INTRODUCTION

Multiple antibiotic resistance in bacteria has been associated with the overexpression of endogenous efflux genes (for reviews see Refs. 1 and 2). In Pseudomonas aeruginosa, active extrusion (3, 4) combined with a tight diffusion barrier at the outer membrane (5, 6) are the main reasons for low specific resistance against diverse antibiotics and other toxic elements (7).

Multidrug resistance in P. aeruginosa is now attributed mainly to the overexpression of three sets of operons. Overexpression of the operon mexA-mexB-oprM is characteristic for nalB type multidrug-resistant mutants, (8, 9), whereas overexpression of the operon mexC-mexD-oprJ (10) is responsible for multidrug resistance in nfxB type mutants (11). A third type designated as nfxC has also been associated with a diminished intracellular drug accumulation (12). The genes coding for the proteins involved in the drug extrusion have been characterized recently (13) and named mexE-mexF-oprN.

A common characteristic of multiantibiotic-resistant mutants in P. aeruginosa is their broad resistance to quinolones, and hence one method generally used to assess multiantibiotic-resistant-type resistance mutations is based on determination of the time course of quinolones accumulation inside the cell. A critical review of the methods used for measuring the accumulation of quinolones discusses the weaknesses of these methods (14). A striking point observed by us and others is the fact that an increase in the MIC¹ of quinolone of several orders of magnitude is often reflected by only about a 2-fold decrease in drug accumulation (3, 11, 12). A high background due to unspecific surface binding combined with the release of a substantial fraction of accumulated quinolone during the long washing period are responsible for these results.

Fluorescent probes, which change their spectroscopic properties upon entering the cell are particularly suitable for uptake experiments because they fluoresce weakly in aqueous environments, but become strongly fluorescent in nonpolar or hydrophobic environments. The possibility of energized extrusion of fluorescent membrane probes was suggested previously (15). Sedgwick and Bragg (16) used DMP to study the role of the cell envelope and of efflux systems in the uptake of lipophilic cations in Escherichia coli acrA mutants that showed hypersusceptibility toward acridine, cationic dyes, detergents, and antibiotics (17, 18) and lacked the ability to extrude these compounds in an energy-dependent manner (19). More recently the combination of several fluorescent probes provided more insight into the extrusion process of the LmrA protein of Lactococcus lactis, which is a member of the ATP-binding cassette superfamily (20).

We previously reported the construction of mutants lacking one, two, or all three components of the efflux system formed by the mexA-mexB-oprM gene products (21). In contrast to MexA-, MexB-, and OprM-deficient mutants constructed by others that were reported to be equally hypersusceptible to many antibiotics (22), we were able to divide our mutants into two groups with different antibiotic susceptibility profiles. All the mutants deficient in OprM exhibited 8-16 times higher susceptibility against fluoroquinolone antibiotics, chloramphenicol, and azthreonam than the parent strain, whereas mutants deficient in MexA, MexB, or MexA,B were only 2-4 times more susceptible to these antibiotics than the parent strain. The findings of the ciprofloxacin accumulation experiments were in agreement with the MIC results, but this method provided little information about uptake kinetics. The use of fluorescent probes combined with genetically defined mutants lacking the subunit protein(s) is therefore a powerful approach to obtaining more insight into the extrusion mechanism associated with mexA-mexB-oprM operon products. In this study, we investigated the response of DMP, TMA-DPH, NPN, and ethidium bromide to the lack of mexA, mexB, or oprM gene product in P. aeruginosa.

EXPERIMENTAL PROCEDURES

The following reagents were obtained from the indicated sources and used without further purification: DMP, 1-(4-trimethylammoniumphenyl)-6-phenyl-1,3,5-hexatriene (TMA-DPH), and carbonylcyanide m-chlorophenylhydrazone (CCCP) were purchased from Sigma, and ethidium bromide (EtBr) and NPN were purchased from Wako Chemicals.

Strains used in this study are listed in Table I. All the strains were grown at 37 °C in LB medium (1% tryptone, 0.5% yeast extracts, 0.5% NaCl, 5 mM MgCl₂, pH 7.2). Experimental cultures inoculated at 10% with a fully grown overnight culture were rotated at 200 rpm at 37 °C for 4 h. The mutants lacking MexA, MexB, OprM, MexA-MexB, and MexA-MexB-OprM were abbreviated to MexA, MexB, OprM, MexA,B, and MexA,B-OprM, respectively.

