Importance of Real-Time Assays To Distinguish Multidrug Efflux Pump-Inhibiting and Outer Membrane-Destabilizing Activities in Escherichia coli

ABSTRACT

The constitutively expressed AcrAB multidrug efflux system of Escherichia coli shows a high degree of homology with the normally silent AcrEF system. Exposure of a strain with acrAB deleted to antibiotic selection pressure frequently leads to the insertion sequence-mediated activation of the homologous AcrEF system. In this study, we used strains constitutively expressing either AcrAB or AcrEF from their normal chromosomal locations to resolve a controversy about whether phenylalanylarginine β-naphthylamide (PAβN) inhibits the activities of AcrAB and AcrEF and/or acts synergistically with antibiotics by destabilizing the outer membrane permeability barrier. Real-time efflux assays allowed a clear distinction between the efflux pump-inhibiting activity of PAβN and the outer membrane-destabilizing action of polymyxin B nonapeptide (PMXBN). When added in equal amounts, PAβN, but not PMXBN, strongly inhibited the efflux activities of both AcrAB and AcrEF pumps. In contrast, when outer membrane destabilization was assessed by the nitrocefin hydrolysis assay, PMXBN exerted a much greater damaging effect than PAβN. Strong action of PAβN in inhibiting efflux activity compared to its weak action in destabilizing the outer membrane permeability barrier suggests that PAβN acts mainly by inhibiting efflux pumps. We concluded that at low concentrations, PAβN acts specifically as an inhibitor of both AcrAB and AcrEF efflux pumps; however, at high concentrations, PAβN in the efflux-proficient background not only inhibits efflux pump activity but also destabilizes the membrane. The effects of PAβN on membrane integrity are compounded in cells unable to extrude PAβN.

IMPORTANCE The increase in multidrug-resistant bacterial pathogens at an alarming rate has accelerated the need for implementation of better antimicrobial stewardship, discovery of new antibiotics, and deeper understanding of the mechanism of drug resistance. The work carried out in this study highlights the importance of employing real-time fluorescence-based assays in differentiating multidrug efflux-inhibitory and outer membrane-destabilizing activities of antibacterial compounds.

INTRODUCTION

Multidrug resistance among human bacterial pathogens remains a grave social concern. Numerous strategies have been proposed to curtail the rampant increase in multidrug resistance among human pathogens, ranging from the effective integration of pharmacokinetic and pharmacodynamic parameters and implementation of antimicrobial stewardship (1) to the development of novel antibiotics based on either an existing or a novel chemical scaffold, exploitation of new cellular targets, and directly tackling the cellular mechanisms that confer multidrug resistance (2). Efflux of antibiotics from the cell is one of the common mechanisms of antibiotic resistance in bacteria, with resistance developing when the rate of drug efflux across the membrane exceeds that of drug influx (3).

Bacterial genomes encode several membrane-bound multidrug efflux systems (4, 5). These systems are usually under the control of an intricate regulatory network, which, in response to the presence of drug and other stress molecules, increases the overall efflux activity and decreases the influx capacity (6). One of the most extensively studied multidrug efflux systems of the resistance-nodulation-division (RND) family is the AcrA, AcrB, and TolC complex of Escherichia coli proteins (7, 8, 9). AcrB is the inner membrane trimeric drug-proton antiporter (10, 11, 12). TolC is an outer membrane channel protein whose periplasmic aperture is critical to the removal of drugs from the cell (13, 14). AcrA, an inner membrane lipoprotein (15), through its interactions with both TolC and AcrB in the periplasm, completes the efflux pump assembly and catalyzes the opening of the TolC channel (16, 17, 18).

There are two main reasons why AcrAB-TolC is the most studied tripartite efflux system in E. coli: first, it is the only efflux system with broad substrate specificity that is constitutively expressed at high levels, and second, it and its homologs are frequently upregulated in drug-resistant Gram-negative isolates with clinical relevance (5). Mutants lacking a functional copy of any of the corresponding genes of the AcrAB-TolC complex display increased susceptibility to a diverse group of antibiotics, detergents, and dyes (19). The E. coli genome also encodes four other antibiotic efflux pump systems belonging to the RND family: AcrD, AcrEF, MdtABC, and MdtEF (4). Of these, only the AcrEF system appears to be functionally analogous to AcrAB (4, 20). However, the acrEF genes are normally not expressed or are expressed at low levels (21) so that deletion of acrEF exerts little effect on the drug susceptibility phenotype (19). Although a repressor-encoding gene, acrS, is located adjacent to acrEF, it does not regulate acrEF expression (22). Instead, AcrS, when artificially overexpressed from a plasmid replicon, represses expression of distally located acrAB (23). Thus, acrAB is under the control of two repressors, in addition to various global regulators (6), while acrEF expression appears to be under the negative control of a single global regulator, H-NS (24). Spontaneous hns mutations have not been obtained among drug-resistant revertants of a strain lacking acrAB, presumably due to acute pleiotropy (25). Instead, selections demanding antibiotic resistance in the ΔacrAB background frequently lead to resistant colonies bearing an insertion sequence (IS) upstream of the acrEF coding region (26, 27, 28). These illegitimate recombination events presumably activate acrEF expression by creating an artificial promoter and/or eliminating H-NS-mediated repression of the acrEF gene. It is unclear under what physiological conditions acrEF is expressed or whether mutations are the only means by which acrEF expression is enhanced.

