Transcriptomic signatures of population divergence in two independent evolutionary lineages
We identified many genes differentially expressed between high- and low-predation populations in each river drainage. The absolute number of population DE transcripts was smaller in the Aripo drainage as compared to the Quare drainage, but the number of developmental and interaction differences was greater in the Aripo drainage. The Quare dataset was 1.5X larger than the Aripo dataset (40 versus 60 total samples) and the identification of a greater number of DE genes in the Quare dataset is therefore likely in part related to greater statistical power. This difference may also be biological in origin, as high- and low-predation populations in the Quare drainage show greater genetic divergence than those in the Aripo drainage (Willing et al., 2010). Indeed, we note that the larger number of genes in the Quare dataset come from a large difference the number of genes with population, but not rearing or interaction, effects. In sum, we suggest that the difference in the number of differentially expressed transcripts between datasets is likely a combination of greater statistical power in the larger Quare dataset and the greater degree of genetic divergence in the Quare as compared to the Aripo river drainage (Willing et al., 2010).
When we compared DE gene sets between drainages, we found a small, but significant, number of population and rearing DE genes shared across lineages. However, the majority of DE genes were non-shared across drainages for both population (Aripo: 78%; Quare: 96%) and rearing (Aripo: 94%; Quare: 92%) effects. Only rearing genes showed a significant association in expression direction, with 67% of genes concordantly differentially expressed between lineages. In contrast, only 52% of the genes that diverged when low-predation fish colonized high-predation habitats in both drainages were concordantly differentially expressed. We suggest that these patterns across drainages point to a small number of core genes that exhibit predictable, plastic expression responses upon colonization of low-predation environments, but that lineage-specific selection pressures, differences in genetic background, non-adaptive processes (e.g. drift, inbreeding, founder effects), and alternative compensatory gene expression responses give rise to largely non-overlapping, non-concordant expression differences associated with parallel phenotypic adaptation across drainages.
A previous study in guppies performed a similar comparison of gene expression changes associated with adaptation to low-predation environments (Ghalambor et al., 2015) and found a strong signal of concordant differential expression in genes differentially expressed based on population of origin. Whereas the present study compared long-term natural population divergence across drainages, Ghalambor et al. (2015) characterized early stages of adaptation of experimentally introduced low-predations populations derived from founders from a single high-predation source population within the same drainage. These contrasting findings in comparisons of population pairs within the same drainage versus across drainages highlight the impacts of standing genetic variation within the source population on mechanisms of divergence (Feiner, Rago, While, & Uller, 2017; Thompson, Osmond, & Schluter, 2019), particularly at early stages of evolution (Barrett & Schluter, 2008): while alternative transcriptional ‘solutions’ are possible, shared genetic background appears to bias evolutionary outcomes toward shared patterns.
Both adaptive and non-adaptive processes may contribute to the combination of shared and distinct transcriptional mechanisms we find associated with parallel, adaptive life-history, morphological, and behavioral phenotypes across lineages in guppies. First, as described above, differences in standing genetic variation likely influence which mechanisms are available to selection in response to common environmental conditions in different drainages (Barrett & Schluter, 2008; Thompson et al., 2019). Second, low-predation populations are typically established by a very small number of individuals (Barson et al., 2009; Fraser et al., 2015; Willing et al., 2010), making them susceptible to the unpredictable, non-adaptive influences of founder’s effects, genetic drift, and/or inbreeding on gene expression divergence – although we note that we found no more evidence for shared mechanisms among genes most likely under selection than among all diverged genes (i.e. genes with PST > FST). Third, the large number of significantly evolved genes that did not overlap between drainages may also represent adaptive responses to drainage- or site-specific environmental factors other than predation (Fitzpatrick, Torres-Dowdall, Reznick, Ghalambor, & Chris Funk, 2014; Zandonà et al., 2011). Finally, alternative compensatory or homeostatic gene expression responses may arise in response to any of the above factors, leading to alternative transcriptional configurations associated with similar higher-level phenotypes. In other words, genetic similarity among ancestral populations may channel low-predation populations within the same drainage toward shared transcriptional solutions (as in Ghalambor et al. 2015), while differences in standing genetic variation, drainage-specific environmental conditions, founder effects, and alternative compensatory changes could result in distinct mechanistic paths to arrive at shared organism-level phenotypes. We cannot definitely distinguish causal from non-adaptive and compensatory gene expression differences under these scenarios – indeed, it is likely a combination of these factors that contribute to distinct transcriptional patterns associated with parallel adaptation. Nonetheless, in either case, alternative transcriptional patterns suggest that mechanistic flexibility and ‘many-to-one’ mapping of gene expression to organism level phenotypes may facilitate adaptation.