Table I. Bacterial strains Strains Relevant properties Reference PAO4290 leu-10, argF10, aph-9004, FP Matsumoto collection TNP070  mexA derivative of PAO4290 21 TNP071  mexB derivative of PAO4290 21 TNP072  oprM derivative of PAO4290 21 TNP073  mexA, mexB derivative of PAO4290 21 TNP074  mexA, oprM derivative of PAO4290 21 TNP075  mexB, oprM derivative of PAO4290 21 TNP076  mexA, mexB, oprM derivative of PAO4290 21

Cells were harvested at 7000 × g for 10 min, washed once with 100 mM NaCl-50 mM sodium phosphate buffer (pH 7.0), and suspended again in the same buffer at A₆₀₀ = 0.1 in the presence of 0.05% of glycerol. Cell suspension at a higher cell density leads to rapid oxygen depletion in the uptake medium, resulting in respiration-related artifacts, especially during NPN and DMP accumulation experiments (23, 24). Experimental measurements were generally performed within 2 h after cell preparation.

Fluorescence measurements were performed at 30 °C. DMP, TMA-DPH, and NPN were dissolved in absolute methanol and EtBr in water to the final concentrations indicated in the figure legends. CCCP was also dissolved in absolute methanol. Control experiments indicated that the presence of methanol at the maximal experimental concentration of 2.5% did not have a significant effect. Fluorescence emission intensity was measured with a Hitachi 650-10S fluorometer equipped with a Lauda RM6 circulating water bath, and the data were registered on a Hitachi 056 recorder. Excitation and emission wavelengths for DMP were, respectively, 467 and 557 nm, for TMA-DPH 350 and 425 nm, for EtBr 520 and 590 nm, and for NPN 340 and 420 nm. Slit widths were all set at 5 nm for excitation and at 10 nm for emission.

RESULTS

The neutral probe NPN has been used to monitor changes in bacterial membranes from the energized to nonenergized state (23) or to monitor outer membrane permeabilization (25). The addition of NPN to the cell suspension leads to a rapid increase in the fluorescence intensity at 420 nm in comparison with fluorescence without cells (Fig. 1A). Initial rapid increments within a period of less than 1 min were higher in the deletion mutants when compared with the parent strain, but this level was nearly the same for all mutants. A second slower incremental phase at 0.5-4 min could be observed in all mutants. Eventual fluorescence intensity in the OprM and MexA or MexB mutants were, respectively, 3.7 and 1.7 times higher than that in the parent strain. This second phase was absent in the parent strain due to active efflux of the dye. In parallel with the increase in emission intensity, a blue shift in the emission spectra was observed (Fig. 1B) reflecting the movement of the probe to a more hydrophobic environment. These results suggested that the lowest fluorescence of NPN in the wild type cell may be due to lower accumulation of the dye resulting from extrusion of NPN by the MexA,B-OprM efflux system. Conversely, the highest and intermediate fluorescence intensities in the mutant lacking OprM (also MexA,B-OprM) and MexA or MexB, respectively, may have been attributable to a total or partial dysfunction of the Mex pump.

Increase of DMP fluorescence intensity in the presence of cells is attributed to movement of the dye molecules from the head group region of the cytoplasmic membrane bilayer into the region of the fatty acyl chains (24). An instantaneous rise in fluorescence emission at 567 nm upon addition of the probe to the cell suspension was followed by a slower phase in the OprM, MexA, or MexB (Fig. 2A). The rate of increment for OprM was significantly higher than for MexA or MexB, whereas no significant variation in the fluorescence intensity was detected in the parent strain. The final fluorescence level attained in the OprM mutant was about twice as high as that in the MexA or MexB mutants and 7 times higher than that in the parent strain. These results suggest that DMP may have also been efficiently extruded in the wild type strain. In the mutants, the dye accumulates at higher levels and fluoresces strongly, which is most likely attributable to pump dysfunction. These observations strongly suggest the involvement of mexA-mexB-oprM gene products in DMP and NPN extrusion. In contrast to NPN, the fluorescence emission spectrum and the increase in DMP fluorescence in the deletion mutants was not accompanied by a shift of the emission maximum (Fig. 2B). The differences in NPN and DMP fluorescence intensity increment among various kinds of mutants was correlated with the MICs of the antibiotics in these cells, reported elsewhere (21).

