Clinical Strains of Pseudomonas aeruginosa Overproducing MexAB-OprM and MexXY Efflux Pumps Simultaneously

Bacterial strains, media, and growth conditions.Twelve clinical isolates of Pseudomonas aeruginosa isolated between 1997 and 2000 in the university-affiliated Hospital of Besançon (eastern France) were selected because of their nonenzymatic resistance to both β-lactams and aminoglycosides. These strains, which belonged to serotypes O:1, O:4, O:11, and O:14, were found to be genotypically distinct by random amplified polymorphic DNA analysis (18). They were isolated from urine (WL22, WL24, 1217, 1237, 1562, and 2172), tracheal aspirates (1113, 1727, and 1738), blood catheters (1250 and 2151), or surgical wounds (2085). Bacteria were cultured at 37°C in either Luria-Bertani broth (LB), Mueller-Hinton broth with adjusted concentrations of Ca²⁺ and Mg²⁺ (MHB; BBL, Cockeysville, Md.), or Mueller-Hinton agar plates (MHA; Bio-Rad, Ivry-sur-Seine, France). The wild-type P. aeruginosa strain PAO1 (K. Stover) was used as the susceptible reference strain throughout the study, and its MexAB-OprM-overproducing nalB mutant strain PT629 (a gift from Thilo Köhler, Gevena, Switzerland) served as a control in gene expression experiments.

Mutants constitutively overproducing MexXY were obtained by incubating strain PAO1 for 2 h in MHB containing 2 μg of gentamicin per ml (1× MIC) and then plating the bacteria on selective MHA plates supplemented with 2 μg of cefepime per ml (1× MIC). Western blot and nucleotide sequencing experiments showed that one resistant clone (named Mut-Gr1) overproduced MexY as a result of a single mutation (C₃₀₇→T) that introduced a stop codon in the coding sequence of the repressor gene mexZ (36a). A mutant overproducing both MexAB-OprM and MexXY was obtained by plating an exponential-phase culture of Mut-Gr1 on selective MHA plates supplemented with 10 μg of aztreonam per ml. Reverse transcription-PCR and nucleotide sequencing showed that this double mutant, named ATM4, overproduced MexA as a result of a single-amino-acid substitution (Ala₆₁→Pro) in the repressor MexR.

Bacterial resistance to antibiotics.Clinical strains of P. aeruginosa potentially overexpressing the two systems MexAB-OprM and MexXY were retrospectively selected from our laboratory collection on the basis of their drug susceptibility profiles. Typically, MexAB-OprM overproducers exhibit reduced susceptibilities to most β-lactams except imipenem, with MICs of aztreonam at least fourfold higher than that of ceftazidime (39). On the other hand, MexXY-overexpressing mutants show decreased susceptibilities to gentamicin, tobramycin, amikacin, isepamycin, and cefepime, with MICs at least twofold higher than that for wild-type susceptible strains (unpublished observation). None of these isolates displayed typical resistance profiles involving MexCD-OprJ (resistance to cefpirome and hypersusceptibility to both ticarcillin and aminoglycosides) or MexEF-OprN (resistance to imipenem and hypersusceptibility to both ticarcillin and aminoglycosides).

Susceptibility testing was performed by the standard microdilution method in MHB with bacterial inocula of about 2.5 × 10⁵ CFU per ml (2). Some of the antibiotics tested were kindly provided by Eli Lilly (tobramycin), Bristol-Meyers Squibb (amikacin, aztreonam, and cefepime), and Glaxo SmithKline (ticarcillin and ceftazidime). Isoelectrofocusing experiments with crude bacterial lysates (6, 22) allowed the selection of 12 strains showing no detectable β-lactamase activity except for faint, wild-type expression of the chromosomally encoded enzyme AmpC (data not shown). The mechanisms of resistance to aminoglycosides in these isolates were deduced from the levels of susceptibility to kanamycin, tobramycin, gentamicin, amikacin, isepamicin, netilmicin, neomycin, 5-epinetilmicin, 2′-netilmicin, 6′-netilmicin, apramycin, and fortimicin, as determined with the aminoglycoside resistance test kit provided by the Schering Plough Research Institute. Apramycin and fortimicin were obtained as titrated powders from Helm (Hamburg, Germany) and Kyowa Hakko Kogyo (Tokyo, Japan), respectively.

