Adaptive tuning of mutation rates allows fast response to lethal stress in Escherichia coli

Toon Swings; Bram Van den Bergh; Sander Wuyts; Eline Oeyen; Karin Voordeckers; Kevin J Verstrepen; Maarten Fauvart; Natalie Verstraeten; Jan Michiels

doi:10.7554/eLife.22939

. 2017 May 2;6:e22939. doi: 10.7554/eLife.22939

Adaptive tuning of mutation rates allows fast response to lethal stress in Escherichia coli

Toon Swings ¹, Bram Van den Bergh ¹, Sander Wuyts ¹, Eline Oeyen ¹, Karin Voordeckers ^1,², Kevin J Verstrepen ^1,², Maarten Fauvart ^1,³, Natalie Verstraeten ¹, Jan Michiels ^1,^*

Editor: Wenying Shou⁴

PMCID: PMC5429094 PMID: 28460660

Abstract

While specific mutations allow organisms to adapt to stressful environments, most changes in an organism's DNA negatively impact fitness. The mutation rate is therefore strictly regulated and often considered a slowly-evolving parameter. In contrast, we demonstrate an unexpected flexibility in cellular mutation rates as a response to changes in selective pressure. We show that hypermutation independently evolves when different Escherichia coli cultures adapt to high ethanol stress. Furthermore, hypermutator states are transitory and repeatedly alternate with decreases in mutation rate. Specifically, population mutation rates rise when cells experience higher stress and decline again once cells are adapted. Interestingly, we identified cellular mortality as the major force driving the quick evolution of mutation rates. Together, these findings show how organisms balance robustness and evolvability and help explain the prevalence of hypermutation in various settings, ranging from emergence of antibiotic resistance in microbes to cancer relapses upon chemotherapy.

DOI: http://dx.doi.org/10.7554/eLife.22939.001

Research Organism: E. coli

eLife digest

A cell’s DNA can acquire errors over the course of its lifetime. These errors, known as mutations, are often harmful and can cripple the cell. However, some mutations are needed to enable a cell or organism to adapt to changes in its environment. Since there is a trade-off between acquiring beneficial mutations versus harmful ones, cells carefully balance how often they acquire new mutations.

Cells have several mechanisms that limit the number of mutations by correcting errors in DNA. Occasionally these repair mechanisms may fail so that a small number of cells in a population accumulate mutations more quickly than other cells. This process is known as “hypermutation” and it enables some cells to rapidly adapt to changing conditions in order to avoid the entire population from becoming extinct.

So far, studies on hypermutation have largely been carried out under conditions that are mildly stressful to the cells, which only cause low frequency of hypermutation. However, little is known about the role of this process in cells under near-lethal levels of stress, for example, when drugs target bacteria or cancer cells in the human body.

Swings et al. studied hypermutation in populations of a bacterium called Escherichia coli exposed to levels of alcohol that cause the bacteria to experience extreme stress. The experiments show that hypermutation occurs rapidly in these conditions and is essential for bacteria to adapt to the level of alcohol and avoid extinction. Populations of bacteria in which hypermutation did not occur were unable to develop tolerance to the alcohol and perished. Further experiments show that an individual population of bacteria can alter the rate of mutation (that is, how often new mutations arise) several times as a result of changing stress levels.

The findings of Swings et al. suggest that hypermutation can rapidly arise in populations of cells that are experiencing extreme stress. Therefore, this process may pose a serious risk in the development of drug resistant bacteria and cancer cells. In the future, developing new drugs that target hypermutation may help to fight bacterial infections and cancer.

DOI: http://dx.doi.org/10.7554/eLife.22939.002

Introduction

Theory predicts that the optimal mutation rate depends on several different factors, including genome size and effective population size. For example, unicellular organisms, such as viruses and bacteria, exhibit per-base mutation rates that are inversely correlated with their population and genome sizes, whereas multicellular organisms with much larger genomes and smaller effective population sizes may have higher per-base mutation rates (Lynch, 2010; Lynch et al., 2016). While mutations are necessary to adapt to new and stressful environments (Barrick and Lenski, 2013; Wiser et al., 2013), random changes in an organism’s DNA are rarely beneficial, but more often neutral or slightly deleterious (Elena and Lenski, 2003a; Perfeito et al., 2007; Eyre-Walker and Keightley, 2007). Consequently, mutation rates are strictly balanced by the trade-off between the need for mutations to adapt and the concomitant increase in genetic load (Sniegowski and Raynes, 2013; Lynch, 2011; Wielgoss et al., 2013; Desai and Fisher, 2011). This trade-off between adaptability and adaptedness is believed to be responsible for the low genomic mutation rates usually observed in organisms, while a further decrease in mutation rate is restricted by the energy needed to increase and maintain high replication fidelity (de Visser, 2002; Sung et al., 2016; Sniegowski et al., 2000; Ram and Hadany, 2014). Furthermore, the power of random drift will limit selection on even lower mutation rates when additional increases in replication fidelity are insufficiently advantageous. Due to this drift-barrier, mutation rates are believed to evolve to an equilibrium where the strength of selection matches the power of drift (Lynch et al., 2016).

An organism’s cellular mutation rate is generally considered to be near-constant (Lynch, 2010; Drake, 1991), yet the optimal mutation rate has been reported to depend on the environment (Elena and de Visser, 2003b; Rando and Verstrepen, 2007). Wild-type bacteria grown under optimal conditions typically have low mutation rates in the order of 10⁻³ mutations per genome per generation (Lee et al., 2012; Drake et al., 1998). Under these conditions, hypermutators with weakly (10-fold) or strongly (100–10.000-fold) increased mutation rates occur only sporadically (de Visser, 2002; Marinus, 2010). Despite low frequencies of hypermutators in laboratory populations (Boe et al., 2000), a much higher prevalence is observed in natural bacterial populations (Gross and Siegel, 1981; Hall and Henderson-Begg, 2006), such as clinical isolates of pathogenic E. coli (Matic et al., 1997; Denamur et al., 2002; LeClerc et al., 1996), Pseudomonas aeruginosa (Marvig et al., 2013, 2015; Oliver, 2015; Ferroni et al., 2009), Salmonella (LeClerc et al., 1996), Staphylococcus aureus (Iguchi et al., 2016) among others (Negri et al., 2002; Gould et al., 2007; Rajanna et al., 2013) and in nearly all A. baumannii strains adapting to severe tigecycline stress (Hammerstrom et al., 2015). In addition, high frequency of hypermutation is also documented in eukaryotic pathogens including the malaria-causing parasite Plasmodium falciparum (Lee and Fidock, 2016; Gupta et al., 2016) and the fungal pathogen Candida glabrata (Healey et al., 2016). Moreover, hypermutation plays an important role in cancer development and proliferation, as it helps to overcome different barriers to tumor progression (Wang et al., 2016; Bielas et al., 2006; Roberts and Gordenin, 2014). These observations suggest the natural occurrence of situations in which higher mutation rates confer a selectable advantage. This is especially obvious in harsh environments, where near-lethal stress requires swift adaptation of at least some individuals to avoid complete extinction of the population (Bell and Gonzalez, 2011). Adaptation sufficiently rapid to save a population from extinction is called evolutionary rescue. This phenomenon is widely studied in the light of climate change and adaptation of declining populations to new environments (Lindsey et al., 2013). It occurs when a population under stress lacks sufficient phenotypic plasticity and can only avoid extinction through genetic change (Gonzalez et al., 2013). Evolutionary rescue depends on different factors such as the population size, genome size, mutation rate, degree of environmental change and history of the stress (Gonzalez et al., 2013; Gonzalez and Bell, 2013). By increasing the supply of mutations, hypermutation might also be crucial to enable evolutionary rescue for populations under near-lethal stress.

Despite the high prevalence of hypermutation in clinical settings, current knowledge is lacking on the long-term fate of mutators and their specific role in survival under near-lethal stress conditions. Previous studies exploring the costs and benefits of mutators mostly focused on mild stresses. Both experimental evidence and theory show that mutators can readily increase in frequency in a population through second-order selection. In this case, a mutator hitchhikes along with a sporadically occurring, beneficial mutation that thrives under natural selection (Gentile et al., 2011; Woods et al., 2011; Giraud et al., 2001; Shaver et al., 2002; Mao et al., 1997; Sniegowski et al., 1997). This process relies on different elements, such as initial mutator frequency (Tenaillon et al., 1999), the relative timing of the emergence of one or multiple beneficial mutations (Tanaka et al., 2003), the degree of environmental change or selection strength (Mao et al., 1997; Pal et al., 2007), the mutational spectrum (Couce et al., 2013) and the strength of the specific mutator (Loh et al., 2010). Although hypermutation can readily spread in a population by means of hitchhiking when adaptation is required, long-term evolution experiments also show selection against hypermutation (Lynch, 2011; Tröbner and Piechocki, 1984). These results demonstrate that the actual mutation rate of a population is prone to change by evolution. However, our current knowledge on the long-term dynamics of hypermutation and the mechanisms underlying changes in mutation rate remains fragmentary. Specifically, conditions under which the spread of mutators is inhibited or the increased mutation rate is reversed, remain largely unexplored (Raynes and Sniegowski, 2014).

The aim of the current study was to better understand the dynamics of hypermutation under near-lethal, complex stress. Therefore, we used E. coli exposed to high ethanol stress as a model system (Goodarzi et al., 2010; Nicolaou et al., 2012). Here, multiple mutations epistatically interact and diverse evolutionary trajectories can lead to adaptation to high ethanol concentrations (Voordeckers et al., 2015). In our study, we found an unexpected flexibility in cellular mutation rates as a response to changes in selective pressure. First, we used a defined collection of mutators with distinct mutation rates to identify a range of optimal mutation rates to enable rapid growth under high ethanol stress. Next, experimental evolution revealed an essential role for hypermutation for de novo adaptation to high ethanol stress. While hypermutation quickly and recurrently arose concurrent with increases in ethanol concentrations, mutation rates rapidly declined again once cells were adapted to the stress. Interestingly, we identified cellular mortality as the major force that drives fast evolution of mutation rates. In summary, our results shed new light on the dynamics of mutation rate evolution and help explain why maintaining high mutation rates is limited in time.

Results

Hypermutation enables rapid growth under high ethanol stress

Little is known on the role of hypermutators under complex, near-lethal stress conditions. In these conditions, growth rates are low and the probability to accumulate an adaptive mutation is strongly limited. We postulate that mutator mutants yield variable benefits under these conditions, depending on their mutation rates. To verify this hypothesis, a collection of E. coli mutants displaying a range of mutation rates (Figure 1—figure supplement 1) was grown in 5% EtOH. At this concentration, ethanol almost completely inhibits growth and drastically reduces the carrying capacity of a wild-type culture, indicating extreme stress (Figure 1—figure supplement 2).

Growth rate and lag time reflect the fitness of a strain in a specific environment (Stepanyan et al., 2015; Hammerschmidt et al., 2014). These growth parameters are contingent upon the initial population size. On the one hand, the effect of a rare beneficial mutation on growth rate and lag time is mitigated by a large initial population size (Mao et al., 1997). On the other hand, the effect of a beneficial mutation on the growth dynamics is amplified by a small initial population size, as this limits the generation of beneficial mutants. Therefore, we tested growth of wild-type and mutator strains in the presence of 5% EtOH both for a small (10⁴ CFU/ml) and a large (10⁷ CFU/ml) population size. In the latter condition, we observed strongly overlapping growth curves. Small initial population sizes, however, led to highly dispersed growth curves, pointing to an important contribution of mutation rates to the adaptive capacity under ethanol stress (Figure 1a). Surprisingly, large initial populations lead to a lower yield compared to small initial population sizes. The growth from the small inoculum is likely driven by adaptive mutations, while the effect of a beneficial mutation might be mitigated when starting with a large inoculum. Moreover, we expect that a mutant occurring in case of a small initial inoculum size will have more time to manifest ( $\log_{2} (d i l u t i o n f a c t o r : 100 000) = \pm 16.61$ generations), compared to the mutant occurring in case of a large initial population size ( $\log_{2} (100) = \pm 6.67$ generations), possibly leading to the observed higher yield.

Figure 1—figure supplement 1. — (a) In the left panel, the large initial population size (10⁷ CFU/ml) mitigates the effect on growth of emerging beneficial mutations. Growth curves of both wild type and mutators are overlapping except for all replicates of the ∆*dnaQ* mutant. In the right panel, we observed highly dispersed growth curves. The effect of a beneficial mutation manifests itself due to the small initial population size (10⁴ CFU/ml). The blue line and shading represents the sigmoidal fit of the wild-type growth curves (n = 3, fit using Gompertz equation with 95% c.i. (shading), see Equation 1 in Materials and methods section), while the grey lines represent growth curve of separate replicates for each mutator mutant (b) Growth rates of all strains in the presence of 5% EtOH were measured both starting from a large initial population size of 10⁷ cells per ml (left) and a small initial population size of 10⁴ cells per ml (right). No significant difference was observed between the growth rates of the wild type and mutants in the case of a large starting population, indicating no direct fitness effect caused by the deletion of mutator genes (except for the ∆*dnaQ* mutant) (mean ± s.d., n = 3, repeated measures ANOVA with post hoc Dunnett correction, ****p<0.001). When starting from a small initial population, growth rates of all mutator mutants increased compared to the wild type (mean ± s.d., n = 3, two-sided Student’s t-test, *p<0.1; **p<0.05; ***p<0.01; ns: not significant), indicative of the occurrence of adaptive mutations as an indirect benefit for hypermutation under complex, near-lethal stress.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.003

The lag times calculated from the growth curves reveal a window of beneficial mutation rates for growth under 5% EtOH (Figure 1—figure supplement 3). Strikingly, 2- to 70-fold increased mutation rates (e.g. in a ∆mutS mutant) are more advantageous under these conditions than mutants with lower or higher mutation rates (e.g. in ∆xthA or ∆dnaQ mutants, respectively). In the absence of ethanol, we did not observe differences in lag time among the wild type and mutator mutants. This suggests a crucial role for hypermutation for rapid growth under near-lethal stress by supplying the population more rapidly with (a combination of) beneficial mutations (Figure 1—figure supplement 3).

Growth rates, in turn, when calculated from a large initial density of 10⁷ cells per ml, did not differ significantly in the presence of 5% EtOH, suggesting no direct fitness effect for hypermutation in high ethanol conditions (Figure 1b). However, when growth rates were determined for each mutant starting from small initial population sizes, we observed higher growth rates for all mutator mutants relative to the wild type, demonstrating the emergence of adaptive mutations (Figure 1b). These data indicate that the advantage of hypermutation under ethanol stress can be attributed mainly to second-order selection, following the beneficial effects of novel mutations relative to possible direct effects of the mutator mutation itself. To corroborate these results, we determined relative fitness (Van den Bergh et al., 2016) from direct competition experiments between the wild type and mutants with contrasting mutation rates (∆mutM, ∆mutS and ∆dnaQ). These tests demonstrate the fitness advantage of hypermutation under 5% EtOH. In accordance with the data of the lag time, the relative fitness compared to the wild type is high for the ∆mutS strain while it does not differ from one for the ∆mutM and ∆dnaQ mutants (Figure 1—figure supplement 4). These results confirm that the advantage of hypermutation under near-lethal stress can be attributed to the rapid emergence of beneficial mutations, enabling fast adaptation to avoid extinction.

Long-term adaptation to high ethanol stress in E. coli is contingent upon hypermutation

Our results suggest an essential role for hypermutation in evolution under near-lethal stress. To further extend these observations to a wild-type population, we set up a long-term evolution experiment aimed at adapting E. coli to high percentages of ethanol. We serially transferred 20 parallel E. coli lines founded by a non-mutator ancestor for approximately two years (more than 500 generations). To maintain near-lethal ethanol concentrations throughout the adaption process, populations were incubated in gradually increasing ethanol concentrations (Figure 2a; Figure 2—figure supplement 1). Although ethanol tolerance increased in all populations, only eight out of 20 lines developed tolerance to very high (7% or more) ethanol concentrations (Figure 2b), while the other 12 lines recurrently died out and only developed tolerance to relatively low ethanol concentrations (6% or lower). These results suggest the presence of a critical factor inherent to those eight lines that underlies their increased ethanol tolerance.

Figure 2. — (a) Setup of the evolution experiment with increasing percentage of EtOH. Initially, ancestral cells were grown in the presence of 5% EtOH, the condition that mimics near-lethal stress (Figure 1—figure supplement 2). Populations that grew until exponential phase were transferred to fresh medium while simultaneously increasing EtOH concentrations with 0.5% (for full details, see Materials and methods). (b) Evolutionary outcome of 20 independent parallel lines. Eight parallel lines evolved to high EtOH tolerance (shown in red). The other 12 lines were only able to acquire low EtOH-tolerance levels (shown in blue). For each line, the relative time (in generations) it spent growing on a certain percentage of EtOH is shown.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.008

Figure 2—figure supplement 1. — (a) Setup of the evolution experiment with increasing percentage of EtOH. Initially, ancestral cells were grown in the presence of 5% EtOH, the condition that mimics near-lethal stress (Figure 1—figure supplement 2). Populations that grew until exponential phase were transferred to fresh medium while simultaneously increasing EtOH concentrations with 0.5% (for full details, see Materials and methods). (b) Evolutionary outcome of 20 independent parallel lines. Eight parallel lines evolved to high EtOH tolerance (shown in red). The other 12 lines were only able to acquire low EtOH-tolerance levels (shown in blue). For each line, the relative time (in generations) it spent growing on a certain percentage of EtOH is shown.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.008

To explore the fundamental difference between high ethanol tolerant lines and low ethanol tolerant lines, we used fluctuation assays to determine the population mutation rate. The results clearly show that lines can be divided in two groups with mutation rates either higher or lower than the wild-type mutation rate. This subdivision perfectly corresponds to the difference in ethanol-tolerance levels (Figure 3a). In conclusion, even though mutator mutants occur spontaneously in the population, these data suggest that hypermutation underlies adaptation to high ethanol levels in such a way that only lines with a higher mutation rate than the wild-type mutation rate are able to evolve high ethanol tolerance (Figure 3b).

Figure 3. — (a) The population mutation rate of parallel evolved lines relative to the wild type mutation rate is shown (mean ± 95% c.i., see Materials and methods). Two different groups can clearly be distinguished according to the higher than wild-type () or lower than wild-type () mutation rate. This subdivision is in accordance with the difference in endpoint EtOH tolerance levels (Figure 2b). All mutation rates were significantly different from the wild type (p<0.001; two-sided Student’s t-test on the absolute number of mutational events as calculated by FALCOR, assuming equal cell densities [see Materials and methods]) (b) For correlation analysis, all parallel lines were subdivided in two groups according to their higher or lower than wild type mutation rate. Spearman correlation analysis resulted in a highly significant positive correlation (p<0.001). Lines with a mutation rate lower or equal than the wild-type mutation rate are therefore correlated with lower ethanol tolerance, whereas lines with a higher mutation rate than the wild-type mutation rate are correlated with high ethanol tolerance. In conclusion, these data suggest that hypermutation is necessary for adaptation to high EtOH stress.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.012

Inline graphic — (a) The population mutation rate of parallel evolved lines relative to the wild type mutation rate is shown (mean ± 95% c.i., see Materials and methods). Two different groups can clearly be distinguished according to the higher than wild-type () or lower than wild-type () mutation rate. This subdivision is in accordance with the difference in endpoint EtOH tolerance levels (Figure 2b). All mutation rates were significantly different from the wild type (p<0.001; two-sided Student’s t-test on the absolute number of mutational events as calculated by FALCOR, assuming equal cell densities [see Materials and methods]) (b) For correlation analysis, all parallel lines were subdivided in two groups according to their higher or lower than wild type mutation rate. Spearman correlation analysis resulted in a highly significant positive correlation (p<0.001). Lines with a mutation rate lower or equal than the wild-type mutation rate are therefore correlated with lower ethanol tolerance, whereas lines with a higher mutation rate than the wild-type mutation rate are correlated with high ethanol tolerance. In conclusion, these data suggest that hypermutation is necessary for adaptation to high EtOH stress.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.012

Mixed pools and one characterized clone of the endpoints of all high tolerant and two low tolerant lines were subjected to whole genome resequencing. In all lines that developed high ethanol tolerance, a prominently high number of mutations was present, compared to the number of mutations in the low tolerant lines (Figure 2—figure supplement 2). This confirms the existence of a hypermutation phenotype in the highly tolerant lines. Furthermore, the observed mutational spectrum reveals the typical pattern expected for methyl-directed mismatch repair (MMR) mutators, in which case transitions are strongly favored over transversions (Schaaper and Dunn, 1987) (Figure 2—figure supplement 3a). We therefore scanned the population sequence data for mutations in genes involved in DNA replication and repair. We found different mutS mutations in six out of eight highly ethanol-tolerant lines (lines E1, 3, 9, 12, 19 and 20). In the remaining highly tolerant lines E5 and E14, as well as in E9, we found fixed mutations in mutL, while line E9 additionally acquired a mutation in mutH, strongly suggesting a deficient MMR pathway (mutS,L,H) as the main cause of increased mutation rates under near-lethal ethanol stress (Figure 2—figure supplement 3b). In addition, mutations in other possible mutator genes (xthA, mutY and uvrD) appeared later on in evolution (i.e. after the occurrence of the MMR mutations).

To confirm the role of MMR, we evaluated the selective advantage of a specific mutS point mutation (G100A) originating from one of the highly-tolerant lines (E1), in the presence of 5% EtOH. This mutation is located near the mismatch recognition site of MutS and causes an approximately 10-fold increase in mutation rate (Figure 4a). The mutS_G100A and ∆mutS strains were competed directly against the wild-type strain under 5% EtOH (Figure 4—figure supplement 1a). The frequency of both mutator strains increased for all initial cell ratios, resulting in rapid fixation in the population after one growth cycle (Figure 4b). Additionally, we observed a distinct pattern for both mutation rate variants. Calculation of the relative fitness (Van den Bergh et al., 2016) (Figure 4—figure supplement 1b) confirms our previous results (Figure 1, Figure 1—figure supplement 3) showing that a ∆mutS mutant is more fit under 5% EtOH stress as compared to variants with a lower mutation rate, such as the mutS_G100A mutant.

