Real-Time Monitoring of nfxB Mutant Occurrence and Dynamics in Pseudomonas aeruginosa Biofilm Exposed to Subinhibitory Concentrations of Ciprofloxacin

ABSTRACT

Biofilm infections caused by Pseudomonas aeruginosa are frequently treated with ciprofloxacin (CIP); however, resistance rapidly develops. One of the primary resistance mechanisms is the overexpression of the MexCD-OprJ pump due to a mutation in nfxB, encoding the transcriptional repressor of this pump. The aim of this study was to investigate the effect of subinhibitory concentrations of CIP on the occurrence of nfxB mutants in the wild-type PAO1 flow cell biofilm model. For this purpose, we constructed fluorescent reporter strains (PAO1 background) with an mCherry tag for constitutive red fluorescence and chromosomal transcriptional fusion between the P_mexCD promoter and gfp leading to green fluorescence upon mutation of nfxB. We observed a rapid development of nfxB mutants by live confocal laser scanning microscopy (CLSM) imaging of the flow cell biofilm (reaching 80 to 90% of the whole population) when treated with 1/10 minimal biofilm inhibitory concentration of CIP for 24 h and 96 h. Based on the observed developmental stages, we propose that nfxB mutants emerged de novo in the biofilm during CIP treatment from filamentous cells, which might have arisen due to the stress responses induced by CIP. Identical nfxB mutations were found in fluorescent colonies from the same flow cell biofilm, especially in 24-h biofilms, suggesting selection and clonal expansion of the mutants during biofilm growth. Our findings point at the significant role of high-enough antibiotic dosages or appropriate combination therapy to avoid the emergence of resistant mutants in biofilms.

INTRODUCTION

Chronic biofilm infections are recognized as a serious and increasing health care problem. As many as 60% of infections are caused by biofilm-growing microorganisms, having high costs for society and the quality of life of patients. Biofilms represent a fertile ground for the development of antibiotic resistance due to their persistence and, consequently, repeated antibiotic treatment without eradication of the bacteria (1).

Pseudomonas aeruginosa is an important opportunistic pathogen that causes persistent infections that are difficult to eradicate. Examples of such infections include chronic urinary tract infection due to catheters, chronic wounds, ventilated-associated pneumonia in intubated patients, and chronic pulmonary disease in patients with cystic fibrosis or chronic obstructive lung disease. P. aeruginosa has particular capabilities to adapt in stressful environments, especially by developing antibiotic resistance.

Biofilms display the phenomenon of antibiotic tolerance, which means reduced susceptibility to antibiotics due to biofilm-specific features such as physiological heterogeneity (2). It is distinct from antibiotic resistance, which arises via mutational mechanisms or horizontal gene transfer and thus is genetically determined (3). The antibiotic tolerance phenomenon has been the preferred focus in studies concerning the failure of treatment of biofilm infection, while conventional genetically determined resistance mechanisms remain relatively unexplored.

In general, mutational processes play the most important role in the acquired resistance of P. aeruginosa, with a single mutation often leading to high-level, clinically relevant resistance (4). There is a growing body of evidence that biofilm growth provides a specific niche for mutational events. For instance, endogenous oxidative stress in biofilms of P. aeruginosa has been shown to promote mutagenesis via the mutagenic repair of double-strand breaks in DNA (5). In addition, it is becoming clear that low-dose antibiotic treatment itself promotes the development of mutational resistance (6, 7). Conditions of low antibiotic levels might arise due to inappropriate dosages but also due to different distributions depending on the body compartments (6). For example, depending on the route of administration (intravenous or oral versus inhaled), P. aeruginosa would be exposed to low levels of antibiotics in different parts of the body (8).

Antibiotics, and especially fluoroquinolones, have been shown to promote genetic variation in planktonic cells by means of different pathways involving oxidative stress, the SOS response, or acting directly on DNA (7).

Ciprofloxacin (CIP), the only antipseudomonal drug available for oral administration, is extensively used to treat biofilm infections caused by P. aeruginosa; however, resistance to this antibiotic is easily acquired (4). CIP resistance can be mediated by mutations in the target genes encoding DNA gyrase (gyrA and gyrB) or DNA topoisomerase IV (parC and parE) and/or the efflux systems, for example, MexCD-OprJ (4). A variety of substrates can be accommodated by MexCD-OprJ, including antibiotics used in clinics, such as fluoroquinolones, macrolides, and glycylcyclines (9), as well as biocides like triclosan and chlorhexidine, dyes, detergents, and organic solvents (10). Efflux-mediated resistance due to the overexpression of MexCD-OprJ is commonly found in resistant P. aeruginosa isolates from chronic lung infection in cystic fibrosis (CF) patients and occasionally in non-CF patients (11–13).

The expression of the mexCD-oprJ operon is negatively regulated by the NfxB repressor protein. Due to the activity of NfxB, the mexCD-oprJ operon is normally silent in wild-type cells. So far, the only basis for the overexpression of the pump known is mutation of the nfxB gene (11). Different spectra of mutations occur in the entire length of the gene and lead to a compromised activity of the NfxB repressor (14).

The unique features of MexCD-OprJ pump regulation make it suitable for the reporter system, where the switch to the reporter signal is a loss-of-function mutation in nfxB (14). A reporter system based on the transcriptional fusion between P_mexCD and the lux operon was applied previously to study the mutational spectrum in different genetic backgrounds of planktonic P. aeruginosa (14) as well as the emergence of resistance during exposure to subinhibitory concentrations of CIP on solid medium (15).

nfxB mutants have been isolated from in vitro biofilm models after azithromycin treatment (16) as well as from a mouse model of chronic lung infection (17). It is worrisome that such mutants were also detected in biofilms treated with pharmacokinetic/pharmacodynamic (PK/PD)-adjusted doses of ciprofloxacin (18). It has been suggested that the overexpression of efflux pumps is the first step in CIP resistance development in biofilms (18). When such first-step mutants in the mutator background are fixed in the population, they seem to promote the emergence of quinolone target mutations (gyrA or gyrB) (18). So far, the genetic basis of CIP resistance, in both biofilm (16–18) and planktonic (14) P. aeruginosa populations, has been investigated by the selection of mutants on CIP-containing plates.

