Drug-stimulated Nucleotide Trapping in the Human Multidrug Transporter MDR1

The human multidrug transporter (MDR1 or P-glycoprotein) is an ATP-dependent cellular drug extrusion pump, and its function involves a drug-stimulated, vanadate-inhibited ATPase activity. In the presence of vanadate and MgATP, a nucleotide (ADP) is trapped in MDR1, which alters the drug binding properties of the protein. Here, we demonstrate that the rate of vanadate-dependent nucleotide trapping by MDR1 is significantly stimulated by the transported drug substrates in a concentration-dependent manner closely resembling the drug stimulation of MDR1-ATPase. Non-MDR1 substrates do not modulate, whereas N-ethylmaleimide, a covalent inhibitor of the ATPase activity, eliminates vanadate-dependent nucleotide trapping. A deletion in MDR1 (Δ amino acids 78–97), which alters the substrate stimulation of its ATPase activity, similarly alters the drug dependence of nucleotide trapping. MDR1 variants with mutations of key lysine residues to methionines in the N-terminal or C-terminal nucleotide binding domains (K433M, K1076M, and K433M/K1076M), which bind but do not hydrolyze ATP, do not show nucleotide trapping either with or without the transported drug substrates. These data indicate that vanadate-dependent nucleotide trapping reflects a drug-stimulated partial reaction of ATP hydrolysis by MDR1, which involves the cooperation of the two nucleotide binding domains. The analysis of this drug-dependent partial reaction may significantly help to characterize the substrate recognition and the ATP-dependent transport mechanism of the MDR1 pump protein.

Overexpression of the human multidrug transporter (MDR1 1 or P-glycoprotein) is responsible for the phenomenon of multiple drug resistance in various cancer cell types. MDR1 is an integral plasma membrane protein that acts as an ATP-dependent efflux pump to reduce the intracellular concentration of diverse hydrophobic compounds (reviewed in Refs. [1][2][3]. MDR1 belongs to the superfamily of the ATP-binding cassette (ABC) transporters and contains a tandem repeat of transmembrane domains and conserved nucleotide-binding motifs connected by a central "linker" region (1,4,5).
MDR1 exhibits an ATP hydrolytic activity closely related to its drug transport function. This ATPase activity is significantly stimulated by the transported substrate drugs but is blocked by low concentrations of vanadate (6 -10). Senior and co-workers (11)(12)(13)(14) demonstrated that in the presence of vanadate and MgATP, similar to the effect observed with myosin and other related ATPases (see Ref. 15), MDR1 forms a strong complex with a radioactively labeled nucleotide. Vanadate stops the full catalytic cycle of MDR1, most probably by replacing inorganic phosphate, and stabilizes a protein-trapped form of a nucleotide, which was found to be exclusively ADP (11). Experimentally, this nucleotide trapping can be followed by using [␣-32 P]ATP as an energy donor substrate, because the MDR1-trapped labeled nucleotide is not removable by washings even in the presence of high concentrations of MgATP and/or MgADP. A covalent MDR1 labeling occurs if the photoaffinity analog, 8-azido-[␣-32 P]ATP is used in the trapping reaction, followed by UV light treatment.
It has been documented (12) that the formation of enzymebound (trapped) ADP from MgATP occurs randomly at the two nucleotide binding sites of MDR1, and whereas one MDR1 molecule is capable for the binding of two MgATP molecules, the saturation stoichiometry for the trapped ADP/MDR1 is one to one. The ADP-associated form of MDR1 may represent a high energy intermediate of the protein, required for the drugpump function (13).
In their experiments, Senior and colleagues (11)(12)(13) did not observe a modulation by the transported drug substrates either of the vanadate-dependent nucleotide trapping, or of the release of the trapped nucleotide in MDR1. In contrast, a substrate stimulation of the labeled nucleotide trapping has recently been reported in the case of MRP, another ABC transporter involved in drug resistance (16). Moreover, when studying the drug substrate interactions with MDR1, Urbatsch and Senior (17) and Dey et al. (18) found a significant inhibition of drug binding after vanadate-dependent nucleotide trapping, indicating a strong interaction between drug binding and ATP hydrolysis.