Due to their high clinical relevance, inhibitors have been sought to reduce or abolish the activity of multidrug-resistant efflux pumps (for reviews, see references 29, 30, and 31). Phenylalanylarginine β-naphthylamide (PAβN) was one of the first lead compounds that showed potent inhibitory activity against a number of RND pumps, including AcrB (32). The authors showed that in strains expressing RND efflux pumps, PAβN causes minimum damage to membrane integrity and potential, whereas in strains lacking a major efflux pump, it can impose some toxic effects (20, 32). However, the conclusion that PAβN acts principally as an efflux pump inhibitor was questioned by three recent publications (33, 34, 35). One of the main objections stems from the observation that PAβN causes elevated toxicity, by damaging the membrane, in cells lacking major efflux pathways. AcrAB and AcrEF share high amino acid sequence identity (66.49% and 77.56%, respectively) and thus are expected to respond very similarly. It was therefore somewhat surprising that two reports concluded that PAβN is an AcrB-specific inhibitor (36, 37).

In this paper, we sought to resolve the controversy about PAβN′s cellular action by employing strains constitutively expressing AcrAB or AcrEF efflux pumps. We used two separate real-time assays designed specifically to measure efflux- or outer membrane-destabilizing activity in vivo. PAβN activities were compared with that of polymyxin B nonapeptide (PMXBN), which is known to destabilize the outer membrane permeability barrier (38). The data revealed that PAβN′s main cellular action, which is completed within 60 s after its addition, is to inhibit efflux pump activities of AcrAB and AcrEF. In contrast, its outer membrane-destabilizing action was found to be weak and occurred well after the inhibition of efflux. PMXBN acted in a fashion opposite to that of PAβN, i.e., it severely disrupted the outer membrane permeability barrier without inhibiting efflux.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

The bacterial strains used in the study are listed in Table 1 and were derived from MC4100 Δara (39) and MC4100 Δara ΔacrAB::scar (40). Luria broth (LB) was prepared from LB Broth EZMix Powder (Lennox). LB agar (LBA) medium contained LB plus 1.5% agar (Becton Dickinson). When necessary, novobiocin and erythromycin (5 μg/ml each) were added to the LBA medium. All cultures were grown at 37°C for the durations specified in Results below. Carbonyl cyanide 3-chlorophenylhydrazone (CCCP), Nile red, N-phenyl-1-naphthylamine (NPN), PAβN, PMXBN, and vancomycin hydrochloride were purchased from Sigma-Aldrich. Nitrocefin was obtained from BioVision. All other chemicals were of analytical grade. NPN and nitrocefin were dissolved in 95% ethanol and dimethyl sulfoxide, respectively. Nile red was dissolved in methanol.

TABLE 1.

Bacterial strains used in this study

Strain	Characteristics	Reference or source
RAM1292	MC4100 Δara714	39
RAM2370	RAM1292 ΔacrAB::scar	40
RAM2371	RAM2370 acrEF↑ (IS2; isolate 017)	This study
RAM2372	RAM2370 acrEF↑ (IS2; isolate 045)	This study
RAM2374	RAM2370 × ΔacrAB::Kmr (via P1 transduction)	This study
RAM2375	RAM2371 × ΔacrAB::Kmr (via P1 transduction)	This study
RAM2376	RAM2372 × ΔacrAB::Kmr (via P1 transduction)	This study
RAM2377	RAM2374 × zba::Tn10 (via P1 transduction) ΔacrAB::Kmr (ΔAcrAB)	This study
RAM2378	RAM2374 × zba::Tn10 (via P1 transduction) acrAB+ (AcrAB+)	This study
RAM2379	RAM2376 × zba::Tn10 (via P1 transduction) ΔacrAB::Kmr (ΔAcrAB AcrEF↑)	This study
RAM2381	RAM2376 × zba::Tn10 (via P1 transduction) acrAB+ (AcrAB+ AcrEF↑)	This study
RAM2521	RAM2377/pBR322 (Apr)	This study
RAM2523	RAM2378/pBR322 (Apr)	This study
RAM2525	RAM2379/pBR322 (Apr)	This study

Isolation of E. coli mutants expressing the chromosomal acrEF genes.

Drug-resistant revertants of an E. coli strain expressing a mutant AcrB protein defective in proper interaction with other efflux pump components (17) were isolated on medium containing erythromycin and novobiocin. In some instances, the reversion mutation mapped outside the acrAB and tolC loci. P1 transduction of the ΔtolC::Tc^r or ΔacrF::Cm^r allele into these drug-resistant revertants revealed that the drug resistance phenotype was dependent on TolC and AcrF. This observation and the prevailing knowledge that acrEF expression can be activated by the IS in the absence of acrAB (22, 26, 27, 28) prompted us to examine the acrEF promoter region by PCR and DNA sequencing. In several revertants, PCR amplification of the promoter region produced a DNA fragment that was roughly 1.3 kb larger than that amplified from the parental or a wild-type strain. DNA sequence analysis confirmed the presence of an IS2 element 90 bp upstream of the acrE start codon in three independent revertants. Interestingly, this was also the location of the IS2 element reported previously (26, 27, 28), indicating a likely hot spot for the integration of IS2. Purified envelopes were analyzed for the presence of AcrE by Western blotting using polyclonal antibodies raised against AcrA, which shares 66.49% amino acid sequence identity with AcrE. AcrE was readily detected from the strain carrying the IS2 element upstream of acrEF (data not shown). The activated acrEF allele is referred to as acrEF↑.

DNA methods.

The acrSE intergenic region and the region encompassing all of acrS and part of acrE were amplified from the chromosome by PCRs using two different forward primers complementary to regions approximately 300 (5′-CACCTCATGACTATTTATACGAGAGGC-3′) and 1,200 (5′-CAGGCTCAGGTAATGATTCGC-3′) nucleotides upstream of the acrE ATG codon and a reverse primer complementary to a region between the 61st and 69th codons of acrE (5′-GCGAACTTCGGCTATACGATAAGC-3′). DNA sequence analysis of the PCR-amplified products was accomplished using the acrE reverse primer.