In the present study, we decided to investigate the emergence of nfxB mutants in real time during P. aeruginosa biofilm treatment with subinhibitory concentrations of CIP. For this purpose, we constructed a fluorescent transcriptional reporter for nfxB mutant detection and employed biofilm growth in flow cell reactors, where the biofilm grows attached to the glass surface in a flow cell chamber (19). This model allowed direct in situ visualization of nfxB mutant development by means of confocal laser scanning microscopy (CLSM).

We show for the first time the fast emergence of nfxB mutants at the single-bacterial-cell level in P. aeruginosa biofilms exposed to subinhibitory concentrations of CIP. The observed mutant cells exhibit stress-induced phenotypes (filaments) and are able to form microcolonies that rapidly overtake the biofilm population.

RESULTS

Construction and testing of P. aeruginosa reporter strains for nfxB mutant detection.

The fluorescent reporter strains were constructed in two major steps. First, wild-type strain PAO1 and its isogenic PAOΔnfxB (PAONB) strain overexpressing MexCD-OprJ were tagged with mCherry at the chromosomal attB site by using a mini-CTX vector (20). The transcription of mCherry is driven by the P_trc promoter, which is derived from trp and lac promoters; thus, mCherry is constitutively expressed in P. aeruginosa unless the LacI repressor is provided in trans. Such tagging alleviated the need for additional staining prior to CLSM imaging of the biofilms. We then integrated our mini-Tn7 construct harboring P_CD-gfp+. This reporter was derived from the mini-Tn7-P_CD-lux vector by inserting the green fluorescent protein (GFP) gene gfp+ (21) instead of the lux operon downstream of the P_CD promoter. gfp+ encodes a stable GFP-positive (GFP+) variant with increased fluorescence intensity due to mutations improving the folding efficiency and the fluorescence yield of GFP (21). Consistent with our expectations, the positive-control nfxB knockout strain gave rise to highly fluorescent colonies, while wild-type PAO1 had only a basal level of green fluorescence, probably resulting from a basal level of expression or autofluorescence (Fig. 1A). When quantified in planktonic cultures, the level of green fluorescence was ∼10 times higher in the ΔnfxB strain overexpressing MexCD-OprJ, with the difference becoming even more pronounced in the stationary phase due to the accumulation of the protein (Fig. 1B and C).

FIG 1.

Characteristics of the P. aeruginosa fluorescent reporter strains. (A) Colonies of PAO1 and its isogenic ΔnfxB strain with the P_CD-gfp+ reporter grown on LB agar plates and examined under an epifluorescence microscope with blue excitation light (green fluorescence) and green excitation light (red fluorescence). (B) The green fluorescent signal together with the OD₆₀₀ was monitored during 24 h of planktonic growth of PAO1 and the ΔnfxB mutant with the P_CD-gfp+ reporter. Raw fluorescence values were divided by the corresponding OD₆₀₀ at each time point. The fluorescence of parent strains (resulting from the autofluorescence background) was subtracted from the fluorescence of reporter strains. The means and 95% confidence intervals based on 5 replicates are shown. (C) Growth curves of the reporter strains. The values represent the means and 95% confidence intervals of data for five replicate wells in a microtiter plate, where the cultures were grown and monitored.

Before moving to the biofilm models, we first tested the validity of the reporter system in planktonic cultures by population analysis. The frequencies of nfxB mutants within CIP-resistant subpopulations were <0.006% on 0.1 mg/liter, 6% on 0.2 mg/liter, 85% on 0.5 mg/liter, and 55% on 1 mg/liter CIP. No growth occurred on 2 mg/liter CIP, and no nfxB mutants were observed on plates with 0.05 mg/liter CIP and without any CIP. We verified the presence of different spectra of nfxB mutations in all 22 fluorescent isolates selected from plates with different concentrations of CIP. In contrast, all 6 nonfluorescent colonies had a nonmutated nfxB gene, proving that the reporter enables the effective detection of nfxB mutants (see Table S1 in the supplemental material).

Low-dose CIP treatment leads to rapid development of nfxB mutants in PAO1 flow cell biofilms.

In order to investigate the effect of CIP treatment on PAO1 biofilms, we employed a well-established in vitro model: biofilm growth in flow cell reactors. The major goal of using this model was to visually monitor the evolution of the biofilm during CIP treatment using CLSM. We hypothesized that mutants would appear in the upper layer of the biofilm, where the cells are metabolically active. To determine how the response depends on the age of the biofilm, we employed two different experimental setups: a 24-h biofilm treated for 24 h and a 72-h biofilm treated for 96 h with the same concentration of CIP, 0.2 μg/ml. This concentration represents 1/10 of the minimal biofilm-inhibitory concentration established in vitro by PK/PD-adjusted concentrations (18).

CLSM images of 24-h PAO1-mCherry-P_CD-gfp+ biofilms prior to CIP treatment in all the imaging areas showed only red fluorescence resulting from the constitutive expression of mCherry (Fig. 2). In contrast, the biofilm formed by the positive-control strain PAOΔnfxB-mCherry-P_CD-gfp+ was brightly green due to the expression of gfp+ (see Fig. S2 in the supplemental material). Even though the sizes of the inoculum were similar (∼5 × 10⁶ CFU), the ΔnfxB mutant biofilm was significantly (P = 0.002) thinner than was the wild-type biofilm. The biomass of the 24-h biofilm formed by PAOΔnfxB-mCherry-P_CD-gfp+ was 0.5083 ± 0.2653 μm³/μm² (mean ± standard error of the mean [SEM]), compared to 2.402 ± 0.03630 μm³/μm² for PAO1-mCherry-P_CD-gfp+. As expected, CIP treatment for 24 h at subinhibitory concentrations did not significantly decrease the biomass of either PAOΔnfxB-mCherry-P_CD-gfp+ (0.3505 ± 0.03914 μm³/μm² [mean ± SEM]) or PAO1-mCherry-P_CD-gfp+ (3.279 ± 0.8554 μm³/μm² [mean ± SEM]). However, green fluorescent cells, indicative of nfxB mutants, grown as microcolonies surrounded by single and extremely elongated wild-type cells with red fluorescence were observed (Fig. 2). In most cases, nfxB mutants formed small aggregates, but we also found quite huge multicellular structures (∼80 μm in height) that were likely to develop into mushroom-shaped structures. We did not detect any cells with increased green fluorescence or filamentous cells in an untreated control PAO1 biofilm grown for 48 h.