In the present paper, we demonstrate that the addition of transported drug substrates significantly increases the rate of vanadate-dependent nucleotide trapping in MDR1, when this process is studied under properly selected experimental conditions. To assay the reactions of the wild-type and mutant MDR1, the transport protein and its variants were expressed using the baculovirus-Spodoptera frugiperda (Sf9) insect cell system and characterized by analyzing their nucleotide trapping in isolated membrane preparations. The experiments presented clearly show that the vanadate-dependent nucleotide trapping in MDR1 reflects a drug-dependent partial reaction of the transport cycle, which is significantly modulated by sitedirected mutations in the pump protein. Our experiments also indicate a close cooperation of the two nucleotide binding sites in the drug-dependent trapping of nucleotides. 32 P]ATP (666 GBq/mmol) and [␣-32 P]ATP (111 TBq/mmol) were obtained from ICN Biomedicals.
The virus-infected Sf9 cells were suspended in a low ionic strength medium (containing 50 mM Tris-HCl, pH 7.0, 50 mM mannitol, 2 mM EGTA, 10 g/ml leupeptin, 8 g/ml aprotinin, 0.5 mM phenylmethylsulfonyl fluoride, 2 mM dithiothreitol) and disrupted using a glass-Teflon homogenizer. Membrane fractions were isolated by repeated centrifugations and homogenizations, and the membrane protein concentrations were determined as described in Ref. 6.
Electrophoresis and Immunoblotting-Membranes were suspended in a disaggregation buffer (6). Samples (20 l) were run on 6% Laemmlitype gels and electroblotted onto polyvinylidene difluoride membranes. Quantitative estimation of the expression of human MDR1 was performed using the polyclonal anti-MDR1 antibody 4077 (22), and a secondary antibody (anti-rabbit peroxidase-conjugated IgG; 20,000 ϫ dilution, Jackson Immunoresearch), as described in Ref. 21. Horseradish peroxidase-dependent luminescence (ECL, Amersham Pharmacia Biotech) was determined by luminography and quantitated by the Bio-Rad phosphorimager system.
Measurement of ATPase Activity-ATPase activity sensitive to vanadate was measured in isolated membranes as described in Ref. 6.
Vanadate-dependent Nucleotide Trapping-Isolated Sf9 cell membranes (100 g protein) were incubated for 30 s to 10 min at 37°C in a reaction buffer containing 50 mM Tris-KCl (pH 7. counter, whereas in the case of 8-azido-[␣-32 P]ATP labeling, the washed pellet was resuspended in 20 l of Tris-EGTA buffer, placed in a drop on a Parafilm-covered glass plate, cooled, and kept on ice. The samples were irradiated for 10 min with a UV lamp ( max about 250 nm) at a distance of 3 cm. Thereafter the membranes were collected in 40 l of the electrophoresis buffer, and the samples were run on 6% Laemmlitype gels. The proteins were electroblotted onto polyvinylidene difluoride membranes as described above and the blots were dried and subjected to autoradiography in a phosphorimager (Bio-Rad). The identity of the 32 P-azido-nucleotide labeled bands was assured by immunostaining with anti-MDR1 specific antibody (antibody 4077) on the same blot.

RESULTS
Nucleotide Trapping in MDR1 from Mg-8-azido-ATP-In this set of experiments, we examined the vanadate-dependent nucleotide trapping in isolated Sf9 cell membranes expressing the wild-type MDR1 by using the photoaffinity ATP analog, 8-azido-[␣-32 P]ATP. 8-Azido-ATP was reported to be an efficient energy donor substrate for the MDR1-ATPase, although both the K m and the V max values were found to be lower for Mg-8-azido-ATP than for MgATP (23). Labeling was performed in the presence of sodium orthovanadate as described under "Materials and Methods" for the time intervals and nucleotide concentrations indicated in Figs. 1 and 2. Verapamil and 5-fluorouracil (5FU) were used as test drugs because verapamil is a well-known substrate and activator of the MDR1 ATPase, whereas 5FU is not transported by MDR1 and does not stim-ulate MDR1-ATPase activity (1,6,24). The addition of 1 mM MgATP to the reaction medium abolished 8-azido-nucleotide trapping in MDR1, indicating a competition of MgATP and Mg-8-azido-ATP at the specific nucleotide binding sites (data not shown). Fig. 1A shows an autoradiogram, and Fig. 1B shows the corresponding immunoblot of the isolated Sf9 cell membranes labeled with 8-azido-[␣-32 P]ATP under the vanadate-dependent nucleotide trapping conditions. The major labeled protein band in the Sf9 cell membranes after 2 min of incubation at 37°C with 5 M Mg-8-azido-[␣-32 P]ATP is human MDR1. A weak labeling was also observed in the 70 -80-kDa proteolytic fragments of MDR1, which were formed during the labeling, washing, and irradiation procedure and which were also visible as reactive bands on the immunoblots (Fig. 1B). The addition of verapamil greatly increased the labeling of MDR1, whereas this labeling was unaffected by 5FU.