Protein methods.

Whole-cell envelopes were isolated from cultures grown overnight by the French press lysis method as described previously (41). Proteins were analyzed by mini-sodium dodecyl sulfate (SDS)-polyacrylamide (11%) gel electrophoresis (PAGE) and transferred onto Immobilon-P polyvinylidene difluoride (PVDF) membranes (Millipore). The membranes were blocked overnight in 5% (wt/vol) nondairy cream. After blocking, the membranes were incubated with primary antibodies for 1.5 h. The primary antibodies used were raised in rabbits against 6×His-tagged AcrA and TolC-MBP (both at 1:10,000 dilution). After incubation with primary antibodies, the membranes were washed twice for 15 min each time and incubated for 1 h with the secondary antibody (goat anti-rabbit horseradish peroxidase [HRP]-conjugated IgG). TolC was visualized from the same membrane blot previously used to probe AcrA. Detection of HRP-conjugated secondary antibodies was performed using ImmunoStar HRP substrate (Pierce). Protein bands were visualized with the Bio-Rad Molecular Imager ChemiDoc XRS System.

Antibiotic susceptibility assay.

MICs of chloramphenicol, deoxycholate, erythromycin, nalidixic acid, novobiocin, PAβN, SDS, and vancomycin were determined by the 2-fold serial dilution method using 96-well microtiter plates. Approximately 10⁵ cells were used in each well containing 200 μl of LB or LB supplemented with different amounts of the inhibitor. The plates were incubated for 16 to 18 h at 37°C on a gently rocking platform. Optical densities were measured at 600 nm (OD₆₀₀) using a VersaMax ELISA Microplate Reader from Molecular Devices. MICs were determined from at least three independent cultures. The MIC values were determined as the lowest concentration of antibiotics/inhibitors at which the bacterial culture failed to reach an OD₆₀₀ of 0.1. Standard errors were no greater than 7%, except for the ΔacrAB strain against nalidixic acid, where it was 14%.

Efflux assays.

Efflux of NPN in live bacterial cells was carried out essentially as described by Lomovskaya et al. (32) with some modifications. Cultures grown overnight were centrifuged, and the pellets were washed with potassium phosphate buffer (20 mM, pH 7.0) containing 1 mM MgCl₂ (KPO₄-MgCl₂ buffer) and centrifuged again. The washed cell pellets were resuspended in KPO₄-MgCl₂ buffer. The cell suspension, at 4 × 10⁸ cells/ml, was treated with 100 μM CCCP for 10 min at room temperature, after which the cells were pelleted and washed twice with KPO₄-MgCl₂ buffer and then resuspended in the same buffer. NPN was added to a final concentration of 10 μM, and the cells were incubated at room temperature for 15 min and then transferred into a quartz cuvette and placed in a Varian Cary Eclipse fluorescence spectrophotometer. For NPN, the fluorescence intensity was measured every 1 s using excitation and emission wavelengths of 340 nm and 410 nm, respectively. The excitation and emission slit widths were set at 5 nm. At the 100-s time point, efflux of NPN was initiated by adding glucose (50 mM final concentration), and changes in fluorescence intensities were measured for 200 s. NPN efflux was blocked by adding PAβN (20-μg/ml, approximately 40 μM, final concentration) 200 s after the initiation of efflux. The slopes (m) [m = (y₂ − y₁)/(x₂ − x₁)] resulting from decrease or increase in NPN fluorescence were calculated and expressed as fluorescence intensity units per second (FI/s). Nile red assays were carried out essentially as described for NPN, with some modifications incorporated according to the method of Bohnert et al. (42), Briefly, the cells were incubated with 10 μM CCCP for 15 min at room temperature and then for an additional 15 min at 37°C with 10 μM Nile red. After incubation with the dye, the cells were pelleted by centrifugation for 5 min and then resuspended in KPO₄-MgCl₂ buffer. Fluorescence was measured immediately using excitation and emission wavelengths of 552 nm and 636 nm, respectively, with slit widths set at 5 mm.

Nitrocefin hydrolysis assay.

A breach in the outer membrane permeability barrier was assessed by an assay involving the hydrolysis of a chromogenic substrate of β-lactamase, nitrocefin (43). Cells harboring the bla-bearing pBR322 plasmid (44) were grown overnight in LB supplemented with 50 μg/ml ampicillin. The next day, the cultures were diluted 1:50 in the same medium and grown for 2 h to mid-log phase (OD₆₀₀ ≈ 0.5). The cells were pelleted and washed with KPO₄-MgCl₂ buffer and centrifuged. The washed cell pellets were resuspended in KPO₄-MgCl₂ buffer. Assays were carried out in triplicate in 96-well microtiter plates with 10⁸ cells (0.2 ml). When required, PAβN or PMXBN was added to a final concentration of 40 μM, followed by nitrocefin (100 μM). Absorbance at 486 nm was recorded every 15 s for 30 min immediately after the addition of nitrocefin.

RESULTS

Drug phenotypes of AcrAB⁺ and AcrEF↑ strains in the absence and presence of PAβN.

