Hypoxia Increases Antibiotic Resistance in Pseudomonas aeruginosa through Altering the Composition of Multidrug Efflux Pumps

Abstract

Antibiotic resistance is a significant and developing problem in general medical practice and a common clinical complication in cystic fibrosis patients infected with Pseudomonas aeruginosa. Such infections occur within hypoxic mucous deposits in the cystic fibrosis lung; however, little is known about how the hypoxic microenvironment influences pathogen behavior. Here we investigated the impact of hypoxia on antibiotic resistance in P. aeruginosa. The MICs of a selection of antibiotics were determined for P. aeruginosa grown under either normoxic or hypoxic conditions. The expression of mRNAs for resistance-nodulation-cell division (RND) multidrug efflux pump linker proteins was determined by real-time PCR, and multidrug efflux pump activity was inhibited using Phe-Arg β-naphthylamide dihydrochloride. The MIC values of a subset of clinically important P. aeruginosa antibiotics were higher for bacteria incubated under hypoxia than under normoxia. Furthermore, hypoxia altered the stoichiometry of multidrug efflux pump linker protein subtype expression, and pharmacologic inhibition of these pumps reversed hypoxia-induced antibiotic resistance. We hypothesize that hypoxia increases multidrug resistance in P. aeruginosa by shifting multidrug efflux pump linker protein expression toward a dominance of MexEF-OprN. Thus, microenvironmental hypoxia may contribute significantly to the development of antibiotic resistance in P. aeruginosa infecting cystic fibrosis patients.

INTRODUCTION

Hypoxia and inflammation are coincidental events in a number of chronic infectious diseases (16). The metabolic demands of infiltrating immune cells and multiplying pathogens along with vascular dysfunction associated with chronic inflammation contribute to tissue hypoxia under such conditions (4). However, rather than simply being a bystander feature of the microenvironment, hypoxia plays an important effector role in influencing gene expression in host cells and invading pathogens alike and can have a significant impact on the development of both infection and inflammation (17). While the impact of decreased oxygen tension on bacterial virulence has been investigated for intestinal pathogens such as Shigella flexneri (12), little is known about its effects on antibiotic resistance in Pseudomonas aeruginosa.

Cystic fibrosis (CF) is the most common severe autosomal recessive disease in Caucasians. Chronic lung disease is the major determinant of long-term survival in cystic fibrosis patients, and P. aeruginosa, an opportunistic pathogen, can cause life-threatening infections. Approximately 80% of adult cystic fibrosis patients suffer from pulmonary P. aeruginosa infection, which is associated with increased morbidity and mortality (15). In cystic fibrosis, P. aeruginosa grows within thick mucous secretions that accumulate within the airway lumen, and the low oxygen permeability of these biofilms results in the establishment of a hypoxic microenvironment (20).

Treatment of P. aeruginosa infections in cystic fibrosis is complicated by antibiotic resistance, and eradication of P. aeruginosa from the cystic fibrosis lung is generally unachievable in persistent infection (15). As the development of new antimicrobial agents has diminished over the last decades, antimicrobial resistance is a growing global problem, and new strategies to combat panresistant bacteria are required. A key mechanism of antibiotic resistance is the expulsion of antibiotics through multidrug resistance (MDR) efflux systems belonging to the resistance-nodulation-division (RND) family. These pumps play an important role in intrinsic and acquired multidrug resistance (14). Four such pumps, MexAB-OprM, MexCD-OprJ, MexEF-OprN, and MexXY-OprM, have been well characterized in P. aeruginosa.

While increased antibiotic resistance has been reported previously for P. aeruginosa grown under anaerobic conditions or in biofilms (19), in this study we investigated whether hypoxia, independently of the complex environment of the biofilm, influences the resistance of P. aeruginosa to antibiotics. A more detailed understanding of the mechanisms of P. aeruginosa antibiotic resistance in cystic fibrosis lung disease will identify new therapeutic targets for antimicrobial therapy.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The P. aeruginosa control strain (ATCC 27853) and clinical strains from chronically infected cystic fibrosis patients (S8263, S8269, S8276, S8279) were cultured in cation-adjusted Mueller-Hinton II broth (MHB II) (Becton Dickinson Microbiology Systems, Cockeysville, MD). The clinical strain S8263 was resistant to all routinely used antipseudomonal antibiotics (ceftazidime [TAZ], piperacillin-tazobactam [P-T], meropenem, aztreonam [AZT], tobramycin, gentamicin, amikacin, ciprofloxacin). The clinical strains S8269 and S8276 were susceptible to all tested antibiotics, whereas S8279 showed a mixed antibiotic susceptibility pattern. All clinical isolates were cultured from sputum specimens of cystic fibrosis patients infected by P. aeruginosa for at least 1 year. Bacteria were incubated at 30°C (for analysis of RND multidrug efflux pump expression) or 37°C (for antibiotic resistance) under normoxia (21% oxygen) or hypoxia (1% oxygen) in a hypoxia chamber (Invivo₂ 400 hypoxia workstation; Ruskinn Technology Limited, Bridgend, United Kingdom).

