Role of GSH in Estrone Sulfate Binding and Translocation by the Multidrug Resistance Protein 1 (MRP1/ABCC1)*

Multidrug resistance protein 1 (MRP1/ABCC1) is an ATP-dependent efflux pump that can confer resistance to multiple anticancer drugs and transport conjugated organic anions. Unusually, transport of several MRP1 substrates requires glutathione (GSH). For example, estrone sulfate transport by MRP1 is stimulated by GSH, vincristine is co-transported with GSH, or GSH can be transported alone. In the present study, radioligand binding assays were developed to investigate the mechanistic details of GSH-stimulated transport of estrone sulfate by MRP1. We have established that estrone sulfate binding to MRP1 requires GSH, or its non-reducing analogue S-methyl GSH (S-mGSH), and further that the affinity (Kd) of MRP1 for estrone sulfate is 2.5-fold higher in the presence of S-mGSH than GSH itself. Association kinetics show that GSH binds to MRP1 first, and we propose that GSH binding induces a conformational change, which makes the estrone sulfate binding site accessible. Binding of non-hydrolyzable ATP analogues to MRP1 decreases the affinity for estrone sulfate. However, GSH (or S-mGSH) is still required for estrone sulfate binding, and the affinity for GSH is unchanged. Estrone sulfate affinity remains low following hydrolysis of ATP. The affinity for GSH also appears to decrease in the post-hydrolytic state. Our results indicate ATP binding is sufficient for reconfiguration of the estrone sulfate binding site to lower affinity and argue for the presence of a modulatory GSH binding site not associated with transport of this tripeptide. A model for the mechanism of GSH-stimulated estrone sulfate transport is proposed.

The development of resistance to chemotherapy remains one of the biggest obstacles to the successful treatment of cancer. One of the mechanisms by which tumors acquire resistance is the overexpression of membrane transport proteins such as multidrug resistance protein 1 (MRP1 2 or ABCC1) (1)(2)(3). MRP1 is a member of the C subfamily of the ABC superfamily of transport proteins. It is expressed in most tissues in the body and utilizes energy from ATP hydrolysis to transport a wide variety of structurally and functionally distinct molecules across the plasma membrane. Many MRP1 substrates are conjugated organic anions such as LTC 4 (4,5), E 2 17␤G (6, 7), and estrone sulfate (8). In addition to transporting these endogenously formed metabolites, MRP1 has an important role in detoxification and protection of normal tissues from xenobiotics (9 -12). However, this is problematic when MRP1 is expressed at elevated levels in cancer cells, because it causes resistance to anticancer agents, including anthracyclines, plant alkaloids, and antifolates, by transporting them out of cells.
An unusual feature of MRP1 (and some other members of the ABCC subfamily) is the requirement for the reducing tripeptide GSH to be present for transport of several of its substrates to occur, including certain anticancer drugs (5,8,(13)(14)(15)(16)(17). GSH is an essential cellular antioxidant, which reacts with electrophilic molecules to form conjugated hydrophilic metabolites that can be more readily eliminated from the body. Prior to elimination, such metabolites must be actively effluxed from the cells in which they are formed. MRP1 can transport many such GSH-conjugated anions (18); however, the role of GSH in MRP1-mediated transport is not limited to forming conjugates. For example, transport of the Vinca alkaloid vincristine is markedly enhanced by GSH, and the interaction appears to be co-transport with the drug without formation of a conjugate (14). Estrone sulfate transport by MRP1 is also enhanced by GSH; however, in contrast to vincristine transport, GSH only stimulates the process and is not itself transported (8). Some MRP1 inhibitors such as the tricyclic isoxazole derivative LY475776 also require GSH to photolabel to MRP1 (19), yet transport of some drugs, such as the antifolate methotrexate occur independently of GSH (20). In contrast, GSH can be transported alone, and some xenobiotics, including verapamil, or bioflavonoids such as apigenin, stimulate GSH efflux without being transported themselves (21,22). Cellular GSH/oxidized glutathione levels are important for regulating the redox status of the cell and modulating responses to oxidative stress. Efflux of GSH in response to cell damage has been found to be part of an apoptotic signaling pathway (23)(24)(25)(26), and it has recently been demonstrated that this efflux can be mediated by MRP1 (26,27). However, it is not the reducing ability of GSH that is important for its interactions with MRP1, because non-reducing analogues such as ophthalmic acid or S-mGSH can functionally substitute for GSH (8,14,16,17,19). Thus it is clear that the role of GSH in MRP1-mediated transport is very complex, but as yet relatively little is known about the mechanism of interaction of GSH with this transporter.
