Mutation bias and GC content shape antimutator invasions

Abstract

Mutators represent a successful strategy in rapidly adapting asexual populations, but theory predicts their eventual extinction due to their unsustainably large deleterious load. While antimutator invasions have been documented experimentally, important discrepancies among studies remain currently unexplained. Here we show that a largely neglected factor, the mutational idiosyncrasy displayed by different mutators, can play a major role in this process. Analysing phylogenetically diverse bacteria, we find marked and systematic differences in the protein-disruptive effects of mutations caused by different mutators in species with different GC compositions. Computer simulations show that these differences can account for order-of-magnitude changes in antimutator fitness for a realistic range of parameters. Overall, our results suggest that antimutator dynamics may be highly dependent on the specific genetic, ecological and evolutionary history of a given population. This context-dependency further complicates our understanding of mutators in clinical settings, as well as their role in shaping bacterial genome size and composition.

Fig. 1

Frequency trajectories of antimutator alleles invading well-adapted, mutator populations. Lines represent 100 independent simulations for each condition. a Invasion dynamics under various values of the mutation rate of the resident mutator (grey, m = 1000; magenta, m = 300; blue, m = 100; green, m = 30; light grey, m = 1), and a fixed fitness cost of deleterious mutations (s_d = 0.064). b Invasion dynamics under various values of the fitness cost of deleterious mutations (from left to right, s_d equals: 0.064, 0.032, 0.016, 0.008, 0.004, 0.002, 0.001), and a fixed mutation rate of the resident mutator (m = 1000). Other parameters as described in Methods

Fig. 2

Mutational spectrum effects on the invasion speed of antimutator alleles. Points represent the effective selection coefficient (s_eff) of the invading antimutator alleles, averaged from 200 independent simulations. Panels correspond to different values of the mutation rate of the resident mutator (a, m = 30; b, m = 100; c, m = 300; d, m = 1000). Within each panel, lines depict different values for the fitness cost of deleterious mutations (from top to bottom, s_d equals: 0.064, 0.032, 0.016, 0.008, 0.004, 0.002, 0.001). Mutational spectrum effects refers to the differential propensity of mutators to produce deleterious mutations with different fitness cost. We modelled this effect as a multiplicative factor (κ) that modifies s_d in the mutator background, such that when κ < 1 mutators produce milder deleterious mutations than antimutators, when κ = 1 there is no difference between backgrounds, and when κ > 1 mutations are more harmful in mutators. The basal deleterious mutation rate (u_d) is set to 2 × 10⁻⁴ (other parameters as described in Methods)

Fig. 3

Impact of the deleterious and lethal mutation rate on antimutator dynamics. Panels a and b show antimutator fitness under two extreme values of the basal deleterious mutation rate (u_d = 0.5 × 10⁻⁴ and u_d = 8 × 10⁻⁴, respectively). Points and colours follow the same convention as in Fig. 2. Panel c shows the change in the antimutator’s effective selection coefficient (s_eff) for various values of the basal deleterious mutation rate (from top to bottom, u_d equals: 1.6 × 10⁻³, 8 × 10⁻⁴, 4 × 10⁻⁴, 2 × 10⁻⁴, 1 × 10⁻⁴, 0.5 × 10⁻⁴). Fold change refers to the change in s_eff from κ = 0.25 to κ = 4. Panels d and e show the effects on antimutator dynamics of spectrum-driven differences in the propensity to produce lethal mutations, under two extreme values of the basal lethal mutation rate (u_l = 0.2 × 10⁻⁵ and u_l = 6.4 × 10⁻⁵, respectively). Points and colours as in Fig. 2. Panel f shows the change in s_eff for various values of the basal lethal mutation rate (from top to bottom, u_l equals: 6.4 × 10⁻⁵, 3.2 × 10⁻⁵, 1.6 × 10⁻⁵, 0.8 × 10⁻⁵, 0.4 × 10⁻⁵, 0.2 × 10⁻⁵). Fold change is defined as in c. In all cases, the mutation rate of the resident mutator was fixed to a single value (m = 300). Other parameters as described in Methods

Fig. 4

Protein-disrupting effects of mutations caused by different mutators in different genomes. Colours correspond to predictions for mutY^– (red), mutT^– (blue) and Mismatch Repair^– (green) mutators. For comparison, the effects of a unbiased spectrum are highlighted with a grey background. a Grantham scores in five different bacterial species for which hypermutability is of particular interest. These species are arranged, from left to right, according to increasing GC content. b Average Grantham scores across a panel of species with genomes spanning a wide range of GC compositions. Boxplots as defined by default. c Average BLOSUM100 scores across the same panel. Details about these genomes are shown in Supplementary Table 1. Source data are provided as a Source Data file