The MexJK Efflux Pump of Pseudomonas aeruginosa Requires OprM for Antibiotic Efflux but Not for Efflux of Triclosan

Bacterial strains and media.

The bacterial strains and plasmids used in this study are shown in Table 1. Luria-Bertani (LB) medium from Gibco (Gaithersburg, Md.) was routinely used as the rich medium for all bacterial strains. The minimal medium used was M9 (28) supplemented with 0.2% Casamino Acids (Difco, Detroit, Mich.). Antibiotics used in selection medium for Escherichia coli were as follows: ampicillin, 100 μg/ml; spectinomycin, 150 μg/ml; and tetracycline, 15 μg/ml. Those used for P. aeruginosa were as follows: carbenicillin, 200 μg/ml; spectinomycin, 400 μg/ml; tetracycline, 25 μg/ml; and triclosan, 25 to 50 μg/ml. Spontaneous Tri^r mutants were selected by plating dilutions of PAO238 cells on Pseudomonas isolation agar (Difco) (this formulation contains 25 μg of triclosan per ml), as previously described (5). Antibiotics were from commercial sources. Triclosan was a gift from KIC Chemicals (Armonk, N.Y.); hexachlorophene and dichlorophene [2,2′-methylenebis(4-chlorphenol)] were purchased form Sigma (St. Louis, Mo.) and TCI America (Portland, Oreg.), respectively.

Antimicrobial susceptibility testing.

MICs were determined by the twofold broth microdilution technique according to National Committee for Clinical Laboratory Standards guidelines (30) or by the E-test method (AB Biodisk, Piscataway, N.J.) (ciprofloxacin and tetracycline only).

General DNA and mutagenesis procedures.

All routine DNA procedures were performed by previously described methods (17, 39). For disruption of the mexJ gene, pPS1152 was electroporated (8, 9) into strain PAO238-1, and tetracycline-resistant (Tc^r) colonies were selected. The mexJ::Tn〈TET-1〉 mutants were identified as Tc^r and carbenicillin susceptible (Cb^s). Gene replacement at the mexJ locus was confirmed by using the inserts from pPS1150 and a 627-bp HindIII-SalI fragment from pBR322ΔAP (41), respectively, as mexJK- and tet-specific probes. A chromosomal ΔmexJKL deletion was isolated by deleting a 1,414-bp XhoI-NcoI fragment from pPS1168 and replacing it with a gentamicin resistance (Gm^r)-FRT (Flp recombinase target site) cassette from pPS856 (17). The resulting pPS1169 was electroporated into PAO238-1, and Gm^r colonies were selected. The ΔmexJKL colonies possessed a Gm^r Cb^s phenotype. Similarly, a chromosomal ΔmexL deletion strain was isolated by deleting a 371-bp XhoI fragment from pPS1175 and replacing it with a Gm^r-FRT cassette from pPS856, followed by return of the deletion into the PAO238 chromosome via conjugation using the gene replacement vector pEX18Ap and sucrose counterselection (17). Flp recombinase-mediated excision was achieved by utilizing a previously described procedure (17). The mutant genotypes were confirmed by PCR and Southern analysis with the inserts from pPS1168 (mexJK′L), pPS1175 (mexL), and pPS856 (Gm^r-FRT) as the probes. Genomic Southern blotting with biotin-labeled probes was performed as previously described (17). Mutants containing unmarked chromosomal Δ(mexXY) deletions were isolated by using previously described plasmid constructs and procedures (5) and were verified by PCR analysis.

Cloning and analysis of the mexJK-containing region.

For the cloning of DNA fragments carrying the Tri^r-conferring chromosomal region from PAO238-1, the bacteriophage mini-D3112-based in vivo cloning method of Darzins and Casadaban (6) was used with the following modifications. Strain PAO238-1 was lysogenized with D3112cts (6). The resulting strain was transformed with pADD948, and a lysate was prepared by heat induction (6). Cells of recipient PAO238 were grown overnight and infected with the mixed lysate as previously described (40). The samples were spread on LB plates containing 30 μg of triclosan per ml and incubated at 30°C. Colonies growing on these plates after 24 to 36 h were purified on the same medium and analyzed for the presence of recombinant plasmids. Cb^r- and Tri^r-conferring plasmids were confirmed by electroporation of PAO238. One plasmid was retained and named pJ22. The Tri^r-encoding region on this plasmid was localized by in vitro transposon mutagenesis with the EZ::TN〈TET-1〉 kit from Epicentre (Madison, Wis.) according to the manufacturer's protocol. An aliquot of the mutagenized plasmid pool was electroporated into PAO238, and Cb^r Tc^r transformants were selected and screened for loss of Tri^r. DNA templates were prepared by using the Qiagen miniprep kit, and the transposon insertion sites were determined by automated nucleotide sequencing. Sequencing reactions were primed by using the TET-1 FP-1 and TET-1 RP-1 primers from the EZ::TN 〈TET-1〉 mutagenesis kit. Homologous sequences were identified by using online BLAST searches of the National Library of Medicine databases.

