Effect of Ciprofloxacin Concentration on the Frequency and Nature of Resistant Mutants Selected from Pseudomonas aeruginosa mutS and mutT Hypermutators^,

Abstract

The rapid emergence of drug resistance upon treatment of Pseudomonas aeruginosa infections with fluoroquinolones is a serious concern. In this study, we report the effect of hypermutability on the mutant selection window for ciprofloxacin (CIP) by comparing the hypermutator MPAO1 mutS and mutT strains with the wild-type strain. The mutant selection window was shifted to higher CIP concentrations for both hypermutators, presenting the mutS strain with a broader selection window in comparison to the wild-type strain. The mutation prevention concentrations (MPC) determined for mutT and mutS strains were increased 2- and 4-fold over the wild-type level, respectively. In addition, we analyzed the molecular bases for resistance in the bacterial subpopulations selected at different points in the window. At the top of the window, the resistant clones isolated were mainly mutated in GyrA and ParC topoisomerase subunits, while at the bottom of the window, resistance was associated with the overexpression of MexCD-OprJ and MexAB-OprM efflux pumps. Accordingly, a greater proportion of multidrug-resistant clones were found among the subpopulations isolated at the lower CIP concentrations. Furthermore, we found that the exposure to CIP subinhibitory concentrations favors the accumulation of cells overexpressing MexCD-OprJ (due to mutations in the transcriptional repressor NfxB) and MexAB-OprM efflux pumps. We discuss these results in the context of the possible participation of this antibiotic in a mutagenic process.

INTRODUCTION

The treatment of Pseudomonas aeruginosa infections represents a major therapeutic challenge due to the intrinsic and acquired resistance to a wide range of antimicrobial agents. Natural resistance mechanisms include constitutive expression of AmpC β-lactamase and multidrug efflux pumps combined with low outer membrane permeability (28, 40). The rapid acquisition and selection of resistance mutations considerably compromise the efficacy of treatments. Subpopulations with multidrug-resistant phenotypes reach up to 20% of bacterial isolates in chronic infections with Pseudomonas, and they are frequently associated with a hypermutator genetic background (4, 23).

The enrichment of resistant mutants is a likely consequence of using antibiotic concentrations placed within the mutant selection window: i.e., the range between the concentration that blocks growth of the majority of susceptible pathogens (MIC) and the concentration that blocks growth of first-step-resistant mutants (mutation prevention concentration [MPC]) (45). Given the mutation rates of bacterial pathogens, second-step-resistant mutants (carrying two resistance mutations) are expected to arise rarely among infective populations (on the order of 1 per 10¹² to 1 per 10¹⁶ cells) (45). Thus, it is generally accepted that to avoid the selection of resistant clones, drug concentrations must be kept above the MPC for a given bacterial strain. However, an important aspect that has not been taken into consideration is the effect of hypermutability on the mutant selection window. Hypermutator strains display increased mutation rates, and as a consequence, they can generate double mutants more frequently. In particular, P. aeruginosa hypermutator strains are found in a high proportion among chronic infections, reaching up to 30% of the clinical isolates (27, 34, 42). The majority of these hypermutable isolates are deficient in the mismatch repair system (MRS), mostly through alterations in mutS, and they display significantly higher antibiotic resistance levels in relation to nonmutator isolates (34, 35). Mutants defective in the 8-oxoguanine (GO) repair system have also been detected among infective populations (25).

Drug therapies for P. aeruginosa infections are currently limited to the use of a few antibiotics, including fluoroquinolones, such as ciprofloxacin (CIP). Fluoroquinolones are broad-spectrum antibiotics that inhibit DNA replication by interacting with DNA gyrase (coded for by gyrA and gyrB) and DNA topoisomerase IV (coded for by parC and parE) (9). Four pumps that efflux CIP in Pseudomonas have been described: MexAB-OprM, MexCD-OprJ, MexEF-OprN, and MexXY-OprM, which are regulated by the MexR, NfxB, MexT, and MexZ proteins, respectively (1, 21). In the clinical setting, high incidence and levels of resistance to CIP and other fluoroquinolones have been well documented (13). The molecular bases of resistance to these antibiotics are mutations in gyrA and parC and mutations in the transcriptional regulators of multidrug efflux pumps (43). Long-term treatment with CIP frequently results in the selection of clones containing two or more resistance mutations, providing increased levels of resistance (16, 18). The relative contributions of target mutations (gyrA and parC) and efflux upregulation to CIP resistance in clinical P. aeruginosa isolates are not well established (11).

Previous work has indicated that bacterial exposure to different concentrations of fluoroquinolones results in the selection of different resistant populations with distinct properties in terms of the molecular bases of resistance and resistance potentials (24, 47). In addition, a general relationship between fluoroquinolone concentration and the restriction of colony growth has been previously established for clinical isolates of P. aeruginosa (14). An analysis of the distribution of MPC values for CIP among fluoroquinolone-susceptible clinical isolates of P. aeruginosa showed that these values ranged mainly between 0.5 and 8 μg ml⁻¹. However, the mutator phenotypes of isolates were not analyzed in that study (14).

