Reversible Transport by the ATP-binding Cassette Multidrug Export Pump LmrA ATP SYNTHESIS AT THE EXPENSE OF DOWNHILL ETHIDIUM UPTAKE*

The ATP dependence of ATP-binding cassette (ABC) transporters has led to the widespread acceptance that these systems are unidirectional. Interestingly, in the presence of an inwardly directed ethidium concentration gradient in ATP-depleted cells of Lactococcus lactis , the ABC multidrug transporter LmrA mediated the reverse transport (or uptake) of ethidium with an apparent K t of 2.0 (cid:1) M . This uptake reaction was competitively inhibited by the LmrA substrate vinblastine and was significantly reduced by an E314A substitution in the membrane domain of the transporter. Similar to efflux, LmrA-mediated ethidium uptake was inhibited by the E512Q replacement in the Walker B region of the nucle-otide-binding domain of the protein, which strongly reduced its drug-stimulated ATPase activity, consistent with published observations for other ABC transporters. The notion that ethidium uptake is coupled to the catalytic cycle in LmrA was further corroborated by studies in LmrA-containing cells and proteoliposomes in which reverse transport of ethidium was associated with the net synthesis of ATP. Taken together, these data demonstrate that the conformational changes required for drug transport by LmrA are (i) not too far from equilibrium under ATP-depleted conditions to be reversed by appropriate changes in ligand concentrations and (ii) not necessarily coupled to ATP hydrolysis, but associated with a reversible catalytic cycle. These findings LmrA

The ATP dependence of ATP-binding cassette (ABC) transporters has led to the widespread acceptance that these systems are unidirectional. Interestingly, in the presence of an inwardly directed ethidium concentration gradient in ATP-depleted cells of Lactococcus lactis, the ABC multidrug transporter LmrA mediated the reverse transport (or uptake) of ethidium with an apparent K t of 2.0 M. This uptake reaction was competitively inhibited by the LmrA substrate vinblastine and was significantly reduced by an E314A substitution in the membrane domain of the transporter. Similar to efflux, LmrA-mediated ethidium uptake was inhibited by the E512Q replacement in the Walker B region of the nucleotide-binding domain of the protein, which strongly reduced its drug-stimulated ATPase activity, consistent with published observations for other ABC transporters. The notion that ethidium uptake is coupled to the catalytic cycle in LmrA was further corroborated by studies in LmrA-containing cells and proteoliposomes in which reverse transport of ethidium was associated with the net synthesis of ATP. Taken together, these data demonstrate that the conformational changes required for drug transport by LmrA are (i) not too far from equilibrium under ATP-depleted conditions to be reversed by appropriate changes in ligand concentrations and (ii) not necessarily coupled to ATP hydrolysis, but associated with a reversible catalytic cycle. These findings and their thermodynamic implications shed new light on the mechanism of energy coupling in ABC transporters and have implications for the development of new modulators that could enable reverse transportassociated drug delivery in cells through their ability to uncouple ATP binding/hydrolysis from multidrug efflux.
ATP-binding cassette (ABC) 1 transporters utilize the energy generated from ATP hydrolysis to translocate a wide variety of molecules (e.g. lipids, polysaccharides, proteins, small nutrients) across cellular membranes. These molecular pumps are found in all phyla and form one of the largest protein families (1). ABC transporters are central to many inherited disorders such as cystic fibrosis, adrenoleukodystrophy, Zellweger syndrome, and progressive familial intrahepatic cholestasis (2). The ABC superfamily also includes multidrug efflux proteins such as the multidrug resistance P-glycoprotein MDR1 (also termed ABCB1), the multidrug resistance-associated protein MRP1 (also termed ABCC1), and the breast cancer resistance protein (also termed ABCG2), the overexpression of which is a cause of resistance of human tumors to anticancer drugs (3). Homologs of human P-glycoprotein are found in prokaryotic organisms (4); and among these, LmrA in Lactococcus lactis represents a useful model for P-glycoprotein (5), as LmrA can functionally substitute for its medically important counterpart in human lung fibroblast cells (6).
ABC transporters are minimally composed of four domains: two hydrophobic integral membrane domains (MDs) and two nucleotide-binding domains (NBDs) (7). In the P-glycoprotein MDR1 gene, the four domains are encoded on a single polypeptide, whereas in lmrA, a single MD is fused to a single NBD. To comprise the four core domains, the half-transporter LmrA has been shown to function as a homodimer (8). The NBDs in Pglycoprotein and dimeric LmrA bind and hydrolyze ATP in a catalytic reaction in which the highly conserved Walker A and B motifs and the ABC signature sequence are involved (9). The MDs in these transporters contain multidrug-binding sites and provide a translocation pathway for drugs across the membrane (10).
At present, the catalytic cycle at the NBDs is believed to involve the following partial reactions: (i) ATP binding, (ii) ATP hydrolysis and the formation of a transition state complex in which both ADP and P i remain bound, (iii) release of P i , and (iv) release of ADP (11). The NBDs interact in catalysis and appear to form a single transition state complex involving liganding from catalytic side chains of both NBDs. Support for this view comes from (i) cross-linking, fluorescence energy transfer, and cryoelectron microscopy studies showing that the two NBDs in P-glycoprotein can be very close (12)(13)(14)(15)(16); (ii) mutational and photo cleavage analysis of P-glycoprotein, LmrA, and the maltose transporter in Escherichia coli demonstrating the involvement of both NBDs in ATP hydrolysis (8,17,18); and (iii) recent x-ray crystallographic data for the vitamin B 12 transporter BtuCD in E. coli (19) and MJ0796 (20) revealing nucleotide binding at the interface between the two NBDs.
