Phenothiazines and Thioxanthenes Inhibit Multidrug Efflux Pump Activity in Staphylococcus aureus

Efflux-related multidrug resistance (MDR) is a significant meansby which bacteria can evade the effects of selected antimicrobialagents. Genome sequencing data suggest that Staphylococcus aureusmay possess numerous chromosomally encoded MDR efflux pumps,most of which have not been characterized. Inhibition of thesepumps, which may restore clinically relevant activity of antimicrobialagents that are substrates for them, may be an effective alternativeto the search for new antimicrobial agents that are not substrates.The inhibitory effects of selected phenothiazines and two geometricstereoisomers of the thioxanthene flupentixol were studied usingstrains of S. aureus possessing unique efflux-related MDR phenotypes.These compounds had some intrinsic antimicrobial activity and,when combined with common MDR efflux pump substrates, resultedin additive or synergistic interactions. For S. aureus SA-1199B,which overexpresses the NorA MDR efflux pump, and for two additionalstrains of S. aureus having non-NorA-mediated MDR phenotypes,the 50% inhibitory concentration (IC₅₀) for ethidium effluxfor all tested compounds was between 4 and 15% of their respectiveMICs. Transport of other substrates was less susceptible toinhibition; the prochlorperazine IC₅₀ for acriflavine and pyroninY efflux by SA-1199B was more than 60% of its MIC. Prochlorperazineand trans(E)-flupentixol were found to reduce the proton motiveforce (PMF) of S. aureus by way of a reduction in the transmembranepotential. We conclude that the mechanism by which phenothiazinesand thioxanthenes inhibit efflux by PMF-dependent pumps is multifactorialand, because of the unbalanced effect of these compounds onthe MICs and the efflux of different substrates, may involvean interaction with the pump itself and, to a lesser extent,a reduction in the transmembrane potential.

INTRODUCTION

Top

Introduction

Efflux-related multidrug resistance (MDR) has become appreciatedas a significant complicating factor in the chemotherapy ofbacterial infections. This is most evident for Pseudomonas aeruginosa,an organism that possesses several three-component efflux systemsconsisting of an integral cytoplasmic membrane drug-proton antiporterof the resistance-nodulation-division family, a channel-formingouter membrane protein, and a periplasmic protein that linksthe previous two proteins (34). The most extensively studiedof these efflux systems is MexAB-OprM, which along with MexXY-OprMis constitutively expressed and is involved in the intrinsicresistance to multiple drugs observed for wild-type strains(23, 34).

Gram-positive bacteria also possess efflux pumps that are capableof conferring clinically relevant resistance to antimicrobialagents such as macrolides, tetracyclines, fluoroquinolones,and selected dyes and disinfectants (39). Most of the pumpsthat transport drugs in gram-positive bacteria are members ofthe major facilitator superfamily (MFS), which are membrane-basedpolypeptides with 12 or 14 transmembrane helices. MFS proteinsemploy the proton motive force (PMF) as the energy source fordrug translocation. These pumps can transport a single classof drug (TetK of Staphylococcus aureus) or multiple drugs (QacAand NorA of S. aureus, Bmr of Bacillus subtilis) (32). The NorAprotein is capable of translocating hydrophilic fluoroquinolonesand monocationic dyes and disinfectants. One study has shownthat it may be responsible for at least 10% of antiseptic resistancein methicillin-resistant strains of S. aureus (30).

Bacterial genome sequencing projects have revealed that as genomesize increases, so does the number of putative drug efflux transporters.S. aureus has a genome size of 2.8 Mb and possesses approximately253 open reading frames encoding putative transport pumps (22).Of these, perhaps 17 may be MDR efflux pumps. In addition toNorA, our laboratory has identified two unique and non-NorA-relatedMDR efflux phenotypes in S. aureus that support the existenceof multiple genes encoding such pumps in the genome of thisorganism (14, 15). These pumps could complicate therapy of S.aureus infections.