FIG 2.

Rapid development of nfxB mutants in a 24-h-old PAO1 flow cell biofilm treated with low-dose CIP. The biofilms of PAO1-mCherry-P_CD-gfp+ were grown in flow cell chambers with a continuous flow of minimal medium for 24 h at 37°C and then grown for an additional 24 h with 0.2 μg/ml CIP (24 h +CIP for 24 h) or without CIP (24 h −CIP for 24 h). Red shows wild-type cells (elongated or filamentous) due to the constitutive expression of mCherry, and green shows nfxB mutants due to the expression of GFP via the P_CD-gfp+ reporter. z-stacks were generated by using a Zeiss LSM 710 microscope and processed with Imaris 8.2 software (Bitplane). Top-row images show orthogonal 3D views, and middle and bottom (for treated biofilms only) rows show perspective 3D views of biofilms with an overlay of red and green fluorescence. The images shown are representative of results for two independent flow cell channels.

The data from the above-described experiment did not clarify the location in the biofilm where the nfxB mutants initially appeared. As older biofilms have been shown to be better differentiated in metabolically active and dormant subpopulations (22), we hypothesized that treatment of older 72-h biofilms would serve better to determine the location of mutagenesis. However, this experiment also could not reveal in which part of the biofilm the nfxB mutants occur. We observed the occurrence of single filamentous cells, few of which already had green fluorescence 24 h after treatment initiation. Different levels of GFP+ accumulation were visible in different filaments, suggesting that these cells acquired mutations at different time points (Fig. 3). Later, brightly green nfxB mutants, whose cell size returned to normal, gradually overtook the biofilm population, representing up to 80 to 90% of the harvested population after 4 days of treatment (Fig. 4). The sequence of the developmental steps was the same for all three independently observed PAO1 biofilms (see Fig. S4 in the supplemental material), even though there was a delay in microcolony formation in one of them, probably due to the stochastic nature of mutability. Importantly, we did not observe filaments in any of the biofilms formed by the positive-control strain (PAOΔnfxB-mCherry-P_CD-gfp+) (Fig. S3).

FIG 3.

Development of nfxB mutants in a 72-h-old PAO1 flow cell biofilm during treatment with low-dose CIP. The biofilms of PAO1-mCherry-P_CD-gfp+ were grown in three independent channels of flow cell chambers with a continuous flow of minimal medium for 72 h and then treated with 0.2 μg/ml CIP for a total of 96 h. Imaging by CLSM was done every 24 h. At least 3 images were taken per channel at every time point. Red represents wild-type cells due to the constitutive expression of mCherry, and green shows nfxB mutants due to the expression of GFP+ via the P_CD-gfp+ reporter. z-stacks were generated by using a Zeiss LSM 710 microscope and processed with Imaris 8.2 software (Bitplane). The images show orthogonal 3D biofilm views (left) or a perspective view (right) with an overlay of red and green channel fluorescence.

FIG 4.

Fraction of nfxB mutants in the 72-h-old PAO1 flow cell biofilm population at different time points of CIP treatment. Shown are the fractions of biomass expressing gfp in 72-h biofilms formed by PAO1-mCherry-P_CD-gfp+ and PAOΔnfxB-mCherry-P_CD-gfp+ at different time points of treatment with CIP. The positive control had significantly higher biomass with green fluorescence at time zero (*, P < 0.0001) and at 48 h (**, P = 0.01) than that of the wild type. After 72 h and 96 h of CIP treatment, the fraction of green fluorescent biomass in the PAO1-mCherry-P_CD-gfp+ (representing nfxB mutants) biofilm approaches that of the positive control PAOΔnfxB-mCherry-P_CD-gfp+.

PAO1 nfxB mutants selected in flow cell biofilms have low-level ciprofloxacin resistance and are likely to have a clonal origin.

The fluorescent isolates from harvested biofilms exhibited a stable phenotype upon restreaking onto Luria-Bertani (LB) plates, confirming that the green fluorescence seen in the flow cell biofilms was not due to a transient biofilm-dependent induction of the pump. The MICs of CIP for the fluorescent isolates were in the range of 0.5 to 1.5 μg/ml, which are 5 to 15 times higher than those of wild-type strain PAO1 and approach the MIC of the PAOΔnfxB strain (Table 1). This suggested that no additional mutations in target genes (e.g., gyrA, gyrB, or parC) had been acquired during treatment, as they would usually confer high-level resistance. In general, MICs for isolates of younger biofilms were slightly lower (14/20 isolates had MICs of 0.5 to 0.75 μg/ml, and 6/20 isolates had MICs of 1.0 μg/ml) than those for 72-h biofilms (10/15 with MICs of 1.0 μg/ml and 3/15 with MICs of 1.5 μg/ml).

TABLE 1.

Characteristics of isolates from CIP-treated PAO1-mCherry-P_CD-gfp+ flow cell biofilms^e