Before the discovery of active drug efflux systems, the outer membrane barrier was considered the main cause of low specific antimicrobial resistance in Gram-negative bacteria including P. aeruginosa (26). To evaluate the importance of the outer membrane barrier and the role of OprM in the overall accumulation rate, the impact of outer membrane permeabilization was investigated. DMP accumulations in PAO4290 and its Mex operon knock-out mutants in the presence and the absence of EDTA are presented in Fig. 3. If we assume that DMP fluorescence emission observed in the parent strain to be the minimum value and the emission in the OprM strain to be the maximum value, the addition of EDTA to PAO4290 resulted in a fluorescence increment of only 15% of the maximum value. The MexA,B mutant reached a fluorescence intensity of about 45% of that in the OprM mutant, which in the presence of EDTA was increased to 90%. The effect of EDTA on DMP fluorescence in the OprM mutant is reflected by a rise of only 10%. Small fluorescence increments in the presence of EDTA for PAO4290 and OprM were attributable to destruction of the outer membrane barrier. Furthermore, a small increment of DMP fluorescence in the parent strain after the addition of EDTA suggests that the ability to extrude the substrate compound is still maintained under these conditions. These results suggest that increased accumulation of the dye in OprM mutant is not due to a higher periplasmic dye concentration, because outer membrane permeabilization allows instantaneous equilibration between the outer medium and the periplasmic space. Consequently, we concluded that MexA,B and OprM form a functional unit during the extrusion process, in which OprM plays an indispensable role. We previously explained the lower MICs of antibiotics in OprM mutants compared with MexA and MexB by assuming the presence of unidentified inner membrane components that act as substitutes for MexA,B. Addition of EDTA showed the highest effect in the MexA,B mutant, leading us to believe that the interaction between OprM and the putative inner membrane pump is susceptible to this treatment.

The amphiphilic character arising from attachment of the positively charged phenyltrimethylammonium group to diphenylhexatriene allows orientation of this probe in the lipid bilayer perpendicular to the plane of the cytoplasmic membrane. Therefore, partitioning of TMA-DPH into the lipid bilayer was shown to be a biphasic process that resulted from rapid insertion of the dye into the outer leaflet of the cytoplasmic membrane, followed by slower transbilayer movement to the inner leaflet of the membrane (20). Addition of TMA-DPH to the cell suspensions resulted in rapid increase of the fluorescence intensity, followed by a slower incremental phase leading to the maximum level within about 2 min in the MexA or MexB mutant and the wild type strain (Fig. 4A). The mutant lacking OprM took a much longer time than the others to equilibrate. The first component of the biphasic process is similar in all strains and the difference in the fluorescence increase mainly occurs during the second phase. The slower second phase was absent in the parent strain, high in the OprM mutant and at an intermediate level in the MexA and MexB strains (Fig. 4). The results clearly showed that TMA-DPH is efficiently extruded via the Mex extrusion machinery, probably from the inner leaflet of the cytoplasmic membrane. Raising the TMA-DPH concentration by the factor of two or more in the parent strain brought the fluorescence intensity to the same level as that caused by half that concentration in the M mutant (Fig. 4B). This means that accumulation in the parent strain takes place only in the outer leaflet of the cytoplasmic membrane, whereas the dye accumulates in the OprM mutant in both the inner and outer leaflet of the inner membrane.