Inhibition of MDR efflux pumps can restore the activities ofantimicrobial agents that are substrates for these proteins.Inhibitors of the MexAB-OprM efflux system have been identifiedand examined in vitro and in animal models (24; D. Griffith,O. Lomovskaya, V. Lee, and M. Dudley, Abstr. 39th Intersci.Conf. Antimicrob. Agents Chemother., abstr. F1268, p. 327, 1999).Novel inhibitors of the NorA pump have also been identifiedbut have only been evaluated in vitro (1, 11, 25). Verapamiland reserpine are capable of inhibiting the NorA pump, but theconcentrations required are too high to be clinically relevant.The identification and development of safe and effective inhibitorsof bacterial efflux pumps are needed.

Phenothiazine and thioxanthene compounds, which are dopaminereceptor antagonists and calmodulin inhibitors and are usedclinically as neuroleptic and antiemetic agents, have been shownto have modest but broad antimicrobial activities (18-20). MICsare generally above clinically relevant concentrations, buttissue levels can be severalfold higher than those found inthe serum, and inhibitory concentrations may be achieved atthe site of infection (3).

The thioxanthenes demonstrate geometric stereoisomerism; inthe clinical setting the cis form possesses neuroleptic activity,but both the cis and trans forms have roughly equal antibacterialpotency (17, 18, 33). The combination of subinhibitory concentrationsof these compounds or phenothiazines with multiple standardantimicrobial agents commonly results in a synergistic effectagainst many species of bacteria (19, 20, 40). Phenothiazinesand thioxanthenes also have been shown to inhibit the functionof eukaryotic MDR efflux pumps such as p-glycoprotein and maybe useful as resistance-modulating agents for the treatmentof some drug-resistant tumors (8, 9).

The mechanism(s) by which phenothiazines and thioxanthenes exerttheir antimicrobial effect and/or potentiate the activity ofsuch a vast array of other antimicrobial agents against bacteriais not completely understood. It has been suggested that someof the potentiation of activity may be the result of inhibitionof efflux pumps. Chlorpromazine has been shown to affect potassiumflux across the membrane in both S. aureus and the yeast Saccharomycescerevisiae and to alter the transmembrane potential in Leishmaniadonovani (6, 21, 42). Either of these actions may affect thefunction of transport proteins that depend upon the PMF to energizesubstrate translocation.

In this study, the activities of several phenothiazine and thioxanthenecompounds against S. aureus were investigated. Of particularinterest was their potential inhibitory effects on efflux-relatedresistance phenotypes, particularly that conferred by NorA.We found that all tested compounds had modest intrinsic antistaphylococcalactivity, augmented the potency of common efflux pump substratesagainst S. aureus strains possessing different MDR efflux-relatedresistance mechanisms, and blocked NorA function and non-NorA-relatedefflux phenotypes in a concentration-dependent manner.

MATERIALS AND METHODS

Top

Materials and Methods

Media and reagents. The structures of the phenothiazine and thioxanthene compoundsinvestigated are shown in Fig. 1. Unless otherwise noted, allreagents were the highest grade available and were obtainedfrom Sigma Chemical Co., St. Louis, Mo. cis(Z)-flupentixol andtrans(E)-flupentixol were obtained from H. Lundbeck A/S, Copenhagen,Denmark. Cation-adjusted Mueller-Hinton broth (CAMHB) was usedas a growth medium (Mueller-Hinton II broth; Becton Dickinsonand Co., Cockeysville, Md.).

View larger version (16K):
[in this window]
[in a new window]
FIG. 1. Structures of phenothiazine and thioxanthene derivatives.

Bacterial strains. The strains of S. aureus employed are listed in Table 1. SA-1199,SA-K1712, and SA 8325-4 are the parent strains of SA-1199B,SA-K1748, and SA-K2068, respectively.

View this table:
[in this window]
[in a new window]
TABLE 1. Study strains

Determination of antimicrobial susceptibilities. MICs were determined by microdilution techniques according toNCCLS guidelines (28). For SA-1199B, SA-K1748, and SA-K2068,the effect of combining 20 µg of reserpine/ml or one-fourththe determined MIC of each phenothiazine and flupentixol stereoisomeron the MICs of the antimicrobial agents tested also was determined.All susceptibility tests were performed at least in duplicate.