Biofilma	Isolate	MIC (μg/ml)b	Mutation	Effect on NfxB protein
FCB1 (24 h + CIP for 24 h)	FCB 2	0.75	Ins 259A	Frameshift, no stopc
	FCB 3	0.75	Ins 259A	Frameshift, no stop
	FCB 4	0.75	Ins 259A	Frameshift, no stop
	FCB 5	0.75	Ins 259A	Frameshift, no stop
	FCB 6	0.75	Ins 259A	Frameshift, no stop
	FCB 7	0.75	Ins 259A	Frameshift, no stop
	FCB 8	0.75	Ins 259A	Frameshift, no stop
	FCB 1	0.75–1.0	Ins 259A	Frameshift, no stop
FCB2 (24 h + CIP for 24 h)	FCB 9	0.75	A600C	Stop>Cys
	FCB 10	0.75	A151C	Thr51Pro
	FCB 12	0.75	A151C	Thr51Pro
	FCB 11	0.75–1.0	A151C	Thr51Pro
	FCB 13	1.0	A151C	Thr51Pro
	FCB 14	1.0	A151C	Thr51Pro
	FCB 15	1.0	Δ(115) 1 bp	Frameshift, early stop (46 aa)
	FCB 16	1.0	A151C	Thr51Pro
FCB3d (24 h + CIP for 24 h)	FCB 17	0.38–0.5	A600C	Stop>Cys
	FCB 18	0.5	A600C	Stop>Cys
	FCB 19	0.5	A151C	Thr51Pro
	FCB 20	0.75	A600C	Stop>Cys
FCB4 (72 h + CIP for 96 h)	FCB 26	1.0	T440C	Phe147Ser
	FCB 27	1.0	Δ(164–165) 2 bp	Frameshift, no stop
	FCB 28	1.0	Δ(485–491) 7 bp	Frameshift, no stop
	FCB 29	1.0	Δ(485–491) 7 bp	Frameshift, no stop
	FCB 30	1.0	C509A	Ala170Glu
FCB5 (72 h + CIP for 96 h)	FCB 31	0.5	A600C	Stop>Cys
	FCB 34	0.5	A600C	Stop>Cys
	FCB 32	1.0	Δ252 (1 bp)	Frameshift, early stop (94 aa)
	FCB 35	1.0	Δ252 (1 bp)	Frameshift, early stop (94 aa)
	FCB 33	1.5	G381A	Trp127Stop
FCB6 (72 h + CIP for 96 h)	FCB 36	1.0	G143A	48 Ser>Asn
	FCB 37	1.0	G143A	48 Ser>Asn
	FCB 39	1.0	G143A	48 Ser>Asn
	FCB 38	1.5	G381A	Trp127stop
	FCB 40	1.5	G143A	48 Ser>Asn

^aIndependent biofilms grown for 24 h or 72 h and then treated with 0.2 μg/ml CIP for 24 h or 96 h, respectively.

^bMIC of CIP as determined by Etest.

^cStop codon not detected in the sequenced region of the nfxB amplicon.

^dCombination of two biofilms.

^eIns, insertion; aa, amino acids.

The types of acquired nfxB mutations in colonies exhibiting an increased-fluorescence phenotype are presented in Table 1. Identical mutations were found in isolates from the same biofilm, especially in 24-h biofilms, suggesting selection and clonal expansion. Many detected mutations led to frameshifts, which either introduced the early stop codon or resulted in the extended NfxB variants. Point mutations led to amino acid changes in positions known to be critical for DNA binding (Ser48Asn and Thr51Pro) or multimerization (Phe147Ser), as observed previously by Purssell and Poole (23).

DISCUSSION

The unique regulation of the MexCD-OprJ pump by loss-of-function mutations in the repressor harboring nfxB makes it an attractive reporter system to study the evolutionary pathways of CIP resistance in P. aeruginosa. Previously, a chromosomal luminescence-based reporter was used to confirm the presence of the nfxB mutation in selected planktonic isolates prior to sequencing or to compare the spectra of spontaneous and mutagen-induced mutations (14). However, it has not been applied for biofilm evolutionary studies, which are of great importance considering the relatively unexplored area of antibiotic resistance development in biofilms.

For the first time, we successfully applied a chromosomal GFP-based reporter to study the occurrence of nfxB mutants in an in vitro biofilm model of P. aeruginosa. Previously, there were attempts to assess the level of basal MexCD-OprJ expression in PAO1 biofilms by using a P_CD-gfp fusion on a high-copy-number plasmid (24). However, we argue that a chromosomal reporter that is stably integrated into the chromosome is superior to a plasmid-based reporter due to several reasons. First, the expression of mexCD-oprJ is subject to tight and strict regulation. Thus, the presence of only one additional copy of the NfxB binding site (P_CD promoter) rules out the possibility that NfxB could be titered out by the need of binding to multiple copies of the P_CD promoter, which would be likely if the reporter was present on a high-copy-number plasmid. In this case, the increase in the reporter signal could not be directly related to the inactivation of the NfxB repressor. Furthermore, there is no need for antibiotic pressure for reporter maintenance, as the mini-Tn7 element is stably integrated at a neutral chromosomal site (25). Such a feature is crucial when the effects of antibiotic treatment are investigated, as was the case for this study.

The combination of a fluorescent reporter and live CLSM imaging enabled us to observe the emergence and development of nfxB mutants during treatment with CIP at subinhibitory concentrations. The major finding of this study is that exposure to low doses of CIP leads to the rapid emergence of nfxB mutants in biofilms, with mutants comprising 80 to 90% of the biofilm population after 4 days of treatment. The development of these mutants followed certain stages in the flow cell biofilm model, and these stages were dependent on the age of the biofilm. While we observed microcolonies formed by nfxB mutants already after 24 h of CIP treatment in younger (24-h-old) biofilms, an additional stage of only filamentous cells, a few of which had already started to accumulate GFP, was seen preceding the formation of microcolonies by nfxB mutants in 72-h-old biofilms. Red filamentous cells were present in both cases, suggesting that this is a common initial step. Contrary to the previous belief that filamentous cells represent severely sick and dying members of bacterial populations, it has been found that different pathogenic bacteria, for example, uropathogenic Escherichia coli, Mycobacterium tuberculosis, and Legionella pneumophila (filamentation observed in in vitro biofilms), employ filamentation as a survival strategy under harsh conditions, including but not limited to antibiotic treatment (26). Moreover, the filaments seem to provide the environmental niche for mutagenesis, as was recently observed for E. coli (27). In that study, when single E. coli cells were challenged with subinhibitory concentrations of CIP, the induction of the SOS response led to filamentation. Time-lapse studies showed that these filaments gave rise to resistant cells of a normal size, leading the authors of that study to propose that filamentation is an initial stage in CIP resistance development (27). The SOS response has also been shown to have a crucial role in CIP-induced mutagenesis in P. aeruginosa (28). The filamentation phenotype is exerted in a process dependent on Lon protease via the action of the SulA protein, which is a cell division inhibitor (28). Therefore, our direct observations of CIP-resistant nfxB mutant development in P. aeruginosa biofilms, taken together with previously reported data on planktonic E. coli and P. aeruginosa, suggest the possibility of SOS-mediated resistance development in P. aeruginosa biofilms during treatment with CIP. However, we cannot exclude that the observed stress filamentation is SOS independent. In order to further elucidate the specific mechanism, time-lapse studies monitoring the same filament in a biofilm over the course of treatment would be useful, especially if we could couple it with a reporter for the SOS response or other stress responses.