The sulfhydril reagent N-ethylmaleimide (NEM), by reacting with cysteine residues at the nucleotide binding sites, was shown to be a powerful covalent inhibitor of the MDR1 ATPase and drug transport function but did not affect primary nucleotide or drug binding (10). As demonstrated in Fig. 1A, nucleotide trapping was fully inhibited by 500 M N-ethylmaleimide. There was no measurable 32 P incorporation into membranes expressing ␤-galactosidase. As estimated from measurements of 32 P incorporation in the washed membranes and in the excised bands of MDR1 from the blots, UV treatment resulted in the cross-linking of 40 -50% of the trapped 32 P activity in each experiment. Fig. 2A represents the 8-azido-ATP concentration dependence of nucleotide trapping by MDR1, measured in the presence or absence of 50 M verapamil after 2 min of incubation at 37°C. Verapamil stimulation of labeling was 3-5-fold at 2-5 M 8-azido-ATP, and although the level of nucleotide trapping increased, the percentage of drug stimulation was much less at higher 8-azido-ATP concentrations. A saturation of the 8-azidonucleotide trapping was observed at about 100 M Mg-8-azido-ATP, both with and without the addition of verapamil. Fig. 2B shows the time course of 32 P-nucleotide trapping when the isolated Sf9 cell membranes were incubated with 5 M 8-azido-[␣-32 P]ATP at 37°C. As shown, the addition of 50 M verapamil increased the labeling 4 -7-fold at 1 min, whereas less stimulation was observed at increasing time periods (after 5 min, the degradation of ATP significantly reduced the rate of further labeling). In each set of experiments, both ␤-galactosidaseexpressing and NEM-pretreated MDR1-expressing membranes served as controls. Altogether, these experiments strongly suggest that the rate of nucleotide trapping in MDR1 from 8-azido-ATP is significantly increased by the addition of a known drug substrate of this transporter, whereas N-ethylmaleimide pretreatment blocks the formation of a trapped nucleotide.
Nucleotide Trapping in MDR1 from MgATP-The physiological energy donor for drug transport by MDR1 is MgATP; therefore, we performed experiments similar to those described above by using Mg[␣-32 P]ATP. Because nucleotide cross-linking could not be found in this case (see also Ref. 12), nucleotide trapping was measured by counting the membrane-bound radioactivity after repeated washings of the membranes in the presence of 200 M sodium orthovanadate and 10 mM MgATP at 4°C (see "Materials and Methods"). It is to be noted that in contrast to verapamil, 5-fluorouracil had no effect on vanadate-dependent nucleotide trapping at any MgATP concentration examined (not shown), and 32 P incorporation was significantly lower in the control, ␤-galactosidase-expressing membranes. Nucleotide trapping was significantly reduced in the NEM-pretreated, MDR1-expressing membranes, and in the presence of this MDR1 transport and ATPase inhibitor, 32 P incorporation was not influenced by verapamil or by other MDR1 substrate drugs (see below). Fig. 3B shows the time course of 32 P-nucleotide trapping when the MDR1-expressing isolated Sf9 cell membranes were incubated with 50 M [␣-32 P]ATP at 37°C. We found that 50 M verapamil increased the labeling 4 -6 times at 30 s to 1 min, whereas significantly less verapamil stimulation was observed after longer incubation periods. Again, a low level of 32 P-nucleotide trapping was found in the NEM-pretreated membranes and in the control, ␤-galactosidase-expressing membranes; the level of trapping was not influenced by verapamil.
