Mutation bias shapes the spectrum of adaptive substitutions

ABSTRACT Evolutionary adaptation often occurs via the fixation of beneficial point mutations, but different types of mutation may differ in their relative frequencies within the collection of substitutions contributing to adaptation in any given species. Recent studies have established that this spectrum of adaptive substitutions is enriched for classes of mutations that occur at higher rates. Yet, little is known at a quantitative level about the precise extent of this enrichment, or its dependence on other factors such as the beneficial mutation supply or demographic conditions. Here we address the extent to which the mutation spectrum shapes the spectrum of adaptive amino acid substitutions by applying a codon-based negative binomial regression model to three large data sets that include thousands of amino acid changes identified in natural and experimental adaptation in S. cerevisiae , E. coli , and M. tuberculosis . We find that the mutation spectrum has a strong and roughly proportional influence on the spectrum of adaptive substitutions in all three species. In fact, we find that by inferring the mutation rates that best explain the spectrum of adaptive substitutions, we can accurately recover species-specific mutational spectra obtained via mutation accumulation experiments. We complement this empirical analysis with simulations to determine the factors that influence how closely the spectrum of adaptive substitutions mirrors the spectrum of amino acid variants introduced by mutation, and find that the predictive power of mutation depends on multiple factors including population size and the breadth of the mutational target for adaptation. SIGNIFICANCE STATEMENT How do mutational biases influence the process of adaptation? Classical neo-Darwinian thinking assumes that selection alone determines the course of adaptation from abundant pre-existing variation. Yet, theoretical work shows that under some circumstances the mutation rate to a given variant may have a strong impact on the probability of that variant contributing to adaptation. Here we introduce a statistical approach to analyzing how mutation shapes protein sequence adaptation, and show that the mutation spectrum has a proportional influence on the changes fixed in adaptation observed in three large data sets. We also show via computer simulations that a variety of factors can influence how closely the spectrum of adaptive substitutions mirrors the spectrum of variants introduced by mutation.