Skip to main content
PLOS Genetics logoLink to PLOS Genetics
. 2025 Mar 28;21(3):e1011627. doi: 10.1371/journal.pgen.1011627

Quantitative analysis of relationship between mutation rate and speed of adaptation under antibiotic exposure in Escherichia coli

Atsushi Shibai 1,#, Minako Izutsu 1,2,3,4,#, Hazuki Kotani 1, Chikara Furusawa 1,5,*
Editor: Megan Behringer6
PMCID: PMC11975134  PMID: 40153704

Abstract

Mutations are the ultimate source of biological evolution that creates genetic variation in populations. Mutations can create new advantageous traits but can also potentially interfere with pre-existing organismal functions. Therefore, organisms may have evolved mutation rates to appropriate levels to maintain or improve their fitness. In this study, we aimed to experimentally quantify the relationship between the mutation rate and evolution of antibiotic resistance. We conducted an evolution experiment using 12 Escherichia coli mutator strains with increased mutation rates and five antibiotics. Our results demonstrated that the rate of adaptation generally increased with higher mutation rates, except in a single mutator strain with the highest mutation rate, which exhibited a significant decline in evolutionary speed. To further elucidate these findings, we developed a simple population dynamics model that successfully recapitulated the observed dependence of adaptation speed on mutation rate. These findings provide important insights into the evolution of mutation rate accompanied by the evolution.

Author summary

Bacteria evolve to survive in changing environments, but how does the rate of genetic mutations influence their ability to adapt, especially to antibiotics? In this study, we explored how different mutation rates affect the speed at which Escherichia coli bacteria develop antibiotic resistance. By engineering bacterial strains with varying mutation rates and exposing them to five different antibiotics, we found that, in most cases, higher mutation rates led to faster adaptation. However, in one strain with an extremely high mutation rate, the ability to evolve was significantly reduced. Using a computational model, we confirmed this pattern and discovered that different antibiotics influence bacterial evolution in unique ways. Our findings help explain why some bacteria rapidly develop drug resistance while others struggle, providing valuable insights into antibiotic resistance. Understanding this balance between mutation and adaptation could lead to better strategies for controlling and preventing resistant infections.

Introduction

Mutations have the ability to create new advantageous traits that are favored by natural selection, while having the potential to interfere with pre-existing organismal functions. Mutations exert a double-edged-sword influence on organisms, conferring either beneficial or deleterious effects [1], suggesting that mutation rates dynamically change over the course of evolution [1–3]. Specifically, under certain selection pressures, the speed of adaptive evolution can be accelerated by increased mutation rates, as evidenced in multiple studies [4–7]. These findings suggest that alleles modifying mutation rates can be subject to positive selection under certain conditions, or alternatively, they may become fixed in the population through hitchhiking on other beneficial mutations [4,8]. For instance, in a long-term experimental evolution study of asexual Escherichia coli populations, cells exhibiting significantly elevated mutation rates, termed ’mutators’, have been observed [9,10]. In another example, an antibiotic selection experiment involved a mutator E. coli strain that initially exhibited an increased mutation rate due to the knockout of one gene. Subsequent increases in the mutation rate, caused by mutations in additional genes, were repeatedly observed [7]. However, when the mutation rate is excessively high, the overall fitness may decrease because deleterious mutations arise more frequently than beneficial mutations in most cases [2,10] and populations cannot avoid the accumulation of deleterious mutations. When a high mutation rate leads to a decrease in fitness, natural selection reduces the mutation rate. Indeed, several studies have demonstrated a reduction in the mutation rate during evolution [3,11,12]. The beneficial and harmful effects of mutations imply a complex and non-linear relationship between the mutation rate and the speed of fitness increase.

The experimental evolution of asexual bacterial populations aided by whole-genome sequencing is a powerful tool for investigating the effects of mutation rates on evolution [1,3–6]. Bacterial strains with different mutation rates can be prepared by deleting genes related to DNA replication or repair mechanisms as the mutator strains. Using these strains, we evaluated how changes in mutation rates affect the course of evolution. For example, Sprouffske et al. conducted experimental evolution using engineered E. coli strains with four different mutation rates [13]. These strains evolved for 3,000 generations in minimal medium without explicit stressors. Although populations with higher mutation rates had greater genetic diversity, this diversity benefited only when the mutation rate was modestly high. The study demonstrated that the highest mutation rates they used were not optimal for adaptation to the environment during long-term cultivation or for stress tolerance in novel environments after evolution.

Although the question of how evolutionary dynamics depend on mutation rates is important and has been the focus of many studies, the extent to which this relationship is influenced by a selective environment remains uncertain. This relationship is expected to depend on multiple factors including the distribution of fitness effects, population size, and the type of selection pressure. Analyzing the mutation rate dependency of evolutionary dynamics allows us to capture the contributions of these factors, providing a better understanding of how mutation rates evolve in nature and in laboratories. For example, the evolution of antibiotic resistance in microorganisms has been extensively studied in both laboratory and clinical settings [14–18] while the association between mutation rates and those events remains obscure. Given that recent studies have revealed drug resistance can be achieved by loss-of-function mutations [18] and studies that show the fitness cost of resistant genotypes [19,20], it is likely that the distribution of fitness effects in environments that contain drugs is qualitatively different from that in environments without drugs, and hypermutator genotypes might be favored by selection in the former environments. Examining how evolutionary dynamics under antibiotic treatment depend on mutation rates will provide valuable insights into the broader mechanisms underlying the emergence of antibiotic resistance and the evolution of mutation rate itself.

In this study, we aimed to experimentally quantify the mutation-rate dependency of the evolution of antibiotic resistance. A recent study by Gifford et al. examined the advantage of a mutator sub-population for the evolution of drug resistance [21]. They changed the frequency of the mutator strain in the evolving populations while using the same mutator strain made by mutS deletion for all populations, which might cause biases on the mutation spectra [22]. To avoid capturing biased trends, we constructed 12 E. coli mutator strains with elevated mutation rates by deleting genes with different mutation spectra and combinations of those genes. It is important to note that the emergence of such multilocus mutators has been previously observed in antibiotic selection experiments [7]. Subsequently, starting with these mutator strains, we conducted evolution experiments using five different antibiotics, each with a distinct mechanism of action. The results revealed that the speed of adaptation, as measured by the rate of increase in the minimum inhibitory concentration (MIC), increased approximately linearly with mutation rate, except for a single mutator strain with the highest mutation rate, which exhibited a significant decrease in adaptation speed. Additionally, the results demonstrated that this dependency varied between bacteriostatic and bactericidal antibiotics. We successfully reproduced these mutation-rate dependencies using numerical simulations of a population dynamics model. These findings offer important insights into the evolution of mutation rates accompanying evolutionary processes.

Results

Construction of mutator strains

We used the E. coli MDS42 strain as the wild-type (WT) and generated knockout mutants of the mutS, mutH, mutL, mutT, and dnaQ genes, denoted as S, H, L, T, and Q, respectively. The mutS, mutH, and mutL genes are involved in the mismatch repair machinery [23], mutT plays a role in maintaining replication fidelity [24], and dnaQ codes for the epsilon subunit of DNA polymerase III [25]. Deletion of these genes causes a loss in replication fidelity and an increase in mutation rates. Additionally, we created seven double-gene-knockout strains and obtained 12 mutator strains, which are listed in Table 1. The sequence of gene deletions during the construction of the mutator strains is summarized in S1 Fig in S1 File.

Table 1. Summary of MA experiments. The two right-hand columns display the numbers of synonymous (Syn) and nonsynonymous (NSyn) base-pair substitutions (BPS) identified, respectively, while the mean and standard deviation of three replicates are presented. All strains except LQ had identical numbers of MA rounds among the three replicate MA lines, respectively. The column of Generations shows the number of elapsed generations estimated from the number of MA rounds and the recorded colony size of each round, while the mean and standard deviation of three replicates are presented. A list of all detected mutations and an extended table showing the number of intergenic BPS and short Indels are shown in S1 and S2 Tables, respectively.

Strain Abbr. MA rounds Generations No. of Syn BPSs No. of NSyn BPSs
MDS42 WT 66 1685.7 ± 7.0 0.67 ± 0.58 0.67 ± 0.58
MDS42ΔmutH H 23 613.4 ± 1.7 4.33 ± 3.21 11.67 ± 1.53
MDS42ΔmutL L 66 1675.5 ± 0.8 8.33 ± 2.31 15.00 ± 1.00
MDS42ΔmutS S 66 1691.3 ± 9.3 11.00 ± 2.00 24.33 ± 1.15
MDS42ΔdnaQ Q 66 1651.0 ± 5.9 8.67 ± 1.15 23.33 ± 2.89
MDS42ΔmutT T 66 1679.7 ± 13.6 27.33 ± 2.31 157.67 ± 20.21
MDS42ΔmutLΔmutH LH 66 1651.4 ± 13.2 10.67 ± 1.53 15.00 ± 2.00
MDS42ΔmutSΔmutH SH 66 1691.1 ± 9.4 13.33 ± 5.13 21.00 ± 6.56
MDS42ΔmutLΔdnaQ LQ 69,61,61 1461.2 ± 96.0 545.67 ± 60.72 939.67 ± 81.5
MDS42ΔmutHΔmutT HT 23 610.8 ± 0.6 15.67 ± 5.03 76.33 ± 6.03
MDS42ΔmutLΔmutT LT 23 614.7 ± 2.5 16.00 ± 2.65 84.67 ± 14.57
MDS42ΔmutSΔmutT ST 23 605.0 ± 4.0 13.33 ± 4.51 54.33 ± 10.97
MDS42ΔdnaQΔmutT QT 23 611.7 ± 2.6 13.33 ± 2.08 63.33 ± 7.09

To quantify the mutation rate, we conducted mutation accumulation (MA) experiments using mutator strains as ancestors. Specifically, we propagated three lineages for each ancestor as single colonies on an agar plate medium for 23-69 passages. We estimated the number of generations during the MA experiment by establishing a relationship between colony size and cell number [26–28]. Subsequently, we sequenced the samples at the end of the MA experiment to detect point mutations that had accumulated in the genome. The total number of identified base-pair substitutions (BPS) is summarized in Table 1.

We calculated the mutation rates per generation for each strain following a previous study [29]. Initially, we determined the proportion of each pattern of synonymous base-pair substitutions. Subsequently, we estimated the genome-wide mutation rate by dividing the number of accumulated synonymous mutations in a genome by the number of generations and then normalizing it with the frequency of possible mutational patterns. Additionally, the frequency of insertions and deletions (indels) was calculated.

In this analysis, the mutation rate was calculated based on the number of mutations observed at the endpoint of the MA experiments. This calculation assumes that the number of mutations accumulates linearly during the MA experiments. While the dynamics of mutation accumulation could potentially be more complex and variable, we adopt this simple assumption as it represents the least biased approach. Supporting this assumption, a previous study [26] demonstrated that during MA experiments spanning 6,000 generations, the mutation rates for the first 3,000 generations and the entire 6,000 generations were not significantly different. This observation is consistent with our assumption that the number of mutations increases linearly during our MA experiments.

As shown in Fig 1, our results showed that the mutator strains exhibited mutation rates approximately 6–400 times higher than that of the WT. The base-pair substitution patterns varied depending on the knockout gene(s), as shown in the pie charts in Fig 1, consistent with previous studies [24,27,30–33]. Specifically, disruption of the mismatch repair mechanism (ΔmutS, ΔmutH, ΔmutL) increased A:T to G:C and G:C to A:T substitutions and indels, whereas knockout of the mutT gene increased only A:T to C:G. Additionally, for each mutator strain, we calculated the dN/dS ratio, which represents the ratio of nonsynonymous to synonymous substitution rates. We found that the dN/dS ratios did not differ from 1 for all strains except WT, which had scarcely accumulated mutations (S2A Fig in S1 File). A dN/dS ratio close to 1 suggests that selection had little effect on our MA experiments.

Fig 1. Mutation rates of mutator E. coli strains.

Fig 1

Each dot with error bars represents the mean and standard deviation of the mutation rates observed in replicate MA lineages. The BPS rate (closed circle) was estimated using only synonymous mutations to exclude the effects of purifying selection. The indels rate (open circle) was also calculated only from the number of intergenic indels and normalized by dividing by the fraction of the length of the intergenic region in the genome (11%). The pie charts above the dots show the distribution of substitution patterns identified in the mutator strains, excluding the wild-type strain, which had as few as two mutations.

