Phylogeographical autocorrelation of phenotypic evolution in honey bees (Apis mellifera L.).

The analysis of phenotypic divergence among local populations within a species has been traditionally performed in a spatial context, although advances in genetic analysis using mtDNA have permitted a simultaneous evaluation of geographical and historical patterns of variation, so-called phylogeographical analysis. In this paper, we combine these two dimensions of variation (geographical space and phylogenetic history) to evaluate patterns of phenotypic evolution in honey bees (Apis mellifera L.). Data on 39 phenotypic traits, derived from 417 colonies grouped into 14 subspecies, were analysed using autocorrelation methods. Mantel tests indicated that the relationship between phenotypic divergence, estimated by Euclidean distances among subspecies' morphological centroids, was significant both when compared to geographical distance (r=0.371; P < 0.01) and to genetic distance (estimated as sequence divergence (%) in a mtDNA region encompassing part of the NADH dehydrogenase subunit 2 and isoleucine transfer RNA (r=0.329; P < 0.01)). For the analysis of each trait, the effects of the geographical co-ordinates (latitude and longitude of subspecies geographical range) and of the phylogenetic patterns (defined by eigenvectors of the genetic distance matrix) on phenotypic variation were simultaneously analysed using an extension of a recently developed model, called Phylogenetic Eigenvector Regression (PVR). In general terms, the partial regression slopes indicated that the variation in the characters traditionally associated with adaptive processes, such as body and wing size, were better explained by geographical position. However, characters usually thought to be neutral, such as wing venation angle, were more associated with phylogeny. This is expected because PVR can be interpreted as a partition model, in which adaptive variation tends to be independent of phylogeny (and, in this case, associated with geography). In addition, the first principal component derived from the expected values of the model for each trait, which can be interpreted as the phenotypic variation predicted by phylogeny, is more structured in a north-south cline than are the original data, supporting an adaptive interpretation. The phylogeographical autocorrelation analyses performed in this study show that different traits are more related to one of the two dimensions of variation (geography and phylogeny), and these patterns can furnish insights into the nature of phenotypic evolution in these organisms.