The DrrAB Efflux System of Streptomyces peucetius Is a Multidrug Transporter of Broad Substrate Specificity*

Background: DrrAB is dedicated to export of doxorubicin in Streptomyces peucetius, an organism that produces this anticancer drug. Whether this prototype system can export other drugs has not been investigated. Results: DrrAB exports multiple drugs efficiently. Conclusion: Substrate specificity of DrrAB overlaps with known bacterial and human multidrug resistance proteins. Significance: This study suggests common mechanisms and origin for DrrAB and other MDR proteins. The soil bacterium Streptomyces peucetius produces two widely used anticancer antibiotics, doxorubicin and daunorubicin. Present within the biosynthesis gene cluster in S. peucetius is the drrAB operon, which codes for a dedicated ABC (ATP binding cassette)-type transporter for the export of these two closely related antibiotics. Because of its dedicated nature, the DrrAB system is believed to belong to the category of single-drug transporters. However, whether it also contains specificity for other known substrates of multidrug transporters has never been tested. In this study we demonstrate under both in vivo and in vitro conditions that the DrrAB system can transport not only doxorubicin but is also able to export two most commonly studied MDR substrates, Hoechst 33342 and ethidium bromide. Moreover, we demonstrate that many other substrates (including verapamil, vinblastine, and rifampicin) of the well studied multidrug transporters inhibit DrrAB-mediated Dox transport with high efficiency, indicating that they are also substrates of the DrrAB pump. Kinetic studies show that inhibition of doxorubicin transport by Hoechst 33342 and rifampicin occurs by a competitive mechanism, whereas verapamil inhibits transport by a non-competitive mechanism, thus suggesting the possibility of more than one drug binding site in the DrrAB system. This is the first in-depth study of a drug resistance system from a producer organism, and it shows that a dedicated efflux system like DrrAB contains specificity for multiple drugs. The significance of these findings in evolution of poly-specificity in drug resistance systems is discussed.

Multidrug resistance (MDR) 2 has emerged as a major clinical problem in recent years both for the treatment of infectious diseases and for chemotherapy of cancer. Although many different mechanisms for drug resistance are known, a common strategy consists of active efflux of drugs from the cells (1). Drug transporters are categorized into either single-drug efflux systems (which are specific for a drug or a group of drugs) or multidrug efflux systems that exhibit a broad specificity and can transport structurally and functionally unrelated compounds. These proteins function as either primary active (belonging to the ATP binding cassette superfamily) or secondary active transporters (2). The phenomenon of multidrug resistance was first characterized in mammalian cancer cells, where exposure to anticancer drugs was seen to result in overexpression of ABC-type efflux pumps, such as P-glycoprotein (Pgp) and MRP1 (2). These proteins have since been shown to transport hundreds of structurally unrelated compounds, including amphipathic anti-cancer drugs, peptides, and fluorescent dyes, etc., thus conferring MDR in cancer cells. MDR is also widespread among bacteria; the best known ABC family members include the bacterial proteins LmrA and LmrCD in Lactococcus lactis, Sav1866 in Staphylococcus aureus, and MsbA in Escherichia coli.
Although most of the ABC proteins mentioned above have served as useful models to characterize and understand the basis of MDR (2,3), the most extensive biochemical analysis of the nature of multidrug specificity has been carried out with P-glycoprotein (4,5). Together the analyses carried out by many different groups suggested that Pgp contains a large drug binding chamber that can accommodate several drugs simultaneously. It was also suggested that the drug binding chamber in Pgp is lined by several transmembrane (TM) helices, including TMs 4 -6 in TMD1 and TMs 9 -12 in TMD2 (4). Most recently, the crystal structure of Pgp confirmed many findings of the biochemical analysis and showed that Pgp indeed contains a large and flexible drug binding cavity made of mostly hydrophobic and aromatic residues (6). Different drugs were seen to interact with residues in different parts of the flexible cavity, mostly through hydrophobic interactions, thus providing an explanation for the poly-specific nature of Pgp. The crystal structure of Pgp also revealed that some drugs were bound to a single site, whereas some others bound to two different locations within the cavity. The drug binding cavity of Pgp was found to reside within the cell membrane, and it showed the presence of two portals formed by TMs 4 and 6 and TMs 10 and 12, which allow direct entry of hydrophobic molecules from the membrane (6). This observation supports previous models, which proposed that Pgp can extract drugs directly from the lipid bilayer and remove them by a hydrophobic vacuum cleaner mechanism (7). The crystal structure of another MDR protein, AcrB, a secondary-active transporter, is also available. This structure also showed the presence of a very large central drug binding cavity, which could accommodate several ligand molecules simultaneously (8). However, the drug binding cavity, although present in the membrane, was found to be accessible to the periplasmic domain. Thus, in contrast to Pgp (where the drug portals are located in the membrane), the periplasmic region of AcrB seems to play a major role in determining substrate specificity, suggesting differences between the mechanisms of different MDR proteins.
The bacterial ABC drug transporters that have also been studied in significant detail include the E. coli ABC transporter MsbA (responsible for export of Lipid A, the core moiety of LPS), L. lactis homologue LmrA, and S. aureus Sav1866 (3,9,10). All of these proteins have been shown to transport multiple drugs. Interestingly, MsbA was found to contain overlapping substrate specificity with LmrA and Sav1866 (11). Some newly identified members, such as VcaM from Non-O1 Vibrio cholerae and the YccC (BmrA) from Bacillus subtilis were also shown to transport multiple drugs (12,13). Therefore, significant progress has indeed been made in understanding the phenomenon of multidrug resistance. However, the available information is based on the analysis of only a handful of drug transporters described above. Most other annotated drug transporters, especially those found in the antibiotic/drug producer organisms, have not been analyzed for their ability to confer MDR. Moreover, not much is known about why and how the ability to confer multidrug resistance evolves.