To address the question of whether or not TMA-DPH accumulates in the inner leaflet of the cytoplasmic membrane and is then extruded from this site by the extrusion system, we performed downhill efflux experiments. Cells were preloaded with concentrated TMA-DPH, and then the suspension was rapidly diluted. We observed a very rapid single-phase decrease in fluorescence of the wild type strain (Fig. 5). On one hand, the fluorescence decrease in ABM cells was preceded by a transient rise in fluorescence; then there was a slow decrease until the main equilibration was reached (Fig. 5). We interpreted these results to mean that transient increase in fluorescence represents slow movement of the dye from the cytoplasm to the inner leaflet then to the outer leaflet, because the mutant lacking the extrusion proteins cannot extrude the dye. In the wild type strain, the Mex pump takes up the dye from the inner leaflet very rapidly, and therefore, a sharp decrease in fluorescence represents only the diffusion from the outer leaflet that also occurs in the MexAB-OprM mutant. A downhill extrusion profile of TMA-DPH in the potassium cyanide-poisoned wild type cell is similar but slightly higher than that in the MexAB-OprM mutant, confirming that the extrusion was driven by the cellular energy.

In contrast to DMP, NPN, and TMP-DPH uptake, which monitors mainly accumulation of the dye to the cytoplasmic membrane, EtBr uptake reflects the accumulation of this dye into the cytoplasmic compartment. Because the MIC of EtBr in the wild type strain of P. aeruginosa is over 200 µg/ml, it is conceivable that a fraction of this high resistance is due to efflux. All the mutants showed markedly higher fluorescence than in the wild type strain (Fig. 6). In contrast to the hydrophobic membrane probes, EtBr did not show intermediate levels of uptake rates in the MexA, MexB, or MexA,B mutant. These results suggest the following possibilities: (i) the Mex extrusion machinery takes up the substrate also from the cytoplasm; and (ii) intact extrusion assembly is necessary for mediating extrusion from the cytoplasm. Addition of CCCP to the mixture induces a sharp increase in EtBr accumulation, confirming the proton motive force dependence of the extrusion process (Fig. 6, curve d).

To determine whether MexA,B-OprM-mediated efflux of the fluorescent dyes was saturable, we examined the effect of external dye concentration on the fluorescence increase (Fig. 7). Fluorescence intensity of DMP in the buffer solution remained constant in the range of 20-80 µM. A maximum saturation in fluorescence increase was reached in the case of the OprM or MexA,B-OprM mutant at a concentration of 50 µM. Increasing external dye concentration further results in the quenching of fluorescence. Fluorescence intensity in the parent strain reaches a steady saturation level at an external concentration of about 50-60 µM, which is at about half the fluorescence intensity of the ABM mutant. We calculated the apparent K_m value of DMP extrusion from the difference in the fluorescence intensity between the ABM mutant and the wild type strain and found the value to be about 10 µM. An apparent K_m of 0.2 mM for a putative carrier-mediated active efflux system was determined by Cohen et al. (27) by using proton motive force-dependent uptake of norfloxacin into everted inner membrane vesicle from E. coli.

Concentration-dependent fluorescence increase of DMP in PAO4290 (), TNP073 (), and TNP076 () and in sodium phosphate buffer without cells ().

DISCUSSION

Fluorescence probes with different intracellular binding modes were used to assess the role of the subunit protein of the Mex extrusion machinery in substrate extrusion by monitoring real time accumulation. The increase in fluorescence with DMP, NPN, and TMA-DPH were shown to be directly related to their accumulation in the cytoplasmic membrane (23, 28). In the present study, the highest level of NPN, DMP, and TMA-DPH accumulation was observed in the mutant lacking OprM, the intermediate level was observed in the mutant lacking MexA or MexB, and the lowest level was observed in the wild type strain (Figs. 1, 2, 3, 4). This order of fluorescence probe accumulation was a perfect match with the MICs of fluoroquinolones, chloramphenicol, and tetracycline in these mutants and the parent strain (21), suggesting the possibility that a low level of antibiotic accumulation may be a major factor causing the resistance. A correlation between the deletion of mexA-mexB-oprM operon products and MIC results was firmly established previously (21).

A conspicuous fact is that permeabilization of the outer membrane by EDTA treatment has a minor effect on the induction of a higher level of DMP accumulation in the wild type or OprM when compared with the effect on MexA,B (Fig. 3). This result indicates that the synergistic effect between the outer membrane barrier and extrusion activity in protecting the cell against lipophilic compounds is still valid, but the role of this barrier in the case of hydrophobic low molecular weight molecules, like DMP (molecular weight, 380) and NPN (molecular weight, 219), is less extensive than expected and defers the outer membrane permeability to a secondary role.

