Resistance and Virulence of Pseudomonas aeruginosa Clinical Strains Overproducing the MexCD-OprJ Efflux Pump

Pseudomonas aeruginosa PAO1, the genome of which has entirely been sequenced, was used as the reference wild-type strain throughout the study (23). EryR is a spontaneous mexCD-oprJ-overexpressing mutant of PAO1 selected on erythromycin (14). FE60 is a ΔmexXY derivative of PAO1 hypersusceptible to aminoglycosides (10). Plasmid pEXΔXYR (ticarcillin and gentamicin resistance), used previously to construct FE60, was employed here to inactivate genes mexXY in EryR (yielding mutant FK06), as described by El'Garch et al. (10). Two hundred five nonreplicate isolates of P. aeruginosa showing low-level resistance to ciprofloxacin (MICs ranging from 0.5 to 4 μg/ml) were collected between January and December 2006 from non-cystic fibrosis patients at the teaching hospital of Besançon (France). During this period of time, a total of 1,632 isolates of P. aeruginosa were isolated, of which 61.6% (n = 1,006) were susceptible to ciprofloxacin, 1.8% (n = 29) were intermediate, and 36.6% (n = 597) were resistant according to the CLSI breakpoints (6). One hundred ten of the 205 strains with decreased susceptibility to ciprofloxacin (53.6%) were further randomly selected to analyze MexCD-OprJ pump expression. The four MexCD-OprJ-overproducing strains, 3308 (serotype O:6), 2439 (O:5), 2126 (O:6), and 1956 (O:2), identified in this study were found to be genotypically distinct by the random amplified polymorphic DNA method (data not shown) (44). Bacterial culturing was performed at 37°C in liquid (Mueller-Hinton broth [MHB]) or on solid (Mueller-Hinton agar [MHA]) medium with adjusted concentrations of Ca²⁺ and Mg²⁺ (Becton Dickinson Microbiology Systems, Cockeysville, MD). Propagation of plasmid vector pMT45 (ticarcillin resistance) and its nfxB recombinant derivative pNF225 was achieved by adding 150 μg/ml ticarcillin to the media (52). Both plasmids were purified from strain PK1013E (52) and transferred into the P. aeruginosa strains by electroporation, as described by Choi et al. (4).

The MICs of selected antibiotics were determined by the conventional serial twofold dilution method in MHA by using a Steers multiple inoculator and bacterial inocula of approximately 10⁴ CFU per spot, as recommended previously (30). The inoculated plates were incubated at 37°C and examined visually after 18 and 24 h.

Overnight cultures in MHB were serially diluted and spread on the surface of MHA in order to obtain 10 to 20 distinct colonies per plate after 24 h of incubation at 37°C. Five colonies of each P. aeruginosa strain were subsequently collected with the agar underneath and individually transferred to tubes containing 10 ml sterile saline buffer. Once serially diluted, these suspensions were used to inoculate MHA plates for CFU counting.

Induced and uninduced AmpC β-lactamase activities (nmol nitrocefin·min⁻¹·mg protein⁻¹) for strains PAO1, EryR, 3308, 2439, 2126, and 1956 were determined by spectrophotometric testing of crude French press lysates, as previously described (18). Measurements were done in triplicate for each strain. Variations in the results from one experiment to another were less than 15%. Induction of AmpC β-lactamase activity was obtained by the addition of 50 μg/ml cefoxitin to the MHB medium 3 h prior to the end of incubation. It should be mentioned that preliminary isoelectrofocusing experiments performed on the crude lysates revealed only the presence of a faint band at a pI of ≥8.5, corresponding to AmpC in the selected strains (18).