Effects of Various Drugs on the Nucleotide Trapping in MDR1- Fig. 4 shows the detailed examination of the effects of various drugs on the nucleotide trapping in MDR1-expressing membranes incubated at 37°C for 2 min with 5 M 8-azido-[␣-32 P]ATP (Fig. 4A) 4B). As shown in Fig. 4A, covalent labeling after nucleotide trapping from 8-azido-ATP was found to be significantly stimulated by the MDR1 substrate drugs verapamil, cyclosporine A (CsA), or calcein AM (CaAM), whereas free calcein (Ca free) and 5FU, which are not MDR1 substrates (see Refs. 1, 2, 6 -8, and 24), had no effect on this phenomenon. Fig. 4B demonstrates the drug concentration dependence of nucleotide trapping in MDR1 from Mg[␣-32 P]ATP. Again, verapamil, CsA, vincristine, and rhodamine 123 significantly stimulated nucleotide trapping in a concentration-dependent manner, whereas free calcein and 5FU had no effect. The stimulation of nucleotide trapping was evoked by much lower concentrations of CsA, verapamil, or calcein AM than of, for example, rhodamine 123. This pattern closely corresponds to the drug concentration dependence of the MDR1-ATPase activity (see Refs. 6, 8, 10, and 25), although the stimulatory effect of cyclosporine A was much more pronounced than in the MDR1-ATPase activity measurements (see "Discussion").
Nucleotide Trapping in a Substrate Affinity Mutant of MDR1-A mutant form of MDR1 with a 20-amino acid deletion in its first extracellular loop (⌬aa 78 -97 MDR1), has been shown to have a reduced drug transport capacity (27) and a low level of ATPase activity, stimulated only by extremely high concentrations (above 200 M) of verapamil (20). Nucleotide trapping in this mutant MDR1, expressed in Sf9 cell mem-branes, was followed in the presence of 5 M Mg-8-azido-ATP and 200 M vanadate for 2 min at 37°C, and covalent labeling was achieved by UV light treatment.
As shown in Fig. 5, in the ⌬aa 78 -97 MDR1 protein, nucleotide trapping was almost negligible without added drugs, and 50 M verapamil or 20 M calcein AM (which produced a maximum stimulation of nucleotide trapping in the wild-type MDR1) had little effect on this labeling. However, the addition of high concentrations of verapamil (300 M) produced a significant labeling even in the mutant MDR1, approaching the level seen in the wild-type protein (Fig. 5C).
Nucleotide Trapping in Nucleotide Binding Site Mutants of MDR1-In the following experiments, we examined nucleotide trapping with 8-azido-[␣-32 P]ATP (followed by photo-crosslinking) in MDR1 variants in which essential lysine residues in the Walker A motifs were mutated. These lysines were replaced by methionines either in the N-terminal ABC domain (K433M), in the C-terminal ABC domain (K1076M), or in both ABC domains (K433M/K1076M). These mutant MDR1s, when expressed in Sf9 cells, were shown to demonstrate significant 8-azido-ATP binding but no drug transport or drug-stimulated ATPase activity (21). Because, in these mutant proteins, nucleotide binding may also be altered at low MgATP concentra- In the experiments shown in Fig. 6, isolated Sf9 membranes were incubated at 37°C with 200 M vanadate in the presence of either 5 M Mg-8-azido-ATP for 2 min (Fig. 6A), or 50 M Mg-8-azido-ATP for 10 min (Fig. 6B). The reaction medium contained either no drug or 50 M verapamil. In these experiments, ␤-galactosidase-expressing membranes, wild-type MDR1, and its NEM-pretreated form served as comparisons. As shown in Fig. 6, none of the mutant MDR1 variants performed any nucleotide trapping, even if higher 8-azido-ATP concentrations and longer incubation periods were used. Fig. 7 shows the autoradiograms (Fig. 7A) and the immunoblots (Fig. 7B) for some of the experimental data presented in Fig. 6. As shown, there was no significant azido-ATP trapping in either the K433M or K1076M mutant MDR1 proteins, whereas the corresponding immunoblots ensured that the isolated Sf9 cell membranes contained about equal amounts of the MDR1 variants in all experiments.

