Discussion
Inferring the genomic basis of adaptive trait variation, including the interplay between adaptive divergence and gene flow, remains a central endeavour in evolutionary biology. To increase rigor, data on both selective agents and organisms’ multivariate phenotypes are needed when identifying regions of the genome potentially implicated in adaptation . We show that there was phenotypic divergence in several traits of threespine stickleback from Mývatn, Iceland, but little evidence for genome wide population genomic structure, suggesting extensive gene flow. However, a combination of GWA and landscape genomics approaches allowed us to isolate genomic regions associated with both environmental and phenotypic variation, suggesting genomic divergence in response to natural selection in face of gene flow.
Phenotypic divergence in the face of gene flow
Mývatn stickleback showed spatial divergence in several traits, some of which was associated with ecological variation. Notably, stickleback were smaller at the warm shore site, had more armour plates and relatively longer first dorsal spines in the North than in the South basin, and relatively longer, but fewer, gill rakers in the rocky shore site than in the rest of the lake. These results align partially with previous findings from the same population sampled in 2009 , which found spatial phenotypic divergence in body size (N basin: larger), spine length (N basin: longer spines) and gill raker gap width, but no divergence in lateral plate number. The number of armour plates and length of spines are important predator defence traits in stickleback , with increased predation pressure by birds or fish selecting for more armour plates and/or longer spines . Mývatn stickleback represent a low plated morph (range 3 to 10) typical of freshwater environments , but the higher plate number and longer spines in the N than the S basin indicates stronger predator induced selection in the N basin. This was further supported by phenotype-environment association analyses that showed that Mývatn stickleback had longer second dorsal spines where piscivorous waterfowl density was higher, and relatively longer pelvic spines in deeper water. The latter may reflect variation in predation pressure by altering the occurrence and visibility to visual predators (e.g., salmonids; ).
Lake fish, including stickleback, often diverge along the benthic-limnetic axis (reviewed in Wagner & Seehausen 2014), whereby fish that specialise on a benthic diet have fewer and shorter gill rakers than those on a limnetic diet . Interestingly, we show that gill raker divergence (i.e., shorter and more versus longer and less gill rakers) did not follow a typical benthic-limnetic divergence. Phenotypic variation in gill raker morphology often reflects diet mediated trait variation and can facilitate resource polymorphism, from individual specialization (Bolnick et al. 2003) to spatial divergence among contrasting environments. Whilst we found some spatial divergence in gill raker number and length, gill raker morphology was not associated with any of our environmental measures (including prey abundances) suggesting that gill rakers may be responding along an unmeasured axis of divergence. Moreover, our findings contrast to some extent with who found divergence between Mined and Warm habitats in gill raker number and gap width rather than gill raker number and length. This discrepancy might suggest that spatiotemporal variation in availability of stickleback prey in Mývatn , may induce fluctuating selection – an interesting target for future research.
Despite observed phenotypic divergence, we found no evidence for genome-wide divergence, suggesting a single panmictic population and phenotypic divergence in face-of-gene flow. Although this aligns with results of , they contrast with those of who found genetic divergence between two sites and suggested the presence of two morphs ‘Lava’ and ‘Mud’. As above, this discrepancy in genetic divergence may reflect temporal changes in the extent of spatial genetic divergence and the loss of previous ecotypes, potentially as a result of strong population fluctuations in stickleback population size . Notably, population demographic analyses indicate that the high-density North basin periodically subsidises the low-density South basin through dispersal , which could result in periodic gene flow. However, it is also possible that differences in the resolution of both spatial sampling and sequencing may have resulted in the discrepancies between studies.