The outer membrane component of efflux systems in P. aeruginosa is thought to act as a sort of exit channel, because the whole extrusion machinery dysfunctions in the mutants carrying at least OprM. Furthermore, the intermediate level in the MIC of antibiotics in MexA,B suggested the possibility of a two-step extrusion by which OprM can extrude drugs directly from the periplasm. However, DMP accumulation experiments in outer membrane permeabilized cells demonstrated that MexA,B-OprM forms a tight functional unit during extrusion, because the EDTA treatment of the wild type strains only increased fluorescence a little (see "Results" and Fig. 3).

The mechanism of extrusion includes at least two steps. In the first step, the dyes and antibiotics located in the inner membrane are recognized by the functional unit composed of MexA and MexB. The second step leads to the release of the drug into the outer medium through OprM. Interaction between the inner membrane components (MexA and MexB) and OprM is the crucial step in triggering the transport of these compounds. We previously suggested the existence of other proteins located in the cytoplasmic membrane, capable of substituting the function of MexA,B (21). The putative functional homologue of MexA,B differs probably in specificity toward different substrates. In the case of ethidium bromide, the accumulation proceeded at the same rate and extent in all mutants (Fig. 6), suggesting that this putative exporter does not recognize this cytoplasmic probe. The MIC test of several hydrophilic -lactam antibiotics like azthreonam, cefsulodin, and cefoperazone in the MexA,B mutants did not show intermediate susceptibility, suggesting that the complementary system might not be able to initiate the extrusion of these compounds (21).

The interaction between outer and inner membrane proteins during extrusion is not specific to P. aeruginosa. Although suspected for some time, there is now a growing amount of evidence that outer membrane protein TolC is the functional counterpart of OprM for AcrA,B-mediated extrusion in E. coli (29). Deletion of TolC results in hypersensitivity to hydrophobic agents, even in AcrA,B overexpressed mutants.

The fact that membrane probes are particularly "sensitive" to mex-operon products is an indication that a crucial step in active extrusion occurs during the membrane crossing step. Bolhuis et al. (20) showed previously that the fluorescent membrane probe TMA-DPH is distributed into the phospholipid bilayer via a biphasic process. A first rapid entry of TMA-DPH into the outer leaflet is followed by a slower transbilayer movement to the inner leaflet. The second component of this process is absent when extrusion occurs in energized L. lactis cells. A similar behavior of TMA-DPH accumulation was observed in P. aeruginosa in which the slower transfer to the inner leaflet was notably higher in the knock-out mutants (Figs. 4 and 5). The role of MexA,B-OprM in removing membrane bound TMA-DPH was particularly obvious in the downhill extrusion experiments (Fig. 5). The slow diffusion step of the amphiphilic dye from the inner leaflet to the outer leaflet before rapid release from the membrane surface was absent in the parent strain. These results suggest the existence of a common mode of action between MexA,B-OprM and LmrA, the ATP-dependent multidrug resistance pump of L. lactis, which is the extrusion of dye that takes place from the inner leaflet of the cytoplasmic membrane.

Because most of the extruded compounds are hydrophobic, it is assumed that the recognition site of the extrusion system possesses lipophilic properties. A recognition site for hydrophobic compounds located in a lipophilic environment has many advantages when compared with a site located in an aqueous environment. Lipophilic interaction sites exposed to the aqueous phase would require complex solvations to stabilize the recognition site (30). Initiation of the extrusion process would necessitate complex interactions between the extruded substrate, the water molecules, and the recognition site. This would in turn impose strict structural conditions on the extruded compound. Hence, a hydrophobic site located in a hydrophobic environment can exist without solvation and stay accessible to a broad range of substrates.

Our results agree with previously reported observations on the human multidrug transporter P-glycoprotein (31, 32) for the MDR-related protein MRP (33) and finally the ATP-dependent drug extrusion in L. lactis. The MDR system driven by the MexA,B-OprM machinery is probably more complex than that described above, because the proteins connecting functions of two membranes are involved. However, well characterized genetic and biochemical systems in P. aeruginosa serve as excellent models for studying the extrusion of noxious compounds. We believe that these studies eventually contribute to understanding of MDR in all living organisms. We assume that a common mechanism drives the first step of the extrusion in all of these systems.