By repeating the cycle of mutagenesis and selection, the Escherichia coli dnaQ49 mutator acquired high level resistance to ampicillin (30 000 μg ml−1), streptomycin (26 000 μg ml−1) and ofloxacin (3000 μg ml−1). Under the strong pressure of ofloxacin, dnaQ49 also followed the history of mutations in the gyrase and topoisomerase IV genes previously observed in clinical isolates of quinolone-resistant E. coli. The results of these in vitro experiments suggest that naturally existing mutators may participate in the rapid acquisition of resistance to various antibiotics in patients. A possible mechanism for the occurrence of this adaptability is discussed with special reference to the property of mutagenesis accompanying DNA replication.
Natural mutators exist in Escherichia coli and pathogenic bacteria [1–3] and mutator strains spontaneously appear in long-term cultures of E. coli under a weak selection pressure . These observations suggest that mutators may contribute to the rapid acquisition of resistance to antibiotics in clinical cases. In order to test this prediction, we chose the E. coli dnaQ49 mutator. dnaQ49 displays hypermutation rates, and one amino acid substitution in the 3′-5′ exonuclease subunit of the DNA Pol III in the wild-type E. coli gives rise to the dnaQ49 phenotype, which very likely also occurs in pathogenic bacteria. The adaptability of the mutator to various antibiotics was examined. A plausible mechanism for exhibiting the strong adaptability is discussed from the viewpoint of mutagenesis accompanying DNA replication.
2Materials and methods
2.1Bacterial strains and mutagenesis
The E. coli KH1370 strain with a normal mutation rate, which is the isogenic wild-type of dnaQ49 (KH1366; a recA+ derivative of KH1116), was used as a control . Both dnaQ49 and KH1370 have a β-lactamase gene, ampC  and do not carry plasmids. The dnaQ49 strain has a normal mutation rate at 24°C, while at 37°C the rate increases to 1000–10 000 times compared to the wild-type . As a control for the mutagenesis by dnaQ49, N-methyl-N′-nitro-N-nitrosoguanidine (MNNG) was used, which alkylates bases.
dnaQ49 or KH1370 was grown in a 5-ml culture tube to a cell density of 4×108 ml−1 at 24°C, washed with Luria broth (LB) by centrifugation, and then suspended in 5 ml of LB (4×102 cells ml−1). The dnaQ49 mutator function was induced by shifting the culture temperature to 37°C until the cell density became about 4×108 ml−1 (usually 24 h). Mutations in the wild-type KH1370 were introduced by treatment with MNNG. A freshly prepared MNNG solution was added to the cell suspension (4×102 cells ml−1) to give various concentrations and these mixtures (5 ml) were incubated for 24 h at 37°C. The mutagenized cells were washed three times and spread on LB agar plates containing various concentrations of ampicillin, and the plates were incubated at 24°C for 2–3 days. Isolated single colonies from the plates, which represented the upper limit ampicillin concentration in which the cells could form colonies, were inoculated into 5 ml of LB containing the same concentration of ampicillin. These cultures were grown to a cell density of 4×108 ml−1 at 24°C, washed, suspended in 5 ml of LB without ampicillin at a density of 4×102 ml−1, and then incubated at 37°C for 24 h for the second mutagenesis. MNNG solutions were added to the KH1370 cell suspensions to give the same concentration used in the first mutagenesis and the cultures (5 ml, 4×102 cells ml−1) were incubated at 37°C for 24 h. These mutagenesis/selection cycles were repeated.
2.3Culture in liquid medium
For the first mutagenesis, the same procedure was followed as described above. After mutagenesis for 24 h, the cells were washed, suspended in 5 ml of LB (4×107 cells ml−1) containing various concentrations of ampicillin or other antibiotics and incubated at 24°C until the cell density reached a value of OD550=0.4. In the successive cycles, mutagenesis was carried out in the absence or presence of the upper limit of antibiotic in which the cells could grow well. By the sixth cycle of mutagenesis/selection, mutagenesis with or without antibiotic gave essentially identical results. Mutagenesis after the seventh cycle was carried out in the presence of ampicillin, since the presence of ampicillin produced variants with higher resistance compared with mutagenesis without the antibiotic.