MRP1 function is often evaluated by measuring ATP-dependent uptake of a radiolabeled substrate into inside-out membrane vesicles prepared from cells expressing the transporter. Transmembrane transport of a substrate assayed in this way gives an overall measure of a multistep process. First, the substrate binds to a high affinity site on the cytosolic side of the membrane. This substrate binding site is subsequently reoriented to the extracellular face of the membrane where the binding affinity is decreased so that the substrate may be released. The final step of the transport cycle is the return of the binding site to its initial high affinity state (28). For MRP1 (and other ABC transporters), these steps of the transport cycle are coupled to the binding and/or hydrolysis of ATP, but the precise details of this coupling are not known. Nor is it known when and how GSH participates in this process. Because conventional vesicular transport assays do not readily allow the individual constituent steps of the transport cycle to be examined, we have developed a radioligand binding assay to investigate directly the binding of estrone sulfate to MRP1, to measure the effect of GSH on the binding of this substrate, and to determine how the binding of estrone sulfate and GSH is affected by nucleotide binding and hydrolysis.
Cell Culture and Membrane Preparation-The H69 small cell lung cancer cell line and its doxorubicin-selected, MRP1-overexpressing H69AR-derivative cell line were cultured as described previously (29). To prepare plasma membranes, cells (ϳ5 ϫ 10 8 ) were harvested, resuspended in buffer 1 (50 mM Tris, pH 7.4, 250 mM sucrose) containing 250 M CaCl 2 and protease inhibitors, and subjected to argon cavitation (250 p.s.i., 5 min, 4°C). Unexploded cells and debris were removed by low speed centrifugation (750 ϫ g, 10 min, 4°C), and the supernatant was loaded onto a 35% sucrose cushion. Following centrifugation (100,000 ϫ g, 1 h, 4°C), plasma membranes were harvested from the sucrose interface, diluted with buffer 2 (50 mM Tris, pH 7.4, 25 mM sucrose), and centrifuged again. The membrane pellet was resuspended in buffer 1 at ϳ10 mg of protein per ml and stored at Ϫ70°C for up to 3 months. Protein concentrations were measured using a Bio-Rad Bradford assay, with bovine serum albumin as a standard.
[ 3 H]Estrone Sulfate Binding Assays-Membrane protein (10 g) was allowed to equilibrate with [ 3 H]estrone sulfate in the presence or absence of organic anion substrates and/or nucleotides in a total volume of 50 l of hypotonic buffer 3 (50 mM HEPES, pH 7.4) at 23°C for 1 h. The use of hypotonic buffer ensures that the membrane vesicles are burst open and, thus, that the MRP1 protein is accessible from both sides of the membrane. Following equilibration, 1.5 ml of ice-cold buffer 4 (20 mM HEPES, pH 7.4, 20 mM MgCl 2 ) was added, and samples were filtered immediately through a PerkinElmer Life Sciences unifilter GF/B plate using a Packard Filtermate Harvester, and washed twice with buffer 4. Tritium bound to the filter was determined using a PerkinElmer Life Sciences Top Count NXT Microplate Scintillation counter.
Displacement/Stimulation Assays-Binding assays were used to measure the displacement of estrone sulfate binding by other substrates of MRP1 or the increase in estrone sulfate binding in the presence of GSH analogues. Membrane proteins (10 g) were incubated with [ 3 H]estrone sulfate (35-40 nM) in the presence of (a) 3 mM S-mGSH and various concentrations of LTC 4 (3 nM to 10 M), E 2 17␤G (300 nM to 3 mM), or methotrexate (1 M to 3 mM) or (b) various concentrations of GSH or S-mGSH (1 M to 10 mM), before filtering as described above. DTT (10 mM) was present in all experiments containing GSH to prevent oxidation to oxidized glutathione. Data were fitted with a sigmoidal dose-response curve by non-linear regression analysis (Equation 1), where B is the amount of estrone sulfate bound, B max is maximal binding at this concentration of estrone sulfate, B min is nonspecific binding, [S] ϭ concentration of substrate, and EC 50 is the substrate concentration at half-maximal estrone sulfate binding.