We then quantified the growth rates of the mutator strains in M9 minimum medium without antibiotics. As depicted in S2B Fig in S1 File, the growth rate decreases with an increasing mutation rate, a pattern also observed in other studies [34]. This observed decrease in growth rate is thought to be caused, at least in part, by an increase in mutation rate. However, the potential contributions of other mechanisms, such as replication inhibition and activation of proofreading mechanisms, are also suggested [35].

Quantifying speed of adaptation under antibiotics

To elucidate the relationship between mutation rate and evolutionary dynamics, we conducted experimental evolution of 12 mutator strains and the wild-type strain using five antibiotics with distinct mechanisms of action, namely chloramphenicol (CP), trimethoprim (TP), amikacin (AMK), cefixime (CFIX), and ciprofloxacin (CPFX). We maintained four replica lines for each strain-antibiotic combination, resulting in a total of 260 individually evolving lines (13 strains × 5 antibiotics × 4 replicas). The cells were cultured in 200 µL of M9 medium supplemented with 20 amino acids (M9+AA medium) in a 96-well microtiter plate, to which each antibiotic was added in a two-fold dilution series (Fig 2A). Cultivation began with a fixed initial cell concentration (OD620 value of 3×10-4, corresponding to approximately 2×105 cells). After 24 h of incubation, cells were collected from the wells with the highest drug concentration among the wells with OD values above a certain threshold (OD620=0.03). The collected cells were then transferred to fresh medium containing the antibiotic dilution series, with the initial OD620 value of 3×10-4. We repeated this serial transfer procedure for nine days, corresponding to approximately 70 generations of cells. The number of generations during experimental evolution differed depending on the size of the overshoot from the threshold OD at each end of the incubation. S5 Table shows the precise number of generations of evolutionary lines calculated using OD values.

Fig 2. Experimental evolution under antibiotic selection pressure.

Fig 2

(A) Schematic representation of the experimental procedure. A 96-well microtiter plate was prepared with two-fold serial dilutions of antibiotics. After 24 hours of cultivation, cells from the well with the highest drug concentration, in which the cell concentration (OD620) exceeded 0.03, were transferred to fresh medium with a drug concentration gradient. In this study, we designated the drug concentration of the selected well (green asterisk) as the minimal inhibitory concentration (MIC). The serial transfer cultures were iterated for 9 days. (B) Example of MIC changes during experimental evolution. The vertical axis shows the MIC for trimethoprim (TP). The black and blue lines represent the data for the WT and ΔmutS (S) strains, respectively. For each strain, data from four replicate series are overlaid. Data for all combinations of strains and drugs are presented in S3 Fig.

In this study, we defined the minimum inhibitory concentration (MIC) as the drug concentration in the well from which the cells were transferred. Fig 2B shows typical examples of the time series of MIC under TP selection, whereas all MIC time series of 260 evolutionary lines are presented in S3 Fig in S1 File. Using these time series MIC data, we evaluated the speed of drug resistance evolution under antibiotic treatment by determining the change in log2-transformed MIC per day, which was obtained by comparing the MIC values of the first and last days of the 9-day evolutionary experiment. We assumed that MIC directly corresponds to fitness, disregarding the non-monotonic relationship between MIC and fitness as well as the dependency of the MIC-fitness relationship on drug concentration. We assumed that MIC directly corresponds to fitness, disregarding the non-monotonic relationship between MIC and fitness as well as the dependency of the MIC-fitness relationship on drug concentration. To assess the reproducibility of our observations, we conducted a replicate experiment over a shorter period of five days. Except for AMK, which exhibited relatively small absolute values of MIC doubling rates, the results demonstrated clear correlations between the adaptation speeds in the 9-day and 5-day experiments (S4 Fig in S1 File). This emphasizes the reproducibility of our adaptation speed estimates.

Fig 3 illustrates the relationship between the mutation rate and the rate of adaptation for each antibiotic. In this analysis, the mutation rate was calculated based on the frequency of BPSs. As observed, the rate of evolutionary adaptation generally increased with mutation rate, except for the LQ strain, which had the highest mutation rate and often exhibited significantly slower adaptation speeds. A similar pattern is also observed when the mutation rate is calculated using the combined frequency of BPSs and indels (S5 Fig in S1 File).

Fig 3. The relationship between mutation rate and MIC doubling rate.

Fig 3

The dots represent experimental observations of 13 strains and four replicate serial transfer cultures. To prevent overlap of data points, small Gaussian noise (mean = 0, standard deviation = 0.05) was added to the y-coordinates.

The dependency on mutation rate varied with the antibiotics used for selection. For CP and TP selections, significant decreases in MIC doubling rates were observed in the LQ strain compared to the other five strains with mutation rates close to 0.1 (p<0.01; Mann-Whitney U-test). Conversely, no significant decreases were observed for the other three drugs. CP and TP are known as bacteriostatic drugs, while AMK, CFIX, and CPFX are classified as bactericidal drugs. Thus, our results suggest that the effect of mutation rate on antibiotic resistance evolution depends on the mechanisms of action of the drugs used for selection.

Analysis of mutations fixed in evolved E. coli populations.

To investigate whether the profiles of mutated genes vary depending on mutation spectra, we sequenced the genomes of evolved resistant populations. Specifically, we analyzed 24 populations after nine days of antibiotic selection, encompassing three mutator strains (L, Q, and T), two antibiotics (CP and CFIX), and four independently evolved populations for each condition. The fixed mutations identified in these populations are summarized in S7 Table.

Our analysis revealed that the mutation spectra of these fixed mutations were consistent with those observed in mutation accumulation experiments (S6 Fig in S1 File). For example, A:T to C:G mutations were significantly enriched in the mutT deletion strain, aligning with its known mutational signature.

Furthermore, this analysis demonstrated that mutations commonly associated with antibiotic resistance evolution are also prevalent in mutator populations. Under CFIX selection, mutations in ampC, which encodes β-lactamase, and ompR, which is involved in the regulation of porin protein, were detected in multiple evolved populations. These mutations are well-documented for their role in conferring resistance to β-lactams, including CFIX [36,37].

Similarly, in populations evolved under CP selection, mutations were frequently identified in acrB, which encodes a multidrug efflux pump, as well as in acrR, marR, and soxR, which regulate its expression. These mutations are also known to contribute to resistance against a broad spectrum of antimicrobial agents [38].

Population dynamics model explains the relationship between mutation rate and speed of adaptation.

To elucidate the mutation rate dependency of adaptation speed, we employed a simple population dynamics model of multistep resistance evolution that incorporates pharmacodynamic modeling [39,40]. In this model, E. coli populations can grow and increase their cell number up to a certain carrying capacity, while accumulating mutations during their replication at a given mutation rate. Specifically, we considered the sequential accumulation of mutations that confer additive beneficial effects on growth in the presence of antibiotics (Fig 4A). Population dynamics are described by the following deterministic differential equations:

Fig 4. Parameter estimation by a population dynamics model.

Fig 4

(A) Schematic representation of the model with multi-step resistance evolution. The mutated subpopulations (Mi) with i mutations proliferate with the growth rate μ, mutate with the rate pb, and die at the death rate γi. The death rate depends on the benefit of mutation bi conferred per mutation, as represented in Eq. (3). (B) The effect of antibiotics on the death rate. For the cases of β=0.5 and β=2, the death rates are plotted as a function of drug concentration, where b0 is set to 1. The death rate curve shifts towards the right due to selection, as depicted by the right green lines in the case of β=2. (C) The simulated relationship between mutation rate and MIC doubling rate. The population dynamics model with mutation accumulation was utilized to simulate the increase in MIC observed in serial transfer experiments for the cases of β=0.5 and β=2. The parameters were set to K=0.2, b0=0.5, and ϵ=0.5, respectively. (D) The estimated beta value for each drug. The mean value of β estimated through 100 bootstrap resampling is represented by the bars, while the error bars represent the standard error.

dMidt=μpbMi−11−MK+μ(1−pb)Mi1−MK−γiMi (1)

with M=∑iMi. Here, Mi represents the population size of E. coli cells that have accumulated i beneficial mutations, μ is the growth rate, γi is the death rate with i mutations, and K is the carrying capacity, respectively. The variables and parameters used in this model are summarized in Table 2. The probability of acquiring a beneficial mutation, pb, is assumed to be proportional to the mutation rate, p, such that pb=αp, where α is the ratio of beneficial mutations to total mutations. The first term on the right side of Eq. (1) represents the influx of population caused by a single beneficial mutation. For simplicity, we do not allow reverse mutations that decrease the number of beneficial mutations. Note that the expected number of mutations during the 9-day evolution experiment with the LQ strain, which had the highest mutation rate among the mutator strains we used, was ~70. Therefore, the chance of reverse mutation, that is, positional overlapping of these ~70 mutations on the genome size of 4Mbp, was very low.

Table 2. Summary of variables and parameters used in the population dynamics model.
Symbol Description
M i population size of cells having i beneficial mutations
M total population size, M=∑iMi
μ growth rate of cells, μ=μ0exp−δp
γi death rate of cells with i mutations
K carrying capacity of environment
p mutation rate
α ratio of beneficial mutations to total mutations
pb beneficial mutation rate, pb=αp
μ0 maximum growth rate
δ coefficient for deleterious effect of mutations on cellular growth
x log2-transformed concentration of antibiotics
β sensitivity of the death rate to the addition of antibiotics
bi resistance of cells having i beneficial mutations
b0 resistance of non-mutated cells
ϵ benefit of a single cell on resistance

By measuring the growth rate of the constructed mutant strains, we observed that the growth rate μ decreased as the mutation rate p increased, as shown in S2B Fig in S1 File owing to the deleterious effects caused by the gene deletion. We fit the data in S2B Fig in S1 File with the following relationship:

μ=μ0exp−δp (2)

and obtained μ0=0.90 per hour and δ=0.54, which represents the maximum growth rate and the coefficient for deleterious effect of mutations on cellular growth. Note that the observed reduction in growth rate may not be fully explained by high mutation rates alone. Although the mechanism for the growth defects cannot be quantified by mutation fixation rates, for simplification, we have assumed that the contributions of these non-observable factors are proportional to the observed mutation rate. Under this assumption, Eq. (2) describes the relationship between population growth rate and mutation rate, incorporating both the effects of high mutation rates and non-observable factors. We then used these fitted parameters for the numerical simulations. Furthermore, for simplicity, we designed our simulation such that the evolutionary variable of a strain was only drug resistance. Therefore, the growth rates of the strains were maintained throughout the simulations.

In our model, the addition of antibiotics resulted in an increase in the cell death rate. However, the acquisition of beneficial mutations can enable cells to overcome the effects of drugs. We assumed that the combined effect of antibiotics and beneficial mutations can be represented by the following death rate:

γi=μx/biβx/biβ+1 (3)

where x is the log2-transformed concentration of antibiotics and the parameter β represents the sensitivity of the death rate to the addition of antibiotics (see Fig 4B). At sufficiently high drug concentrations, the death rate γi converges to the growth rate μ, and the cells are unable to proliferate further. We do not consider the case where μ<γi, as a subpopulation with μ<γi is rapidly diluted in our serial transfer culture and becomes negligible. bi represents the benefit resulting from the accumulation of i mutations, which corresponds to the log2-transformed drug concentration that reduces the growth rate of the subpopulation by half (close to the MIC in our experimental setting). In bacterial experimental evolution under antibiotics, the MIC often increases exponentially (see S3 Fig in S1 File and [41,42]). To model the exponential increase of MIC by mutations, we simply assume the following additive effect of mutations on the benefit: bi=b0+iϵ, where b0 and ϵ are constant parameters representing the resistance of non-mutated cells and the benefit of a single mutation, respectively. Note that the unit of benefit bi is the log-transformed concentration; thus, the above additive effect of the mutations represents an exponential increase in the MIC. We simplified our model by neglecting the individuality of mutations and assuming that each mutation has the same beneficial effect and cost. Additionally, we neglected any positive or negative epistasis between the mutations.

