Identifying patterns of selection at synonymous sites in a Pseudomonas phylogeny.

A growing body of evidence suggests that synonymous substitutions, DNA sequence changes that do not alter the encoded amino acid and so should be selectively neutral, can be under strong selection. Yet our ability to identify putative non-neutral synonymous substitutions from comparative sequence data is limited. Here, we use Bayesian stochastic character mapping to contrast substitution rates and patterns for each position along two bacterial genes, followed by hierarchical clustering on principal components (HCPC) and consensus clustering to robustly identify synonymous sites showing signatures of selection. Specifically, we compare gtsB from Pseudomonas fluorescens, known from experimental studies to harbour synonymous mutations with strongly beneficial fitness effects, with a more conserved gene in the same operon, gtsD. We find that synonymous four-fold degenerate sites can vary from highly constrained, characterized by low rates and constrained patterns of substitution comparable to neighboring nonsynonymous second codon positions, to weakly constrained, exhibiting high rates and diverse patterns of substitution indicative of weak purifying selection characteristic of nearly neutrally evolving sites. Highly constrained sites account for 8% and 7.5% of the four-fold degenerate sites in gtsB and gtsD, respectively, and include some positions previously characterized through experiments as beneficial, suggesting they are sites that experience non-neutral selection. Patterns of codon usage, protein structure, or local sequence motifs do not explain the variation in substitution patterns between constrained and unconstrained sites. Combined, our site-specific approach suggests non-neutral synonymous mutations can occur at an appreciable frequency and may often contribute to adaptation.