Strong selection is poorly aligned with genetic variation in Ipomoea hederacea

The multivariate evolution of populations is the result of the interactions between natural selection, drift, and the underlying genetic structure of the traits involved. Covariances among traits bias responses to selection, and the multivariate axis which describes the greatest genetic variation is expected to be aligned with patterns of divergence across populations. An exception to this expectation is when selection acts on trait combinations lacking genetic variance, which limits evolutionary change. Here we used a common garden field experiment of individuals from 57 populations of Ipomoea hederacea to characterize linear and nonlinear selection on five quantitative traits in the field. We then formally compare patterns of selection to previous estimates of within-population genetic covariance structure (the G-matrix) and population divergence in these traits. We found that selection is poorly aligned with previous estimates of genetic covariance structure and population divergence. In addition, the trait combinations favoured by selection were generally lacking genetic variation, possessing approximately 15-30% as much genetic variation as the most variable combination of traits. Our results suggest that patterns of population divergence are likely the result of the interplay between adaptive responses, correlated response, and selection favoring traits lacking genetic variation.