Even though all mutator mutations were 100% fixed in the population, we were able to detect a vast amount of low frequency mutations present in the population (Figure 2—figure supplement 2). The graph reveals the difference between total amount of mutations (>10% frequency) and number of ‘fixed’ mutations (>75% frequency). Only a fraction of the variants are fixed in the hypermutating lines, suggesting a complex population structure with different subpopulations (Pulido-Tamayo et al., 2015). Consequently, the population mutation rate will reflect the average genomic mutation rate of the entire population, containing different subpopulation that possibly display above- or below-average mutation rates. This may already explain the discrepancy between the 20-fold increased endpoint mutation rate of line E1 (Figure 3a) and the 10-fold increased clonal mutation rate caused by the mutS_G100A mutation identified in that same line (Figure 4a). Furthermore, these data suggest that mutation rate can vary along with population structure throughout the evolution rather than being a fixed rate after the occurrence and spread of one mutator mutation.

Dynamics in mutation rate underlie evolution to high ethanol tolerance

To understand dynamics of mutation rates, we analyzed the occurrence of variations in mutation rate during evolution towards high ethanol tolerance by measuring the genomic mutation rate of populations sampled at different time points during the evolution experiment. The results reveal a dynamic pattern of rapidly altering mutation rates (Figure 5a). Furthermore, there is a significant, positive correlation (p<0.05) between changes in ethanol tolerance and differences in mutation rate between two consecutive time points (Figure 5b). Next, we measured the mutation rate of two selected intermediate point (IM2 and IM3, Figure 6a) in the presence of 7% ethanol. The relative fold-change decrease in mutation rate between these two points was unaffected by the presence of ethanol compared to the ±7-fold change in the absence of ethanol, strongly suggesting that ethanol itself does not change the dynamics caused by differences in the genomic mutation rate (Figure 5—figure supplement 1). In conclusion, increased mutation rate co-occurs with increased ethanol tolerance, likely because mutator mutations hitchhike along with a beneficial mutation. This is in line with results from our previous competition experiments (Figures 1, 4), showing that direct selection on adaptive mutations increases ethanol tolerance while the hitchhiking of mutator mutations increases the mutation rate in the population accordingly.

Figure 6. — (a) The specified intermediate time points from high EtOH-tolerant line E5 were selected based on their different mutation rate (Figure 5). (b) The death rate constant for each intermediate point is shown (mean, n = 3, one-phase exponential decay fitting on the decrease in viable cell count, see Materials and methods, Figure 6—figure supplement 1). The death rate constants were statistically compared using a one-way ANOVA with post hoc Tukey correction (*p<0.05). (c) The percentage dead cells as determined by live-dead staining (see Materials and methods) is shown for each intermediate time point. The fractions of dead cells were statistically compared using a two-sided Student’s t-test (mean, n = 3, *p<0.05). An increase in mutation rate coincides with both an increase in death rate constant and an increase in fraction of dead cells in the population. A subsequent decrease in mutation rate coincides with a decrease in death rate constant and fraction of dead cells. These data suggest that higher mortality is the cost of hypermutation when adaptation to a certain level of EtOH stress is achieved.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.017

Figure 6—figure supplement 1. — (a) The specified intermediate time points from high EtOH-tolerant line E5 were selected based on their different mutation rate (Figure 5). (b) The death rate constant for each intermediate point is shown (mean, n = 3, one-phase exponential decay fitting on the decrease in viable cell count, see Materials and methods, Figure 6—figure supplement 1). The death rate constants were statistically compared using a one-way ANOVA with post hoc Tukey correction (*p<0.05). (c) The percentage dead cells as determined by live-dead staining (see Materials and methods) is shown for each intermediate time point. The fractions of dead cells were statistically compared using a two-sided Student’s t-test (mean, n = 3, *p<0.05). An increase in mutation rate coincides with both an increase in death rate constant and an increase in fraction of dead cells in the population. A subsequent decrease in mutation rate coincides with a decrease in death rate constant and fraction of dead cells. These data suggest that higher mortality is the cost of hypermutation when adaptation to a certain level of EtOH stress is achieved.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.017

Cellular mortality is the underlying force driving evolution of mutation rates

Remarkably, the mutation rate decreases quickly in the long-term evolution experiment during periods when the concentration of ethanol is kept constant (Figure 5a). This fast decrease can either be explained by reversion of mutator mutations or by the accumulation of compensatory suppressor mutations (Wielgoss et al., 2013). We tested the former by targeted sequencing of the mutations that were acquired in the MMR genes in intermediate points before and after the decrease in mutation rate. No such reversions of mutator mutations were found. These results therefore suggest that suppressor mutations have accumulated in the ethanol-tolerant mutator lines. To unravel the benefit of a lower mutation rate and the cost of hypermutation when high ethanol tolerance is reached and thus selective pressure for ethanol tolerance is not further increased, we selected intermediate time points (IM1, IM2 and IM3) with contrasting mutation rates (Figure 6a). These intermediate points were used to determine cell viability under high ethanol stress and to extract relevant growth parameters by fitting a bacterial growth equation to the growth dynamics (See Materials and methods). Cell death was determined both by quantifying viable cells and by live-dead staining. Surprisingly, all tested intermediate points showed a very fast decrease in viable cell count when entering the stationary phase (Figure 6—figure supplement 1). This decrease in viable cells is explained by a genuine increased death rate in the population since cells are in stationary phase. We fitted this decrease in viable cells with an exponential decay function and extracted the death rate constant (See Methods). The fitting data demonstrate an association between an increase in mutation rate and an increase in death rate constant (Figure 6b). In addition, death rates significantly decline as mutation rates decrease. These results were confirmed by live-dead staining and subsequent flow cytometry analysis (Figure 6c). Moreover, a constant death rate was measured throughout the growth cycle of the strain (Figure 6—figure supplement 2). Intermediate evolved strains with increased ethanol tolerance are thus characterized by high death rates which are dependent on the mutation rates. Likely, these strains have accumulated a high genetic load throughout the evolution experiment. Our data now suggest that a further buildup of genetic load and a higher chance to acquire a lethal mutation cause increased mortality, which results in a selective pressure per se. Strains with lower mutation rates resulting from compensatory mutations can increase their fitness due to decreased death rates. To corroborate these data, we used time to grow and optical density data from the initial evolution experiment to calculate the relative fitness of IM3 (low mutation rate) compared to IM2 (high mutation rate) in the presence of 7.5% ethanol. We used the relative fitness to calculate the theoretical number of selection rounds necessary for IM3 to fix in a population of IM2 and compared it to the actual number of selection rounds (Figure 6—figure supplement 3). The discrepancy between those two values suggests that fixation happened faster than theoretically possible given the calculated relative fitness. Additional mutations that occurred between IM2 and IM3 might affect the speed of selection. Further buildup of genetic load and a continuous higher chance of acquiring a lethal mutation, will speed up the elimination of the high mutation rate genotype (IM2) and enhance the fixation rate of the low mutation rate genotype (IM3).

To confirm the role of cellular mortality as modulator of mutation rates, we selected an evolved intermediate point of line E9, with a high mutation rate resulting from fixed MMR mutations (Figure 2—figure supplement 3). Next, we re-evolved this population for 150 generations on the same percentage of ethanol, without further increasing this concentration when adaptation occurs, to mimic and prolong plateau conditions experienced in the original evolution experiment. We observed a fast decrease in population mutation rate (Figure 7a). Live-dead staining on both initial and endpoints, shows a decrease in cellular mortality linked to the decrease in mutation rate (Figure 7b). Further, we increased the mutation rate again by deleting the mutS gene (Figure 7c). By monitoring the number of viable cells, we observed a much higher death rate for the strain with increased mutation rate (END ∆mutS) (Figure 7d). Moreover, the strain with a low mutation rate rapidly outcompetes the ∆mutS strain in direct competition under 7% EtOH (Figure 7—figure supplement 1). Again, we used the relative fitness to calculate the theoretical number of selection rounds needed for the END strain to outcompete the END ΔmutS strain (Figure 7—figure supplement 2). In contrast to Figure 6—figure supplement 3 the calculated rounds now correspond to the actual observed rounds. Here, both strains are isogenic apart from the mutS deletion, so the fitness only reflects the benefit of the anti-mutator (an intact mutS gene) that leads to a lower mutation rate and lower mortality. In summary, these results show that mutation rate and mortality are crucial factors to explain the fast increase of genotypes with a low mutation rate and mortality when the strain is already adapted to the environment.

Figure 7. — (a) A selected intermediate point of evolved highly EtOH-tolerant line E9 with an increased mutation rate was evolved during approximately 150 generations on the same percentage of 7% EtOH. After 150 generations the genomic mutation rate, measured in the absence of ethanol, significantly decreased to almost the ancestral mutation rate (mean ± 95% c.i., ***p<0.001, see Materials and methods). (b) Measuring the percentage of dead cells revealed a higher death rate in the START point with high mutation rate compared to the END point with a lower mutation rate (two-sided Student’s t-test, mean ± s.d., n = 3, **p<0.01). (c) To confirm the role of mortality as modulator of cellular mutation rate, the *mutS* gene was deleted in the END point with a low mutation rate. This deletion caused a significant increase in mutation rate (mean ± 95% c.i., ***p<0.001, see Materials and methods). However, the increase in mutation rate is less pronounced as for the *mutS* deletion mutant in the clean wild-type background, suggesting the presence of mutations that not only compensate for the original mutator mutation in E9 but also for a deletion of *mutS*. (d) The number of viable cells decreases significantly for both low and high mutation rate variants during growth on 7% EtOH, although this decrease in the strain with a low mutation rate is less compared to the strain with a high mutation rate (two-sided Student’s t-test, mean ± s.d., n = 3, **p<0.01, ***p<0.001). These results show a lower mortality for a strain with a lower mutation rate, resulting in a competitive advantage in an EtOH environment to which the strain is already adapted.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.021

Figure 7—figure supplement 1. — (a) A selected intermediate point of evolved highly EtOH-tolerant line E9 with an increased mutation rate was evolved during approximately 150 generations on the same percentage of 7% EtOH. After 150 generations the genomic mutation rate, measured in the absence of ethanol, significantly decreased to almost the ancestral mutation rate (mean ± 95% c.i., ***p<0.001, see Materials and methods). (b) Measuring the percentage of dead cells revealed a higher death rate in the START point with high mutation rate compared to the END point with a lower mutation rate (two-sided Student’s t-test, mean ± s.d., n = 3, **p<0.01). (c) To confirm the role of mortality as modulator of cellular mutation rate, the *mutS* gene was deleted in the END point with a low mutation rate. This deletion caused a significant increase in mutation rate (mean ± 95% c.i., ***p<0.001, see Materials and methods). However, the increase in mutation rate is less pronounced as for the *mutS* deletion mutant in the clean wild-type background, suggesting the presence of mutations that not only compensate for the original mutator mutation in E9 but also for a deletion of *mutS*. (d) The number of viable cells decreases significantly for both low and high mutation rate variants during growth on 7% EtOH, although this decrease in the strain with a low mutation rate is less compared to the strain with a high mutation rate (two-sided Student’s t-test, mean ± s.d., n = 3, **p<0.01, ***p<0.001). These results show a lower mortality for a strain with a lower mutation rate, resulting in a competitive advantage in an EtOH environment to which the strain is already adapted.

**DOI:** http://dx.doi.org/10.7554/eLife.22939.021

Discussion

By using experimental evolution, we observed rapid emergence of hypermutation during de novo adaptation to near-lethal ethanol stress. While mutators only sporadically occur in laboratory evolution experiments using mild stress conditions (Sniegowski et al., 1997; Barrick et al., 2009; Sandberg et al., 2014), all highly ethanol-tolerant lines in our study acquired a hypermutation phenotype. We provide evidence that lethal environments trigger a shift in the optimal balance between keeping a constant genetic load and mutational supply towards a higher supply rate. Despite the burden of additional, possibly lethal mutations, the increased mutational supply enables fast adaptation of at least some individuals and rescues the population from extinction (Bell and Gonzalez, 2011; Lindsey et al., 2013; Gonzalez et al., 2013; Gonzalez and Bell, 2013). Unexpectedly, by measuring the mutation rate at different time points during evolution, we found a highly dynamic mutation rate that recurrently increases as a response to enhanced ethanol pressure and decreases again once cells are adapted to the stress.

The rise of hypermutation during adaptation to near-lethal ethanol stress is possibly linked to the idea of second-order selection as suggested by the growth rate and lag time measured for a collection of mutator mutants under 5% ethanol stress (Figure 1; Figure 1—figure supplement 3). In addition, it has been reported previously that random occurrence of mutator mutations in the population is facilitated by a wider and less deleterious-mutation-biased distribution of fitness effects in changing environments (Hietpas et al., 2013). As a consequence of the lowered relative mutational load, populations in harsh environments may thus consist of cells with various mutation rates as they tolerate more hypermutators, possibly offering a valid additional explanation for the increased mutation rate observed in the ethanol tolerant lines (Figure 3). However, we observed that mutator mutations are fixed in the end point populations (Figure 2—figure supplement 3) and that non-mutator lines are not able to adapt to high tolerance levels (Figure 3), rather pointing to the former hypothesis of second-order selection or even a combination of the two explanations. In this scenario, the initial emergence of mutators is facilitated in populations exposed to severe stress (Hietpas et al., 2013) followed by hitchhiking of the mutator mutation along a physically linked (combination of) beneficial mutation(s) where selection act on (Woods et al., 2011). Additionally, this would mean that mutator mutations do not have a direct selective advantage themselves, but instead are only beneficial through enabling rapid adaptation by increasing the mutational supply rate. However, the following elements in our results might also suggest potential direct effects of the mutator mutations. First, mainly point mutations were identified in the mismatch repair genes (Figure 2—figure supplement 3) although inactivation of a gene is more likely to occur, given the high competitive benefit of the ΔmutS strain compared to the mutS_G100A strain in the presence of 5% ethanol (Figure 4—figure supplement 1). This would suggest selection of specific changes in the mechanism of the mismatch repair pathway. Second, both the ΔmutS and the mutS_G100A strains still increase in frequency when competed against the wild type in a ratio of 1:1000 or lower (Figure 4). Given the 10- to 50-fold increased mutation rate, a mutator subpopulation at a ratio of 1:1000 or lower should be too small to have an increased chance of acquiring a beneficial mutation compared to the wild-type subpopulation. These data suggest direct beneficial effects of MMR mutations (Cooper et al., 2012; Torres-Barceló et al., 2013; Nowosielska and Marinus, 2008) that, combined with second-order selection, can explain our observations. In addition, we confirmed that any increase in mutation rate, irrespective of the disrupted cellular system, can confer a selective benefit. Therefore, these direct effects, which are usually the result of disruptions of one specific system or even of one specific gene, may influence which specific mutator mutations eventually spread, but will only have a limited effect on the initial selection of hypermutation compared to the direct effect of linked beneficial mutations. However, at later stages these direct effects possibly affect the fate of hypermutators by lowering the cost of the extended buildup of genetic load.

Surprisingly, in addition to the increase in mutation rate, we found that hypermutator states are transitory and mutation rates decrease again once cells are adapted to the stressful environment. Current knowledge of the mechanisms underlying these changes in mutation rate remains largely fragmentary (Raynes and Sniegowski, 2014). In this study, we identified cellular mortality as major modulator of the population mutation rate. A higher mutation rate is linked to a higher mortality, probably due to the extended buildup of genetic load and increased probability of acquiring a lethal mutation. However, mutation rate is not the only factor affecting the mortality. Some direct effects of already accumulated mutations might explain the inconsistency between the 2-fold difference in mortality between IM2 and IM3 with a 10-fold difference in mutation rate (Figure 6) and the more than 10-fold difference in mortality between END and END ΔmutS with only a 2.5-fold difference in mutation rate (Figure 7). Notwithstanding this non-linearity, differences in mortality confer a selectable pressure that favors strains with lower mutation rates when cells are adapted to the environment. Selection of lower mutation rate genotypes that arise in an adapted population of high mutation rate genotypes is probably enhanced by the faster decrease of the high mutation rate genotypes compared to the low mutation rate genotypes due to their differences in mortality. Therefore, these findings might explain the recurrent mutation rate alterations observed in our evolution experiment. Nevertheless, the speed of mutation rate alterations clearly differs from an earlier report, showing that a single gradual decrease in mutation rate, due to invasion of a mutY anti-mutator in a mutT mutator line, occurred over a relatively long time span of at least 1000 generations (Wielgoss et al., 2013). While it was difficult to observe fitness benefits of anti-mutators under these less restrictive stress conditions, we here report the observation of much higher benefits under near-lethal ethanol conditions, allowing rapid, mortality-driven changes of the mutation rate.

Finally, by analyzing growth characteristics of a panel of mutators, we observed a range of mutation rates enabling fast growth under near-lethal ethanol stress. These results substantiate the theoretical modelling work of Bjedov et al. that predicts the highest fixation probability for a 10- to 100-fold increased mutation rate and decreasing fixation probabilities for weaker or stronger mutators (Bjedov et al., 2003) (Figure 1—figure supplement 3). Furthermore, our work extends the findings by Loh et al. showing that different PolA mutants with altered mutation rates predominate after serial passage in a fluctuating environment (Loh et al., 2010). In contrast to this study, we used genes of several distinct cellular pathways to enhance mutagenesis. That way, we were able to demonstrate that even minor alterations in mutation rate, irrespective of the targeted cellular system, can confer a competitive advantage under near-lethal, complex stress.

Both these observations corroborate the idea that moderate mutators will be more easily selected for, because their benefit is higher than low mutation rate variants and their long-term cost is lower than high mutation rate variants. The identification of mostly point mutations leading to amino acid changes and not to nonsense mutations in the MMR genes during evolution similarly suggests selection for mild increases in the mutation rate (such as shown for the mutS_G100A mutant).

Interestingly, the occurrence of hypermutation under extreme stress is not only limited to prokaryotes, such as E. coli. Previously, hypermutators were also observed in S. cerevisiae during evolution under ethanol stress (Voordeckers et al., 2015), in the malaria-causing parasite Plasmodium falciparum (Lee and Fidock, 2016; Gupta et al., 2016), the fungal pathogen Candida glabrata (Healey et al., 2016) and in temozolomide-treated, relapsed glioblastoma tumors (Wang et al., 2016). These examples demonstrate the relevance of hypermutation in eukaryotes exposed to severe stress. However, the lower emergence of mutators compared to our study may be explained by the ploidy of eukaryotic cells and their larger genetic arsenal (Thompson et al., 2006), which allows for more alternative adaptive routes to cope with stress.

Even though we mainly focused on increased mutation rates in this study, the 12 slowly-mutating, low ethanol tolerant lines might be an interesting starting point for further research. The lack of hypermutation in these lines seems to impede further adaptation to high ethanol concentrations. Sequence analysis of two of these lines revealed the presence of mutations in rpoZ (subunit of the RNA polymerase) and infB (protein chain initiation factor) in lines E4 and E17, respectively. Since ethanol is toxic through its effect on transcription and translation (Haft et al., 2014), disruption of the transcription or translation machinery due to these mutations might cause increased sensitivity to higher ethanol levels. This would prevent further growth and the possibility of acquiring a mutator allele or any other mutation. Although we have no evidence supporting that the rpoZ and infB mutations are causal for the decreased mutation rate, these mutations are interesting and might explain the lack in further adaptive improvement in lines E4 and E17 (Figure 3a).

In conclusion, while an organism’s mutation rate is generally considered a slowly-evolving parameter, we demonstrate an unexpected flexibility in cellular mutation rates matching changes in selective pressure to avoid extinction under near-lethal stress. Bacteria undergoing antibiotic treatment or cancer cells exposed to chemotherapy are prime examples of cells exposed to stressful conditions. Therefore, hypermutation should be considered a risk for both the development of multidrug resistance in pathogenic bacteria (Hammerstrom et al., 2015; Jolivet-Gougeon et al., 2011; Blázquez, 2003; Chopra et al., 2003) and cancer relapses as recently shown (Wang et al., 2016). Targeting hypermutation could pave the way not only for the development of novel anti-cancer therapies, but also for containing the spread of multidrug tolerant pathogens and even for the generation of robust, stress-resistant strains for use in various industrial processes.

Materials and methods

Bacterial strains and culture conditions

E. coli SX4, SX25, SX43 and SX43 ∆venus, all derived from BW25993 (Datsenko and Wanner, 2000), were used in this study. SX4 is characterized by a genomic tsr-venus fusion inserted in the lacZ gene under control of the lac-promoter (Yu et al., 2006). SX43 is a derivative of SX4 in which the kanamycin-resistance cassette (Km^R) was removed by Flp-mediated recombination (Van den Bergh et al., 2016; Cherepanov and Wackernagel, 1995). The tsr-venus fusion results in a polarly localized fluorescent Venus tag. E. coli SX25 expresses a genomic venus marker inserted in the lacZ gene under control of the lac-promoter (Yu et al., 2006). SX43 ∆venus is a non-fluorescent variant constructed by P1vir transduction (Thierauf et al., 2009) using the lacI::Km^R Keio mutant (Keio collection number JW0336) (Van den Bergh et al., 2016; Baba et al., 2006) as donor. All strains were grown in an orbital shaker at 200 rpm and 37°C in liquid lysogeny broth (LB) medium or on LB agar plates.

Construction of deletion mutants

Target genes to increase the genomic mutation rate were selected based on their various roles in DNA replication and repair (Supplementary file 1A). Hypermutating variants of the ancestor were generated by P1vir transduction (Thierauf et al., 2009) to the SX43 ancestor using the corresponding Keio deletion mutants as donor strains (Baba et al., 2006) (RRID:SCR_002303). Transductants were subsequently selected on kanamycin resistance. Correct deletion of the target genes in positive colonies was confirmed by PCR (Supplementary file 1B). The mutS deletion was introduced using the protocol described by Datsenko and Wanner (Datsenko and Wanner, 2000). In short, a kanamycin-resistance cassette flanked by FRT sites was amplified from the pKD4 plasmid using primers with homologous ends complementary to the flanking sequences of the mutS gene (Supplementary file 1B). This PCR product was electroporated in the ancestor in which the λ-red genes for homologous recombination were expressed from the pKD46 plasmid. Positive colonies were selected on kanamycin resistance and correct deletion of the mutS gene was assessed by PCR.