Checkerboard combination studies were performed as describedpreviously using strains SA-1199B, SA-K1748, and SA-K2068 (7).Combinations tested included norfloxacin, ethidium bromide (EtBr),or tetraphenylphosphonium bromide (TPP) plus either prochlorperazineor trans(E)-flupentixol. Fractional inhibitory concentrations(FICs) were calculated as follows: (MIC of drug A or B in combination)/(MICof drug A or B alone). The FIC index was calculated by addingthe FIC values. FIC indices were interpreted as follows: $<=$ 0.5,synergy; 1.0, additivity; 2.0, indifferent; and $>=$ 4.0, antagonism.

Ethidium, acriflavine, and pyronin Y efflux. The loss of ethidium cation from S. aureus strains loaded withEtBr was determined fluorometrically as previously described(14). Experiments were performed in duplicate, and the resultswere expressed as the mean total efflux over a 5-min time course.Acriflavine and pyronin Y efflux was determined in exactly thesame fashion by using excitation and emission wavelengths appropriatefor each compound. The effect of increasing concentrations ofreserpine, each of the phenothiazines, and each geometric stereoisomerof flupentixol on the EtBr efflux of SA-1199B was determined.We also determined the effect of increasing concentrations ofprochlorperazine and reserpine on the EtBr efflux of SA-K1748and SA-K2068, as well as on the acriflavine and pyronin Y effluxof SA-1199B.

${Delta}$ pH determination. A modification of the method of Zhang et al. was employed tomeasure the transmembrane pH gradient (41). CAMHB (pH adjustedto 7.0 [CAMHB7]) was used throughout the procedure. SA-1199Bwas grown in CAMHB7 to an optical density at 660 nm of 0.8.Cells were recovered by centrifugation, washed, and then concentrated20-fold in ice-cold CAMHB7. Aliquots (1 ml) were placed on iceuntil used. Cells were warmed to 37°C in a shaking waterbath for 10 min, and at zero time 1 µCi of [¹⁴C]salicylicacid, used as a pH probe, was added (12 mCi/mmol; Sigma). Aliquots(50 µl) were removed at frequent intervals over a 20-mintime period and filtered through 0.45-µm-pore-size nitrocellulosemembranes, and the membranes were washed once with 2 ml of phosphate-bufferedsaline (pH 7.0). Additional aliquots were removed at varioustimes for the determination of protein content (Bio-Rad proteinassay kit; Bio-Rad Laboratories, Hercules, Calif.). Radioactivityretained on filters was measured in a scintillation counter,and nonspecific binding of salicylic acid to the filters wassubtracted. The experiment was repeated in duplicate, and resultswere expressed as mean micrograms of salicylic acid accumulatedper milligram of cell protein. The effects of prochlorperazine(30 µM), reserpine (20 µg/ml [33 µM]), andcarbonyl cyanide m-chlorophenyl hydrazone (CCCP; 100 µM),each added 10 min prior to the addition of salicylic acid, onsalicylic acid accumulation were also determined. A cell volumeof 4.2 µl/mg of cell protein was used (determined as previouslydescribed [26]), and change in pH ( ${Delta}$ pH) was calculated accordingto the method of Rottenberg (35).

