Abstract
Fitness landscapes map the relationship between genotypes and fitness. However, most fitness landscape studies ignore the genetic architecture imposed by the codon table and thereby neglect the potential role of synonymous mutations. To quantify the fitness effects of synonymous mutations, we used a new software based on Bayesian Monte Carlo Markov Chain methods and estimated selection coefficients from deep sequencing data obtained across 9 amino-acid positions from Hsp90 in Saccharomyces cerevisiae. This work demonstrates how topology and topography of the codon fitness landscape change when synonymous effects are considered. This impacts how populations traverse fitness space as well as their likelihood of reaching a global optimum, in particular in a stressful environment. Finally, we show that residue position, mRNA stability, and codon frequency are predictors of synonymous effect size. Together these results highlight the role of synonymous mutations in adaptation and demonstrate the potential mis-inference when they are neglected in fitness landscape studies.