The expression levels of operons mexAB-oprM, mexCD-oprJ, mexEF-oprN, mexGHI-opmD, mexJK, mexVW, and mexXY were assessed by reverse transcription-real-time PCR (RT-PCR) with the fluorescent dye Sybr green (Qiagen Sciences, MD) in a RotorGene RG3000 apparatus (Corbett Research, Sydney, Australia), essentially as described by Dumas et al. (9). The primers used for the amplication of genes mexB (MexB1, MexB2), mexC (MexC3, MexC4), mexE (MexE4, MexE5), mexG (MexG1, MexG2), mexJ (MexJ1, MexJ2), mexV (MexV1, MexV2), and mexY (MexY1A, MexY1B) have been reported elsewhere (9, 20, 27, 37). The gene expressions were normalized to that of the housekeeping gene uvrD (28). PAO1 was used as the wild-type control (values set at 1) and well-characterized gain-of-efflux derivatives as positive controls: PT629 for MexAB-OprM, EryR for MexCD-OprJ, PAO7H for MexEF-OprN, Mut-GR1 for MexXY, and PAO318 for MexJK (20). Induction of the mexXY operon in the P. aeruginosa strains and their transformants harboring either plasmid pMT45 or pNF225 (ticarcillin resistance) was achieved by adding chloramphenicol or gentamicin at one-half the MIC to the bacterial cultures for 4 h (27). Comparison of CFU counts on drug-free MHA and on MHA supplemented with ticarcillin at the end of the induction period in each experiment showed that more than 95% of the bacterial population still harbored pMT45 or pNF225 in the absence of selective pressure for plasmid maintenance. The RT-PCR data reported in the text are means of at least three independent determinations, with experiment-to-experiment variations being lower than 10%.

Plasmid DNAs were purified by using the Wizard Plus SV Minipreps kit (Promega France, Charbonnière, France) according to the manufacturer's protocol. To search for possible mutations, the nfxB gene was amplified with the pair of primers nfxB1 (5′-ACGCGAGGCCAGTTTTCT-3′) and nfxB2 (5′-ACTGATCTTCCCGAGTGTCG-3′) and then sequenced on both strands by using an AB Prism 3130 genetic analyzer (Applied Biosystems, Courtaboeuf, France) at the Institut Fédératif de Recherche IFR133, Besançon, France. The quinolone resistance-determining regions of genes gyrA, gyrB, parC, and parE were amplified and sequenced in the clinical strains, as reported by Llanes et al. (40). All the quinolone resistance-determining region sequences turned out to be identical to that of wild-type strain PAO1. Amplification of exoU and exoS genes was performed as described previously (3).

One-step spontaneous mutants were selected on MHA containing 1 μg per ml ciprofloxacin, from three genotypically different, drug-susceptible strains of P. aeruginosa, including the reference strain PA14 (55) and two previously characterized bacteremic isolates (3). The selective plates were inoculated with ca. 10⁸ log phase bacteria and incubated at 37°C for 24 h. Ten resistant clones from each strain were analyzed by RT-PCR to identify those overexpressing gene mexC. Alteration of gene nfxB in the selected clones was subsequently confirmed by nucleotide sequencing. Mutants KJ0707 (from PA14) and KJ0711 (from 19.1) were found to contain nonsense mutations in nfxB (A344G and T241C, respectively), while mutant KJ0705 (from 4.2) appeared to harbor a 8-bp deletion from position 137 to 144. In parallel, a new nfxB mutant of strain PAO1, named KJ0702, exhibiting a nonsense mutation (T154C) in nfxB, was isolated under the same conditions.

Bacterial membranes were isolated, subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and analyzed by Western blotting with MexY- and MexB-specific polyclonal antisera (diluted 1:20,000 and 1:1,000, respectively), as already reported (22).

Elastase activity was determined by an elastin Congo red (ECR) assay (61) of supernatants of cultures grown for 7 h in PB medium (11). Briefly, 50 μl of supernatant was added to 0.95 ml of 0.1 M Tris (pH 7.4) and 1 mM CaCl₂ with an excess of ECR (4-mg/ml reaction volume; Elastin Company, Owensville, MO). After 18 h of incubation at 37°C, samples were centrifuged and degradation of ECR was determined by measuring absorbance at 495 nm in the supernatant. Elastase activity was expressed as the ratio optical density at 495 nm (OD₄₉₅)/OD₆₀₀ of the culture.