Furthermore, exposure to subinhibitory concentrations of CIP was shown to promote the development of low-level resistance to different unrelated antibiotics (12). This phenomenon was also observed with other classes of antibiotics, including aminoglycosides and β-lactams (15, 20).

We have previously reported that, when determined at a particular CIP concentration (2-fold higher than the MIC), CIP resistance frequencies for the hypermutable mutS and mutT P. aeruginosa strains (defective in the MRS and GO repair system, respectively) are approximately 1,000-fold increased above the wild-type level (31). In the present study, we performed a population analysis of the wild-type strain and the hypermutator mutT and mutS strains to determine the effect of hypermutability on the size and nature of CIP-resistant subpopulations. We determined the CIP mutant selection window for wild-type and mutS- and mutT-deficient P. aeruginosa strains. In addition, we analyzed the molecular bases of resistance selected and the occurrence of multidrug resistance phenotypes among CIP-resistant clones isolated at different concentrations between the lower and upper limits of this window. Finally, we report the effect of the exposure to CIP subinhibitory concentrations in solid media on the emergence of nfxB mutants from wild-type, mutT, and mutS P. aeruginosa strains.

MATERIALS AND METHODS

Bacterial strains, plasmids, and media.

The bacterial strains and plasmids used in this study are listed in Table S1 in the supplemental material. The P. aeruginosa MPAO1 wild-type strain and its derivative mutT and mutS strains (deficient in the expression of endogenous mutT and mutS, respectively) were kindly provided by Michael Jacobs from the University of Washington Genome Center (17). Transposon insertion within the corresponding genes was confirmed by PCR analysis following the manufacturer's instructions.

To prepare inocula, bacteria were routinely cultured on Luria-Bertani (LB) (38) agar plates from frozen stocks and subcultured in LB liquid medium overnight at 37°C with shaking at 250 rpm. Plates for susceptibility tests were prepared with Mueller-Hinton agar (Biokar Diagnostics).

Determination of fractions of cells recovered in plates with ciprofloxacin.

To determine the fractions of cells recovered on LB plates containing different ciprofloxacin (CIP) concentrations, approximately 5 × 10³ cells from an overnight single-colony culture of wild-type or mutT or mutS MPAO1 strains were inoculated in 2 ml of LB medium and incubated at 37°C for 16 h. Aliquots from successive dilutions of these subcultures were plated onto LB plates to measure the total number of viable cells and onto LB plates containing CIP at the following concentrations: 0.06, 0.08, 0.1, 0.2, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.25, 1.5, 2.0, 2.25, 2.5, 3.0, 3.5, 4.0, and 5.0 μg ml⁻¹. Doses from 0.06 to 0.2 μg ml⁻¹ were used for the wild-type strain, while doses from 2.5 to 5.0 μg ml⁻¹ were used for mutT and mutS strains. Colonies growing after 36 h of incubation were counted, and the number of cells growing in CIP-containing plates was expressed as a fraction of the total number of cells applied to each plate.

Subpopulations selected at CIP concentrations close to the MIC were tested for regrowth on the same CIP concentration after being grown in drug-free plates. This assay was conducted with 150 colonies isolated at CIP concentrations of 0.06, 0.08, and 0.5 μg ml⁻¹ for the wild-type strain, and 150 colonies were isolated at CIP concentrations of 0.4, 0.5, and 0.6 μg ml⁻¹ for each hypermutator strain. The fraction of cells that regrew at the same CIP concentration initially used for selection was almost indistinguishable (97% to 100%) from those determined at the initial selective step. This control was included in order to rule out the possibility that the high fraction of cells recovered at CIP concentrations close to the MIC was influenced by experimental artifacts.

Estimation of the mutation prevention concentration.

The mutation prevention concentration (MPC) was determined for wild-type, mutT, and mutS MPAO1 strains as described by Zhao and Drlica (45). MPC was obtained as the minimal antibiotic concentration at which no resistant clone was recovered when 1 × 10¹⁰ cells were applied onto selective plates. This population size is considered to be substantially larger than that normally present in an infection (30). To estimate the MPC, approximately 1 × 10¹⁰ cells from 10 independent overnight cultures of each strain were applied to agar plates supplemented with increasing concentrations of CIP (specified above), and the number of cells recovered as CFU was measured after overnight incubation at 37°C.

Isolation of ciprofloxacin-resistant mutants and genotypic analysis.

CIP-resistant mutants were selected by plating dilutions of independent overnight cultures of the wild-type strain onto LB plates containing 0.5 and 1 μg ml⁻¹ CIP and of mutT and mutS strains onto LB plates containing 0.5 and 2 μg ml⁻¹ CIP. Data corresponding to the selection of mutT and mutS strains at 1 μg ml⁻¹ CIP were taken from our previous work (31). After incubation at 37°C for 36 h, between 10 and 20 clones derived from each culture and selective condition were subsequently isolated on LB plates. Sequence analysis of gyrA, parC, and nfxB was conducted as previously described (31). The sequences of mexZ, mexT, and mexR were also determined for those isolates selected at 0.5 μg ml⁻¹ CIP that had no mutations in any of the three previously analyzed genes. For this purpose, primers that covered the entire open reading frames of genes were designed (see Table S1 in the supplemental material). DNA sequencing was performed at the University of Chicago CRC-DNA Sequencing Facility.