We utilized a population dynamics model to simulate the adaptive evolution of antibiotic resistance in our serial transfer setup (Fig 2). In the initial condition, non-mutated cells were inoculated into a series of 2-fold dilutions of antibiotics with a constant initial density denoted as M0= 3×10-4 and Mi= 0 for i≠ 0, in units of OD620 which corresponds to approximately 2×105 cells in our experimental settings. After 24 h of cultivation, cells were sampled from an environment with the highest drug concentration, in which the total cell density exceeded a certain threshold (M=∑iMi> 0.03), diluted, and subsequently transferred to a fresh environment for the next round of cultivation. This simulation procedure corresponds to our experimental design (Fig 2A) and was iterated nine times, resulting in an increase in minimum inhibitory concentration (MIC) similar to that observed in Fig 2B.

Fig 4C presents typical simulation results for the MIC doubling rate as a function of the mutation rate. As shown in the figure, the curve is significantly influenced by β, which represents sensitivity to antibiotics. When β is equal to 0.5, the MIC increases significantly with an increase in the mutation rate before experiencing a drastic decrease when the mutation rates become extremely high. Conversely, when the sensitivity was high, as indicated by the steeper kill curve in Fig 4B, the change in the MIC with increasing mutation rates was smaller. This difference can be explained by the variation in effective population sizes between the two conditions. In the case of death rate with low steepness, cells with fewer beneficial mutations have a greater chance of growth, and these cells can also acquire further beneficial mutations. Therefore, the speed of adaptation can increase substantially with an increase in the mutation rate. In contrast, when the sensitivity curve is steep, only a small fraction of cells with beneficial mutations can grow, and MIC improvement is primarily governed by the time required for the mutated cells to take over the population. Hence, in this case, the effect of the increased mutation rate on the speed of MIC improvement was relatively small.

We used this population dynamics model to fit the experimental data and to elucidate the evolutionary dynamics presented in Fig 5. The experimentally estimated values of K, μ0, and δ were used as fixed parameters, while ϵ, b0 and β were used as fitting parameters to minimize the residual sum of squares between experimental and simulated adaptation speeds for each antibiotic. Because of the difficulty in experimentally estimating parameter α, which represents the ratio of beneficial mutations, we assigned an arbitrary value and confirmed that the simulated results remained qualitatively unchanged within a certain range. The green solid lines in Fig 5 represent the average of the simulated results under the fitted parameters, with the light green range indicating the standard error estimated by bootstrap random resampling of the experimental data.

Fig 5. The fitting results of the population dynamics model to experimental observations.

Fig 5

The blue dots represent experimental observations shown in Fig 3 and the green solid line shows the mean of simulated MIC changes generated with the parameters estimated from 100 bootstrap resampling, while the light green region represents the estimated standard error.

Our analysis revealed an interesting finding that the estimated sensitivity β significantly differed among antibiotics (Fig 4D), with relatively low values for CP and TP and high values for the other three antibiotics. This result is consistent with the classification of bacteriostatic and bactericidal antibiotics, where the former is expected to have shallow killing curves. However, interpreting the estimated beneficial effect of each mutation represented by ϵ (S7 Fig in S1 File) remains difficult and may be related to the physical characteristics of the drugs. For instance, drugs with high estimated ϵ values typically have smaller molecular weights (e.g., the molecular weights of CP, TP, and CPFX are approximately 300), whereas those with lower values tend to have larger molecular weights (e.g., the molecular weight of AMK is 586 and that of CFIX is 453). Regarding the observed correlation between ϵ values and the molecular weight of antibiotics, we propose the following hypothesis: One of the main mechanisms for acquiring antibiotic resistance in this experimental setting involves the activation of efflux pumps. The activity of these pumps may be influenced by the molecular weight of the drugs, implying that mutation effects might similarly depend on molecular weight. Currently, there is no experimental data to substantiate this relationship; however, we believe these parameter estimates offer a novel perspective for further analysis.

In our model simulations, with β=0.5, as observed with drugs such as CP and TP, doubling the drug concentration was predicted to have a minimal impact on the population growth rate, suggesting that most cells retain their ability to proliferate (S8 Fig in S1 File). Conversely, at β=4, which is a characteristic of bactericidal drugs, a doubled drug concentration significantly decreased the growth rate, indicative of a decrease in population size. Such variations in population dynamics in response to changes in drug concentration may influence the observed mutation-rate dependency on the speed of adaptation.

Discussion

In this study, we conducted quantitative analysis to investigate how changes in mutation rates affect the evolution of antibiotic resistance in E. coli. To achieve this goal, we generated a library of E. coli mutator strains using single and double deletions of genes related to DNA replication and error correction. We quantified the mutation rates of these mutators using MA experiments, which resulted in a range of mutation rates caused by gene deletions. We then performed evolution experiments starting from these mutators by serial transfer culture under five antibiotics and quantified the rate of MIC increase over time. The results demonstrated that, for nearly all mutator strains, the rate of adaptation positively correlated with mutation rate. However, an exception was observed in the LQ strain, which exhibited the highest mutation rate but frequently showed a significant reduction in the adaptation rate (Fig 3). This finding highlights that the relationship between mutation rate and evolutionary adaptation rate is potentially non-monotonic under certain conditions.

The observed reduction in the adaptation rate of the LQ strain is associated with its growth defect (S2B Fig in S1 File). However, the underlying mechanism responsible for this growth defect remains unclear. One possible explanation is that the simultaneous deletion of the mutL and dnaQ genes exerts a synthetic effect, thereby impairing growth. Alternatively, this study posits a model in which the accumulation of deleterious mutations and other detrimental effects, proportional to the mutation rate, ultimately leads to a reduction in growth rate. Both hypotheses are equally plausible and provide potential explanations for the observed decrease in the evolutionary rate due to growth impairment. However, no experimental data are currently available to distinguish between these two possibilities.

Nevertheless, our population dynamics model, which assumes the accumulation of deleterious mutations, provides a general framework for describing the relationship between growth inhibition and evolutionary rate. This approach may offer a suitable method for characterizing the interplay between mutation rate and evolutionary adaptation rate. Furthermore, previous studies investigating the relationship between mutation rate and growth rate across various mutator strains and culture conditions have consistently demonstrated that an elevated mutation rate frequently results in a decline in growth rate [34]. These observations suggest that a spectrum of deleterious mutations may collectively contribute to the observed growth defects.

Our results indicate that the relationship between the speed of adaptation and increasing mutation rates was dependent on the selective agent. When using bacteriostatic drugs, a modest mutation rate led to a significant increase in the speed of adaptation, whereas bactericidal drugs exhibited less dependency of adaptation speed on the mutation rate. This difference provides novel insights into the conditions for the emergence of resistant strains. In cases where a two-fold change in antibiotic concentration used in the evolution experiment has a relatively small effect on lethality, that is, if it has a lower β coefficient in our killing curve model (Eq. (3)), a greater number of cells have survived after the drug increase and obtained the opportunity to acquire beneficial mutations, resulting in stronger selection favoring high mutation rates under the drug treatment. Conversely, if sensitivity to a drug shows a steep response, only a small fraction of cells survive and have a chance to acquire beneficial mutations. In this case, with a smaller effective population size, the benefit of increasing the mutation rate for evolution of resistance is limited. In both cases, with extremely high mutation rates, such that each individual in a population obtains one or more mutations, selection cannot remove deleterious mutations regardless of the population size. This can be a possible scenario that the LQ strain showed little or no advantage in MIC improvement. While the relationship between the mutation rate and the speed of adaptive evolution has been previously explored [5,13], these analyses were conducted under a single selection environment. In contrast, the current study employed various drugs with distinct action mechanisms, implying that the effective population size under the influence of drug selection may impact the association between mutation rate and adaptation speed. Notably, our results showed that the speed of adaptation to amikacin addition did not increase significantly with higher mutation rates. This result may shed light on the mechanism responsible for the infrequent emergence of amikacin-resistant strains [43].

There may be alternative explanations for the observed differences in adaptive evolution under bacteriostatic and bactericidal drugs. For example, the size or nature of the mutational targets leading to resistance acquisition may differ among these drugs, as discussed in previous studies [44,45]. Although our previous research demonstrated that the number of fixed mutations did not significantly differ among evolved E. coli strains under these drugs [42], a more detailed analysis of the fixed mutations from the evolved strains obtained in this study could provide valuable insights into the differences in the evolution of resistance to bacteriostatic and bactericidal agents.

Recently, Lukačišinová et al. conducted drug-resistance evolution experiments in E. coli using an automated platform to quantify the evolvability of 98 strains with gene deletions [46]. Their study demonstrated epistasis between resistance mutations and the genetic background by identifying gene deletion strains that affect the evolution of drug resistance. The gene deletion library they used included mutL and mutT, which showed a moderately high mutation rate in our MA experiment. Additionally, they used Tetracycline and Chloramphenicol, which are bacteriostatic drugs. Their experiment also showed that deletion of mutL and mutT generally accelerated the evolution of resistance to these drugs, supporting our conclusion. Note that they selected gene deletions by ensuring that they did not bring significant growth reduction to the cells in drug-free media. This gene selection contrasts with our research which focused on the impact of higher mutation rates alongside growth reduction on evolvability. Thus, their study elaborates on the effect of gene deletions mainly involved in efflux pumps on drug resistance and resistance evolution, as well as the types of mutations that occur in the evolved strain. On the other hand, our numerical approach was able to quantify various factors, such as mutation rate, growth rate, and features of the selective environment, and to predict the effect of changes in the mutation rate on the evolution of resistance for any known selective agent.

As shown in Fig 1, the results of MA experiment indicate that A:T→C:G mutations become predominant due to the deletion of the mutT gene. Such changes in the mutation spectrum could potentially affect the rate of evolution. Indeed, the evolved strains with a relatively high evolutionary rate, approximately a mutation rate of 0.1, are those lacking the mutT gene, as demonstrated in Fig 3. The frequent occurrence of A:T→C:G mutations in these strains may contribute to their high evolutionary speeds. To explore this possibility, we analyzed fixed mutations in antibiotic-resistant E. coli strains obtained in a previous study of experimental evolution [42]. We hypothesized that if A:T→C:G mutations significantly contribute to resistance acquisition, an enrichment of these mutations in antibiotic-resistant strains would also be expected in the evolutionary dynamics of non-mutator strains. However, as shown in S8 Table, A:T→C:G mutations are not consistently enriched; instead, the distribution of mutations aligns with the base-pair substitution profile of wild-type E. coli [26]. This suggests that A:T→C:G mutations in mutT deletion strains do not necessarily increase the likelihood of resistance enhancement or accelerate the evolution of drug resistance.

The analysis of the mutational spectrum also revealed that the ratio of indels to BPSs varied depending on the knockout gene(s). As shown in Fig 1, when the LQ strain is excluded as an outlier, mutator strains with a high fixation rate of BPSs do not necessarily exhibit a correspondingly high frequency of indels. Given that indels often have a more pronounced impact on phenotypic changes compared to BPSs and are more likely to result in lethal mutations, substantial variation in the ratio of indels to BPSs across different mutation rates may influence the evolutionary dynamics of distinct mutator strains. Given the limited amount of data and the challenges associated with conducting statistical analyses, it is difficult to draw definitive conclusions from this study. Future research with sufficient data should be conducted to further explore the relationship between the mutational spectrum in mutators and the evolution of drug resistance.

The mutL deletion strains exhibited lower growth rates compared to mutator strains with similar mutation rates (S2B Fig in S1 File), suggesting that the growth reduction in these strains cannot be solely attributed to mutation rate. The underlying mechanism of the slow growth observed in mutL deletion strains remains unclear. The mechanism underlying the slow growth observed in mutL deletion strains remains unclear. One possible explanation involves an imbalance between the functions of the MutL and MutS proteins. As reported [47], MutL and MutS coordinately contribute to mismatch repair; MutS binds specifically to mismatched DNA, while MutL helps to release the bound MutS. This suggests that maintaining a proper balance between MutS and MutL is critical for efficient mismatch repair. The deletion of the mutL gene may disrupt this balance, leading to an excess of unbound MutS, which could contribute to the observed growth defect.

In the experimental evolution under antibiotics, there exists a possibility that selection may favor reversing growth defects in ancestral strains rather than enhancing antibiotic resistance [46]. To investigate this in our experimental evolution with mutator strains, we measured the growth rates of three strains (WT, S, and LQ) both before and after experimental evolution. As presented in S9 Fig in S1 File, the growth rates remained largely unchanged. This finding strongly suggests that the observed increases in MIC during our experiments were not attributable to recovery in growth rates but were likely driven by adaptations related to antibiotic resistance.