In this study we analyzed the ABC transporter DrrAB, which confers self-resistance to two related anticancer antibiotics doxorubicin (Dox) and daunorubicin in the producer soil organism Streptomyces peucetius. The genes for this system are present in an operon located within the gene cluster for biosynthesis of Dox and daunorubicin; therefore, they code for a dedicated transporter for these antibiotics. The DrrAB system represents the simplest form of an ABC drug transporter, which is assembled from two molecules each of DrrA (the catalytic subunit) and DrrB (integral membrane subunit) (14). In the mammalian Pgp, the two catalytic and two integral membrane domains are naturally fused into a single large polypeptide, possibly the result of an evolutionary gene fusion event (15). Both DrrAB and Pgp confer resistance to the anticancer agents Dox and daunorubicin; DrrAB in the producer organism and Pgp in cancer cells. Therefore, the overall structure and function of the DrrAB transporter bears significant similarity to Pgp even though these two transporters belong to different classes of ABC proteins (16). Subcloning of the drrAB locus in E. coli was previously shown to confer doxorubicin resistance in this host (17). It was also previously shown that the DrrAB system confers resistance to Dox by an energy-dependent efflux mechanism (18). However, it is not known if this system is specific for Dox and daunorubicin, or if, like Pgp, it can also recognize and transport multiple drugs. Because this is a prototype drug resistance mechanism found in the producer organism, analysis of this system could shed light on the nature of substrate specificity and elucidate how the ability to confer multidrug resistance evolves in proteins.
In this paper we provide in-depth characterization of drug transport by the DrrAB system and show that, contrary to the generally held assumption, this system forms a multidrug transporter. Using both E. coli whole cells and inside-out membrane vesicles (IOVs), it is shown that the DrrAB system can efficiently transport not only Dox but also Hoechst 33342 (H 33342) and ethidium bromide (EtBr), two substrates most commonly used to establish the MDR phenotype (3,10,19). We also found that the DrrAB-mediated Dox efflux is inhibited by a number of other well characterized MDR substrates, such as verapamil, rifampicin, vinblastine, and colchicine, suggesting that these drugs are also substrates of the DrrAB pump. Interestingly, DrrAB-mediated efflux could be coupled to the energy of either ATP or GTP hydrolysis, and, as expected, the function of this transporter was found to be completely independent of the proton motive force (pmf). Because multiple drugs were found to inhibit Dox efflux by the DrrAB system, kinetics analysis was carried out to understand the mechanism of inhibition and interaction of drugs with DrrAB. Our studies revealed that inhibition of Dox efflux by H 33342 and rifampicin occurs by a competitive mechanism, whereas verapamil inhibits Dox transport by a non-competitive mechanism, suggesting that the DrrAB transporter may contain at least two drug binding sites. The findings of this paper demonstrate for the first time that the dedicated Dox transport system, DrrAB, can recognize and transport multiple drugs. This study highlights overlaps between the substrate specificity of the DrrAB system and Pgp and points to a common mechanism and perhaps origin for most MDR proteins.
In Vivo Dox Efflux in Cells-E. coli LE392⌬uncIC cells (Table  1) containing either vector pSU2718 or pDX101(pSU2718/ drrAB) were grown in 200 ml of triethanolamine medium (50 mM triethanolamine HCl, pH 6.9, 15 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 1 mM MgSO 4 ) supplemented with 0.5% (w/v) glycerol, 2.5 g/ml thiamine, 0.5% (w/v) peptone, and 0.15% (w/v) succinate to mid-log phase and induced with 0.1 mM IPTG for 1 h. The harvested cells were washed twice and resuspended in 100 l of triethanolamine buffer. 10 l of the cell suspension from above was incubated in 3 ml of triethanolamine medium containing 10 M doxorubicin and 5 mM 2,4-dinitrophenol for 11 h at 37°C. The loaded cells were washed twice with 0.1 mM MOPS buffer, pH 7.0, and resuspended in 3 ml of MOPS buffer containing 2 mM MgSO 4 . The fluorescence spectra were recorded on an Alphascan-2 spectrofluorometer (Photon Technology Int., London, Ontario, Canada). The excitation wavelength for doxorubicin was 480 nm, and emission was monitored at 590 nm. The excitation and emission slit widths were both set at 1.00 mm, and a time-based script was run. After an initial recording of fluorescence for 100 s at 37°C, energy was provided in the form of either 20 mM glucose or 20 mM succinate, and recording was continued for an additional 400 s. The rate of Dox efflux was determined from the slope in the steady-state range (300 -500 s). Where indicated, 10 mM sodium fluoride was added as inhibitor of ATP synthesis.
In Vivo Ethidium Bromide Efflux in Cells-E. coli LE392⌬uncIC cells containing the indicated plasmid were grown, induced, and loaded with various concentrations (1 M to 100 M) of EtBr as described above except that the loading time used was 1 h at 37°C. The loaded cells were washed twice with 0.1 mM MOPS, pH 7.0, and resuspended in 3 ml of MOPS buffer containing 2 mM MgSO 4 . EtBr efflux from loaded cells was measured fluorometrically (excitation, 500 nm; emission, 580 nm) on an Alphascan-2 spectrofluorometer (Photon Technology Int.). After 100 s, energy was provided in the form of 20 mM glucose. The recording was continued for additional 400 s. The rate of EtBr efflux was determined from the slope in the steady-state range (300 -500 s).
Preparation of IOVs-E. coli LE392⌬uncIC cells containing the indicated plasmids were grown in 1 liter of LB medium at 37°C until mid-log phase and induced with 0.25 mM IPTG at 37°C for 3 h. The cells pellet was resuspended in 20 ml of 1ϫ PBS buffer, pH 7.4, and lysed with French press at 16,000 p.s.i. twice. The membrane fraction was prepared according to the previously published protocol (14), except that the membrane vesicles were washed twice with 20 ml of 1ϫ PBS buffer.