Cloning and sequencing of mexL.

For PCR amplification of the mexL coding region from genomic and plasmid DNA templates, two primers were designed. Primer mexL-up (5′-CGTTCGAaTTCTTATACTGGGCGG) contained a single base mismatch (lowercase a) and introduced an EcoRI site (underlined) 70 bp upstream of the mexL start codon. Similarly, mexL-down (5′-ACTGGGTCGAcCACTGGGACATC) contained a single mismatch (lowercase c) and introduced a SalI site (underlined) 108 bp downstream of mexL. PCRs were performed with Pfu DNA polymerase (Stratagene). Reaction mixtures (50 μl) contained 1× Pfu buffer (Stratagene), 100 ng of each primer, 10 ng of plasmid DNA or 100 ng of chromosomal DNA, and 2.5 U of Pfu. The reaction mixture was subjected to the following thermal cycles: one cycle at 96°C for 5 min; 30 cycles (for plasmid DNA) or 35 cycles (for chromosomal DNA) of 96°C for 45 s, 60°C for 45 s, and 72°C for 1 min; and a final extension at 72°C for 10 min. Sequences of PCR fragments were determined by using the same primers employed for amplification. For complementation analyses, the PCR fragments were digested with EcoRI and SalI, gel purified, and then cloned between the same sites of pUCP20T (46) to yield pPS1153.

Construction of oprM, oprJ, and oprN expression vectors.

The oprM, oprJ, and oprN genes were PCR amplified from genomic DNA templates by using primers containing base mismatches (indicated by lowercase letters in each primer sequence) that introduced new restriction sites after PCR amplification. For oprM, the forward primer oprM-up (5′-CGAGGaaTtCAAGCAGCAGGCGTCCGT) introduced an EcoRI site (underlined) upstream of the oprM start codon and ribosome-binding site. The reverse primer oprM-down (5′-ACGCCAaGcTtAGGGTCGGCGTTCTTG) introduced a HindIII site (underlined) downstream of oprM. For oprJ, the forward primer oprJ-up (5′-GCAGCAAGCttGCACCCATCGAAC) introduced a HindIII site (underlined) upstream of the oprJ start codon and ribosome-binding site. The reverse primer oprJ-down (5′-CACCGgaTCCCACACGTTTACC) introduced a BamHI site (underlined) downstream of oprJ. The forward primer for oprN, primer oprN-up (5′-CGCGAAGCttGCCGCGCCGCCA), introduced a HindIII site (underlined) upstream of the oprN start codon and ribosome-binding site. The oprN reverse primer, oprN-down (5′-GAGTGGtCGAcTTCCATCGGCCG) introduced a SalI site (underlined) downstream of oprN. PCRs were performed with Taq DNA polymerase (Gibco). Reaction mixtures (50 μl) contained 1× Taq buffer (Gibco), 30 pmol of each primer, 100 ng of chromosomal DNA, and 2.5 U of Taq. For amplification of oprM, the reaction mixture was subjected to the following thermal cycles: one cycle at 96°C for 5 min; 35 cycles of 96°C for 45 s, 71°C for 45 s, and 72°C for 2 min; and a final extension at 72°C for 10 min. The oprJ and oprN genes were amplified using the same denaturation and extension conditions, except that the annealing conditions were changed to 60°C for 45 s and 70°C for 45 s for oprJ and oprN, respectively. The PCR fragments were digested with the appropriate enzymes, purified from an agarose gel, and ligated either to EcoRI-HindIII-digested pVLT35 for oprM, to HindIII-BamHI-digested pAK1900 for oprJ, or to HindIII-SalI-digested pAK1900 for oprN. This procedure generated pPS1180, pPS1162, and pPS1163, which express oprM, oprJ, and oprN, respectively, from P_tac (pVLT35) or P_lac (pAK1900). Expression of OprM, OprJ, and OprN was verified in E. coli and P. aeruginosa by Western blot analysis with specific antibodies, as previously described (3, 5).

