data suggest that individuals from L224 and L373 were most similar in terms of transcriptomic profiles, with L260 and L223 being the most different from the other lakes. Results of the hierarchical clustering and PCA overall suggest that there were clear transcriptomic differences among individuals collected from different lakes and that these differences can be detected nonlethally in epidermal fish mucus.