Delineating Mutation Bias and Selection during Plant Development

Plants accumulate germline mutations through meristematic lineages where somatic purifying selection could remove deleterious coding mutations and produce perceived mutation biases, like lower genic mutation rates. Here we describe a quantitative framework that uses mutation spectra and genome architecture to separate missing coding mutations from true genic hypomutation for selection-corrected estimates of mutation biases. In simulations of plant meristems subject to selection and mutation bias, only extremely strong dominant mutations produced gene-body patterns in the same direction as empirical estimates of hypomutation. However, after accounting for this selection, we can recover the true underlying mutation rate variation. We apply this to re-analysis of Arabidopsis mutation-accumulation datasets. After selection, we see [~]50% lower mutation rates in gene bodies and an additional [~]15-30% reduction in broadly expressed genes, consistent with mutation rate bias. This framework provides a quantitative method to compare mutation rates across the genome while explicitly controlling for selection on coding sequence mutations.