Antimicrobial susceptibility testing.

Antibiotic susceptibility testing was performed by broth microdilution using a Sensititre GNX2F susceptibility plate (Trek Diagnostic Systems, East Grinstead, United Kingdom). To determine the role of efflux pumps in antimicrobial susceptibility, the pump inhibitor Phe-Arg β-naphthylamide dihydrochloride (final concentration, 20 μg/ml; Sigma-Aldrich, Dorset, United Kingdom) was used (8). To inhibit hydroxylases, dimethyloxalylglycine (DMOG; 1 mM; Cayman Chemicals, Ann Arbor, MI) was used. Dimethyl sulfoxide (DMSO; Sigma-Aldrich, Dorset, United Kingdom) was used as a vehicle control.

Analysis of RND multidrug efflux pump expression levels.

Six milliliters of preconditioned MHB II medium was inoculated with 3 × 10⁸ bacteria and was then incubated for 6 to 24 h under either normoxia or hypoxia. Bacteria were harvested by centrifugation for 2.5 min at 15,000 rpm. Total RNA was isolated with the RiboPure Bacteria kit (Ambion, Texas) according to the manufacturer's instructions.

Real-time quantification of cDNA was carried out on an ABI Prism 7900HT sequence detection system (Applied Biosystems, Warrington, United Kingdom) using the Sybr green PCR master mix (2×). Primers for partial amplification of genes encoding the membrane fusion proteins MexA, MexC, MexE, and MexX were used as published previously (7). The PCR was run for 2 min at 50°C and 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 1 min at 60°C. Control samples without the cDNA template or reverse transcriptase were run in parallel. Data were normalized for each gene and are presented as mexE/mexC and mexE/mexA ratios.

Statistical analysis.

MIC values are presented as descriptive statistics, and MICs were compared with the Kruskal-Wallis test and Duncan posttest (GraphPad Prism, version 5) (10). RND multidrug efflux pump expression levels were compared using an unpaired 2-tailed Student t test. P values of <0.05 were considered statistically significant.

RESULTS

P. aeruginosa grown under hypoxia displays increased antibiotic resistance.

P. aeruginosa formed confluent monolayers on Mueller-Hinton plates during overnight incubation under either normoxia or hypoxia. Under hypoxic conditions, P. aeruginosa grew as colorless colonies rather than as standard green colonies, indicating impaired pyocyanin production during growth under hypoxic conditions (Fig. 1A). We investigated the susceptibility of P. aeruginosa to a range of antibiotics under hypoxia compared to normoxia by using broth microdilution plates. MIC values for 21 antibiotics following exposure to either normoxia or hypoxia were determined. Among antibiotics tested by the broth microdilution method, penicillin and cephalosporin antibiotics demonstrated increased MIC values under hypoxia compared to normoxia (Fig. 1B). These antibiotics included cefotaxime (FOT), TAZ, cefepime (FEP), AZT, ticarcillin-clavulanic acid (TIM), and P-T (Fig. 1B). Other antibiotics tested, including aminoglycosides, carbapenems, polymyxins, quinolones, and tetracyclines, showed similar MIC values in normoxic and hypoxic cultures (Fig. 1B).

Fig 1.

Altered antibiotic resistance in P. aeruginosa under hypoxic conditions. (A) Mueller-Hinton agar plates were inoculated with ATCC 27853, which was grown for 20 h in a normoxic (N) or a hypoxic (H) (1% oxygen) environment. (B) Determination of MICs for ATCC 27853 at 21% (N) and 1% (H) oxygen by broth microdilution of 21 antibiotics using a Sensititre GNX2F susceptibility plate. AMI, amikacin; GEN, gentamicin; TOB, tobramycin; FOT, cefotaxime; TAZ, ceftazidime; FEP, cefepime; AZT, aztreonam; TIM, ticarcillin-clavulanic acid; P/T, piperacillin-tazobactam; MERO, meropenem; ETP, ertapenem; DOR, doripenem; IMI, imipenem; COL, colistin; POL, polymyxin B; LEVO, levofloxacin; CIP, ciprofloxacin; DOX, doxycycline; MIN, minocycline; TGC, tigecycline; SXT, trimethoprim-sulfamethoxazole. Data are mean MICs ± standard errors of the means for 3 independent experiments.

