Transfer RNA genes experience exceptionally elevated mutation rates

Transfer RNAs (tRNAs) are a central component for the biological synthesis of proteins, and they are among the most highly conserved and frequently transcribed genes in all living things. Despite their clear significance for fundamental cellular processes, the forces governing tRNA evolution are poorly understood. We present evidence that transcription-associated mutagenesis and strong purifying selection are key determinants of patterns of sequence variation within and surrounding tRNA genes in humans and diverse model organisms. Remarkably, the mutation rate at broadly expressed cytosolic tRNA loci is likely between seven and ten times greater than the nuclear genome average. Furthermore, evolutionary analyses provide strong evidence that tRNA genes, but not their flanking sequences, experience strong purifying selection, acting against this elevated mutation rate. We also find a strong correlation between tRNA expression levels and the mutation rates in their immediate flanking regions, suggesting a simple new method for estimating individual tRNA gene activity. Collectively, this study illuminates the extreme competing forces in tRNA gene evolution, and implies that mutations at tRNA loci contribute disproportionately to mutational load and have unexplored fitness consequences in human populations.\n\nSignificance StatementWhile transcription-associated mutagenesis (TAM) has been demonstrated for protein coding genes, its implications in shaping genome structure at transfer RNA (tRNA) loci in metazoans have not been fully appreciated. We show that cytosolic tRNAs are a striking example of TAM because of their variable rates of transcription, well-defined boundaries and internal promoter sequences. tRNA loci have a mutation rate approximately seven-to tenfold greater than the genome-wide average, and these mutations are consistent with signatures of TAM. These observations indicate that tRNA loci are disproportionately large contributors to mutational load in the human genome. Furthermore, the correlations between tRNA locus variation and transcription implicate that prediction of tRNA gene expression based on sequence variation data is possible.