In Vitro Dox Efflux in IOVs and Kinetic Analysis-250 g of IOVs were resuspended in 3 ml of 1ϫ PBS buffer, pH 7.4, supplemented with 0.1 mg/ml creatine kinase and 5 mM creatine phosphate. Dox was added to a final concentration of 1.0 M or as indicated. The fluorescence spectra were recorded on an Alphascan-2 spectrofluorometer with an excitation wavelength of 480 nm and emission wavelength of 590 nm. The excitation and emission slit widths were set to 1.00 mm, and data were collected at 0.1-s intervals. After 100 s, the detection was paused, 1 mM Mg 2ϩ and 1 mM ATP, pH 7.5, were added into the reaction, and the detection was continued for additional 400 s. Where indicated, ATP was substituted with 1 mM GTP or 5 mM NADH. The rate of Dox transport was determined from the In Vitro Hoechst 33342 Efflux in IOVs-The DrrAB-mediated efflux of H 33342 was studied in IOVs, as described above for Dox Efflux with some modifications. Briefly, 250 g of IOVs were resuspended in 3 ml of 1ϫ PBS buffer, pH 7.4, supplemented with 0.1 mg/ml creatine kinase and 5 mM creatine phosphate and various concentrations of H 33342 (0.1-2.5 M).
The excitation and emission wavelengths of H 33342 were 355 and 457 nm, respectively. The rate of H 33342 transport was determined from the initial slope of the linear range between100 s and 200 s.
Vanadate Inhibition of Dox Efflux in IOVs-250 g of IOVs were resuspended in 3 ml of 1ϫ PBS buffer, pH 7.4, supplemented with 0.1 mg/ml creatine kinase, 5 mM creatine phosphate, 1 M Dox, and various concentrations of sodium o-vanadate (0 -100 M). The measurement of Dox efflux was performed as described above. The rate of Dox transport was determined from the slope of the linear range between 100 and 200 s.
Determination of IC 50 -250 g of IOVs were resuspended in 3 ml of 1ϫ PBS buffer, pH 7.4, supplemented with 0.1 mg/ml creatine kinase and 5 mM creatine phosphate, 1 M Dox, and various concentrations of the inhibitory drug. The measurement of Dox efflux was performed as described above. The rate of Dox transport was determined from the slope of the initial linear range between 100 and 200 s. Designating the efflux rate of the sample without inhibitor as 1.0, the relative rate of each sample was calculated. The average data of three independent experiments were plotted by "scatter plot with simple error bars" in Sigma Plot 11.0 software and fitted by the dynamic curve fit (y ϭ ae Ϫbx , x is the concentration of inhibitor; and y is the relative rate). The IC 50 value was determined based on the concentration of the drug that brings ϳ50% inhibition of the DrrAB-mediated Dox efflux at a Dox concentration of 1 M.  concentrations of H 33342 (0, 0.2, 0.6, or 0.8 M) were studied. Similar assays were set up to study kinetics of Dox inhibition by different drugs. Initial rate of Dox transport was determined as described above. The rate of Dox transport obtained with 1 M Dox and 0 M H 33342 (or another drug) was designated as 1.0. The relative rates were then calculated for each efflux curve, and the data were plotted by Lineweaver-Burk plot using Sig-maPlot kinetics software in single substrate/single inhibitor kinetics format. The error bars represent three independent experiments. The type of inhibition was determined based on the AICc value (Akaikes Information Criterion corrected); lower AICc values correspond to better fits to the data. Point Mutations in DrrA-Site-directed mutagenesis of the drrA gene was carried out by a QuikChange multisite-directed mutagenesis kit (Stratagene, La Jolla, CA). Using pDX101 (pSU2718/drrAB) plasmid as the template, Gln 197 , located in the Switch motif of DrrA, was changed to histidine. The resulting plasmid was named pDX101(Q197H). Another plasmid, generated by substituting Tyr 198 to arginine, was named pDX101(Y198R). Double mutations of Gln 197 -Tyr 198 to His 197 -Arg 198 and His 197 -His 198 were also created, and these plasmids were named pDX101(Q197H/Y198R; HR) and pDX101(Q197H/Y198H; HH). Other mutations used in this study were described previously (Table 1).

Kinetics of Dox Efflux Inhibition by Known MDR Substrates-
ATPase Activity in IOVs-7.5 g of IOVs expressing either wild type DrrAB or DrrAB with mutations in the Switch motif were incubated in 1 ml of reaction mixture containing 50 mM MOPS buffer, pH 7.5, 1 mM dithiothreitol, 10 l PK/LDH enzyme (Sigma), 5 mM ATP, 0.25 mM NADH, and 1.25 mM phosphor(enol)pyruvic acid at 37°C for 10 min as described previously (20). The reaction was started by the addition of 2.5 mM MgCl 2 . The optical density at 340 nm was monitored for 10 min using the Shimadzu UV1601 spectrophotometer and the UV probe 2.20 Kinetics software. The slope of the linear portion of each curve (between 200 and 400 s) was used to calculate ATPase activity. The activity of the control IOVs (without DrrAB) was subtracted from the activity of each test sample to obtain DrrAB-specific activity. Relative activity was then calculated by dividing the activity of each sample by the activity of the wild type sample.

Characterization of the DrrAB-mediated Dox Efflux under in Vivo and in Vitro Conditions-An in vivo assay for studying
DrrAB-mediated Dox efflux was reported previously (18,21) and is shown in Fig. 1A. In this study, we established conditions to study DrrAB-mediated Dox efflux under in vitro conditions using E. coli IOVs (Fig. 1B). Both in vivo and in vitro assays utilize the fluorescent nature of Dox to measure efflux in E. coli LE392⌬uncIC cells (or IOVs). This strain of E. coli contains a deletion in the unc genes; as a result it is unable to carry out synthesis of ATP using proton gradients or establish a proton gradient by hydrolysis of ATP (22). Therefore, it is possible to establish conditions where only the pmf or ATP is available as a source of energy (described below). Dox is fluorescent in solution; however, its accumulation inside the cells results in quenching of its fluorescence, and its efflux results in an increase in fluorescence intensity (23).