Measurements of fluorescence in cell suspensions.

Cells for fluorescence measurements were grown overnight in M9 medium (28) supplemented with 1% thiamine and 0.2% Casamino Acids. The cells were washed in 100 mM NaCl-50 mM sodium phosphate buffer (pH 7.0) as previously described (31). The washed cells were adjusted to an optical density at 540 nm of 0.1 in the wash buffer containing 0.05% of glycerol and used for fluorescence measurements within 2 h. The substrate used in fluorescence measurements was 2-(4-diethylaminostyryl)-1-methylpyridinium iodide (DMP), and it was added to cell suspensions to a final concentration of 25 μM. The fluorescence emission intensity at room temperature was measured with a Fluorolog-3 fluorometer (Instruments S.A., Inc., Edison, N.J.), and data were recorded and analyzed with DataMax and Igor Pro software. Excitation and emission wavelengths for DMP were, respectively, 467 and 557 nm. Slit widths were set at 5 nm for excitation and at 10 nm for emission.

Construction of a mexJ-lacZ fusion.

A plasmid-borne fusion was constructed by ligating a 655-bp BamHI-XhoI fragment from pPS1176 between the same sites of pTZ110 (44) to form pJZ110. The fused fragment contained 470 bp of mexJ, the 94-bp mexJ-mexL intergenic region, and 71 bp of mexL coding sequence. The fusion plasmid was transferred to P. aeruginosa by electroporation. β-Galactosidase activity was measured and activity units were calculated as described by Miller (28).

Isolation and characterization of triclosan-resistant mutants.

When triclosan-susceptible cells of the Δ(mexAB-oprM) Δ(mexCD-oprJ) strain PAO238 (Table 1) were exposed to 25 μg of triclosan per ml, Tri^r derivatives were obtained at a frequency of 10⁻⁸. Four randomly picked Tri^r derivatives, PAO238-1 to PAO238-4, were further analyzed, and all of them were highly Tri^r (MICs of at least 128 μg/ml). None of them was resistant to hexachlorophene (MIC, 8 μg/ml) or dichlorophene (MICs, 32 to 64 μg/ml), two other antimicrobials with structures similar to that of triclosan. For comparison, in PAO1 dichlorophene was found to be a good substrate for MexAB-OprM (MIC, 128 μg/ml), but hexachlorophene (MIC, 4 μg/ml) was not. In addition, none of the mutants were resistant to any of the antibiotics tested (see footnote a to Table 2), presumably due to lack of an Opr channel (see below). Analysis of total outer membrane proteins did not reveal overproduction of any novel proteins (data not shown). Although the MexEF-OprN system supports the efflux of triclosan, in Western blots none of the four mutants expressed detectable levels of OprN (data not shown). This finding was not surprising, since MexEF-OprN is not derepressible in PAO238 because this PAO background harbors an 8-bp insertion in the mexEF-oprN activator mexT (23). These preliminary results indicated that the system(s) responsible for the Tri^r observed in these mutants was novel and most likely did not involve overproduction of a new outer membrane protein. Since all four Tri^r PAO238 derivatives behaved similar in preliminary analyses, we decided to concentrate on one isolate, PAO238-1, and to decipher the mechanism(s) responsible for its Tri^r phenotype.

Cloning of the triclosan resistance determinant from PAO238-1.

The phage D3112-based in vivo cloning technique was used to clone a ∼32-kb chromosomal DNA fragment from PAO238-1. The resulting plasmid, pJ22, conferred Tri^r when transferred to the susceptible parent strain PAO238. In vitro transposon mutagenesis was used to localize the Tri^r determinant in the chromosomal insert of pJ22. The relatively high frequency (∼13%) with which the transposon-mutagenized plasmid population lost their ability to confer Tri^r indicated that a relatively large region on pJ22 was required for the Tri^r phenotype. To further localize the Tri^r-encoding determinant to the mexJK region, the entire operon with its flanking DNA sequences was subcloned on a 6,945-bp NotI fragment (Fig. 1). As a control, the same fragment containing the 1,674-bp Tn〈TET-1〉 insertion in mexJ was also subcloned. When transformed into PAO238, the plasmid containing the unmutagenized NotI fragment conferred Tri^r, but the vector control and the plasmid containing mexJ::Tn〈TET-1〉 did not.

