Fitness effects of new mutations are small and heavily confounded with non-genetic sources of variation in Escherichia coli

The distribution of fitness effects (DFE) of new mutations underpins our understanding of molecular evolution, mutation load and the maintenance of quantitative genetic variation. Most direct estimates in microbes rely on mutation-accumulation (MA) lines that harbour many mutations, so that only the mean and variance of the DFE can be inferred reliably and non-genetic ('environmental') differences between lines are usually ignored. Here we generated 192 Escherichia coli MA lines that accumulated no mutations, one mutation or more than one mutation in similar proportions. Comparing the growth rates of lines with mutations with the mutation-free controls allowed us to partition fitness variation into genetic and non-genetic components. We estimate the average selection coefficient of a single mutation is unlikely to be less than -0.0020 (with 95% credibility) - a value that is small, not significantly different from zero, and largely concordant with most previous estimates from multi-mutant MA studies, although those earlier results have often been misinterpreted as implying much larger deleterious effects. Crucially, systematic non-genetic effects among lines were an order of magnitude larger than mutational effects, and failure to model them would have biased the mean selection coefficient downwards by an order of magnitude. Our study demonstrates that spontaneous mutations in E. coli are typically only mildly deleterious and that rigorous controls for environmental effects are essential for unbiased inference of the DFE in microbial systems.