Determination of near-lethal EtOH percentage

To assess the percentage of EtOH needed to expose cells to near-lethal stress, the growth dynamics of wild-type E. coli in different levels (0–5% (v/v)) of EtOH were studied. Data were fitted to a Gompertz equation for bacterial growth dynamics (see Equation 1) to extract relevant growth parameters (Figure 1—figure supplement 2). 5% EtOH was considered a breakpoint concentration as an abruptly increased doubling time and decreased carrying capacity is observed under these conditions. To ensure the proper percentages of ethanol added to the medium we used the Alcolyser beer analyzing system (Anton Paar GmbH, Austria).

Determination of lag time and growth rate

Lag times and growth rates of selected mutants were determined using the Bioscreen C system for automated monitoring of microbial growth (Bioscreen C MBR, Oy Growth Curves AB Ltd., Finland) (RRID:SCR_007172). Cells were grown in 10 × 10 well Honeycomb microplates with shaking at 37°C and optical density at A_600nm was automatically measured every 15 min. Growth in each well was monitored for five days. First, the optical density of each preculture was equalized to approximate equal numbers of cells for all mutants. Next, dilution series were made ranging from 10⁻¹ to 10⁻⁴. To test the effect of different starting amounts of cells on the lag time in the presence of EtOH, 20 µl of each dilution was used to inoculate 180 µl of LB medium supplemented with 5% EtOH. These dilutions correspond to the different inoculum sizes as shown in Figure 1 and Figure 1—figure supplement 3. For every mutant all dilutions were tested in biological triplicate. Additionally, each well was covered with 100 µl of mineral oil (BioReagent, Sigma-Aldrich, MO, USA) to prevent evaporation of EtOH as previously described (Zaslaver et al., 2006). As previously shown (Kaplan et al., 2008; Zaslaver et al., 2004; Ronen et al., 2002) mineral oil has no significant effect on growth or aeration. Growth curves were fitted using the widely accepted Gompertz equation as previously described (Zwietering et al., 1990). In this equation, y₀ is the starting density, y_M is the carrying capacity of the population, SGR is the specific growth rate (h⁻¹) and LT is the lag time (h). Log₁₀ of the optical density values (595 nm) were used to fit with this equation and subsequently extract all growth parameters.

y (x) = y_{0} + y_{M} * e^{[- e^{((2.718 * \frac{S G R}{y_{M}}) * (L T - x) + 1)}]}

(1)

Equation 1: Gompertz equation for fitting of bacterial growth dynamics.

Competition assay

To determine the relative fitness of a mutant compared to the wild type, we conducted direct competition experiments. All mutator mutants carried the Venus fusion that could be detected by excitation at 530 nm, while the ancestor SX43 ∆venus was not fluorescent. The relative fitness of all mutants was assayed in triplicate. Both non-fluorescent ancestor and fluorescent mutator mutants were revived from a glycerol stock stored at −80°C and grown overnight at 200 rpm and 37°C. These overnight cultures were first diluted to an A_595nm of 0.5 to obtain an equal cell quantity for all cultures prior to the competition experiment. Next, equal amounts of ancestor and mutant were diluted 1000x in 50 ml LB medium, containing 5% EtOH. To avoid EtOH evaporation, cultures were grown in flaks with a rubber-sealed screw cap. To start the competition experiment with different mutator versus ancestor ratios, corresponding amounts of both strains were mixed. The initial ratios were verified by flow cytometry using a BD Influx cell sorter equipped with a 488 nm laser (200 mW) and standard filter sets (530/40 nm for Venus detection). To assess relative fitness, mixed populations were incubated for 48 hr in the presence of 5% EtOH. Ratios of both ancestor and wild type were subsequently determined by flow cytometry. For each sample at least 100.000 cells were counted and each cell was tallied as ancestor or mutator based on their fluorescence intensity. Loss of fluorescence in the mutator mutants was accounted for by measuring fluorescence and loss of fluorescence in the individually grown ancestor and all mutator mutants. First, the fraction false positive non-fluorescent mutators was quantified for each mutant. This fraction was subsequently used to correct the measured mutator versus wild type ratios to account for non-fluorescent mutators.

As the results of this test reflect ratios of ancestor and mutant in the population before and after growth on 5% EtOH, the proportion of a less fit population is expected to decline relative to the other. Therefore, we calculated the relative fitness, W, as previously described (Equation 2) (Van den Bergh et al., 2016). This equation is based on a discrete time-recurrence equation that describes the spread of mutant in a haploid population thereby defining relative fitness based on differences in survival between two strains (Van den Bergh et al., 2016; Otto and Day, 2007). This equation allows to calculate the relative fitness, W_A, of a mutant A compared to its competitor, based on the difference between the final detected proportion (A_end) and the initial proportion (A_start) of the mutant in the population, given a certain number of selection rounds n. Significance of difference from 1, where the mutant has no benefit over the wild type, was determined using a repeated measures ANOVA with post hoc Dunnett correction. An F-test was used to assess the difference in variance between the groups that were statistically compared. In order to confirm marker neutrality, we competed the fluorescent ancestor (SX43) and non-fluorescent ancestor (SX43 ∆venus) against each other both in the absence and presence of EtOH. No significant difference from one was observed using a one-sample Student’s t-test (data not shown). Additionally, the neutrality of the Venus marker was already confirmed in a previous study (Van den Bergh et al., 2016).

W_{A} = \frac{1}{{[\frac{((1 - A_{e n d}) \times A_{s t a r t})}{(A_{e n d} \times (1 - A_{s t a r t})}]}^{(\frac{1}{n})}}

(2)

Equation 2: Relative fitness of mutant A compared to its competitor

Experimental evolution

The 20 parallel populations of the evolution experiment originated from independent colonies of the ancestral strains SX4 or SX25. Odd lines were founded by SX4, while even lines were founded by SX25 to enable detection of cross-contamination between parallel lines. All strains were initially grown in LB medium containing 5% EtOH. This percentage was found to mimic near-lethal stress (Figure 1—figure supplement 2). The culture volume during the evolution experiment was 50 ml and dilution was 100-fold at each transfer. Ethanol tolerance of a population was measured as the ability to grow in liquid medium in the presence of a certain percentage of ethanol to an optical density of at least 0.2. Intermediate time points, sampled every transfer to fresh medium, were stored in a −80° glycerol stock. Growth in exponential phase was maintained throughout the evolution experiment to select for growth rate and lag time only and to minimize potential, unwanted effects and genomic changes due to stress that would additionally be experienced by nutrient limitation. Two different parameters were used to monitor the evolution of the independent lines. First, the optical density was measured. A strain with an A_595nm around 0.2 was assumed to be in exponential phase. Second, the time needed for the strain to achieve exponential growth at an A_595nm around 0.2 was used to determine the degree of adaptation. When the population reached exponential growth within 24 hr, we assumed it was fully adapted to a certain percentage of EtOH. We transferred the population to new LB medium containing 0.5% more EtOH than in the previous step. By increasing the percentage of EtOH during the adaptation of the populations, the near-lethal stress was maintained. In later stages of the evolution experiment at very high EtOH concentrations of 7.5% or more, an increase of 0.5% was found to be excessive. Therefore, we increased the percentage with 0.25% EtOH starting from 7.5% EtOH tolerance. If the strain needed more than 24 hr but less than 14 days to reach exponential growth, we assumed that it was not yet completely adapted to a certain concentration of EtOH. Therefore, we transferred the strain to new LB medium containing the same percentage of EtOH as in the previous step. If the strain needed more than 14 days to grow, we assumed this population died out. Therefore, we revived the previously stored time point and used it to restart the evolving line in new LB medium with 0.5% EtOH less than the tolerance level of this time point (Figure 2—figure supplement 1). The minimal optical density of 0.2 upon transfer resulted in an average final cell density of 5.4 × 10⁸ CFU/ml in a volume of 50 ml. For each passage, we consequently transferred 500 µl or approximately 2.73 × 10⁸ CFUs (N₀). The average number of generations (g) for each growth cycle, estimated based on the optical density reached upon transfer, is 6.67. Taking these values together, we can calculate an estimated effective population size (N_e) of 1.82 × 10⁹ using the formula (Lenski et al., 1991⁹¹) N_e = gN₀. Finally, the number of generations are estimates calculated with a previously described equation, assuming equal growth of the entire population based on optical density and time (Wiser et al., 2013) (Equation 3). In this equation CFU_i is the number of viable cells at the start of each cycle, while CFU_e is the number of viable cells at the end of each cycle and c is the total number of cycles. The number of viable cells was estimated using optical density (A_595nm) values. This calculation does not specifically take into account the death rate of the cells. However, since the optical density reflects both living and death cells in the culture, calculating the number of generations using the OD values is more accurate compared to calculations based on CFU counts. Indeed, when using the viable cell count data of IM1, IM2 and IM3 (Figure 5), we found a 1.27-fold (±0.29) underestimation of the number of generations when using the number of viable cells as compared to using OD values.

\sum_{c = 1}^{n} \log_{2} (\frac{C F U_{e}}{C F U_{i}})

(3)

Equation 3: Estimation of the number of generations per cycle based on the initial and final number of viable cells.

Fluctuation assay

The genomic mutation rate of strains of interest was estimated using a Luria-Delbrück fluctuation assay. This assay is commonly used to determine the spontaneous mutation rate at different loci in the genome, where mutations cause easily-scored phenotypic changes. Acquiring rifampicin resistance through mutations in the rpoB gene was used as a measurable marker. This protocol was adapted from the one described by Jeffrey Barrick (http://barricklab.org/twiki/bin/view/Lab/ProtocolsFluctuationTests). The selected strains were revived from a glycerol stock stored at −80°C and grown overnight in an orbital shaker at 200 rpm and 37°C. All strains were tested in at least two independent biological replicates. The number of cells in each culture was determined using an optical density versus cell count standard curve and was subsequently equalized over all tested strains. Next, the equalized cultures were diluted 100 times in fresh LB medium and grown in an orbital shaker at 200 rpm and 37°C for 2–3 hr until the optical density at 595 nm reached 0.2–0.4. At this optical density the final cell density in solution did not exceed 2–4 × 10⁸ CFU/ml. These preconditioned strains were then diluted in LB medium to a density of 5000 cells per ml, which is denoted as the master inoculum mix. The master inoculum mix was divided in replicate cultures of 200 µl in separate Eppendorf tubes or a 96-well plate. For each strain at least 30 replicate parallel cultures, divided over minimum two biological repeats were used to determine the number of spontaneous mutants. These cultures were grown for 24 hr and plated on LB agar supplemented with 100 µg/ml rifampicin to determine the number of spontaneous mutants that arose during the growth period. Additionally, for each biological repeat, at least four cultures were grown for 24 hr, diluted and plated on LB agar to determine the total number of viable colonies. The colonies on the non-selective LB agar plates were counted after 24 hr, while colonies on the selective, rifampicin plates were first counted after 48 hr and again after 72 hr. The number of mutants divided by the total number of cells gives a mutation rate estimate of the tested strain. The occurrence of rifampicin resistance conferring mutations in the rpoB is extrapolated to estimate the global genomic mutation rate. In recent years, many improvements were made to the statistical estimation of mutation rates based on the number of mutants and the total number of cells. We used the Ma-Sandri-Sarkar Maximum Likelihood Estimation method as implemented in the Fluctuation Analysis Calculator (FALCOR, http://www.keshavsingh.org/protocols/FALCOR.html) (Hall et al., 2009). For the statistical analysis on the mutation rate estimates, 95% confidence intervals, calculated by the FALCOR, were compared. In the case of confidence interval overlap, mutation rates were statistically compared using a two-sided Student’s t-test on the normally distributed absolute number of mutational event as calculated by FALCOR (Hall et al., 2009). To ensure correct comparison of the mutation rates, we verified that the population densities at the time of plating on rifampicin did not differ significantly. If this was not the case, the test was repeated. The statistical difference between the population densities was measured using a one-way ANOVA with post-hoc Tukey correction. We found no significant difference for any intermediate point compared to the average density and to the density of the other time points. We therefore avoid possible population density effects (Krašovec et al., 2014).

To study the change of mutation rates during adaptive evolution, we performed a correlation analysis between the difference in mutation rate and the difference in EtOH tolerance at each time point. Since EtOH tolerance during evolution only increases by 0.25% or 0.5%, we considered the difference in EtOH tolerance as a discrete ordinal variable. Therefore, we used the non-parametric Spearman method to determine the significance of the correlation between the difference in EtOH and the difference in mutation rate between consecutive time-points (Graphpad Prism 6, CA, USA).

Whole-genome sequencing and identification of mutations

High-quality genomic DNA was isolated from overnight cultures of the ancestor and end points of evolved lines (DNeasy Blood and Tissue kit, Qiagen). We isolated genomic DNA from both mixed pools and one characterized clone of each high ethanol tolerant line and two low ethanol tolerant lines. Both Figure 2—figure supplement 2 and Figure 2—figure supplement 3 represent the results from the analysis of the pooled sequence data. Concentration and purity of the DNA were determined using Nanodrop analysis (Thermo Fisher Scientific, MA, USA), gel electrophoresis and Qubit analysis (Thermo Fisher Scientific). Libraries were prepared at GeneCore (EMBL, Heidelberg, Germany) (RRID:SCR_004473) using the NEBNext kit with an average insert size of 200 bp. The DNA libraries were multiplexed and subjected to 100-cycle paired-end massive parallel sequencing with the Illumina HiSeq2000 (RRID:SCR_010233) (GeneCore, EMBL, Heidelberg, Germany). CLC Genomics Workbench version 7.6 (RRID:SCR_011853) (https://www.qiagenbioinformatics.com) was used for analysis of the sequences. Following quality assessment of the raw data, reads were trimmed using quality scores of the individual bases. The quality limit was set to 0.01, and the maximum allowed number of ambiguous bases was set to 2. Reads shorter than 15 bases were discarded from the set. The trimmed reads were mapped (mismatch cost = 2; insertion cost = 3; deletion cost = 3; length fraction = 0.8; similarity fraction = 0.8) to the E. coli MG1655 reference genome (NC_000913.1) using the CLC Assembly Cell 4.0 algorithm yielding an average coverage of approximately 150x. Finally, mutations in all samples were detected using the CLC Fixed Ploidy Variant Detector. To exclude mutations in the SX4 ancestor compared to the MG1655 reference genome, we compared the mutations of all evolved lines with the SX4 ancestor.

Mortality assay

To assess the rate at which cells die during growth, we made growth curves using optical density measurements with concurrent viable cell determination. The ancestor and selected evolved intermediate time points were directly inoculated from a frozen glycerol stock in 50 ml LB medium containing no EtOH, 5% EtOH, or 6.5% EtOH. Each strain was tested in triplicate. All flasks were subsequently grown at 200 rpm and at 37°C. At 30 different time points during a 90 hr timespan the optical density was measured and samples were taken for CFU determination. For each sample, a dilution series was made and appropriate dilutions were plated on LB agar plates using an EddyJet2 spiral plater (IUL Instruments, Spain). Agar plates were grown 48 hr at 37°C and the CFU/ml was determined using the Flash and Go automatic colony counter (IUL Instruments). During growth, the number of CFU/ml initially increases exponentially but then flattens and decreases again. The colony count data corresponding to the decrease in CFU/ml were fitted using an exponential decay function (Equation 4) in GraphPad Prism 6. In this function, k is the death rate constant. For all samples, this constant was determined. Statistical significance of the difference between the death rate constants of two consecutive evolved time points was determined using a two-tailed Student’s t-test.

y = (y_{m a x} - y_{m i n}) * e^{- k * x} + y_{m i n}

(4)

Equation 4: Exponential decay function with k the decay constant

Live-dead staining

To measure the amount of dead cells at a certain time point in a population we used the LIVE/DEAD BacLight Bacterial viability kit (Thermo Fisher Scientific). The selected strains were revived from a frozen glycerol stock and grown overnight in an orbital shaker at 200 rpm and 37°C. Overnight cultures were diluted to an A_595nm of 0.5. Next, 1 µl propidiumiodide (20 mM, Thermo Fisher Scientific) per 1 ml diluted culture was added, vortexed to mix the propidiumiodide homogeneously and incubated in the dark at room temperature for 10 min. Propidiumiodide can only penetrate the cell when the membrane is disrupted, as is the case in dead cells, and can be detected by excitation at 620 nm. Therefore, the amount of dead cells in a population could be determined by flow cytometry. All strains were tested at least in triplicate. To measure the number of dead cells throughout the different growth phases, the selected strain was inoculated at different time points ranging from 48 hr to 10 hr prior to flow cytometry analysis. The amount of dead cells was determined as previously described. Statistical significance was determined using a two-sided Student’s t-test.

Acknowledgements

TS is a fellow of the Agency for Innovation by Science and Technology (IWT). The research was supported by the KU Leuven Research Council (PF/10/010; PF/10/07; IDO/09/010; IDO/13/008; CREA/13/019; DBOF/12/035; DBOF/14/049), Interuniversity Attraction Poles-Belgian Science Policy Office IAP-BELSPO) (IAP P7/28), ERC (241426), Human Frontier Science Program (HFSP) (RGP0050/2013), FWO (G047112N, KAN2014 1.5.222.14), Flanders Institute for Biotechnology (VIB) and the European Molecular Biology organization (EMBO). We thank S. Xie for providing the ancestor strains.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Funding Information

This paper was supported by the following grants:

Agentschap voor Innovatie door Wetenschap en Technologie Strategic Basic Research Fellowship,121525 to Toon Swings.
Fonds Wetenschappelijk Onderzoek Postdoctoral Fellowship,12O1917N to Bram Van den Bergh.
Fonds Wetenschappelijk Onderzoek Postdoctoral Fellowship,1249117N to Karin Voordeckers.
Onderzoeksraad, KU Leuven PF/10/010 to Kevin J Verstrepen, Jan Michiels.
H2020 European Research Council 241426 to Kevin J Verstrepen.
Human Frontier Science Program RGP0050/2013 to Kevin J Verstrepen.
Vlaams Instituut voor Biotechnologie to Kevin J Verstrepen.
European Molecular Biology Organization to Kevin J Verstrepen.
Onderzoeksraad, KU Leuven IDO/09/010 to Kevin J Verstrepen, Jan Michiels.
Onderzoeksraad, KU Leuven DBOF/12/035 to Kevin J Verstrepen, Jan Michiels.
Onderzoeksraad, KU Leuven DBOF/14/049 to Kevin J Verstrepen, Jan Michiels.
Onderzoeksraad, KU Leuven CREA/13/019 to Maarten Fauvart.
Fonds Wetenschappelijk Onderzoek KAN2014 1.5.222.14 to Maarten Fauvart.
Federaal Wetenschapsbeleid Interuniversity Attraction Poles Programme P7/28 to Jan Michiels.
Fonds Wetenschappelijk Onderzoek G047112N to Jan Michiels.
Onderzoeksraad, KU Leuven IDO/13/008 to Jan Michiels.

Additional information

Competing interests

The authors declare that no competing interests exist.

Author contributions

TS, Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing, T.S. conceptualized the study, designed and performed the experiments, analyzed the data and wrote the manuscript.

BVdB, Conceptualization, Supervision, Methodology, Project administration, Writing—review and editing, B.V.D.B. discussed the results and edited the manuscript.

SW, Data curation, Investigation, S.W. helped in performing the experiments.

EO, Data curation, Investigation, E.O. helped in performing the experiments.

KV, Conceptualization, Funding acquisition, Project administration, Writing—review and editing, K.V. conceptualized the study, discussed the results and edited the manuscript.

KJV, Conceptualization, Funding acquisition, Project administration, Writing—review and editing, K.J.V. conceptualized the study, discussed the results and edited the manuscript.

MF, Conceptualization, Supervision, Funding acquisition, Project administration, Writing—review and editing, M.F. conceptualized the study, designed the experiments, discussed the results and edited the manuscript.

NV, Conceptualization, Supervision, Funding acquisition, Investigation, Methodology, Project administration, Writing—review and editing, N.V. conceptualized the study, designed the experiments, discussed the results and edited the manuscript.

JM, Conceptualization, Supervision, Funding acquisition, Investigation, Methodology, Project administration, Writing—review and editing, J.M. conceptualized the study, designed the experiments, discussed the results and edited the manuscript.

Additional files

Supplementary file 1. Addtional tables with primer sequences and selected mutator genes.

