Overlap in identity and expression direction of DE gene sets within lineages
To determine if transcripts exhibiting expression plasticity in response to rearing with predators were more likely to also exhibit evolutionary divergence between high- and low-predation populations, we evaluated overlap in transcripts differentially expressed between populations and between rearing conditions as well as their directions of expression change. To test the hypothesis that initial plastic responses following colonization of low-predation environments might drive subsequent divergence, we specifically compared pattern of ancestral plasticity (i.e. high-predation fish reared with and without predators) to genetic divergence (high- versus low-predation fish reared in the novel predator-free environment) within each drainage. We used post-hoc tests of simple effects to identify DE genes in these comparisons and estimate log-fold gene expression changes. We then used log-fold expression changes to determine whether gene expression differences were in the same/concordant direction (i.e. upregulated in high-predation populations and in response to rearing with predator cues, or down-regulated in high-predation populations and in response to rearing with predator cues) or opposite/non-concordant direction (upregulated in high-predation populations but down-regulated in response to rearing with predators, or vice versa).
Within each drainage, we compared the overlap in genes with both statistically significant genetic and plastic expression changes. We compared the association between directions of plasticity and divergence (concordant vs. non-concordant) in (1) those genes with both significant ancestral plasticity and significant population differences, and (2) all transcripts with significant population differences in expression, including those without significant expression plasticity. The second comparison addressed the possibility that even subthreshold expression plasticity in response to rearing environment could influence patterns of expression divergence, especially given the high false negative rate in our conservative DE analyses. Because these comparisons rely on two comparisons to a single group (high predation fish reared without exposure to predator cues), the estimates of plasticity and divergence are dependent. We adapted a parametric bootstrap method (Efron & Tibshirani, 1993) similar to the approach implemented by (W. C. Ho & Zhang, 2019) but accounting for conditional probabilities. We tested whether the conditional probabilities of gene expression divergence differed for genes that exhibited ancestral plasticity and those that did not. We further tested whether conditional probabilities for upregulated or downregulated genes in the low predation population compared to the high predation population differed depending on the direction of plasticity. We calculated parametric bootstrap confidence intervals for these differences in conditional probabilities as detailed in the Supplemental Methods, with confidence intervals that do not include zero indicating association between divergence and plasticity.