${Delta}$ ${psi}$ determination. Accumulation of the membrane permeant cation TPP was used tomeasure ${Delta}$ ${psi}$ , the transmembrane electrical potential, by employinga modification of the procedure of Silverman et al. (36). Todetermine if TPP efflux by SA-1199B had a confounding effecton these experiments (TPP is known to be a good substrate forthe NorA MDR efflux pump [13]), SA-K1712 (in which norA is insertionallyinactivated; see Table 1) was also analyzed. Organisms weregrown overnight in CAMHB and then diluted 1:1,000 into freshprewarmed CAMHB and grown with shaking for an additional 2 hat 37°C. Cultures then were shifted to room temperatureand grown with shaking for another 2 h, followed by removalof several 350-µl portions to Eppendorf tubes embeddedin ice. These samples were warmed for 5 min to room temperaturein a shaking water bath prior to use. The protein content ofone aliquot was determined as described previously. One aliquotwas treated with 8% butanol (vol/vol) for 10 min prior to labelingto obtain a background value. Cells were labeled with 1 µM[³H]TPP (27 Ci/mmol; Amersham Biosciences Corp., Piscataway,N.J.) for 10 min in the presence or absence of prochlorperazine(30 µM), trans(E)-flupentixol (50 µM; SA-K1712 only),reserpine (33 µM), or CCCP (100 µM). Samples thenwere filtered through GF/C filters (Whatman, Maidstone, UnitedKingdom), which were washed once with 2 ml of 0.15 M NaCl. Radioactivityretained on filters was measured in a scintillation counter,and background counts were subtracted. Internal cell volumewas estimated as described previously, and the Nernst equation, ${Delta}$ ${psi}$ = -(RT/F) · ln([TPP⁺]_in/[TPP⁺]_out), was used to calculate ${Delta}$ ${psi}$ . The experiment was repeated three times, and results wereexpressed as the mean ± standard deviation ${Delta}$ ${psi}$ (in millivolts).

RESULTS

Top

Results

Antimicrobial susceptibilities and effects of inhibitors. For the sake of simplicity, reserpine, the phenothiazines, andboth flupentixol stereoisomers are referred to as inhibitors.MICs for the study strains are shown in Table 2. Within theconcentration range employed, some intrinsic antimicrobial activitywas measurable for all inhibitors except reserpine. Thioridazinewas more potent than the other phenothiazines against all strainsexcept SA-1199 and SA-1199B. The remaining four strains allare derived from SA 8325-4, so it is not surprising that theincreased susceptibility to thioridazine was consistent amongthem. Against these same strains, the trans isomer of flupentixolwas consistently twofold more potent than the cis form. Alsoof note is the fact that susceptibilities to phenothiazinesand thioxanthenes essentially were unaffected by the presenceof an efflux-related MDR phenotype, as MICs for parent and mutantstrains differed by no more than twofold.

View this table:
[in this window]
[in a new window]
TABLE 2. Susceptibilities of study strains

These data illustrate the effect that overexpression of NorAhas on norfloxacin, acriflavine, EtBr, pyronin Y, and TPP MICs(Table 2, compare SA-1199 with SA-1199B data), although someof the increase in the norfloxacin MIC for SA-1199B is relatedto its GrlA substitution (A116E). NorA overexpression had noeffect on the MICs of gentamicin, linezolid, nafcillin, or tetracycline,which are not substrates of NorA. SA-K1712 is an SA 8325-4 derivativewith norA insertionally inactivated by tetK, which encodes a14-transmembrane segment MFS transporter (10). The effect ofa single chromosomal copy of tetK can be seen by comparing thetetracycline MICs for SA-K1712 and its parent, SA 8325-4.

The SA-K1712 derivative SA-K1748 possesses a non-NorA-mediatedinducible MDR phenotype (14). The resistance profile of thisstrain includes organic cations such as EtBr and TPP, but itdoes not include fluoroquinolones. The raised norfloxacin MICin this strain is the result of a GrlA substitution (S80F) andnot drug efflux. SA-K2068 is an SA 8325-4 derivative selectedby exposure of the organism to gradually increasing concentrationsof both EtBr and moxifloxacin. It demonstrates a non-NorA-mediatedMDR phenotype that includes efflux-related resistance to selectedfluoroquinolones, including the C₈-methoxy fluoroquinolonesmoxifloxacin and gatifloxacin, as well as EtBr and TPP (15).As for the NorA-overexpressing strain (SA-1199B), neither SA-K1748nor SA-K2068 showed any significant change in susceptibilityto gentamicin, linezolid, nafcillin or, in the case of SA-K2068,tetracycline.