DOI: http://dx.doi.org/10.7554/eLife.22939.024

elife-22939-supp1.docx^{(23.2KB, docx)}

DOI: 10.7554/eLife.22939.024

Major datasets

The following dataset was generated:

Swings T,Michiels J,2016,Highly ethanol tolerant Escherichia coli with hypermutation phenotype,https://www.ncbi.nlm.nih.gov/bioproject/PRJNA344553,Publicly available at the NCBI SRA repository (accession no: PRJNA344553)

References

Baba T, Ara T, Hasegawa M, Takai Y, Okumura Y, Baba M, Datsenko KA, Tomita M, Wanner BL, Mori H. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Molecular Systems Biology. 2006;2:2006.0008. doi: 10.1038/msb4100050. [DOI] [PMC free article] [PubMed] [Google Scholar]
Barrick JE, Lenski RE. Genome dynamics during experimental evolution. Nature Reviews Genetics. 2013;14:827–839. doi: 10.1038/nrg3564. [DOI] [PMC free article] [PubMed] [Google Scholar]
Barrick JE, Yu DS, Yoon SH, Jeong H, Oh TK, Schneider D, Lenski RE, Kim JF. Genome evolution and adaptation in a long-term experiment with Escherichia coli. Nature. 2009;461:1243–1247. doi: 10.1038/nature08480. [DOI] [PubMed] [Google Scholar]
Bell G, Gonzalez A. Adaptation and evolutionary rescue in metapopulations experiencing environmental deterioration. Science. 2011;332:1327–1330. doi: 10.1126/science.1203105. [DOI] [PubMed] [Google Scholar]
Bielas JH, Loeb KR, Rubin BP, True LD, Loeb LA. Human cancers express a mutator phenotype. PNAS. 2006;103:18238–18242. doi: 10.1073/pnas.0607057103. [DOI] [PMC free article] [PubMed] [Google Scholar]
Bjedov I, Tenaillon O, Gérard B, Souza V, Denamur E, Radman M, Taddei F, Matic I. Stress-induced mutagenesis in bacteria. Science. 2003;300:1404–1409. doi: 10.1126/science.1082240. [DOI] [PubMed] [Google Scholar]
Blázquez J. Hypermutation as a factor contributing to the acquisition of antimicrobial resistance. Clinical Infectious Diseases. 2003;37:1201–1209. doi: 10.1086/378810. [DOI] [PubMed] [Google Scholar]
Boe L, Danielsen M, Knudsen S, Petersen JB, Maymann J, Jensen PR. The frequency of mutators in populations of Escherichia coli. Mutation Research/Fundamental and Molecular Mechanisms of Mutagenesis. 2000;448:47–55. doi: 10.1016/S0027-5107(99)00239-0. [DOI] [PubMed] [Google Scholar]
Cherepanov PP, Wackernagel W. Gene disruption in Escherichia coli: tcr and KmR cassettes with the option of Flp-catalyzed excision of the antibiotic-resistance determinant. Gene. 1995;158:9–14. doi: 10.1016/0378-1119(95)00193-A. [DOI] [PubMed] [Google Scholar]
Chopra I, O'Neill AJ, Miller K. The role of mutators in the emergence of antibiotic-resistant bacteria. Drug Resistance Updates. 2003;6:137–145. doi: 10.1016/S1368-7646(03)00041-4. [DOI] [PubMed] [Google Scholar]
Cooper LA, Simmons LA, Mobley HL. Involvement of mismatch repair in the reciprocal control of motility and adherence of uropathogenic Escherichia coli. Infection and Immunity. 2012;80:1969–1979. doi: 10.1128/IAI.00043-12. [DOI] [PMC free article] [PubMed] [Google Scholar]
Couce A, Guelfo JR, Blázquez J. Mutational spectrum drives the rise of mutator bacteria. PLoS Genetics. 2013;9:e1003167. doi: 10.1371/journal.pgen.1003167. [DOI] [PMC free article] [PubMed] [Google Scholar]
Datsenko KA, Wanner BL. One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. PNAS. 2000;97:6640–6645. doi: 10.1073/pnas.120163297. [DOI] [PMC free article] [PubMed] [Google Scholar]
de Visser JA. The fate of microbial mutators. Microbiology. 2002;148:1247–1252. doi: 10.1099/00221287-148-5-1247. [DOI] [PubMed] [Google Scholar]
Denamur E, Bonacorsi S, Giraud A, Duriez P, Hilali F, Amorin C, Bingen E, Andremont A, Picard B, Taddei F, Matic I. High frequency of mutator strains among human uropathogenic Escherichia coli isolates. Journal of Bacteriology. 2002;184:605–609. doi: 10.1128/JB.184.2.605-609.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
Desai MM, Fisher DS. The balance between mutators and nonmutators in asexual populations. Genetics. 2011;188:997–1014. doi: 10.1534/genetics.111.128116. [DOI] [PMC free article] [PubMed] [Google Scholar]
Drake JW. A constant rate of spontaneous mutation in DNA-based microbes. PNAS. 1991;88:7160–7164. doi: 10.1073/pnas.88.16.7160. [DOI] [PMC free article] [PubMed] [Google Scholar]
Drake JW, Charlesworth B, Charlesworth D, Crow JF. Rates of spontaneous mutation. Genetics. 1998;148:1667–1686. doi: 10.1093/genetics/148.4.1667. [DOI] [PMC free article] [PubMed] [Google Scholar]
Elena SF, Lenski RE. Evolution experiments with microorganisms: the dynamics and genetic bases of adaptation. Nature Reviews Genetics. 2003a;4:457–469. doi: 10.1038/nrg1088. [DOI] [PubMed] [Google Scholar]
Elena SF, de Visser JA. Environmental stress and the effects of mutation. Journal of Biology. 2003b;2:12. doi: 10.1186/1475-4924-2-12. [DOI] [PMC free article] [PubMed] [Google Scholar]
Eyre-Walker A, Keightley PD. The distribution of fitness effects of new mutations. Nature Reviews Genetics. 2007;8:610–618. doi: 10.1038/nrg2146. [DOI] [PubMed] [Google Scholar]
Ferroni A, Guillemot D, Moumile K, Bernede C, Le Bourgeois M, Waernessyckle S, Descamps P, Sermet-Gaudelus I, Lenoir G, Berche P, Taddei F. Effect of mutator P. aeruginosa on antibiotic resistance acquisition and respiratory function in cystic fibrosis. Pediatric Pulmonology. 2009;44:820–825. doi: 10.1002/ppul.21076. [DOI] [PubMed] [Google Scholar]
Gentile CF, Yu SC, Serrano SA, Gerrish PJ, Sniegowski PD, Chao L, Cox EC. Competition between high- and higher-mutating strains of Escherichia coli. Evolution. 2011;7:125–134. doi: 10.2307/2408181. [DOI] [PMC free article] [PubMed] [Google Scholar]
Giraud A, Matic I, Tenaillon O, Clara A, Radman M, Fons M, Taddei F. Costs and benefits of high mutation rates: adaptive evolution of bacteria in the mouse gut. Science. 2001;291:2606–2608. doi: 10.1126/science.1056421. [DOI] [PubMed] [Google Scholar]
Gonzalez A, Bell G. Evolutionary rescue and adaptation to abrupt environmental change depends upon the history of stress. Philosophical Transactions of the Royal Society B: Biological Sciences. 2013;368:20120079. doi: 10.1098/rstb.2012.0079. [DOI] [PMC free article] [PubMed] [Google Scholar]
Gonzalez A, Ronce O, Ferriere R, Hochberg ME. Evolutionary rescue: an emerging focus at the intersection between ecology and evolution. Philosophical Transactions of the Royal Society B: Biological Sciences. 2013;368:20120404. doi: 10.1098/rstb.2012.0404. [DOI] [PMC free article] [PubMed] [Google Scholar]
Goodarzi H, Bennett BD, Amini S, Reaves ML, Hottes AK, Rabinowitz JD, Tavazoie S. Regulatory and metabolic rewiring during laboratory evolution of ethanol tolerance in E. coli. Molecular Systems Biology. 2010;6:378. doi: 10.1038/msb.2010.33. [DOI] [PMC free article] [PubMed] [Google Scholar]
Gould CV, Sniegowski PD, Shchepetov M, Metlay JP, Weiser JN. Identifying mutator phenotypes among fluoroquinolone-resistant strains of Streptococcus pneumoniae using fluctuation analysis. Antimicrobial Agents and Chemotherapy. 2007;51:3225–3229. doi: 10.1128/AAC.00336-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
Gross MD, Siegel EC. Incidence of mutator strains in Escherichia coli and coliforms in nature. Mutation Research Letters. 1981;91:107–110. doi: 10.1016/0165-7992(81)90081-6. [DOI] [PubMed] [Google Scholar]
Gupta DK, Patra AT, Zhu L, Gupta AP, Bozdech Z. DNA damage regulation and its role in drug-related phenotypes in the malaria parasites. Scientific Reports. 2016;6:23603. doi: 10.1038/srep23603. [DOI] [PMC free article] [PubMed] [Google Scholar]
Haft RJ, Keating DH, Schwaegler T, Schwalbach MS, Vinokur J, Tremaine M, Peters JM, Kotlajich MV, Pohlmann EL, Ong IM, Grass JA, Kiley PJ, Landick R. Correcting direct effects of ethanol on translation and transcription machinery confers ethanol tolerance in bacteria. PNAS. 2014;111:E2576–E2585. doi: 10.1073/pnas.1401853111. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hall BM, Ma CX, Liang P, Singh KK. Fluctuation analysis CalculatOR: a web tool for the determination of mutation rate using Luria-Delbruck fluctuation analysis. Bioinformatics. 2009;25:1564–1565. doi: 10.1093/bioinformatics/btp253. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hall LM, Henderson-Begg SK. Hypermutable bacteria isolated from humans--a critical analysis. Microbiology. 2006;152:2505–2514. doi: 10.1099/mic.0.29079-0. [DOI] [PubMed] [Google Scholar]
Hammerschmidt K, Rose CJ, Kerr B, Rainey PB. Life cycles, fitness decoupling and the evolution of multicellularity. Nature. 2014;515:75–79. doi: 10.1038/nature13884. [DOI] [PubMed] [Google Scholar]
Hammerstrom TG, Beabout K, Clements TP, Saxer G, Shamoo Y. Acinetobacter baumannii repeatedly evolves a Hypermutator phenotype in response to Tigecycline that effectively surveys evolutionary trajectories to resistance. PLoS One. 2015;10:e0140489. doi: 10.1371/journal.pone.0140489. [DOI] [PMC free article] [PubMed] [Google Scholar]
Healey KR, Zhao Y, Perez WB, Lockhart SR, Sobel JD, Farmakiotis D, Kontoyiannis DP, Sanglard D, Taj-Aldeen SJ, Alexander BD, Jimenez-Ortigosa C, Shor E, Perlin DS. Prevalent mutator genotype identified in fungal pathogen candida glabrata promotes multi-drug resistance. Nature Communications. 2016;7:11128. doi: 10.1038/ncomms11128. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hietpas RT, Bank C, Jensen JD, Bolon DN. Shifting fitness landscapes in response to altered environments. Evolution. 2013;67:3512–3522. doi: 10.1111/evo.12207. [DOI] [PMC free article] [PubMed] [Google Scholar]
Iguchi S, Mizutani T, Hiramatsu K, Kikuchi K. Rapid acquisition of Linezolid resistance in Methicillin-Resistant Staphylococcus aureus: role of hypermutation and homologous recombination. PLoS One. 2016;11:e0155512. doi: 10.1371/journal.pone.0155512. [DOI] [PMC free article] [PubMed] [Google Scholar]
Jolivet-Gougeon A, Kovacs B, Le Gall-David S, Le Bars H, Bousarghin L, Bonnaure-Mallet M, Lobel B, Guillé F, Soussy CJ, Tenke P. Bacterial hypermutation: clinical implications. Journal of Medical Microbiology. 2011;60:563–573. doi: 10.1099/jmm.0.024083-0. [DOI] [PubMed] [Google Scholar]
Kaplan S, Bren A, Zaslaver A, Dekel E, Alon U. Diverse two-dimensional input functions control bacterial sugar genes. Molecular Cell. 2008;29:786–792. doi: 10.1016/j.molcel.2008.01.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
Krašovec R, Belavkin RV, Aston JA, Channon A, Aston E, Rash BM, Kadirvel M, Forbes S, Knight CG. Mutation rate plasticity in rifampicin resistance depends on Escherichia coli cell-cell interactions. Nature Communications. 2014;5:3742. doi: 10.1038/ncomms4742. [DOI] [PMC free article] [PubMed] [Google Scholar]
LeClerc JE, Li B, Payne WL, Cebula TA. High mutation frequencies among Escherichia coli and Salmonella pathogens. Science. 1996;274:1208–1211. doi: 10.1126/science.274.5290.1208. [DOI] [PubMed] [Google Scholar]
Lee AH, Fidock DA. Evidence of a mild mutator phenotype in cambodian Plasmodium falciparum malaria parasites. PLoS One. 2016;11:1–14. doi: 10.1371/journal.pone.0154166. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lee H, Popodi E, Tang H, Foster PL. Rate and molecular spectrum of spontaneous mutations in the bacterium Escherichia coli as determined by whole-genome sequencing. PNAS. 2012;109:E2774–E2783. doi: 10.1073/pnas.1210309109. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lenski RE, Rose MR, Simpson SC, Tadler SC. Long-Term experimental evolution in Escherichia coli. I. adaptation and divergence during 2,000 generations. The American Naturalist. 1991;138:1315–1341. doi: 10.1086/285289. [DOI] [Google Scholar]
Lindsey HA, Gallie J, Taylor S, Kerr B. Evolutionary rescue from extinction is contingent on a lower rate of environmental change. Nature. 2013;494:463–467. doi: 10.1038/nature11879. [DOI] [PubMed] [Google Scholar]
Loh E, Salk JJ, Loeb LA. Optimization of DNA polymerase mutation rates during bacterial evolution. PNAS. 2010;107:1154–1159. doi: 10.1073/pnas.0912451107. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lynch M, Ackerman MS, Gout JF, Long H, Sung W, Thomas WK, Foster PL. Genetic drift, selection and the evolution of the mutation rate. Nature Reviews Genetics. 2016;17:704–714. doi: 10.1038/nrg.2016.104. [DOI] [PubMed] [Google Scholar]
Lynch M. Evolution of the mutation rate. Trends in Genetics. 2010;26:345–352. doi: 10.1016/j.tig.2010.05.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lynch M. The lower bound to the evolution of mutation rates. Genome Biology and Evolution. 2011;3:1107–1118. doi: 10.1093/gbe/evr066. [DOI] [PMC free article] [PubMed] [Google Scholar]
Mao EF, Lane L, Lee J, Miller JH. Proliferation of mutators in A cell population. Journal of Bacteriology. 1997;179:417–422. doi: 10.1128/jb.179.2.417-422.1997. [DOI] [PMC free article] [PubMed] [Google Scholar]
Marinus MG. DNA methylation and mutator genes in Escherichia coli K-12. Mutation Research/Reviews in Mutation Research. 2010;705:71–76. doi: 10.1016/j.mrrev.2010.05.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
Marvig RL, Johansen HK, Molin S, Jelsbak L. Genome analysis of a transmissible lineage of Pseudomonas aeruginosa reveals pathoadaptive mutations and distinct evolutionary paths of hypermutators. PLoS Genetics. 2013;9:e1003741. doi: 10.1371/journal.pgen.1003741. [DOI] [PMC free article] [PubMed] [Google Scholar]
Marvig RL, Sommer LM, Molin S, Johansen HK. Convergent evolution and adaptation of Pseudomonas aeruginosa within patients with cystic fibrosis. Nature Genetics. 2015;47:57–64. doi: 10.1038/ng.3148. [DOI] [PubMed] [Google Scholar]
Matic I, Radman M, Taddei F, Picard B, Doit C, Bingen E, Denamur E, Elion J. Highly variable mutation rates in commensal and pathogenic Escherichia coli. Science. 1997;277:1833–1834. doi: 10.1126/science.277.5333.1833. [DOI] [PubMed] [Google Scholar]
Negri MC, Morosini MI, Baquero MR, del Campo R, Blázquez J, Baquero F. Very low cefotaxime concentrations select for hypermutable Streptococcus pneumoniae populations. Antimicrobial Agents and Chemotherapy. 2002;46:528–530. doi: 10.1128/AAC.46.2.528-530.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
Nicolaou SA, Gaida SM, Papoutsakis ET. Exploring the combinatorial genomic space in Escherichia coli for ethanol tolerance. Biotechnology Journal. 2012;7:1337–1345. doi: 10.1002/biot.201200227. [DOI] [PubMed] [Google Scholar]
Nowosielska A, Marinus MG. DNA mismatch repair-induced double-strand breaks. DNA Repair. 2008;7:48–56. doi: 10.1016/j.dnarep.2007.07.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
Oliver A. Clinical relevance of Pseudomonas aeruginosa hypermutation in cystic fibrosis chronic respiratory infection. Journal of Cystic Fibrosis. 2015;14:e1–e2. doi: 10.1016/j.jcf.2014.12.009. [DOI] [PubMed] [Google Scholar]
Otto SP, Day T. A Biologist’s Guide to Mathematical Modeling in Ecology and Evolution. Princeton University Press; 2007. [Google Scholar]
Pal C, Maciá MD, Oliver A, Schachar I, Buckling A. Coevolution with viruses drives the evolution of bacterial mutation rates. Nature. 2007;450:1079–1081. doi: 10.1038/nature06350. [DOI] [PubMed] [Google Scholar]
Perfeito L, Fernandes L, Mota C, Gordo I. Adaptive mutations in bacteria: high rate and small effects. Science. 2007;317:813–815. doi: 10.1126/science.1142284. [DOI] [PubMed] [Google Scholar]
Pulido-Tamayo S, Sánchez-Rodríguez A, Swings T, Van den Bergh B, Dubey A, Steenackers H, Michiels J, Fostier J, Marchal K. Frequency-based haplotype reconstruction from deep sequencing data of bacterial populations. Nucleic Acids Research. 2015;43:e105. doi: 10.1093/nar/gkv478. [DOI] [PMC free article] [PubMed] [Google Scholar]
Rajanna C, Ouellette G, Rashid M, Zemla A, Karavis M, Zhou C, Revazishvili T, Redmond B, McNew L, Bakanidze L, Imnadze P, Rivers B, Skowronski EW, O'Connell KP, Sulakvelidze A, Gibbons HS. A strain of Yersinia pestis with a mutator phenotype from the republic of Georgia. FEMS Microbiology Letters. 2013;343:113–120. doi: 10.1111/1574-6968.12137. [DOI] [PubMed] [Google Scholar]
Ram Y, Hadany L. Stress-induced mutagenesis and complex adaptation. Proceedings of the Royal Society B: Biological Sciences. 2014;281:20141025. doi: 10.1098/rspb.2014.1025. [DOI] [PMC free article] [PubMed] [Google Scholar]
Rando OJ, Verstrepen KJ. Timescales of genetic and epigenetic inheritance. Cell. 2007;128:655–668. doi: 10.1016/j.cell.2007.01.023. [DOI] [PubMed] [Google Scholar]
Raynes Y, Sniegowski PD. Experimental evolution and the dynamics of genomic mutation rate modifiers. Heredity. 2014;113:375–380. doi: 10.1038/hdy.2014.49. [DOI] [PMC free article] [PubMed] [Google Scholar]
Roberts SA, Gordenin DA. Hypermutation in human cancer genomes: footprints and mechanisms. Nature Reviews Cancer. 2014;14:786–800. doi: 10.1038/nrc3816. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ronen M, Rosenberg R, Shraiman BI, Alon U. Assigning numbers to the arrows: parameterizing a gene regulation network by using accurate expression kinetics. PNAS. 2002;99:10555–10560. doi: 10.1073/pnas.152046799. [DOI] [PMC free article] [PubMed] [Google Scholar]
Sandberg TE, Pedersen M, LaCroix RA, Ebrahim A, Bonde M, Herrgard MJ, Palsson BO, Sommer M, Feist AM. Evolution of Escherichia coli to 42 °C and subsequent genetic engineering reveals adaptive mechanisms and novel mutations. Molecular Biology and Evolution. 2014;31:2647–2662. doi: 10.1093/molbev/msu209. [DOI] [PMC free article] [PubMed] [Google Scholar]
Schaaper RM, Dunn RL. Spectra of spontaneous mutations in Escherichia coli strains defective in mismatch correction: the nature of in vivo DNA replication errors. PNAS. 1987;84:6220–6224. doi: 10.1073/pnas.84.17.6220. [DOI] [PMC free article] [PubMed] [Google Scholar]
Schaaper RM, Radman M. The extreme mutator effect of Escherichia coli mutD5 results from saturation of mismatch repair by excessive DNA replication errors. The EMBO Journal. 1989;8:3511–3516. doi: 10.1002/j.1460-2075.1989.tb08516.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
Shaver AC, Dombrowski PG, Sweeney JY, Treis T, Zappala RM, Sniegowski PD. Fitness evolution and the rise of mutator alleles in experimental Escherichia coli populations. Genetics. 2002;162:557–566. doi: 10.1093/genetics/162.2.557. [DOI] [PMC free article] [PubMed] [Google Scholar]
Sniegowski P, Raynes Y. Mutation rates: how low can you go? Current Biology. 2013;23:R147–R149. doi: 10.1016/j.cub.2013.01.018. [DOI] [PubMed] [Google Scholar]
Sniegowski PD, Gerrish PJ, Johnson T, Shaver A. The evolution of mutation rates: separating causes from consequences. BioEssays. 2000;22:1057–1066. doi: 10.1002/1521-1878(200012)22:12<1057::AID-BIES3>3.0.CO;2-W. [DOI] [PubMed] [Google Scholar]
Sniegowski PD, Gerrish PJ, Lenski RE. Evolution of high mutation rates in experimental populations of E. coli. Nature. 1997;387:703–705. doi: 10.1038/42701. [DOI] [PubMed] [Google Scholar]
Stepanyan K, Wenseleers T, Duéñez-Guzmán EA, Muratori F, Van den Bergh B, Verstraeten N, De Meester L, Verstrepen KJ, Fauvart M, Michiels J. Fitness trade-offs explain low levels of persister cells in the opportunistic pathogen Pseudomonas aeruginosa. Molecular Ecology. 2015;24:1572–1583. doi: 10.1111/mec.13127. [DOI] [PubMed] [Google Scholar]
Sung W, Ackerman MS, Dillon MM, Platt TG, Fuqua C, Cooper VS, Lynch M. Evolution of the Insertion-Deletion mutation rate across the tree of Life. Genes|Genomes|Genetics. 2016;6:2583–2591. doi: 10.1534/g3.116.030890. [DOI] [PMC free article] [PubMed] [Google Scholar]
Tanaka MM, Bergstrom CT, Levin BR. The evolution of mutator genes in bacterial populations: the roles of environmental change and timing. Genetics. 2003;164:843–854. doi: 10.1093/genetics/164.3.843. [DOI] [PMC free article] [PubMed] [Google Scholar]
Tenaillon O, Toupance B, Le Nagard H, Taddei F, Godelle B. Mutators, population size, adaptive landscape and the adaptation of asexual populations of bacteria. Genetics. 1999;152:485–493. doi: 10.1093/genetics/152.2.485. [DOI] [PMC free article] [PubMed] [Google Scholar]
Thierauf A, Perez G, Maloy AS. Generalized transduction. Methods in Molecular Biology. 2009;501:267–286. doi: 10.1007/978-1-60327-164-6_23. [DOI] [PubMed] [Google Scholar]
Thompson DA, Desai MM, Murray AW. Ploidy controls the success of mutators and nature of mutations during budding yeast evolution. Current Biology. 2006;16:1581–1590. doi: 10.1016/j.cub.2006.06.070. [DOI] [PubMed] [Google Scholar]
Torres-Barceló C, Cabot G, Oliver A, Buckling A, Maclean RC. A trade-off between oxidative stress resistance and DNA repair plays a role in the evolution of elevated mutation rates in bacteria. Proceedings of the Royal Society B: Biological Sciences. 2013;280:20130007. doi: 10.1098/rspb.2013.0007. [DOI] [PMC free article] [PubMed] [Google Scholar]
Tröbner W, Piechocki R. Selection against hypermutability in Escherichia coli during long term evolution. MGG Molecular & General Genetics. 1984;198:177–178. doi: 10.1007/BF00328720. [DOI] [PubMed] [Google Scholar]
Van den Bergh B, Michiels JE, Wenseleers T, Windels EM, Boer PV, Kestemont D, De Meester L, Verstrepen KJ, Verstraeten N, Fauvart M, Michiels J. Frequency of antibiotic application drives rapid evolutionary adaptation of Escherichia coli persistence. Nature Microbiology. 2016;1:16020. doi: 10.1038/nmicrobiol.2016.20. [DOI] [PubMed] [Google Scholar]
Voordeckers K, Kominek J, Das A, Espinosa-Cantú A, De Maeyer D, Arslan A, Van Pee M, van der Zande E, Meert W, Yang Y, Zhu B, Marchal K, DeLuna A, Van Noort V, Jelier R, Verstrepen KJ. Adaptation to high ethanol reveals complex evolutionary pathways. PLoS Genetics. 2015;11:e1005635. doi: 10.1371/journal.pgen.1005635. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wang J, Cazzato E, Ladewig E, Frattini V, Rosenbloom DI, Zairis S, Abate F, Liu Z, Elliott O, Shin YJ, Lee JK, Lee IH, Park WY, Eoli M, Blumberg AJ, Lasorella A, Nam DH, Finocchiaro G, Iavarone A, Rabadan R. Clonal evolution of glioblastoma under therapy. Nature Genetics. 2016;48:768–776. doi: 10.1038/ng.3590. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wielgoss S, Barrick JE, Tenaillon O, Wiser MJ, Dittmar WJ, Cruveiller S, Chane-Woon-Ming B, Médigue C, Lenski RE, Schneider D. Mutation rate dynamics in a bacterial population reflect tension between adaptation and genetic load. PNAS. 2013;110:222–227. doi: 10.1073/pnas.1219574110. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wiser MJ, Ribeck N, Lenski RE. Long-term dynamics of adaptation in asexual populations. Science. 2013;342:1364–1367. doi: 10.1126/science.1243357. [DOI] [PubMed] [Google Scholar]
Woods RJ, Barrick JE, Cooper TF, Shrestha U, Kauth MR, Lenski RE. Second-order selection for evolvability in a large Escherichia coli population. Science. 2011;331:1433–1436. doi: 10.1126/science.1198914. [DOI] [PMC free article] [PubMed] [Google Scholar]
Yu J, Xiao J, Ren X, Lao K, Xie XS. Probing gene expression in live cells, one protein molecule at a time. Science. 2006;311:1600–1603. doi: 10.1126/science.1119623. [DOI] [PubMed] [Google Scholar]
Zaslaver A, Bren A, Ronen M, Itzkovitz S, Kikoin I, Shavit S, Liebermeister W, Surette MG, Alon U. A comprehensive library of fluorescent transcriptional reporters for Escherichia coli. Nature Methods. 2006;3:623–628. doi: 10.1038/nmeth895. [DOI] [PubMed] [Google Scholar]
Zaslaver A, Mayo AE, Rosenberg R, Bashkin P, Sberro H, Tsalyuk M, Surette MG, Alon U. Just-in-time transcription program in metabolic pathways. Nature Genetics. 2004;36:486–491. doi: 10.1038/ng1348. [DOI] [PubMed] [Google Scholar]
Zwietering MH, Jongenburger I, Rombouts FM, van 't Riet K. Modeling of the bacterial growth curve. Applied and Environmental Microbiology. 1990;56:1875–1881. doi: 10.1128/aem.56.6.1875-1881.1990. [DOI] [PMC free article] [PubMed] [Google Scholar]