In Vivo Dox Efflux-The basic strategy for studying efflux under in vivo conditions consists of loading of the de-energized cells with Dox (18) followed by the addition of an energy source, which is expected to result in efflux of Dox and an increase in its fluorescence (Fig. 1A). Under the conditions used in our experimental system, use of succinate as energy will generate only proton motive force, whereas glucose will generate both pmf and ATP, therefore allowing us to discriminate between the energy sources used by the DrrAB system. The data in Fig. 1A compare the rate of Dox efflux in DrrAB-containing cells in the presence of glucose or succinate. E. coli LE392⌬uncIC cells (containing empty vector) were used as a negative control in these experiments. Two conclusions can be made from the data shown in Fig. 1A. First, the rate of Dox efflux by the DrrABcontaining cells in the presence of glucose is about 5-fold higher as compared with the rate in control cells (Fig. 1A.2, compare  columns 1 and 4). A small increase in Dox efflux efficiency seen in control cells on the addition of glucose is likely due to the action of one of the several MDR pumps known to be present in E. coli (1). Secondly, in contrast to glucose, use of succinate as energy showed no increase in fluorescence intensity in DrrABcontaining cells ( . These studies show that doxorubicin efflux by the DrrAB pump is solely ATPdependent, and pmf is not required for this process, thus highlighting similarities between DrrAB and other MDR proteins of the ABC superfamily (19,24,25).
In Vitro Dox Efflux-To understand the kinetics of Dox transport, an in vitro Dox efflux assay was optimized using inside-out membrane vesicles, as described under Experimental Procedures." In this assay the vesicles were mixed with Dox, and efflux was initiated by the addition of ATP and Mg 2ϩ . Because of the inverted nature of the IOVs, DrrAB-mediated efflux results in accumulation of Dox inside the vesicles, which is seen as quenching of its fluorescence (Fig. 1B.1). The data in Fig. 1B.1 show that the addition of ATP/Mg 2ϩ first results in a rapid nonspecific decrease in Dox fluorescence due to interaction between Dox and ATP (23). This quick phase is then followed by a slower rate of quenching, which corresponds to the DrrAB-dependent efflux in the vesicles. The initial rate of Dox efflux was, therefore, determined from the linear slope of the fluorescence spectra between 100 and 200 s. As seen in Fig. 1, B.1 and B.2, Dox transport activity in vesicles containing DrrAB was found to be almost 10 -12-fold higher as compared with the control vesicles prepared from cells containing vector alone ( Fig. 1B.2, compare columns 1 and 4). The absence of either Mg 2ϩ (column 2) or ATP (column 3) resulted in the failure of these vesicles to transport Dox. Use of NADH as an energy source also did not support Dox efflux (column 6), once again confirming that proton motive force is not used by the DrrAB proteins as a source of energy. Surprisingly, when GTP was  (21). Dox efflux was measured fluorometrically (excitation, 480 nm; emission, 590 nm) on an Alphascan-2 spectrofluorometer (Photon Technology Int.). Energy was provided in the form of 20 mM glucose or 20 mM succinate at 100 s (marked with an arrow), and fluorescence was monitored for an additional 400 s. 10 mM sodium fluoride (NaF) was added to the samples where indicated. A.1, curve 1, pDx101(DrrAB)/glucose; curve 2, pDx101 (DrrAB)/succinate; curve 3, pDx101 (DrrAB)/glucose/NaF; curve 4, pSU2718 (vector)/glucose; curve 5, pSU2718/succinate; curve 6, pSU2718/glucose/NaF. A.2, quantitative presentation of the Dox efflux data shown in A.1. The slope of the linear portion of each curve shown in A.1 was calculated. The slope of curve 1 was designated as 1.0. Relative slope of each curve was then obtained by dividing the slope of the curve by the slope of curve 1. The average data obtained from three independent experiments are shown in the histogram. B, in vitro analysis of DrrAB-mediated Dox efflux using IOVs. E. coli LE392⌬uncIC cells containing either vector or pDX101 (DrrAB) were grown to mid-log phase and induced with 0.25 mM IPTG at 37°C for 3 h. The membrane fraction was prepared (14), and Dox efflux in DrrAB-containing IOVs was performed, as described under "Experimental Procedures." B.1, in vitro Dox efflux assay was carried out using 250 g of IOVs in the presence of 1 M Dox, 0.1 mg/ml creatine kinase, and 5 mM creatine phosphate in 3 ml of PBS buffer, pH 7. used as an energy source instead of ATP, even higher transport activity was observed (column 5). Dox-dependent ATP and GTP binding to DrrA was shown previously (17). Together these data indicate that either ATP or GTP (and Mg 2ϩ ) can serve as a source of energy for the Dox transport function of DrrAB.
Kinetic Analysis of Dox Efflux in Vitro-To understand the kinetics of Dox efflux by the DrrAB system, efflux was analyzed (as shown in Fig. 1B.1) at a wide range of Dox concentrations. The initial rate (slope between 100 and 200 s) of each efflux curve was determined. Efficiency of Dox efflux was then calculated, as described under "Experimental Procedures" and in the figure legends. The data were fitted to a hyperbola using the Michaelis-Menten equation (with an R 2 of 0.89), yielding an apparent K m of 0.39 M and V max of 1000 (arbitrary units (a.u.)) (Fig. 1C). These data showed a linear increase in the rate of Dox transport at concentrations ranging between 0.1 and 1 M, which became saturated after 3 M. The Dox transport data could also be fitted equally well by the Hill equation; the implication of this finding is discussed later.