Saturation Binding Isotherms-To determine parameters of maximal binding capacity (B max ) and binding dissociation constant (K d ), 10 g of membrane protein was incubated with various concentrations of [ 3 H]estrone sulfate (10 nM to 30 M). This assay was done in the presence of 0.3, 1, 3, or 10 mM S-mGSH or GSH (plus 10 mM DTT) in the presence or absence of 1 mM E 2 17␤G. Nonspecific binding (in the presence of 1 mM E 2 17␤G) was subtracted from total binding to determine specific binding, and specific data were fitted with a one-site binding hyperbola (Equation 2), where B is the amount of estrone sulfate bound, B max is maximal binding capacity for estrone sulfate, [L] is the concentration of estrone sulfate, and K d is the binding dissociation constant. The commercially available radioligand could not be used to generate [ 3 H]estrone sulfate concentrations above 3 M, and this was therefore achieved by supplementing with unlabeled estrone sulfate as previously described (30).
Association Kinetics-The rate of association of estrone sulfate with MRP1 in the presence of S-mGSH was determined by using the following three protocols. Comparison of the data from each protocol was used to ascertain the order of binding of GSH and estrone sulfate to MRP1. Firstly, membrane proteins (10 g) were preincubated with 3 mM S-mGSH for 30 min in the presence or absence of 1 mM E 2 17␤G (for determination of nonspecific estrone sulfate binding). [ 3 H]Estrone sulfate (40 nM) was added at selected time points (0 -60 min) and then stopped by addition of buffer 4 and filtration as described above. Secondly, [ 3 H]estrone sulfate (40 nM) and S-mGSH (3 mM) were mixed, with or without E 2 17␤G (1 mM), and membrane proteins (10 g) were added at selected times (0 -60 min) before filtering. Lastly, membrane proteins (10 g) were preincubated with [ 3 H]estrone sulfate (40 nM) for 30 min, with or without 1 mM E 2 17␤G, and S-mGSH (3 mM) was added at selected times (0 -60 min) before filtering. For each of the above three protocols, specific estrone sulfate binding was calculated by subtracting nonspecific binding (in the presence of 1 mM E 2 17␤G) from total binding (in the absence of E 2 17␤G), for each time point. An exponential association curve (Equation 3) was fitted to the specific binding data by non-linear regression, where B is the estrone sulfate bound, B max is maximal binding at this concentration of estrone sulfate, t is time, and k obs is the observed rate constant.
Addition of Nucleotides-AMPPNP or ATP␥S was added to membrane proteins to a final concentration of 4 mM in the presence of 5 mM MgSO 4 . Vanadate trapping of nucleotide was achieved by incubating membrane proteins at 37°C for 30 min with 4 mM ATP, 5 mM MgSO 4 , and 1 mM sodium orthovanadate (prepared as a 100 mM stock solution, pH 10, and boiled prior to use (31)). Excess ATP was removed by centrifugation (25,000 ϫ g, 15 min, 4°C), and the membranes were resuspended in buffer 3 containing 1 mM orthovanadate, ready for use in the estrone sulfate binding assays. Membrane proteins were subjected to the same conditions but without ATP as a control and were included in each assay to ensure there was no spurious effect of the nucleotidetrapping conditions.
[ 3 H]LTC 4 Transport to Measure Vanadate-induced Nucleotide Trapping Efficiency-Vanadate-induced nucleotide trapping by MRP1 was carried out as described above, except buffer 1 (sucrose buffer) was used in place of buffer 3 (hypotonic buffer), and following centrifugation to remove unbound ATP, the vesicles were resuspended in buffer 1 without vanadate and immediately used in LTC 4 transport assays. As before, a sample subjected to the same conditions but without ATP was prepared as a control. [ 3 H]LTC 4 transport was carried out as described previously (5,32). Briefly, 50 nM [ 3 H]LTC 4 (20 nCi per point), 10 mM MgCl 2 , and 4 mM ATP (plus an ATP-regenerating system consisting of creatine kinase and creatine phosphate) or 4 mM AMP were incubated with 2 g of membrane protein at 23°C for 1 min. The reaction was stopped by dilution in ice-cold buffer 1 and rapidly filtered as described for the binding assays. Uptake in the presence of AMP was subtracted from uptake in the presence of ATP to determine ATP-dependent transport. The residual ATP-dependent [ 3 H]LTC 4 uptake in the vanadate-trapped sample was compared with the control sample as a measure of vanadate-induced nucleotide trapping efficiency.