During the drug transport cycle, the MDs are thought to undergo a sequence of conformational changes during which their drug-binding sites alternately face the intracellular and extracellular sides of the plasma membrane with a change in binding affinity. In LmrA, the high and low affinity drugbinding sites are found on the inner and outer membrane surfaces, respectively (8). P-glycoprotein and LmrA both exhibit a basal rate of ATP hydrolysis, which is stimulated by up to 6-fold in different preparations by drugs and other transport substrates, indicating the occurrence of long-range conformational coupling between the catalytic cycle at the NBDs and the drug transport cycle at the MDs (10). Indeed, secondary and tertiary structural changes induced by nucleotide binding and/or hydrolysis and upon the binding of drugs have been detected in LmrA and P-glycoprotein by Fourier transform infrared spectroscopy (21,22) and in P-glycoprotein by crosslinking of cysteine residues introduced into the protein (23), immunoreactivity of the protein with monoclonal antibody UIC2 (24), partial trypsin digestion (25), intrinsic tryptophan fluorescence measurements (22,26), fluorescence labeling of the ABC domains (27), and electron cryomicroscopy of twodimensional protein crystals (28), indicating the existence of different P-glycoprotein conformations associated with different steps in the catalytic and drug transport cycles. Although the detailed molecular mechanism of the conformational coupling between ATP hydrolysis and substrate transport in ABC transporters remains to be elucidated, various coupling models have been proposed (8,9,19,29,30). Aside from differences in the mechanistic details, implicit in all models is the notion of unidirectional movement of substrate: upon ATP hydrolysis, substrates are either pumped out of or into the cell depending on the physiological role of the transporter as an exporter or importer, respectively.
ATP is a "high energy" compound in that the equilibrium between ATP and ADP ϩ P i is on the side of ATP hydrolysis at the usual levels of ATP, ADP, and P i in cells that generate metabolic energy. Thus, under these conditions, a reaction toward ATP hydrolysis tends to occur at the NBDs of ABC transporters, which is coupled to the movement of substrate from one side of the membrane to the other. However, like any chemical reaction, these reactions may, in principle, be reversible. Here, we summarize evidence that (i) LmrA is able to mediate the reverse transport (or uptake) of ethidium in ATPdepleted L. lactis cells in the presence of an inwardly directed ethidium concentration gradient and that (ii) the LmrA-mediated uptake of ethidium is coupled to the catalytic cycle in the direction of ATP synthesis.

MATERIALS AND METHODS
Construction of the E512Q Mutant lmrA Gene-The lmrA gene in pNHLmrA (31) was ligated as an NcoI-SacI fragment into the E. coli vector pGEM-5Zf (Promega), yielding pGHLmrA. Site-directed mutagenesis was performed on pGMHLmrA using the QuikChange kit (Stratagene) with primers 5Ј-GCTTGATCAAGCAACAGCC-3Ј and 5Ј-GGCTGTTGCTTGATCAAGC-3Ј for E512Q. The mutant lmrA gene was then subcloned into pNZ8048 as an NcoI-XbaI fragment, yielding pNHLmrA-E512Q. The cloned PCR products were sequenced to ensure that only the intended changes were introduced.
Bacterial Strains, Plasmids, and Growth Conditions-L. lactis strains NZ9000 and NZ9700 were grown at 30°C in M17 medium (Difco) supplemented with 20 mM glucose (32). Chloramphenicol (5 g/ml) was added as required. Fresh media were inoculated with a 1:50 dilution of overnight cultures of L. lactis strain NZ9000 harboring the nisin A-inducible vectors pNHLmrA (31), pNHLmrA-E314A (33), pNHLmrA-E512Q, or empty pNZ8048, encoding N-terminally His 6tagged wild-type LmrA, mutant LmrA with a glutamate-to-alanine substitution at position 314 or a glutamate-to-glutamine substitution at position 512, or no protein, respectively. The cells were grown at 30°C to an A 660 of 0.6 -0.8. Protein expression was then induced for 1 h at the same temperature by the addition of 10 ng of nisin A/ml of culture or a 1:1000 dilution of the supernatant of the nisin A-producing L. lactis NZ9700 strain (32).