Inhibition of Dox Transport Activity by Sodium o-Vanadate-Vanadate (Vi) is a known inhibitor of the ATPase activity of ABC proteins (26 -29). Because it functions as an analog of P i , the ADP⅐Vi⅐Mg 2ϩ complex is trapped in the ATP binding pocket after a single catalytic turnover, thus blocking further hydrolysis of ATP as well as drug transport. To determine if the DrrAB system is inhibited by vanadate, Dox transport was measured in IOVs in the absence or presence of increasing amounts of vanadate (5-100 M). The data in Fig. 1D show that vanadate is a potent inhibitor of the Dox transport of DrrAB, with an IC 50 ϭ 11 M. A complete inhibition of the Dox transport activity was seen at 100 M vanadate. These data are consistent with the previously reported studies on the inhibitory effect of vanadate on Pgp and MsbA (28,30).
Point Mutations in the Nucleotide Binding Domain (NBD) of DrrA Compromise Dox Transport Activity-The N-terminal nucleotide binding domain of DrrA contains a 200-amino acidlong ABC cassette consisting of all the conserved motifs (Walker A, Walker B, signature motif/C-loop, Q-loop, and the Switch motif/H-loop) involved in ATP binding and hydrolysis ( Fig. 2A.1). We previously reported that mutations in the conserved residues of DrrA confer Dox sensitivity (31). Mutations in Walker A, as expected, also compromised ATP binding (31). Here we evaluated the effect of several mutations on DrrABmediated Dox efflux in IOVs. The data in Fig. 2B show that, as expected, single point mutations in Walker A (such as G44A, G44S, or K47R), signature (S141R), or the Walker B (E165Q) motif of DrrA result in a drastic effect on Dox transport activity. Because the sequence of the Switch motif of DrrA is different from most other ABC proteins ( Fig. 2A), this region was analyzed in greater detail as described below. Most ABC proteins normally contain a highly conserved histidine residue in their Switch motif followed by an arginine, histidine, or a lysine ( Fig.  2A.1). The conserved histidine of the Switch motif and a conserved glutamate immediately following the Walker B motif (both are shown as highlighted areas in Fig. 2) are together believed to be critical for formation of the active sites in ABC proteins. Zaitseva et al. (32) recently proposed that these two residues together form a catalytic dyad that functions in substrate-assisted catalysis. However, despite the high conservation of these two residues in ABC proteins, deviations in the sequence of these motifs are sometimes seen. Most commonly, a glutamine replaces the histidine in the Switch motif, and an aspartate replaces the glutamate in the Walker B region. When present (for example, in TAP1 and LmrC), these deviations are seen to result in asymmetrical ATP binding pockets with one site being catalytically non-functional (33,34). Interestingly, however, despite the presence of the non-canonical glutamine residue (Gln 197 followed by Tyr 198 , resulting in the QY sequence) in the Switch motif of DrrA (and its close prokaryotic homologs, Fig. 2A.2), it is able to form a functional drug transporter with DrrB, as seen in Fig. 1. Note that the close eukaryotic homologs of DrrA most often contain an HH sequence in the Switch region ( Fig. 2A.3), and both prokaryotic and eukaryotic homologs contain the conserved glutamate in the Walker B region (specifically E165 in DrrA) (Fig. 2, A.2 and A.3).
Because histidine is conserved in the Switch of most ABC proteins, we wondered if the DrrAB transporter will become more efficient if Gln 197 is substituted with a histidine or if the QY sequence is changed to the commonly occurring sequence HR or HH. Surprisingly, we found that the Q197H mutation in DrrA produces a drastic effect on Dox efflux; however mutation of Tyr 198 to Y198R retains about 35% Dox efflux function (Fig.  2C). Interestingly, a double mutation HR (Q197H/Y198R) also resulted in Dox transport activity of ϳ38%, indicating that the second mutation partially masked the harmful effect of the Q197H single mutation (Fig. 2C). By contrast, the double mutation HH (Q197H/Y198H) exhibited extremely low Dox efflux (Fig. 2C). To understand the role of Gln 197 and Tyr 198 in catalysis, the effect of the above-mentioned mutations on the ATPase activity of the DrrAB complex was determined. We found that although Q197H produces a drastic effect on ATPase activity, both HR and HH double mutations show significant ATPase activity (43 and 61%, respectively) (Fig. 2, D and E), indicating that the second mutation in each case partially compensates for the negative effect of the Q197H mutation on catalysis. Overall the findings in Fig. 2 indicate that residues Gln 197 and Tyr 198 in the Switch motif of DrrA function together and that the QY sequence works much better than the HH or HR sequence for the overall Dox efflux function of the DrrAB complex. Interestingly, the HH double mutant still exhibits 60% ATP hydrolysis activity, which suggests that either the QY or the HH sequence could participate in the formation of functional catalytic sites in DrrA. However, the HH allele seems to be defective in specific communication between DrrA and DrrB, resulting in significantly reduced Dox efflux (Ͻ5%). Therefore, the context in which the Switch motif functions in different ABC proteins may determine the nature of this motif. The conserved glutamate Glu 165 present near the Walker B region in DrrA served as a control in these experiments. Analysis of the E165Q mutation showed a drastic effect on both hydrolysis of ATP and Dox efflux (Fig. 2, B-D), which is consistent with the critical role of this residue previously reported in literature (35). How a glutamine residue participates in the formation of functional ATP binding pockets in DrrA is still an open question. However, we can draw from the analogous sit-uations present in other ATP/GTP-binding proteins (including RecA and H-ras p21) that contain a catalytic glutamine residue (Gln 194 in RecA and Gln 61 in H-ras p21) in their switch II domain in the position corresponding to Gln 197 in DrrA (36 -38). The three-dimensional structures of RecA and H-ras indicate that these glutamines show similar interactions with the catalytic glutamate and ␥-phosphate of ATP to those seen with histidine in the ABC proteins, therefore suggesting that the DrrAB Is a Multidrug Transporter MAY 2, 2014 • VOLUME 289 • NUMBER 18 glutamine functions in a manner similar to histidine in producing the overall conformation of the active site. This is consistent with our observation that either QY or HH sequence can indeed participate in formation of the catalytic sites in DrrA, although the HH sequence is deficient in energy transduction. In summary, the analyses in Fig. 2 confirm that a functional nucleotide binding domain of DrrA is essential for Dox efflux by the DrrAB system. The analysis of the Switch mutations, in particular, also indicates that the in vitro efflux assay described here can provide a sensitive and a valuable approach for elucidating the role of critical residues in the catalytic and drug transport function of the DrrAB complex.