eLife. 2017 May 2;6:e22939. doi: 10.7554/eLife.22939.030

Decision letter

Editor: Wenying Shou¹

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Adaptive tuning of mutation rates allows fast response to lethal stress in Escherichia coli" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom Wenying Shou (Reviewer #1), is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Diethard Tautz as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Olivier Tenaillon (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This article shows that mutation rate can be highly dynamic during evolution. The work is interesting. However, there are major issues that will need to be clarified and addressed. Details are provided in the reviewer comments below. Please note that depending on your response, the reviewers might come to the conclusion that additional experiments could be necessary.

Reviewer #1:

This review is from the point of view of a "general audience" member who is informally interested in mutators. This paper showed that when E. coli was subjected to lethal ethanol stress, mutators evolved in multiple lines. Those mutator lines (but not lines with normal mutation rates) could tolerate lethal levels of ethanol. Mutation rates in these mutator lines were then reduced in later stages of evolution.

Authors claim that "an organism's mutation rate is often considered to be a slowly-evolving parameter." I am not sure whether this claim is true since Wielgoss et al., 2013 showed that a hyper-mutator evolved in one of the 12 Lenski lines and that it was later replaced by milder mutators. The difference is that in this study, reduction of mutation rate is through compensatory mutations instead of replacement of one type of mutator with another type. Hammerstrom et al., 2015 (not cited) showed that hypermutators can repeatedly evolve in response to tigecycline (via transposon insertion). Thus, repeated evolution of hypermutator (in contrast to the stochastic appearance of mutator in one of Lenski lines) is not new either. However, there are nice experiments in this paper, and the dynamics of mutation rate is also tracked with fine resolution.

Main points:

Subsection “Long-term adaption to high ethanol stree in E. coli is contingent upon hypermutation”: "evolutionary dead ends" – how do you know that? Have you introduced mutator mutations into dead-end lines and observed lack of improved ethanol survival? If so, it will be interesting to know why the "dead-end" lines failed to evolve mutator phenotypes.

Quantification method of ethanol tolerance needs to be described. Figure 5B: Units? I am in fact not sure about the point of Figure 5B given that high ethanol tolerance can be achieved with low mutation rate at later stages of evolution.

I do not find increased death rate associated with mutators particularly interesting, without authors explaining to me the fundamental difference between slower birth, increased death, or a combination of the two. After all, one can view birth and death as "net-growth".

I do not understand the point of Table 1. Why not recalculate estimated number of mutations based on the dynamics of mutation rate and see whether it matches observed number of mutations?

It will be interesting to know which mutations directly promote ethanol tolerance (if you already have the data).

Reviewer #2:

In their manuscript entitled "Adaptive tuning of mutation rate allows fast response to lethal stress in E. coli", Toon Swings and collaborators study the evolution of mutation rate under adaptation to ethanol. They show that increased mutation rate is rapidly selected for in the early stages of the adaptation, and more surprisingly they uncover a rapid decay of mutation rate that seems to be directly linked to some increased mortality.

The data are solid, quantitative and very interesting, especially the connection between mortality and mutation rates. My comments are mostly on some of the interpretations of the results, and on some experimental details

In the Introduction, it is not fully clear if ethanol 5% is a real lethal stress: mutations are required to survive the first growth cycle. Based on the lag time is seems so. If it is that case the paper would benefit from referring to the filed of rescue experiments. These are fashionable experiments that study adaptation in conditions in which it is essential for survival. A paragraph (or even the title) on the link between the experiment and that field would increase the readership of the paper.

The paper is clearly showing a new factor that may explain changes in mutation rate: mortality. However, the whole interpretation of the phenomenon is still linked to the idea of second order selection and load. This is really in line with the first theoretical data on mutators and second order selection that neglected all potential direct effects. More recent work have put more emphasis on potential direct effects and I think the discussion should introduce some of these ideas and eventually challenge them. Some results here suggest that eventually some direct effect of mutators may be at play, not in the early selection of mutators but eventually in their costs.

First in Figure 4 even at extremely low frequencies, mutator alleles are selected for. This is quite surprising as at some point mutator subpopulation should be to low to generate any beneficial mutations. Indeed if we assume a 1000-fold effect on transitions, mutator should be outcompeted by wild type at a ratio of 1/10000 or so. So Figure 4 suggest a direct effect. Also mentioned in the text the fact that only point mutations were found in mutS and no inactivation is also an argument for a more complex effect than just increase in mutation rate. (inactivation should be much more likely to occur (though some deletions may not be detected at population level with the pipeline used) and as they have an impressive benefit in competition they should invade immediately…

Even in the very nice experiments linking mortality and mutation rate, a 10 fold reduction in mutation rate from IM2 to IM3 lead to a two fold decrease in mortality or less in dead cells. So mutation rate is not the only player. Mutation could have accumulated that lead to the mild differences between IM2 and IM3, But in Figure 7, the author found that a three fold change in mutation rate has this time a drastic effect on survival (more than 10 times)And the growth curves are completely different.

It may be extremely difficult to go after mechanistic effects, but discussing that both primary selection and second order selection may be at play will benefit the paper.

MMR have been shown to be beneficial because they allow a fast switching of flagella expression though increased recombination, have been found to be toxic when too many mismatches are present in the cell such that they lead to double strand break… and few other recent papers have suggested that evolution of mutation rate may also be linked to direct effect. This does not decrease in any way the value of the experiments done, but is worth mentioning in the discussion.

Of note, with respect to potential rise of mutators, the experiments with a subset of mutators involved in different mechanisms is clearly a proof that increased mutation rate is selected for. Then the fact that only MMR clones are found and no KO is found suggests that there are eventually selection of more moderate effects or of some specific mecanisms.

Overall it is a very nice study that clearly shows a very dynamic and fast evolution of mutation rate that is extremely relevant to understand evolution in stressful conditions.

Reviewer #3:

The authors present an interesting set of experiments showing surprisingly high dynamics of changes in the mutation rate of E. coli populations adapting to near-lethal ethanol concentrations. Several previous microbial evolution experiments have demonstrated the rapid rise of mutator mutants via genetic linkage with beneficial mutations during adaptation to novel conditions and stresses. Some studies also found declines of the mutation rate later during adaptation, when the ratio of beneficial to deleterious mutations declines, due to the selection of anti-mutator genotypes. However, the selection of genotypes with lower mutation rates is thought to be driven by much smaller selection coefficients than the selection of genotypes with increased mutation rates. Mutators hitchhike with beneficial mutations, which can have large fitness benefits under stressful conditions, while the fitness benefit of anti-mutators is proportional to the difference in mutation load between mutator and anti-mutator – with the difference in the fraction of lethal mutations as an upper limit. Given a wild-type mutation in of ~0.003 per genome per generation (e.g. Drake et al., 1998), mutS mutators have a mutation rate of ~0.1 (10-100 fold higher). This sets the upper limit to the selective benefit of a change back from mutator to wild-type mutation rates to 10% if all mutations were lethal, suggesting that rapid declines of mutation rates are theoretically possible under conditions where lethal mutations are frequent. While it has been very difficult to detect fitness benefits of anti-mutators under less stressful conditions (e.g. Wielgoss et al., 2013), the authors of the present study claim to have observed much higher benefits under near-lethal conditions, allowing not only rapid increases, but also subsequent rapid decreases of the mutation rate. If true (but see below), this finding is sufficiently interesting and novel to publish in a general high-impact journal, such as eLife.

We have three main comments:

Essential information is missing at several places, which prevents both a basic understanding of the population dynamics during the experiment and a comparison of the present findings with those of previous work on the evolution of mutation rates. First, no information on the effective population size of the evolving populations is given: what culture volumes, which dilutions, what final cell densities were obtained and what was the effective population size (~inoculum size x number of generations per growth cycle)? Second, given the high mortality of cells, how were numbers of generations calculated? Dead cells have also been produced and they should be taken into account for the estimates of the total numbers of generations. Third and perhaps most importantly, it was unclear whether fluctuation tests and genome sequence analyses have been done on individual clones or mixed population samples. As a result, we don't know how to interpret Figure 5 and Table 1: do mutation rates in Figure 5 reflect mutation rates of fixed genotypes or are these a mix of mutator and wild-type mutation rates of a polymorphic population? Similar for the frequency of observed mutations in Table 1: are these based on sequences of random clones picked form these populations or on mixed population DNA? The variation in mutation rate shown in Figure 5A suggests that these reflect population mutation rates, since changes are too small to reflect pure genotype effects if these involve mutator mutants such as mutS (see Figure 4). This distinction is crucial for the possible mechanisms that should be considered for explaining the dynamics in mutation rate: if high mutation-rate genotypes fix in the population, later observed low mutation-rate genotypes must derive from the former genotypes, whereas if they are *not* fixed, later observed low mutation-rate genotypes may derive from ancestral genotypes that were still present in the population when the increase in mutation rate was measured.

Our second comment is about the surprising rapid selection of genotypes with lower mutation rates at several times during evolution (Figure 5). As sketched above, to understand whether natural selection has driven these changes (and assuming that they reflect the mutation rate of the population, not of individual clones!), two pieces of information are required. First, the fitness consequences of these changes should be quantified in direct competition experiments between genotypes with high and low mutation rate under the conditions prevalent during evolution. Now only the production of dead cells is compared (Figure 6), but no attempt is made to translate mortality estimates into fitness values, either empirically (by running these competitions) or theoretically (by using growth models to predict these effects). The decline of mutation rate during 150 generations of continues evolution for population E9 (Figure 7), and its correlation with lower production of dead cells, are suggestive, but do not solve the riddle how mortality rates translate into fitness under the selective conditions. Second, fitness estimates of high and low mutation-rate genotypes should then be used to predict the dynamic of decline of the mutation rate, to see if this mechanism may indeed explain the dynamics shown in Figure 5.

Our third comment is on Figure 1. First, why would small inocula lead to much higher yields than large inocula? This is surprising, but remains unexplained. Second, the identity of the growth curves should be better explained. What do the grey lines reflect, individual replicates or mean curves for each mutator strain? The number of lines suggest the former, but the text refers to "the" growth curve for dnaQ, suggesting they reflect mean curves. It may be best to show mean growth curves or fitted growth models per strain, with different colors per strain. How are the different populations (of different size) generated? Do they come from the same 'batch' which are either less or more diluted? Or does the 10⁷ dilution come from a later time point of the smaller dilutions? The physiological state of the cells (e.g. early exponential phase, later exponential phase) at the moment that they are inoculated is known to affect the subsequently generated growth curves. Or perhaps, can larger population sizes 'soak' the ethanol, and thus effectively decrease the concentration in the environment? (see e.g. doi: 10.1098/rsbl.2012.0569)

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Adaptive tuning of mutation rates allows fast response to lethal stress in Escherichia coli" for further consideration at eLife. Your revised article has been favorably evaluated by Diethard Tautz (Senior editor). The Reviewing editor Wenying Shou and one of the original reviewers (Olivier Tenaillon) are satisfied with your responses to their critiques. The original third reviewer was not able to re-review. Instead, two scientists familiar with mutation rates co-reviewed your work and your response.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

Reviewers #4 and 5:

This is a study with complicated design and large amount of work. The authors have addressed (or at least tried) most concerns from the last round of revisions. However, as new reviewers, we have some additional points that need the authors' explanations. The paper could be accepted given satisfying responses.

1) "A higher mutation rate is linked to a higher mortality, probably due to the extended buildup of genetic load and increased probability of acquiring a lethal mutation". There could be a more straight-forward explanation: Krasovec et al., (2014 Nature Communications) showed that mutation rate from fluctuation assay inversely correlates with population density, due to cell-cell interactions. For the same experimental evolution line at different intermediate time points, higher mortality could reduce population density-given the same inoculum size for fluctuation assay, which in turn increases mutation rate. This alternative interpretation needs to be addressed in the current experimental system.

2) While the authors interpret the data as evidence to support the idea that cells evolve hypermutation to avoid extinction under near-lethal stress, we should also consider a more straight-forward alternative.

In the population, mutants with various mutation rates are generated continuously. When the environment turns stressful, the distribution of fitness effects of mutations accordingly becomes wider and less deleterious-mutation-biased due to the decreased population-average fitness (e.g. Hietpas et al., 2013). As a result, the mutation load of hypermutators relative to WT becomes smaller and hypermutators are more likely to survive or even prosper (if hitchhiked with beneficial mutations). The argument can be reversed when organisms become more adapted in the new environment to explain the decline of hypermutators. In this way, the dynamic pattern of mutation rates can be understood as a passive balance between influx (by random mutations) and outflux (due to accumulated genetic load) of mutators without evoking any active response. It would help if the authors can clarify if and why their data fit better with their claim than the above alternative interpretation.

Note that the above argument is assuming the fitness effects are due to secondary mutations, thus the authors' discovery of potential direct effects of mutator-generating mutations (e.g.mutS in Figure 1—figure supplement 4) is not affected. However, in the current wording the potential direct benefit of these mutations does not seem to be the main message they try to make impact. It is also not clear why the authors claim "…these direct effects…may not directly influence early selection of hypermutation, but may become decisive in its lower cost at later stages."

eLife. 2017 May 2;6:e22939. doi: 10.7554/eLife.22939.031

Author response

Reviewer #1:

This review is from the point of view of a "general audience" member who is informally interested in mutators. This paper showed that when E. coli was subjected to lethal ethanol stress, mutators evolved in multiple lines. Those mutator lines (but not lines with normal mutation rates) could tolerate lethal levels of ethanol. Mutation rates in these mutator lines were then reduced in later stages of evolution.

Authors claim that "an organism's mutation rate is often considered to be a slowly-evolving parameter." I am not sure whether this claim is true since Wielgoss et al., 2013 showed that a hyper-mutator evolved in one of the 12 Lenski lines and that it was later replaced by milder mutators. The difference is that in this study, reduction of mutation rate is through compensatory mutations instead of replacement of one type of mutator with another type. Hammerstrom et al. (2015) (not cited) showed that hypermutators can repeatedly evolve in response to tigecycline (via transposon insertion). Thus, repeated evolution of hypermutator (in contrast to the stochastic appearance of mutator in one of Lenski lines) is not new either. However, there are nice experiments in this paper, and the dynamics of mutation rate is also tracked with fine resolution.

We thank the reviewer for the appreciation of our work and the detailed revision and suggestions that were made to further improve the manuscript.

Wielgoss et al., 2013 indeed reported the occurrence of a hypermutator and a later decrease in mutation rate in one of the Lenski lines. However, in the light of their work, we still believe our statement of mutation rate being a slowly-evolving parameter is true as this increase and subsequent decrease happened over a long period of time ( ± 40K generations or 20 years) and was observed in only one replicate population. In addition, as also pointed out by reviewer 3, much lower selection coefficients are thought to drive selection on genotypes with lower mutation rates and were observed in studies without stress or under mild stress conditions (Wielgoss et al., 2013; Sniegowski, Gerrish and Lenski, 1997; Barrick et al., 2009; Sandberg et al., 2014). Altogether, we believe that it is valid to state that an organism’s mutation rate is often considered to be a slowly evolving parameter.

Thank you for pointing out the Hammerstrom et al. study that shows repeated evolution of A. baumannii hypermutators in response to tigecycline treatment (Hammerstrom et al., 2015). It demonstrates that our observation of hypermutation in all high ethanol tolerant lines is not limited to near-lethal ethanol stress, but underlies a broader phenomenon where cells exposed to extreme stress (including ethanol or antibiotics) favor higher mutation rates to avoid extinction. As this excellent work substantiates the clinical relevance of our study, we have cited reference 36 and discussed it both in the Introduction and Discussion sections of the revised manuscript.

However, we believe that our results go far beyond the study of Hammerstrom et al. We not only describe rapid, repeated evolution of hypermutation in each high ethanol tolerant line, but also the evolution of multiple, consecutive increases and decreases of the mutation rates within the same evolutionary line. To our knowledge, such a highly dynamic change of mutation rate over time has never been observed before. To stress the difference between the repeated evolution of hypermutation in all high ethanol tolerant lines and the recurrent increases and decreases observed within a line, we changed the Abstract to emphasize the major findings of our study.

“[…]In contrast, we demonstrate an unexpected flexibility in cellular mutation rates as a response to changes in selective pressure. We show that hypermutation independently evolves when different Escherichia coli cultures adapt to high ethanol stress. Furthermore, hypermutator states are transitory and repeatedly alternate with decreases in mutation rate. Specifically, mutation rates rise when cells experience higher stress and decline again once cells are adapted.[…]”

Main points:

Subsection “Long-term adaption to high ethanol stree in E. coli is contingent upon hypermutation”: "evolutionary dead ends" – how do you know that? Have you introduced mutator mutations into dead-end lines and observed lack of improved ethanol survival? If so, it will be interesting to know why the "dead-end" lines failed to evolve mutator phenotypes.

We did not introduce a mutator mutation back into these dead-end lines. Therefore, we have removed and rephrased the term “evolutionary dead-ends” to describe these low ethanol tolerant lines in the manuscript.