We first determined the MICs of various inhibitors (Table 2) using strains that either lack acrAB (ΔAcrAB; RAM2377), express acrAB (AcrAB⁺; RAM2378), or express acrEF in the absence of acrAB (ΔAcrAB AcrEF↑; RAM2379) (Table 1). Although AcrAB⁺ and AcrEF↑ strains behaved similarly in the presence of all seven compounds, the AcrEF↑ strain consistently showed a 2-fold-higher MIC of chloramphenicol and nalidixic acid and a 4-fold higher MIC of erythromycin than the isogenic AcrAB⁺ strain (Table 2). The only exception was SDS, which had a 2-fold-higher MIC for the AcrAB⁺ strain than for the AcrEF↑ strain (Table 2). We then determined the MICs of antibiotics and SDS in the presence of 10 μg/ml PAβN (Table 2). Note that the MIC of PAβN alone, which is also a substrate of the AcrAB and AcrEF pumps (Table 2), was found to be greater than 160 μg/ml in a strain expressing AcrAB or AcrEF. The presence of PAβN significantly reduced the MICs of chloramphenicol, erythromycin, nalidixic acid, novobiocin, and SDS, all of which are known substrates of the AcrAB pump. However, unlike these five inhibitors, the presence of PAβN produced no change in the MIC of the nonsubstrate antibiotic, vancomycin, in AcrAB⁺ and AcrEF↑ strains. The presence of PAβN did, however, reduce the MIC of vancomycin 2-fold in a ΔacrAB strain. The effect of PAβN on the MIC of deoxycholate could not be determined due to the appearance of a white precipitate upon mixing of the two chemicals in LB.

TABLE 2.

MICs of antibiotics and detergents

Inhibitor	MICa (μg/ml) for strain with genotype:
	ΔacrAB	acrAB+	acrEF↑
Chloramphenicol	1	4 (1)	8 (1)
Deoxycholate	800	≥1,600	≥1,600
Erythromycin	2	64 (2)	256 (4)
Nalidixic acid	1	4 (0.5)	8 (0.5)
Novobiocin	2	128 (4)	128 (4)
PAβN	20	>160	>160
SDS	≤25	1,600 (≤25)	800 (≤25)
Vancomycin	160 (80)	160 (160)	160 (160)

^aMICs in the presence of PAβN (10 μg/ml) are shown in parentheses.

Based on the MIC data, we chose three antibiotics—erythromycin, nalidixic acid, and chloramphenicol—to further evaluate MICs in the presence of low levels of PAβN to minimize any potential side effects of PAβN on membranes. These antibiotics have different sizes and biochemical properties: erythromycin is a relatively large (molecular weight, 733.93) and hydrophobic molecule compared to nalidixic acid, which is small (molecular weight, 232.34), amphipathic, and water soluble. Chloramphenicol is also small (molecular weight, 323.13) but, due to its hydrophobic nature, only moderately soluble in water. Because of their small size, nalidixic acid and chloramphenicol can readily penetrate the outer membrane through the porin proteins (45, 46, 47). In contrast, passage of erythromycin through the outer membrane is expected to be impeded due to its large size and hydrophobicity (45, 47). (Note that the fact that the ΔacrAB mutant displays hypersusceptibility to erythromycin [Table 2] shows that entry of erythromycin into the cell is not completely blocked in the wild-type cell.) Therefore, if the potentiating effect of PAβN is principally due to its destabilizing effect on the outer membrane permeability barrier, it would be expected to act synergistically and preferentially with erythromycin and not necessarily with nalidixic acid and chloramphenicol, which can already readily cross the unperturbed membrane (45, 46, 47). On the other hand, if PAβN acts as an inhibitor of AcrAB and AcrEF, it should act synergistically with all three antibiotics. The PAβN concentrations used in the antibiotic MIC assays were 1.25, 2.5, and 5 μg/ml. The MIC data showed that PAβN acted synergistically with all three antibiotics (Table 3). A slight difference from chloramphenicol may reflect nonoverlapping binding sites of PAβN and chloramphenicol (48). Alternatively, rapid outer membrane permeation of uncharged chloramphenicol compared to charged nalidixic acid or bulky erythromycin makes inhibition of efflux less significant for chloramphenicol than for the other two antibiotics. These results suggest that at low concentrations, PAβN′s potentiating effect on the three antibiotics tested is possibly due to its inhibitory action against the RND pump proteins and not to membrane destabilization.

TABLE 3.

MICs of antibiotics in the presence of different amounts of PAβN

Bacterial strain genotype	Antibiotic	MIC (μg/ml) with PAβN concn (μg/ml):
		0	1.25	2.5	5.0
acrAB+	Erythromycin	64	32	16	4
	Nalidixic acid	4	1	0.5	0.5
	Chloramphenicol	4	2	2	2
acrEF↑	Erythromycin	256	128	64	16
	Nalidixic acid	8	2	1	1
	Chloramphenicol	8	4	2	2

Real-time efflux assays with live cells.

The MIC data only indirectly reflect the efflux activities of the two pumps and the effects of PAβN. Moreover, prolonged incubation of cells with PAβN, even when present in small quantities, could potentially damage the envelope in addition to inhibiting the efflux activity. Therefore, to directly measure the efflux activities of the two pumps and the effect of PAβN on their activities, we conducted real-time assays with live cells by employing a fluorescent dye, NPN. NPN has been used previously to probe efflux activity (32, 49, 50).

NPN efflux assays were carried out according to a published report (32) with some modifications as detailed in Materials and Methods. Prior to carrying out NPN efflux assays, we performed a control experiment to ensure that the amount of NPN used did not lead to NPN self-quenching upon its accumulation inside the cell. For this, we incubated efflux-deficient ΔacrAB cells, which result in maximum accumulation of NPN inside the cell, for 15 min and then measured fluorescence either directly (Fig. 1A, main graph) or after washing the cells with KPO₄-MgCl₂ buffer (Fig. 1A, inset). NPN fluorescence was measured with excitation and emission wavelengths set at 340 nm and 410 nm, respectively. As can be seen from Fig. 1A, the use of NPN from 0 μM (ethanol only) to 20 μM produced a linear fluorescence output proportional to the amount used, showing no self-quenching by NPN at these concentrations. In subsequent experiments, 10 μM NPN was used.

