Reply to Massey: Drift does influence mutation-rate evolution.

Although Massey argues that proteome size (P), not the effectivepopulation size (Ne), is the primary determinant of mutation-rate evolution (1), nothing inhis commentor inhisprior work (2) contradicts our conclusions (3). First, although we noted how the mutation rate/nucleotide site/generation (μ) scales with Ne and P of a species, we did not rank the relative importance of these sources of variation, which are functions of the variance harbored by the samples in an analysis (in this case, for both Ne and P). This issue can be seen easily in figure 2 of Massey (1), where the addition of viruses expands the range of variation in P 100-fold beyond that in our study (3). Massey (1) concludes that P accounts for 94% of the variance in μ among taxa (including viruses), whereas our analysis of cellular species reduces this proportion to ∼41%. Additional caveats to assigning relative levels of causality to the factors driving the evolution of μ are: (i) Ne and P are not phylogenetically independent variables, and (ii) the substantially greater sampling variance for Ne than for P will reduce the slope and correlation of a regression of μ on Ne (3). Second, Massey (1) argues that a reduced correlation between the mutation rate per cell-division and Ne weakens our proposal that mutation-rate evolution is influenced by drift. No general relationship is expected between the mutation rate per germ-line/cell division (μd = μ/x, where x is the number of germ-line cell divisions/generation) and Ne under existing evolutionary theory, which postulates that selection on μ is an indirect effect of linked deleterious mutations, the influence of which is only felt after expression. For a multicellular organism with germ-line sequestration, selection operates on the mutation rate per sexual generation, and we expect the indirect selective effect to be accommodated by reducing μd to μ/x (assuming that replication fidelity has not reached its biophysical limits). In principle, multicellularity may impose direct selection on the mutation rate, as somatic mutations have immediate effects (3–5). However, even though high somatic-mutation rates make the direct-selection hypothesis plausible, the limited data are inconsistent with germ-line μd scaling with 1/P in multicellular species (when our data are used, the correlation is very close to zero). Because there is variation in x among taxa (which is not proportional to P or Ne), the expectation is that a regression of μd on Ne will be weaker than one of μ on Ne, just as found by Massey. Finally, a per cell-division argument does not apply to the mostly unicellular species that we have analyzed, and yet the scaling of μ and Ne is similar in unicellular and multicellular species. Third, Massey claims that we provide no support for an inverse relationship between μ and P. However, figure 1C in our work (3) shows that μP scales with ∼1/Ne, which is equivalent to saying that μ scales with ∼1/ (NeP). Although we conclude that the ability of selection to influence μ depends on both the power of drift and the effective target size for deleterious mutations, the importance of proteome-size variation has been made by other authors (2, 6). Should future work with additional key taxa uphold the inverse scaling between μ and NeP, we will have achieved a fairly general theory for one of the key evolutionary-genetic features of species.