Preparation of Membrane Vesicles-The preparation of inside-out membrane vesicles of L. lactis was performed as described previously (34). For the preparation of right-side-out membrane vesicles, a 1-liter culture of nisin-induced L. lactis cells containing pNHLmrA was harvested, washed with 100 mM potassium P i (pH 7.0), and resuspended in 5 ml of the buffer with Complete protease inhibitor (Roche Applied Science, Mannheim, Germany). The cell suspension was incubated with 10 mM MgSO 4 and 10 mg/ml lysozyme at 30°C for 1 h with mild shaking. While stirring, 4.8 ml of 0.75 M K 2 SO 4 was added, and the cell suspension was incubated for 2 min at 30°C. Following the addition of 36 ml of 100 mM potassium P i (pH 7.0) containing 10 g/ml DNase and 10 g/ml RNase and a further incubation at 30°C for 20 min, 0.9 ml of 1 M K-EDTA (pH 7.0) was added. The lysed cells were then incubated for 10 min at 30°C before the addition of 0.6 ml of 1 M MgSO 4 . The suspension was centrifuged at 48,200 ϫ g for 30 min at 4°C. The resulting pellet was resuspended in 48 ml of 50 mM potassium P i (pH 7.0) and 10 mM MgSO 4 and centrifuged at 750 ϫ g for 1 h at 4°C to remove cell debris and unbroken cells. The supernatant was carefully removed and spun at 48,200 ϫ g for 30 min at 4°C. The right-side-out membrane vesicles in the pellet were resuspended in 1 ml of 50 mM potassium P i (pH 7.0) containing 10% glycerol and stored in liquid nitrogen. The right-side-out orientation of the membrane vesicles was tested through the accessibility of the intracellular His 6 tag at the amino terminus of LmrA by digestion with proteinase K. Membrane vesicles were incubated for 30 min at 20°C with proteinase K at a protein/protease ratio of 5:1. The digestion reaction was stopped by the addition of 20% trichloroacetic acid, and the samples were separated by SDS-PAGE and immunoblotted with anti-pentahistidine antibody (QIAGEN Inc.).
Ethidium Transport in ATP-depleted Cells-For ethidium uptake measurements, the cell pellet of 50 ml of culture of nisin-induced L. lactis cells was resuspended in 50 mM potassium P i (pH 7.0) containing 5 mM MgSO 4 . The cell suspension was then incubated in the presence of 0.5 mM dinitrophenol for 30 min at 30°C to deprive the cells of metabolic energy. The cells were washed four times with 50 ml of 50 mM potassium P i (pH 7.0) supplemented with 5 mM MgSO 4 , resuspended in the same buffer to a final A 660 of 5.0, and kept on ice. In each individual experiment, cells were diluted to a final A 660 of 0.5 in 2 ml of 50 mM potassium P i (pH 7.0) supplemented with 5 mM MgSO 4 in a 3-ml glass cuvette. The fluorescence was then followed at 30°C in a PerkinElmer Life Sciences LS 55B fluorimeter using excitation and emission wavelengths of 500 and 580 nm, respectively, and slit widths of 5 and 10 nm, respectively. After 60 s, ethidium bromide was added to the desired final concentration to each sample, as indicated in the figure legends; and the increase in ethidium fluorescence was followed over time. For kinetic analysis of the LmrA-mediated uptake of ethidium in ATPdepleted cells (see Figs. 2 and 3C), the uptake rate observed in control cells was subtracted from the rate measured in LmrA-expressing cells.
To assess the effect of a competing substrate on ethidium transport in whole cells, vinblastine sulfate was included in the transport buffer at the concentrations indicated in the legend to Fig. 3 after steady-state equilibrium had been reached after ϳ1 h (Fig. 3A) or 60 s prior to the addition of ethidium bromide (see Fig. 3, B and C).
Ethidium Transport in Membrane Vesicles-Inside-out membrane vesicles prepared from L. lactis cells expressing LmrA were diluted to a final protein concentration of 10 g/ml in 2 ml of 50 mM potassium P i (pH 7.0) supplemented with 5 mM MgSO 4 . Basal fluorescence was followed for 60 s before the addition of ethidium bromide to a final concentration to 0.5 M. The settings on the fluorimeter were as indicated under "Ethidium Transport in ATP-depleted Cells." ATPase Assay-The ATPase activity was based on a colorimetric ascorbic acid/ammonium molybdate assay to measure the liberation of P i from ATP as described (35). Inside-out membrane vesicles were used at a final protein concentration of 1 mg/ml in 20 mM K-HEPES (pH 7.4) containing 5 mM MgSO 4 and 5 mM NaATP. The assay mixtures were prepared on ice and contained ethidium at the concentrations indicated in Fig. 5A. After incubation for 10 min at 30°C, the reactions were stopped by the addition of 0.2 mM ascorbic acid solution, after which the P i concentration in each sample was determined at A 690 using P i standard solutions for calibration. The ATPase activity measurements in the presence of 1 mM orthovanadate were obtained in parallel and subtracted from the readings.
Reconstitution of Purified LmrA in Proteoliposomes-The solubilization of His 6 -tagged LmrA with n-dodecyl ␤-D-maltoside, the purification of the solubilized protein on nickel-nitrilotriacetic acid resin, and the reconstitution of the purified protein were performed as described previously (34), except that Triton X-100 was used (up to 3 mM) to destabilize the liposomes instead of n-dodecyl ␤-D-maltoside to obtain the unidirectional reconstitution of the protein in an inside-out orientation (33). The proteoliposomes were resuspended to 10 mg of membrane protein/ml in 50 mM potassium P i (pH 8.0).