FIG 1.

(A) Fluorescence outputs at different NPN concentrations. ΔacrAB cells grown overnight and washed with KPO₄-MgCl₂ were treated with 0, 2.5, 5, 7.5, 10, and 20 μM NPN for 15 min, and then fluorescence was measured either directly (main graph) or after washing the cells in KPO₄-MgCl₂ buffer (inset). The excitation and emission wavelengths were set at 340 nm and 410 nm, respectively. a.u., arbitrary units. (B) NPN efflux assays. The preparation of the cells used in the efflux assay is described in Materials and Methods. Efflux of NPN was initiated 100 s after the start of fluorescence measurement by adding 50 mM glucose. A rapid loss of fluorescence intensity shows AcrAB- or AcrEF-mediated efflux of NPN. NPN fluorescence was measured as described for panel A. The slopes (m; FI/s; ±standard deviations) were determined with Varian Cary kinetics software and were calculated from at least two independent assays.

We used isogenic strains that either had acrAB deleted (ΔacrAB), expressed wild-type acrAB (acrAB⁺), or had acrAB deleted but expressed acrEF from the chromosome due to the nonpolar insertion of an IS2 element 90 bp upstream of the acrE ATG codon (ΔacrAB acrEF↑). Efflux of NPN, preloaded in cells deenergized by CCCP treatment, was initiated by the addition of glucose, which reenergizes the membrane. NPN fluoresces weakly in an aqueous environment but strongly in a nonpolar environment of the cell (51) (see below). As expected, no significant reduction in the NPN fluorescence intensity was observed in the ΔacrAB strain after the addition of glucose (Fig. 1B). A weak decline in the fluorescence intensity (m = −0.07 ± 0.01 FI/s) after glucose addition is likely due to the combined efflux activities of other weakly expressed RND efflux pumps. The mean time for 50% NPN efflux in ΔacrAB cells was determined to be greater than 300 s. In contrast, a sharp decline in the NPN fluorescence intensity, with an m value of −4.6 ± 0.27 FI/s, was observed in cells expressing wild-type AcrAB (Fig. 1B). The mean time for 50% NPN efflux was 15 s based on two independent assays. Moreover, the lowest NPN fluorescence intensity, a drop of 91% from the preefflux intensity, was reached in just over 50 s after the initiation of efflux. In cells expressing the AcrEF pump, the average mean time for 50% NPN efflux was 18 s (m = −2.79 ± 0.01 FI/s), and it took approximately 100 s after the initiation of efflux s to reach the lowest fluorescence intensity (an 85% drop from the preefflux intensity) (Fig. 1B). Thus, cells expressing AcrAB consistently showed slightly better NPN efflux activity than those expressing AcrEF. The different outcomes of MIC and NPN efflux assays could reflect differences in substrate preferences, mechanisms of efflux by the two pumps, and/or their expression levels.

Effect of PAβN on NPN efflux.

We then examined the effects of PAβN on AcrAB- and AcrEF-mediated NPN efflux activities. In their original paper, Lomovskaya et al. (32) used a close analog of PAβN, MC-002,595, because PAβN′s own fluorescence interfered with the assay. We reevaluated this potential limitation under our NPN efflux assay conditions and in our strain background. We first conducted control experiments to test whether the presence of PAβN would quench NPN fluorescence either in KPO₄-MgCl₂ buffer (Fig. 2A) or in KPO₄-MgCl₂ buffer containing CCCP-treated wild-type cells (Fig. 2B). Samples containing NPN, PAβN or both were excited at 340 nm, and the emission spectra were measured from 370 nm to 520 nm. All the samples produced an extremely weak fluorescence signal that peaked at 385 nm (Fig. 2A). Apart from this common peak, buffer containing PAβN alone produced no measurable fluorescence (Fig. 2A). In contrast, NPN in buffer produced a modest fluorescence peak at 470 nm (Fig. 2A). The solution containing both chemicals produced an emission spectrum identical to that with NPN alone, showing that in solution, PAβN neither produces its own fluorescence nor interferes with that of NPN (Fig. 2A). In the presence of cells, NPN fluorescence increased dramatically, and the peak shifted to 410 nm (Fig. 2B). In contrast, the cell suspension containing PAβN produced no or very weak fluorescence after incubation for 1 min or 15 min, respectively; however, after 30 min, a modest fluorescence signal, also peaking at 410 nm, emerged (Fig. 2B). This signal was likely generated by β-naphthylamine (β-NA) upon internalization of PAβN (52). Just as was found in the phosphate buffer, the presence of both PAβN and NPN in the same cell suspension produced fluorescence spectra similar to that obtained from cells containing NPN alone, showing that the incubation of 40 μM PAβN with cells for as long as 15 min neither contributes to its own fluorescence nor interferes with that of NPN. From these control experiments, we concluded that when employing real-time assays, experiments involving PAβN-mediated inhibition of NPN efflux and lasting for 15 min can be conducted without concern for interference by PAβN or its metabolite with NPN fluorescence.

FIG 2.

Comparison of spectral properties of NPN and PAβN. (A) Measurement of fluorescence in KPO₄-MgCl₂ buffer containing NPN (10 μM), PAβN (40 μM), or both. Emission spectra (370 nm to 520 nm) were measured after exciting samples at the 340-nm wavelength. (B) NPN (10 μM) was added to one aliquot of wild-type cells grown overnight, which were treated with CCCP, washed, and resuspended in KPO₄-MgCl₂ buffer. After incubating the cells at room temperature with NPN for 15 min, fluorescence was measured as for panel A and then again after adding PAβN (40 μM) to the same aliquot and further incubating for 1, 15, and 30 min. Fluorescence was measured from the second aliquot of CCCP-treated cells after incubation with PAβN alone (40 μM) for 1, 15, and 30 min at room temperature.