Hypoxia differentially alters multidrug efflux pump linker protein expression in P. aeruginosa.

We have observed previously that hypoxia affects MDR1 gene expression and chemotherapeutic drug resistance in mammalian cells (5). The expression of bacterial multidrug efflux pumps is an antimicrobial resistance mechanism exploited by P. aeruginosa and affecting various antibiotic classes. Therefore, we investigated whether hypoxia-induced multidrug efflux pump expression may contribute to changes in antibiotic susceptibility in P. aeruginosa. For the analysis of pump expression, we used quantitative real-time PCR. A correlation between mRNA expression and efflux protein expression has been demonstrated previously (21).

The expression levels of four MDR efflux pumps were analyzed following a 6- to 24-h exposure to normoxia or hypoxia. Interestingly, the stoichiometry of multidrug efflux pump linker protein expression changed to a dominance of MexE expression over MexA and MexC expression following either 6 or 24 h of hypoxia (Fig. 2A and B). This was due primarily to decreased MexA and MexC expression, with MexE demonstrating sustained expression. Expression of the MexX linker protein gene could not be detected under these conditions (data not shown).

Fig 2.

Altered stoichiometry of RND multidrug efflux pump composition under hypoxic conditions. The expression of the genes (mexA, mexC, and mexE) encoding the linker molecules for the RND multidrug efflux pumps MexAB-OprM, MexCD-OprJ, and MexEF-OprN in P. aeruginosa under normoxia (N) and hypoxia (H) was investigated by real-time PCR. (A and B) ATCC 27853 was grown under N or H for 6 h (A) or 24 h (B). Expression of the mexA, mexC, and mexE genes was detected by real-time PCR. Relative gene expression was normalized and expressed as mexE/mexA and mexE/mexC ratios. Values are means ± standard errors of the means for 3 independent experiments (*, P < 0.05). (C) Clinical P. aeruginosa strains were exposed to N and H for 6 h, and expression of the genes encoding the linker molecules was detected by real-time PCR. Relative gene expression was expressed as mexE/mexA and mexE/mexC ratios. Values are means ± standard errors of the means for 2 independent experiments (*, P < 0.05).

To determine whether clinical P. aeruginosa strains isolated from cystic fibrosis patients demonstrated a similar change in pump isoform stoichiometry, we evaluated MexA, MexC, and MexE pump expression in response to hypoxia in four P. aeruginosa strains with different antibiotic susceptibility profiles isolated from cystic fibrosis patients. All cystic fibrosis patients had been chronically infected with P. aeruginosa for more than 1 year. Three of the four cystic fibrosis clinical isolates investigated in this study revealed similar changes in pump stoichiometry in response to hypoxia (Fig. 2C).

Increased MIC values in hypoxia are normalized by the multidrug efflux pump inhibitor Phe-Arg β-naphthylamide dihydrochloride.

To investigate the existence of a functional link between antibiotic resistance and changed efflux pump activity in hypoxia, we used Phe-Arg β-naphthylamide dihydrochloride, an inhibitor of RND efflux pumps (8). Phe-Arg β-naphthylamide dihydrochloride reversed the hypoxia-induced resistance of P. aeruginosa to a selection of antibiotics (Fig. 3). This led us to hypothesize that changes in efflux pump activity contributed to increased antibiotic resistance in hypoxia. MIC values in normoxia remained unchanged in the presence of the pump inhibitor (data not shown).

Fig 3.

Increased antibiotic resistance in hypoxia is reversed by a multidrug efflux pump inhibitor. MICs for ATCC 27853 were determined by broth microdilution using a Sensititre GNX2F susceptibility plate under normoxia (N), hypoxia (H), or hypoxia in the presence of the efflux pump inhibitor (H+EPI) Phe-Arg β-naphthylamide dihydrochloride (20 μg/ml). A selection of CF-relevant antibiotics for which the EPI decreased the MIC under hypoxia is shown. Values are means ± standard errors of the means for 3 independent experiments (*, P < 0.05).

Antibiotic susceptibility changes in hypoxia are not mediated by inactivation of hydroxylase activity.

In mammalian cells, a key signaling event in response to hypoxia is the inhibition of oxygen-dependent hydroxylases, leading to activation of the hypoxia-inducible factor (HIF) and downstream activation of HIF-responsive genes (9). Inhibition of such hydroxylases can be achieved by exposure of cells to DMOG, a pan hydroxylase inhibitor (6). While multiple hydroxylases have been described in several microbes, including P. aeruginosa (18), their role in oxygen sensing in the HIF pathway is thought to be restricted to metazoans (11). We investigated the question of whether hypoxia-induced changes in MIC values can be mimicked by exposure of bacteria to DMOG. There was no difference in antibiotic susceptibilities between untreated cells and cells treated with DMOG, indicating that 2-oxoglutarate-dependent hydroxylase inhibition did not participate in hypoxia-mediated changes in antibiotic susceptibilities (Fig. 4). These data are consistent with the report that HIF hydroxylases are not primary oxygen sensors in bacteria (11).