“[…]Although ethanol tolerance increased in all populations, only eight out of 20 lines developed tolerance to very high (7% or more) ethanol concentrations (Figure 2B), while the other 12 lines recurrently died out and only developed tolerance to relatively low ethanol concentrations (6% or lower).[…]”

“[…]If the strain needed more than 14 days to grow, we assumed this population died out.[…]”

“[…]If the strain did not show growth in a 14-days timespan, we assumed that the line died out and we revived the previous stored intermediate point to restart the evolution.[…]”

While we have no direct evidence, introducing a mutator mutation might possibly allow these lines to further adapt to higher stress levels. To have a better understanding of why these lines did not evolve a mutator phenotype, we sequenced 2 of the low tolerant lines (E4 & E17). Both of these lines accumulated a mutation in fabA, a gene involved in fatty acid biosynthesis that has already been linked to higher ethanol tolerance (Dombek and Ingram, 1984; Luo, et al., 2009). However, line E4 additionally contained a mutation in rpoZ and line E17 accumulated an additional deletion of infB. rpoZencodes a subunit of RNA polymerase and plays a role in transcription, while infB encodes a protein chain initiation factor and plays a role in translation. It was recently shown that ethanol both affects transcription and translation in the cell (Haft, et al., 2014). While speculative, the effect of these specific mutations might severely impair growth of these low-tolerant lines on higher percentages of ethanol preventing the occurrence of mutator mutations or others and thus further evolutionary improvement. We have included this in the Discussion section:

“[…]Even though we mainly focused on increased mutation rates in this study, the 12 slowly-mutating, low ethanol tolerant lines might be an interesting starting point for further research. The lack of hypermutation in these lines seems to impede further adaptation to high ethanol concentrations. Sequence analysis of two of these lines revealed the presence of mutations in rpoZ (subunit of the RNA polymerase) and infB (protein chain initiation factor) in lines E4 and E17, respectively. Since ethanol is toxic through its effect on transcription and translation (Haft, et al., 2014), disruption of the transcription or translation machinery due to these mutations might cause increased sensitivity to higher ethanol levels. This would prevent further growth and the possibility of acquiring a mutator allele or any other mutation. Although we have no evidence supporting that the rpoZ and infB mutations are causal for the decreased mutation rate, these mutations are interesting and might explain the lack in further adaptive improvement in lines E4 and E17 (Figure 3A).[…]”

Ethanol tolerance of a population was measured as the ability to grow in liquid medium in the presence of a certain percentage of ethanol. When the population grew and reached an optical density of 0.2 or more in the presence of a certain percentage of ethanol it was considered tolerant to that percentage. To ensure the proper percentages of ethanol were added to the medium, we used the Alcolyser beer analyzing system (Anton Paar GmbH, Austria). We now clarified the quantification of ethanol tolerance and measurement of ethanol concentration in the Materials and methods section.

“[…]To ensure the proper percentages of ethanol added to the medium we used the Alcolyser beer analyzing system (Anton Paar GmbH, Austria).[…]”

“[…]Ethanol tolerance of a population was measured as the ability to grow in liquid medium in the presence of a certain percentage of ethanol to an optical density of at least 0.2.[…]”

We have also added units to Figure 5B.

Data shown in Figure 5B prove that the dynamics in mutation rate underlie the evolution of ethanol tolerance. There is no correlation between the absolute mutation rate in each point and the absolute ethanol tolerance in each point. Instead, this figure shows that increases in mutation rate between two consecutive points are significantly correlated with increases in ethanol tolerance between those two time points, indicating that a mutator mutation (that causes the increase in mutation rate) hitchhikes with a beneficial mutation (that causes increased ethanol tolerance) and that these beneficial mutations have higher probability to occur in a mutator background. Moreover, decreases in mutation rate (negative Δ mutation rate) are correlated with periods of constant ethanol pressure (Δ ethanol tolerance of zero), showing that mutation rate decreases again once a population is able to grow in the presence of a certain ethanol percentage. This correlation information is lacking from Figure 5A, showing the relevance of the extra panel.

We agree that the growth of a population depends on both the birth and the death of the cells in the population. Therefore, we added some lines to the Results section to clarify that our observations are linked to increased death rather than to slower birth.

“[…]Surprisingly, all tested intermediate points showed a very fast decrease in viable cell count when entering the stationary phase (Figure 6—figure supplement 1). This decrease in viable cells is explained by a genuine increased death rate in the population since cells are in stationary phase.[…]”

Also, by plating growing populations at different time points, we were able to observe a decrease in viable cell (i.e. negative net-growth) starting from early stationary phase. Because birth of cells is already absent or low in stationary phase, the decrease in viable cells is explained by increased death of the cells. Moreover, we were able to confirm an increased death by life- dead staining (Figure 6; Figure 6—figure supplement 1; Figure 6—figure supplement 2).

I do not understand the point of Table 1. Why not recalculate estimated number of mutations based on the dynamics of mutation rate and see whether it matches observed number of mutations?

Table 1 was meant to show the discrepancy between the expected number of mutations based on the endpoint mutation rate and the actual number of fixed mutations effectively observed in the genomic sequence of the endpoints. It is an interesting suggestion to recalculate the number of mutations based on the dynamics in mutation rate. We tried to calculate the area under the curve (AUC) in Figure 5A to generate an estimated number of mutations over a period of time, taking into account the fluctuations in mutation rate at each time point. However, this calculation is oversimplified. The genomic mutation rate, measured by fluctuation assays, is not suitable for the calculation of estimates that will be compared to the true number of fixed mutations. The genomic mutation rate only reflects the number of mutations occurring in the population, but it does not give an estimate of the number of mutations that will eventually fix in the population. Generating a proper approximation of the expected number of mutations would require estimates of the beneficial mutation rate and the fixation probability of a mutation in a given environment. It would be extremely difficult to infer this in our specific setup as we for example, did not use a constant environment but instead continuously increased the percentage of ethanol. Also, calculations of the necessary parameters to generate a valid estimate would vary for each time point, making it extremely complex. Therefore, we decided to remove the column with the estimated number of mutations and replace the table by an additional supplemental figure, showing both the total number of variants in the population and the number of fixed mutations (Figure 2—figure supplement 2). This additional figure nicely shows that the mutational profile of mutator populations is complex with several low frequency mutations that belong to different subpopulations, possibly having higher or lower mutation rates than the average population mutation rate.

It will be interesting to know which mutations directly promote ethanol tolerance (if you already have the data).

The identification of mutations that directly promote ethanol tolerance is extremely challenging in our dataset, given the likely high number of passenger mutations among few driver mutations. However, we are currently developing a method to analyze such highly complex mutational datasets, often resulting from hypermutator evolution. By looking at parallelism between evolved lines we were already able to identify the fatty-acid biosynthesis pathway, encoded by the fab genes, as a main target in the initial adaptation to high ethanol stress. The identification of this pathway serves as a true positive for our analysis since changing the fatty acid composition of the membrane was previously reported as a prime mechanism to promote ethanol tolerance (Dombek and Ingram, 1984; Luo, et al., 2009).

Reviewer #2:

In their manuscript entitled "Adaptive tuning of mutation rate allows fast response to lethal stress in E. coli", Toon Swings and collaborators study the evolution of mutation rate under adaptation to ethanol. They show that increased mutation rate is rapidly selected for in the early stages of the adaptation, and more surprisingly they uncover a rapid decay of mutation rate that seems to be directly linked to some increased mortality. The data are solid, quantitative and very interesting, especially the connection between mortality and mutation rates. My comments are mostly on some if the interpretations of the results, and on some experimental details

In the Introduction, it is not fully clear if Ethanol 5% is a real lethal stress: mutations are required to survive the first growth cycle. Based on the lag time is seems so. If it is that case the paper would benefit from referring to the filed of rescue experiments. These are fashionable experiments that study adaptation in conditions in which it is essential for survival. A paragraph (or even the title) on the link between the experiment and that field would increase the readership of the paper.

We agree with the reviewer and have included the concept of evolutionary rescue and the link with hypermutation in the Introduction and Discussion sections.

“[…]This is especially obvious in harsh environments, where near-lethal stress requires swift adaptation of at least some individuals to avoid complete extinction of the population (Bell and Gonzalez, 2011). Adaptation sufficiently rapid to save a population from extinction leads to so-called evolutionary rescue. This phenomenon is widely studied in the light of climate change and adaptation of declining populations to new, changing environments (Lindsey, et al., 2013). It occurs when a population under stress lacks sufficient phenotypic plasticity and can only avoid extinction through genetic change (Gonzalez, et al., 2013). Evolutionary rescue depends on different factors such as the population size, genome size, mutation rate, degree of environmental change and history of the stress (Gonzalez, et al., 2013; Gonzalez and Bell, 2012). By increasing the supply of mutations, hypermutation might also be crucial to enable evolutionary rescue for populations under lethal stress.[…]”

“[…]Despite the burden of additional, possibly lethal mutations, the increased mutational supply enables fast adaptation of at least some individuals and rescues the population from extinction (Bell and Gonzalez, 2011; Lindsey et al., 2013; Gonzalez et al., 2013; Gonzalez and Bell, 2013).[…]”

Evolutionary rescue is indeed an interesting phenomenon that is highly relevant for adaptation to lethal stress. 5% ethanol almost completely inhibits growth of the wild type and drastically reduces the carrying capacity (Figure 1—figure supplement 2). The different lag times observed in both the wild type and the mutator mutants indeed suggest that mutations are necessary to enable growth in the presence of 5% ethanol, which would consequently imply that it is a real lethal stress. However, to certify this claim we would need to sequence the strain after growth on 5% ethanol to see if we can actually identify the adaptive mutations that enabled growth and rescued the population from extinction.

Of note, previous studies pointed out that increasing ethanol tolerance is complex, involves interaction between multiple genes and pathways and requires multiple mutations (Goodarzi, et al., 2010; Nicolaou, et al., 2012; Voordeckers, et al., 2015). The need for multiple mutations to achieve higher ethanol tolerance combined with the severely impaired growth under 5% ethanol drastically limits the possibility to acquire adaptive mutations. Therefore, we believe that increasing the mutation rate is the only possibility to generate enough genetic variation in time to enable adaptation and avoid extinction. Although evolutionary rescue largely depends on the mutation rate (Gonzalez, et al., 2013; Gonzalez and Bell, 2012), the occurrence of hypermutation in relation to evolutionary rescue has not been discussed in detail. Possibly because hypermutation-enabled evolutionary rescue requires two mutational events: the occurrence of the mutator mutation and the occurrence of an adaptive mutation that rescues the population. The occurrence of hypermutation in all our high ethanol tolerant lines and the recurrent increases and decreases in mutation rate that match the changes in ethanol pressure during evolution demonstrate that hypermutation is crucial to allow fast adaptation of at least some individuals and rescue the population from extinction under near-lethal stress conditions.

First in Figure 4 even at extremely low frequencies, mutator alleles are selected for. This is quite surprising as at some point mutator subpopulation should be to low to generate any beneficial mutations. Indeed if we assume a 1000 fold effect on transitions, mutator should be outcompeted by wild type at a ratio of 1/10000 or so. So Figure 4 suggest a direct effect. Also mentioned in the text the fact that only point mutations were found in mutS and no inactivation is also an argument for a more complex effect than just increase in mutation rate. (inactivation should be much more likely to occur (though some deletions may not be detected at population level with the pipeline used) and as they have an impressive benefit in competition they should invade immediately…

It may be extremely difficult to go after mechanistic effects, but discussing that both primary selection and second order selection may be at play will benefit the paper.

We thank the reviewer for this interpretation of the data and for his very interesting suggestion to address the role of possible direct effects of MMR on the selection of hypermutation under high ethanol stress. We have now included an extensive discussion of possible directs effects to the manuscript.

“[…]These data indicate that the advantage of hypermutation under ethanol stress can be attributed mainly to second-order selection, following the beneficial effects of novel mutations relative to possible direct effects of the mutator mutation itself.[…]”

“[…] However, the following elements in our results might also suggest potential direct effects of the mutator mutations. First, mainly point mutations were identified in the mismatch repair genes (Figure 2—figure supplement 3) although inactivation of a gene is more likely to occur, given the high competitive benefit of the ΔmutS strain compared to the mutS_G100A strain in the presence of 5% ethanol (Figure 4—figure supplement 1). This would suggest selection of specific changes in the mechanism of the mismatch repair pathway. Second, both the ΔmutS and the mutS_G100A strains still increase in frequency when competed against the wild type in a ratio of 1:1000 or lower (Figure 4). Given the 10- to 50-fold increased mutation rate, a mutator subpopulation at a ratio of 1:1000 or lower should be too small to have an increased chance of acquiring a beneficial mutation compared to the wild-type subpopulation. These data suggest direct beneficial effects of MMR mutations (Cooper, et al., 2012; Torres-Barceló, et al., 2013; Nowosielska and Marinus, 2007) that combined with second-order selection can explain our observations. In addition, we confirmed that any increase in mutation rate, irrespective of the disrupted cellular system, can confer a selective benefit. Therefore these direct effects, which are usually the result of disruptions in one specific system or even in one gene, may not directly influence early selection of hypermutation, but may be obvious in its cost at later stages. […]”

“[…] Some direct effects of already accumulated mutations might explain the inconsistency between the 2-fold difference in mortality between IM2 and IM3 with a 10-fold difference in mutation rate (Figure 6) and the more than 10-fold difference in mortality between END and END ΔmutS with only a 2.5-fold difference in mutation rate (Figure 7). […]”

Apart from the possible direct effects as described above that might lead to direct selection of mostly MMR mutators, point mutations in the MMR genes might also be selected because they only moderately increase the mutation rate compared to the mutation rate caused by a full knock- out (Figure 4). This would suggest that moderately increased mutation rates are more beneficial because they have a lower cost at later stages compared to higher mutation rate variants. To elaborate on this point, we added a paragraph on this to our Discussion section:

“[…]Both these observations corroborate the idea that moderate mutators will be more easily selected for, because their benefit is higher than low mutation rate variants and their long-term cost is lower than high mutation rate variants. The identification of mostly point mutations leading to amino acid changes and not to nonsense mutations in the MMR genes during evolution similarly suggests selection for mild increases in the mutation rate (such as shown for the mutS_G100A mutant). […]”

Reviewer #3:

[…]

We have three main comments:

We included all the requested information in the manuscript.

“[…]The culture volume during the evolution experiment was 50 ml and dilution was 100- fold at each transfer[…].”

“[…]The minimal optical density of 0.2 corresponds to a minimal final cell density of 5.4x10⁸ CFU/ml in a volume of 50 ml. For each passage, we consequently transferred 500 µl or at least 2.73x10⁸ CFUs. The average number of generations for each growth cycle, estimated based on the optical density reached upon transfer, is 6.67. This value corresponds to the theoretical value resulting from log₂(dilution factor (100)) = 6.67. Taking these values together, we can calculate an estimated effective population size of 1.82x10⁹ CFU. […]”

Second, given the high mortality of cells, how were numbers of generations calculated? Dead cells have also been produced and they should be taken into account for the estimates of the total numbers of generations.

The number of generations was calculated based on the optical density measured at each intermediate time point. First, we calculated the corresponding number of CFU/ml for the measured optical density using an equation derived from the OD₅₉₅-CFU/ml standard curve. Using this CFU/ml value we calculated the total number of CFUs present in the 50 ml culture volume. Since we diluted 100x at each transfer, we could use the total number of CFUs from the previous time point to calculate the inoculum size. Finally, the increase in number of CFUs was calculated by dividing the final number of CFU by the inoculated number of CFU. The log base 2 of this value gives the number of generations (see subsection “Experimental evolution and Equation 3, Materials and methods).

Although this value is a good approximation of the number of generations as OD measurements also record dead cells, we are aware that this calculated value does not specifically take into account the formation of dead cells. Therefore, we added a paragraph to the manuscript explaining how we estimated the number of generations while mentioning the problem of the unknown death rate in the population.

“[…]The number of viable cells was estimated using optical density (A_595nm) values. This calculation does not specifically take into account the death rate of the cells. However, since the optical density reflects both living and death cells in the culture, calculating the number of generations using the OD values is more accurate compared to calculations based on CFU counts. Indeed, when using the viable cell count data of IM1, IM2 and IM3 (Figure 5), we found a 1.27-fold ( ± 0.29) underestimation of the number of generations when using the number of viable cells as compared to using OD values. […]”

Figure 6 —figure supplement 1B shows that the OD only decreases in late stationary phase, when a large proportion of the population is already dead. Given the fact that we transferred the cultures during the evolution experiment in exponential phase at an OD of at least 0.2, this density likely reflect all cells, both living and death in the culture. Hence, generations calculated from these density values will closely approximate the true number of generations. Nonetheless, to calculate the exact number of generations we would need to measure the death rate of each time point separately and extend the growth model with a death rate component. Given the elaborative nature of such experiments this would be practically unfeasible within a reasonable timeframe.

Third and perhaps most importantly, it was unclear whether fluctuation tests and genome sequence analyses have been done on individual clones or mixed population samples. As a result, we don't know how to interpret Figure 5 and Table 1: do mutation rates in Figure 5 reflect mutation rates of fixed genotypes or are these a mix of mutator and wild-type mutation rates of a polymorphic population? Similar for the frequency of observed mutations in Table 1: are these based on sequences of random clones picked form these populations or on mixed population DNA? The variation in mutation rate shown in Figure 5A suggests that these reflect population mutation rates, since changes are too small to reflect pure genotype effects if these involve mutator mutants such as mutS (see Figure 4). This distinction is crucial for the possible mechanisms that should be considered for explaining the dynamics in mutation rate: if high mutation-rate genotypes fix in the population, later observed low mutation-rate genotypes must derive from the former genotypes, whereas if they are *not* fixed, later observed low mutation-rate genotypes may derive from ancestral genotypes that were still present in the population when the increase in mutation rate was measured.

All fluctuation assays were done on mixed population samples. Therefore, one possible way of explaining the rapid decrease in mutation rate is by a shift in the population structure where a subpopulation with wild-type mutation rate that was still present, fixes again in the population. However, sequencing analysis on mixed pools using the low frequency variant caller (CLC Genomics Workbench) showed that all found mismatch repair mutator mutations (shown in Figure 2—figure supplement 2) became 100% fixed in the population and remained fixed even after a decrease in mutation rate was observed. Therefore, low mutation rate genotypes must have emerged from the former high mutation rate genotypes. Consequently, selection on low mutation rate genotypes must drive the rapid decrease in mutation rate observed during periods when ethanol concentrations remained constant. The speed of fixation of a mutant with a lower mutation rate and lower mortality might be enhanced by a lower production of dead cells and consequently faster increase in proportion in the population compared to the population with high mutation rate and high mortality.

In addition, we both sequenced the mixed pool and one selected clone from each pool. Table 1 is now replaced by Figure 2—figure supplement 2 that represents the mutations found in the mixed pool sample. Both the total amount of variants (frequency >10%) in each pool as well as the number of “fixed” (>75% frequency) mutations are shown. We now specifically added mention of population mutation rate in the manuscript to clarify the origin of the mutation rate data. Furthermore, we added a paragraph explaining the relevance for interpretation of the data.

“[…]Consequently, the population mutation rate will reflect the average genomic mutation rate of the entire population, containing different subpopulations that possibly display above- or below-average mutation rates. This may explain the discrepancy between the 20-fold increased endpoint mutation rate of line E1 (Figure 3A) and the 10-fold increased clonal mutation rate caused by the mutS_G100A mutation identified in that same line (Figure 4A). Furthermore, these data suggest that mutation rate can vary along with population structure throughout the evolution rather than being a fixed rate after the occurrence and spread of one mutator mutation. […]”

We continuously increased the percentage of ethanol during evolution. Since IM2 and IM3 were able to grow on the same percentage of ethanol (7.5%), it would only make sense to compete these intermediate time points against each other to determine fitness values linked to the mortality. However, the difference in number of generations between IM2 and IM3 exceeds 70 and it is plausible that in this time frame other adaptive mutations, anti-mutator mutations and additional genetic load have accumulated. For that reason, we believe that it is highly unlikely that we can directly link a fitness value from a competition between IM2 and IM3 to the mortality. To address this comment we therefore chose to calculate theoretical fitness consequences, rather than running direct competitions.

Accurate prediction of the effect of mortality on the fitness would necessitate setting up a new growth model including a parameter to account for the effect of the death rate. This would require running new experiments to gather enough data to build and substantiate the new model and possibly even a new collaboration with a more theoretically oriented lab. Therefore, we here used the available data and existing growth models to simulate a simplified competition experiment and calculate fitness value and number of selection rounds needed for a lower mutation rate genotype to fix in a population.

First, we determined the different growth parameters, such as growth rate (GR) and carrying capacity (y0). To this end, we used the time to grow in the presence of 7.5% EtOH and final cell density measurements originating from our evolution experiment. This leads to the following parameters for both IM2 and IM3 (Data tables available upon request).

	IM2	IM3
SGR (h^-1)	0.0973	0.392
Y₀	0.005	0.00363
Y_M	0.292	0.392

Open in a new tab

Next, we used these values to run a theoretical competition experiment between IM2 and IM3. Therefore, we used the Gompertz growth model (shown below) to calculate the theoretical number of CFU/ml for both strains after a 24 hour growth cycle (y(24)). Here, we made two assumptions. First, we were not able to derive a lag time (LT) value from the data of our evolution experiment. Therefore, we reasonably assumed equal lag times for IM2 and IM3, which is also suggested by the growth curves made for determination of the death rate (Figure 6—figure supplement 1). Second, we assumed that the individual growth dynamics of each strain equals their respective growth dynamics in direct competition with each other.

y (x) = y_{0} + y_{M} * e^{[- e^{((2.718 * \frac{S G R}{y_{M}}) * (L T - x) + 1)}]}

	IM2	IM3
CFU/ml_(o)	3.95E+06	3.95E+06
CFU/ml₍₂₄₎	1.73E+08	6.23E+08

Open in a new tab

Next, we used these CFU/ml values to calculate the relative fitness of IM3 (low mutation rate) compared to IM2 (high mutation rate), using the discrete time-recurrence equation that describes the spread of a mutant in a haploid population (Author response image 1). For A_start and A_end we used the proportions of IM2 and IM3 in the population determined by the number of CFU/ml for both strains after 24 hours as calculated in the previous step. This calculation resulted in an average relative fitness benefit of 3.61 for the IM3 strain with a low mutation rate compared to the IM2 strain with a high mutation rate.

Author response image 1. — **DOI:** http://dx.doi.org/10.7554/eLife.22939.025

Finally, we used this relative fitness to estimate the number of selection rounds necessary for fixation of IM2 in a population of IM3 genotypes. The effective population size in our evolution experiment was in the order of 10⁹ CFUs. If we assume that a mutant occurring in the population is present in a 1 over 10⁹ ratio, then we can calculate the number of selection rounds needed to increase in frequency to 99% of the population, given that this mutation is not lost due to random drift. Therefore, we solved the relative fitness equation for n as shown in Author response image 1 using both the relative fitness (3.61) and start (10^-9) and end (0.99) ratios of the competing strain.