Disk diffusion susceptibility assays.

Disk diffusion tests were performed with Neosensitabs tablets (Rosco, Denmark) and monodisks from Britania (Argentina) according to CLSI guidelines (7). Inocula were prepared from overnight LB cultures and then adjusted to match the turbidity of a 0.5 MacFarland standard. A 50-μl aliquot of this dilution was applied to Mueller-Hinton plates, and disks were then placed on the surface. Zones of growth inhibition were measured after overnight incubation at 37°C. The following disks of antibiotics were used: fluoroquinolones (5 μg ciprofloxacin and 5 μg levofloxacin), β-lactams (100 μg piperacillin, 30 μg ceftazidime, 10 μg imipenem, 10 μg meropenem, and 30 μg aztreonam), macrolides (60 μg erythromycin and 15 μg azithromycin), aminoglycosides (30 μg amikacin, 300 μg streptomycin, 250 μg gentamicin, and 10 μg tobramycin), and 10 μg colistin. This analysis was conducted at least with seven independent clones isolated at each CIP concentration.

Construction of luminescent reporters for the transcriptional derepression of mexCD-oprJ and mexAB-oprM operons.

To identify nfxB mutants among P. aeruginosa populations, a transcriptional fusion of the promoter of mexCD-oprJ operon to the lux operon was designed and inserted in the chromosome of wild-type, mutT, and mutS MPAO1 strains using a broad-host-range mini-Tn7 vector (6). For this purpose, a DNA fragment comprising the intergenic region between nfxB and mexC was amplified from MPAO1 using primers P_{CD_}Fw and P_{CD_}Rv (see Table S1 in the supplemental material) and cloned upstream of the lux operon in the SpeI-BamHI restriction sites of pUC18-mini-Tn7T-Gm-lux vector. The insertion of this transcriptional fusion at the Tn7 attachment (attTn7) site in the bacterial chromosome was achieved by cotransformation with pTNS2 helper plasmid and selection with gentamicin. Chromosomal insertions were confirmed by PCR as recommended in the published protocol (6). Luminescence was measured directly from agar plates in a NightOWL LB 983 luminometer (Berhold Technologies).

Similarly, to evaluate the overexpression of this efflux pump among mutT- and mutS-derived CIP-resistant isolates selected at 0.5 μg ml⁻¹ CIP and carrying no mutations in nfxB or the topoisomerase genes, a luminescent reporter of MexAB-OprM expression was designed. To this end, a sequence including the intergenic region between mexR and mexA was amplified with primers P_{AB_}Fw and -Rv (see Table S1 in the supplemental material) and cloned upstream of the lux operon in pUC18-mini-Tn7T-Gm-lux vector, as described above. This construction was introduced in the chromosome of wild-type, mutT, and mutS strains, as well as in the mutS- and mutT-derived strains selected at 0.5 μg ml⁻¹ CIP (mentioned above).

Effect of ciprofloxacin on the emergence of nfxB mutants.

To determine the influence of CIP at subinhibitory levels on the emergence of nfxB mutants, reporter wild-type, mutT, and mutS P. aeruginosa strains containing the nfxB-mexC intergenic region in fusion with the lux operon were used. Twenty independent overnight cultures of each strain were diluted and approximately 300 CFU from each culture were applied onto LB solid media with or without the addition of CIP at subinhibitory levels (0.06 μg ml⁻¹ for the wild-type strain and 0.4 μg ml⁻¹ for the mutT and mutS strains). Luminescence of colonies growing after 24, 48, and 72 h was measured, and the most luminescent colony from each plate was further isolated in LB solid media to verify the stability of luminescent phenotypes. PCR colony and DNA sequencing analysis of nfxB were performed for about 6 to 10 of the most luminescent clones derived in this way from each strain and condition.

RESULTS

Mutant selection windows for wild-type and hypermutator P. aeruginosa strains.

Fractions of cells recovered from wild-type, mutT, and mutS MPAO1 cultures when applied to solid media containing different CIP concentrations are shown in Fig. 1. The minimal drug concentration required to inhibit the growth of 99% of the bacterial population (MIC₉₉) and the mutant prevention concentration (MPC) are indicated by arrowheads and asterisks for each strain in Fig. 1, respectively. Numerical values for both parameters are also listed in the legend to Fig. 1.

Fig. 1.

Fraction of cells recovered in LB plates containing ciprofloxacin (CIP) at different concentrations from the wild-type (WT; ●), mutT (○), and mutS (▾) MPAO1 strains (see Materials and Methods). The MICs required to inhibit the growth of 99% of the bacterial population (MIC₉₉) were determined from the plots (indicated by arrowheads). These values are 0.08, 0.45, and 0.52 μg ml⁻¹ for the wild-type, mutT, and mutS strains, respectively. Symbols marked with asterisks represent the CIP concentration at which no colony is recovered when 1 × 10¹⁰ cells are applied to the plates (mutant prevention concentration [MPC]). MPC values are 1.25, 2.25, and 5 μg ml⁻¹ for the wild-type, mutT, and mutS strains, respectively.

