Evolution of mutation rates in hypermutable populations of Escherichia coli propagated at very small effective population size

Abstract

Mutation is the ultimate source of the genetic variation-including variation for mutation rate itself-that fuels evolution. Natural selection can raise or lower the genomic mutation rate of a population by changing the frequencies of mutation rate modifier alleles associated with beneficial and deleterious mutations. Existing theory and observations suggest that where selection is minimized, rapid systematic evolution of mutation rate either up or down is unlikely. Here, we report systematic evolution of higher and lower mutation rates in replicate hypermutable Escherichia coli populations experimentally propagated at very small effective size-a circumstance under which selection is greatly reduced. Several populations went extinct during this experiment, and these populations tended to evolve elevated mutation rates. In contrast, populations that survived to the end of the experiment tended to evolve decreased mutation rates. We discuss the relevance of our results to current ideas about the evolution, maintenance and consequences of high mutation rates.

Keywords: antimutators; asexual populations; evolution of mutation rates; indirect selection; mutation accumulation; mutators.

Figure 1.

Maximum-likelihood estimates of mutation rates to streptomycin resistance and nalidixic acid resistance and their 95% confidence intervals for the LB (a,b) and MG (c,d) populations. Black markers denote populations that survived to the end of the experiment; red markers denote the populations that went extinct. (Some populations could not be assayed for mutation rate and are not represented: see §2.) Solid horizontal lines give the estimated mutation rate in the ancestral strain; upper and lower dotted lines represent its 95% CI.

Figure 2.

Endpoint fitness versus endpoint mutation rate to nalidixic acid resistance in the MG (a) and LB (b) populations that survived to the end of the experiment. Fitness is represented by the relative number of colonies obtained when plating identical dilutions of an evolved population and the ancestor (relative CFU count). The correlation between endpoint fitness and endpoint mutation rate is non-significant in the MG populations (p-value = 0.298) and significant in the LB populations (p-value = 0.045). The grey lines represent the least-squares fit regression lines for both datasets.