Horizontal gene transfer may explain variation in {\theta}s

Martincorena et al. estimated synonymous diversity ({\theta}s = 2N{\mu}) across 2,930 orthologous gene alignments from 34 Escherichia coli genomes, and found substantial variation among genes in the density of synonymous polymorphisms. They argue that this pattern reflects variation in the mutation rate per nucleotide ({\mu}) among genes. However, the effective population size (N) is not necessarily constant across the genome. In particular, different genes may have different histories of horizontal gene transfer (HGT), whereas Martincorena et al. used a model with random recombination to calculate {\theta}s. They did filter alignments in an effort to minimize the effects of HGT, but we doubt that any procedure can completely eliminate HGT among closely related genomes, such as E. coli living in the complex gut community. Here we show that there is no significant variation among genes in rates of synonymous substitutions in a long-term evolution experiment with E. coli and that the per-gene rates are not correlated with {\theta}s estimates from genome comparisons. However, there is a significant association between {\theta}s and HGT events. Together, these findings imply that {\theta}s variation reflects different histories of HGT, not local optimization of mutation rates to reduce the risk of deleterious mutations as proposed by Martincorena et al.