The direct effect of estrone sulfate, GSH, and/or DTT on trypsin activity per se was assayed using the model substrate BAEE. Diphenylcarbamyl chloride-treated trypsin (500 units ml Ϫ1 in 1 mM HCl) was incubated alone, or in the presence of 50 M estrone sulfate, 10 mM DTT, 10 mM GSH, 10 mM S-mGSH or combinations of these reagents at 23°C for 30 min. The trypsin (200 l) was then added to 0.25 mM BAEE (3 ml), and trypsin activity was determined as the rate of increase of absorbance at 250 nm over a 10-min period. 3 Data Analysis-All non-linear regression analyses were carried out using GraphPad Prism 3.0 (San Diego, CA). Data sets contain a minimum of three independent experiments, and data are depicted as means Ϯ S.E. Statistical comparisons were carried out using one-way analysis of variance with a Tukey post-hoc test. Differences were considered statistically significant when p was Ͻ0.05.

Requirement of GSH for Specific Estrone Sulfate
Binding-Radioligand binding assays were developed to study the interaction of estrone sulfate and GSH with MRP1. Approximately 250,000 dpm of [ 3 H]estrone sulfate was added to each reaction (10 g of membrane), of this Ͻ1% was bound by the filters. As shown in Fig. 1A, the level of [ 3 H]estrone sulfate binding to H69AR (MRP1 ϩ ) membranes in the absence of GSH was ϳ1,100 dpm, the same as that to membranes from parental H69 (MRP1 Ϫ ) cells, indicating low level nonspecific binding to non-MRP1 components of the membrane preparation. Consistent with this conclusion, the amount of estrone sulfate binding could not be displaced by the MRP1 substrate E 2 17␤G. The inability to measure specific estrone sulfate binding could also not be overcome by increasing concentrations of [ 3 H]estrone sulfate, at least up to 10 M (data not shown). In contrast, the addition of GSH produced a substantial increase in [ 3 H]estrone sulfate binding to the MRP1 ϩ membranes while having no effect on the MRP1 Ϫ membranes (Fig. 1A), indicating that specific binding of estrone sulfate to MRP1 occurs in the presence of GSH. The specificity of this GSH-induced binding was confirmed by the ability of excess E 2 17␤G to reduce [ 3 H]estrone sulfate binding to the nonspecific level seen in the absence of GSH while having no effect on the MRP1 Ϫ membranes. The specific binding of estrone sulfate to MRP1 could also be readily measured in the presence of the non-reducing GSH analogue S-mGSH. Together these data indicate that membranes from H69AR cells display a specific binding site for estrone sulfate but only in the presence of GSH (or S-mGSH). The lack of binding to H69 membranes indicates that the binding to membranes from resistant cells was to the MRP1 transporter. Support for the specific binding component as MRP1 was provided through heterologous displacement assays using several other known substrates of MRP1. Thus, LTC 4 (IC 50 0.95 M), E 2 17␤G (IC 50 90 M), and methotrexate (IC 50 1.8 mM) all produced concentration-dependent reductions in estrone sulfate binding to levels observed in the H69 (MRP1 Ϫ ) membranes (Fig. 1B). The order of potency to displace estrone sulfate binding matched the relative uptake affinities for MRP1 displayed by these substrates (5,6,20). 3 Available at www.sigmaaldrich.com/sigma/enzyme%20assay/t9253enz.pdf. Estrone Sulfate Binding to MRP1 Is Greater in the Presence of S-mGSH than GSH-The increase in estrone sulfate binding to MRP1 in the presence of increasing concentrations of GSH or S-mGSH is shown in Fig. 2. The level of estrone sulfate binding in the presence of S-mGSH was significantly higher than in the presence of GSH itself, a difference consistent with the relative abilities of these two tripeptides to stimulate estrone sulfate vesicular transport by MRP1 (8,34). This difference is not due to the presence of DTT, which is added to all experiments using GSH to prevent oxidation to oxidized glutathione, because comparable results are obtained with S-mGSH whether supplemented with DTT or not (data not shown). Despite differences in the amount of estrone sulfate bound, the potencies of GSH and S-mGSH to increase estrone sulfate binding were comparable, with EC 50 values of 0.75 Ϯ 0.12 mM and 0.77 Ϯ 0.06 mM, respectively. These observations indicate that MRP1 displays no difference in its affinity for GSH or S-mGSH.