The experiments and simulations presented in this study encompass numerous limitations and unvalidated hypotheses. For instance, in our population dynamics model, we assumed bi=b0+iϵ, which means that the log-transformed MIC linearly increases with the number of beneficial mutations. This assumption is grounded in observations from our nine-day experimental evolution shown in S3 Fig in S1 File, where the log-transformed MIC increased approximately linearly over time across many culture lines. However, for some lineages, the log-transformed MIC shows a saturation curve, indicating that mutations fixed later have a diminishing effect. This trend has been observed in several experimental evolution studies, including Lenski’s long-term experimental evolution [48]. To account for this, we have modified our model to include a saturation curve for the increase in fitness due to mutations, as bi=b0+iϵS/i+S. Here, S is a parameter that influences the shape of the saturation curve. When S is small, the increase in log-transformed MIC saturates with only a few mutations, whereas a larger S results in an almost linear increase, akin to the original model. Using this modified model, we estimated other parameters using a genetic algorithm. S10 Fig in S1 File illustrates how the average residuals between the model predictions and observed data points change when parameter S is fixed. As depicted, the residuals decrease as S increases. This result supports the appropriateness of using the assumption in the original model, which corresponds to the case of large S.

In addition to the examples mentioned above, there are various other assumptions that require verification. For instance, we assume that the mutation rate remains constant throughout the experimental evolution, although this assumption lacks empirical verification. Previous studies have demonstrated that mutations that alter the mutation rate can be fixed during mutation accumulation experiments [49]. For other examples, the functional forms of equations (1) and (2) assumed in the simulation, the constancy of ϵ, and the consideration of only the death rate when using a common growth rate should be subjected to future experimental validation.

The evolution of mutation rates and the effects of new mutations on reproductive success have been well studied [2,50] while pathogen/drug resistance has been used merely as a marker trait to estimate mutation rates using the fluctuation test [51,52]. Our study builds on previous observations of the relationship between the mutation rate and the speed of adaptation by quantifying specific features of a selective environment that controls the shape of this relationship. Further experimental and theoretical investigations would help to better understand not only the evolution of drug-resistant microbes but also other evolutionary challenges, such as the emergence of drug-resistant cancer cells [53].

Methods

Bacterial strains and media

The insertion sequence (IS)-free E. coli K-12 substrain MDS42 [54], purchased from Scarab Genomics (Madison, Wisconsin, USA), was used as the wild-type (WT) strain in this study. Using this WT strain, we constructed 12 knockout strains of genes related to the suppression of mutations (mutS, mutH, mutL, mutT, dnaQ), as listed in Table 1. Single- and double-gene knockouts were performed using the λ-Red homologous recombination method [55]. In this study, we utilized an IS-free strain to ensure reliable resequencing analysis, as determining the precise positions of IS element insertions can be challenging when using short-read sequencing technologies. E. coli cells were cultured in modified M9 minimal medium supplemented with 20 amino acids (M9+AA medium) which includes 17.1 g/L Na2HPO4·12H2O, 3.0 g/L KH2PO4, 0.5 g/L NaCl, 2.0 g/L NH4Cl, 5.0 g/L glucose, 14.7 mg/L CaCl2·2H2O, 123.0 mg/L MgSO4·7H2O, 2.8 mg/L FeSO4·7H2O, and 10.0 mg/L thiamine hydrochloride (pH 7.0), and 20 canonical amino acids (0.02 mM Tyr and 0.05 mM of the remaining 19 amino acids: Gly, Ala, Asn, Gln, Leu, Ile, Met, Ser, Pro, Thr, Val, Phe, Trp, Asp, Glu, Arg, His, Lys and Cys) [34]. In this study, to enhance the reproducibility of cultures and facilitate future analysis of medium components during resistance evolution, we opted for a synthetic medium with defined components rather than a rich medium like LB, which contains unspecified components. Moreover, to improve the robustness of E. coli against mutations [34], we employed a minimal medium supplemented with amino acids.

Measurement of growth rate

To measure the growth rate, cells precultured to yield an initial OD620 of 0.001 were inoculated into a 96-well microtiter plate containing 200 μL M9+AA medium. The cells were then cultured at 34 °C with shaking at 300 rpm, and the cell density was measured at OD620 using a FilterMax F3 microplate reader (Molecular Devices) at 30-minute intervals. The specific growth rate was calculated as the slope of the exponential growth curve determined by data points with OD620 values under the condition 0.005 <OD620 < 0.03, according to the standard Malthusian growth model.

Mutation accumulation (MA) experiment

E. coli cells were subcultured as single colonies on M9+AA agar medium and incubated at 34 °C. A colony was randomly selected and streaked onto fresh agar medium every two days for a total of 23 or 61-69 passages. The number of generations per passage was estimated based on colony diameter, which has been shown to be a reliable indicator [26–28]. Specifically, we used a regression curve to estimate colony-forming units from the diameter and then calculated the number of generations based on the assumption that all cells in a colony were derived from a single cell from the previous passage. The adopted relationship used in this study was (number of generations) = 24.51 + 2 log2 (colony diameter in mm), which is a reuse of that experimentally obtained in our previous study [28]. After each passage, we cultured the remaining cells of the selected colony in M9+AA liquid medium and stored them in 15% glycerol. We kept three MA lines for each ancestral strain, resulting in 39 MA lines. S3 Table shows the colony size for each round recorded during the MA experiment.

Whole-genome resequencing

Glycerol-stocked cells were grown in M9+AA medium until they reached full growth, and were then harvested as a cell pellet by centrifugation. Genomic DNA was extracted from cells using a DNeasy Blood & Tissue Kit (Qiagen). For library preparation prior to sequencing, a Nextera XT kit (Illumina) was used in a paired-end (2×300 bp) setting. The Illumina MiSeq platform was used to sequence the libraries using the MiSeq Reagent Kit v3, which provided 600 cycles (Illumina). The obtained short reads were mapped to the reference genome (GenBank accession: AP012306.1), and point mutations accumulated during the MA experiment were detected using Breseq [56] (Ver.0.30.1) with bowtie2 (Ver.2.1.0) and R (Ver.3.2.4). The known common mutations in our ancestral wild-type MDS42 cells (S4 Table) were omitted from the detected mutations in this study. Also, the mutations in the genome of each mutator ancestor cell in the MA experiment was omitted from the detected mutations in the post-MA cell lineages.

Calculation of mutation rate and dN/dS

We calculated the base-pair substitution (BPS) rate as the mutation rate using the numbers and spectra of synonymous mutations accumulated during MA and codon-usage information on the ancestral genome, following a previous publication [29]. First, we classified the accumulated synonymous substitutions into each of the six possible types, and divided their numbers by the number of sites in the genome where each type of substitution is synonymous, resulting in the number of substitutions per site, excluding the selective effect. Then, we multiplied the number of synonymous, nonsynonymous, and non-coding BPS sites by the number of substitutions per site and summed the results for the six types of substitutions. Finally, we divided this by the number of MA generations to obtain the genome-wide base substitution rates. To obtain dN/dS, we first calculated dN by dividing the number of nonsynonymous substitutions by the number of nonsynonymous sites, which was normalized considering the spectrum of BPS in the same manner as calculating the BPS rate. Similarly, we calculated dS by dividing the number of synonymous substitutions by the number of synonymous sites, which was normalized considering the spectrum of BPS.

Experimental evolution under antibiotics

The cells were cultured in 200 µL of M9+AA liquid medium in a 96-well microtiter plate. During the experiment, each culture line was exposed to 12 concentrations of antibiotics, corresponding to 11 wells with a two-fold dilution series, and one drug-free well. For CP, the well with the highest drug concentration was 600 ng/µl. Similarly, 200 ng/µL TP, 400 ng/µL AMK, 40 ng/µL CFIX, and 4 ng/µL CPFX. Cell concentration was quantified by measuring the optical density at 620 nm (OD620). Cells were inoculated into these 12 wells at an initial OD620 of 3×10-4. After 24 h of cultivation at 34 °C with shaking at 300 rpm, cells were sampled from the well with the highest drug concentration among the wells that showed an OD620 value greater than 0.03. Cultivation at 34°C was chosen to reduce the growth rate compared to the standard condition of 37°C, thereby facilitating subculturing during the exponential growth phase in our experimental settings. The sampled cells were diluted to an initial OD620 value of 3×10-4 and inoculated into fresh medium in 12 wells of a new plate, which was then incubated. Four independent culture lines were maintained in parallel for each combination of strain and antibiotic. A serial transfer experiment was conducted for nine days. After that we stored the evolved samples in 15% glycerol at -80°C. Some of the samples were then subjected to whole-genome resequencing analysis in the same manner as the samples from the MA experiment. The described experimental operations were performed using an automated culture system comprising a Biomek NXP laboratory automation workstation (Beckman Coulter) in a clean booth, STX44 automated shaker incubator (LiCONiC), LPX220 plate hotel (LiCONiC), and a FilterMax F3 microplate reader (Molecular Devices) [57].

Parameter estimation

Custom C programs were used to implement the population dynamics model described in equation (1) and optimize the model parameters. Specifically, a genetic algorithm was employed to estimate the values of the parameters ϵ, b0 and β. The fitness function for the genetic algorithm is defined as the Euclidean distance between the experimentally observed MIC doubling rate and the rate obtained from simulations of t population dynamics model. The population size for the genetic algorithm was set to 100, and the parameter sets with the top 5% highest fitness were selected. each generation, a random change of 0.5% was applied to each parameter. The mean and standard errors of the parameters were estimated using bootstrap resampling of the experimental data, wherein the same number of experimental data points was resampled with allowed duplication.

Supporting information

S1 File

S1 Fig. Mutator Construction Procedure. Each box represents either a mutator strain or the wild-type strain. Arrows depict the lineage from parent to offspring in the construction of mutators. The list of mutations on the right details those shared between each parent-child pair (refer to S1 Table for the complete list of mutations). The values indicated in each strain box represent the growth rates (1/h; mean ± SD). S2 Fig. (A) Neutrality in mutation accumulation. The dN/dS ratio was calculated for each mutator strain and the wild-type strain. Error bars show the standard deviations between the MA lineages. (B) Relationship between growth rate and mutation rate. The horizontal error bars show the standard deviation across MA lineages, whereas the vertical error bars represent the standard deviation among replicate experiments in growth rate measurement. The sample sizes in the growth rate measurements were n=20 for wild-type strains and n=10 for mutant strains. The color and fill pattern of the markers correspond to those in Fig 3. S3 Fig. Experimental Evolution of Mutator Strains Under Antibiotics. For all combinations of drugs and strains, the changes in MIC over time are plotted. Each plot overlays data from four replicate series. Dashed lines represent the minimum and maximum MIC values attainable within the constraints of our experimental setup. S4 Fig. Reproducibility of the adaptation speed quantification. The MIC doubling rates were estimated by conducting independent experimental evolution trials with varying duration (9 days and 5 days, respectively). Each dot and error bar show the mean and standard deviation of MIC fold change per day, calculated from the data in S3 Fig. The black solid diagonal line means y=x identity line, while the blue line shows linear regression without intercept. The linear regression coefficient and corresponding R2 value were computed for these data points (N=13). S5 Fig. The relationship between the mutation rate, calculated by the sum of fixed BPSs and indels, and the MIC doubling rate is shown. Each dot represents an experimental observation from 13 strains across four replicate serial transfer cultures. To prevent overlap of data points, small Gaussian noise (mean = 0, standard deviation = 0.05) was added to the y-coordinates. S6 Fig. Mutation Spectra in Resistant Strains. The distribution of substitution patterns and indels calculated by 24 resistant strains (three ancestor mutators, two antibiotics, and four replicates) is presented. Each dot represents the mutation fixation rate during the experimental evolution calculated based on BPSs (closed circle) or indels (open circle). S7 Fig. Estimated parameter ϵ representing the beneficial effect of each mutation. The mean value estimated through 100 bootstrap resampling is represented by the bars, while the error bars represent the standard error. S8 Fig. A doubled drug concentration with large β remarkably reduces the growth rate. We defined the net growth rate as μ-γ where μ is the drug-free growth rate and γ is the death rate by drug treatment. μx is the net growth rate with drug concentration x, μ2x is the net growth rate with drug concentration 2x, and μmut2x is the net growth rate with drug concentration 2x for cells with one mutation of beneficial effect ϵ. We assume here a situation where the growth rate and dilution rate are balancing at the drug concentration x, i.e., μx=ln100/24≅0.19h−1. The black dashed line represents μx=0.19. We used ϵ=0.9 and bi=1 for these simulations. S9 Fig. Changes in growth rate during resistance evolution. Growth rates for three strains (WT, S, and LQ) were quantified both before and after experimental evolution. For each selection drug, two independently evolved strains were analyzed. The bars represent the standard deviation of three replicate measurements. S10 Fig. Relationship between the parameter S and the residual of fitting. The parameter S, which describes the saturation effect of mutation accumulation in the form bi=b0+iϵS/i+S, is plotted against the residual of the fitting. The modified model incorporating the saturation effect was used to fit the experimental data shown in Fig 3. As observed, the residuals decrease with increasing S, suggesting that the saturation effect is negligible.