DrrAB-mediated Dox Efflux Is Inhibited by Multiple MDR
Substrates-To determine if the DrrAB transporter can recognize and bind other known MDR substrates, inhibition of Dox efflux by different drugs was investigated in IOVs. These assays were carried out at a wide range of the inhibitor concentrations while maintaining a constant concentration of Dox, as described under "Experimental Procedures" (determination of IC 50 values). The data in Fig. 3 show that many known MDR drugs, including H 33342 (Fig. 3A), verapamil (Fig. 3B), and rifampicin (Fig. 3C), inhibit DrrAB-mediated Dox efflux with high efficiency. The data in Fig. 3, panel D, summarize the inhibitory effects of many different MDR substrates and indi-  (16). B, effect of point mutations in Walker A, Walker B, or the Signature motif of DrrA on DrrAB-mediated Dox efflux in IOVs. E. coli LE392⌬uncIC cells containing either pSU2718 vector, pDX101 (drrAB in pSU2718), pDX102 (drrA only, in pSU2718), pDX103 (drrB only, in pSU2718), or pDX101 containing mutations of Walker A (G44A, G44S, K47R), Signature motif (S141R), or the Walker B region (E165Q) were induced with IPTG, and the IOVs were prepared as described under "Experimental Procedures." The initial rate of Dox efflux was determined for each sample, and the relative slopes were calculated as described for Fig. 1. The average data obtained from three independent experiments are shown in the histogram. C, effect of point mutations in the Switch motif of DrrA on DrrAB-mediated Dox efflux in IOVs. E. coli LE392⌬uncIC cells containing either pSU2718 vector, pDX101 (drrAB in pSU2718), or pDX101 with mutations of the Walker B region (E165Q) or the Switch motif (Q197H, Y198R, Q197H/Y198H, Q197H/Y198R) were induced with IPTG, and the IOVs were prepared as described under "Experimental Procedures." The initial rate of Dox efflux was determined for each sample, and the relative slopes were calculated as described for Fig. 1.  cate that the IC 50 values vary dramatically for different substrates. For example, H 33342, vinblastine, verapamil, and rifampicin show relatively low IC 50 values, whereas much higher concentrations of EtBr, quinidine, and colchicine were required for the same level of inhibition. These data suggest that the DrrAB system has multiple substrates that bind with varying affinities. However, whether these drugs bind to the same site or if there are multiple drug binding sites cannot be ascertained from these data. To understand the nature of drug binding, kinetics of DrrAB-mediated Dox efflux was studied at multiple Dox concentrations in the presence of several concentrations of each inhibitory drug, which is described below.
Kinetics of Dox Inhibition by MDR Drugs-The data in Fig.  1C showed that the rate of Dox efflux by DrrAB-containing IOVs increases linearly between 0.1 and 1 M Dox concentration; therefore, four evenly distributed concentrations of Dox within this range (0.25, 0.50, 0.75, and 1.0 M) were used in the kinetics analysis. At each concentration of Dox, efflux was measured in the presence of four different concentrations of H 33342, verapamil, or rifampicin. These drugs were chosen based on their low IC 50 values seen in inhibition studies (Fig.  3D) (because Dox and EtBr together were incompatible in the fluorescence-based assays due to their overlapping excitation/ emission spectra, this combination could not be used in this study). The initial slope (100 -200 s) for each Dox efflux curve was determined. The slope of the Dox efflux curve at 1.0 M Dox and 0 concentration of the inhibitor was designated as 1, which was then used to calculate the relative slope of each curve obtained in the presence of the inhibitor, as described under Experimental Procedures." The relative slopes were plotted by the Lineweaver-Burk plots using the Sigma Plot kinetics software (Fig. 4). The data in Fig. 4A indicate that the inhibition of Dox efflux by H 33342 is characteristic of competitive inhibition; the K m of DrrAB-mediated Dox transport increased at increasing concentrations of H 33342, whereas the V max remained unchanged. These data suggest that Dox and H 33342 may bind to the same site in DrrAB. The apparent inhibition constant (K i for H 33342) was found to be 0.6 M, which corresponds well to the IC 50 value of H 33342 (Fig. 3, panel D). When studying the inhibition of Dox transport by rifampicin, a similar pattern was observed that indicated a competitive inhibition between Dox and rifampicin (Fig. 4C). However, the kinetics of inhibition of Dox transport by verapamil showed a different pattern (Fig. 4B). The V max decreased at increasing concentrations of verapamil, whereas K m remained unaltered, indicating a non-competitive inhibition between Dox and verapamil. These data suggest that verapamil may bind to both the unliganded DrrAB and the binary DrrAB-Dox complex. The K i value for verapamil inhibition was found to be 9.4 M, which also matches well to the IC 50 of verapamil (Fig. 3, panel D). The observed competitive and non-competitive inhibitions of DrrAB-mediated Dox transport in IOVs imply that DrrAB must contain at least two drug binding sites.