Characterization of nfxB mutants isolated in this study shows that various different types of mutations in different positions of nfxB lead to a compromised activity of NfxB, as determined by fluorescent reporter signals as well as resistance to CIP.

The effect of CIP treatment suggests that mutation of nfxB is the preferred primary mechanism for low-level CIP resistance in biofilms. The observation that the MICs for nfxB mutants isolated from younger biofilms were slightly lower than those for 72-h biofilms can be explained by different mutational spectra of nfxB in these two groups of isolates (Table 1). This is in accordance with data from previous studies that showed that the antibiotic resistance phenotype depends on the degree of derepression of the MexCD-OprJ pump caused by various changes in NfxB (29). However, we did not observe the emergence of high-level resistance conferred by mutations in DNA gyrase or DNA topoisomerase IV in PAO1 biofilms. In favor of this model, it was shown previously that a first-step mutation in nfxB provides the ground for second-step mutations in gyrA or gyrB only in the PAOΔmutS mutator and not in wild-type biofilms (18). The isolation of first-step resistant PAO1 nfxB mutants after CIP treatment has also been documented in a mouse model of lung infection (17). nfxB mutants were also isolated after P. aeruginosa treatment with low doses of the macrolide antibiotic azithromycin, which is also a substrate for MexCD-OprJ (16). In a planktonic evolution study of P. aeruginosa with subinhibitory concentrations of CIP, nfxB mutants were also detected but did not seem to be fixed in the evolved populations. In that particular experimental setup, target mutations in gyrA and gyrB were primarily selected, leading to high-level resistance (30).

It has been shown that the overexpression of the MexCD-OprJ pump impairs all forms of motility and alters virulence, among other physiological changes (13, 31, 32). In addition, it might cause increased pumping of needed metabolites because of the requirement for proton motive force (31). All of these observations point to a degree of impaired fitness of nfxB mutants compared to wild-type P. aeruginosa. In this study, we also observed that the ΔnfxB strain formed a significantly thinner biofilm than the one formed by the wild-type PAO1 strain. However, in planktonic experimental evolution studies, it has been shown that evolved resistant mutants do not have a lower fitness than the ancestor population (30, 33), suggesting the occurrence of compensatory mutations. Taking into account the heterogenous biofilm environment, the compensatory mutations required for persistence and a decrease of the fitness cost would probably be different in biofilms than in planktonic populations. Identification of the compensatory mutations would require whole-genome sequence analysis of the persisting resistant mutants.

Nevertheless, the fact that these mutants are frequently isolated from CF lungs (12) and are occasionally also found in non-CF patients (13) provides evidence that they are still pathogenic and might have an underappreciated role in the primary stages of resistance development, especially in biofilms, as we have seen in our study. With the perspective of future studies, our fluorescence-based reporter could also be applied to evolutionary studies in vivo in animal models. In addition, the live dynamics of nfxB mutant development could be monitored under different treatment regimes, for example, using combination treatment or monitoring the effect of antibiotic withdrawal.

The effects of nonlethal, low-dose antibiotic treatment had been neglected for a long time. In this study, we used subminimal biofilm-inhibitory concentrations of CIP that quickly selected for nfxB mutants overtaking the biofilm population. The results of this study provide evidence that the effects of low-dose treatment are relevant not only to planktonically grown bacteria but also to biofilms, and a minimal selective concentration (34) should also be determined for biofilm-growing bacteria. Altogether, our findings point at the significant role of high-enough antibiotic dosages or, alternatively, appropriate combination therapy to avoid the emergence of resistant mutants in biofilms.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

Escherichia coli and Pseudomonas aeruginosa strains as well as plasmids used in this study are shown in Table 2. P. aeruginosa strains from frozen glycerol (20%) stocks were cultured on LB agar (2%) plates and incubated aerobically overnight at 37°C. For some experiments, a trace element solution (Btrace) with or without 2% agar was used: 0.1 ml of trace metals (200 mg/liter CaSO₄·2H₂O, 200 mg/liter FeSO₄·7H₂O, 20 mg/liter MnSO₄·H₂O, 20 mg/liter CuSO₄·5H₂O, 20 mg/liter ZnSO₄·7H₂O, 10 mg/liter CoSO₄·7H₂O, 12 mg/liter NaMoO₄·H₂O, 5 mg/liter H₃BO₃) was added to 1 liter of ultrapure H₂O. E. coli strains from frozen glycerol stocks were grown on ABTG agar plates (Btrace, 2% agar, 10% A-10 phosphate buffer [pH 6.7], thiamine [25 mg per liter of medium], 0.2% glucose, 0.5% Casamino Acids, and trace elements [1/1,000]). Growth medium and conditions specific for the experiment are indicated below. When needed, the antibiotics gentamicin (GEN) (gentamicin sulfate; Sigma-Aldrich), tetracycline (TET) (tetracycline hydrochloride; Sigma-Aldrich), CIP (ciprofloxacin hydrochloride; Bayer), and ampicillin (AMP) (ampicillin sodium salt; Sigma-Aldrich) were used at the concentrations indicated for each experiment specifically.

TABLE 2.