Affinity and Binding Capacity of MRP1 for Estrone Sulfate-To determine the affinity and maximal binding capacity of MRP1, saturation binding isotherms for estrone sulfate, in the presence of either GSH or S-mGSH (3 mM), were determined (Fig. 3). A statistically significant difference between the affinity (K d ) for estrone sulfate binding to MRP1 was observed in the presence of 3 mM GSH (1.48 Ϯ 0.06 M) compared with 3 mM S-mGSH (0.59 Ϯ 0.05 M, p Ͻ 0.001). However, there was no difference in the maximal binding capacity for estrone sulfate (B max 83 Ϯ 8 pmol mg Ϫ1 with GSH versus 87 Ϯ 7 pmol mg Ϫ1 with S-mGSH). The difference in affinity accounts for the higher level of estrone sulfate binding in the presence of S-mGSH compared with GSH shown in Fig. 2.
The effects of varying GSH and S-mGSH concentrations on the affinity and binding capacity of MRP1 for estrone sulfate are summarized in Table 1. As the concentration of GSH increased, a corresponding increase in the estrone sulfate maximal binding capacity (B max ) was observed. However, there was no change in the affinity (K d ) for estrone sulfate. Similar results were observed with S-mGSH. This suggests that, as the concentration of GSH (or mGSH) is increased, more of the tripeptide is available to bind to MRP1, and therefore more MRP1 molecules can bind estrone sulfate. Notably no differences in the B max for estrone sulfate in the presence of GSH or S-mGSH at the same concentration were observed. These data provide additional support for the conclusion that there is no difference in the affinity of MRP1 for GSH or S-mGSH. It was not possible to directly measure the K d for GSH using radioligand binding assays because of the low affinity of MRP1 for GSH (millimolar range (35)) and the limiting concentration and specific activity of commercially available tritiated GSH.
Association Kinetics for Estrone Sulfate Binding to MRP1-To determine whether the order of binding of GSH/S-mGSH and estrone sulfate to MRP1 was important, the rate of association of estrone sulfate to MRP1 was measured under three different conditions: 1) preincubation of MRP1 with S-mGSH before addition of estrone sulfate, 2) addition of S-mGSH and estrone sulfate to MRP1 at the same time, and 3) preincubation of MRP1 with estrone sulfate prior to the addition of S-mGSH. As shown in Fig. 4, the rate of association under the first set of conditions is markedly faster than under the second or third set of conditions, and the latter two cannot be distinguished from each other. The level of estrone sulfate binding reached half of the equilibrium level in 2 min when MRP1 was preincubated with S-mGSH, approximately half the time required when S-mGSH and estrone sulfate were added at the same time or when S-mGSH was added after estrone sulfate (associa-    tion parameters are summarized in Table 2). These observations indicate that the GSH or S-mGSH must bind to MRP1 first before estrone sulfate can bind.

GSH Binding Causes a Conformational Change in MRP1-
The mechanism by which GSH binding allows estrone sulfate to bind MRP1 is not clear. One possibility is that GSH binding causes a conformational change in MRP1 that opens up a binding site for estrone sulfate. Altered sensitivity to proteolysis has been previously reported to reflect conformational changes in several different membrane proteins (36 -39). Consequently, limited trypsin digestion of MRP1 was carried out in the presence and absence of GSH and estrone sulfate. As shown in Fig. 5A, digestion of MRP1 with increasing concentrations of trypsin produces first a larger 80-kDa COOH-proximal fragment (C1) and then the smaller 36-to 40-kDa COOH-terminal fragment (C2), as detected using mAb MRPm6 (COOH-terminal epitope) and characterized previously (19,40).