Using this formula we calculated n necessary for IM3 to fix in a population of IM2. This resulted in an average of approximately 20 (19.74) selection rounds needed for IM3 to fix in a population with initially only IM2 genotypes. Assuming 6.67 generations per selection round this comes down to an average of approximately 132 generations. Consequently, the theoretical time needed to fix based on the calculated fitness parameters exceeds the actual time of 10 selection rounds or 70 generations observed between intermediate points IM2 and IM3. We added a new supplementary graph to show the comparison between calculated and observed number of selection rounds needed for IM3 to fix (Figure 6—figure supplement 3 and subsection “Cellular mortality is the underlying force driving evolution of mutation rates”). The discrepancy between these two values suggests a possible direct effect of further adaptive mutations or of mutator mutations themselves accumulated during consecutive growth steps between IM2 and IM3, as also suggested by Reviewer #2. These two strains not only differ in their mortality and mutation rate, but also in a further buildup of genetic load and possible mutations that allowed further adaptation to the 7.5% ethanol stress. Therefore, it is difficult to link the theoretical fitness calculations directly to the effect of anti-mutators or mortality alone as other factors are present that might influence these values.

Data shown in Figure 7, Figure 7—figure supplement 1 and Figure 7—figure supplement 2 tackle this problem further. Here, we continuously evolved a hypermutator sample from an intermediate time point to reduce the mutation rate. The reduction in mutation rate was accompanied by a reduction in the generation of dead cells. Through the introduction of the ΔmutS allele we now obtained both a low mutation rate variant (END) and a high mutation rate variant (END ΔmutS) that only differed in their mutS allele (and their mutation rate), but are otherwise isogenic. This allowed to calculate the same parameters as we did for IM2 and IM3, but now the results should only represent the effect of the changes in mutation rate and its consequences on the generation of dead cells. To theoretically simulate our competition experiment we calculated the number of selection rounds needed for the END ΔmutS strain to decrease in frequency from 50% to less than 1%. This calculation resulted in an average number of selection rounds of 2.5. Indeed, the direct competition between END and END ΔmutS starting from a 1:1 ratio demonstrates that after 2 selection rounds the frequency of the END ΔmutS strains was reduced to less than 10% (Figure 7—figure supplement 1). Since both competed strains only differ in their mutation rate and mortality, these data confirm that mutation rate and mortality are crucial factors to explain the fast increase of genotypes with a low mutation rate and mortality when the strain is already adapted. To clarify this part in the manuscript, we included an additional graph indicating that there is no significant difference between a theoretical calculation of selection rounds based on the fitness parameter of the strains and observed number of selection rounds necessary for fixation of the END (low mutation rate) strain in a direct competition experiment (Figure 7—figure supplement 2 and subsection “Cellular mortality is the underlying force driving evolution of mutation rates”).

Our third comment is on Figure 1. First, why would small inocula lead to much higher yields than large inocula? This is surprising, but remains unexplained. Second, the identity of the growth curves should be better explained. What do the grey lines reflect, individual replicates or mean curves for each mutator strain? The number of lines suggest the former, but the text refers to "the" growth curve for dnaQ, suggesting they reflect mean curves. It may be best to show mean growth curves or fitted growth models per strain, with different colors per strain. How are the different populations (of different size) generated? Do they come from the same 'batch' which are either less or more diluted? Or does the 10⁷ dilution come from a later time point of the smaller dilutions? The physiological state of the cells (eg early exponential phase, later exponential phase) at the moment that they are inoculated is known to affect the subsequently generated growth curves. Or perhaps, can larger population sizes 'soak' the ethanol, and thus effectively decrease the concentration in the environment? (see e.g. doi: 10.1098/rsbl.2012.0569)

All inocula originated from the same batch: The wild type and tested mutants were grown overnight and subsequently diluted to initiate the growth starting from different inoculum sizes. Therefore, we can exclude possible effects of the physiological state of the cell on the growth dynamics. However, we have two hypotheses on the difference in yield between high and low initial inocula.

First, the time axis is different for the two inoculum sizes. We monitored the growth curves starting from a large initial inoculum size for 24 hours compared to 120 hours when starting with a small initial inoculum size. The question here is whether the initial growth on 5% ethanol requires adaptive mutations. We hypothesized that the effect of adaptive mutations would be mitigated by a large initial population size, which is suggested by the results of the experiment and which means that the majority of the population is able to grow without adaptive mutations. However, we did not sequence any strains after the first growth cycle on 5% ethanol to confirm whether adaptive mutations are necessary or not. In addition, we expect more generations in case of the small initial inoculum size (log₂(dilution factor:100000)=16.61 generations) compared to the case of a large initial inoculum size (log₂(100)=6.67 generations). Therefore, a mutant occurring in case of a small initial inoculum size will have more time to manifest, possibly leading to the observed higher yield.

Second, growth of a population in the presence of high stress might be facilitated by soaking the ethanol as suggested by the reviewer. Soaking has been described for P. aeruginosa that produce a bacteriocin to eliminate possible competitors. The producer cells often have receptors to translocate and neutralize their own bacteriocin, thereby reducing the overall effect of the toxin (Inglis, et al., 2012). Soaking of the ethanol might also happen in larger populations thereby reducing the effective concentration. Ethanol soaking and degradation can happen through mutations in the alcohol dehydrogenase gene adhE and has been linked to higher ethanol tolerance before (Goodarzi, et al., 2010).

We changed the text to address the surprising observation of higher yields of small inocula compared to lower yields of large inocula.

“[…] Surprisingly, large initial populations lead to a lower yield compared to small initial population sizes. The growth from the small inoculum is likely driven by adaptive mutations, while the effect of a beneficial mutation might be mitigated when starting with a large inoculum. Moreover, we expect that a mutant occurring in case of a small initial inoculum size will have more time to manifest (log₂(dilution factor ∶ 100 000) = ± 16.61 generations), compared to the mutant occurring in case of a large initial population size (log₂(100) = ± 6.67 generations), possibly leading to the observed higher yield[…].”

We changed the caption of the figure to clarify the identity of the gray lines in the revised manuscript. The gray lines represent individual replicates of the mutator strains.

“[…]The blue line and shading represents the sigmoidal fit of the wild-type growth curves (n=3, fit using Gompertz equation with 95% c.i. (shading), see Equation 1 in methods section), while the grey lines represent growth curve of separate replicates for each mutator mutant. […]”

The chosen visualization was inspired by a recent paper (Peters, et al., 2016). Showing all the replicates highlights the similarity or diversity in great detail between the growth curves for large and small inocula, respectively. We tried visualizing the average growth curves as shown in Author response image 2, but these graphs did not entirely capture the dispersive nature of the growth curves, even though it is still clear that mutS, mutL, uvrD, mutH, mutT and mutM mutants grow much faster than the wild type when starting from a small initial inoculum size.

Author response image 2. — **DOI:** http://dx.doi.org/10.7554/eLife.22939.026

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Reviewers #4 and 5:

1) "A higher mutation rate is linked to a higher mortality, probably due to the extended buildup of genetic load and increased probability of acquiring a lethal mutation". There could be a more straight-forward explanation: Krasovec et al. (2014 Nature Communications) showed that mutation rate from fluctuation assay inversely correlates with population density, due to cell-cell interactions. For the same experimental evolution line at different intermediate time points, higher mortality could reduce population density-given the same inoculum size for fluctuation assay, which in turn increases mutation rate. This alternative interpretation needs to be addressed in the current experimental system.

This is an interesting suggestion to explain the observed differences in the mutation rate. However, we believe that it is not entirely applicable to our results. We have convincing data showing that mutator mutations occurred prior to changes in mortality and caused the increase in mutation rate, not the way around (Figure 4). We were not only able to identify these mutations (Figure 2—figure supplement 3) but we also show that an increased mutation rate, caused by these mutations, is necessary to enable adaptation under near-lethal ethanol stress. This was demonstrated both with a constructed set of mutation rate variants (Figure 1) as well as in an experimental evolution experiment (Figures 2 and 3). Therefore, we believe that the increases in mutation rate were caused by genuine mutational changes in DNA repair genes and not as a consequence of an increased death rate and subsequent reduction in population density.

Moreover, we carefully measured the population density while performing the fluctuation assays. This was done to avoid possible population density effects on the mutation rate, which were elegantly demonstrated by the Krašovec et al. study. We clarified this in the Materials and methods section.

“[…]To ensure correct comparison of the mutation rates we verified that the population densities at the time of plating on rifampicin did not differ significantly. If this was not the case, the test was repeated. The statistical difference between the population densities was measured using a one-way ANOVA with post-hoc Tukey correction. We found no significant difference for any intermediate point compared to the average density and to the density of the other time points. We therefore avoid possible population density effects (Krašovec, et al., 2014, Nat. Commun.). […]”

The final mean densities for each time point of E5 (Figure 5A) are now given in panel A of Author response image 3. Calculation of the Pearson’s correlation coefficient between the absolute population density and the corresponding mutation rate for each time point (panel B of Author response image 3) demonstrated that the population density did not significantly affect the mutation rate in our study. This is also shown by the dashed line representing the linear fitting through the data points which has a slope not significantly different from zero (P = 0.2318). Next, in case the population density would have influenced the mutation rate in our study, an increase in population density should be correlated with a decrease in mutation rate. However, the Pearson’s correlation coefficient between the relative difference in population density between two consecutive time points and the corresponding difference in mutation rate between those two points (panel C of Author response image 3) was not significant (P = 0.511). In conclusion, by carefully performing the fluctuation tests when population densities of the strains were not significantly different, we clearly exclude possible effects of cell density on the mutation rate

In the population, mutants with various mutation rates are generated continuously. When the environment turns stressful, the distribution of fitness effects of mutations accordingly becomes wider and less deleterious-mutation-biased due to the decreased population-average fitness (e.g. Hietpas et al., 2013). As a result, the mutation load of hypermutators relative to WT becomes smaller and hypermutators are more likely to survive or even prosper (if hitchhiked with beneficial mutations). The argument can be reversed when organisms become more adapted in the new environment to explain the decline of hypermutators. In this way, the dynamic pattern of mutation rates can be understood as a passive balance between influx (by random mutations) and outflux (due to accumulated genetic load) of mutators without evoking any active response. It would help if the authors can clarify if and why their data fit better with their claim than the above alternative interpretation.

We thank the reviewers for pointing out this alternative interpretation of the data. However, we believe that this interpretation at least partially overlaps with our current interpretation of the data.

Our results corroborate the idea that only cells with increased mutation rate can generate enough genetic diversity in a short time period to avoid extinction and enable adaptation under near-lethal ethanol stress conditions (potentially aided by a wider and less deleterious- mutation-biased distribution of fitness effects). However, we do not claim that the occurrence of hypermutation itself is an active process. Therefore, our interpretation is largely in line with the interpretation proposed by the reviewers. Indeed, mutants with variations in mutation rate are continuously generated in a population. Moreover, in changing environments the occurrence of hypermutable variants might be facilitated as more mutations are beneficial due to the decreased average population fitness (Hietpas, et al., 2013, Evolution). As suggested by the reviewers, this could mean that, under such near-lethal stress conditions, the genetic load of hypermutable variants is less deleterious-mutation biased, thereby possibly facilitating the emergence of mutators. Here, we moreover believe that these mutators are crucial to enable adaptation and avoid extinction. Since we observed hypermutation in all evolved high ethanol- tolerant lines, but not in the low ethanol-tolerant lines. We have strong data supporting the idea that only cells with a higher mutation rate can rapidly acquire adaptive mutations to avoid extinction. Additionally, our results also confirm that the death rate and mutation rate of a population are linked. These data suggest that higher mortality becomes the cost of hypermutation when a population is already adapted (potentially resulting from a distribution of fitness effects that became again more narrow and deleterious-mutation-biased). Consequently, mutants with a lower mutation rate and lower mortality will have a higher benefit and take over the population, thereby lowering the overall mutation rate. In conclusion, we state that our interpretation largely overlaps with the one proposed by the reviewers. To clarify this in the manuscript we adjusted the text accordingly.

“[…] In conclusion, even though mutator mutants occur spontaneously in the population, these data suggest that hypermutation is a prerequisite, as a driving force, to adapt to high ethanol levels in such a way that only lines with a higher mutation rate than the wild-type mutation rate are able to evolve high ethanol tolerance (Figure 3B). […]”

“[…] The rise of hypermutation during adaptation to near-lethal ethanol stress is linked to the idea of second-order selection as suggested by the growth rate and lag time measured for a collection of mutator mutants under 5% ethanol stress (Figure 1; Figure 1—figure supplement 3) and might be facilitated by a wider and less deleterious-mutation-biased distribution of fitness effects in stressful environments (Hietpas, et al., 2013 Evolution). […]”

Note that the above argument is assuming the fitness effects are due to secondary mutations, thus the authors' discovery of potential direct effects of mutator-generating mutations (e.g. mutS in Figure 1—figure supplement 4) is not affected. However, in the current wording the potential direct benefit of these mutations does not seem to be the main message they try to make impact. It is also not clear why the authors claim "…these direct effects…may not directly influence early selection of hypermutation, but may become decisive in its lower cost at later stages."

We acknowledge that the possible direct effects of mutator mutations might play a role, even though our data (Figure 1) strongly support the idea that mutators increase in frequency by second-order selection on linked, beneficial mutations. The direct effects might influence which specific mutator mutation eventually spread, but, for the initial adaptation, the effect of hypermutation (with its indirect effects of linked beneficial mutations) clearly is much more important than the (different) small direct effects of specific mutator mutations. To clarify this part, we changed the Discussion section of the manuscript accordingly:

“[…]Therefore, these direct effects, which are usually the result of disruptions of one specific system or even of one specific gene, may influence which specific mutator mutations eventually spread, but will only have a limited effect on the initial selection of hypermutation compared to the direct effect of linked beneficial mutations. However, at later stages these direct effects possibly affect the fate of hypermutators by lowering the cost of the extended buildup of genetic load. […]”

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary file 1. Addtional tables with primer sequences and selected mutator genes.

DOI: http://dx.doi.org/10.7554/eLife.22939.024

elife-22939-supp1.docx^{(23.2KB, docx)}

DOI: 10.7554/eLife.22939.024

[bib1] Baba T, Ara T, Hasegawa M, Takai Y, Okumura Y, Baba M, Datsenko KA, Tomita M, Wanner BL, Mori H. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Molecular Systems Biology. 2006;2:2006.0008. doi: 10.1038/msb4100050. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib2] Barrick JE, Lenski RE. Genome dynamics during experimental evolution. Nature Reviews Genetics. 2013;14:827–839. doi: 10.1038/nrg3564. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib3] Barrick JE, Yu DS, Yoon SH, Jeong H, Oh TK, Schneider D, Lenski RE, Kim JF. Genome evolution and adaptation in a long-term experiment with Escherichia coli. Nature. 2009;461:1243–1247. doi: 10.1038/nature08480. [DOI] [PubMed] [Google Scholar]

[bib4] Bell G, Gonzalez A. Adaptation and evolutionary rescue in metapopulations experiencing environmental deterioration. Science. 2011;332:1327–1330. doi: 10.1126/science.1203105. [DOI] [PubMed] [Google Scholar]

[bib5] Bielas JH, Loeb KR, Rubin BP, True LD, Loeb LA. Human cancers express a mutator phenotype. PNAS. 2006;103:18238–18242. doi: 10.1073/pnas.0607057103. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib6] Bjedov I, Tenaillon O, Gérard B, Souza V, Denamur E, Radman M, Taddei F, Matic I. Stress-induced mutagenesis in bacteria. Science. 2003;300:1404–1409. doi: 10.1126/science.1082240. [DOI] [PubMed] [Google Scholar]

[bib7] Blázquez J. Hypermutation as a factor contributing to the acquisition of antimicrobial resistance. Clinical Infectious Diseases. 2003;37:1201–1209. doi: 10.1086/378810. [DOI] [PubMed] [Google Scholar]

[bib8] Boe L, Danielsen M, Knudsen S, Petersen JB, Maymann J, Jensen PR. The frequency of mutators in populations of Escherichia coli. Mutation Research/Fundamental and Molecular Mechanisms of Mutagenesis. 2000;448:47–55. doi: 10.1016/S0027-5107(99)00239-0. [DOI] [PubMed] [Google Scholar]

[bib9] Cherepanov PP, Wackernagel W. Gene disruption in Escherichia coli: tcr and KmR cassettes with the option of Flp-catalyzed excision of the antibiotic-resistance determinant. Gene. 1995;158:9–14. doi: 10.1016/0378-1119(95)00193-A. [DOI] [PubMed] [Google Scholar]

[bib10] Chopra I, O'Neill AJ, Miller K. The role of mutators in the emergence of antibiotic-resistant bacteria. Drug Resistance Updates. 2003;6:137–145. doi: 10.1016/S1368-7646(03)00041-4. [DOI] [PubMed] [Google Scholar]

[bib11] Cooper LA, Simmons LA, Mobley HL. Involvement of mismatch repair in the reciprocal control of motility and adherence of uropathogenic Escherichia coli. Infection and Immunity. 2012;80:1969–1979. doi: 10.1128/IAI.00043-12. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib12] Couce A, Guelfo JR, Blázquez J. Mutational spectrum drives the rise of mutator bacteria. PLoS Genetics. 2013;9:e1003167. doi: 10.1371/journal.pgen.1003167. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib13] Datsenko KA, Wanner BL. One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. PNAS. 2000;97:6640–6645. doi: 10.1073/pnas.120163297. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib14] de Visser JA. The fate of microbial mutators. Microbiology. 2002;148:1247–1252. doi: 10.1099/00221287-148-5-1247. [DOI] [PubMed] [Google Scholar]

[bib15] Denamur E, Bonacorsi S, Giraud A, Duriez P, Hilali F, Amorin C, Bingen E, Andremont A, Picard B, Taddei F, Matic I. High frequency of mutator strains among human uropathogenic Escherichia coli isolates. Journal of Bacteriology. 2002;184:605–609. doi: 10.1128/JB.184.2.605-609.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib16] Desai MM, Fisher DS. The balance between mutators and nonmutators in asexual populations. Genetics. 2011;188:997–1014. doi: 10.1534/genetics.111.128116. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib17] Drake JW. A constant rate of spontaneous mutation in DNA-based microbes. PNAS. 1991;88:7160–7164. doi: 10.1073/pnas.88.16.7160. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib18] Drake JW, Charlesworth B, Charlesworth D, Crow JF. Rates of spontaneous mutation. Genetics. 1998;148:1667–1686. doi: 10.1093/genetics/148.4.1667. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib19] Elena SF, Lenski RE. Evolution experiments with microorganisms: the dynamics and genetic bases of adaptation. Nature Reviews Genetics. 2003a;4:457–469. doi: 10.1038/nrg1088. [DOI] [PubMed] [Google Scholar]

[bib20] Elena SF, de Visser JA. Environmental stress and the effects of mutation. Journal of Biology. 2003b;2:12. doi: 10.1186/1475-4924-2-12. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib21] Eyre-Walker A, Keightley PD. The distribution of fitness effects of new mutations. Nature Reviews Genetics. 2007;8:610–618. doi: 10.1038/nrg2146. [DOI] [PubMed] [Google Scholar]

[bib22] Ferroni A, Guillemot D, Moumile K, Bernede C, Le Bourgeois M, Waernessyckle S, Descamps P, Sermet-Gaudelus I, Lenoir G, Berche P, Taddei F. Effect of mutator P. aeruginosa on antibiotic resistance acquisition and respiratory function in cystic fibrosis. Pediatric Pulmonology. 2009;44:820–825. doi: 10.1002/ppul.21076. [DOI] [PubMed] [Google Scholar]

[bib23] Gentile CF, Yu SC, Serrano SA, Gerrish PJ, Sniegowski PD, Chao L, Cox EC. Competition between high- and higher-mutating strains of Escherichia coli. Evolution. 2011;7:125–134. doi: 10.2307/2408181. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib24] Giraud A, Matic I, Tenaillon O, Clara A, Radman M, Fons M, Taddei F. Costs and benefits of high mutation rates: adaptive evolution of bacteria in the mouse gut. Science. 2001;291:2606–2608. doi: 10.1126/science.1056421. [DOI] [PubMed] [Google Scholar]

[bib25] Gonzalez A, Bell G. Evolutionary rescue and adaptation to abrupt environmental change depends upon the history of stress. Philosophical Transactions of the Royal Society B: Biological Sciences. 2013;368:20120079. doi: 10.1098/rstb.2012.0079. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib26] Gonzalez A, Ronce O, Ferriere R, Hochberg ME. Evolutionary rescue: an emerging focus at the intersection between ecology and evolution. Philosophical Transactions of the Royal Society B: Biological Sciences. 2013;368:20120404. doi: 10.1098/rstb.2012.0404. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib27] Goodarzi H, Bennett BD, Amini S, Reaves ML, Hottes AK, Rabinowitz JD, Tavazoie S. Regulatory and metabolic rewiring during laboratory evolution of ethanol tolerance in E. coli. Molecular Systems Biology. 2010;6:378. doi: 10.1038/msb.2010.33. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib28] Gould CV, Sniegowski PD, Shchepetov M, Metlay JP, Weiser JN. Identifying mutator phenotypes among fluoroquinolone-resistant strains of Streptococcus pneumoniae using fluctuation analysis. Antimicrobial Agents and Chemotherapy. 2007;51:3225–3229. doi: 10.1128/AAC.00336-07. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib29] Gross MD, Siegel EC. Incidence of mutator strains in Escherichia coli and coliforms in nature. Mutation Research Letters. 1981;91:107–110. doi: 10.1016/0165-7992(81)90081-6. [DOI] [PubMed] [Google Scholar]