A novel efflux system is responsible for triclosan resistance in PAO238-1.

Nucleotide sequence analysis of one of the Tri^s transposon pJ22 insertion mutants revealed that it contained a single transposon insertion in a gene that was annotated as PA3677 in the published genome sequence (50). This gene (1,103 bp), which we named mexJ (Fig. 1), encodes the periplasmic MFP of an RND-type efflux system. The transposon in pJ22 is inserted between codons 110 and 111 of mexJ. Separated by 4 nucleotides from mexJ and in the same transcriptional orientation is another 3,077-bp gene, mexK, which encodes PA3676, a putative transmembrane protein sharing significant homology with other inner membrane transporter proteins of RND-type efflux systems of the AcrBF family. The proposed mexJK operon lies at 4.12 Mb on the PAO1 genome.

MexJ is 37 to 51% similar to MFPs from other P. aeruginosa RND efflux systems and most similar (51 and 48%, respectively) to PA156 and PA157, two MFPs shared by the same efflux pump with divalent cation transporter homology (50). Like other MFPs of the RND efflux protein family, MexJ is most likely a lipoprotein, since it contains a signal peptidase II cleavage site (FLAACGNG) at the end of a 20-amino-acid NH₂-terminal signal sequence. The role of acylation and membrane anchoring in MFP function remains unclear because many MFPs, including P. aeruginosa MexA (55) and E. coli AcrA (56), are functional as nonacylated proteins.

The inner membrane transporter protein MexK is 41 to 64% similar to transporters from other P. aeruginosa RND systems and most related (64% similar) to PA158, the transporter component of an efflux system with divalent cation transporter homology. Based on amino acid comparisons, MexJK is thus most related to PA156-PA157-PA158 than to the other known or proposed RND-type drug efflux systems in this bacterium. Although PA156-PA157-PA158 and the related CzrCBA divalent cation efflux pump were previously excluded from the list of putative RND-type drug efflux systems (50), our data suggest that they are even more related to MexJK than some of the known Mex proteins and therefore should be included in this list.

To verify that the mexJK operon was indeed responsible for triclosan efflux in PAO238-1, a chromosomal mexJ mutant was constructed by transferring the mexJ::Tn〈TET-1〉 to the PAO238-1 chromosome. The resulting mutant strain, PAO298, became Tri^s and behaved the same as the subsequently constructed ΔmexJKL mutant PAO314 (Table 2).

The MexJK efflux system requires OprM for efflux of antibiotics but not triclosan.

The molecular architecture of the MexJK operon is very similar to that of the MexXY system in that both systems do not contain their own Opr channel. Since it has been shown that MexXY requires OprM for antibiotic efflux (5, 25), we reasoned that PAO238-1 did not efflux antibiotics since it was lacking OprM. To test this hypothesis, PAO238-1 was electroporated with the OprM-expressing plasmid pPS1180 and its vector control, pVLT35. Only PAO238-1 containing pPS1180 effluxed tetracycline, erythromycin, and ciprofloxacin (Table 2), and to a lesser extent gentamicin and fusidic acid, but not carbenicillin, trimethoprim, dichlorophene, and hexachlorophene (data not shown).

