Induced mutations: A novel tool to study phenotypic integration and evolutionary constraints in Arabidopsis thaliana

Adaptive phenotypic evolution is constrained by the availability of genetic variation and the patterns of phenotypic integration and genetic architect ure among characters. Phenotypic and genotypic variance/covariance matrices provide a convenient quantitative summary of such patterns and are widely used in evolutionary studies. These matrices are, for mathematical tractabi lity, assumed to remain constant or change only proportionally in the cours e of evolution. However, this assumption is tenuous over the medium and lon g term and cannot possibly hold at all taxonomic scales. One important sour ce of genetic change is mutations, but low natural rates of mutation make t hem difficult to study at the population level. We used ethyl-methane-sulph onate to induce mutations in a flowering plant and assess their impact on t he means and phenotypic variances of seven morphological and life-history t raits as well as the covariances among these characters. We found that only one trait mean changed in response to mutagenesis, but that mutations gene rated new variance in all traits in a roughly dose-dependent fashion. Commo n principal components analyses of the matrices showed that the degree of d ivergence from a non-mutagenized control population is roughly dose-depende nt. This overall divergence of covariance structure is the product of idios yncratic mutation-induced changes in particular pairwise correlations among traits. One well-studied correlation between time to flowering and the num ber of rosette leaves at flowering persists in all mutagenic treatments. We discuss these findings and their implications for quantitative genetic mod els that assume constant variance/covariance matrices.