Relationships between environmental variables and mucus transcriptomic profiles
As the transcriptome responds to environmental changes, differences and similarities in environmental variables may explain mucus transcriptome variation among lake trout populations (Oomen & Hutchings, 2017). The high degree of transcriptomic similarity among L224 and L373 populations could potentially be explained by the fact that both lakes were fairly oligotrophic whereas L260 and L223, while still oligotrophic, tended to be more productive. However, comparison of results from the water quality PCA and transcriptome PCA suggests there are environmental factors not accounted for in the study. For example, while L260 and L373 were most similar in their water quality characteristics (Figure 6), lake trout from those lakes deviated from each other in terms of their transcriptomic profiles (Figure 2). Therefore, it is likely that there were other abiotic and biotic variables influencing gene expression changes other than those taken into account in the water quality PCA, such as pathogen loading, epigenetics, or food availability.
While PCA was used to characterize similarity in water quality profiles among lakes, a linear mixed model was used to quantify the percent variance in transcriptome-wide gene expression explained by those water quality variables. As the linear mixed model accounted for 68.1% of transcriptional variance, results suggest that variation in water quality contributes to transcriptional differences among lake trout populations (Figure 7). Conductivity, chlorophyll a, alkalinity, and dissolved oxygen explained the greatest proportion of variance within the transcriptome, suggesting these variables had the greatest influence on gene expression differences among lakes (Figure 7). Variables identified by the linear mixed model as accounting for the most transcriptomic variation among lake trout populations were also among the variables identified in the PCA as driving inter-lake water quality differences. However, a median of 31.9% of transcriptomic variance was not explained by the water quality variables in the model. Therefore, both the water quality PCA and linear mixed model suggest that the variables considered in the present study did not fully explain transcriptional variance and, thus, other variables could be driving transcriptomic differences among populations albeit similar water quality conditions. As suggested by differential gene expression data, an additional source of transcriptional variance among populations may stem from presence of pathogens. For example, despite their fairly similar water quality profiles, the divergence in transcriptomic profiles of L223 and L260 lake trout may be partially explained by the strong immune response in L223. Overall, it is challenging to robustly quantify the effects of environmental variables on gene expression patterns as transcription may be affected by even small changes in environment (Alvarez et al., 2015).
Aside from environmental factors, epigenetic modifications are another potential source of gene expression variation among populations, with studies showing that genes may be transgenerationally dysregulated following exposure to stressors in prior generations. Transgenerational alterations in the transcriptome are well-illustrated in a study which examined responses to increased temperature across generations in common reef fish (Acanthochromis polyacanthus ) (Veilleux et al., 2015). Compared to controls, fish that were transgenerationally exposed to higher temperatures showed upregulation of metabolic, immune, and stress-responsive genes, likely an adaptive mechanism to maintain performance and cope with higher temperatures. Epigenetic modifications are highly possible in lake trout populations at IISD-ELA as L260 and L223 have been used for whole-lake experiments in the past. In an effort to understand population-level impacts of environmental estrogens, L260 was subject to whole-lake ethynylestradiol additions for three years starting in 2001, resulting in feminization and a near extirpation of fathead minnow (Pimephales promelas ) as well as population declines of lake trout (Kidd et al., 2007; Palace et al., 2009). To simulate the effects of acid precipitation on freshwater lakes, sulfuric acid was added to L223 over a three year period from 1976 to 1978 (Schindler, Wagemann, Cook, Ruszczynski, & Prokopowich, 1980), resulting in disappearance of fathead minnow, increased lake trout embryonic mortality and deformities, and decreased survival of lake trout due to emaciation (Kennedy, 1980; Mills, Chalanchuk, Mohr, & Davies, 1987; Nero & Schindler, 1983; Schindler & Turner, 1982). Interestingly, L224 and L373, which have not been subject to experimental modifications, had highly similar transcriptomic profiles, whereas the experimentally-modified L260 and L223 diverge from all other lakes (Figure 2); these observations suggest that epigenetic modifications in response to past stressful conditions may contribute to the observed transcriptional variations.
As the transcriptome reflects both short-term plastic responses and transgenerational plasticity (Oomen & Hutchings, 2017), it is difficult to adequately discern whether transcriptional variation among populations stems from environmental changes or transgenerational alterations. Although concrete identification of explanatory variables requires further development, RNA sequencing of mucus provides meaningful information on transcriptomic variation among fish populations and can be used to identify molecular differences driving variation in a nonlethal manner.