(DOCX)

pgen.1011627.s001.docx (1.4MB, docx)
S1 Table. List of mutations identified in mutation accumulation (MA) experiments.

(XLSX)

pgen.1011627.s002.xlsx (927.8KB, xlsx)
S2 Table. Extended summary of MA experiments.

(XLSX)

pgen.1011627.s003.xlsx (11.1KB, xlsx)
S3 Table. Colony-size records during MA experiments.

(XLSX)

pgen.1011627.s004.xlsx (22.6KB, xlsx)
S4 Table. Common mutations in ancestors which were omitted from the detected mutations in this study.

(XLSX)

pgen.1011627.s005.xlsx (11.1KB, xlsx)
S5 Table. Number of generations during the 9-day experimental evolution.

(XLSX)

pgen.1011627.s006.xlsx (13.9KB, xlsx)
S6 Table. Number of common mutations among ancestral strains.

(XLSX)

pgen.1011627.s007.xlsx (10.7KB, xlsx)
S7 Table. List of mutations in drug-resistant evolved strains.

(XLSX)

pgen.1011627.s008.xlsx (72.9KB, xlsx)
S8 Table. Mutational spectrum of wild-type E. coli during drug resistance evolution experiments and drug-free MA experiment in previous studies.

(XLSX)

pgen.1011627.s009.xlsx (14.8KB, xlsx)

Acknowledgments

We thank Dr. Saburo Tsuru for the fruitful discussions.

Data Availability

The resequencing analysis data have been deposited at the DDBJ Sequence Read Archive (https://www.ddbj.nig.ac.jp/dra/index-e.html) under accession number DRA016357 and DRA019544.

Funding Statement

This work was supported by Japan Society for the Promotion of Science (JSPS; https://www.jsps.go.jp/english/) KAKENHI Grant Numbers 22K21344, 23K27164, 24H01798 (to C. F.), 19K16114 and 21K15077 (to A. S.), and Japan Science and Technology Agency (JST; https://www.jst.go.jp/EN/) ERATO Grant Number JPMJER1902 (to C. F.). The funder had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

  • 1.Wielgoss S, Barrick JE, Tenaillon O, Wiser MJ, Dittmar WJ, Cruveiller S, et al. Mutation rate dynamics in a bacterial population reflect tension between adaptation and genetic load. Proc Natl Acad Sci U S A. 2013;110(1):222–7. doi: 10.1073/pnas.1219574110 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Lynch M, Ackerman MS, Gout J-F, Long H, Sung W, Thomas WK, et al. Genetic drift, selection and the evolution of the mutation rate. Nat Rev Genet. 2016;17(11):704–14. doi: 10.1038/nrg.2016.104 [DOI] [PubMed] [Google Scholar]
  • 3.Raynes Y, Sniegowski PD. Experimental evolution and the dynamics of genomic mutation rate modifiers. Heredity (Edinb). 2014;113(5):375–80. doi: 10.1038/hdy.2014.49 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Taddei F, Radman M, Maynard-Smith J, Toupance B, Gouyon PH, Godelle B. Role of mutator alleles in adaptive evolution. Nature. 1997;387(6634):700–2. doi: 10.1038/42696 [DOI] [PubMed] [Google Scholar]
  • 5.Arjan JA, Visser M, Zeyl CW, Gerrish PJ, Blanchard JL, Lenski RE. Diminishing returns from mutation supply rate in asexual populations. Science. 1999;283(5400):404–6. doi: 10.1126/science.283.5400.404 [DOI] [PubMed] [Google Scholar]
  • 6.Wiser M, Ribeck N, Lenski R. Long-term dynamics of adaptation in asexual populations. Science. 2013;342:1364–7. [DOI] [PubMed] [Google Scholar]
  • 7.Elgrail M, Sprouffske K, Dartey J, Garcia A. Emergence of a multilocus mutator genotype in mutator Escherichia coli experimental populations under repeated lethal selection. Journal of Evolutionary Biology. 2024;37:346–52. [DOI] [PubMed] [Google Scholar]
  • 8.Chao L, Cox EC. Competition between high and low mutating strains of Escherichia Coli. Evolution. 1983;37(1):125–34. doi: 10.1111/j.1558-5646.1983.tb05521.x [DOI] [PubMed] [Google Scholar]
  • 9.Desai MM, Fisher DS. The balance between mutators and nonmutators in asexual populations. Genetics. 2011;188(4):997–1014. doi: 10.1534/genetics.111.128116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Sniegowski PD, Gerrish PJ, Johnson T, Shaver A. The evolution of mutation rates: separating causes from consequences. Bioessays. 2000;22:1057–66. [DOI] [PubMed] [Google Scholar]
  • 11.McDonald MJ, Hsieh Y-Y, Yu Y-H, Chang S-L, Leu J-Y. The evolution of low mutation rates in experimental mutator populations of Saccharomyces cerevisiae. Curr Biol. 2012;22(13):1235–40. doi: 10.1016/j.cub.2012.04.056 [DOI] [PubMed] [Google Scholar]
  • 12.Swings T, Van den Bergh B, Wuyts S, Oeyen E, Voordeckers K, Verstrepen KJ, et al. Adaptive tuning of mutation rates allows fast response to lethal stress in Escherichia coli. Elife. 2017;6:e22939. doi: 10.7554/eLife.22939 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Sprouffske K, Aguilar-Rodríguez J, Sniegowski P, Wagner A. High mutation rates limit evolutionary adaptation in Escherichia coli. PLoS Genet. 2018;14(4):e1007324. doi: 10.1371/journal.pgen.1007324 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Peterson E, Kaur P. Antibiotic Resistance Mechanisms in Bacteria: Relationships Between Resistance Determinants of Antibiotic Producers, Environmental Bacteria, and Clinical Pathogens. Front Microbiol. 2018;9:2928. doi: 10.3389/fmicb.2018.02928 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Toprak E, Veres A, Michel J-B, Chait R, Hartl DL, Kishony R. Evolutionary paths to antibiotic resistance under dynamically sustained drug selection. Nat Genet. 2011;44(1):101–5. doi: 10.1038/ng.1034 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Imamovic L, Ellabaan MMH, Dantas Machado AM, Citterio L, Wulff T, Molin S, et al. Drug-Driven Phenotypic Convergence Supports Rational Treatment Strategies of Chronic Infections. Cell. 2018;172(1–2):121–134.e14. doi: 10.1016/j.cell.2017.12.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Andersson DI, Nicoloff H, Hjort K. Mechanisms and clinical relevance of bacterial heteroresistance. Nat Rev Microbiol. 2019;17(8):479–96. doi: 10.1038/s41579-019-0218-1 [DOI] [PubMed] [Google Scholar]
  • 18.Maeda T, Iwasawa J, Kotani H, Sakata N, Kawada M, Horinouchi T, et al. High-throughput laboratory evolution reveals evolutionary constraints in Escherichia coli. Nat Commun. 2020;11(1):5970. doi: 10.1038/s41467-020-19713-w [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Andersson DI, Hughes D. Antibiotic resistance and its cost: is it possible to reverse resistance?. Nat Rev Microbiol. 2010;8(4):260–71. doi: 10.1038/nrmicro2319 [DOI] [PubMed] [Google Scholar]
  • 20.Suzuki S, Horinouchi T, Furusawa C. Phenotypic changes associated with the fitness cost in antibiotic resistant Escherichia coli strains. Mol Biosyst. 2016;12(2):414–20. doi: 10.1039/c5mb00590f [DOI] [PubMed] [Google Scholar]
  • 21.Gifford DR, Berríos-Caro E, Joerres C, Suñé M, Forsyth JH, Bhattacharyya A, et al. Mutators can drive the evolution of multi-resistance to antibiotics. PLoS Genet. 2023;19(6):e1010791. doi: 10.1371/journal.pgen.1010791 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Sane M, Diwan GD, Bhat BA, Wahl LM, Agashe D. Shifts in mutation spectra enhance access to beneficial mutations. Proc Natl Acad Sci U S A. 2023;120(22):e2207355120. doi: 10.1073/pnas.2207355120 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Junop MS, Yang W, Funchain P, Clendenin W, Miller JH. In vitro and in vivo studies of MutS, MutL and MutH mutants: correlation of mismatch repair and DNA recombination. DNA Repair (Amst). 2003;2(4):387–405. doi: 10.1016/s1568-7864(02)00245-8 [DOI] [PubMed] [Google Scholar]
  • 24.Fowler RG, Schaaper RM. The role of the mutT gene of Escherichia coli in maintaining replication fidelity. FEMS Microbiol Rev. 1997;21(1):43–54. doi: 10.1111/j.1574-6976.1997.tb00344.x [DOI] [PubMed] [Google Scholar]
  • 25.Fijalkowska IJ, Schaaper RM. Mutants in the Exo I motif of Escherichia coli dnaQ: defective proofreading and inviability due to error catastrophe. Proc Natl Acad Sci U S A. 1996;93(7):2856–61. doi: 10.1073/pnas.93.7.2856 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Lee H, Popodi E, Tang H, Foster P. Rate and molecular spectrum of spontaneous mutations in the bacterium Escherichia coli as determined by whole-genome sequencing. Proceedings of the National Academy of Sciences of the United States of America. 2012;109:E2774–83. doi: 10.1073/pnas.1208824109 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Tsuru S, Ishizawa Y, Shibai A, Takahashi Y, Motooka D, Nakamura S, et al. Genomic confirmation of nutrient-dependent mutability of mutators in Escherichia coli. Genes Cells. 2015;20(12):972–81. doi: 10.1111/gtc.12300 [DOI] [PubMed] [Google Scholar]
  • 28.Maeda T, Shibai A, Yokoi N, Tarusawa Y, Kawada M, Kotani H, et al. Mutational property of newly identified mutagen l-glutamic acid γ-hydrazide in Escherichia coli. Mutat Res. 2021;823:111759. [DOI] [PubMed]
  • 29.Shibai A, Takahashi Y, Ishizawa Y, Motooka D, Nakamura S, Ying B-W, et al. Mutation accumulation under UV radiation in Escherichia coli. Sci Rep. 2017;7(1):14531. doi: 10.1038/s41598-017-15008-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Foster PL, Niccum BA, Popodi E, Townes JP, Lee H, MohammedIsmail W, et al. Determinants of Base-Pair Substitution Patterns Revealed by Whole-Genome Sequencing of DNA Mismatch Repair Defective Escherichia coli. Genetics. 2018;209(4):1029–42. doi: 10.1534/genetics.118.301237 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Foster PL, Lee H, Popodi E, Townes JP, Tang H. Determinants of spontaneous mutation in the bacterium Escherichia coli as revealed by whole-genome sequencing. Proc Natl Acad Sci U S A. 2015;112(44):E5990-9. doi: 10.1073/pnas.1512136112 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Niccum BA, Lee H, MohammedIsmail W, Tang H, Foster PL. The Spectrum of Replication Errors in the Absence of Error Correction Assayed Across the Whole Genome of Escherichia coli. Genetics. 2018;209(4):1043–54. doi: 10.1534/genetics.117.300515 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Garushyants SK, Sane M, Selifanova MV, Agashe D, Bazykin GA, Gelfand MS. Mutational Signatures in Wild Type Escherichia coli Strains Reveal Predominance of DNA Polymerase Errors. Genome Biol Evol. 2024;16(4):evae035. doi: 10.1093/gbe/evae035 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Ishizawa Y, Ying B-W, Tsuru S, Yomo T. Nutrient-dependent growth defects and mutability of mutators in Escherichia coli. Genes Cells. 2015;20(1):68–76. doi: 10.1111/gtc.12199 [DOI] [PubMed] [Google Scholar]
  • 35.Stefan A, Reggiani L, Cianchetta S, Radeghieri A, Gonzalez Vara y Rodriguez A, Hochkoeppler A. Silencing of the gene coding for the epsilon subunit of DNA polymerase III slows down the growth rate of Escherichia coli populations. FEBS Letters. 2003;546:295–9. [DOI] [PubMed] [Google Scholar]
  • 36.Jacoby GA. AmpC beta-lactamases. Clinical Microbiology Reviews. 2009;22(2):161–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Zhou G, Wang Q, Wang Y, Wen X, Peng H, Peng R, et al. Outer Membrane Porins Contribute to Antimicrobial Resistance in Gram-Negative Bacteria. Microorganisms. 2023;11(7):1690. doi: 10.3390/microorganisms11071690 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Blanco P, Hernando-Amado S, Reales-Calderon JA, Corona F, Lira F, Alcalde-Rico M, et al. Bacterial Multidrug Efflux Pumps: Much More Than Antibiotic Resistance Determinants. Microorganisms. 2016;4(1):14. doi: 10.3390/microorganisms4010014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Igler C, Rolff J, Regoes R. Multi-step vs. single-step resistance evolution under different drugs, pharmacokinetics, and treatment regimens. Elife. 2021;10:e64116. doi: 10.7554/eLife.64116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Yu G, Baeder DY, Regoes RR, Rolff J. Predicting drug resistance evolution: insights from antimicrobial peptides and antibiotics. Proc Biol Sci. 2018;285(1874):20172687. doi: 10.1098/rspb.2017.2687 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Marciano DC, Wang C, Hsu T-K, Bourquard T, Atri B, Nehring RB, et al. Evolutionary action of mutations reveals antimicrobial resistance genes in Escherichia coli. Nat Commun. 2022;13(1):3189. doi: 10.1038/s41467-022-30889-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Suzuki S, Horinouchi T, Furusawa C. Prediction of antibiotic resistance by gene expression profiles. Nat Commun. 2014;5:5792. doi: 10.1038/ncomms6792 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Nageeb WM, Hetta HF. The predictive potential of different molecular markers linked to amikacin susceptibility phenotypes in Pseudomonas aeruginosa. PLoS One. 2022;17(4):e0267396. doi: 10.1371/journal.pone.0267396 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Callens M, Rose CJ, Finnegan M, Gatchitch F, Simon L, Hamet J, et al. Hypermutator emergence in experimental Escherichia coli populations is stress-type dependent. Evol Lett. 2023;7(4):252–61. doi: 10.1093/evlett/qrad019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Couce A, Guelfo JR, Blázquez J. Mutational spectrum drives the rise of mutator bacteria. PLoS Genet. 2013;9(1):e1003167. doi: 10.1371/journal.pgen.1003167 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Lukačišinová M, Fernando B, Bollenbach T. Highly parallel lab evolution reveals that epistasis can curb the evolution of antibiotic resistance. Nat Commun. 2020;11(1):3105. doi: 10.1038/s41467-020-16932-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Acharya S, Foster P, Brooks P, Fishel R. The coordinated functions of the E. coli MutS and MutL proteins in mismatch repair. Molecular Cell. 2003;12233–46. [DOI] [PubMed] [Google Scholar]
  • 48.Barrick JE, Yu DS, Yoon SH, Jeong H, Oh TK, Schneider D, et al. Genome evolution and adaptation in a long-term experiment with Escherichia coli. Nature. 2009;461(7268):1243–7. doi: 10.1038/nature08480 [DOI] [PubMed] [Google Scholar]
  • 49.Ragheb MN, Thomason MK, Hsu C, Nugent P, Gage J, Samadpour AN, et al. Inhibiting the Evolution of Antibiotic Resistance. Mol Cell. 2019;73(1):157–165.e5. doi: 10.1016/j.molcel.2018.10.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Eyre-Walker A, Keightley PD. The distribution of fitness effects of new mutations. Nat Rev Genet. 2007;8(8):610–8. doi: 10.1038/nrg2146 [DOI] [PubMed] [Google Scholar]
  • 51.Luria SE, Delbrück M. Mutations of Bacteria from Virus Sensitivity to Virus Resistance. Genetics. 1943;28(6):491–511. doi: 10.1093/genetics/28.6.491 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Sniegowski P, Gerrish P, Lenski R. Evolution of high mutation rates in experimental populations of E. coli. Nature. 1997;387:703–5. [DOI] [PubMed] [Google Scholar]
  • 53.Mansoori B, Mohammadi A, Davudian S, Shirjang S, Baradaran B. The Different Mechanisms of Cancer Drug Resistance: A Brief Review. Adv Pharm Bull. 2017;7(3):339–48. doi: 10.15171/apb.2017.041 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Pósfai G, Plunkett G 3rd, Fehér T, Frisch D, Keil G, Umenhoffer K, et al. Emergent properties of reduced-genome Escherichia coli. Science. 2006;312. [DOI] [PubMed] [Google Scholar]
  • 55.Datsenko KA, Wanner BL. One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc Natl Acad Sci U S A. 2000;97(12):6640–5. doi: 10.1073/pnas.120163297 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Deatherage DE, Barrick JE. Identification of mutations in laboratory-evolved microbes from next-generation sequencing data using breseq. Methods Mol Biol. 2014;1151:165–88. doi: 10.1007/978-1-4939-0554-6_12 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Horinouchi T, Minamoto T, Suzuki S, Shimizu H, Furusawa C. Development of an automated culture system for laboratory evolution. J Lab Autom. 2014;19:478–82. [DOI] [PubMed] [Google Scholar]