Determination of ATP Synthesis-His 6 -LmrA-containing proteoliposomes (20 g of protein) were pre-equilibrated for 60 min in the presence of 20 M ethidium bromide and subsequently diluted 100-fold in a buffered mixture composed of (i) 100 l of 50 mM potassium P i (pH 8.0) containing 5 mM MgSO 4 and 5 mM ADP and (ii) 100 l of luciferase reagent (ATP bioluminescence assay kit HSII; Roche Applied Science). L. lactis cells were incubated in 50 mM potassium P i (pH 8.0) containing 5 mM MgSO 4 and 0.5 mM dinitrophenol and washed with the same buffer without dinitrophenol as described under "Ethidium Transport in ATPdepleted Cells." One-ml aliquots of the cell suspension (A 660 ϭ 0.7) were then incubated with 20 M ethidium for 5 min at 30°C and lysed using the lysis solution supplied with the ATP bioluminescence assay kit. The amount of ATP present in the cell lysates or formed in proteoliposome suspensions over time was measured by bioluminescence in accordance with the manufacturer's instructions. ATP standard solutions were used to convert arbitrary bioluminescence units into molarity of ATP. Bioluminescence was measured using a luminometer sample housing with 1-s integration time. The photon multiplier tube (modal no. 9899A) was supplied by Electron Tubes. A cooled housing (FACT50) maintained temperatures of the photon multiplier at Ϫ13°C, and photon counts were processed by a CT2 counter time module (RS 232).
Data Analysis-All statistical analyses were performed using Student's paired t test with a 95% confidence interval for the sample mean. Where indicated, n refers to the number of independent observations with different cell batches and (proteo)liposome preparations.

RESULTS
LmrA-mediated Ethidium Uptake-Ethidium transport in L. lactis cells can be measured by monitoring the fluorescence of the intracellular ethidium-polynucleotide complex. In previous work, we demonstrated that, in the presence of glucose, cells expressing LmrA exhibit an enhanced rate of ethidium efflux compared with non-expressing control cells, which is due to ATP-dependent extrusion of ethidium by LmrA (5,31). Surprisingly, the starvation of L. lactis cells in the presence of the protonophore dinitrophenol with a reduction of the intracellular ATP concentration from 9.0 Ϯ 0.2 mM to 7.1 Ϯ 0.9 M (n ϭ 3) allowed cells expressing LmrA to take up ethidium at an enhanced rate compared with non-expressing control cells when an inwardly directed ethidium concentration gradient was imposed (Fig. 1A). The cells and the recombinant LmrA protein were shown to be functional after the de-energization procedure, as the addition of glucose again elicited the characteristic ATP-dependent efflux of ethidium from LmrA-expressing cells and the passive membrane potential (interior negative)-driven ethidium uptake in the control (Figs. 1A and 5C) (5). These data point to a role of LmrA in drug influx in ATP-depleted cells. Also, when an outwardly directed ethidium concentration gradient was imposed through a 100-fold dilution of ethidium-loaded cells in buffer without ethidium, ATPdepleted cells expressing LmrA exhibited an enhanced rate of ethidium efflux compared with non-expressing control cells (Fig. 1B). Hence, the downhill transport of ethidium was enhanced in both transport directions in ATP-depleted cells expressing LmrA compared with non-expressing control cells.
To confirm that the enhanced ethidium accumulation in LmrA-expressing cells (Fig. 1A) was due to LmrA-mediated uptake of ethidium and binding to chromosomal DNA rather than binding to LmrA directly, the ethidium fluorescence was measured in LmrA-containing right-side-out membrane vesicles lacking intravesicular DNA. The orientation of the membrane vesicles was verified by testing the accessibility of the intracellular amino-terminal His 6 tag of LmrA to digestion by proteinase K. The protease was unable to cleave the amino terminus of LmrA from the external side of the membrane vesicles, allowing the detection of the His 6 tag by immunoblotting (data not shown). In contrast, in control experiments with inside-out membrane vesicles, an immunosignal was not obtained, as proteinase K was able to cleave the amino terminus of LmrA from the external side. However, the addition of ethidium to right-side-out membrane vesicles containing LmrA did not elicit the increase in ethidium fluorescence as observed in ATP-depleted cells expressing LmrA (data not shown). This result confirmed that the increased ethidium fluorescence in cells ( Fig. 1A) was not due to the binding of ethidium to LmrA, but to the intercalation of ethidium in cellular DNA.
Kinetic analysis of ethidium uptake in LmrA-expressing cells revealed (i) an LmrA-mediated uptake reaction, the rate of which increased as a function of the ethidium concentration in a saturable fashion with an apparent K t of 2.0 Ϯ 0.3 M ethidium and an apparent V max of 0.23 Ϯ 0.02 arbitrary units (a.u.)/min (n ϭ 3) (Fig. 2), and, superimposed, (ii) a passive uptake reaction which, similar to ethidium uptake in the control cells, increased linearly with the ethidium concentration and exhibited no saturation in the micromolar concentration range (Fig. 2, inset). Only the passive ethidium uptake reaction was observed in non-expressing control cells (Fig. 2, inset). Interestingly, similar to the observations on LmrA-mediated uptake of ethidium, the rate of downhill LmrA-mediated ethidium efflux increased as a function of the ethidium concentration in a saturable fashion with an apparent K t of 2.4 Ϯ 0.2 M ethidium and an apparent V max of 0.08 Ϯ 0.02 a.u./min (n ϭ 3) (Fig. 2). Hence, the kinetic parameters for LmrA-mediated ethidium efflux in ATP-depleted cells were in a similar range as those obtained for LmrA-mediated ethidium uptake.