The DrrAB System Is a Multidrug Transporter-To determine whether the DrrAB system can actually transport other drugs in addition to Dox, two substrates (H 33342 and EtBr) commonly used to establish the MDR phenotype (10,19) were tested in IOVs as described under "Experimental Procedures." In addition, five other fluorescent dyes or drugs (including Hoechst 34580, Hoechst 33258, quinine, tetramethylrhodamine, and rhodamine 123) were also tested. Of these substrates, H 33342 and EtBr were successfully transported by the DrrAB system, as described below.
H 33342 is a cell-permeable dye that is fluorescent only when bound to the cell membrane and loses its fluorescence in aqueous environment. Thus a decrease in fluorescence is expected when H 33342 is transported by the DrrAB system from the membrane to the inner aqueous environment in the IOVs. EtBr, on the other hand, becomes fluorescent when bound to DNA inside the cells. Therefore, DrrAB-mediated efflux of EtBr was studied in preloaded cells, as described under "Experimental Procedures." EtBr efflux under this condition is expected to result in quenching of fluorescence. The data in Fig. 5 show that both H 33342 (Fig. 5A.1) and EtBr (Fig. 5B.1) are transported efficiently by the DrrAB system. In the case of H 33342, the rate of transport was seen to decrease after an initial linear phase of transport. This is likely due to depletion of H 33342 in the membrane and due to passive rebinding of H 33342 to the membrane from the aqueous phase, as shown earlier in the case of Pgp (19). To determine the kinetic parameters for the transport of H 33342 and EtBr, efflux of each substrate was analyzed at a wide range of concentrations. The initial rate of H 33342 transport in DrrAB-containing IOVs was measured between 0.1 and 2.5 M (Fig. 5A.2). The efficiency of H 33342 transport was calculated as described earlier for Dox (Fig. 1C). The data were plotted using the Michaelis-Menten equation by the Sigma Plot kinetics software and could be fitted to a hyperbola with an R 2 of 0.90, yielding an apparent K m of 0.78 M and V max of 16000 (arbitrary units). To study the kinetics of EtBr transport, energy-depleted cells were loaded with different concentrations of EtBr (ranging from 1 to 100 M) for 1 h, and efflux was initiated by the addition of 20 mM glucose (Fig. 5B.2). The data fitted to the hyperbola (Michaelis-Menten equation) with an R 2 of 0.97, yielding an apparent K m of 21 M and V max of 550 (arbitrary units). The data in Fig. 5 show that both H 33342 and EtBr are transported efficiently by the DrrAB system. A comparison of the K m and V max values for transport of Dox, H 33342, and EtBr is shown in Fig. 5, panel C. Five other fluorescent substrates (Hoechst 34580, Hoechst 33258, quinine, tetramethylrhodamine, and rhodamine 123) tested in this study were not found to be exported by the DrrAB system (data not shown). In general, the transport data agreed with the inhibition data shown in Fig. 3, panel D. Interestingly, of the three Hoechst dyes tested, only H 33342 was transported. Recent studies showed that a certain balance between the hydrophobic and hydrophilic forces in the structure of the target substrate is essential for optimum efflux by Pgp (39). Because H 33342 is the most amphiphilic of the three dyes, this may explain why DrrAB is able to preferentially transport H 33342 but not the other two dyes.

DISCUSSION
Multidrug resistance is conferred by the action of specialized proteins in the cell membrane that have the ability to carry out energy-dependent efflux of structurally and functionally unrelated hydrophobic compounds. This phenomenon defies the concept of single enzyme-single substrate (the lock and key model), which is applicable to most other enzymes. Since the discovery of multidrug resistance in human cancer cells and bacteria (40 -42), two important questions have remained at the forefront of this field; 1) what is the basis of poly-specific drug recognition and transport by MDR proteins? 2) how and why does MDR evolve? Extensive biochemical analysis of Pgp (1,4,5) as well as the availability of three-dimensional structures of the drug-bound forms of Pgp, AcrB, BmrR, and QacR (6,8,43,44) has contributed significantly to understanding the molecular basis of poly-specific drug binding. These studies indicate that the MDR proteins contain an extensive (and flexible) drug binding pocket that can accommodate multiple drugs simultaneously via interactions with specific (sometimes overlapping) aromatic and polar residues present in different parts of the pocket. The three-dimensional structure of Pgp in particular also provides support for the hydrophobic vacuum cleaner model, which was proposed early on by Gottesman and co-workers (7,15) to explain the mechanism of multidrug transport. According to this model, Pgp substrates (which are primarily cationic, lipid soluble, planar molecules) partition spontaneously into the membrane and are picked up by Pgp directly from the membrane (instead of from the cytoplasm) and then removed by its ability to flip drugs from the inner to the outer leaflet. This model, which has now been extended to many other MDR proteins (45), makes a clear distinction between the classical solute transporters that remove molecules from an aqueous compartment and the MDR proteins that are proposed to function as flippases (46). A clear relationship between the ability of drugs to partition into the membrane and their binding and transport by Pgp has since been documented (47)(48)(49). Additional evidence in support of this model comes from 1) the ability of most MDR proteins studied to date to extrude the lipophilic dye H 33342 directly from the membrane (3,19,50,51), thus making it an ideal molecule to establish the MDR phenotype and 2) the ability of many MDR proteins to flip-flop lipids (52)(53)(54)(55). Together, these observations led to the idea that hydrophobic drugs and lipids may use a common pathway or flipping mechanism for extrusion.
Despite significant progress in understanding the basis of poly-specific drug binding, the origin of multidrug resistance still remains an enigma. It is generally accepted that most of the antibiotic and drug resistance genes found in clinical settings have their origins in the environmental bacteria (1). Indeed many antibiotic producing organisms contain dedicated export systems to protect themselves against their own antibiotics (56,57). Understanding these systems could hold the key to unraveling the origin and evolution of multidrug specificity. It is gen-erally assumed that the dedicated systems found in producer organisms belong to the category of single-drug transporters as opposed to multidrug transporters (2); however, their substrate specificity has never been investigated in any significant detail, thus providing the impetus for this study.