The effects of subinhibitory concentrations of each inhibitoron MICs of selected pump substrates for strains expressing efflux-relatedMDR phenotypes are shown in Table 3. In general, all inhibitorsreduced MICs by fourfold or more with the exception of SA-K1748,where norfloxacin susceptibility was minimally affected. Asstated previously, norfloxacin resistance in this strain isnot related to efflux but rather to a GrlA substitution (14).Differential effects were noted for some of the inhibitors;for example, with respect to the phenothiazines, fluphenazine,thioridazine, and prochlorperazine appeared most efficient inreducing the MICs of EtBr and TPP for SA-K1748 and SA-K2068.

View this table:
[in this window]
[in a new window]
TABLE 3. Effects of inhibitors on susceptibilities of strains with MDR phenotypes^a

Inhibitors generally reduced the MICs of gentamicin, nafcillin,linezolid, and tetracycline no more than twofold (data not shown).An exception to this was observed for SA-K1748, where reserpine,chlorpromazine, and both flupentixol stereoisomers (the onlyinhibitors tested against this strain) reduced tetracyclineMICs by four- to eightfold. It must be remembered that thisstrain possesses the tetK determinant, and it appears as thoughthese inhibitors may have some effect on the function of theTetK transporter.

Checkerboard combination study results are shown in Table 4.The combination of norfloxacin, EtBr, or TPP with either prochlorperazineor trans(E)-flupentixol was either additive or synergistic againstS. aureus strains with efflux-related MDR phenotypes.

View this table:
[in this window]
[in a new window]
TABLE 4. Combination studies

Effect of inhibitors on efflux. Figure 2 illustrates the effects of increasing concentrationsof inhibitors on EtBr efflux (Fig. 2A) and on acriflavine andpyronin Y efflux (Fig. 2B) of SA-1199B. In almost every casea concentration-dependent effect was observed, with the exceptionof reserpine-mediated inhibition of acriflavine and pyroninY efflux. The concentrations required to cause at least a 50%inhibition (IC₅₀) of EtBr efflux were 10 µM for reserpine,thioridazine, and prochlorperazine, whereas the remainder ofthe tested compounds did not result in this degree of inhibitionuntil a concentration of 30 µM was used. For both SA-K1748and SA-K2068, the IC₅₀ of reserpine and prochlorperazine forEtBr efflux was 10 µM (data not shown). With respect toacriflavine and pyronin Y efflux of SA-1199B, the IC₅₀s forprochlorperazine were 125 and 100 µM, respectively, whereasreserpine did not achieve this degree of inhibition at the highestconcentration tested (100 µM, near the limit of solubility).

View larger version (17K):
[in this window]
[in a new window]
FIG. 2. Effects of inhibitors on substrate efflux of SA-1199B (data are means of duplicate experiments). The horizontal line indicates the concentration necessary to inhibit efflux by 50%. (A) Inhibition of ethidium efflux. •, reserpine; ${circ}$ , chlorpromazine; ${blacktriangledown}$ , fluphenazine; ${triangledown}$ , thioridazine; ${blacksquare}$ , prochlorperazine; ${square}$ , cis(Z)-flupentixol; ${diamondsuit}$ , trans(E)-flupentixol. (B) Inhibition of acriflavine and pyronin Y efflux. • and ${circ}$ , inhibition of acriflavine by prochlorperazine and reserpine, respectively; ${blacktriangledown}$ and ${triangledown}$ , inhibition of pyronin Y by prochlorperazine and reserpine, respectively.

The concentrations of inhibitors required to produce $>=$ 50% inhibitionof EtBr efflux were significantly lower than their respectiveMICs. The ratio of the IC₅₀ for each compound to its MIC forSA-1199B ranged from a low of 0.04 (for thioridazine) to a highof 0.15 (for fluphenazine). Thus, for all phenothiazines andboth flupentixol stereoisomers concentrations of significantlyless than one-fourth the MIC reduced NorA-mediated EtBr effluxby at least half. Similar results were observed for the inhibitionof EtBr efflux by prochlorperazine versus SA-K1748 and SA-K2068;the IC₅₀-to-MIC ratio for this compound was 0.06 for both strains.However, the effect of prochlorperazine on acriflavine and pyroninY efflux of SA-1199B was much less, with IC₅₀-to-MIC ratiosof 0.76 and 0.61, respectively.