Decision Letter 0

Megan Behringer

4 Sep 2024

Dear Dr Furusawa,

Thank you very much for submitting your Research Article entitled 'Quantitative Analysis of Saturating Relationship between Mutation Rate and Speed of Adaptation under Antibiotic Exposure in Escherichia coli' to PLOS Genetics.

The manuscript was fully evaluated at the editorial level and by independent peer reviewers. All three reviewers agreed that understanding how mutation rates influence antimicrobial resistance is an important topic and that the work presented is interesting. However, the reviewers noted several substantial concerns about the current manuscript. Based on the reviews, we will not be able to accept this version of the manuscript, but we would be willing to review a much-revised version. We cannot, of course, promise publication at that time.

Should you decide to revise the manuscript for further consideration here, your revisions should address the specific points made by each reviewer. We will also require a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript.

Notably, multiple reviewers voiced reservations about the MutL/DnaQ double mutant and if this combination results in an initial fitness defect independent of mutation rate that would alter its ability to evolve antimicrobial resistance. They note that there may be an inherent defect in the MutL background and that if the other mutants were built upon the MutL strain, this could impact the authors results. Thus, in addition to responding to this concern provide a more complete description of ancestral mutations and how strains are constructed. Additionally, multiple reviewers also commented on the importance of considering how mutator alleles affect not just mutation rate, but also mutation spectrum. The authors may also want to consider the effects of indels as repressor loss of function is a valid source of resistance. Lastly, there is a concern about model fitting, the underlying assumptions of the models used, and how outliers may drive associations. To this end, I also encourage the authors to increase accessibility to the raw data, as providing the data underlying the figures either in a supplementary table or through digital repositories can aid the reviewers in validating the authors’ conclusions.

If you decide to revise the manuscript for further consideration at PLOS Genetics, please aim to resubmit within the next 60 days, unless it will take extra time to address the concerns of the reviewers, in which case we would appreciate an expected resubmission date by email to plosgenetics@plos.org.

If present, accompanying reviewer attachments are included with this email; please notify the journal office if any appear to be missing. They will also be available for download from the link below. You can use this link to log into the system when you are ready to submit a revised version, having first consulted our Submission Checklist .

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please be aware that our data availability policy  requires that all numerical data underlying graphs or summary statistics are included with the submission, and you will need to provide this upon resubmission if not already present. In addition, we do not permit the inclusion of phrases such as "data not shown" or "unpublished results" in manuscripts. All points should be backed up by data provided with the submission.

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine  (PACE) digital diagnostic tool.  PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org.

PLOS has incorporated Similarity Check , powered by iThenticate, into its journal-wide submission system in order to screen submitted content for originality before publication. Each PLOS journal undertakes screening on a proportion of submitted articles. You will be contacted if needed following the screening process.

To resubmit, log into your Editorial Manager account and select the option 'Revise Submission' in the 'Submissions Needing Revision' folder.

We are sorry that we cannot be more positive about your manuscript at this stage. Please do not hesitate to contact us if you have any concerns or questions.

Yours sincerely,

Megan Behringer

Guest Editor

PLOS Genetics

Justin Fay

Section Editor

PLOS Genetics

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: The authors’ major conclusion is that there is an optimal mutation rate for the evolution of drug resistances and that a mutation rate higher than the optimum appears to inhibit that evolution. The authors’ simulations of the results lead to the interesting conclusion that high drug sensitivity minimizes the effect of mutation rate on the evolution of drug resistance because fewer cells with beneficial mutation survive. That is, what is important in determining the relationship between mutation rate and speed of evolution is the effective population size under selection. This conclusion will inform our understanding and help to direct further experimentation on the evolution of drug resistances

I do have three major comments. The authors should modify their discussion to incorporate the possibilities that the following factors are also contributing to the results.

1. The main conclusion rests on the observation that the strain with the highest mutation rate failed to have the highest rate of evolution to drug resistances (Figs 3 & 5). But, that strain was a mutLdnaQ double, and it is clear from Fig S1B that the mutLdnaQ strain had a significantly lower growth rate than any other strain even in the absence of drugs. This result suggests that the strain is simply “sick” and that is at least part of the reason it failed to evolve drug resistance. While the authors attribute its growth defect to its high mutation rate, there is another good reason that it might be sick. dnaQ encodes the epsilon subunit of Pol III, the replicative DNA polymerase. Epsilon is part of the core and interacts with the polymerase subunit, alpha, and with theta, the third member of the core. Previous work has shown that interaction of alpha and epsilon is required for fully processive DNA synthesis (Studwell, and O’Donnell, 1990 JBC 265: 1171), and the high rate of deletion formation in dnaQ mutants may be due to these replication defects (Saveson & Lovett, 1997, Genetics 146:457). dnaQ null mutant strains rapidly acquire suppressor mutations that allow DNA replication to recover. The original identified suppressor was in dnaE, which encodes alpha (Lancy et al, 1989, J. Bact 171:5572), but other suppressors are possible, and they might have other phenotypic effects. Clearly, with the loss of MutL, the high mutation rate in the dnaQ mutant is further increased, but the data in this paper do not convincingly show that it is solely the high mutation rate, and not also the loss of other properties due to replication defects, that are responsible for the poor evolution of the mutLdnaQ stain.

2. As shown in Fig. 1 and discussed in lines 137ff; the various strains have different mutational spectra. In particular, the MMR mutants enhance transitions, whereas the presence of the mutT deletion in any strain results in a high rate of AT to CG transversions. This is completely as expected from previous reports. What is striking is that in Fig. 3A&B the strains with the highest evolved drug resistances were mutT mutants (this is assuming I can correctly identify the symbols, some of which are nearly identical). This result suggests that these strains could become drug resistant not just because of their high mutation rate, but also because AT to CG transversions are more likely to give drug resistance than transitions. This possibility could be because the mutT mutants produce a greater proportion of nonsynonymous mutations than the other mutators (Table 1& S2).

3. The three strains with the lowest growth rates are mutL, mutLmutQ, and mutLmuH. This suggests that the mutH and dnaQ alleles were recombined into the mutL single mutant strain that had a growth defect. The sequences of the genetic manipulations are not described, but I tried to confirm this assumption by looking at the mutations labeled “ancestor” of the individual strains in Table S1. I found only an A335V mutant in malP common to all the strains with mutL, including mutTmutL, which did not appear to have a growth defect. So I cannot attribute the low growth rates solely to a pre-existing mutation in the mutL parent before all the other mutant alleles were recombined in, but there does appear to be a growth defect common to all the mutL strains but mutTmutL, suggesting the defect existed at some point in the mutL parent.

Related to this comment: what are the mutations labeled “ancestor”? If, for example, mutS was the starting strain into which mutH and mutT were placed, should not all the “ancestor” mutations in mutS also appear in mutSmutH and mutSmutT? I found only seven mutations common to mut S and mutSmutH, and only four of these were in mutSmutT. This is confusing and the reader should not have to examine the ancestor mutations to understand the genetic manipulations. The authors should describe the constructions more completely and define what the ancestor mutations are.