3.1Acquisition of ampicillin resistance
Fig. 1 summarizes the results of the colony assay for ampicillin resistance in the three times repeated experiments. The wild-type KH1370 showed a weak resistance to ampicillin (MIC=1 μg ml−1) due to the genomic β-lactamase gene. dnaQ49 showed a marked increase in drug resistance with repeating cycles. It finally developed colonies on agar plates containing 10 000 μg ml−1 of ampicillin at the fourth and fifth cycles (Exp. 2), 6000 μg ml−1 at the fifth (Exp. 1) and 6000 μg ml−1 at the seventh cycle (Exp. 3), respectively. In contrast, MNNG-treated KH1370, at all concentrations tested (1, 3, 6, 10, 30, 60 and 100 μg ml−1), did not produce colonies in the presence of drug concentrations higher than 100 μg ml−1 even after repeated mutagenesis/selection cycles. Extinction of populations was often observed, irrespective of the MNNG doses used. In Exp. 1, two clones survived after the fifth round of mutagenesis/selection in the cases of 1 and 6 μg ml−1 of MNNG treatment, and the maximal tolerated doses (MTD) were 60 and 30 μg ml−1 of ampicillin, respectively. In Exp. 2, one clone survived after the fifth round in the case of 1 μg ml−1 of MNNG treatment, and the MTD was 30 μg ml−1. In Exp. 3, two clones survived after the seventh round in the cases of 6 and 10 μg ml−1 of MNNG treatment, and their MTD was 100 μg ml−1 (data for MNNG not shown in Fig. 1). The intact dnaQ+ cells became extinct after the third round in Exp. 2. However, in Exp. 1 and 3, the dnaQ+ formed colonies after the fifth and seventh rounds, and their MTD was 60 μg ml−1 (Fig. 1).
The result of culturing dnaQ49 in a liquid medium was more striking, in that we have so far obtained variants proliferating in 30 000 μg ml−1 of ampicillin. These super ampicillin-resistant bacteria were established after only eight mutagenesis/selection cycles. Identification of the mutations which may contribute to the hyper-resistance remains to be examined. A preliminary experiment, however, shows that four amino acid substitutions were observed in ampC in the variant which grew at 10 000 μg ml−1 (n= 2): 31Thr (ACA)→Gln (CAA), 32Ile (ATT)→Phe (TTT), 148Leu (CTG)→Val (GTG) and 238Met (ATG)→Val (GTG). The detailed characterization of this variant will be described elsewhere.
This highly ampicillin-resistant E. coli was sensitive to other antibiotics, but also highly resistant to cefotaxime which is an improved β-lactam (Table 1). MIC values of the highly ampicillin-resistant variant to other antibiotics are equivalent to or less than those of the parental dnaQ49.
Table 1. Sensitivity of the super ampicillin-resistant dnaQ49 variant to various antibiotics
aGrowing even in the presence of 30 000 μg ml−1 of ampicillin during prolonged incubation time.
MIC (μg ml−1)
3.2Acquisition of resistance to other antibiotics
Highly resistant variants to other antibiotics were also obtained in liquid culture from dnaQ49 by the same procedure. We have established variants with high resistance to streptomycin (MIC=2048 μg ml−1) after the 22nd round, nalidixic acid (2048 μg ml−1) after the 23rd round and ofloxacin (1024 μg ml−1) after the 63rd round of mutagenesis/selection. The MIC values of each drug in the parent dnaQ49 were 1, 1 and 0.0156 μg ml−1, respectively. By prolonging the culture periods, the cells could proliferate although growth rates were very low. The streptomycin-resistant variant proliferated in 26 000 μg ml−1 of streptomycin, the nalidixic acid-resistant one in 7000 μg ml−1 of nalidixic acid and the ofloxacin-resistant one in 3000 μg ml−1 of ofloxacin.