Production of rhamnolipid was assessed on square-well plates (59) based on M9 minimal medium in which NH₄Cl was replaced by 0.1% glutamate, and supplemented with 0.2% glucose and MgSO₄ (2 mM final concentration). Three-microliter droplets containing ca. 10⁶ CFU were placed at equal distances on the dried plates, which were incubated for 24 h at 37°C and then for 24 h at 25°C. The rhamnolipid production was estimated semiquantitatively by measuring the diameter of the blue halos forming around bacterial spots.

Activity of phospholipase C in the supernatants of stationary-phase cultures was quantified spectrophotometrically by using NPPC (p-nitrophenyl-phosphorylcholine) as a chromogenic substrate, according to the method developed by Berka et al. (2). Briefly, 2 ml of tryptose minimal medium was inoculated with 200 μl of a bacterial suspension calibrated at a McFarland standard of 0.4 in saline buffer. After 24 h of incubation at 37°C, the culture supernatant was recovered and treated with 10 mg of decolorizing carbon. Ten-microliter fractions of medium were then added to 90 μl of a reaction mixture containing 250 mM Tris-HCl (pH 7.2), 10 mM NPPC, 60% (wt/vol) glycerol, and 1 μM ZnCl₂. Absorbance at 405 nm of triplicate samples was recorded after 1 h of incubation at 37°C.

In addition to eliciting phospholipase C production, the low-phosphate tryptose minimal medium was found to elicit strong production of pyocyanin in the various strains of P. aeruginosa. The redox active dye in the supernatants of 24-h cultures was quantified spectrophotometrically, as described by Essar et al. (11). Briefly, 1-ml fractions of supernatant were extracted twice with 1 ml chloroform, and pyocyanin was reextracted from the solvent in 1 ml of 0.2 HCl. Absorbance of this solution was measured at 520 nm.

The T3SS-dependent cytotoxicity of the strains was assayed with the murine macrophage cell line J774 (ATCC), grown in Dulbecco's modified Eagle medium (Gibco) supplemented with heat-inactivated fetal calf serum (Gibco) at 10%. The cells were seeded in 24-well culture plates at 3 × 10⁵ cells per well 20 h before infection. The bacterial strains were grown overnight in Luria broth, diluted to an OD₆₀₀ of 0.1, and grown for an additional 3 h to an OD₆₀₀ of approximately 1.0. Macrophages were infected with bacteria at a multiplicity of infection of 5. Cytotoxicity was assessed at 2 h and 3 h postinfection by determination of lactate dehydrogenase (LDH) release into infected supernatants using a cytotoxicity detection kit (Roche), as described previously (7). The 100% value represented the LDH released from uninfected cells lysed with Triton X-100.

Strain 3308 was isolated from a 70-year-old woman with long-term history of acute myeloid leukemia and admitted to the hospital for pneumonia of unknown origin. The strain was recovered from the sputum and a single blood culture after 32 days of hospitalization and multiple courses of chemotherapy, including a 9-day per os treatment with ciprofloxacin (1.5 g per day) and cefpodoxime (400 mg per day). The patient's death, which occurred 1 month later despite administration of adequate antipseudomonal antibiotics (2 g per day ceftazidime and 750 mg per day amikacin for 19 and 3 days, respectively), was caused by a septic shock not documented bacteriologically, concomitant with unfavorable evolution of her leukemia. Strain 2439 was isolated from a purulent surgical wound in a 39-year-old male admitted to the hospital for multiple trauma, after 11 days of single therapy with intravenous ciprofloxacin (1 g per day). The patient's fatal outcome 1 month later was not attributed to P. aeruginosa. Strain 2126 was isolated from a 51-year-old outpatient male with severe external otitis, after a 10-day intravenous course of ciprofloxacin (800 mg per day) and topical application of ofloxacin. The clinical evolution was favorable, and the patient recovered from his infection. Finally, strain 1956 was recovered from a 65-year-old man admitted to the hospital for cerebrovascular atherosclerotic disease. This patient was transferred 48 days after admission to the medical intensive care unit and intubated because of obstructive bronchitis. The nfxB mutant was isolated in a tracheal aspirate after 11 days of ciprofloxacin (400 mg per day) and ceftriaxone (1 g per day). The patient improved under treatment and was transferred to a tertiary-care center, where he died 1.5 month later from severe sepsis and pneumonia not documented bacteriologically.