MIC₉₉ and MPC values determined for the hypermutator strains were higher than those obtained for the wild-type strain, resulting in mutant selection windows shifted to higher CIP concentrations. Both hypermutator strains showed similar MIC₉₉ values that were ∼6-fold higher than those of the parental strain. MPCs were 2- and 4-fold increased over the wild-type level for mutT and mutS strains, respectively. As shown in Fig. 1, the mutS strain exhibited the broadest mutant selection window.

The relationship between the fraction of cells recovered and the CIP concentration showed the typical stepwise dependence previously described in fluoroquinolone resistance studies on Mycobacterium and Staphylococcus aureus (8, 47). Each plateau region spanning between the MIC₉₉ and MPC in these curves probably represents distinct types of resistant mutants, as indicated by the sequencing analysis of variants selected at 0.5, 1.0, and 2.0 μg ml⁻¹ CIP.

Molecular bases of resistance at different ciprofloxacin concentrations.

To identify the molecular mechanisms of resistance to CIP, the sequences of gyrA, parC, and nfxB were determined for resistant clones isolated from wild-type, mutT, and mutS MPAO1 strains at CIP concentrations of 0.5, 1, and 2 μg ml⁻¹. These results are summarized in Table 1. All of the resistant variants isolated from the wild-type strain at 0.5 and 1 μg ml⁻¹ CIP were mutated in nfxB. For this strain, no resistant clone was recovered at 2 μg ml⁻¹. A high proportion of nfxB mutants were also found among the isolates selected at 0.5 and 1 μg ml⁻¹ CIP for the mutT strain. The mutS strain also showed a high proportion of nfxB mutants at 0.5 μg ml⁻¹ CIP; however, the proportion of nfxB mutants substantially diminished among the resistant population isolated at 1 μg ml⁻¹ CIP. When the selection pressure increased (2 μg ml⁻¹ CIP for both hypermutator strains), the representation of gyrA mutants increased to nearly 100% of the resistant mutant population.

Table 1.

Mutations detected in ciprofloxacin-resistant isolates derived from P. aeruginosa strains

MPAO1 strain type	CIP concn (μg ml−1)	No. of clones analyzed	Mutated gene(s) (% of clones mutated)a	Mutation(s) (no. of occurrences)b
Wild type	0.5	10	nfxB (100)	ins C at nt 40–41 (1), A→C (Thr39Pro) (4), ΔA at nt 115 (1), G→A (Arg42His) (2), G→C (Ala181Pro) (1), many substitutions from nt 485 (1)
	1.0	5	nfxB (100)	T→A (Leu14Gln) (1), A→C (Thr39Pro) (1), ΔT at nt 464, Δ520–530 (1), Δ127–402 (1)
mutT mutant	0.5	14	nfxB (43)	A→C (Ser36Arg) (1), A→C (His57Pro) (1), A→C (Gln131Pro) (2), T→G (Leu122Arg) (1), T→G (Phe177Cys) (1)
			ND (57)
	1.0	20	nfxB (70)	A→C (Ser36Arg) (2), A→C (His57Pro) (1), A→C (Thr145Pro) (2), A→C (Gln175Pro) (2), A→C (His179Pro) (4), Δ341–351 (3)
			gyrA (5)	G→T (Asp87Tyr) (1)
			nfxB/gyrA (10)	A→C (Ser36Arg)/A→C (Asp87Ala) (1), A→C (Thr145Pro)/A→C (Glu153Ala) (1)
			ND (15)
	2.0	12	gyrA (83)	G→A (Gly81Asp) (1), C→T (Thr83Ile) (6), G→A (Asp87Asn) (2), G→T (Asp87Tyr) (1)
			ND (17)
mutS mutant	0.5	13	nfxB (31)	T→C (Pro26Leu) (1), A→G (His87Arg) (1), T→C (Leu88Pro) (1), T→G (Met176Arg) (1)
			gyrA (15)	C→T (Thr83Ile) (1), G→A (Glu153Lys) (1)
			ND (54)
	1.0	10	gyrA (40)	C→T (Thr83Ile) (4)
			nfxB/gyrA (10)	C→T (Ala38Val)/G→A (Asp87Asn) (1)
			gyrA/parC (50)	C→T (Thr83Ile)/A→G (Tyr31Cys) (1), C→T (Thr83Ile)/C→T (Ser87Leu) (3), C→T (Thr83Ile)/A→G (Glu91Gly) (1)
	2.0	10	gyrA (90)	C→T (Thr83Ile) (9)
			gyrA/parC (10)	C→T (Thr83Ile)/G→A (Val107Met) (1)

^aThe percentage of clones mutated in nfxB and in the QRDR of gyrA and/or parC that were selected at different CIP concentrations is indicated. “ND” indicates clones with no mutations in any of these genes.