Effects of inhibitors on ${Delta}$ pH and ${Delta}$ ${psi}$ . At pH 7.0, the ${Delta}$ pH of SA-1199B in CAMHB was +0.6 to +0.8 overthe entire 20-min time course of the experiment (data not shown).The addition of reserpine or prochlorperazine had no effect,but the ${Delta}$ pH was completely abolished by CCCP. The ${Delta}$ ${psi}$ of SA-1199Bin CAMHB was -148 ± 2 mV, and in the presence of prochlorperazine,reserpine, and CCCP ${Delta}$ ${psi}$ was -105 ± 6, -158 ± 1, and0 mV, respectively. For SA-K1712 the results were -146 ±3 mV for CAMHB and -164 ± 1 mV for reserpine. At theconcentrations employed, prochlorperazine, trans(E)-flupentixol,and CCCP reduced ${Delta}$ ${psi}$ of SA-K1712 to zero.

DISCUSSION

Top

Discussion

Bacterial efflux pumps clearly contribute to the increasingproblem of MDR. Identification of inhibitors of efflux pumpsfor which antimicrobial agents are substrates is an active areaof research in both the pharmaceutical and academic sectors(1, 11, 24, 25). The study of inhibitors of the MexAB-OprM effluxsystem is perhaps the most advanced, but to our knowledge onlylimited in vivo testing has been done and no human toxicologyof promising candidates has been performed. Compounds that inhibitthe activity of NorA also have been described, including theproton pump inhibitor omeprazole, but only in vitro testinghas been done. In addition, the concentration of omeprazolerequired was above what is clinically attainable (1). Obviously,necessary characteristics of a pump inhibitor are that it beactive at an achievable serum drug concentration and that ithas minimal toxicity.

The search for candidate efflux pump inhibitors can be costlyand time consuming. It makes reasonable sense to examine selecteddrugs already in clinical use as potential inhibitors. The pharmacokineticsand the toxicities of these compounds generally are known, obviatingthe need to determine these parameters. The choice of candidatesmay be difficult, but perhaps can be simplified by examiningdrugs for which some data already exist regarding additive orsynergistic antibacterial effects when combined with compoundsknown to be efflux pump substrates, such as fluoroquinolones,tetracyclines, macrolides, or quaternary amine-containing antisepticagents (39). Such data were available for the phenothiazinesand thioxanthenes and prompted the present study (17-20, 40).

We found that phenothiazines and two geometric stereoisomersof the related thioxanthene class possessed some intrinsic antistaphylococcalactivities, albeit at concentrations above those which are clinicallyachievable. This activity was unaffected by the presence ofa variety of efflux-related phenotypes, indicating that thesecompounds are not substrates for the involved pumps themselves.We found some evidence of stereoselectivity, which was strainrelated, in that trans(E)-flupentixol was twofold more potentthan the cis isomer against SA 8325-4 and its derivatives. Therewere no differences in the potencies of the flupentixol compoundsversus SA-1199 or SA-1199B, and each was equipotent in inhibitionof EtBr efflux in SA-1199B. The dissociation between the antibacterialand the anti-EtBr efflux effect (trans isomer equipotent orsuperior) and the neuroleptic effect (only present in the cisisomer) has been observed previously and is an important considerationin the possible design of more-potent inhibitors of the thioxantheneclass (17).

More interesting than intrinsic antistaphylococcal activitywas the fact that when combined at one-fourth their MICs withcommon efflux pump substrates all inhibitors augmented the antimicrobialactivity of those substrates. The greatest effects in this regardwere observed for EtBr and TPP, and the most potent inhibitorswere fluphenazine, thioridazine, and prochlorperazine, especiallyagainst strains SA-K1748 and SA-K2068. Checkerboard testingverified the above observations in that an additive or synergisticeffect was observed against all strains tested expressing efflux-relatedMDR phenotypes by combining prochlorperazine or trans(E)-flupentixolwith pump substrates.