Vinblastine Competes with Ethidium for Uptake by LmrA-Vinblastine is a chemotherapeutic agent and a substrate of LmrA and the human P-glycoprotein MDR1 (6) and competes with ethidium for ATP-dependent efflux by LmrA in L. lactis (Fig. 3A). The effect of vinblastine on the ethidium uptake in ATP-depleted lactococcal cells expressing LmrA was compared with its effect on ethidium uptake in non-expressing control cells. As shown in Fig. 3B, the passive diffusion of 0.5 M ethidium into ATP-depleted control cells lacking LmrA was not affected by the presence of 5 M vinblastine. However, under similar conditions, the rate of ethidium uptake in LmrAexpressing cells was reduced by ϳ30% in the presence of vinblastine (Fig. 3B). When vinblastine was added to LmrAexpressing or control cells after ethidium uptake had reached a steady state, no effect was observed on the fluorescence level of ethidium (data not shown). These results confirmed that the reduced accumulation of ethidium in LmrA-expressing cells was not based on the displacement of ethidium from chromosomal DNA by the non-fluorescent vinblastine. Instead, the data pointed to a direct inhibition of the LmrA-mediated ethidium uptake by vinblastine.
The mechanism of this inhibition was further examined. The initial rate of ethidium influx in ATP-depleted LmrA-expressing cells was measured at ethidium concentrations ranging from 0.5 to 5 M and compared with the rate in control cells in the absence or presence of 5 or 10 M vinblastine (Fig. 3C). Linear regression analysis of the kinetic data in a Lineweaver-Burk plot yielded R 2 values of Ͼ0.95 and showed that vinblastine competitively inhibited LmrA-mediated ethidium influx with a K i of 2.7 Ϯ 0.5 M vinblastine (n ϭ 4). Hence, as in ATP-dependent efflux, vinblastine and ethidium also competed for uptake by LmrA into cells.
Mutation in the MD of LmrA Reduces the Rate of Ethidium Uptake-To further demonstrate the ability of LmrA to mediate ethidium influx, the effect of an E314A substitution in the MD of LmrA on the rate of ethidium uptake in ATP-depleted L. lactis cells was studied. Recently, we observed that E314A mutant LmrA exhibits a reduced rate of ATP-dependent ethidium efflux (33). Even though the expression levels of LmrA and E314A mutant LmrA were equivalent in L. lactis (data not shown), the rate of ethidium influx in ATP-depleted cells expressing E314A mutant LmrA was significantly lower and comparable with the ethidium influx rate observed in non-expressing control cells (Fig. 4). This result strongly supports the notion that LmrA mediates ethidium uptake in ATPdepleted lactococcal cells. BmrA. Mutations at these positions have been reported to reduce ATP hydrolysis by the mentioned ABC transporters due to the inhibition of the post-hydrolytic release of ADP (36 -38) and/or deletion of the catalytic base for ATP hydrolysis (39). Likewise, E512Q mutant LmrA exhibited significantly reduced ethidiumstimulated vanadate-sensitive ATPase activity (Fig. 5A) and ethidium efflux activity (Fig. 5B), but retained the ability to interact with the photoaffinity compound [ 3 H]azidopine (data not shown). Surprisingly, the ATP-depleted cells expressing E512Q mutant LmrA also exhibited a reduced rate of ethidium uptake compared with cells expressing the wild-type protein (Fig. 5C), even though the expression levels of the wild-type and mutant LmrA proteins were identical (data not shown). These data pointed to a coupling between LmrA-mediated ethidium uptake and the catalytic cycle at the NBD of the protein.

LmrA ATPase Mutant Exhibits a Reduced Rate of Ethidium
Ethidium Uptake by LmrA Is Associated with ATP Synthesis-To further study the coupling between the drug transport and catalytic cycles in LmrA during ethidium uptake, the ATP levels in ATP-depleted L. lactis cells were measured in a luciferase-driven assay after a 5-min incubation of the cells in the absence or presence of 20 M ethidium in the extracellular buffer (Fig. 6). Surprisingly, a significant level of ATP synthesis was detected in LmrA-expressing cells in the presence of ethidium, but not in its absence. No net ATP synthesis was observed in non-expressing control cells under these conditions. The ethidium-dependent LmrA-associated synthesis of ATP could be blocked by preincubation of cells for 5 min in buffer containing the LmrA inhibitor nicardipine at 5 M (6) before the addition of ethidium. An intermediate level of ATP synthesis was observed in E512Q mutant LmrA-expressing cells in the presence of ethidium (Fig. 6), consistent with the intermediate rate of ethidium uptake in these cells compared with LmrA-expressing cells or control cells (Fig. 5C). Hence, the data in intact cells provided evidence that the uptake of ethidium by LmrA was associated with the net synthesis of ATP.