^bThe specific mutations and the number of occurrences of the mutation determined in each case are shown. nt, nucleotide; ins, insertion; Δ, deletion.

The nfxB mutants derived from the wild-type strain showed several types of mutations including deletions, insertions, and substitutions. Among the mutS-derived clones, several types of transition mutations were detected in nfxB, while gyrA mutations were almost exclusively a C→T transition leading to Thr83Ile substitution in GyrA. As observed in this and other studies (16), Thr83Ile mutation in GyrA appears to confer the highest levels of CIP resistance. In addition, this is the mutation most frequently selected among CIP-resistant clinical isolates (16, 18). On the other hand, among the mutT-derived clones, the majority of nfxB and gyrA mutations detected at 0.5 and 1 μg ml⁻¹ CIP were the typical A→C transversions observed in a mutT mutator background (29). Alternatively, the gyrA mutations selected at 2 μg ml⁻¹ CIP were mainly base transitions (essentially C→T transition leading to Thr83Ile substitution).

Interestingly, the percentage of gyrA parC double mutants among the resistant variants isolated from mutS strain at 1 and 2 μg ml⁻¹ CIP was higher than it would be expected considering the probability of occurrence of two independent mutational events. The enhanced percentage of double mutants is the most likely cause of the significant increase in MPC observed for this strain. Further experiments are being carried out to evaluate the definitive cause of this phenomenon.

Low-level ciprofloxacin resistance is mainly associated with the overexpression of efflux pumps.

Around half of the clones selected at 0.5 μg ml⁻¹ CIP had no detectable mutations in any of the sequenced genes for both mutator strains (Table 1). These subgroups of variants were analyzed to determine if MexAB-OprM, MexXY-OprM, and MexEF-OprN efflux pumps were overexpressed. To test this, the sequences of the transcriptional regulators of these efflux pumps (MexR and MexZ repressors and MexT activator, respectively) were determined. As indicated in Table 2, mexZ was not mutated in any of the analyzed clones, while mexT was mutated in all the clones. However, because these mexT mutations were also found in parental wild-type, mutT, and mutS MPAO1 strains, it is unlikely that they are responsible for selection at 0.5 μg ml⁻¹ CIP.

Table 2.

Sequence analysis of genes encoding the transcriptional regulators MexZ, MexT, and MexR

MPAO1 strain	CIP concn (μg ml−1)	Change(s) in gene encoding transcriptional regulatora:
		mexZ	mexT	mexR
Wild type	0	None	Δ229–236; T→A (Phe172Ile)	None
mutT mutant	0	None	Δ229–236; T→A (Phe172Ile), T→C (Ile337Thr)	T→G (Leu95Arg)
	0.5	None	Δ229–236; T→A (Phe172Ile), T→C (Ile337Thr)	T→G (Leu95Arg)
mutS mutant	0	None	Δ229–236; T→A (Phe172Ile), T→C (Ile337Thr)	None
	0.5	None	Δ229–236; T→A (Phe172Ile), T→C (Ile337Thr)	17% C→T (Arg91Cys)

^amexZ, mexT, and mexR sequences were determined for six CIP-resistant variants isolated at 0.5 μg ml⁻¹ CIP that were not mutated in nfxB or in the topoisomerase genes. Except for the C→T transition, the mutations shown were observed in 100% of the analyzed variants. The sequence analysis of parental MPAO1 strains is also included.

On the other hand, mexR was mutated in 17% of the mutS-derived variants (C→T transition leading to Arg91Cys substitution). Unexpectedly, a mexR mutation (T→G [Leu95Arg]) was found in 100% of the mutT-derived variants, including the parental mutT strain. A comparative analysis of the mutT, wild-type, and mutS strains carrying a luminescent reporter demonstrated that this mexR mutation increased the activity of the mexAB-oprM promoter (see Fig. S1-a in the supplemental material). Notably, however, the MexAB-OprM pump is not fully expressed in the parental mutT strain. When this luminescent reporter was introduced into the mutT- and mutS-derived clones selected at 0.5 μg ml⁻¹ CIP, which were not mutated in nfxB or the topoisomerase genes, we found that 2 out of 5 mutT variants (8 and 10 in Table S2 in the supplemental material) and 2 out of 4 mutS variants (6 and 7 in Table S2) exhibited an increased activity of the mexAB-oprM promoter (relative to each parental strain) (see Fig. S1-b and -c). This observation agrees with previous data demonstrating that MexR is not the only regulator of mexAB-oprM expression (39).

Clones with decreased susceptibilities to β-lactams and macrolides prevail among subpopulations selected at low CIP concentrations.

The sensitivity to antibiotics of different classes was assayed for clones isolated from mutT and mutS strains at 0.5 and 2 μg ml⁻¹ CIP using antibiogram tests (see Materials and Methods). The diameter of the growth inhibition areas and the criteria assumed for susceptibility comparisons in this study are described in Table S2 in the supplemental material. In relation to the parental mutT strain, the percentage of mutT-derived variants showing a reduced susceptibility to at least one of the β-lactam antibiotics assayed was 54% for the clones selected at 0.5 μg ml⁻¹ CIP and 14% for those selected at 2 μg ml⁻¹ CIP. The corresponding percentages for clones with decreased susceptibilities to macrolides were 73% and 14%. For the mutS-derived variants, percentages of clones showing reduced susceptibilities to β-lactams and macrolides were also higher for clones selected at 0.5 μg ml⁻¹ CIP (55% for both antibiotics in comparison to 28% for those selected at 2 μg ml⁻¹ CIP). No significant difference was observed regarding the sensitivity to aminoglycosides and colistin.