DISCUSSION
The detailed reaction mechanism of ABC transporters, which involves the hydrolysis of MgATP and a concomitant membrane transport of the relevant substrates, is still largely unknown. The discovery of a vanadate-dependent trapping of a nucleotide during the reaction cycle of the multidrug transporter MDR1 (11)(12)(13)(14), opened new avenues for the detailed examination of its reaction mechanism. In the present experiments, we studied the effects of drug substrates on this nucleotide trapping in MDR1, as well as the modulation of this phenomenon by well characterized mutations in the protein.
In the catalytic cycle model originally suggested for MDR1 by Senior et al. (13), MgATP binding to MDR1 is followed by a conversion of this nucleotide to MgADP and P i . Addition of vanadate blocks MDR1-ATPase probably by replacing the released inorganic phosphate and results in the formation of a trapped nucleotide, which is predominantly ADP (12,13). Our experiments indicate that the vanadate-dependent nucleotide trapping by MDR1 most probably represents a partial reaction of the enzyme activity, the rate of which is greatly increased by the transported drug substrates. We suggest that vanadate trapping stabilizes an occluded, enzyme-bound form of the cleaved nucleotide, the formation of which shows a drug concentration dependence closely resembling the drug stimulation of the full MDR1-ATPase cycle.
Judged from MgATP binding studies (see Refs. 28 and 29) and from the MgATP dependence of the MDR1-ATPase (see Refs. 6, 10, and 25), MDR1 binds MgATP with a relatively low affinity, with a K D of about 200 -500 M. The binding of MgATP occurs randomly at the two nucleotide binding sites, and the stoichiometry of MgATP/MDR1 at saturating MgATP concentrations is 2:1 (12). Data in the literature indicate that this MgATP binding is not affected by the presence of drug substrates (1,10,28). Experimental data also suggest that primary drug binding by MDR1 (as measured at 4°C) is similarly unaffected by the presence or absence of MgATP (see Refs. 17 and 30); thus, initial MgATP and drug binding seem to be independent reactions.
In contrast to MgATP binding, ATP hydrolysis requires an interaction of the protein with the transported drug substrate. Human MDR1-ATPase activity, as measured in isolated Sf9 cell membranes, is about 5-6-fold higher in the presence than in the absence of drug substrates, e.g. verapamil (the "basic" MDR1-ATPase activity is probably supported by endogenous lipid-like or other hydrophobic molecules in the membrane preparation). As we demonstrate in this report, the partial reaction of MDR1-ATPase (reflected in the MDR1 nucleotide trapping), is also strongly accelerated by the transported substrates.
In previous studies (11)(12)(13), this drug stimulation of nucleotide trapping was not observed, probably because the experiments were carried out at relatively high ATP concentrations and/or for long measurement periods. Drug substrate stimulation of the nucleotide trapping in our experiments is also restricted to early time periods and unsaturating MgATP (or 8-azido-MgATP) concentrations, that is, under conditions where the drug-dependent acceleration still significantly affects the "titration" of the relevant nucleotide trapping site. It is interesting to note that the differences between the concentration dependence and the time course of nucleotide trapping observed here for Mg-8-azido-ATP and for MgATP can be well explained by the lower K m and V max values of Mg-8-azido-ATP than of MgATP (17,29): although higher MgATP concentrations are required to saturate the nucleotide binding sites, both ATP splitting and nucleotide trapping are faster with MgATP than with Mg-8-azido-ATP. In our present experiments under near-saturating conditions, the molar ratio of nucleotide trapping by MDR1 was estimated to be about 0.4 -0.6, supporting the conclusion of Urbatsch et al. (12) that only one nucleotide is promoted to an occluded state during this partial reaction.