We repeated the NPN efflux experiments shown in Fig. 1B, but with a modification: PAβN was added 200 s after the addition of glucose, when NPN fluorescence was lowest in wild-type cells due to steady efflux activity. The cell suspension buffer always contained 1 mM MgCl₂ to minimize potential damaging action of PAβN on the outer membrane. The addition of 20 μg/ml (∼40 μM) PAβN increased NPN fluorescence, with m values of 2.80 ± 0.42 and 4.46 ± 0.35 FI/s in cells expressing AcrAB and AcrEF, respectively (Fig. 3A). Even the weak decline in NPN fluorescence in the ΔacrAB mutant was blocked by PAβN (m = 0.63 ± 0.08 FI/s), consistent with a broad inhibitory action of PAβN. Interestingly, in ΔacrAB and ΔacrAB acrEF↑ cells the NPN fluorescence intensities reached preefflux levels 100 s after the addition of PAβN and remained high (Fig. 3A). In comparison, in acrAB⁺ cells, NPN fluorescence not only failed to reach the preefflux level, it began to drop again 100 s after the addition of PAβN (Fig. 3A). Together, these data showed that despite responding slightly differently, both AcrAB- and AcrEF-mediated efflux activities were strongly inhibited by PAβN within 60 s after its addition. It is worth noting that although we interpret an increase in NPN fluorescence after the addition of PAβN as reflecting inhibition of NPN efflux, the increase could also result from the combined effects of reduced NPN efflux and increased NPN influx. These possibilities were tested, as discussed below.

FIG 3.

Inhibition of NPN and Nile red efflux by PAβN. (A) NPN efflux was carried out as described in the legend to Fig. 1. PAβN (40 μM) was added 200 s after the initiation of NPN efflux. The slopes (m; FI/s; ±standard deviations) of efflux inhibition were determined with Varian Cary kinetics software and were calculated from at least two independent assays. (B) Nile red efflux assays with wild-type (acrAB⁺) cells were carried out as described in Materials and Methods. PAβN (40 μM or 80 μM) was added 200 s after the initiation of NPN efflux. Two simultaneous fluorescence measurements were made after exciting samples at 340 nm (for PAβN) and 552 nm (for Nile red) and measuring fluorescence at the emission wavelengths of 410 nm (for PAβN) and 636 nm (for Nile red). Ex, excitation; Em, emission.

Because NPN and a metabolite of PAβN, β-NA, have overlapping fluorescence properties (Fig. 2B), their fluorescence spectra in the efflux assay shown in Fig. 3A could not be distinguished in the same sample and at the same time. However, we overcame this limitation by employing Nile red, which, like NPN, is a substrate of AcrAB (42) and fluoresces strongly at 636 nm when excited at 552 nm. The efflux of Nile red in CCCP-treated wild-type cells was initiated by adding glucose, and 200 s post-efflux initiation, PAβN (40 μM or 80 μM) was added to inhibit efflux (Fig. 3B). The addition of 40 μM PAβN weakly inhibited Nile red efflux and produced no measurable fluorescence of its own at the emission wavelength of 410 nm. When added to 80 μM, PAβN strongly inhibited Nile red efflux and contributed 2 units of fluorescence, measured at 410 nm, 50 s after its addition. In contrast to this small increase, NPN fluorescence, also measured at 340-nm excitation and 410-nm emission wavelengths, rose over 10-fold 50 s after the addition of 40 μM PAβN (Fig. 3A). These data show unambiguously that PAβN or its metabolite did not directly contribute to an increase in fluorescence observed during the inhibition of NPN efflux (Fig. 3A).

Effects of outer membrane-destabilizing agents on the NPN efflux assay.

Next, we examined NPN efflux in the presence of EDTA and PMXBN, which are known to disrupt outer membrane integrity (38). These compounds were added after the initiation of NPN efflux by glucose so that their effects on efflux could be monitored in real time. We first tested EDTA, since the positive charge of PAβN could potentially displace some outer membrane-bound cations. However, addition of EDTA from 10 μM to 5,000 μM failed to increase NPN fluorescence significantly (data not shown), indicating that chelation of outer membrane-bound cations is not a critical factor in the PAβN-mediated increase in NPN fluorescence.

We then employed PMXBN, a compound previously used to provide evidence that PAβN, like PMXBN, acts by destabilizing the outer membrane permeability barrier (33, 35). As was done with PAβN, PMXBN was added 200 s after the initiation of NPN efflux. We used the same molar concentrations of PMXBN and PAβN (40 μM) for direct comparison. In the AcrAB⁺ background, addition of PMXBN only weakly elevated NPN fluorescence (Fig. 4A) (m = 0.07 ± 0.02 FI/s), indicating that neither NPN influx nor efflux was significantly altered. In contrast, a sharp increase in NPN fluorescence occurred upon the addition of PAβN (m = 2.05 ± 0.09 FI/s), presumably reflecting complete inhibition of NPN efflux (Fig. 4A). We also conducted the same analysis with the strain expressing AcrEF (Fig. 4B). Interestingly, addition of PMXBN elevated NPN fluorescence at a rate roughly three times greater than that observed in the AcrAB⁺ background (Fig. 4B). Nevertheless, PAβN-mediated increase in NPN fluorescence was significantly greater than that in the presence of PMXBN (Fig. 4B).

FIG 4.