The GenBank accession numbers of nfxB sequences of strains 3308, 2439, 2126, and 1956 are EU525876, EU525877, EU525878, and EU525879, respectively.

In vitro-selected MexCD-OprJ-overexpressing mutants have been reported to exhibit 4- to 16-fold increases in resistance to fluoroquinolones, macrolides, and zwitterionic cephalosporins (e.g., cefpirome and cefepime) compared to wild-type susceptible strains (14, 15, 24, 25, 45, 54). However, the NfxB phenotype may also include a 2- to 8-fold-higher resistance to tetracycline and chloramphenicol concomitant with a 2- to 32-fold-greater susceptibility to aminoglycosides and some β-lactams (e.g., ticarcillin, aztreonam, and imipenem) (15, 24, 45, 46), making its identification rather difficult in clinical strains. To get an insight into the prevalence of nfxB mutants in our hospital, we screened a collection of 110 nonreplicate non-cystic fibrosis clinical isolates showing low-level resistance to ciprofloxacin (MIC from 0.5 to 4 μg/ml), compatible with expression of an efflux-based mechanism. Only four strains (3308, 2439, 2126, and 1956) appeared to overexpress the mexC gene significantly by RT-PCR (Table 1). Sequencing of the nfxB gene, whose product normally downregulates the mexCD-OprJ operon (54), confirmed that these genotypically distinct strains were nfxB mutants and revealed various nucleotide deletions (strains 3308, 2439, and 2126) as well as a missense mutation resulting in an Ala38→Val substitution in the DNA binding domain (helix-turn-helix motif) of the NfxB protein (strain 1956) (51). The in vitro-selected nfxB mutant EryR (14) was found to harbor a nonsense mutation leading to a truncated NfxB peptide. Loss of function of NfxB has previously been associated with frameshift mutations (5, 8, 54) or disruption of gene nfxB by insertion sequences (8), as well as amino acid substitutions lying within (Leu40→Gln, Arg42→Gly, Arg42→His) or outside (Arg82→Leu, His87→Arg, Leu88→Pro) the helix-turn-helix domain of the protein (5, 26, 34, 51, 54). Of particular interest here was the observation that mexC expression in strain 1956 was somewhat lower (from 1.5- to 3.5-fold) than that in strains with disrupted nfxB genes, suggesting that the Ala38→Val substitution might only partially compromise the ability of NfxB to bind to the promoter region of mexCD-oprJ and repress its expression. Consistent with this assumption, sequencing of the intergenic region between nfxB and mexCD-oprJ did not reveal additional alterations in that strain. The involvement of other Mex pumps in addition to MexCD-OprJ in the resistance of the clinical strains was ruled out by RT-PCR experiments that showed that the expression levels of genes mexB, mexE, mexG, mexJ, mexV, and mexY were similar in the four clinical nfxB mutants and PAO1 (data not shown except for mexB and mexY; Table 1).

As shown in Table 2, strains 3308, 2439, and 2126 were phenotypically similar to previously reported type B nfxB mutants (45) including EryR. Compared with wild-type strain PAO1, they exhibited some degree of resistance to ciprofloxacin (8-fold), cefepime (2- to 4-fold), and chloramphenicol (2-fold), with concomitant increased susceptibility to ticarcillin (4- to 8-fold), aztreonam (4-fold), imipenem (4- to 16-fold), and aminoglycosides (2- to 8-fold). As expected from the frameshift mutations found in nfxB, complementation of these bacteria with a plasmid-borne nfxB gene from PAO1 (named pNF225) restored wild-type levels of susceptibility to most drugs. Reminiscent of type A mutants, which produce lower levels of MexCD-OprJ than their type B counterparts (45), strain 1956 appeared to be less resistant to ciprofloxacin than strains 3308, 2439, and 2126 (fourfold increase in MIC; Table 2) and indistinguishable from PAO1 regarding its susceptibility to cefepime and imipenem. Also, consistent with operon mexCD-oprJ being partially derepressed in 1956, complementation of that strain with plasmid pNF225 indeed produced only 2- to 4-fold variations in the MICs of selected antibiotics compared with the 2- to 16-fold changes observed in strains 3308, 2439, and 2126. It should be mentioned that, whatever their A or B NfxB phenotype, all the clinical strains showed the same susceptibility as PAO1 to piperacillin, a poor substrate for MexCD-OprJ (data not shown) (49).