The coupling between LmrA-mediated ethidium uptake and ATP synthesis was also studied in proteoliposomes containing purified, functionally reconstituted, inside-out oriented LmrA (Fig. 7) (33). The proteoliposomes were preloaded with ethidium and diluted in ethidium-free buffer containing a luciferasebased detection system for ATP. The imposition of a 100-fold outwardly directed ethidium concentration gradient in the proteoliposomes was associated with an ATP bioluminescence rate of 7138 Ϯ 310 a.u./s (n ϭ 4), corresponding to the synthesis of ATP up to a concentration of 10 nM in the proteoliposome suspensions during the time course of the experiment (600 s).
Consistent with the reduced ATPase activity of E512Q mutant LmrA (Fig. 5A), the ATP bioluminescence rate increased only to 2203 Ϯ 120 a.u./s (n ϭ 4) in proteoliposomes containing purified E512Q mutant LmrA, corresponding to the synthesis of ATP up to a concentration of 10 pM during the incubation period. In contrast, an ATP bioluminescence rate of 18 Ϯ 10 a.u./s (n ϭ 4) was observed for empty liposomes, indicating that the ATP concentration remained below 1 fM in the control  (n ϭ 4). B, efflux of ethidium by E512Q mutant LmrA. ATP-depleted L. lactis control cells or those expressing LmrA or E512Q mutant LmrA were pre-equilibrated for 1 h with 5 M ethidium. Subsequently, ethidium fluorescence was followed over time before and after the addition of 20 mM glucose (at t ϭ 120 s). C, influx of ethidium into ATP-depleted control cells or cells expressing LmrA or E512Q mutant LmrA. At the first arrow (t ϭ 120 s), 2 M ethidium was added. At the second arrow (t ϭ 1200 s), 20 mM glucose was added to enable the generation of metabolic energy in the cells. cells (Fig. 7). Consistent with the observations on ATP synthesis in intact cells (Fig. 6), the data in (proteo)liposomes provided strong support for the direct role of LmrA in ATP synthesis at the expense of downhill ethidium uptake. DISCUSSION One of the characteristics of all ABC transporters, which distinguishes these proteins mechanistically from secondary active transporters, is their ATP dependence. This property of ABC transporters has led to the generally accepted idea that their transport cycles must operate in a unidirectional fashion, in the direction of ATP hydrolysis. However, because enzymemediated reactions are thermodynamically reversible, the unidirectional transport by ABC transporters could reflect a kinetic irreversibility under physiological conditions rather than a mechanical irreversibility. For the P-type Na ϩ /K ϩ -ATPase in the plasma membrane of mammalian cells, which catalyzes the exchange of 3Na ϩ (efflux) and 2K ϩ (uptake) during an ATPdependent transport cycle, it has been shown that, if the resources of ATP are low in the cell in comparison with the concentrations of ADP and P i , the reaction of ATP hydrolysis is not favorable. If, under these conditions, inverted Na ϩ and K ϩ concentration gradients (inwardly and outwardly directed, respectively) are imposed across the plasma membrane, the direction of the Na ϩ /K ϩ -ATPase transport cycle can be reversed, resulting in the backward movement of Na ϩ and K ϩ with the synthesis of ATP (40,41).
Four observations suggested that multidrug transport by LmrA could be forced to run in a reversed reaction upon the imposition of an inwardly directed ethidium concentration gradient in ATP-depleted L. lactis cells. First, ATP-depleted cells expressing LmrA accumulated ethidium at a faster rate compared with control cells (Fig. 1A). The rate of ethidium uptake saturated at increasing ethidium concentrations with an apparent K t of ϳ2.0 M ethidium (Fig. 2). In control experiments, the ATP-depleted cells could be re-energized in the presence of glucose, enabling active LmrA-mediated efflux of ethidium (Figs. 1A and 5C). Second, the LmrA-mediated uptake of ethidium was competitively inhibited by the LmrA substrate vinblastine with a K t of ϳ2.7 M vinblastine (Fig. 3). Third, ethidium uptake was reduced in cells expressing E314A mutant LmrA, containing an amino acid substitution in the membrane domain of the transporter (Fig. 4); we have recently shown that this mutant LmrA exhibits a reduced efflux rate for cationic drugs compared with wild-type LmrA (33). Finally, the rate of ethidium uptake was reduced in cells expressing mutant LmrA containing the E512Q substitution in the NBD (Fig. 5C). Analogous amino acid substitutions at equivalent positions in various ABC transporters inhibited the operation of the catalytic cycle of these proteins (36 -39). Similarly, E512Q mutant LmrA exhibited a reduced ethidium-stimulated ATPase activity (Fig. 5A) and rate of ethidium efflux (Fig. 5B). The observations on the reduced uptake of ethidium by E512Q mutant LmrA pointed to a coupling between LmrA-mediated ethidium uptake and the catalytic cycle of the protein. Direct evidence for this coupling was obtained from observations on ethidium transport-associated synthesis of ATP in ATP-depleted cells and in proteoliposomes containing purified LmrA (see below).