Most of the clones selected at 0.5 μg ml⁻¹ CIP that showed decreased susceptibilities to β-lactams were not nfxB mutants, and they are likely to be overproducers of the MexAB-OprM efflux pump (see Table S2 in the supplemental material). In fact, some of the clones selected from the mutS strain at 0.5 μg ml⁻¹ CIP harbored mutations in mexR, which were absent in the parental strain (Table 2). NfxB mutants, on the contrary, showed increased susceptibilities to some β-lactam antibiotics, such as imipenem.

Low ciprofloxacin concentration accelerates the emergence of nfxB mutants on solid medium.

Considering the high proportion of cells growing at 0.5 μg ml⁻¹ CIP for mutT and mutS strains (approximately 0.1 to 1% of the total population) (Table 3) and the percentage of nfxB mutants selected at this CIP concentration (Table 1), it is noteworthy that nfxB mutants represent a clearly high percentage of the overall bacterial population (between 0.04 and 0.3%). In order to rule out the possibility of incorporation of preexisting nfxB mutants before overnight incubation, we reduced cell inocula to between 10 and 50 cells and repeated the determination of CIP resistance frequency at 0.5 μg ml⁻¹ CIP. The resistance frequencies determined in this way did not differ from those previously calculated using 5 × 10³ cells in the inoculum. Hence, mutations arise either during overnight growth in liquid LB media or upon cell plating onto CIP plates. To discern between these possibilities, we analyzed the occurrence of nfxB mutants when wild-type and hypermutator strains were grown in plates containing nonselective LB medium or with a CIP concentration that allowed the growth of 70 to 77% of the total CFU applied. To detect nfxB mutants, we constructed a chromosomal luminescent reporter of mexCD-oprJ expression. As a control assay, we introduced the reporter fusion into several nfxB mutant clones previously identified, which clearly showed a higher luminescence than a nonmutant strain (Fig. 2a). As shown in Fig. 2b, the intensity of luminescence was low, homogeneous, and stable among colonies grown in nonselective media for all of the reporter strains. In contrast, colonies with increased luminescence were detected on plates containing CIP (Fig. 2c and d), and their proportion on the plates increased with incubation time. This increased luminescence affected either the whole colony or a small external sector of them (mixed colonies) (Fig. 2c and d). As summarized in Table S3 in the supplemental material, nfxB sequencing analysis of the most luminescent clones arising on each plate indicated that none of the LB-derived clones was mutated in nfxB, while those selected at subinhibitory CIP concentrations harbored nfxB mutations. These mutations were mainly deletions or insertions for clones derived from the wild-type strain: A→C transversions for those derived from the mutT strain and G→A or T→C transitions for the mutS-derived variants. Interestingly, sequence chromatograms showed that for the majority of mutated colonies, each mutation affected a fraction of cells at each colony, as exemplified in Fig. 2f. Moreover, when mixed colonies were further isolated in LB media, they were composed of two distinct populations of cells, with either low or strong luminescence (Fig. 2e). Sequence analysis indicated that the most luminescent population was mutated in nfxB, while the other was not (data not shown). The percentages of nfxB mutants arising in LB plates with or without CIP are shown in Fig. 3.

Table 3.

Fractions of cells recovered in ciprofloxacin-containing plates for wild-type and hypermutable P. aeruginosa strains

CIP concn (μg ml−1)	Fraction of cells recovered for MPAO1 straina
	Wild type	mutT mutant	mutS mutant
0.5	(1.1 ± 0.6) × 10−7	(1.2 ± 0.2) × 10−3	(2 ± 1) × 10−2
1.0	(2 ± 1) × 10−10	(8 ± 0.8) × 10−6	(1.5 ± 0.4) × 10−5
2.0	<1 × 10−11	(2 ± 1) × 10−10	(3 ± 2) × 10−7

^aThe values shown represent the mean of at least 10 independent measurements for each strain and CIP concentration.

Fig. 2.

Effect of ciprofloxacin (CIP) at subinhibitory levels on nfxB mutagenesis. MPAO1 wild-type, mutT, and mutS reporter strains carrying a chromosomal transcriptional fusion of the mexCD-oprJ promoter to the lux operon were used to identify nfxB mutants. (a) Representative luminescence signal of an nfxB mutant (above) and the parental mutT strain (below) carrying the reporter construction. Approximately 300 CFU/plate were grown on LB media without (b) or with (c and d) CIP at subinhibitory concentrations (0.06 and 0.4 μg ml⁻¹ for the wild-type and mutator strains, respectively). Colonies with increased luminescence were observed from day 2 at the CIP plates and, after further isolation on LB medium, were shown to be composed of a mixed population (e). Sequence analysis of nfxB from mixed colonies also indicated the presence of both mutated and not mutated cells in the same colony. Insets in panels c and d present an expanded view of some mixed colonies. A representative chromatogram for one of the colonies grown on CIP plates is shown in panel f; here A→C transversion leading to Lys37Gln substitution does not affect the entire colony population (the residue C resulting from this mutation is underlined).