Specific Comments

57-58: According to the drift-barrier hypothesis articulated in Ref. 2, there is no mechanism for selection to adjust the mutation rate to promote evolution. So this statement is a bit extreme, although the subsequent discussion modifies it.

154: see above about other possible contributions to decreased growth rates

Fig. S1b, Fig 3: It is very difficult to distinguish which strains are indicated by the open circles in these figures. Especially, the symbols for L and HT are nearly identical. The closed circles for SH and LQ also look nearly the same.

179-182: This sentence is difficult to understand

183-185: it would be clearer if the statement excepting the AMK results was first so that the reproducibility statement directly applied to the rest of the experiments.

193: It appears in Fig 3 D & E that the strain with the highest mutation rate, LQ, does have a lower average MIC doubling rate compared to the other mutant strains. The authors might supply some statistics to support this statement

198ff: this reviewer found it easier to follow the mathematical modeling with a table defining the terms, but I still found it difficult. Note that δ is not defined.

220-223: As discussed above, other factors could contribute to growth rate decline. In Fig S1B it is obvious that two strains, L and LH, have close to the same mutation rates as 4 other strains yet have significantly lower growth rates.

290-294: It is not clear why molecular weight of the antibiotic should influence the beneficial effect of a mutation. More likely it is the mechanism of the antibiotic.

351ff: Lukacisinova et al conclude that more sensitive strains evolved resistance faster than more resistant strains. This seems to contradict the conclusions of this paper. Also, Lukacisinova et al argue that using strains without growth defects eliminates the possibility that selection would act on reversing the growth defect instead of antibiotic resistance. This is related to my concern, discussed above. The authors should discuss these differences in approach and conclusions.

372-393: The authors should discuss the results I list in my first comment about dnaQ, viability and suppressors.

Table 1 and S2: A mean of 13.33 synonymous BPS appears 3 times, which seems unusual. Are these correct or was there a miscalculation or typo?

466: AT to CG doesn’t belong in this sentence

Reviewer #2: The manuscript submitted to PLoS Genetics by Shibai et al. reports on the relationship between mutation rate and adaptation rate under antibiotic selection. While the findings are interesting, they are not entirely novel, as previous studies have also shown that a high mutation rate can be costly and does not necessarily lead to better adaptation due to the potential impact on key processes such as replication, transcription, and translation. Although this is an interesting manuscript, it has several weaknesses that need to be addressed before it can be considered for publication.

My major concerns are listed below.

1. When calculating mutation rates, only single nucleotide polymorphisms (SNPs) are considered, which may be problematic. This approach overlooks several loss-of-function mutations, such as indels in a repressor of an efflux operon, that can lead to resistance. Additionally, mutation rates are calculated using only the final time points of evolved cultures, assuming that mutations accumulate monotonically. This assumption might be problematic and not consistent because the rate of adaptation is calculated using continuous phenotype-changing data over time (Figure 2, Figure S2).

2. The rate of adaptation is calculated (Figure S2) by fitting a line to the log(MIC) versus time trajectories. This approach is problematic because adaptation functions can be monotonic, stepwise, or sigmoidal, leading to poor fits in many cases. A better algorithm, such as fitting a log-logistic function, should be used. Alternatively, to be consistent with how mutation rates are calculated, only the net change over 9 days (delta-MIC) can be considered.

3. The assumptions in the population dynamics model are overly simplistic and sometimes problematic. For example, a monotonic increase in the benefit function is assumed as bi=b0+i⋅ϵ. This assumption is difficult to justify in the context of antibiotic resistance evolution. For instance, a single gyrase mutation might be sufficient for bacteria to survive ciprofloxacin, with all subsequent mutations having negligible effects within the experimental range.

4. The authors state, "Moreover, we showed that the relationship between the mutation rate and speed of adaptation is contingent upon the mode of action of the drug involved." However, I am not convinced that this relationship is adequately explained, and I am concerned that the issues highlighted in points #1 and #2 may mislead the authors' conclusions.

5. The claim that growth rates decrease as mutation rates increase is based on the fit in Figure S1, which is unusual and heavily influenced by a single data point.

6. Some of the nomenclature is confusing. For example, what is meant by "mutS mutant"?

Reviewer #3: In this work, the authors have investigated the role of different mutation rates in resistance to five antibiotics. There is a lot I like about this study from an experimental point of view. In particular, I appreciate that the authors performed mutation accumulation to measure the relative mutation rates of the strains. I also like the combination of experiments with parameter estimation.

I was a little disappointed to not see any genetic data from the evolved resistant populations. Particularly, looking at the mutational spectrum from the MA experiments in the different mutators would nicely setup looking at how spectra affects the propensity to evolve different genetic bases for resistance—both in terms of specific mutations within resistance genes, but also the set of genes that acquire mutations. I really think this is a missed opportunity here that would greatly increase the impact of the study.

I would like to have seen some discussion about the epistasis between the mutator alleles themselves. To a first approximation, there are 4 mutation rates here (see Fig3), (1) wild-type (one point), (2) loss of function (LOF) in MutSHL or DnaQ alone (four points), (3) LOF of MutT with or without LOF of MutSHL (six points), and (4) combined LOF of both Mut(SH)L and DnaQ (one point). Changes in MIC were also more or less similar within these clusters.

Joint LOF of DnaQ+Mut(SH)L is clearly an oddity. Given this, it might be an idea to temper the conclusions on the lack of monotonicity, as it’s hard to know whether the evolution of MIC is indeed constrained at higher mutation rates (putatively by the accumulation of deleterious mutations) or because of some synthetically deleterious combination of these alleles. If the authors have this data, it would be nice to compare what the growth rates of the ancestral knockout strains are relative to wild-type, to see if there’s any big defect present already in the LQ strain (or indeed any of the strains). An absence of an initial growth defect would help strengthen the conclusion regarding the non-monotonicity of MIC with mutation rate. The presence of a growth defect would potentially complicate this.

The authors may consider citing a paper from earlier this year by Elgrail et al. (2024, J Evol Biol) on ‘multi-locus mutators’, as it could provide additional support and context for their approach of studying the double knockout strains here. Pubmed link: https://pubmed.ncbi.nlm.nih.gov/38367184/

If I had to find something to comment negatively about, for balance, one area that could be improved was the choice of antibiotics. The authors could alleviate this by mentioning that their study system is for understanding the general principles of resistance evolution, and not specific to particular clinical outcomes. While resistance to each of these can be conferred by spontaneous mutation, mutation isn’t the predominant mechanism of resistance for four out of five of them, outside of laboratory settings. Enzymatic resistance is more relevant in the clinical context for trimethoprim (via acquired dfrA genes), amikacin (aminoglycoside-modifying enzymes), chloramphenicol (chloramphenicol acetyltransferases) and cefixime (beta-lactamases). Ciprofloxacin resistance is, however, acquired by predominantly by mutation. This might not matter for the general readership of PLOS Genetics, but does detract from the applicability to real-world resistance for these particular drugs. To be fair to the authors, they are not explicitly claiming clinical relevance, but there is a tendency toward that thinking implicitly when resistance is studied.

Technical questions:

In general, the experimental design is sound and the analyses performed on the data are appropriate.

1.Why was 34°C used for growth? This is not necessarily problematic, but it differs from others who have largely done experiments at 30°C or 37°C, and this introduces some difficulty when comparing with previous work. Growth temperatures can also affect mutation rates. A brief mention of why might put the readers at ease.

2.Minimal medium with amino acids: Again, no issues with this, but a note about why this medium might be helpful. Particularly, it might help the field to know that rich media such as LB inhibit the effects of some antibiotics e.g. trimethoprim.

3.Insertion sequence (IS)-free E. coli K-12: I don’t think this is a problem, but could the authors please elaborate why they chose to use this interesting strain? I think they have done this because they want to study the pure effects of the mutator alleles. However, IS mobilisation plays an important role in resistance evolution by loss of function, both in laboratory and clinical strains. It would help the reader understanding.

Presentation comments:

Fig1: I like this figure. I assume the pie charts are mutational spectra from the MA experiment, but this isn’t 100% explicit. Might it be helpful in the legend to group the mutations by transition/transversion?

Fig3: I find the colours palette non-intuitive and difficult to distinguish, in particular L/HT (both open points with a blue/green colour), and S/QT (two shades of yellow). The open-yellow Q points are also very hard to see. I can’t make out why some treatments were assigned open points.

I would suggest two things. First, use shapes in addition to colours. For instance, black squares for WT, coloured circles for the single-locus mutators (H,L,S,Q,T) and then coloured triangles for the two-locus mutators (LH,SH, LQ, etc…).

Second, group the labels according to the ‘clusters’ they form on the x-axis. If you group the labels in this way, then you only need to take care that the colours are very distinct within these. There are some nice options for six-colour palettes here: https://coolors.co/palettes/popular/6%20colors

FigS1: Repeat the colour scale legend here as it’s hard for the reader to make them flip between documents to read the figures, especially as it’s the same complicated colour scale as in Fig3.

I hope that the authors find these comments to be useful.

Danna Gifford

**********

Have all data underlying the figures and results presented in the manuscript been provided?

Large-scale datasets should be made available via a public repository as described in the PLOS Genetics data availability policy , and numerical data that underlies graphs or summary statistics should be provided in spreadsheet form as supporting information.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: No:  Raw data for figures was *not* available during review, either as part of the supplementary excel file or in the Data Availability statement

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean? ). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy .

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes:  Danna Gifford

Decision Letter 1

Megan Behringer

10 Jan 2025

PGENETICS-D-24-00811R1

Quantitative Analysis of Saturating Relationship between Mutation Rate and Speed of Adaptation under Antibiotic Exposure in Escherichia coli

PLOS Genetics

Dear Dr. Furusawa,

Thank you for submitting your manuscript to PLOS Genetics. After careful consideration, we feel that it has merit but does not fully meet PLOS Genetics's publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the remaining points raised during the review process.

We appreciate the careful revisions and thoughtful responses provided in this recent submission. However, there are still a couple of issues raised by the reviewers that merit discussion. In addition to the comments below, we also appreciate the newly-added indel data in the mutation rate and spectra results, and believe it elevates the impact of the manuscript. However, these new results deserve further discussion and contextualization. In particular, the increased indel frequency is primarily observed in the lower tier mutators and the LQ outlier, while indels are considerably rare in the higher tier mutators. As indels are expected to result in larger effects on average compared to base pair substitutions, and resistance mutations often arise in essential genes where disruption by indels can be lethal, it is important to consider the implications of the differing mutation spectra on your results.

Please submit your revised manuscript within 30 days Feb 09 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosgenetics@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pgenetics/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

* A rebuttal letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. This file does not need to include responses to formatting updates and technical items listed in the 'Journal Requirements' section below.

* A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

* An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, competing interests statement, or data availability statement, please make these updates within the submission form at the time of resubmission. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

We look forward to receiving your revised manuscript.

Kind regards,

Megan Behringer

Guest Editor

PLOS Genetics

Justin Fay

Section Editor

PLOS Genetics

Aimée Dudley

Editor-in-Chief

PLOS Genetics

Anne Goriely

Editor-in-Chief

PLOS Genetics

Journal Requirements:

1) Please upload all main figures as separate Figure files in .tif or .eps format. For more information about how to convert and format your figure files please see our guidelines:

https://journals.plos.org/plosgenetics/s/figures

2) Please ensure that the funders and grant numbers match between the Financial Disclosure field and the Funding Information tab in your submission form. Note that the funders must be provided in the same order in both places as well.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: The authors have satisfactorily addressed most of my specific comments but my overall opinion of this paper is not changed significantly by the revisions. While the results are interesting, the authors’ interpretations rely heavily on the behavior of one strain, the mutL dnaQ double mutant (LQ). Indeed, the relationship between mutation rates and the increases in MIC of the strains in Figure 3 can be interpreted in every case as linear with one clear outlier, LQ. Likewise, the relationship between mutation rates and growth rates of the strains in Fig. S2B can be interpreted as a straight line, no effect, with LQ again being a clear outlier and L and LH also lying below the line. All three original reviewers brought this up; reviewer #3 was most articulate, pointing out the possibility of “synthetically deleterious combination of these alleles”. The results in Fig. S2B verify this possibility. The authors have addressed this interpretation in the text, justifying their modeling by with the assumption that the effects of any detrimental factors are proportional to the effects of the observed mutation rates. But this assumption has no evidential support. Factors that reduce growth rates may directly or indirectly also reduce mutation rates and/or the rates of MIC increase.