The chemical events at the NBDs in ABC transporters are thought to be stoichiometrically linked to substrate translocation. For LmrA in intact cells, the reaction can be written as shown in Reaction 1, ATP ϩ nEt in 7 ADP ϩ P i ϩ nEt out REACTION 1 in which n refers to the number of ethidium (Et) molecules transported per molecule of ATP hydrolyzed. For this transport reaction, Equation 3 (see "Appendix A") describes the thermodynamic relation between the electrochemical potential of ethidium ions and the Gibbs free energy change for ATP hydrolysis (⌬G ATP3 ADPϩP i ). The kinetic properties of LmrA-mediated ethidium uptake and efflux reactions were analyzed in ATP-depleted cells (Fig. 2). Using the Haldane equation (Equation 4, see "Appendix B"), the kinetic parameters can be related to the thermodynamic equilibrium constant (K eq ) of the LmrAmediated transport reaction, giving a K eq of 0.42, which is close to unity. Evaluation of this result in the context of Equation 3 (see "Appendix A") suggests that the ATP hydrolysis reaction is close to its equilibrium (⌬G ATP3 ADPϩP i Ϸ 0) in ATP-depleted cells as a result of the incubation with the protonophore dinitrophenol. Under these conditions at steady state, the chemical gradient of ethidium will be in equilibrium with the transmembrane potential; and hence, the LmrA-mediated fluxes of ethidium in the direction of uptake and efflux across the membrane will be approximately equal.
Because K eq for the LmrA-mediated ethidium transport is close to unity under ATP-depleted conditions, the amount of free energy required to displace the equilibrium will be relatively small. Consistent with this notion, evidence was obtained for the coupling of LmrA-mediated ethidium uptake in intact cells with the net synthesis of ATP (Fig. 6). The lack of ATP synthesis under these conditions in LmrA-expressing cells in the presence of the LmrA inhibitor nicardipine or in control cells without LmrA provided further support for the direct role of LmrA in the ATP synthesis reaction. The LmrA-mediated synthesis of ATP was also observed during ethidium efflux in proteoliposomes containing purified and functionally reconstituted LmrA in an inside-out orientation (Fig. 7). No ATP synthesis was observed in control liposomes under identical conditions, whereas a reduced rate of ATP synthesis was obtained for E512Q mutant LmrA, consistent with the reduced rates of ethidium uptake and efflux and the reduced drug-stimulated ATPase activity obtained for this mutant in experiments in cells and membrane vesicles (Fig. 5). The increase in the ATP concentration up to 10 nM in suspensions of LmrA-containing proteoliposomes during the first 600 s after the imposition of the outwardly directed electrochemical ethidium gradient of Ϫ120 mV is in a similar range as the transient increase of 5 or 52 nM ATP that can be calculated using Equation 5 (see "Appendix C") assuming that two or one molecule of ATP, respectively, is synthesized per molecule of ethidium transported. The results obtained for LmrA-containing proteoliposomes demonstrate that the LmrA-mediated ATP synthase reaction during ethidium uptake is independent of auxiliary proteins, such as other ATPases present in the plasma membrane of L. lactis.
With the K eq for the LmrA-mediated ethidium transport reaction of close to unity under ATP-depleted conditions, a relatively small change in the free energy would also be required to displace the equilibrium in the direction of ethidium efflux. ATP hydrolysis, with a standard free energy ⌬G 0 of about Ϫ32.6 kJ/mol, could increase the K eq by a factor of Ͼ10 6 in cells generating metabolic energy (Equation 7, see "Appendix D"). Interestingly, ATP binding to the NBD, with a reported dissociation constant K d(MgATP) between 0.7 and 3 mM (11, 42,43) and with a calculated ⌬G between Ϫ14 and Ϫ18 kJ/mol, could increase K eq up to 3000-fold (see "Appendix D"). This notion may relate to published observations on P-glycoprotein in membrane vesicles suggesting that ATP binding, rather than ATP hydrolysis, induces the reorientation of the drugbinding site from the inside surface to the outside surface of the protein, with a concomitant reduction in drug-binding affinity (44). It may also relate to previous observations on MRP1, the cystic fibrosis transmembrane conductance regulator, and sulfonylurea receptor-1 that ATP binding to nonequivalent NBDs may operate as a molecular switch that triggers conformational changes in these ABC proteins (38,45). For ATP binding, the complementary fit between the ligand and its binding site would provide the thermodynamic basis for the catalytic function of the NBD, a principle that is relevant for many enzymecatalyzed reactions. Hence, our calculations predict that K eq could be significantly displaced in the direction of ethidium efflux by both ATP hydrolysis and binding.
The relevance of each of these potential driving forces for LmrA-associated drug efflux will be dependent on the timing of ATP binding and hydrolysis in the transport cycle of the transporter. Overall, the drug transport cycle of ABC multidrug transporters can be depicted in a general scheme, with two conformations of the transporter with drug-binding site(s) exposed to the cytoplasm (E1) or the extracellular space (E2) (10). During drug efflux, the inward facing E1 conformation is converted to the outward facing E2 conformation at a specific stage of the ATP hydrolytic cycle (Step 1). Dissociation of the trans-ported substrate and/or another stage of the ATP hydrolytic cycle converts the E2 conformation back to the original E1 conformation (Step 2). If ATP binding would drive drug translocation in Step 1, then ATP hydrolysis in Step 2 would be a post-translocation event that would impose directionality on the transport cycle and reset the transporter in its original state. This model would be consistent with the "concerted" transport mechanism, recently proposed for P-glycoprotein (46) and the bacterial maltose transporter MalFGK 2 (47), in which the release of substrate from the transporter precedes the formation of the ADP⅐P i -bound transition state. Alternatively, the order of these reactions could be reversed, with ATP binding in Step 2 to reset the transporter and ATP hydrolysis in Step 1 to drive drug translocation. Finally, in the more classical interpretation, Steps 1 and 2 could both be driven by ATP hydrolysis.