[bib30] Gupta DK, Patra AT, Zhu L, Gupta AP, Bozdech Z. DNA damage regulation and its role in drug-related phenotypes in the malaria parasites. Scientific Reports. 2016;6:23603. doi: 10.1038/srep23603. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib31] Haft RJ, Keating DH, Schwaegler T, Schwalbach MS, Vinokur J, Tremaine M, Peters JM, Kotlajich MV, Pohlmann EL, Ong IM, Grass JA, Kiley PJ, Landick R. Correcting direct effects of ethanol on translation and transcription machinery confers ethanol tolerance in bacteria. PNAS. 2014;111:E2576–E2585. doi: 10.1073/pnas.1401853111. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib32] Hall BM, Ma CX, Liang P, Singh KK. Fluctuation analysis CalculatOR: a web tool for the determination of mutation rate using Luria-Delbruck fluctuation analysis. Bioinformatics. 2009;25:1564–1565. doi: 10.1093/bioinformatics/btp253. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib33] Hall LM, Henderson-Begg SK. Hypermutable bacteria isolated from humans--a critical analysis. Microbiology. 2006;152:2505–2514. doi: 10.1099/mic.0.29079-0. [DOI] [PubMed] [Google Scholar]

[bib34] Hammerschmidt K, Rose CJ, Kerr B, Rainey PB. Life cycles, fitness decoupling and the evolution of multicellularity. Nature. 2014;515:75–79. doi: 10.1038/nature13884. [DOI] [PubMed] [Google Scholar]

[bib35] Hammerstrom TG, Beabout K, Clements TP, Saxer G, Shamoo Y. Acinetobacter baumannii repeatedly evolves a Hypermutator phenotype in response to Tigecycline that effectively surveys evolutionary trajectories to resistance. PLoS One. 2015;10:e0140489. doi: 10.1371/journal.pone.0140489. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib36] Healey KR, Zhao Y, Perez WB, Lockhart SR, Sobel JD, Farmakiotis D, Kontoyiannis DP, Sanglard D, Taj-Aldeen SJ, Alexander BD, Jimenez-Ortigosa C, Shor E, Perlin DS. Prevalent mutator genotype identified in fungal pathogen candida glabrata promotes multi-drug resistance. Nature Communications. 2016;7:11128. doi: 10.1038/ncomms11128. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib37] Hietpas RT, Bank C, Jensen JD, Bolon DN. Shifting fitness landscapes in response to altered environments. Evolution. 2013;67:3512–3522. doi: 10.1111/evo.12207. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib38] Iguchi S, Mizutani T, Hiramatsu K, Kikuchi K. Rapid acquisition of Linezolid resistance in Methicillin-Resistant Staphylococcus aureus: role of hypermutation and homologous recombination. PLoS One. 2016;11:e0155512. doi: 10.1371/journal.pone.0155512. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib39] Jolivet-Gougeon A, Kovacs B, Le Gall-David S, Le Bars H, Bousarghin L, Bonnaure-Mallet M, Lobel B, Guillé F, Soussy CJ, Tenke P. Bacterial hypermutation: clinical implications. Journal of Medical Microbiology. 2011;60:563–573. doi: 10.1099/jmm.0.024083-0. [DOI] [PubMed] [Google Scholar]

[bib40] Kaplan S, Bren A, Zaslaver A, Dekel E, Alon U. Diverse two-dimensional input functions control bacterial sugar genes. Molecular Cell. 2008;29:786–792. doi: 10.1016/j.molcel.2008.01.021. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib41] Krašovec R, Belavkin RV, Aston JA, Channon A, Aston E, Rash BM, Kadirvel M, Forbes S, Knight CG. Mutation rate plasticity in rifampicin resistance depends on Escherichia coli cell-cell interactions. Nature Communications. 2014;5:3742. doi: 10.1038/ncomms4742. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib42] LeClerc JE, Li B, Payne WL, Cebula TA. High mutation frequencies among Escherichia coli and Salmonella pathogens. Science. 1996;274:1208–1211. doi: 10.1126/science.274.5290.1208. [DOI] [PubMed] [Google Scholar]

[bib43] Lee AH, Fidock DA. Evidence of a mild mutator phenotype in cambodian Plasmodium falciparum malaria parasites. PLoS One. 2016;11:1–14. doi: 10.1371/journal.pone.0154166. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib44] Lee H, Popodi E, Tang H, Foster PL. Rate and molecular spectrum of spontaneous mutations in the bacterium Escherichia coli as determined by whole-genome sequencing. PNAS. 2012;109:E2774–E2783. doi: 10.1073/pnas.1210309109. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib45] Lenski RE, Rose MR, Simpson SC, Tadler SC. Long-Term experimental evolution in Escherichia coli. I. adaptation and divergence during 2,000 generations. The American Naturalist. 1991;138:1315–1341. doi: 10.1086/285289. [DOI] [Google Scholar]

[bib46] Lindsey HA, Gallie J, Taylor S, Kerr B. Evolutionary rescue from extinction is contingent on a lower rate of environmental change. Nature. 2013;494:463–467. doi: 10.1038/nature11879. [DOI] [PubMed] [Google Scholar]

[bib47] Loh E, Salk JJ, Loeb LA. Optimization of DNA polymerase mutation rates during bacterial evolution. PNAS. 2010;107:1154–1159. doi: 10.1073/pnas.0912451107. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib48] Lynch M, Ackerman MS, Gout JF, Long H, Sung W, Thomas WK, Foster PL. Genetic drift, selection and the evolution of the mutation rate. Nature Reviews Genetics. 2016;17:704–714. doi: 10.1038/nrg.2016.104. [DOI] [PubMed] [Google Scholar]

[bib49] Lynch M. Evolution of the mutation rate. Trends in Genetics. 2010;26:345–352. doi: 10.1016/j.tig.2010.05.003. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib50] Lynch M. The lower bound to the evolution of mutation rates. Genome Biology and Evolution. 2011;3:1107–1118. doi: 10.1093/gbe/evr066. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib51] Mao EF, Lane L, Lee J, Miller JH. Proliferation of mutators in A cell population. Journal of Bacteriology. 1997;179:417–422. doi: 10.1128/jb.179.2.417-422.1997. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib52] Marinus MG. DNA methylation and mutator genes in Escherichia coli K-12. Mutation Research/Reviews in Mutation Research. 2010;705:71–76. doi: 10.1016/j.mrrev.2010.05.001. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib53] Marvig RL, Johansen HK, Molin S, Jelsbak L. Genome analysis of a transmissible lineage of Pseudomonas aeruginosa reveals pathoadaptive mutations and distinct evolutionary paths of hypermutators. PLoS Genetics. 2013;9:e1003741. doi: 10.1371/journal.pgen.1003741. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib54] Marvig RL, Sommer LM, Molin S, Johansen HK. Convergent evolution and adaptation of Pseudomonas aeruginosa within patients with cystic fibrosis. Nature Genetics. 2015;47:57–64. doi: 10.1038/ng.3148. [DOI] [PubMed] [Google Scholar]

[bib55] Matic I, Radman M, Taddei F, Picard B, Doit C, Bingen E, Denamur E, Elion J. Highly variable mutation rates in commensal and pathogenic Escherichia coli. Science. 1997;277:1833–1834. doi: 10.1126/science.277.5333.1833. [DOI] [PubMed] [Google Scholar]

[bib56] Negri MC, Morosini MI, Baquero MR, del Campo R, Blázquez J, Baquero F. Very low cefotaxime concentrations select for hypermutable Streptococcus pneumoniae populations. Antimicrobial Agents and Chemotherapy. 2002;46:528–530. doi: 10.1128/AAC.46.2.528-530.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib57] Nicolaou SA, Gaida SM, Papoutsakis ET. Exploring the combinatorial genomic space in Escherichia coli for ethanol tolerance. Biotechnology Journal. 2012;7:1337–1345. doi: 10.1002/biot.201200227. [DOI] [PubMed] [Google Scholar]

[bib58] Nowosielska A, Marinus MG. DNA mismatch repair-induced double-strand breaks. DNA Repair. 2008;7:48–56. doi: 10.1016/j.dnarep.2007.07.015. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib59] Oliver A. Clinical relevance of Pseudomonas aeruginosa hypermutation in cystic fibrosis chronic respiratory infection. Journal of Cystic Fibrosis. 2015;14:e1–e2. doi: 10.1016/j.jcf.2014.12.009. [DOI] [PubMed] [Google Scholar]

[bib60] Otto SP, Day T. A Biologist’s Guide to Mathematical Modeling in Ecology and Evolution. Princeton University Press; 2007. [Google Scholar]

[bib61] Pal C, Maciá MD, Oliver A, Schachar I, Buckling A. Coevolution with viruses drives the evolution of bacterial mutation rates. Nature. 2007;450:1079–1081. doi: 10.1038/nature06350. [DOI] [PubMed] [Google Scholar]

[bib62] Perfeito L, Fernandes L, Mota C, Gordo I. Adaptive mutations in bacteria: high rate and small effects. Science. 2007;317:813–815. doi: 10.1126/science.1142284. [DOI] [PubMed] [Google Scholar]

[bib63] Pulido-Tamayo S, Sánchez-Rodríguez A, Swings T, Van den Bergh B, Dubey A, Steenackers H, Michiels J, Fostier J, Marchal K. Frequency-based haplotype reconstruction from deep sequencing data of bacterial populations. Nucleic Acids Research. 2015;43:e105. doi: 10.1093/nar/gkv478. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib64] Rajanna C, Ouellette G, Rashid M, Zemla A, Karavis M, Zhou C, Revazishvili T, Redmond B, McNew L, Bakanidze L, Imnadze P, Rivers B, Skowronski EW, O'Connell KP, Sulakvelidze A, Gibbons HS. A strain of Yersinia pestis with a mutator phenotype from the republic of Georgia. FEMS Microbiology Letters. 2013;343:113–120. doi: 10.1111/1574-6968.12137. [DOI] [PubMed] [Google Scholar]

[bib65] Ram Y, Hadany L. Stress-induced mutagenesis and complex adaptation. Proceedings of the Royal Society B: Biological Sciences. 2014;281:20141025. doi: 10.1098/rspb.2014.1025. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib66] Rando OJ, Verstrepen KJ. Timescales of genetic and epigenetic inheritance. Cell. 2007;128:655–668. doi: 10.1016/j.cell.2007.01.023. [DOI] [PubMed] [Google Scholar]

[bib67] Raynes Y, Sniegowski PD. Experimental evolution and the dynamics of genomic mutation rate modifiers. Heredity. 2014;113:375–380. doi: 10.1038/hdy.2014.49. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib68] Roberts SA, Gordenin DA. Hypermutation in human cancer genomes: footprints and mechanisms. Nature Reviews Cancer. 2014;14:786–800. doi: 10.1038/nrc3816. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib69] Ronen M, Rosenberg R, Shraiman BI, Alon U. Assigning numbers to the arrows: parameterizing a gene regulation network by using accurate expression kinetics. PNAS. 2002;99:10555–10560. doi: 10.1073/pnas.152046799. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib70] Sandberg TE, Pedersen M, LaCroix RA, Ebrahim A, Bonde M, Herrgard MJ, Palsson BO, Sommer M, Feist AM. Evolution of Escherichia coli to 42 °C and subsequent genetic engineering reveals adaptive mechanisms and novel mutations. Molecular Biology and Evolution. 2014;31:2647–2662. doi: 10.1093/molbev/msu209. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib71] Schaaper RM, Dunn RL. Spectra of spontaneous mutations in Escherichia coli strains defective in mismatch correction: the nature of in vivo DNA replication errors. PNAS. 1987;84:6220–6224. doi: 10.1073/pnas.84.17.6220. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib72] Schaaper RM, Radman M. The extreme mutator effect of Escherichia coli mutD5 results from saturation of mismatch repair by excessive DNA replication errors. The EMBO Journal. 1989;8:3511–3516. doi: 10.1002/j.1460-2075.1989.tb08516.x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib73] Shaver AC, Dombrowski PG, Sweeney JY, Treis T, Zappala RM, Sniegowski PD. Fitness evolution and the rise of mutator alleles in experimental Escherichia coli populations. Genetics. 2002;162:557–566. doi: 10.1093/genetics/162.2.557. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib74] Sniegowski P, Raynes Y. Mutation rates: how low can you go? Current Biology. 2013;23:R147–R149. doi: 10.1016/j.cub.2013.01.018. [DOI] [PubMed] [Google Scholar]

[bib75] Sniegowski PD, Gerrish PJ, Johnson T, Shaver A. The evolution of mutation rates: separating causes from consequences. BioEssays. 2000;22:1057–1066. doi: 10.1002/1521-1878(200012)22:12<1057::AID-BIES3>3.0.CO;2-W. [DOI] [PubMed] [Google Scholar]

[bib76] Sniegowski PD, Gerrish PJ, Lenski RE. Evolution of high mutation rates in experimental populations of E. coli. Nature. 1997;387:703–705. doi: 10.1038/42701. [DOI] [PubMed] [Google Scholar]

[bib77] Stepanyan K, Wenseleers T, Duéñez-Guzmán EA, Muratori F, Van den Bergh B, Verstraeten N, De Meester L, Verstrepen KJ, Fauvart M, Michiels J. Fitness trade-offs explain low levels of persister cells in the opportunistic pathogen Pseudomonas aeruginosa. Molecular Ecology. 2015;24:1572–1583. doi: 10.1111/mec.13127. [DOI] [PubMed] [Google Scholar]

[bib78] Sung W, Ackerman MS, Dillon MM, Platt TG, Fuqua C, Cooper VS, Lynch M. Evolution of the Insertion-Deletion mutation rate across the tree of Life. Genes|Genomes|Genetics. 2016;6:2583–2591. doi: 10.1534/g3.116.030890. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib79] Tanaka MM, Bergstrom CT, Levin BR. The evolution of mutator genes in bacterial populations: the roles of environmental change and timing. Genetics. 2003;164:843–854. doi: 10.1093/genetics/164.3.843. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib80] Tenaillon O, Toupance B, Le Nagard H, Taddei F, Godelle B. Mutators, population size, adaptive landscape and the adaptation of asexual populations of bacteria. Genetics. 1999;152:485–493. doi: 10.1093/genetics/152.2.485. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib81] Thierauf A, Perez G, Maloy AS. Generalized transduction. Methods in Molecular Biology. 2009;501:267–286. doi: 10.1007/978-1-60327-164-6_23. [DOI] [PubMed] [Google Scholar]

[bib82] Thompson DA, Desai MM, Murray AW. Ploidy controls the success of mutators and nature of mutations during budding yeast evolution. Current Biology. 2006;16:1581–1590. doi: 10.1016/j.cub.2006.06.070. [DOI] [PubMed] [Google Scholar]

[bib83] Torres-Barceló C, Cabot G, Oliver A, Buckling A, Maclean RC. A trade-off between oxidative stress resistance and DNA repair plays a role in the evolution of elevated mutation rates in bacteria. Proceedings of the Royal Society B: Biological Sciences. 2013;280:20130007. doi: 10.1098/rspb.2013.0007. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib84] Tröbner W, Piechocki R. Selection against hypermutability in Escherichia coli during long term evolution. MGG Molecular & General Genetics. 1984;198:177–178. doi: 10.1007/BF00328720. [DOI] [PubMed] [Google Scholar]

[bib85] Van den Bergh B, Michiels JE, Wenseleers T, Windels EM, Boer PV, Kestemont D, De Meester L, Verstrepen KJ, Verstraeten N, Fauvart M, Michiels J. Frequency of antibiotic application drives rapid evolutionary adaptation of Escherichia coli persistence. Nature Microbiology. 2016;1:16020. doi: 10.1038/nmicrobiol.2016.20. [DOI] [PubMed] [Google Scholar]

[bib86] Voordeckers K, Kominek J, Das A, Espinosa-Cantú A, De Maeyer D, Arslan A, Van Pee M, van der Zande E, Meert W, Yang Y, Zhu B, Marchal K, DeLuna A, Van Noort V, Jelier R, Verstrepen KJ. Adaptation to high ethanol reveals complex evolutionary pathways. PLoS Genetics. 2015;11:e1005635. doi: 10.1371/journal.pgen.1005635. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib87] Wang J, Cazzato E, Ladewig E, Frattini V, Rosenbloom DI, Zairis S, Abate F, Liu Z, Elliott O, Shin YJ, Lee JK, Lee IH, Park WY, Eoli M, Blumberg AJ, Lasorella A, Nam DH, Finocchiaro G, Iavarone A, Rabadan R. Clonal evolution of glioblastoma under therapy. Nature Genetics. 2016;48:768–776. doi: 10.1038/ng.3590. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib88] Wielgoss S, Barrick JE, Tenaillon O, Wiser MJ, Dittmar WJ, Cruveiller S, Chane-Woon-Ming B, Médigue C, Lenski RE, Schneider D. Mutation rate dynamics in a bacterial population reflect tension between adaptation and genetic load. PNAS. 2013;110:222–227. doi: 10.1073/pnas.1219574110. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib89] Wiser MJ, Ribeck N, Lenski RE. Long-term dynamics of adaptation in asexual populations. Science. 2013;342:1364–1367. doi: 10.1126/science.1243357. [DOI] [PubMed] [Google Scholar]

[bib90] Woods RJ, Barrick JE, Cooper TF, Shrestha U, Kauth MR, Lenski RE. Second-order selection for evolvability in a large Escherichia coli population. Science. 2011;331:1433–1436. doi: 10.1126/science.1198914. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib91] Yu J, Xiao J, Ren X, Lao K, Xie XS. Probing gene expression in live cells, one protein molecule at a time. Science. 2006;311:1600–1603. doi: 10.1126/science.1119623. [DOI] [PubMed] [Google Scholar]

[bib92] Zaslaver A, Bren A, Ronen M, Itzkovitz S, Kikoin I, Shavit S, Liebermeister W, Surette MG, Alon U. A comprehensive library of fluorescent transcriptional reporters for Escherichia coli. Nature Methods. 2006;3:623–628. doi: 10.1038/nmeth895. [DOI] [PubMed] [Google Scholar]

[bib93] Zaslaver A, Mayo AE, Rosenberg R, Bashkin P, Sberro H, Tsalyuk M, Surette MG, Alon U. Just-in-time transcription program in metabolic pathways. Nature Genetics. 2004;36:486–491. doi: 10.1038/ng1348. [DOI] [PubMed] [Google Scholar]

[bib94] Zwietering MH, Jongenburger I, Rombouts FM, van 't Riet K. Modeling of the bacterial growth curve. Applied and Environmental Microbiology. 1990;56:1875–1881. doi: 10.1128/aem.56.6.1875-1881.1990. [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Adaptive tuning of mutation rates allows fast response to lethal stress in Escherichia coli

Toon Swings

Bram Van den Bergh

Sander Wuyts

Eline Oeyen

Karin Voordeckers

Kevin J Verstrepen

Maarten Fauvart

Natalie Verstraeten

Jan Michiels

Roles

Abstract

eLife digest

Introduction

Results

Hypermutation enables rapid growth under high ethanol stress

Figure 1. Hypermutation favors growth in high EtOH stress through generation of beneficial mutations.

Figure 1—figure supplement 1. Deletion of selected mutator genes causes increased mutation rates under normal growth conditions.

Figure 1—figure supplement 2. 5% EtOH mimics near-lethal stress and leads to a severe decrease in growth rate and a decrease in carrying capacity.

Figure 1—figure supplement 3. Lag times reveal a window of optimal mutation rates for growth in the presence of 5% EtOH.

Figure 1—figure supplement 4. Relative fitness W associated with different mutation rate variants reveals an advantage for the ∆mutS mutator under EtOH stress.

Long-term adaptation to high ethanol stress in E. coli is contingent upon hypermutation

Figure 2. Experimental evolution of E. coli to increasing EtOH concentrations.

Figure 2—figure supplement 1. Flowchart of the experimental evolution experiment to high EtOH tolerance in E. coli.

Figure 2—figure supplement 2. The total number of mutations exceeds the number of fixed mutations in the population.

Figure 2—figure supplement 3. The mutational spectrum of evolved EtOH-tolerant lines corresponds to the mutational spectrum of MMR mutators.

Figure 3. Increased mutation rate underlies evolution of high EtOH tolerance.

Figure 4. mutS mutators are able to outcompete wild-type cells in direct competition under 5% EtOH, irrespective of the mutator population size.

Figure 4—figure supplement 1. Setup mutS mutator competition experiment under 5% EtOH.

Dynamics in mutation rate underlie evolution to high ethanol tolerance

Figure 5. Dynamics in population mutation rate underlie evolution to high EtOH tolerance.

Figure 5—figure supplement 1. Dynamics in mutation rate during evolution are not affected by ethanol itself.

Figure 6. Mortality is the cost of hypermutation in evolved strains when high EtOH tolerance is reached.

Figure 6—figure supplement 1. Number of viable cells decreases more rapidly in highly EtOH-tolerant strains than in the wild type.

Figure 6—figure supplement 2. The death rate is constant during exponential-phase growth of IM1 (Figure 6).

Figure 6—figure supplement 3. Direct effects of acquired mutations possibly cause the discrepancy between calculated and observed number of selection rounds necessary for fixation of IM3.

Cellular mortality is the underlying force driving evolution of mutation rates

Figure 7. Mortality is the cost of hypermutation when adaptation to a certain EtOH stress level is achieved.

Figure 7—figure supplement 1. In direct competition under 7% EtOH the strain with a low mutation rate outcompetes the hypermutating strain.

Figure 7—figure supplement 2. The theoretical number of selection rounds needed for END ΔmutS to fix corresponds to the actual observed number of selection rounds.

Discussion

Materials and methods

Bacterial strains and culture conditions

Construction of deletion mutants

Determination of near-lethal EtOH percentage

Determination of lag time and growth rate

Competition assay

Experimental evolution

Fluctuation assay

Whole-genome sequencing and identification of mutations

Mortality assay

Live-dead staining

Acknowledgements

Funding Statement

Funding Information

Additional information

Competing interests

Author contributions

Additional files

Major datasets

References

Decision letter

Roles

Author response

Author response image 1.

Author response image 2.

Author response image 3.

Associated Data

Supplementary Materials

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases