INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Decreased intracellular accumulation due to active efflux of antibiotics out of bacterial cells is one of the mechanisms that contributes to the failure of therapy with many currently used antibiotics. Both antibiotic-specific and multidrug-resistant pumps were identified. The latter class of transporter proteins can extrude out of the cell a large variety of structurally unrelated compounds with different modes of action. Many of them are currently used antibiotics (15-17, 24-27).

Pseudomonas aeruginosa is an important opportunistic pathogen in which three multicomponent, multidrug-resistant efflux pumps have been identified, namely, Mex-AB-OprM (30, 31), MexCD-OprJ (29), and MexEF-OprN (11). Of the known multidrug-resistant pumps in P. aeruginosa, only MexAB-OprM is expressed at a level sufficient to confer intrinsic multidrug resistance in wild-type cells. Deletion of the mexA, mexB, or oprM gene renders P. aeruginosa more susceptible to multiple antibiotics (6, 31, 38). Multidrug-resistant mutants with increased expression of any of the pumps can easily be isolated and manipulated under laboratory conditions (8, 19, 20, 32).

Fluoroquinolones, primary therapeutic antibiotics for P. aeruginosa, are effluxed by all the known Mex pumps. Mutants with elevated levels of expression of the pumps, which confer increased resistance to fluoroquinolones, have been identified among clinical strains: nalB mutants that overproduce the MexAB-OprM pump (2), nfxB mutants that overproduce MexCD-OprJ (10, 40), and nfxC mutants that overproduce the MexEF-OprN efflux pump (5). This resistance to fluoroquinolones through the overproduction of efflux pumps is distinct from the resistance to fluoroquinolone antibiotics through the mutation of quinolone resistance-determining regions (QRDRs) (9, 28, 39) in DNA gyrase and topoisomerase IV, which are encoded by gyrAB and parCE genes, respectively (9), in many organisms (28) including P. aeruginosa (14, 21).

In this report we show that deletion of efflux pumps reduces the level of resistance to fluoroquinolones even in highly resistant strains with multiple target mutations. We also show that deletion of all three described pumps significantly reduces the frequency of emergence of fluoroquinolone-resistant mutant strains. These results demonstrate the potential effects of inhibition of efflux pumps on the susceptibility to fluoroquinolones.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and media. The bacterial strains used in this study are listed in Table 1. Bacterial cells were grown in Luria (L) broth (1% [wt/vol] tryptone, 0.5% [wt/vol] yeast extract, 0.5% [wt/vol] NaCl) or L agar (L broth plus 1.5% agar) at 37°C. The following antibiotics were added to the media at the indicated concentrations: tetracycline, 20 µg/ml for Escherichia coli and 100 to 150 µg/ml for P. aeruginosa; chloramphenicol, 20 µg/ml for E. coli and 100 µg/ml for P. aeruginosa; gentamicin, 15 µg/ml for both E. coli and P. aeruginosa; HgCl₂, 15 µg/ml for both E. coli and P. aeruginosa; ampicillin, 100 µg/ml for E. coli; and kanamycin, 50 µg/ml for E. coli. L agar was supplemented with 5% (wt/vol) sucrose as required. Levofloxacin (LVX) was synthesized at Daiichi Pharmaceutical Co., Ltd. (Tokyo, Japan). All other antibiotics were purchased from Sigma Chemical Co. (St. Louis, Mo.).

Selection of multidrug-resistant mutants of P. aeruginosa. Selection of multidrug-resistant mutants of P. aeruginosa was performed as described previously (20). The frequency of resistance was determined as the ratio of the numbers of CFU per milliliter that appeared after overnight incubation on antibiotic-containing L agar plates versus the numbers that appeared after overnight incubation on antibiotic-free L agar plates.

Stepwise selection of LVX resistance. Wild-type strain PAM1020 (LVX MIC, 0.25 µg/ml) was plated on LBA plates with LVX at 4× the MIC. The first-generation spontaneous mutants were selected at a frequency of 10⁶ to 10⁷. The same procedure was repeated several times for subsequent generations of mutants, each time with higher concentrations of LVX but still at 4× the MIC. During the next four steps of selection, spontaneous mutants were isolated at a frequency of ca. 10⁸. The highest MIC achieved after five selection steps was 128 µg/ml.

Transductions. Transductions in P. aeruginosa were performed with phage F116L by a previously described protocol (13).

MIC determinations. MICs were determined in 96-well microtiter plates by a standard broth microdilution method (22) in Muller-Hinton broth (Difco). The inoculum was 10⁴ to 10⁵ cells/ml.

DNA manipulations. Plasmid DNA was purified with the RPM Spin Kit (Bio 101, Inc., Vista, Calif.). Chromosomal DNA was prepared by using the Qiagen Blood and Cell Culture Mini Kit (Qiagen Inc., Valencea, Calif.). DNA fragments were gel purified and extracted with the Qiagen Gel Purification Kit or the Bio 101 GeneClean Kit. Restriction enzymes were obtained from New England Biolabs (Beverly, Mass.), and AmpliTaq was obtained from Perkin-Elmer (Branchburg, N.J.). Plasmid DNA was introduced into E. coli strains by electroporation (Bio-Rad Laboratories, Mississauga, Ontario, Canada). All molecular biology techniques were performed according to the manufacturer's instructions or as described by Sambrook et al. (33). PCR was carried out in a Perkin-Elmer GeneAmp 9600 thermal cycler. Typically, 30 cycles of denaturation (30 s at 95°C), annealing (30 s at 55°C), and extension (1 min at 72°C) were used to amplify the DNA used in the construction of operon deletions. Analysis of the sequences of the QRDRs was performed directly with PCR-amplified genomic DNA segments. Cycle sequencing was carried out with the ABI-PRISM fluorescent dye terminator kit (Applied Biosystems Inc., Foster City, Calif.). The QRDR of the gyrA (14) gene was amplified with primers TTATGCCATGAGCGAGCTGGGCAACGACT (29mer) and TTCCGTTGACCAGCAGGTTGGGAATCTT (28mer). The QRDR of the parC (21) gene was amplified with primers ATCTGAGCCTGGAAG (15mer) and AGCAGCACCTCGGAATAG (18mer).

Construction of a recombinant plasmid for deletion of the mexAB-oprM operon. Plasmid pAL232 was constructed to replace a part of the sequence of the mexAB-oprM operon with the chloramphenicol resistance [cat (Cm^r)] gene in the chromosome of P. aeruginosa. Construction was performed as follows. First, we created auxiliary plasmids pAL219, pX1918-Cm, and pMOB3-Gm. pAL219 is a derivative of pNOT19 (34) whose SalI site was removed by the Klenow treatment and religation. pX1918-Cm is a derivative of pX1918-GT (35) in which a BamHI fragment carrying a gentamicin resistance [aacC1 (Gm^r)] gene was replaced with a BamHI fragment carrying the cat gene. The cat gene was obtained from plasmid pMOB3 (34). pMOB3-Gm is a derivative of pMOB3 in which a BamHI fragment carrying a cat gene was replaced with a BamHI fragment carrying an aacC1 gene. The aacC1 gene was obtained from plasmid pX1918-GT. Second, plasmid pAL225 was constructed from pAL219 and pRS14. Plasmid pRS14, obtained from K. Poole (37), contains a 4.3-kb HindIII fragment carrying the mexAB-oprM operon with a 4.1-kb internal deletion (obtained by SacII digestion and religation). Plasmid pAL225 was created by cloning the 4.3-kb HindIII fragment from pRS14 into the HindIII site of pAL219. Third, plasmid pAL231 was created by inserting a 1.6-kb SalI fragment with a cat gene (isolated from plasmid pX1918-Cm) into the unique SalI site of pAL225 located in the oprM gene, close to the remaining SacII site. A final construct, plasmid pAL232, was obtained by ligating the 6.9-kb NotI fragment from pMOB3-Gm into the NotI site of pAL231. Besides the mexAB-oprM sequence being partially replaced with the cat gene, this plasmid also contained sacB and oriT.

Construction of recombinant plasmids for deletion of mexEF-oprN operon. Plasmid pAL241 was constructed to replace a part of the sequence of the mexEF-oprN operon with the mercury resistance [mer (Hg^r)] operon in the chromosome of P. aeruginosa. Construction was performed as follows. A 0.65-kb portion of the mexE gene was amplified from the chromosome and was subsequently ligated into the EcoRI and BamHI sites of pNOT19 to create pAL234. Primers MexE-EcoRI (GCTGAACGAGTGGGACGAATTCAC) and MexE-BamHI (CAGGATCCGGTTGACCTGGTTGTCGA) were used. A 0.98-kb portion of the oprN gene was amplified from the chromosome with primers OprN-BamHI (CGGGATCCAACGATCGCTTCCCGGT) and OprN-HindIII (CTCAAGCTTGGTGCCTTCGCGGTACGGAT). The resulting PCR fragment was ligated into the BamHI and HindIII sites of pAL234 to create pAL237. The 5.5-kb BamHI fragment of pHP45Hg (4) containing the mercury resistance determinant was ligated into the BamHI site of pAL237 to create plasmid pAL239. The mercury resistance determinant therefore separates the two gene fragments. The final construct, pAL241, was obtained by ligating the 6.7-kb NotI fragment of pMOB3 containing the sacB, oriT, and aacC1 genes into pAL239.

Construction of a recombinant plasmid for deletion of the mexCD-oprJ operon. Plasmid pAL224 was constructed to replace a part of the sequence of the mexCD-oprJ operon with the gentamicin resistance gene in the chromosome of P. aeruginosa. Construction was performed as follows. First, we constructed plasmid pAL215. The 0.95-kb EcoRI-BamHI fragment that contains the gene nfxB and the 5' end of the gene mexC and the 0.97-kb BamHI-HindIII fragment containing part of the gene oprJ were inserted into pNOT19 to obtain pAL215. The nfxB-mexC fragment was obtained by chromosomal PCR with the primers NfxB-EcoRI (TTTGAATTCGCCAAGTGCCAGTATCG) and NfxB-BamHI (TTTGGATCCCGATCCTTCCTATTGCACG); the oprJ fragment was obtained by chromosomal PCR with the primers OprJ-BamHI (GGGGGATCCGAGTACGAACTGGACCTC) and OprJ-HindIII (CCCAAGCTTTAGCACCGTTTCCCACAC). Second, a 2.4-kb BamHI fragment from pX1918-GT containing a gentamicin marker was inserted between the nfxB gene fragment and the oprJ gene fragment to create pAL217. The final construct, pAL224, was created by ligation of a 5.3-kb NotI fragment obtained from pMOB3 containing sacB, oriT, and a cat gene into pAL217.

Deletions of efflux pump operons in chromosome of P. aeruginosa. Plasmids pAL224, pAL232, and pAL241 were transformed into E. coli S-17 (36) and were subsequently mobilized into various strains of P. aeruginosa via conjugation. Conjugation was performed as described elsewhere (30). Subsequent sucrose selection rendered strains PAM1360, PAM1536, and PAM1610, which were then used as sources of the mexCD-oprJ::Gm, mexAB-oprM::Cm, and mexEF-oprN::Hg deletions, respectively.

Gene replacement. Strains PAM1106 (PAM1020 mexA::Tc) and PAM1154 (PAM1020 oprM::Hg) were obtained by transducing tetracycline (Tc) or Hg resistance markers from strains K590 (30) or K613 (31), respectively, which were kindly provided by K. Poole. Strain PAM1064 (PAM1020 mexA-phoA::Tc) was constructed as follows. Plasmid pSUP202-mexA-phoA (a gift from K. Poole) contains the mexA-phoA transcriptional fusion inserted into vector pSUP202 (which confers the Tc^r Cb^r Cm^r phenotype), which cannot replicate in P. aeruginosa but which does contain the mob (mobilization) site. This plasmid was mobilized into P. aeruginosa PAM1020. One of the transconjugants, PAM1064, was confirmed by PCR and its antibiotic susceptibility profile to contain a chromosomal mexA-phoA fusion, an intact and functional mexAB-oprM operon, and closely linked plasmid-encoded Tc^r and Cb^r markers.

SDS-PAGE and Western immunoblotting. Sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) was performed by a previously described protocol (6) with 10% (wt/vol) acrylamide in the running gel. Proteins separated by SDS-PAGE were electrophoretically transferred to a nitrocellulose membrane (BA85; Schleicher & Schuell) as described previously (7), with the exception that SDS (0.1% [wt/vol]) was included in the buffer and transfer was carried out at 100 mA for 90 min. The membranes were processed as described previously (6) with murine monoclonal antibodies specific for the OprM, OprJ, or OprN protein (obtained from N. Gotoh) as the primary antibodies and alkaline phosphatase-conjugated goat antibodies to mouse immunoglobulin G as the secondary antibodies (Bio-Rad). The blots were developed with the AP Conjugate Substrate Kit (Bio-Rad) by the manufacturer's protocol.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Creation of isogenic strains overexpressing individual efflux pumps. Strain PAO1(PAM1020) was chosen as a parent strain for all subsequent selection and construction procedures (Table 1). Two types of selections were used: PAM1020 was plated either on LBA plates with levofloxacin at 4× the MIC (1 µg/ml) or on plates with combinations of antibiotics. The latter procedure was based on the previously reported susceptibility profiles for the mutants overexpressing individual efflux pumps (19, 20). Each mutant was profiled with a panel of antibiotics. Mutants that did not show the multidrug-resistant phenotype were tested for the presence of gyrA mutations (QRDRs were PCR amplified and sequenced). Two types of gyrA mutants were isolated: gyrA (Asp87Tyr) and gyrA (Thr83Ile), as exemplified by strains PAM1324 and PAM1548, respectively. Mutants with mutations in nalB (resulting in overexpression of MexAB-OprM) and nfxB (resulting in overexpression of MexCD-OprJ) were isolated at a frequency of 10⁶ to 10⁷, gyrA mutants were isolated at a frequency 10⁸, and nfxC mutants (resulting in overexpression of the MexEF-OprN pump) were isolated at a frequency of 10⁹ to 10¹⁰. Overexpression of individual efflux pumps in multidrug-resistant mutants (nalB, nfxB, and nfxC) was confirmed with monoclonal antibodies (obtained from N. Gotoh) that were raised against the OprM, OprJ, or OprN protein (data not shown). LVX MICs (1 to 2 µg/ml) were comparable for gyrA and pump-overexpressing mutants.

Creation and characterization of isogenic mutants lacking individual efflux pumps and combinations of pumps. In order to model the effects of inhibition of multiple pumps, we have constructed the strain that lacks all three known efflux pumps and strains that lack one or two efflux pumps in various combinations. Strains with deletions of individual operons and the MexAB-OprM/MexCD-OprJ double knockout were reported previously (6, 11, 29-31, 38), and their viabilities were not impaired. Neither the double-deletion mutants nor the triple-deletion mutant that were constructed in the course of our work had detectable growth defects under laboratory conditions (data not shown). These data suggest that a drug that inhibits Mex pumps, singularly or in multiples, will have no antibacterial effect by itself.

Effect of overexpression of various efflux pumps on strains with gyrA mutations. We studied the effects of overexpression of various efflux pumps on strains containing mutations in the target genes. A series of strains with various gyrA mutations that also overexpress efflux pumps was constructed. To do so, we transduced nalB, nfxB, and nfxC mutations from strains PAM1032, PAM1033, and PAM1034, respectively, into PAM1324 with gyrA (Asp87Tyr) or PAM1548 with gyrA (Thr83Ile) mutations. Our results (Table 2) indicate that when both gyrA and efflux pump-overexpression mutations are present in the same strain, the MIC of LVX is increased above the MIC for either mutant alone. The gyrA mutation (Asp87Tyr) increased the LVX MIC fourfold for the strain in which efflux pumps were not overexpressed (compare PAM1020 and PAM1324), while the gyrA mutation (Thr83Ile) resulted in an eightfold increase in the MIC (compare PAM1020 and PAM1548). The same four- or eightfold increase in the MIC due to these mutations was also observed in strains which overexpressed any of these three efflux pumps (Table 2). Since various efflux pumps confer slightly different levels of resistance to LVX to begin with, the MICs of this antibiotic for the resulting transductants were also different.

Effect of mexAB-oprM operon on strains with multiple target mutations. To establish further the contribution of efflux pumps to acquired resistance to fluoroquinolones, we investigated the effects that an efflux pump(s) would have on the strains with multiple target mutations. To obtain such mutants, we used stepwise selection by increasing the concentrations of LVX in the medium. After the first step of selection we obtained both efflux and target-based mutant strains, and all of them had comparable susceptibilities to LVX (MICs, 1 to 2 µg/ml). It is noteworthy that efflux mutants arose at a higher frequency (see above). The stepwise mutants were obtained in the following order: PAM1020 (wild type) > PAM1032 (nalB) > PAM1573 (nalB gyrA [Thr83Ile]) > PAM1582 (nalB gyrA [Thr83Ile] parC [Ser87Leu] gyrA [Asp87Tyr]). For quadruple mutant PAM1609 the LVX MIC was 128 µg/ml (Table 3).

Effect of deleting efflux pump operons on the emergence of clinically relevant resistance to fluoroquinolones. Since overexpression of any of the efflux pumps will lead to increased resistance to LVX, one can hypothesize that the frequency of emergence of resistant variants will be decreased if efflux pumps are inactive. Various deletion mutants were used to test this hypothesis. Selection was performed at 1 µg/ml (4× the MIC for the wild type). The results are presented in Table 4. Deletion of only individual efflux pumps did not alter the frequency of emergence of resistant mutants compared to that for the wild-type strain (despite the low level of resistance of the mexAB-oprM mutant PAM1554, for which the MIC was 0.015 µg/ml). The mutants isolated in this experiment were shown to overexpress the MexCD-OprJ efflux pump (data not shown). Two of the strains that lacked two efflux pumps, either mexAB-oprM mexEF-oprN (PAM1625 [MIC, 0.015 µg/ml]) or mexCD-oprJ mexEF-oprN (PAM1624 [MIC, 0.25 µg/ml]) also demonstrated no alteration in frequency. Mutants overexpressing MexCD-OprJ or MexAB-OprM were isolated from the double-knockout strains (data not shown). When the mexAB-oprM mexCD-oprJ double mutant was used in the selection (PAM1561 [MIC, 0.015 µg/ml]), the frequency was detectable but was significantly decreased. Mutants obtained from PAM1561 were confirmed to overexpress the MexEF-OprN efflux pump (data not shown). However, the frequency of emergence of LVX-resistant mutants was undetectable when the triple-deletion mutant PAM1626 (mexAB-oprM mexCD-oprJ mexEF-oprN [MIC, 0.015 µg/ml]) was used in the selection experiments with LVX at 1 µg/ml). Importantly, no target-based mutations were isolated under these selective conditions. Mutants with a low level of LVX resistance were isolated at a frequency of 10⁸ to 10⁹ when selection was performed with LVX at 4× the MIC (0.05 µg/ml) for the triple-deletion mutant. This frequency is in good accordance with that expected for target-based mutations.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

We have chosen P. aeruginosa and fluoroquinolone antibiotics to evaluate the consequences of inhibition of efflux pumps in this organism. One obvious expectation from inhibition of the efflux pumps became apparent after several groups reported that the MexAB-OprM efflux pump significantly contributes to the high intrinsic resistance in P. aeruginosa (6, 31, 38). It is clear that inhibition of the MexAB-OprM efflux pump alone should decrease the intrinsic resistance of the wild-type strains of P. aeruginosa to many clinically relevant antibiotics that are the substrates of this pump. For example, as we have shown in this report, the susceptibility of the mexAB-oprM deletion mutant to LVX was increased eightfold compared to that of the wild-type strain. It is equally obvious that inhibition of multiple efflux pumps should reverse the acquired fluoroquinolone resistance associated with efflux pump overexpression. Indeed, susceptibility to LVX was increased 64-fold in the mutant that lacks three known efflux pumps (which would be the maximal expected effect of pump inhibition) compared to those for the strains that overexpress efflux pumps. However, the unqualified efficacy of efflux pump inhibitors for use in conjunction with fluoroquinolones may be argued, since efflux is not the sole mechanism of fluoroquinolone resistance and target modification mutations (in gyrase and topoisomerase IV) have been recognized to confer resistance to fluoroquinolones. To assess the relative contributions of the efflux pumps and the target modification in the acquisition of resistance to fluoroquinolones by P. aeruginosa, isogenic strains with various combinations of efflux and target mutations were used.

With these strains, it was demonstrated that overexpression of the mexAB-oprM operon due to a particular nalB mutation resulted in the same relative (eightfold) increase in resistance to LVX whether or not multiple target-based mutations were present in the same strain. This indicates that efflux contributes equally to fluoroquinolone resistance over a wide range of fluoroquinolone concentrations. Deletion of the MexAB-OprM efflux pump from the strain in which this pump was overexpressed resulted in a 64-fold reduction in the LVX MIC, independent of the presence of additional resistance mechanisms. These results indicate that, depending on the level of expression of efflux pumps, inhibition of the efflux pumps should result in 8- to 64-fold reductions in LVX MIC even for strains with target mutations. Analysis of isogenic mutant strains also showed that individual efflux- and target-based mutations resulted in comparable four- to eightfold increases in the LVX MIC.

An important observation that we have made, which is in a good agreement with previously reported results (12), is that frequencies of occurrence of mutants due to pump overexpression are ca. 10-fold higher compared with those due to target-based mutations, at least in the case of the MexAB-OprM and MexCD-OprJ pumps. Therefore, it is conceivable that a high proportion of mutants present among both moderately and highly resistant clinical strains of P. aeruginosa are efflux mediated. Indeed, recently, several laboratories have reported the presence of multiple resistance mechanisms, including efflux, in a single bacterial strain isolated from the clinic (3, 40). These observations further support the notion that an inhibitor of multiple efflux pumps will serve as a good LVX-potentiating agent.

Another important beneficial consequence of inhibition of multiple efflux pumps demonstrated in this report is the decreased frequency of emergence of P. aeruginosa strains with clinically relevant levels of resistance to fluoroquinolones. Specifically, the emergence of clinically relevant resistant mutants for which the LVX MIC is 1 µg/ml was nondetectable (<10¹¹) for the mexAB-oprM mexCD-oprJ mexEF-oprN triple-deletion strain (MIC, 0.015 µg/ml). While inhibition of the efflux pumps should prevent the appearance of efflux-mediated mutants, we also did not obtain strains with increased resistance due to target-based mutations. As we have shown here, in order for the bacteria without efflux pumps to grow under the selective conditions used (LVX at 1 µg/ml), such bacteria are required to acquire simultaneously at least three target-based mutations to attain the necessary level of resistance (Table 3, PAM1640). Multiple target-based mutations are required since, as we have shown in this report, a single target-based mutation provides only a four- to eightfold increase in LVX resistance. Furthermore, the simultaneous acquisition of multiple mutations in a single experiment is an extremely rare event. It is also noteworthy that no additional efflux-based mutants conferring an increase in LVX resistance like that provided by three known efflux-based pumps were selected from the triple-deletion strain. Similar effects of inhibition of efflux pumps on the frequency of emergence of resistance were obtained in experiments with Staphylococcus aureus. When selection for norfloxacin resistance was performed in the presence of the NorA efflux pump inhibitor reserpine (23), a significant decrease in the frequency of emergence of resistance was observed (18).

In conclusion, we have demonstrated that efflux pumps contribute significantly to LVX resistance in P. aeruginosa. Inhibition of efflux pumps will (i) decrease intrinsic resistance, (ii) significantly reverse acquired resistance, and (iii) result in a decreased frequency of emergence of P. aeruginosa strains highly resistant to fluoroquinolones. These results occur only with simultaneous inhibition of multiple efflux pumps in P. aeruginosa. The benefits of broad-spectrum bacterium efflux pump inhibitors for the control of LVX resistance in P. aeruginosa warrant vigorous searches for such inhibitors.