Fig. 3.

Percentage of nfxB mutants arising in solid media with or without the addition of ciprofloxacin (CIP) at subinhibitory levels (gray bars). These percentages are relative to the total bacterial populations, and their numerical values are also indicated inside the bars. The percentages of cells recovered in relation to the total number of cells applied to the plates are also indicated (white bars). In this assay, 20 independent cultures of wild-type, mutT, and mutS strains carrying a luminescent reporter of mexCD-oprJ expression were plated in solid media with or without the addition of CIP at subinhibitory levels (approximately 300 cells at each plate). After 48 h of growth, about 6 to 10 of the most luminescent clones derived from each strain and condition were further isolated in LB medium to determine its nfxB DNA sequence. The nfxB mutations detected in this assay are detailed in Table S3 in the supplemental material.

We conclude that exposure to subinhibitory CIP concentrations on solid media promotes nfxB mutation, possibly by a general mutagenic process, a phenomenon that was already observed for ciprofloxacin and other antibiotics (12, 15).

DISCUSSION

The rapid emergence and selection of resistant P. aeruginosa clones upon treatment with fluoroquinolones significantly reduce the efficacy of these antibiotics. In order to improve drug therapies, it is important to evaluate the effect of drug doses on the number and nature of resistant variants selected, especially considering the case of hypermutator strains. Parameters such as the MIC and the mutant prevention concentration (MPC), which define a concentration window in which resistant mutants are selected, help to address questions regarding effective drug dosing and drug resistance (45).

MIC₉₉ and MPC of CIP obtained for the MPAO1 wild-type strain in this work were 0.08 and 1.25 μg ml⁻¹, respectively. Results presented in this study indicated that mutT and mutS MPAO1 hypermutator strains show mutant selection windows shifted to higher CIP concentrations in relation to the wild-type strain. Both hypermutators exhibited higher MIC₉₉ values (∼6-fold) than the wild-type strain, possibly due to the inclusion in these values of the resistant mutant subpopulation, which is higher among mutator populations (24, 36). It is noteworthy that the parental mutT strain used in this study was originally mutated in mexR. However, since the expression of a mutT copy from a plasmid in this strain restored the wild-type levels of cell recovery in CIP, it is unlikely that the mexR mutation affects the mutant selection window of the mutT strain.

Another important result is that the hypermutator strains showed higher MPC values relative to the wild-type strain. Whereas the mutT strain exhibited a 2-fold-higher MPC, the mutS strain showed a 4-fold-increased MPC, defining a much broader window for selection of resistant mutants.

CIP-resistant subpopulations were previously characterized with clinical isolates of P. aeruginosa (14). For fluoroquinolone-susceptible isolates, the average MIC₉₉ and MPC for CIP were 0.2 and 3 μg ml⁻¹, respectively. MPC values ranged mainly between 0.5 and 8 μg ml⁻¹. Considering the high percentage of mutS strains among P. aeruginosa isolates in chronic infections and the results presented here, it is likely that subpopulations with hypermutator phenotypes may be responsible for the higher MPC values observed among clinical isolates (14).

As observed with other bacterial species (8, 47), resistance to CIP for the P. aeruginosa strains analyzed in this work showed a stepwise dependence on drug concentration. As drug concentrations increased, each successive plateau region correlated with the selection of particular resistant genotypes conferring higher levels of resistance. For the clones isolated at the bottom of the mutant selection window (CIP concentrations near the MIC), resistance was mainly associated with derepression of the MexCD-OprJ efflux pump, achieved through mutations in the transcriptional repressor NfxB. In addition, the analysis of hypermutator clones selected near the MIC that were not mutated in nfxB indicated that about 40% of them overexpressed MexAB-OprM efflux pump in relation to each parental strain. On the other hand, overexpression of MexXY-OprM and MexEF-OprN efflux pumps or mutations in the topoisomerase genes did not appear to be prevalent mechanisms of resistance at low CIP doses. These results are in agreement with previous studies of different bacterial species, including Pseudomonas, Rhodococcus, and Mycobacteria, in which high percentages of resistant clones not mutated in the topoisomerase genes were selected with fluoroquinolones at concentrations near the MIC (24, 33, 47). Those low-level-resistant subpopulations are likely to be composed of clones overexpressing efflux pumps, as was observed in an in vivo model of P. aeruginosa infections (24).

