Introduction

The repeated evolution of similar phenotypes across independent lineages in response to shared environmental conditions (i.e., parallel evolution) provides strong evidence for natural selection (Manceau, Domingues, Linnen, Rosenblum, & Hoekstra, 2010; Rosenblum, Parent, Diepeveen, Noss, & Bi, 2017; Torres-Dowdall et al., 2017). Cases of parallel evolution have been described in a wide array of organisms across the tree of life (Colosimo et al., 2005; Mahler, Ingram, Revell, & Losos, 2013; Rosenblum et al., 2017; Sage, Christin, & Edwards, 2011). Parallel evolution was initially studied on the phenotypic level but recently focus has shifted towards identifying examples on the molecular level (Stern, 2013). Phenotypic parallelism can be the product of mutations in the same gene (Chan et al., 2010; Rosenblum, Rompler, Schoneberg, & Hoekstra, 2010; Steiner, Rompler, Boettger, Schoneberg, & Hoekstra, 2009; Zhen, Aardema, Medina, Schumer, & Andolfatto, 2012) or involve many changes across the genome (Jones, Grabherr, et al., 2012; Ravinet et al., 2016; Rennison, Stuart, Bolnick, & Peichel, 2019) which results in a broad signature of parallel genomic divergence. Identifying examples of genetic and genomic parallelism improved our general understanding of parallel evolution (Arendt & Reznick, 2008; Manceau et al., 2010); repeated use of the same genes or genomic regions can suggest a source of genetic bias or constraint (reviewed in Bolnick, Barrett, Oke, Rennison, & Stuart, 2018) and the reuse of genes or regions can also be leveraged to identify candidate loci important for adaptation.
It is becoming clear that the magnitude of repeatability of genome-wide parallelism varies considerably across study systems (Jones, Chan, et al., 2012; Le Moan, Gagnaire, & Bonhomme, 2016; Ravinet et al., 2016). For example, species pairs of sunflowers that diverged along latitudinal gradients (Renaut, Owens, & Rieseberg, 2014) show high levels of genomic parallelism whereas little evidence for genomic parallelism is found in repeated adaptive radiations of Nicaraguan crater lake cichlid fishes (Kautt, Elmer, & Meyer, 2012). Within a species, population pairs can also vary in their magnitude of parallelism (e.g., Ravinet et al., 2016; Rennison et al., 2019). Threespine stickleback (Gasterosteus aculeatus ) population pairs from adjacent lake and stream habitats in Canada show multiple highly divergent genomic regions. A substantial portion of these divergent regions (37%) is shared among independently evolved lake-stream pairs. In contrast, lake-stream pairs from Europe share only 3% of divergent regions (Feulner et al., 2015; Rennison et al., 2019). In the rough periwinkle (Littorina saxatilis ), parallelism ranges from 8-34 % of outliers, depending on the populations compared (Kess, Galindo, & Boulding, 2018; Ravinet et al., 2016). Both spatial proximity and ecological similarity seem to be key predictors of the overall magnitude of genome-wide parallelism (Morales et al., 2019; Rennison, Delmore, Samuk, Owens, & Miller, 2020). A recent study on threespine stickleback from different areas of their global distribution further emphasized that the demographic history and previous selection can affect levels of genomic repeatability (Fang, Kemppainen, Momigliano, Feng, & Merila, 2020). Taken together, these results suggest that parallelism in genomic differentiation can be substantial but highly context dependent. Despite these research efforts, we currently lack a good understanding of how the geographic context (divergence in allopatry vs. sympatry) may affect patterns of genomic parallelism for populations adapting to similar ecological niches.
In sympatry, the lack of physical barriers allows for gene flow between diverging populations, which can counteract the accumulation of genome-wide differentiation (Coyne & Orr, 2004). Gene flow homogenizes neutral regions of the genome, and only few regions harboring genes under divergent selection are expected to be strongly differentiated when divergence occurs with gene flow, as shown for crows andHeliconius butterflies (Nadeau et al., 2014; Poelstra et al., 2014), although reinforcement could potentially mitigate this effect (Garner, Goulet, Farnitano, Molina-Henao, & Hopkins, 2018). The homogenizing effect of gene flow also reduces the fraction of the genome able to respond to natural selection (Samuk et al., 2017); previous work has shown that in the presence of gene flow, divergence is limited to regions with low rates of recombination (Samuk et al., 2017). Such constraints are not expected in allopatry and the stochastic effects of genetic drift, differences in effective population size and variable ecology may generate more inconsistent patterns of differentiation among allopatric populations. Thus, we predict higher levels of genomic parallelism across sympatric species due to the bias of divergence towards a smaller fraction of the genome and fewer stochastic peaks due to genetic drift.
Threespine stickleback represent an excellent system for studying the genomic signatures of repeated evolution in natural populations across different geographic settings. Stickleback have rapidly adapted to freshwater habitats throughout the northern hemisphere (Bell & Foster, 1994). Newly formed freshwater lakes were independently colonized by marine stickleback after the last ice age, around 10,000 - 12,000 years ago (Bell & Foster, 1994). Within these young lakes, stickleback have repeatedly and independently adapted to novel resources through parallel phenotypic evolution in trophic morphology (Bell & Foster, 1994; Bolnick & Ballare, 2020; Schluter & McPhail, 1992). Lakes vary in size and depth, encompassing different proportions of benthic and limnetic habitat, which affects dietary and habitat availability for stickleback (Bolnick & Ballare, 2020). Accordingly, variation in diet and morphology across allopatric populations is associated with lake size; stickleback mostly feed on littoral invertebrates (benthic prey) in small lakes and pelagic zooplankton (limnetic prey) in large lakes (Bolnick & Ballare, 2020). In medium-sized lakes, stickleback generally have intermediate phenotypes and broader dietary niches (Bolnick & Ballare, 2020). While most lakes are inhabited by a solitary population (morphologically unimodal for most traits and approximately panmictic), in five lakes in British Columbia the colonizing stickleback independently evolved into co-occurring pairs of sympatric benthic and limnetic specialists (Taylor & McPhail, 1999). This repeated divergence in trophic ecology along the benthic-limnetic axis across sympatric and allopatric stickleback populations allows us to study parallelism of genomic differentiation in different geographic settings.
Here, we employ two approaches to map the genomic signatures of sticklebacks’ adaptation to benthic and limnetic habitats. We use FST to detect adaptive divergence between benthic and limnetic sympatric species pairs (Gow, Rogers, Jackson, & Schluter, 2008; Schluter & McPhail, 1992) and among allopatric populations from small benthic and large limnetic lakes (Bolnick & Ballare, 2020). Further, we use genome-wide association (GWA) mapping (Bolnick & Ballare, 2020) for a larger dataset of allopatric lake populations to detect alleles associated with lake size (the proxy for dietary niche). By comparing benthic-limnetic adaptation in different geographic contexts, we were able to quantify the magnitude of parallelism and ask whether the geographic context affects patterns of shared genomic architecture during adaptation to similar niches. Furthermore, it is likely that regions identified to overlap between these datasets contain loci important for adaptation to divergent benthic and limnetic niches, such candidate regions provide opportunities for follow-up work.