Contrasting mutation rates from specific-locus and long-term mutation-accumulation procedures

Until recently, the two predominant ways to estimate mutation rates were the specific-locus method and the mutation-accumulation (Bateman-Mukai) method. Both involve seeding a number of parallel lines from a small, genetically uniform population, growing as long as is feasible but not so long as to allow selection to perturb mutant frequencies, and sometimes using extreme bottlenecks to facilitate the retention of deleterious mutations. In the specific-locus method, mutations are selected according to their specific phenotypes and are confirmed by sequencing. In older versions of the mutation-accumulation method, the increase in variance of a quantitative fitness trait is measured and converted into a mutation rate. More recently, a variation on the mutation-accumulation method has become possible based on phenotype-blind genomic sequencing, which might (or might not) provide improved sampling breadth, usually at the expense of sample size. In a recent study, genomic sequencing was applied to Escherichia coli lines propagated for 40,000 generations and passaged daily via 5,000,000 cells. To mitigate the impact of selection, the only targets employed for rate calculations were putatively neutral synonymous mutations. The mutation rate estimate was about 6-fold lower than obtained previously with a robust specific-locus method. Here I argue that purifying selection acting to shape the strong codon preferences of E. coli is the probable cause of the lower estimate, rather than, for instance, a lower mutation rate in nature than in the laboratory.