3.3Characterization of ofloxacin-resistant variants
Mutations which appeared in the gyrase A (gyrA) and topoisomerase IV (parC) genes in the ofloxacin-resistant dnaQ49 variants are noteworthy (Fig. 2). We found one mutation 83Ser (TCG)→Leu (TTG) in the quinolone resistance-determining region of gyrA (amino acid residues 64–106) from three variants of MIC=0.25, 4 and 16 μg ml−1, and a double mutation of 83Ser→Leu in GyrA plus 80Ser (AGC)→Arg (AGA) in the related region of ParC from two variants of MIC=128 and 256 μg ml−1. Surprisingly, no additional base changes, including silent mutations, were found in these domains of these genes. The order and sites of quinolone resistance mutations are coincident with those reported from clinical isolates of quinolone-resistant E. coli, indicating that dnaQ49 might recapitulate the history of acquisition of mutations in quinolone-treated patients in a constrained period of time. The highest quinolone MIC observed so far in clinical isolates is 100 μg ml−1. Thus, there is a possibility that the site(s) of mutation in forthcoming quinolone-resistant variants of E. coli from quinolone-treated patients might be predicted by analyzing mutations in the current dnaQ49 quinolone-resistant variants.
The mutator phenotype of dnaQ49 was produced by a single point mutation in the 3′-5′ exonuclease subunit of the Pol III gene . Even in E. coli, more than 10 mutators have been identified, indicating that the number of candidate mutations which give rise to mutators can be significantly large. Similar things may happen in the case of pathogenic bacteria in nature. Thus, it can be said that naturally existing mutators of pathogenic bacteria in the human body may participate, at least partly, in the rapid appearance of prominent antibiotic-resistant bacteria.
The question remains as to how the dnaQ49 population exhibits such adaptability and follows the evolutionary history of quinolone resistance in patients. One possibility relates to the mechanism of mutagenesis. It is believed that a main cause of evolution is mutations accompanying DNA replication and the mutations intrinsically arise from naturally occurring tautomers of nucleotide bases. As dnaQ49 lacks the proofreading activity of the 3′-5′ exonuclease at 37°C but other machinery for DNA replication is intact, the characteristics of mutations introduced accompanying DNA replication appear to be no different from those introduced in native E. coli. Therefore, the lack of the proofreading activity may lead to a rapid accumulation of potent mutations caused by base tautomers, contributing to the acceleration of evolution. The above explanation, however, appears to be insufficient to understand fully the present results. This is because it could be speculated that the high mutation rate in dnaQ49 may give rise to the cancellation of the function of beneficial mutations in successive generations. Another explanation can be found in our disparity model of mutagenesis [10–12].
dnaQ49 is estimated to accumulate, on average, 1–10 mutations per genome per replication  and the fidelity difference between the leading and lagging strands is at least 100 times (; and unpublished observations). If so, in each replication, several mutations are inserted exclusively in a hemicircle of the genomic DNA on either side of the origin of replication (ori) synthesized as the lagging strand. No mutations occur in a hemicircle synthesized as the leading strand. In the leading strand hemicircle, developed genotypes are guaranteed in successive generations, but in the lagging strand hemicircle, several additional mutations are inserted in every replication. Thus, the disparity model guarantees the existence of genotypes developed once in the past in the hemicircle of genomic DNA on either side of the ori in successive generations, serving to enlarge the genotypic diversity of the population. Computer simulation showed that when two mutations occurred on average in two daughter DNAs and the fidelity difference between DNA strands was more than 100 times, genotypes that had appeared in the past were guaranteed in successive generations . Thus, the ‘disparity’ mutator may be able to accumulate beneficial stepwise mutations as seen in the quinolone-resistant variants. In contrast, MNNG treatment may introduce mutations fairly evenly throughout the genome regardless of DNA replication. Therefore, beneficial mutations once acquired always face the risk of extinction by additional mutations. To show more precisely the advantage of our disparity model of mutagenesis for the current adaptive evolution, more detailed experiments would be required.
We thank Dr. H. Maki for discussion and the E. coli strains including dnaQ49, Drs. M. Tanaka and K. Sato for useful suggestions, and Drs. F. Rüker and D. Stearns-Kurosawa for critical reading of the manuscript.