Bacterial strains and plasmids used in this study^a

Strain or plasmid	Relevant characteristic(s)	Source or reference
Strains
E. coli
JM105	thi rpsL (Strr) endA sbcB15 sbcC hsdR4 (rK− mK−) Δ(lac-proAB) [F′ traD36 lacIq Δ(lacZ)M15 proA+B+]	Laboratory collection
JBAMG134	JM105/pJBA285; GENr	Jens Bo Andersen
GZA42	JM105/pGZA1; GENr	This study
T121	DH5α/miniCTX2T2.1.-GW::Ptrc-mCherry; TETr	20
P. aeruginosa
PAO1	Wild-type P. aeruginosa strain	17
PAONB (PAOΔnfxB)	PAO1 ΔnfxB::lox; constitutively expresses the MexCD-OprJ pump	16
PAO1-mCherry	PAO1 with mCherry integrated at the attB site via a mini-CTX vector; TETr marker removed via Flp-mediated recombination	This study
PAOΔnfxB-mCherry	PAOΔnfxB with mCherry integrated at the attB site via a mini-CTX vector; TETr marker removed via Flp-mediated recombination	This study
PAO1-mCherry-PCD-gfp+	PAO1-mCherry with mini-Tn7 carrying the transcriptional fusion of PCD-gfp+; GENr	This study
PAOΔnfxB-mCherry-PCD-gfp+	PAOΔnfxB-mCherry with mini-Tn7 carrying the transcriptional fusion of PCD-gfp+; GENr	This study
Plasmids
Mini-Tn7-PCD-lux	pUC18-mini-Tn7T-Gm-PCD-lux; GENr	15
pUX-BF13	Helper plasmid with transposase for integration of the mini-Tn7 element; AMPr	37
pHA51	Mini-CTX2T2.1.-GW::Ptrc-mCherry; TETr	20
pFLP2	Site-specific excision vector with Flp recombinase and sacB; AMPr CARr	38
pJBA285	Donor plasmid for RBSII-gfp+-T0; GENr	Jens Bo Andersen
pGZA1	pUC18-mini-Tn7T-Gm-PCD-gfp+; GENr	This study

^aAMP^r, ampicillin resistance; CAR^r carbenicillin resistance; GEN^r, gentamicin resistance; TET^r, tetracycline resistance; RBSII, strong ribosomal binding site; T₀, transcriptional terminator derived from phage λ; P_CD, mexCD-oprJ promoter with an NfxB binding site.

Construction of P. aeruginosa fluorescent reporter strains. (i) Tagging of strains with mCherry and removal of the TET^r marker.

The mini-CTX-based mCherry (20) integration vector was electroporated into PAO1 and PAOΔnfxB (PAONB) electrocompetent cells. Electrocompetent cells were prepared by washing with 0.3 M sucrose as described previously (25). A total of 500 ng of the mini-CTX vector was electroporated into 100 μl of electrocompetent cells by using an E. coli Pulser Transformation apparatus (Bio-Rad) with settings of 25 mF, 200 W, <5 ms, and 2.5 kV. Phenotypic expression was performed with LB medium overnight at 37°C. Transformants were selected on LB agar plates supplemented with 100 to 150 μg/ml TET. The transcription of mCherry upon integration into the chromosome is driven by the strong P_trc promoter, which is derived from trp and lac promoters; thus, in the absence of LacI, mCherry is constitutively expressed. Successful transformants exhibiting red fluorescence were subjected to the removal of the TET resistance (TET^r) marker. Flp-mediated excision of TET^r and subsequent curing from pFLP2 by counterselection on 10% sucrose were done as described previously (25). A slight modification was needed for the ΔnfxB strain, which is hypersusceptible to β-lactam antibiotics, including carbenicillin (CAR), which was used to select successful transformants harboring the Flp-encoding pFLP2 plasmid. Therefore, for this strain, 50 μg/ml instead of 200 μg/ml CAR was used.

(ii) Construction of delivery vectors with P_CD-gfp reporters.

Mini-Tn7-based delivery vector pGZA1 containing a P_CD-gfp+ transcriptional fusion was constructed as follows. First, the RBSII-gfp+-T₀ (RBSII is a strong synthetic ribosomal binding site, and T₀ is a strong transcriptional terminator derived from phage λ) fragment was PCR amplified from pJBA285 carrying gfp+. The composition of the PCR mixture was as follows: 32.5 μl of MilliQ water, 10 μl of 5× Phusion HF buffer, 1 μl of 10 mM deoxynucleoside triphosphates (dNTPs), 2.5 μl of 10 μM forward primer RBSII-gfp(up)-fw and reverse primer Gfp(down)-rev (Table 3), ∼10 ng of the template plasmid (noted above), and 0.5 μl of 2 U/μl Phusion High Fidelity DNA polymerase (Thermo Scientific, Lithuania). The PCR cycling settings were an initial denaturation step at 98°C for 3 min; 35 cycles of denaturation at 98°C for 10 s, annealing at 66.9°C for 20 s, and extension at 72°C for 15 s; and a final extension step at 72°C for 5 min. The amplified RBSII-gfp+-T₀ fragment was purified by using the Wizard SV Gel and PCR cleanup system (Promega) according to the protocol provided by the manufacturer. The purified fragment was digested with the restriction endonucleases KpnI (Fermentas) and SmaI (Thermo Scientific) sequentially (total reaction volume, 100 μl): a 3-h digestion with 2× KpnI (20 U) in KpnI buffer was first performed at 37°C, the reaction mixtures were then heat inactivated at 80°C for 20 min, and 4× SmaI (40 U) was added for further digestion for 3 h at 30°C. The mini-Tn7-P_CD vector backbone was derived from the pUC18-miniTn7T-Gm-P_CD-lux vector by digesting it with KpnI and SmaI to yield a 4.3-kb fragment ready for the insertion of RBSII-gfp-T₀. Sticky-end ligation of the gel-purified 4.3-kb mini-Tn7-P_CD vector backbone and the RBSII-gfp+-T₀ fragment to yield pGZA1 plasmids was done in a total volume of 20 μl with the following reaction mixture composition: 100 ng of the mini-Tn7-P_CD backbone, 63 or 100 ng of the RBSII-gfp+-T₀ insert (molar ratio of the insert to the vector of 3:1 or 5:1), 2 μl of 10× T4 DNA ligase buffer (Thermo Scientific), 1 μl T4 DNA ligase (Fermentas), and MilliQ water. Reactions were carried out at room temperature (18°C) overnight, and ligase was then inactivated at 65°C for 20 min. The ligation reaction mixture was transformed into JM105 chemically competent cells by using a standard heat shock transformation protocol for E. coli to obtain GZA42. Successful transformants were selected on low-salt LB (4 g/liter NaCl instead of 10 g/liter) agar plates supplemented with GEN at 10 μg/ml. The pGZA1 plasmid was purified (QIAprep Spin Miniprep kit; Qiagen) and first subjected to restriction endonuclease digestion analysis with KpnI and SmaI. The correct sequence of the P_CD-gfp+ insert in pGZA1 was verified by sequencing (Macrogen Inc. sequencing service) of both strands by using four primers: two primers flanking the inserted gfp+ fragment [PmexCD_fw and Gfp_T₀(rev)new1] and two primers internal to the reporter [Gfp_fw_1 and Gfp_(rev)]. Primer sequences can be found in Table 3. The map of the constructed delivery plasmid is provided in Fig. S1 in the supplemental material.