The drug substrate concentration dependence of the MDR1-ATPase activity and that of the nucleotide trapping were found to be similar, that is, a significant stimulation of both reactions was obtained by similar concentrations of verapamil, vincristine, rhodamine 123, or calcein AM. The data obtained for the deletion mutant in the first extracellular loop of MDR1 (⌬aa 78 -97), namely that stimulation could be obtained only with extremely high concentrations of verapamil, also strongly suggest that substrate interactions modify the process of nucleotide trapping similarly to that of the MDR1-ATPase activity. Still, some substrates (such as CsA) that yield only very small stimulation of the ATPase but strongly inhibit its verapamil stimulation (31) produced a much greater stimulation of nucleotide trapping (the inhibition of this reaction at higher CsA concentrations was also apparent). These data indicate that the partial reaction of nucleotide trapping may be efficiently promoted by some substrates that can yield only a low turnover for the MDR1-ATPase activity, which reflects the full catalytic cycle.
Interaction of the multidrug transporter with SH group reagents, such as NEM, preferentially occurs at the two cysteine residues in the two nucleotide binding domains (29,32). Pretreatment of MDR1 with SH group reagents blocks MDR1-ATPase activity (29) but at low concentrations does not inhibit the primary binding of MgATP or drug substrates, as shown by the MgATP-or drug-dependent quenching of an MDR1-bound fluorescent SH-reactive probe (32). A recent communication (33) suggests that NEM may in fact increase drug binding by MDR1 under certain conditions. Because, as reported here, NEM fully blocks drug-stimulated nucleotide trapping, MgATP and primary drug substrate binding most probably occur independently from the following reaction steps, leading to ATP hydrolysis.
The mutant MDR1 proteins, in which lysines in the first (K433M), second (K1076M), or both nucleotide binding domains are replaced by methionines, were demonstrated to bind MgATP less efficiently at low MgATP concentrations (2-5 M) but similarly to the wild-type MDR1 at concentrations above 10 M MgATP (21). None of these MDR1 variants possess measurable ATPase activity (21); thus, a mutation in one ATP binding domain is sufficient to eliminate the catalytic reaction in the whole pump protein (see also Refs. 26 and 34). This finding was interpreted to indicate a strong cooperative inter-action between the two nucleotide binding domains of MDR1. As we demonstrate here, independently of the presence or absence of drug substrates, none of these nucleotide binding site mutants perform the nucleotide trapping reaction. Thus, a mutation in one of the nucleotide binding domains eliminates nucleotide trapping entirely, showing a cooperation between the ABC domains already in this partial reaction.
Based on the above-described features, we propose that the molecular mechanism of the vectorial drug transport by MDR1 is initiated by the independent primary binding of MgATP and the drug substrate. The following conformational changes in the structure of the MDR1 protein induce a drug-dependent cleavage and a concomitant occlusion of a MgADP molecule, whereas the full catalytic cycle involves the transport of drug substrate to the external membrane surface and the full hydrolysis of ATP and the dissociation of ADP and inorganic phosphate. All of the steps after MgATP and drug binding are based on the functional cooperation of the two ATP binding domains and the appropriate drug binding domains within MDR1.
The drug-dependent ATP hydrolysis and occlusion most probably significantly alter the conformation and/or location of the drug binding site(s). As shown by Urbatsch and Senior (17), vanadate-induced nucleotide trapping significantly reduced azidopine labeling in MDR1. Recent experiments of Dey et al. (18) demonstrated the presence of two nonidentical drug-interaction sites in the MDR1 protein. In this study the C-terminal drug-recognition ("on") site was found to be significantly more sensitive to vanadate trapping of nucleotides than the N-terminal ("off") site, and ATP hydrolysis was essential for the vanadate-induced reduction of drug binding. Our experiments and these data collectively suggest that ATP hydrolysis is strongly coupled to the movement of the drug substrate from an "on" to an "off" site within the MDR1 protein, and nucleotide occlusion may coincide with the occlusion of the drug binding site(s), leading to a vectorial movement of the drug substrate.
It should be noted that the above-described stimulation of nucleotide trapping in MDR1 by the substrate drugs may also be efficiently employed for the screening of specific drug interactions with MDR1, because even those substrates (e.g. CsA) that predominantly inhibit the MDR1-ATPase activity may show a strong stimulation of nucleotide trapping. Moreover, the investigation of substrate stimulation of nucleotide trapping in other related proteins, e.g. CFTR, TAP, or MRP, may help to understand the reaction mechanism and transport characteristics of these ABC transporters.