In this study, when the hypermutator strains were challenged with increasing CIP doses, mutations in the quinolone resistance-determining region (QRDR) of the target genes (mainly gyrA mutations) became the most prevalent resistance mechanisms selected. High recovery of gyrA mutants among subpopulations selected at high concentrations of fluoroquinolones was also observed in Mycobacterium and Rhodococcus (33, 47). This phenomenon may be due to the capacity of mutations in the topoisomerases to confer higher levels of fluoroquinolone resistance relative to derepression of the efflux pump alone, as was also observed in different large-scale studies with clinical P. aeruginosa isolates (16, 18). That gyrA mutants were not detected among subpopulations selected near the MIC can be explained by the much smaller proportion of gyrA mutants in the population relative to mutations in the transcriptional repressors of efflux pumps.

The most common resistance mutation observed at the top of the selection window for both hypermutators was the Thr83Ile substitution in GyrA achieved by C→T transition. This mutation is the most frequent fluoroquinolone resistance mutation found among clinical isolates, and it also confers the highest levels of resistance (16, 18). Given that C→T transitions are overrepresented in a mutS-deficient background, it is likely that this GyrA mutation is generated more frequently in the mutS strain, widening the plateau region between 1 and 2 μg ml⁻¹ in the mutant selection window for this strain. In addition, the number of plateau regions or allelic diversity of resistant variants (proportionally significant in the population) appeared to be larger for the mutS strain. These results are the probable consequence of the higher mutation rate of the mutS strain (31) and its corresponding potential to generate clones mutated in two or more genes conferring CIP resistance, which may be able to resist higher CIP concentrations.

Notably, the percentage of the gyrA parC double mutants observed among mutS-derived resistant clones was unexpectedly high, representing 50% of clones selected at 1.0 μg ml⁻¹ CIP (about one double mutant each 10⁵ viable cells) (Table 3). Results presented in a recent work suggest that this phenomenon may be related to an increase in bacterial fitness of gyrA mutants by the fixation of an additional mutation in parC (26). Alternatively, mutations in a single topoisomerase gene may sustain an initial growth of cells in CIP plates, allowing the fixation of further mutations at a higher frequency in the presence of CIP.

By comparing the susceptibilities to other antibiotics for the resistant clones selected from mutT and mutS strains at the bottom or the top of the mutant selection window (0.5 and 2.0 μg ml⁻¹ CIP), we found that a greater proportion of clones with decreased susceptibilities to β-lactams and macrolides were isolated at the bottom of the window. This observation is in agreement with the selection of a high proportion of resistant clones containing mutations that derepress the MexCD-OprJ and MexAB-OprM efflux pumps at CIP concentrations near the MIC. Multidrug-resistant phenotypes among subpopulations selected with low levels of fluoroquinolones were also observed for Mycobacteria and in an in vivo model for Pseudomonas infections (24, 47).

Measurement of the transcriptional derepression of the mexCD-oprJ promoter among colonies developed in solid media (with or without subinhibitory levels of CIP) and the subsequent determination of the nfxB sequence indicated that CIP exposure stimulates the emergence of nfxB mutants. In addition, analysis of the activity of mexAB-oprM promoter in wild-type, mutT, and mutS strains indicated that in the presence of subinhibitory levels of CIP, the frequency of colonies showing a substantial increase in this activity was 5-fold higher than that in the absence of the drug (data not shown). When these resistant clones were isolated in drug-free media, they maintained an increased activity of mexAB-oprM promoter, indicating that they are generated by a mutagenic process. Thus, CIP may be inducing a transient increase in the general mutation rate at subinhibitory concentrations. Fluoroquinolones and other drugs were shown to promote mutagenesis in P. aeruginosa and other bacterial species (12, 15). This phenomenon was associated with DNA oxidation and damage due to an oxidative burst caused by antibiotics (2, 3, 19, 20), mutagenic repair of that damage by means of RecA-dependent recombination (41), and the induction of mutagenic polymerases that participate in the SOS response (32, 37, 44). In the present work, colony staining with diaminobenzidine indicated that colonies growing on LB plates supplemented with CIP accumulated considerably more hydrogen peroxide than those developed in the absence of antibiotic (data not shown). The increase in the frequency of nfxB mutants that we observed on CIP-containing plates might arise from oxidative stress induced by this antibiotic. Further experiments are being conducted in order to discern the dependence of this mutagenic effect of CIP on the oxidative stress, RecA activity, and the induction of the SOS response.

In the present work, we have analyzed for the first time the behavior of P. aeruginosa hypermutant populations exposed to a broad range of CIP concentrations. This constitutes an important set of findings because hypermutator clones are present in high proportion among isolates in chronic infections and because CIP is commonly used in the treatment of infections caused by P. aeruginosa. The general effect of hypermutability on the mutant selection window here described may be extensive in other antibiotic classes and bacterial species. Moreover, this work provides a test of the mutant selection window hypothesis: the presence of subpopulations of first-step mutants is predicted to shift MPC to higher drug concentration, as we demonstrated. The selection window concept is important because in principle it can guide therapy to restrict the emergence of resistance (10, 22, 46). Our results also stress the importance of avoiding low antibiotic doses. We have shown that exposure to subinhibitory CIP concentrations promote the generation of resistant clones that overexpress MexCD-OprJ and MexAB-OprM efflux pumps in P. aeruginosa and that these resistant clones emerge at a higher frequency in hypermutator strains.