The MexXY-OprM efflux system contributes efficiently to the natural resistance of P. aeruginosa to aminoglycosides under standard laboratory growth conditions (1, 48). It has recently been demonstrated that these agents indirectly induce the expression of operon mexXY as a result of their interaction with the ribosome (27). Because the natural resistance to aminoglycosides requires the induced production of proteins MexX and MexY, we tested whether mexXY is still inducible in the nfxB mutants. Table 1 shows that both gentamicin and chloramphenicol at half the MIC were able to induce mexXY expression in strains EryR, 2439, 2126, and 1956. Additional RT-PCR experiments on genotypically different nfxB mutants selected in vitro on ciprofloxacin (KJ0702, KJ0707, KJ0705, and KJ0711 from strains PAO1, PA14, 4.2, and 19.1, respectively) confirmed these results (data not shown). In PAO1 and its nfxB mutant EryR, the drug-induced mexY levels were similar and not significantly influenced by complementation with nfxB plasmid pNF225. Interestingly, the uninduced transcription of mexY was found to be slightly higher in EryR and EryR(pNF225) than in PAO1, a result which correlated with an increased production of protein MexY in EryR (Western blotting data not shown). As in EryR, plasmid pNF225 had minor if any effects on basal or induced expression of mexY in strains 2439, 2126, and 1956. Similar results were obtained for 3308, with the notable difference that this strain expressed barely detectable, albeit still inducible, amounts of mexY mRNA. As confirmation that 3308 is strongly deficient in MexXY, the strain turned out to be as susceptible as ΔmexXY mutant FE60 to aminoglycosides (Table 2). Sequencing experiments on strain 3308 did not reveal particular mutations in gene mexZ (whose product downregulates operon mexXY), in the mexZ-mexXY intergenic region, or in gene PA5471 (whose product is necessary for drug-induced expression of mexXY in strain PAO1) (50), a result which provides indirect evidence for the existence of additional genes regulating mexXY expression.

Altogether these data clearly showed that the nfxB mutation has no influence on the expression levels of mexXY or the inducibility of mexXY by aminoglycosides. To further investigate the contribution of the MexXY-OprM system to the resistance of nfxB mutants to these antibiotics, we attempted to inactivate the mexXY operon with a gene replacement strategy (10). We were unsuccessful with the clinical strains, as the suicide plasmid pEXΔXYR repeatedly integrated into the bacterial chromosome at sites other than the mexXY locus. In contrast, we could easily obtain a ΔmexXY derivative from EryR (named FK06). Susceptibility to aminoglycosides of EryR was poorly affected by the inactivation of MexXY, suggesting that the activity of the pump is compromised as a result of MexCD-OprJ being upregulated (compare FK06 with EryR and FE60 in Table 2). Furthermore, the observation that FK06, once complemented with nfxB plasmid pNF225, exhibited unchanged susceptibility to these products, in contrast to EryR, strongly suggests that in nfxB mutants such as EryR, 2439, 2126, and 1956 the MexXY-OprM-mediated export of aminoglycosides is impaired. Interestingly, other investigators (60) also came to the conclusion that complex factors may influence the contribution of MexXY-OprM to aminoglycoside resistance in clinical strains.