Immunodetection of MexY and OprM.Bacterial membranes were isolated and analyzed by Western blotting with MexY- and OprM-specific antisera as previously described (7).

PCR conditions and DNA sequencing.Chromosomal DNA was extracted with the EZNA bacterial DNA kit (Omega, Doraville, Ga.). The coding sequences of mexR and mexZ as well as the intergenic regions between mexA and mexR (mexA-mexR) and between mexX and mexZ (mexX-mexZ) were amplified and sequenced as previously described (6). PCR amplification of the putative repressor nalC (PA3721) was performed with primers nalC1 and nalC2 (Table 1) in a DNA thermal cycler (Perkin-Elmer, Norwalk, Conn.) for 30 cycles, each cycle consisting of 40 s at 94°C for denaturation, 1 min at 69°C for annealing, and 1 min at 72°C for polymerization. PCR amplicons were sequenced on both strands.

Quantitative real-time reverse transcriptase-PCR.Total RNA were isolated from exponential-phase cultures (A₆₀₀ = 1) with the Qiagen RNeasy protocol (Qiagen, Courtaboeuf, France). The RNA samples were further treated with DNase (RQ1 DNase; Promega, Madison, Wis.) and purified by phenol-chloroform extraction and ethanol precipitation (J.-L. Dumas, C. Van Delden, and T. Köhler, submitted for publication). We reverse transcribed 2 μg of total RNA with ImPromII reverse transcriptase (Promega) according to the supplier's instructions. The mexA and mexX cDNAs were subsequently quantified in a Rotor Gene RG3000 RealTime PCR machine (Corbett Research, Sydney, Australia) with a SybrGreen Quantitect kit (Qiagen) with primers (Dumas et al., submitted for publication) mexA-1/mexA-2 for mexA and mexX-1 for mexX (Table 1). The expression levels of PA3720 (putative gene nalC) were estimated after amplification with primers PA3720-1 and PA3720-2 (Table 1). To correct for differences in the amount of starting materials, the ribosomal rpsL gene was chosen as a reference housekeeping gene (primers rpsL-1 and rpsL-2; Table 1). The results are presented as ratios of gene expression between the target gene (mexA, mexX, or nalC) and the reference gene (rpsL) (29).

Selection of double-efflux clinical strains of P. aeruginosa.Twelve genotypically distinct strains were selected for their resistance profiles to both β-lactams and aminoglycosides, evoking concomitant overproductions of efflux systems MexAB-OprM and MexXY (see Materials and Methods). None of these isolates were derepressed for the chromosomally encoded AmpC β-lactamase or produced transmissible secondary β-lactamases.

As indicated in Table 2, these strains were two- to eightfold more resistant to ticarcillin, aztreonam, and cefepime than was the wild-type reference strain PAO1. All of them remained susceptible to ceftazidime according to the standard breakpoints (MICs, ≤8 μg/ml). The 12 strains also displayed reduced susceptibilities to all 12 aminoglycosides tested (Schering Plough kit), including the enzyme-recalcitrant compounds apramycin and fortimicin (Table 2). This decreased susceptibility strongly suggested the expression of one or several nonenzymatic resistance mechanisms to aminoglycosides in the selected bacteria. Except for the well-known chromosomally encoded APH(3′)-II enzyme, which naturally provides P. aeruginosa with high resistance to kanamycin and neomycin (27), no additional enzymatic mechanisms could be detected in six strains (1113, 1250, 1727, 1738, 2151, and 2172), while the aminoglycoside-modifying enzymes ANT(2")-I, AAC(6′)-II, and AAC(3)-VI were phenotypically identified in isolates WL24, 1237, and 2085, respectively. Finally, three isolates (WL22, 1217, and 1562) showed complex susceptibility profiles, suggesting the synthesis of several modifying enzymes. The resistance of the 12 strains to ciprofloxacin varied greatly, with MICs ranging from 0.5 to 64 μg/ml (Table 2).

Overexpression of mexA and mexX.Upregulation of the MexAB-OprM and MexXY efflux systems in the P. aeruginosa strains was assessed by determining the transcription levels of mexA and mexX by quantitative real-time reverse transcription-PCR (Table 3). Confirming that the selected isolates are double efflux mutants, mexA and mexX appeared to be expressed 2.6- to 34.8-fold and 22- to 312-fold, respectively, more than in PAO1. Control experiments performed in parallel showed that the transcription levels of mexA and mexX in eight randomly chosen, susceptible clinical strains were very similar to that of the reference strain (0.8- ± 0.2-fold and 1- ± 0.3-fold that of PAO1, respectively). These results were in agreement with membrane immunoblots showing increased amounts of MexY and OprM proteins in the mutants (data not shown).

Mutations in repressor gene mexZ.To gain insight into the mutational events responsible for MexXY upregulation in the 12 double-efflux strains, we amplified and sequenced the intergenic region between mexX and mexZ as well as mexZ, the repressor gene of the mexXY operon (C. Vogne, D. Hocquet, J. Ramos Aires, F. El Garch, P. Plésiat, Abstr. 42nd Intersci. Conf. Antimicrob. Agents Chemother., abstr. C1-434, 2002). Compared with the genome sequence of PAO1 (available at www.pseudomonas.com), nine isolates were found to harbor mutations, resulting in (i) single-amino-acid substitutions in MexZ (strains WL22, 1562, 2085, 2151, and 2172) or (ii) production of aberrant peptides (strains 1113, 1217, 1237, and 1250) (Table 3). Most of these mutations (seven of nine) occurred in the helix-turn-helix N-terminal motif of MexZ (between positions 32 and 53), a domain predicted by the EMBOSS algorithm (available at www.pasteur.fr) to be involved in DNA binding. Two strains (1738 and 2085) exhibited single mutations in the mexX-mexZ intergenic region outside the putative promoter sequences of mexZ and mexXY (data not shown). Finally, strains WL24 and 1727 had nucleotide sequences identical to that of PAO1. No correlation could be established between the various MexZ alterations, the expression level of mexX, which varied from 29.6- to 312-fold that of PAO1 (Table 3), and resistance to aminoglycosides (Table 2). Important variations in mexX transcription (22- to 300-fold higher than that in PAO1) were also observed in the three strains harboring intact mexZ genes (WL24, 1738, and 1727).

In contrast to previous observations made on cystic fibrosis isolates of P. aeruginosa (37), stable derepression of mexXY in the double-efflux mutants described here was mostly associated with alterations in mexZ. We thus propose the names agrZ (for aminoglycoside resistance dependent on mexZ) and agrW (unknown locus) for the genotypes of MexXY-overproducing mutants with altered and intact mexZ genes, respectively. In the absence of specific aminoglycoside-modifying enzymes, overexpression of mexXY in both types of mutants as well as in the mexZ-null mutant Mut-Gr1 was associated with low to moderate resistance to all the aminoglycosides tested (MICs increased two- to eightfold) compared to PAO1. Although relevant in some clinical situations where aminoglycosides diffuse poorly at the site of infection (4), these resistance levels are far below that conferred by aminoglycoside-modifying enzymes for major therapeutic agents such as amikacin (e.g., isolates WL22 and 1562) and tobramycin (e.g., WL22 and 1237) (Table 2). The contribution of the MexXY-OprM efflux system to the overall resistance of strains producing aminoglycoside-modifying enzymes remains to be explored.

Mutations in the mexABoprM operon regulatory genes.Sequencing experiments revealed several other mutations in the double-efflux strains (Table 3). One strain (1250) harbored an A→G substitution of unknown significance in the mexA-mexR intergenic region at position −5 (position 1 being the A of the mexR start codon), which is distant from the MexR binding sites (5). MexR amino acid sequences strictly identical to that published by Poole et al. (32) were found in six putative nalC strains (WL24, 1217, 1250, 1562, 1738, and 2172). Five other strains (WL22, 1113, 1237, 2085, and 2151) harbored a Val₁₂₆→Glu substitution in MexR, already observed in susceptible wild-type isolates and considered nonsignificant (28, 39). In this group, four nalB mutants (WL22, 1237, 2085, and 2151) displayed an already known mutation in mexR that results in a Lys₄₄→Met substitution (9), affecting the DNA binding domain of MexR (from residues 37 to 97) (17). Finally, a frameshift mutation (7-bp insertion between nucleotides 355 and 361) was discovered in the mexR sequence of another nalB strain, 1727. Similar genetic events leading to frameshifts in mexR have already been reported in MexAB-OprM-overexpressing mutants (36, 39).

