The divergence of mutation rates and spectra across the Tree of Life

Abstract

Owing to advances in genome sequencing, genome stability has become one of the most scrutinized cellular traits across the Tree of Life. Despite its centrality to all things biological, the mutation rate (per nucleotide site per generation) ranges over three orders of magnitude among species and several-fold within individual phylogenetic lineages. Within all major organismal groups, mutation rates scale negatively with the effective population size of a species and with the amount of functional DNA in the genome. This relationship is most parsimoniously explained by the drift-barrier hypothesis, which postulates that natural selection typically operates to reduce mutation rates until further improvement is thwarted by the power of random genetic drift. Despite this constraint, the molecular mechanisms underlying DNA replication fidelity and repair are free to wander, provided the performance of the entire system is maintained at the prevailing level. The evolutionary flexibility of the mutation rate bears on the resolution of several prior conundrums in phylogenetic and population-genetic analysis and raises challenges for future applications in these areas.

Keywords: drift-barrier hypothesis; effective population size; mutation rate; mutation spectrum; nucleotide composition.

Figure 1. Scaling of base‐substitution mutation rates with genome size and effective population size

(A) Base‐substitution mutation rates per generation ($u_{\mathit{bs}}$) as a function of genome size ($G$, in Mb). Fitted log–log regression for prokaryotes: $y=-8.21-1.98x$, with standard errors (SEs) of 0.18 and 0.30 for the intercept and slope, respectively; $r^2=0.553$ and sample size $n=38$. Fitted regression for unicellular eukaryotes (excluding Chlamydomonas and Emiliana: $y=-8.40-1.10x$) with SEs of 0.40 and 0.27 for the intercept and slope, respectively; $r^2=0.442$ and $n=20$. The dashed line is a reference with slope = −1.0. (B) Group‐specific regressions relating $u_{\mathit{bs}}$ to the effective population size ($N_e$). (C) Relationship between $U_{\mathit{bs}}$ (the product of $u_{\mathit{bs}}$ and the total number of nucleotides contained within protein‐coding sequence) and $N_e$; $y=3.10-0.75x$, with SEs of 0.18 and 0.03 for the intercept and slope, respectively; $r^2=0.861$ and $n=117$. Here, the diagonal dashed lines are references with slopes =−1 All raw data used in these regressions are in Dataset EV1. Regression statistics are summarized in Appendix Tables S1–S3.

Figure 2. Scaling of effective population size and organism size

Relationship between the effective population size ($N_e$) and adult dry weight (in μg). Fitted log–log regression for the entire pool of data (Dataset EV1): $y=6.717-0.199x$, with standard errors (SEs) of 0.054 and 0.007 for the intercept and slope, respectively; $r^2=0.875$, sample size $n=113,$ with $P<{10}^{-53}$. The data for four pathogenic bacteria and two self‐fertilizing nematodes, all of which have unusually low $N_e$, are excluded from this analysis.

Figure 3. Scaling of insertion/deletion and base‐substitution mutation rates

Relationship between the mutation rate for insertion/deletions < 50 bp in length, $u_{\mathit{id}}$, and for base substitutions, $u_{\mathit{bs}}$ across the Tree of Life, both in units of nucleotide site⁻¹ · generation⁻¹. The upper and lower dashed lines denote references for 1.0 and 0.1 ratios of the two rates. The three species with ratios > 1.0 are the yeast Zymoseptoria tritici, the slime mold Dictyostelium, and the nematode Caenorhabditis elegans. Numerical results are tabulated in Dataset EV1.

Figure 4. Aspects of the molecular spectra of base‐substitution mutation rates across the Tree of Life

(A) The phylogenetic dispersion of transition:transversion (ts:tv) ratios and AT mutation bias ($\beta$) across the Tree of Life. The dashed lines denote the respective null values of 0.5 and 1.0 under the hypothesis that all 12 possible base‐substitution mutation types arise at equal rates. Data are only shown for species with $n\ge 20$ observed base substitution mutations. From equation A1.19b in Lynch & Walsh (1998), the coefficient of variation (CV) of a ts:tv estimate equal to $x$ is $\sim \left(1+x\right)/\left(n\sqrt{x}\right)$, which is < 0.10 for $x=1$ and < 0.175 for $x=10$. (B) Genome‐wide AT compositions as a function of the AT mutation bias. The solid line denotes the null expectation under neutrality resulting from the balance between bidirectional mutational pressures, whereas the dashed lines denote the expectations for cases in which selection reduces the fixation probabilities of G/C $\to$ A/T mutations by factors of 0.3 and 0.1 relative to A/T $\to$ G/C mutations (as described in the text).