Population structure
Our high-resolution SNP genealogies revealed genetic structure among marine stickleback from the Atlantic. Specifically, the phylograms based on SNPs both chosen randomly across the genome and filtered stringently to reduce the influence of selection consistently recovered three marine branches (Figure 1c; bootstrap support is given in Figure S1). These branches were formed by the marine individuals from North Uist and Ireland (ARDH, OBSM, IR), the two samples from the North Sea (DE, NL), and stickleback from Canada and Iceland (CA, IS). Within these branches, however, marine fish from a given location generally did not emerge as monophyletic, except for the Canadian individuals collected thousands of kilometers from the nearest sampling locations (IR, IS) (Figure 1b; overall genome-wide genetic differentiation among the marine samples is presented in Table S2). In contrast to the marine fish, genetic structure among our freshwater samples differed fundamentally between the random and neutral SNP panels (Figure 1c). Based on the former, all freshwater stickleback together grouped to a single, well-supported branch distinct from marine fish, and within this freshwater branch, individuals clustered almost perfectly according to acidic versus basic habitat. This ecological structure largely vanished when using SNPs ascertained to reduce the influence of selection. Moreover, contrary to marine stickleback, freshwater individuals almost consistently grouped by sampling location, despite the dramatically smaller geographic distance among the lakes compared to the marine locations (Figure 1b). All these patterns remained qualitatively consistent when using sparser data sets, and when replacing individual-level by synthetic genotypes derived from pooled data (Figure S1). The latter confirms that poolSeq data enable meaningful genealogical analyses at the population level (Haenel et al. 2019).
The weak genetic structure among our marine locations within the three marine branches is consistent with the notion that marine stickleback display large population sizes, that genetic drift is relatively weak (Mäkinen et al. 2006; Hohenlohe et al. 2010; Jones et al. 2012a; Catchen et al. 2013; Roesti et al. 2014; Lescak et al. 2015), and hence that deleterious genetic variation introduced by hybridization with freshwater fish should be eliminated efficiently. Nevertheless, stickleback across the Atlantic clearly do exhibit genetic structure. Assuming gene flow-selection balance as a cause for the maintenance of SGV, we would therefore expect differences in the level of SGV among broad regions within the Atlantic if these regions differed in the input of maladaptive acidic alleles. A further insight into marine stickleback is that with both the random and neutral SNPs, the freshwater populations from North Uist appear genetically no more similar to marine fish sampled in immediate (ARDH, OBSM) or relative (IR) proximity than to the samples from the much more distant marine locations. This implies that at the genome-wide level, any Atlantic marine sample – irrespective of its precise geographic origin (and including off-shore samples like IR; Table S1) – serves as an adequate representation of ancestral Atlantic marine stickleback (see also Kirch et al. 2021).
An intriguing finding emerging from the genealogy is the nearly perfect segregation of stickleback by habitat when using SNPs sampled at random across the genome. At first glance, this may stimulate the interpretation that on North Uist, initially a single freshwater stickleback form evolved, subsequently differentiated into a single acidic and basic ecomorph, and these ecomorphs then split into multiple sub-populations. Apart from being hydrogeographically implausible (see the Supporting Discussion in Haenel et al. 2019), this interpretation is challenged by the genetic structure revealed by the neutral SNPs: the deep separation of freshwater populations on North Uist based on this marker panel indicates that acidic and basic ecomorphs have arisen multiple times independently through the adaptive sorting of ancestral marine SGV (Magalhaes et al. 2016; Haenel et al. 2019; see also Bell et al. 1993). The contrasting results obtained from random versus neutral SNPs in freshwater but not marine stickleback highlight, on the one hand, how deterministically genome-wide polygenic selection and associated hitchhiking during freshwater adaptation can shape genetic population structure and thus confound neutral evolutionary history (see also Berner & Roesti 2017). On the other hand, these results indicate that the genomes of stickleback populations recently adapted to ecologically novel freshwater habitats are much more profoundly shaped by selection than the genomes of the ancestral marine form.