Importance of epistatic genetic variance in developmental quantitative genetics

Abstract

Additive genetic variance, or variance arising from the average effects of alleles, is a key parameter for models predicting evolutionary change such as the Breeders equation. Theoretical models diverge on predictions of how much non-additive genetic variance is likely to exist in natural populations, and empirical estimates of non-additive genetic components of genetic variation are very rare. Knowing the proportions of additive and non-additive genetic variance in outbred populations can thus give insights into the amount of genetic variance available for evolution. Models based on discrete loci suggest that variance associated with genetic interactions (statistical epistasis) will typically be modest. Analyses based on polygenic models wherein epistatic effects arise for non-linear relationships among traits may greatly change our expectations for the amount of non-additive variance in ecologically important traits. In such systems, there is no additive genetic variance at equilibrium (when equilibria exist) for any traits that are directionally selected; but epistatic variance can persist such that all variance at equilibrium is non-additive. I analyse existing optimality models in the literature as an example of nonlinear developmental system, wherein performance traits, which are presumed to be directionally selected, are non-linear functions of other traits (for example related to morphology). I find mixed results for the proportion of additive genetic variance in these systems - substantial epistatic variance in performance can be found if the systems are close to the optimum.

Next, I set up a greenhouse blocked diallel design experiment using the annual plant, fast cycling Brassica rapa to yield estimates of these proportions for life history (phenology and flowering) and morphological traits given a known pedigree. I hypothesise that life history traits, that have been under strong selection, should have susbstantial epistatic variance compared to morphological traits if genetic variance is present. I perform univariate analysis using the “animal” model to decompose the phenotypic variance, and directly estimate additive and non-additive genetic variance components. I find that non-additive genetic variance may make up a substantial proportion of the total genetic variance, when genetic variance is maintained. Proportions of additive genetic variance (given substantial total genetic variance) are highest in stem heights in early development and at bolting. The results confirm the importance of infering epistatic variance so as to accurately interpret the potential for evolutionary change.