These considerations cloud the interpretation of the results. I feel that most readers, if they read carefully, will come to the same conclusion.

One new comment. The authors repeatedly attribute the growth defects of the LQ strain to hyperactivation of mismatch repair (MMR) and reference Stefan et al, Ref. 35, for this possibility. But, they have misinterpreted Ref. 35. In Stefan et al, MMR is not inferred to affect the growth rate, but to act to keep the mutation rate low when cell growth is slow due to the silencing of dnaQ. They conclude: “the ε subunit primarily affects the processivity of DNA polymerase III” and that is what causes the growth defects.

The authors also reference a paper by Iyer et al, 2020, showing that overexpression of MutS causes double-strand breaks and inhibits cell division. However, these effects were with MutS dramatically overinduced. At lower levels (10x normal) MutS overexpression had no growth effect and, in fact, elevated some mutation rates, which effect was suppressed by overexpression of MutL. Also, as far as this reviewer can find, this paper did not appear in a peer-reviewed journal.

Reviewer #2: The authors addressed almost all my concerns, except for their linear model used for benefit change (bi = b0 + i⋅ϵ). However, this does not impact my recommendation for publication, as such differences in scientific opinions are both expected and justified. Therefore, I recommend the manuscript for publication.

Reviewer #3: I thank the authors for their high attention to detail and cleanly presented response to comments.

Regarding the new genetic information, I am surprised the authors did not make a little more about the information. Specifically, the fact that 'expected' mutations in the canonical resistance target, gyrA, were found in the ciprofloxacin populations, and ampC mutations in the cefotaxime populations, I think warrants mention.

Otherwise I am satisfied with the changes the authors have made.

Danna

**********

Have all data underlying the figures and results presented in the manuscript been provided?

Large-scale datasets should be made available via a public repository as described in the PLOS Genetics  data availability policy , and numerical data that underlies graphs or summary statistics should be provided in spreadsheet form as supporting information.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean? ). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review?  For information about this choice, including consent withdrawal, please see our Privacy Policy .

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes:  Danna Gifford

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

Figure resubmission:

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step. If there are other versions of figure files still present in your submission file inventory at resubmission, please replace them with the PACE-processed versions.

Reproducibility:

To enhance the reproducibility of your results, we recommend that authors deposit laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Decision Letter 2

Megan Behringer

17 Feb 2025

Dear Dr Furusawa,

We are pleased to inform you that your manuscript entitled "Quantitative Analysis of Relationship between Mutation Rate and Speed of Adaptation under Antibiotic Exposure in Escherichia coli" has been editorially accepted for publication in PLOS Genetics. Congratulations!

Before your submission can be formally accepted and sent to production you will need to complete our formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Please note: the accept date on your published article will reflect the date of this provisional acceptance, but your manuscript will not be scheduled for publication until the required changes have been made.

Once your paper is formally accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you’ve already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosgenetics@plos.org.

In the meantime, please log into Editorial Manager at https://www.editorialmanager.com/pgenetics/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production and billing process. Note that PLOS requires an ORCID iD for all corresponding authors. Therefore, please ensure that you have an ORCID iD and that it is validated in Editorial Manager. To do this, go to ‘Update my Information’ (in the upper left-hand corner of the main menu), and click on the Fetch/Validate link next to the ORCID field.  This will take you to the ORCID site and allow you to create a new iD or authenticate a pre-existing iD in Editorial Manager.

If you have a press-related query, or would like to know about making your underlying data available (as you will be aware, this is required for publication), please see the end of this email. If your institution or institutions have a press office, please notify them about your upcoming article at this point, to enable them to help maximise its impact. Inform journal staff as soon as possible if you are preparing a press release for your article and need a publication date.

Thank you again for supporting open-access publishing; we are looking forward to publishing your work in PLOS Genetics!

Yours sincerely,

Megan Behringer

Guest Editor

PLOS Genetics

Justin Fay

Section Editor

PLOS Genetics

Aimée Dudley

Editor-in-Chief

PLOS Genetics

Anne Goriely

Editor-in-Chief

PLOS Genetics

www.plosgenetics.org

Twitter: @PLOSGenetics

----------------------------------------------------

Comments from the reviewers (if applicable):

----------------------------------------------------

Data Deposition

If you have submitted a Research Article or Front Matter that has associated data that are not suitable for deposition in a subject-specific public repository (such as GenBank or ArrayExpress), one way to make that data available is to deposit it in the Dryad Digital Repository . As you may recall, we ask all authors to agree to make data available; this is one way to achieve that. A full list of recommended repositories can be found on our website .

The following link will take you to the Dryad record for your article, so you won't have to re‐enter its bibliographic information, and can upload your files directly: 

http://datadryad.org/submit?journalID=pgenetics&manu=PGENETICS-D-24-00811R2

More information about depositing data in Dryad is available at http://www.datadryad.org/depositing. If you experience any difficulties in submitting your data, please contact help@datadryad.org for support.

Additionally, please be aware that our data availability policy  requires that all numerical data underlying display items are included with the submission, and you will need to provide this before we can formally accept your manuscript, if not already present.

----------------------------------------------------

Press Queries

If you or your institution will be preparing press materials for this manuscript, or if you need to know your paper's publication date for media purposes, please inform the journal staff as soon as possible so that your submission can be scheduled accordingly. Your manuscript will remain under a strict press embargo until the publication date and time. This means an early version of your manuscript will not be published ahead of your final version. PLOS Genetics may also choose to issue a press release for your article. If there's anything the journal should know or you'd like more information, please get in touch via plosgenetics@plos.org .

Acceptance letter

Megan Behringer

PGENETICS-D-24-00811R2

Quantitative Analysis of Relationship between Mutation Rate and Speed of Adaptation under Antibiotic Exposure in Escherichia coli

Dear Dr Furusawa,

We are pleased to inform you that your manuscript entitled "Quantitative Analysis of Relationship between Mutation Rate and Speed of Adaptation under Antibiotic Exposure in Escherichia coli" has been formally accepted for publication in PLOS Genetics! Your manuscript is now with our production department and you will be notified of the publication date in due course.

The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript.

Soon after your final files are uploaded, unless you have opted out or your manuscript is a front-matter piece, the early version of your manuscript will be published online. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers.

Thank you again for supporting PLOS Genetics and open-access publishing. We are looking forward to publishing your work!

With kind regards,

Anita Estes

PLOS Genetics

On behalf of:

The PLOS Genetics Team

Carlyle House, Carlyle Road, Cambridge CB4 3DN | United Kingdom

plosgenetics@plos.org | +44 (0) 1223-442823

plosgenetics.org | Twitter: @PLOSGenetics

Associated Data

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

    Supplementary Materials

    S1 File

    S1 Fig. Mutator Construction Procedure. Each box represents either a mutator strain or the wild-type strain. Arrows depict the lineage from parent to offspring in the construction of mutators. The list of mutations on the right details those shared between each parent-child pair (refer to S1 Table for the complete list of mutations). The values indicated in each strain box represent the growth rates (1/h; mean ± SD). S2 Fig. (A) Neutrality in mutation accumulation. The dN/dS ratio was calculated for each mutator strain and the wild-type strain. Error bars show the standard deviations between the MA lineages. (B) Relationship between growth rate and mutation rate. The horizontal error bars show the standard deviation across MA lineages, whereas the vertical error bars represent the standard deviation among replicate experiments in growth rate measurement. The sample sizes in the growth rate measurements were n=20 for wild-type strains and n=10 for mutant strains. The color and fill pattern of the markers correspond to those in Fig 3. S3 Fig. Experimental Evolution of Mutator Strains Under Antibiotics. For all combinations of drugs and strains, the changes in MIC over time are plotted. Each plot overlays data from four replicate series. Dashed lines represent the minimum and maximum MIC values attainable within the constraints of our experimental setup. S4 Fig. Reproducibility of the adaptation speed quantification. The MIC doubling rates were estimated by conducting independent experimental evolution trials with varying duration (9 days and 5 days, respectively). Each dot and error bar show the mean and standard deviation of MIC fold change per day, calculated from the data in S3 Fig. The black solid diagonal line means y=x identity line, while the blue line shows linear regression without intercept. The linear regression coefficient and corresponding R2 value were computed for these data points (N=13). S5 Fig. The relationship between the mutation rate, calculated by the sum of fixed BPSs and indels, and the MIC doubling rate is shown. Each dot represents an experimental observation from 13 strains across four replicate serial transfer cultures. To prevent overlap of data points, small Gaussian noise (mean = 0, standard deviation = 0.05) was added to the y-coordinates. S6 Fig. Mutation Spectra in Resistant Strains. The distribution of substitution patterns and indels calculated by 24 resistant strains (three ancestor mutators, two antibiotics, and four replicates) is presented. Each dot represents the mutation fixation rate during the experimental evolution calculated based on BPSs (closed circle) or indels (open circle). S7 Fig. Estimated parameter ϵ representing the beneficial effect of each mutation. The mean value estimated through 100 bootstrap resampling is represented by the bars, while the error bars represent the standard error. S8 Fig. A doubled drug concentration with large β remarkably reduces the growth rate. We defined the net growth rate as μ-γ where μ is the drug-free growth rate and γ is the death rate by drug treatment. μx is the net growth rate with drug concentration x, μ2x is the net growth rate with drug concentration 2x, and μmut2x is the net growth rate with drug concentration 2x for cells with one mutation of beneficial effect ϵ. We assume here a situation where the growth rate and dilution rate are balancing at the drug concentration x, i.e., μx=ln100/24≅0.19h−1. The black dashed line represents μx=0.19. We used ϵ=0.9 and bi=1 for these simulations. S9 Fig. Changes in growth rate during resistance evolution. Growth rates for three strains (WT, S, and LQ) were quantified both before and after experimental evolution. For each selection drug, two independently evolved strains were analyzed. The bars represent the standard deviation of three replicate measurements. S10 Fig. Relationship between the parameter S and the residual of fitting. The parameter S, which describes the saturation effect of mutation accumulation in the form bi=b0+iϵS/i+S, is plotted against the residual of the fitting. The modified model incorporating the saturation effect was used to fit the experimental data shown in Fig 3. As observed, the residuals decrease with increasing S, suggesting that the saturation effect is negligible.

    (DOCX)

    pgen.1011627.s001.docx (1.4MB, docx)
    S1 Table. List of mutations identified in mutation accumulation (MA) experiments.

    (XLSX)

    pgen.1011627.s002.xlsx (927.8KB, xlsx)
    S2 Table. Extended summary of MA experiments.

    (XLSX)

    pgen.1011627.s003.xlsx (11.1KB, xlsx)
    S3 Table. Colony-size records during MA experiments.

    (XLSX)

    pgen.1011627.s004.xlsx (22.6KB, xlsx)
    S4 Table. Common mutations in ancestors which were omitted from the detected mutations in this study.

    (XLSX)

    pgen.1011627.s005.xlsx (11.1KB, xlsx)
    S5 Table. Number of generations during the 9-day experimental evolution.

    (XLSX)

    pgen.1011627.s006.xlsx (13.9KB, xlsx)
    S6 Table. Number of common mutations among ancestral strains.

    (XLSX)

    pgen.1011627.s007.xlsx (10.7KB, xlsx)
    S7 Table. List of mutations in drug-resistant evolved strains.

    (XLSX)

    pgen.1011627.s008.xlsx (72.9KB, xlsx)
    S8 Table. Mutational spectrum of wild-type E. coli during drug resistance evolution experiments and drug-free MA experiment in previous studies.

    (XLSX)

    pgen.1011627.s009.xlsx (14.8KB, xlsx)
    Attachment

    Submitted filename: Response_to_Reviewers.docx

    pgen.1011627.s011.docx (1.1MB, docx)
    Attachment

    Submitted filename: Response_to_Reviewers.pdf

    pgen.1011627.s012.pdf (169.5KB, pdf)

    Data Availability Statement

    The resequencing analysis data have been deposited at the DDBJ Sequence Read Archive (https://www.ddbj.nig.ac.jp/dra/index-e.html) under accession number DRA016357 and DRA019544.


    Articles from PLOS Genetics are provided here courtesy of PLOS

    RESOURCES