Transformation of EryR, FK06, 3308, 2439, 2126, and 1956 with plasmid pNF225 completely restored wild-type levels of resistance to aztreonam and imipenem (Table 2), thereby indicating that the increased susceptibility to β-lactams of all the nfxB mutants is linked to the alteration of nfxB. We therefore tested whether this phenotype is due to the reduced expression of efflux system MexAB-OprM and/or of chromosomally encoded β-lactamase AmpC, as previously reported for laboratory mutants (14, 47). All the nfxB mutants studied here, including KJ0702, KJ0707, KJ0705, and KJ0711 (data not shown), expressed gene mexB at comparable levels (from 0.4- to 1.4-fold that of PAO1), which remained unchanged upon complementation with pNF225 despite restoration of the wild-type susceptibility to β-lactams (Tables 1 and 2). Confirming these data, Western blot experiments revealed the presence of similar amounts of protein MexB in the mutants and their respective pNF225 transformants (not shown). Furthermore, the AmpC activities (uninduced versus induced with 50 μg/ml cefoxitin) for PAO1 (13 versus 1,820 nmol nitrocefin min⁻¹ mg⁻¹ protein), EryR (29 versus 3,110 nmol nitrocefin min⁻¹ mg⁻¹ protein), 3308 (20 versus 2,670 nmol nitrocefin min⁻¹ mg⁻¹ protein), 2439 (19 versus 2,680 nmol nitrocefin min⁻¹ mg⁻¹ protein), 2126 (33 versus 5,140 nmol nitrocefin min⁻¹ mg⁻¹ protein), and 1956 (11 versus 4,550 nmol nitrocefin min⁻¹ mg⁻¹ protein) were roughly similar. Thus, in apparent contrast to the conclusions of other studies (14, 47), but reminiscent of our results with MexXY-OprM (see above), these data suggest that, while normally produced, efflux pump MexAB-OprM has partially lost its ability to efficiently transport β-lactam antibiotics in nfxB mutants.

Upregulation of efflux pumps has been reported to impact bacterial virulence in strain PAO1 (12, 33). For instance, MexAB-OprM- and MexCD-OprJ-overproducing mutants of PAO1 produce less pyocyanin, pyoverdine, casein protease, and elastase than their wild-type parent (12, 57); they also exhibit a reduced capacity to invade cell monolayers and to kill leukemic mice (16) or nematodes such as Caenorhabditis elegans (57). Recently, a MexCD-OprJ-overproducing derivative of PAO1 was found to be partially deficient in the T3SS, a needle-like export machinery that allows direct injection of various cytotoxins into eucaryotic cells (38). Whether these partial virulence defects are clinically significant in human infections remains to be confirmed, as MexAB-OprM-upregulated strains are very prevalent in hospitalized patients and may be involved in severe infections (17, 20, 21, 39). In vitro assays showed that EryR, 3308, 2439, 2126, and 1956 differed quite significantly in terms of rhamnolipid, elastase, phospholipase C, and pyocyanin production, as well as in terms of T3SS-dependent cytotoxicity (compare 2439 and 1956 in Table 3). The effects of MexCD-OprJ upregulation on the virulence of P. aeruginosa also appeared to be variable and dependent on the genetic background of the strains (compare EryR to PAO1, KJ0707 to PA14, KJ0705 to 4.2, and KJ0711 to 19.1 in Table 3). All the nfxB mutants of the study produced smaller colonies on MHA than PAO1 or their respective wild-type parents (Table 3), strongly suggesting that when overproduced this pump represents a significant burden for P. aeruginosa. However, since strains 3308, 2439, 2126, and 1956 were involved in confirmed cases of infection (see Materials and Methods for more clinical data), it seems evident that some nfxB mutants may remain pathogenic, at least under certain clinical situations.

Many studies have highlighted the fact that fluoroquinolones readily select for nfxB mutants in vitro (24, 25, 32, 46) as well as in infected animals (29, 31, 43). Although very few clinical nfxB strains have been characterized to date (24, 25, 62), the circumstances of their selection in vivo have rarely been documented (56). Review of the treatments administered to the patients with strains 3308, 2439, 2126, and 1956 clearly showed that these nfxB mutants emerged under long-term single therapy (as cefpodoxime and ceftriaxone are quite inefficient against P. aeruginosa) with fluoroquinolones (mostly ciprofloxacin), as in the case described by Reinhardt et al. (56). Whether these mutants evolved from initially susceptible strains or were selected in the hospital environment by antiseptics such as triclosan (5) prior to infecting patients could not be determined precisely.