Because the mexJK-overproducing strain PAO238-1 contains a wild-type mexXY operon, whose expression was recently shown to be inducible by several of its antibiotic substrates (25), we considered that inducible MexXY may have been at least partly responsible for the observed multidrug resistance phenotype observed in PAO238-1 expressing cloned OprM. To test this notion and to assess the true contribution of MexJK to multidrug resistance, we constructed three isogenic derivatives of PAO238-1: the Δ(mexJKL) strain PAO314, expressing only inducible MexXY; the Δ(mexXY) strain PAO327, which expresses only MexJK; and finally the Δ(mexJKL) Δ(mexXY) strain PAO325, expressing neither efflux system. These strains were then transformed with either the vector control pVLT35 or the OprM-expressing pPS1180, and MICs of various drugs were determined (Table 2). As expected, PAO314 effluxed the known MexXY-OprM substrates ciprofloxacin, erythromycin, tetracycline, and gentamicin (not shown) only when transformed with pPS1180 but did not efflux triclosan because it is not a MexXY-OprM substrate (5). In contrast, PAO327 expressing OprM from pPS1180 no longer effluxed ciprofloxacin and gentamicin (not shown) but still effluxed tetracycline and erythromycin, albeit more weakly than PAO314. Efflux of erythromycin and tetracycline in this strain was dependent on OprM. Levels of triclosan efflux were the same as those seen in PAO238-1 and were independent of OprM. Lastly, the Δ(mexJKL) Δ(mexXY) strain PAO325 effluxed neither triclosan nor any of the antibiotics tested. These results indicated that inducible MexXY, in concert with OprM, was mostly responsible for the multidrug resistance phenotype of PAO238-1, although MexJK contributed considerably to the triclosan, erythromycin, and tetracycline resistance of this strain. Neither the MexJK nor the MeXY system functioned with either OprJ or OprN, although the respective proteins were expressed in the cells used for MIC determinations, as indicated by Western blot analysis (data not shown).

The outer membrane channel requirement for MexJK was further assessed by using DMP. This fluorescent probe fluoresces intensely when present in nonpolar or hydrophobic environments but fluoresces weakly in aqueous environments (31). As previously shown (31), DMP is efficiently extruded from cells by the MexAB-OprM system expressed in PAO1 (Fig. 2). Even cells overexpressing OprM in the absence of other known pump proteins (PAO314/pPS1180) showed some DMP efflux. Efficient extrusion of DMP via MexJK in PAO238-1 required OprM, since a strain expressing MexJK alone (PAO238-1) behaved similarly to the Δ(mexJK) mutant PAO314. Although all of the cells studied for DMP efflux contained a functional MexXY system, its expression was not induced because the cells were not pregrown in inducing antibiotics. Therefore, DMP efflux measured in these cells was due to MexJK.

MexJ and MexK are both required for a functional triclosan efflux system.

Because MexJK effluxed triclosan independently of the expression of a known Opr channel, we reasoned that MexK may function in triclosan export independently of a MFP, similarly to AcrD, an RND aminoglycoside pump of E. coli (37). To test this hypothesis, plasmids overproducing MexJ and MexK were independently and in combination transformed into strain PAO325, which expresses no efflux pump either because of deletion of the structural genes (MexAB-OprM, MexCD-OprJ, MexXY, and MexJK) or because of a regulatory mutation (MexEF-OprN). PAO325 cells expressing MexJ or MexK alone did not efflux triclosan (Table 3), and neither did uninduced cells containing both the MexJ and MexK expression plasmids. However, cells overproducing MexJ and MexK after induction with IPTG (isopropyl-[beta]-d-thiogalactopyranoside) effluxed triclosan at levels observed in the MexJK-overproducing mutants PAO238-1 and PAO318 (Table 2).

The mexJK operon is negatively regulated by the mexL gene product.

The mexJK operon is separated by 94 bp from an upstream open reading frame, PA3678 or mexL, which is transcribed divergently from mexJK (Fig. 1). PA3678 was annotated as a putative transcriptional regulator of the tetracycline repressor (TetR) family, since it contains the signature motif of this transcription regulator family. We therefore considered that mexL might encode a repressor of mexJK expression. If this were true, then mexL would most likely be mutated in PAO238-1 and on pJ22 but should be wild type in PAO238. The mexL gene was PCR amplified from PAO238, PAO238-1, and pJ22. Whereas the mexL sequence from PAO238 was identical to the one established for PAO1, the mexL genes from PAO238-1 and pJ22 contained a single base change. This missense mutation replaced an alanine with a glutamic acid residue in the first helix of the putative helix-turn-helix domain (Fig. 1). The cloned mexL gene from PAO1 carried on pPS1153 restored Tri^s on PAO238-1 (Table 2). Strain PAO318 containing a 371-bp chromosomal ΔmexL, which removes the helix-turn-helix coding sequence, constitutively expressed mexJK, as evidenced by high-level triclosan resistance (Table 2) and high-level expression of a mexJ-lacZ transcriptional fusion (Fig. 3). Cloned mexL restored triclosan susceptibility on PAO318 transformed with pPS1153 (Table 2).