TABLE 3.

Primers used in this study

Primer	Sequence (5′–3′)	Description
RBSII-gfp(up)-fw	ATATACCCGGGTCTAGAATTAAAGAGGAGAAATTAAGCATG	RBSII-gfp-T0 amplification (forward)
Gfp(down)-rev	TATATGGTACCCTCCTGAAAATCTCGCCAAGCTAGC	RBSII-gfp-T0 amplification (reverse)
Pmexcd_fw	TTGCTGTTGACAAAGGGAATC	Sequencing
Gfp_fw_1	CAGTGGAGAGGGTGAAGGTGA	Sequencing
Gfp-To(rev) new1	TCCTGAAAATCTCGCCAAGC	Sequencing
Gfp_(rev)	CCATAACCGAAAGTAGTGACAAGTG	Sequencing
nfxB fw3	TGACACACCCGACCGTTG	nfxB amplification and sequencing
nfxB rev1	TCGGTCCGTGCCATGC	nfxB amplification and sequencing
PglmS-down	GCACATCGGCGACGTGCTCTC	Verification of correct mini-Tn7 element integration
PTn7R	CACAGCATAACTGGACTGATTTC	Verification of correct mini-Tn7 element integration

(iii) Integration of the mini-Tn7-P_CD-gfp+ construct into mCherry-tagged strains.

Delivery plasmid pGZA1 was coelectroporated with helper plasmid pUX-BF13 into mCherry-tagged electrocompetent cells of PAO1 and PAOΔnfxB according to established procedures (25). Transformants were selected on LB agar plates supplemented with 30 μg/ml GEN. The correct integration of the mini-Tn7 element was confirmed by PCR with primer pair Pglms-down and PTn7R (Table 3), which amplified a 272-bp product in the case of correct chromosomal integration of the reporter downstream of the glmS gene. The composition of the PCR mixture was the same as the one described previously (25), and cycling was done with an initial denaturation step at 96°C for 5 min, 35 cycles of denaturation at 96°C for 45 s, annealing at 60°C for 45 s, and extension at 72°C for 45 s; and a final extension step at 72°C for 10 min.

Testing of the P_CD-gfp+ reporter in planktonic cultures.

Cultures of fluorescent reporter strains grown overnight (PAO1-mCherry-P_CD-gfp+ and PAOΔnfxB-mCherry-P_CD-gfp+) were diluted 100× in defined ABtrace medium (450 ml Btrace, 10% A-10 phosphate buffer [pH 6.7], 0.2% glucose, 0.5% Casamino Acids, 0.01 mM FeCl₃), which has lower autofluorescence than LB medium. PAO1-mCherry was used as a control for cellular autofluorescence. Diluted cultures were transferred (200 μl) to five replicate wells in a black 96-well microtiter plate with a flat bottom (Nunc, Thermo Fisher Scientific). The optical density at 600 nm (OD₆₀₀) and GFP+ (green) fluorescence (excitation 485 nm and emission 535 nm) were measured for 24 h (68 cycles in total, with shaking) at 37°C by using an Infinite F200 Pro plate reader (Tecan). The measurements were controlled via the Magellan V 7.2 program. Relative fluorescence was calculated as the ratio of fluorescence/OD₆₀₀, and the background green fluorescence of PAO1-mCherry was subtracted at every time point.

Population analysis of planktonic PAO1 cultures.

Cultures of PAO1-mCherry-P_CD-gfp+ grown overnight were diluted (10⁻⁷) in fresh LB medium and grown for 18 h at 37°C with shaking (200 rpm). The initial inoculum was kept low (∼500 cells) in order to avoid incorporating preexisting nfxB mutants from the culture grown overnight. Grown cultures were serially diluted in saline (0.9% NaCl), and appropriate dilutions were plated onto antibiotic-free LB plates (for total CFU counts) or supplemented with different CIP concentrations, 0.05, 0.1, 0.2, 0.5, 1.0, and 2.0 μg/ml, to recover CIP-resistant subpopulations. After 24 h (or 48 h if the colonies were too small to count) of incubation at 37°C, recovered colonies were counted.

Green fluorescence was observed under an epifluorescence microscope (Leitz Aristoplan) with blue excitation light (488 nm; fluorescein isothiocyanate [FITC] filter). The percentage of nfxB mutants in different CIP-resistant subpopulations was calculated as (number of gfp+ colonies/total number of recovered colonies) × 100%. The ratio of the number of fluorescent colonies to the total number of cells applied to the CIP plates was calculated to determine the number of nfxB mutants that can be selected from the total population. Randomly selected colonies (five fluorescent and two without fluorescence) recovered with different CIP concentrations were restreaked onto LB agar plates to confirm the stable fluorescent/nonfluorescent phenotype. Selected isolates were frozen as glycerol stocks (LB medium with 20% glycerol) at −80°C.

Biofilm experiments in flow cell reactors. (i) Growth of the flow cell biofilms.