Altogether, our data demonstrate that (i) the LmrA-mediated drug transport reaction is not too far from equilibrium under ATP-depleted conditions to be displaced in the forward or reversed reaction by appropriate changes in ligand concentrations; and hence, (ii) the direction of transport by LmrA is not restricted by a mechanical irreversibility. In addition, the conformational changes required for drug transport by LmrA (e.g. the reorientation of drug-binding sites) do not require ATP hydrolysis per se and can even be coupled to ATP synthesis during reverse transport. Much remains to be learned about the molecular mechanisms by which ABC transporters operate. A better understanding of these mechanisms may lead to the development of effective modulators of ABC transporters in the clinical setting. In this context, the intrinsic reversibility of LmrA may represent a more general property of ABC multidrug transporters. If so, similar to previous conclusions (48), our findings imply the potential development of modulators that could enable reverse transport-associated drug delivery in cells through their ability to uncouple ATP binding/hydrolysis from multidrug efflux.
Addendum-Recent studies on a truncated LmrA protein lacking the NBD (LmrA-MD) have shown that the MD of the transporter functions as a secondary active multidrug uptake system without the requirement of ATP (33). The observations on drug efflux and uptake by full-length LmrA presented in this work show that the reversibility of transport by LmrA-MD is conserved in LmrA. APPENDIX A: Energy Coupling to Transport-For Reaction 1 in intact cells, we can write the equilibrium condition as shown in Equation 1, in which ⌬G ATP3 ADPϩP i is the equilibrium constant for ATP hydrolysis, {Et ϩ } is the electrochemical activity of ethidium ions, and n is the number of ethidium ions transported per molecule of ATP hydrolyzed. In various transport models proposed in the literature, n is equal to 0.5 or 1 (8,9,19,29,30 in which ⌬ Et ϩ is the electrochemical potential of ethidium, R is the gas constant, T is the temperature (in kelvin), ⌬G 0 is the standard free energy change in ATP hydrolysis, and ⌬G ATP3 ADPϩP i is the free energy change of ATP hydrolysis under physiological conditions. Due to the depletion of ATP in the cells in the presence of dinitrophenol, RT ln([ADP][P i ]/ [ATP]) will approximate Ϫ⌬G 0 , yielding a ⌬G ATP3 ADPϩP i close to zero.

B: Haldane Equation-
The relationship between an enzyme's kinetic constants and the thermodynamic equilibrium constant K eq of the reaction it catalyzes is given by Equation 4, Eq. 4) in which K eq refers to the overall equilibrium constant for the consecutive reactions involved in the translocation of ethidium; V max and K t refer to the maximal transport rate and the ethidium concentration that gives 1/2V max , respectively; and the superscripts e and u refer to ethidium efflux and uptake, respectively. Using the kinetic parameters determined for LmrA-mediated efflux and uptake, K eq ϭ (0.08 a.u./min ϫ 2. in which F is the Faraday constant. With (i) the imposition of an outwardly directed electrochemical ethidium gradient of Ϫ120 mV in proteoliposomes containing inside-out oriented LmrA, which is equivalent to an inwardly directed ethidium gradient in intact cells; (ii) ADP and P i concentrations of 5 and 50 mM, respectively; (iii) a ⌬G 0 for ATP hydrolysis of about Ϫ32.6 kJ/mol; and (iv) the transport of one ethidium ion/molecule of ATP hydrolyzed (n ϭ 1), yielding a transient increase in the ATP concentration of 52 nM. For n ϭ 0.5, the ATP concentration would increase to 5 nM. D: Displacement of Equilibrium by ATP Binding/Hydrolysis-The free energy change ⌬G of the transport reaction in ATP-depleted cells is a function of the displacement of the reaction from the equilibrium, which is described by Equation 6, ⌬G ϭ ϪRT ln (qK eq ) ( E q . 7 ) which can be rewritten as Equation 7, q ϭ e Ϫ(⌬G/RT) K eq (Eq. 8) in which ⌬G refers to the free energy change in the transport reaction, K eq refers to the overall equilibrium constant, and q refers to the -fold displacement of the reaction from the equilibrium. When transport would be coupled to ATP hydrolysis with a ⌬G 0 of about Ϫ32.6 kJ/mol, q ϭ e Ϫ(Ϫ32,600/(8.314ϫ298)) /0.42 ϭ 1.2 ϫ 10 6 . For ATP binding to the NBDs of P-glycoprotein with an experimentally determined association constant K A(MgATP) of 1.25 ϫ 10 3 M Ϫ1 (11), the ⌬G of the binding reaction equals ϪRT ln K A ϭ Ϫ8.314 ϫ 298 ϫ ln 12,500 ϭ Ϫ23.4 kJ/mol. When coupled to LmrA-mediated efflux, K eq could be displaced by a factor of q ϭ e Ϫ(Ϫ23,400/(8.314ϫ298)) /0. 42  q would still be equal to e Ϫ(Ϫ14,100/(8.314ϫ298)) /0.42 ϭ 705.