Fig 4.

Hydroxylase inhibition fails to increase antibiotic resistance. MICs for ATCC 27853 were determined by broth microdilution using a Sensititre GNX2F susceptibility plate in the presence of the hydroxylase inhibitor DMOG (1 mM) or the vehicle control (DMSO). Data are mean MICs ± standard errors of the means for 3 independent experiments.

DISCUSSION

Hypoxia has been shown to be an important modulator of the virulence of the intestinal pathogen Shigella flexneri (12). In chronic infection of the CF lung, P. aeruginosa produces alginate and grows in dense bacterial populations, forming mucoid biofilms (1). Complex microbial communities, including anaerobic bacteria, have been detected in sputa from CF patients (3). Although decreased oxygen tensions have been demonstrated in mucous layers in the CF lung, the effects of hypoxia on antibiotic resistance have not been investigated independently of biofilm formation. Our data show that exposure to hypoxia induces selective antibiotic resistance in P. aeruginosa. The antibiotic resistance of P. aeruginosa in response to hypoxia may be important not only for the treatment of infections in CF lung disease but also for the management of infections in other chronic respiratory diseases, such as bronchiectasis and chronic obstructive pulmonary disease (COPD). In these diseases, chronic inflammation with remodeling of lung tissue and excessive mucous production generates growth conditions comparable to those in CF, and not surprisingly, P. aeruginosa is one of the major bacterial pathogens encountered.

The analysis of RND efflux pump expression levels in P. aeruginosa revealed a shift in pump expression toward MexEF-OprN under hypoxic conditions in strain ATCC 27853 as well as in 3 out of 4 clinical isolates of P. aeruginosa from CF patients. The β-lactam antibiotics (aztreonam, ceftazidime), β-lactam–inhibitor combinations (piperacillin–tazobactam), tetracyclines, and trimethoprim have all been shown to be substrates for the MDR efflux pumps MexAB-OprM, MexCD-OprJ, and MexEF-OprN (14). Aminoglycosides are believed to be transported exclusively by MexXY-OprM (14), so it was not surprising that no effect of hypoxia on tobramycin susceptibilities was observed, since we could not detect MexXY-OprM. The failure of hypoxia to increase ciprofloxacin MICs was surprising given that ciprofloxacin had been described as a substrate for all three pumps: MexAB-OprM, MexCD-OprJ, and MexEF-OprN (14). One potential explanation for this phenomenon could be differences in the substrate specificities of pumps expressed by various P. aeruginosa strains. Since altered efflux pump expression is one of a number of antibiotic resistance mechanisms used by P. aeruginosa, we cannot exclude the possibility that other bacterial resistance mechanisms are also affected by hypoxia. Normalization of increased MIC values in the presence of an efflux pump inhibitor indicates that alterations in efflux pump activities are indeed associated with the observed changes in antimicrobial susceptibility.

Altered expression of bacterial RND efflux pumps has been seen in response to exposure to antimicrobials both in vivo and in vitro (2, 13), where antibiotic exposure selects for mutation events in genetic regulators suppressing pump expression. Clinical isolates with increased efflux pump expression but without identifiable regulatory mutations had been identified previously (2), and no inducers of efflux pump expression apart from substrate exposure have been described so far. Here we describe hypoxia as a novel regulator of RND efflux pumps in P. aeruginosa.

Hydroxylases are key regulatory elements in the cellular response of eukaryotic cells to hypoxia. In P. aeruginosa, pharmacologic hydroxylase inhibition failed to increase MIC values over those for normoxia, suggesting that the increased MIC levels in hypoxia are independent of hydroxylase activity.

Since secondary metabolites have been postulated to be the natural substrates for the pumps (14), it is possible that efflux pump genes contain hypoxia-responsive elements as part of the bacterial adaptive response to hypoxia. Thus, the extrusion of antibiotics via MDR efflux pumps in response to hypoxia could be a secondary effect of the adaptation of the bacterial metabolism to decreased oxygen values. Our finding suggests that bacterial isolates could appear sensitive in vitro on testing in the laboratory yet function as resistant organisms in vivo once they encounter hypoxic environments. This could be one factor explaining the failure of clinical treatment with appropriate antibiotics, which is observed frequently in the management of chronic respiratory infections.