Thioridazine and prochlorperazine were the most potent of thetested compounds with respect to inhibition of NorA-mediatedEtBr efflux, having IC₅₀-to-MIC ratios of 0.04 and 0.06, respectively,indicating significant NorA inhibition at concentrations wellbelow their respective MICs. Prochlorperazine was also an efficientinhibitor of EtBr efflux against SA-K1748 and SA-K2068, whereits IC₅₀-to-MIC ratio was also 0.06. However, a differentialeffect was observed when the effects of prochlorperazine onNorA-mediated efflux of acriflavine and pyronin Y were examined.IC₅₀ values which were more than 10-fold higher than those determinedfor EtBr efflux were observed. This observation, when combinedwith the differential effect of some inhibitors on MICs of selectedpump substrates (see above), suggests the possibility that thesecompounds may physically interact with the pumps themselves.There is evidence to support such a contention with respectto reserpine for the B. subtilis MDR efflux pump Bmr (2). Suchan interaction may affect pump affinity for one substrate morethan another, perhaps by altering a substrate recognition site,such that export proceeds with variable efficiency for differentsubstrates. Clearly, more work needs to be done to determineif such a physical interaction actually occurs.

Unfortunately, the IC₅₀ values of inhibitors for EtBr, acriflavine,and pyronin Y efflux are above those employed in clinical practice.Even so, prochlorperazine is of particular interest becauseof its reduced central nervous system toxicity compared to otherphenothiazines and its current widespread clinical use as anantiemetic agent. Additionally, this drug has been shown toincrease the retention of doxorubicin in tumor cells recoveredfrom patients receiving both drugs concurrently (37, 38). However,peak prochlorperazine concentrations found in those patientswere approximately half the determined IC₅₀ for EtBr effluxmediated by NorA and the SA-K1748 and SA-K2068 efflux systems,and about 5% of the determined IC₅₀ for NorA-mediated acriflavineand pyronin Y efflux. It is possible that for bacteria maximalpump inhibition is not required to achieve a beneficial effectin vivo. Animal studies examining this issue would seem to bein order. It is also possible that chemical modification ofthe prochlorperazine molecule could result in a compound witheven less central nervous system toxicity than it currentlyhas as well as improved inhibitory activity towards bacterialefflux pumps.

The equivalence of the ${Delta}$ ${psi}$ in the absence of NorA (SA-K1712, -146± 3 mV) and in the presence of its overexpression (SA-1199B,-148 ± 2 mV) indicate that the function of this MDR pumpdoes not interfere with the methodology employed for the determinationof ${Delta}$ ${psi}$ , despite TPP being a good NorA substrate. Our ${Delta}$ ${psi}$ data suggestthat, in addition to a possible direct interaction with pumps,depolarization of the cytoplasmic membrane by phenothiazinesand thioxanthenes may also play a role in efflux pump inhibition.This activity can result in a decrease in the PMF, upon whichmost bacterial drug efflux pumps are dependent for their function.The degree of depolarization appears strain dependent, as demonstratedby the differences observed between SA-1199B and SA-K1712. ThePMF is related to both ${Delta}$ ${psi}$ and ${Delta}$ pH in the following manner: PMF(in millivolts) = ${Delta}$ ${psi}$ - 59 ${Delta}$ pH (at room temperature) (35). If ${Delta}$ pHremains constant, and our determination in the presence of prochlorperazineand reserpine suggests that it does, the PMF of SA-1199B inthe absence of prochlorperazine was -183 mV, and in its presenceit was -140 mV. This modest (23%) PMF change could reduce theefficiency of NorA and other PMF-dependent pumps. The PMF calculationcannot be done for SA-K1712 with the available data, as ${Delta}$ pH wasnot determined for that organism, but the total collapse ofits ${Delta}$ ${psi}$ in the presence of both prochlorperazine and trans(E)-flupentixolwould result in a greater decrease in the PMF than that observedfor SA-1199B.

If membrane depolarization were the main means of pump inhibition,it would be expected that the same, or at least similar, concentrationsof prochlorperazine would inhibit the PMF-driven efflux of allsubstrates to a similar degree. This was not observed for SA-1199B,in which the IC₅₀ of prochlorperazine for acriflavine and pyroninY efflux was at least 10 times that observed for EtBr efflux.In addition, the fact that inhibitors had minimal effects ongentamicin MICs, which is dependent on ${Delta}$ ${psi}$ for its transport, suggeststhat the depolarization effects, which are strain dependent,may be of secondary importance to another means of pump inhibition,such as a direct interaction of inhibitor and pump (5).

Our data indicate that reserpine does not affect pump functionby decreasing the PMF. In fact, the ${Delta}$ ${psi}$ of SA-1199B and SA-K1712was increased in its presence while ${Delta}$ pH remained constant, changesthat would result in an increase in the PMF. The explanationfor this observation is not clear but may be related, at leastin part, to a broad inhibition of many pumps for which TPP isa substrate resulting in somewhat higher intracellular concentrationsof the cation in the presence of the inhibitor and thus an increasein ${Delta}$ ${psi}$ . The facts that prochlorperazine was equipotent to reserpinein the inhibition of EtBr efflux (for all efflux-related MDRstrains) and more potent in the inhibition of acriflavine andpyronin Y efflux (for SA-1199B), combined with the observedopposite effects of these compounds on ${Delta}$ ${psi}$ , are more evidence infavor of a direct interaction with the pumps being the mainmechanism of action of these inhibitors.

Indirect effects on multiple cytoplasmic membrane-based processescould result from perturbations in ${Delta}$ ${psi}$ . The phenothiazines andthioxanthene derivatives we investigated did have small effectson the MICs of antibiotics that are not pump substrates, whichmay be due to the changes in ${Delta}$ ${psi}$ we observed. In fact, the ${Delta}$ ${psi}$ effectmay be the explanation, at least in part, for an earlier observationof a remarkable lowering of the methicillin MIC in a strainof methicillin-resistant S. aureus in the presence of subinhibitoryconcentrations of selected phenothiazines (19). It also hasbeen shown that in the presence of chlorpromazine S. aureuscells undergo ultrastructural changes including abnormal cellwalls, asymmetrical cell divisions, and the appearance of largemesosome-like structures intracellularly (16, 20). It is conceivablethat processes associated with these structural changes couldaffect susceptibility to antibiotics that act either intracellularlyor at bacterial surfaces. The test strains employed in thisstudy were all susceptible to the nonpump substrates that weinvestigated. Had we employed additional strains for which startingMICs to these compounds were higher, a greater effect may havebeen observed. Further work along these lines is warranted.

It is clear that the mechanism by which phenothiazines and thioxanthenesinhibit the function of MDR efflux pumps is multifactorial,including perturbation of membrane energetics and possibly adirect interaction with the transporters themselves. It hasbeen shown that a single amino acid change in human p-glycoproteinaffects the inhibitory activity of both cis(Z)- and trans(E)-flupentixol,suggesting that these stereoisomers directly interact with thatpump (4). Further work would be necessary to establish directevidence that this occurs with bacterial efflux pumps. It alsowould be of great interest to determine whether or not thesecompounds can inhibit the activity of other bacterial drug transporters,such as the MexAB-OprM system of P. aeruginosa, or even ATP-dependentpumps, such as the macrolide efflux pump MsrA of Staphylococcusepidermidis and S. aureus. If so, structure-activity relationshipsfor phenothiazines and thioxanthene derivatives, especiallytrans isomers of the latter compounds, would be a logical steptoward finding a potent, and ultimately minimally toxic, broad-spectrumbacterial MDR efflux pump inhibitor. The evaluation of otherdrugs currently in clinical practice, such as selective serotoninreuptake inhibitors, which have been shown to possess some antimicrobialactivity, as potential efflux pump inhibitors also seems reasonable(27, 29).

ACKNOWLEDGMENTS

This study was supported by VA Research Funds.

FOOTNOTES

* Corresponding author. Mailing address: Department of Internal Medicine, Division of Infectious Diseases, Wayne State University School of Medicine, B4333 John D. Dingell VA Medical Center, 4646 John R, Detroit, MI 48201. Phone: (313) 576-4487. Fax: (313) 576-1112. E-mail: gkaatz{at}juno.com.

REFERENCES

Top