Predictable and divergent change in the multivariate P-matrix during parallel adaptation

Abstract

AbstractAdaptation to replicated environmental conditions can be remarkably predictable, suggesting parallel evolution may be a common feature of adaptive radiation. An open question, however, is how phenotypic variation itself evolves during repeated adaptation. Here, we use a dataset of morphological measurements from 35 populations of threespine stickleback, consisting of 16 parapatric lake- stream pairs and three marine populations, to understand how phenotypic variation has evolved during transitions from marine to freshwater environments, and during subsequent diversification across the lake-stream boundary. We find statistical support for divergent phenotypic covariance (P) across populations, with most diversification of P occurring across freshwater populations. Despite a close correspondence between within-population phenotypic variation and among population divergence, we find that variation in P is unrelated to total variation in population means across the set of populations. Within lake-stream pairs, however, we find that theoretical predictions for microevolutionary change can explain over 30% of the total divergence in P matrices across the habitat boundary. Together, our results indicate that variance evolution occurs primarily in dimensions of trait space with low phenotypic integration, driven by divergence into disparate lake and stream environments, illustrating how conserved and divergent features of multivariate variation can underlie adaptive radiation.