In order to further characterize the mutants harboring intact mexR genes, we sequenced PA3721, the repressor gene of the MexAB-OprM putative activator PA3720-PA3719 (Cao et al., abstr. 430) and assessed the expression of PA3720 by real-time reverse transcription-PCR (Table 3). Compared with PAO1, three recurrent substitutions in repressor protein PA3721 were noticed in 11 of 12 isolates: Gly₇₁→Glu, Glu₁₅₃→Gln, and Ser₂₀₉→Arg. These amino acid differences with PAO1 do not seem to have an impact on PA3721 activity, as they were also present in four of four susceptible wild-type environmental strains of P. aeruginosa (data not shown). In addition, strain 1237, which overexpressed PA3720 only 3.7-fold more than PAO1, appeared to have a Ser₄₆→Ala substitution in the PA3721 repressor. More importantly, three isolates were found to strongly overexpress PA3720 (147- to 224-fold more than in PAO1), suggesting that they are nalC mutants. Two of them harbored single-amino-acid substitutions in PA3721: Met₁₅₁→Thr for isolate 2172, and Leu₆₁→Pro for isolate 1738. The third strain (1250) displayed a large 500-bp deletion in PA3721. Another strain (1727) already characterized as a nalB mutant (see above) and showing moderate overexpression of PA3720 (9.9-fold more than in PAO1) appeared to carry an Asp₇₆→Glu substitution in PA3721. Finally, various differences in the DNA sequences upstream and downstream of the PA3721 gene were observed between five isolates with low levels of PA3720 expression (WL22, 1113, 1237, 2085, and 2151) and PAO1 (Table 3). The effect of these nucleotide changes on PA3720 transcription is unclear.

As for MexXY, the transcription levels of mexA in the isolates were not clearly correlated with particular types of mutations in mexR or PA3721. The MICs of β-lactams such as ticarcillin and aztreonam, which are known to be good substrates for MexAB-OprM (21), also did not vary with the degree of mexA overexpression (Table 2). This observation is not surprising by itself, as multiple factors are involved in determining the susceptibility of a given isolate to antibiotics (e.g., outer membrane permeability, affinity of drug targets, degree of inducibility of AmpC enzyme, etc.) (25). In contrast to previous results obtained with reference strain PAO1 (36), the clinical nalC mutant isolates reported here and elsewhere (39) did not appear to be less resistant than the nalB mutants to β-lactams.

In addition to well known nalB strains (32, 36, 39), this work identified three nalC mutants and one nalB nalC mutant among the 12 MexAB-OprM/MexXY-overproducing strains selected. This indicates that nalC mutants with alterations in PA3721, like those initially characterized in vitro from PAO1 (Cao et al., abstr. 430), are rather prevalent among resistant clinical isolates. As indicated in Table 3, mutations in both mexR and PA3721 could have additive effects on mexA expression, as nalB nalC isolate 1727 displayed the highest mexA transcription levels (about twofold higher than that of nalB or nalC mutants). A similar observation has been made for a ΔacrR Δmar double mutant of Escherichia coli which expressed greater amounts of acrB mRNA than single mar mutants (10). Interestingly, interplay between mexR and nalC has already been suspected in PAO1 (36). Finally, no less than four P. aeruginosa strains (WL24, 1113, 1217, and 1562), which we propose to call nalD type mutants, appeared to contain no mutation in the known regulatory genes for MexAB and OprM.

Lee et al. reported that concomitant overproduction of two Mex pumps in PAO1 produces additive effects on the resistance to shared antibiotic substrates (14). This tends to suggest that clinical strains of P. aeruginosa may increase their resistance to a given compound by coexpressing two Mex efflux systems. In support of this, the clinical double-efflux mutants studied here appeared to be more resistant to cefepime (Table 2) than mutants overproducing MexAB-OprM (PT629) or MexXY (Mut-Gr1) alone (MIC, 8 to 16 versus 4 μg/ml). Convincing evidence of cooperation between MexXY and MexAB-OprM for the extrusion of common substrates is also provided by the double mutant ATM4, which was two- to fourfold more resistant to cefepime and ciprofloxacin than the single mutants Mut-Gr1 and PT629, respectively (Table 2). Besides this, our results show that superimposition of the resistance profiles conferred by two efflux systems, such as MexAB-OprM (β-lactams) and MexXY (aminoglycosides), may be an efficient way for infectious strains to become less susceptible to numerous antibiotics. In a therapeutic perspective, efflux inhibitors of broad specificity would be potentially interesting to reverse the resistance or prevent the emergence of such double-efflux mutants.