Biofilms were cultivated in flow cells (chamber with an attached glass coverslip) with three parallel channels (dimensions of the channel of 1 by 4 by 40 mm), which were directly attached to flexible silicone tubing (1 mm to 2 mm in diameter), connecting the flow cell to the medium supply upstream and a Watson Marlow 250S peristaltic pump downstream. The assembled system was sterilized by pumping a 0.5% sodium hypochlorite solution and then washing the system with sterile water. A continuous flow of minimal ABtrace medium (for 1 liter of medium, 10% of A-10 phosphate buffer, 1 ml of 1 M MgCl₂, 0.1 ml 1 M CaCl₂, and 0.1 ml of trace metals solution) supplemented with 0.3 mM glucose was maintained overnight at 37°C to acclimatize the system prior to the inoculation of bacteria. The flow cell channels were directly inoculated with PAO1-mCherry-P_CD-gfp+ and positive-control PAOΔnfxB-mCherry-P_CD-gfp+ strains (∼300 μl of cultures grown overnight diluted to an OD₆₀₀ of 0.01) by using a sterile syringe. The flow cells were left upside down without the flow for 1 h to allow the attachment of bacteria to the glass coverslip. A constant flow of medium (3 ml/h) was then maintained by using a Watson Marlow 250S peristaltic pump (Watson Marlow, UK). Replicate biofilms of each strain were grown in 2 to 4 independent flow cells.

(ii) Flow cell biofilm treatment with ciprofloxacin and CLSM imaging.

After 24 h (experiment 1) or 72 h (experiment 2), grown biofilms were challenged with 0.2 μg/ml CIP (corresponding to ∼2× MIC of planktonic PAO1) by switching the medium supply. Antibiotic treatment was continued for 24 h (experiment 1) or 96 h (experiment 2). The biofilms were nondestructively imaged every 24 h (flow cells directly mounted onto the microscope stage) by using a Zeiss LSM 710 confocal laser scanning microscope (Plan-Apochromat 63×/1.40 oil differential interference contrast [DIC] objective; for GFP+ fluorescence, excitation at 488 nm and emission peak at 509 nm; for mCherry, fluorescence excitation at 594 nm and emission peak at 610 nm). Imaging was done at an ∼5-mm distance from the inlet in the flow cell chamber. z-stacks were generated with 1-μm intervals; the number of slices was chosen individually for each imaged area to cover the whole biofilm. At least three z-stacks were taken for each flow cell channel and time point. Minimal z-stack image processing (smoothing, background subtraction, and different three-dimensional [3D] views) was done by using Imaris 8.2 software (Bitplane) and ImageJ (NIH) to crop the area of interest from the whole imaging area.

(iii) Isolates from treated flow cell biofilms.

To isolate nfxB mutants for further characterization, PAO1-mCherry-P_CD-gfp+ biofilms were harvested by pumping 1 to 1.5 ml of a glass bead (212 to 300 μm; Sigma)-saline suspension through the flow cell channels and collected into sterile Eppendorf tubes. The biofilms from each channel (three biofilms per experiment) were isolated separately, unless otherwise specified. Appropriate dilutions of biofilm suspensions were plated onto antibiotic-free LB agar plates, and colonies were examined under an epifluorescence microscope (Leitz Aristoplan) for green fluorescence (FITC filter cube). The frequency of nfxB mutants was expressed as the ratio (percentage) of green fluorescent colonies versus the total number of colonies examined. Depending on the biofilm, 4 to 8 colonies with increased green fluorescence and 1 to 2 nonfluorescent colonies from each individual biofilm were restreaked on LB agar (total of 35 fluorescent and 11 nonfluorescent colonies) to test the stability of the fluorescent phenotype. All tested isolates retained the original phenotype and were frozen as glycerol stocks (LB medium, 20% glycerol) and kept at −80°C before further characterization.

(iv) Statistical analysis.

Quantification of the biomass was performed by using COMSTAT 2 (http://www.comstat.dk/) (35, 36). GraphPad Prism for Windows (version 6.04; GraphPad Software) was used for statistical analysis. A t test was used to analyze the difference in biomass between PAO1-mCherry-P_CD-gfp+ and PAOΔnfxB-mCherry-P_CD-gfp+, and 2-way analysis of variance (ANOVA) for multiple comparisons was used to analyze the fraction of the biomass expressing GFP at different time points of treatment with ciprofloxacin.

Characterization of biofilm isolates. (i) MIC of ciprofloxacin.

MIC values for selected isolates were determined by performing an Etest. Briefly, an inoculum of 10⁵ CFU/ml was spread onto blood agar plates (5% horse blood; SSI Diagnostica, Hillerød, Denmark), and CIP Etest strips (Liofilchem) were applied to the plates. Results were interpreted after overnight incubation at 37°C.

(ii) PCR amplification and sequencing of nfxB.

Chromosomal DNA was extracted (Qiagen) from selected isolates (57 fluorescent [35 biofilm and 22 planktonic] and 10 nonfluorescent [4 biofilm and 6 planktonic]) to amplify nfxB by using primers nfxB fw3 and nfxB rev1 and DyNAzyme Ext DNA polymerase (Thermo Fisher Scientific, Lithuania). The composition of the PCR mixture included 81 μl MilliQ water, 3 μl dimethyl sulfoxide (DMSO), 10 μl 10× Ext buffer (Thermo Fisher Scientific), 2 μl 10 mM dNTP, 1 μl each 50 μM primer, 1 μl template, and 1 μl Ext DNA polymerase. Cycling conditions included an initial denaturation step at 94°C for 3 min; 35 cycles of denaturation at 94°C for 30 s, annealing at 54°C for 30 s, and extension at 72°C for 1 min; and a final extension step at 72°C for 5 min. The sequencing of column-purified (Wizard SV Gel and PCR cleanup system; Promega) nfxB amplicons (amplicon size of 775 bp and open reading frame [ORF] size of 600 bp) was done with the same primers, nfxB fw3 and nfxB rev1, to sequence the entire ORF on both strands (Macrogen Inc. sequencing service). Sequencing analysis was done by using SnapGene software (GSL Biotech LLC) and pairwise sequence alignments (http://www.ebi.ac.uk/Tools/psa/). Note that the numbering of mutations is according to the recently annotated open reading frame of the functional nfxB gene (600 bp; 199 amino acid residues), which has 36 additional base pairs in the 5′ region and, consequently, 12 residues in the N terminus, as determined previously by Purssell and Poole (23).