We show for the first time that the DrrAB system functions as a typical multidrug transporter that can carry out efflux of structurally unrelated substrates, including Dox, EtBr, and H 33342. Inhibition studies further demonstrate that the substrate range of the DrrAB system includes many additional sub- strates, such as verapamil, vinblastine, and rifampicin, among others. Thus the substrate specificity of DrrAB overlaps with most other known MDR proteins, including Pgp (58 -60). We found, however, that unlike Pgp, neither quinine nor rhodamine 123 serves as a substrate for export by the DrrAB system nor did they show inhibition of Dox transport. Therefore, despite the broad substrate range of DrrAB, it appears that the range of substrates recognized by Pgp may still be larger (61), implying that the ability to bind some of these substrates may have evolved later in Pgp. Interestingly, the IC 50 values varied significantly among different drugs, indicating the different binding affinities of each substrate. The highest IC 50 was observed for colchicine, which is consistent with the data previously reported for Pgp (47,62,63). This may be related with low hydrophobicity of colchicines (64), therefore, indicating that the ability to partition into the membrane may be an important factor for transport by both Pgp and DrrAB.
Kinetic characterization of Dox, H 33342, and EtBr transport by the DrrAB system revealed single-site transport kinetics. Furthermore, the K m values of 0.39 and 0.78 M for transport of Dox and H 33342, respectively, reflect high affinity of these substrates for the DrrAB system and correspond well with the K m values seen earlier for Dox transport by Pgp (65) and H 33342 transport by MsbA and LmrA (10). Interestingly, however, we found that the kinetic data for Dox as well as H 33342 could also be fitted by the Hill equation (shown in Fig. 5C) providing the n Hill value of 1.7 for Dox (R 2 ϭ 0.93) and 2.0 for H 33342 (R 2 ϭ 0.95), which suggests co-operative interaction between two or more drug binding sites in DrrAB. It is not clear from these data, however, whether both H 33342 and Dox bind to the same sites in the DrrAB system. Because EtBr transport by DrrAB exhibited only single-site transport kinetics (n Hill ϭ 1.1; R 2 ϭ 0.98) (Fig. 5C), it possibly interacts with only one of the Dox/Hoechst binding sites, or it may bind to a completely different site in DrrAB. The kinetic analyses shown in Fig. 4 provide evidence for competitive inhibition of Dox transport by H 33342 and rifampicin and non-competitive inhibition by verapamil. Therefore, we conclude that both H 33342 and rifampicin bind to the same site(s) as Dox, whereas verapamil may bind to a different site in DrrAB. Evidence for two or more nonidentical drug binding sites has also been obtained previously with MsbA (10, 53), LmrA (25), and Pgp (10,66,67). In Pgp, the two sites were defined as the H and R sites based on their preferential binding to H 33342 and rhodamine, respectively (68). Together, these data point toward the presence of two or more drug binding sites in the DrrAB system. Gleaning from the knowledge previously gained from studies with Pgp and AcrB, these different drug binding sites are likely part of the same binding pocket in DrrB. However, in the absence of a crystal structure of the DrrAB transporter, this conclusion is only tentative.
Even though the present study does not provide an answer to the origin of poly-specific drug recognition, it does indeed show that a simple system like DrrAB (which may be closely related to the ancestral system) already contains the ability to recognize and transport multiple substrates. This leads us to conjecture that the ability to transport multiple drugs is an inherent property of a certain class of proteins that followed a different evo-lutionary path than the classical transporters. This work also opens new questions regarding the mechanism by which the DrrAB system extrudes multiple hydrophobic substrates. The most important of these questions is whether this system relies on a hydrophobic vacuum cleaner-like mechanism and if it is capable of flippase action? Doxorubicin is produced inside S. peucetius cells; therefore, presumably the function of the DrrAB system is to remove it from the cytoplasmic compartment to the outside. Being amphipathic, however, Dox could conceivably partition into the cell membrane and then be picked up by the DrrB protein directly from the membrane, as suggested in an early model proposed for AcrB function (8). This strategy for its removal could also provide protection to the cell from the toxic effects of free Dox inside the cell. The ability to transport H 33342 no doubt indicates that the DrrB protein can indeed extract this fluorescent dye directly from the membrane by a mechanism similar to the one used by Pgp and other proteins (10,19,60). Future studies will, therefore, focus on an in-depth analysis of the mechanism of drug extrusion by the DrrAB system.
This and previous studies on MDR proteins raise many other intriguing questions; for example, is there evolutionary relatedness between MDR proteins and lipid flippases? What is the real difference between the single-drug and multidrug transporters? What makes certain proteins specific for a substrate and some others multi-specific? What is the contribution of high aromatic amino acid content of the TM helices of MDR proteins in conferring poly-specificity? Because the transmembrane helices of Pgp and other MDR proteins, such as the bacterial MDR protein, BMR, are highly enriched in aromatic amino acids (18 and 15.4% for Pgp and bacterial MDR protein, respectively) as compared with the single-drug transporter TetA (9.4%), Pawagi et al. (69) proposed that a greater number of aromatic amino acids may correlate with a decrease in substrate specificity of a transporter by providing additional binding sites for drugs containing aromatic rings in their structure. Moreover, it was shown that mutagenesis of a single residue Ser 941 to Phe in TM11 of Pgp significantly altered its drug efflux profile (70). The recent three-dimensional structure confirmed that the drug binding pocket of Pgp is made up of mostly hydrophobic and aromatic residues (6). Interestingly, we found that the aromatic amino acid content of predicted TM helices of DrrB (71) is also relatively high (15%), which is close to the aromatic content of Pgp (18%) and the bacterial MDR protein (15.4%). Whether this plays a role in conferring multi-specificity in DrrB will be determined in future studies by mutagenesis of specific aromatic and other residues in DrrB.