Coarse- and fine-grained phenotypic divergence among threespine stickleback from alternating lake and stream habitats

Background: Habitat characteristics can vary over small spatial scales at which gene flow is expected to swamp any effect of divergent natural selection. However, fine-grained (‘microgeographic’) adaptive divergence may still be feasible if individuals exhibit dispersal behaviours that improve the match between their phenotype and habitat. For example, threespine stickleback (Gasterosteus aculeatus) from lake and stream habitats maintain differences across a narrow ecotone because of non-random gene flow. However, it is unknown whether dispersal bias might also contribute to even finer-scale divergence within habitats, in response to microhabitat variation within lakes and within streams. Question: Does stickleback morphology co-vary with flow regime within stream populations, controlling for distance from adjoining lake populations? Data: We sampled stickleback along a transect through alternating lake and stream habitats. Within each stream, multiple traps were set at 50 m intervals. We recorded microhabitat data (flow rate and depth) at each trap. We measured morphology (gill rakers, head shape, fin shape, standard length) of more than 900 stickleback captured from these traps. Analysis: We used multivariate analyses of covariance and linear models to test for: (1) phenotypic divergence between lake and stream stickleback, (2) divergence among stream sites as a function of their distance from an adjoining lake, and (3) covariation between local flow regime (at each trap) and the morphology of stickleback captured from that trap. Conclusions: Fish from different flow regimes within a stream show phenotypic variation that is not due to clinal transitions from lake to stream. We found covariation between local flow regime and either fin morphology or gill raker length in different streams. The total effect size of stream microhabitat on morphology was greater than the effect size of habitat (lake vs. stream), for overall multivariate data and for a subset of univariate traits. These findings imply that local adaptation can occur on a finer spatial scale than is typically expected, perhaps as a result of non-random dispersal.