Effects of PAβN and PMXBN on NPN efflux. NPN efflux was carried out as described in the legend to Fig. 1. PAβN or PMXBN (both 40 μM) was added 200 s after the initiation of efflux. Fluorescence intensities from experiments involving PAβN and PMXBN treatments were graphed together for comparison. Slopes (m; FI/s; ±standard deviations) were determined with Varian Cary kinetics software and were calculated from at least two independent assays. The relevant genotypes of cell types used in the assay are shown.

The strong response of PAβN in elevating NPN fluorescence is consistent with the idea that PAβN acts by inhibiting the activities of AcrAB and AcrEF. In contrast, the weak action of PMXBN shows that it neither significantly inhibits NPN efflux nor increases its influx above the rate of efflux. These data show that PAβN′s effect can be readily distinguished from that of PMXBN by employing an assay that monitors efflux in real time.

Assessment of the outer membrane permeability barrier.

The NPN assays described above were designed to specifically monitor the in vivo efflux activities of AcrAB and AcrEF and not the integrity of the outer membrane permeability barrier. Therefore, we employed an assay that directly measures disruption of the outer membrane permeability barrier. The assay uses nitrocefin, which is a chromogenic cephalosporin substrate of the periplasmically localized β-lactamase enzyme (43) and is a typical substrate of AcrB (53). Disruption of the outer membrane allows β-lactamase-mediated cleavage of the β-lactam ring of nitrocefin, resulting in the disappearance of the yellow color of nitrocefin and the appearance of the red color of the hydrolyzed product. Nitrocefin hydrolysis assays were carried out with freshly grown cells harboring the bla-bearing pBR322 plasmid (44) (Table 1). Preincubation of buffer-washed cells with PAβN or PMXBN was minimized so that their effects on the outer membrane could be monitored on a real-time basis. Immediately after the addition of PAβN or PMXBN (40 μM final concentration), nitrocefin was added to a final concentration of 100 μM, and absorbance was recorded at 486 nm for 30 min (Fig. 5). In buffer-washed control cells, the hydrolysis of nitrocefin was monitored without PAβN or PMXBN, and the OD₄₈₆ values of control samples were subtracted from those obtained from cells treated with PAβN or PMXBN (Fig. 5).

FIG 5.

Effects of PAβN and PMXBN on nitrocefin hydrolysis. Nitrocefin hydrolysis assays were carried out as described in Materials and Methods. The slope values, shown as milli-OD units/s, from PMXBN- and PAβN-treated cells were determined from OD₄₈₆ readings obtained between 60 and 300 s and between 300 and 900 s, respectively. The OD₄₈₆ values from PMXBN- and PAβN-treated cells were deducted from those obtained from untreated control cells. The slope values and standard deviations were calculated from three independent experiments.

The nitrocefin hydrolysis assays revealed that PMXBN severely disrupted the outer membrane permeability barrier, while PAβN had relatively little effect (Fig. 5). Quantification of slope values showed that the rates of nitrocefin hydrolysis in PMXBN-treated cells were 5 (AcrAB⁺) and 8 (AcrEF↑) times higher than those seen in the PAβN-treated cells (Fig. 5). Interestingly, PAβN-treated cells from the ΔacrAB strain (RAM2521) showed a nitrocefin hydrolysis rate that was 1.5- and 2-fold higher than those observed in the PAβN-treated AcrAB⁺ (RAM2523) and AcrEF⁺ cells (RAM2525), respectively (Fig. 5). This observation is consistent with the data showing elevated vancomycin susceptibility (Table 2) and shows that the outer membrane-destabilizing effect of PAβN can be compounded when PAβN is not removed from cells lacking the AcrAB or AcrEF efflux pump. Together with the NPN efflux assay, the nitrocefin hydrolysis experiment showed that the primary activity of PAβN is to inhibit multidrug efflux pumps, AcrAB and AcrEF.

DISCUSSION

In this study, we determined the effects of a general RND pump inhibitor, PAβN, on AcrAB and AcrEF activities. We found that PAβN, a broad efflux pump inhibitor of MexAB, MexCD, and MexEF of Pseudomonas aeruginosa and AcrAB of E. coli (32), also inhibits E. coli's AcrEF activity. Our data are consistent with those of Olliver et al. (22), who also found PAβN-mediated inhibition of AcrEF activity in Salmonella enterica serovar Typhimurium, but inconsistent with those of Viveiros et al. (37) and Amaral et al. (36), who concluded that PAβN is an AcrB-specific inhibitor. A reason for this discrepancy could be that their conclusion was based on the lack of a PAβN effect on tetracycline resistance resulting from transient, tetracycline-induced expression of AcrEF and other pump proteins in the absence of AcrAB (37).

We successfully used PAβN in the NPN efflux assay and showed, for the first time, that it is an effective and potent inhibitor of both AcrAB- and AcrEF-mediated NPN efflux. Prior to this study, PAβN′s use in the NPN efflux assay was avoided due to a concern about high background fluorescence (32). Instead, PAβN was used to monitor its intracellular accumulation and subsequent breakdown to highly fluorescent β-NA in strains expressing wild-type or mutant efflux pumps (52, 54). While conversion of PAβN to β-NA is a relatively slow process, e.g., taking 30 min to triple the fluorescence intensity over the background level in the wild-type strain (54) (Fig. 2), the inhibition of NPN efflux by PAβN is immediate, resulting in a 7-fold increase in NPN fluorescence within 75 s after the addition of PAβN (Fig. 3 and 4), even when only adding one-fifth the molar amount used in the accumulation studies (8). Moreover, PAβN alone produced no increase in fluorescence during the same period (Fig. 2B and 3B), indicating that neither the intrinsic fluorescence properties of PAβN nor those of its cleaved products, including β-NA, interfered with the NPN efflux assay. Thus, when used to measure rapid changes in fluorescence due to efflux inhibition, PAβN can be employed in NPN efflux studies without background fluorescence problems.

The robust inhibitory effects of PAβN on AcrAB and AcrEF efflux activities are compatible with the notion that, under our experimental conditions, PAβN first acts by blocking the RND pump activities. In contrast to PAβN, PMXBN, which is expected to be retained in the outer membrane, produces virtually no inhibitory effects on NPN efflux; a slight elevation in NPN fluorescence likely results from increased NPN influx rather than its reduced efflux. However, in an assay specifically designed to monitor a breach in the outer membrane permeability barrier, PMXBN acts far more destructively than PAβN. These contrasting effects of PAβN and PMXBN in real-time assays reiterate their primary activities as an efflux pump inhibitor (PAβN) and a destabilizer of the outer membrane permeability barrier (PMXBN). Unlike the real-time NPN efflux assay, where cells are exposed to PAβN for only a brief period, prolonged incubation of PAβN with cells in the MIC assay can conceivably produce an outcome stemming from both efflux pump inhibition and a breach in the outer membrane permeability barrier. Despite this possibility, we believe that the observed synergistic effect of PAβN, when used at extremely low levels of 1.25 and 2.5 μg/ml, with erythromycin and nalidixic acid (Table 3) is likely due to the inhibition of efflux pump activity rather than a breach in the outer membrane permeability barrier, since at these low concentrations, PAβN causes barely recordable hydrolysis of nitrocefin (data not shown).

Highly sensitive fluorescent probes, including 8-anilino-1-naphthalenesulfonic acid (ANS), fluorescein-di-β-d-galactopyranoside (FDG), and Sytox Green, have been used to gauge PAβN′s membrane-destabilizing effects in E. coli and P. aeruginosa cells (33, 34, 35). In these studies, changes in fluorescence, which indicate internalization of these normally impervious fluorescent compounds, were measured 10 to 20 min (33, 35) or over an hour (34) after incubation with various amounts of PAβN. In contrast, preincubation of cells with PAβN was avoided in this study to minimize any membrane damage prior to the start of real-time assays. Moreover, we found that ANS is a substrate of the RND-type efflux pumps (data not shown); therefore, PAβN-mediated inhibition of the RND pumps is expected to reduce ANS efflux from the envelope, resulting in higher fluorescence intensities, thus complicating interpretation of the data. When FDG, a substrate of AcrB, was used to study the effects of PAβN and other compounds in increasing membrane permeability, the authors noted that as little as 4 μg/ml of PAβN was enough to initiate the release of presumably intracellularly generated fluorescein, the cleaved product of FDG, from a strain lacking TolC (33, 35). However, since efflux-deficient cells, such as those lacking TolC, cannot remove PAβN from the cell, the secondary effects of PAβN are expected to be greatly amplified. The idea that efflux pumps help minimize secondary effects of PAβN is further corroborated by the fact that a drastically higher concentration of PAβN (128 μg/ml) was required to only moderately increase Sytox Green fluorescence in the wild-type E. coli strain (35). In contrast, 32 μg/ml of PAβN, the lowest concentration used by the authors, was sufficient to significantly elevate Sytox Green fluorescence in an E. coli strain lacking AcrB or TolC (35).

Consistent with the idea that AcrAB and AcrEF (when expressed) help minimize secondary effects of PAβN on outer membrane permeability, we found here that in strains lacking the major efflux activity, PAβN causes elevated susceptibility to vancomycin and an increased rate of nitrocefin hydrolysis compared to efflux-proficient strains (Table 2 and Fig. 5). Note that vancomycin, due to its large size (molecular weight, 1,485.71), is normally excluded from entering Gram-negative cells. However, E. coli mutants defective in the biogenesis of the outer membrane display elevated susceptibility to vancomycin (55, 56, 57, 58). Therefore, PAβN clearly has the potential to breach the outer membrane permeability barrier if not expelled from the cells. However, in an efflux-proficient strain, the primary effect of PAβN, when used in small amounts, is not to destabilize the outer membrane permeability barrier but instead to block the activity of the efflux pump. Cells that lack acrAB or do not express acrEF display an 8-times-lower MIC for PAβN than those expressing either of the two pumps (Table 2). This shows that, if not removed from the cell, PAβN causes a toxic effect by disabling some essential cellular function. A breach in the outer membrane permeability barrier does not appear to be the reason for the observed PAβN toxicity, since we observed that the presence of PAβN at the MIC level of an efflux-deficient strain (20 μg/ml) causes a modest 50% increase in the rate of nitrocefin hydrolysis in the ΔacrAB strain compared to the acrAB⁺ strain (Fig. 5). In contrast, PAβN (32 μg/ml) causes a significant increase in the fluorescence intensity of a DNA binding dye, Sytox Green, in a ΔacrB strain, indicating a breach in the inner membrane (35). Therefore, PAβN′s toxic effect in an efflux-deficient background likely stems from destabilization and/or depolarization of the inner membrane potential.

Previous studies have ruled out any significant effect of PAβN in depolarizing the inner membrane in efflux-proficient cells (32, 42). Even when used at 2.5 times (100 μM) the quantity used here (40 μM) in the NPN efflux assay, PAβN reduced the transport of a proton motive force-dependent substrate by only 40% (42). The authors concluded that the efflux-inhibitory activity of PAβN is not the result of energy dissipation. In conclusion, we suggest that PAβN, when used at low levels, specifically inhibits RND pump activities. However, high levels of PAβN in efflux-proficient backgrounds and relatively low levels in efflux-deficient backgrounds can lead to membrane destabilization. The data presented here highlight the importance of using real-time assays to assess primary and secondary effects of efflux-inhibitory and membrane-destabilizing compounds in vivo.