In the presence of 50 M estrone sulfate, there was no change in this digestion profile (data not shown). In the presence of 10 mM GSH (and DTT), however, a different profile was observed (Fig. 5A). Thus, the intact MRP1 protein was more susceptible to tryptic digestion than the control, yet there was also a protective effect against cleavage of the C1 fragment to produce the C2 fragment. This protection of the COOH terminus of MRP1 afforded by GSH agrees with a previously published report (41). A similar effect on the digestion profile was observed when S-mGSH was substituted for GSH (data not shown). Furthermore, the presence of estrone sulfate in addition to GSH did not significantly change the profile compared with GSH alone (data not shown). In contrast to the ability of GSH to protect the COOH terminus of MRP1, GSH had no effect on the trypsin-mediated cleavage of more NH 2proximal regions of MRP1 as detected with mAb MRPr1 (epitope in intracellular loop 3 between the first and second membrane-spanning domains) (data not shown).
The relatively long exposure time (5 min) used for the immunoblot shown in Fig. 5A, gives the appearance of large scale breakdown of MRP1 to the C1 fragment even at very low concentrations of trypsin. However, as seen in Fig. 5B, the short exposure time (30 s) of lane 1 shows this is simply an artifact produced by the overexposure of the higher molecular weight bands to detect the C2 COOH-terminal fragments. The possibility that the effect of GSH (plus DTT) or S-mGSH on the susceptibility of MRP1 to tryptic digestion was due to a direct effect on the activity of trypsin was excluded using a BAEE assay (data not shown). Our results suggest that binding of GSH to MRP1 induces a conformational change that appears specific to the COOH-terminal half of the protein, which subsequently enhances estrone sulfate binding.
Effect of ATP Binding on Estrone Sulfate Binding-The studies above describe estrone sulfate and GSH binding to MRP1 in the nucleotidefree state. To more fully understand the role of GSH in estrone sulfate transport, studies in the presence of nucleotides were undertaken. Thus, binding of estrone sulfate to MRP1 was measured in the presence of AMPPNP (4 mM), a non-hydrolyzable analogue of ATP, at a concentra-   Assay conditions were as follows: 1) membranes were preincubated with S-mGSH before adding ͓ 3 H͔estrone sulfate; 2) S-mGSH and ͓ 3 H͔estrone sulfate were added to membranes at the same time; and 3) membranes were preincubated with ͓ 3 H͔estrone sulfate before adding S-mGSH. b Observed association rate constants (k obs ) and time taken for estrone sulfate binding to reach half the equilibrium binding level (t1 ⁄ 2 ) were determined for ͓ 3 H͔estrone sulfate (40 nM) binding to MRP1 (10 g of membrane protein) (Fig. 4). Data are means Ϯ S.E. from at least three independent experiments. c Denotes a significant difference from assay condition 1 ( p Ͻ 0.001).
tion we showed previously was sufficient to bind to both NBDs of MRP1 (42). As observed with nucleotide-free MRP1 in the absence of GSH, there was only low level nonspecific binding of estrone sulfate, but in the presence of GSH, MRP1-specific binding was readily detectable (Fig.  6A). However, the level of GSH-enhanced MRP1-specific estrone sulfate binding in the presence of AMPPNP was significantly reduced compared with binding in the absence of nucleotide. Comparable results were obtained when 4 mM ATP␥S (a very slowly hydrolyzable ATP analogue) was used instead of AMPPNP.
Further examination of estrone sulfate binding over a full concentration range in the presence of 3 mM S-mGSH showed that AMPPNP binding to MRP1 caused a substantial shift in the binding curve (Fig.  6B). This represents a significant decrease in the affinity for estrone sulfate when AMPPNP binds (K d 2.52 Ϯ 0.12 M) compared with the nucleotide-free state (K d 0.59 Ϯ 0.05 M). However, no change in the B max for estrone sulfate was observed. Similar results were obtained when GSH was used instead of S-mGSH, i.e. a significant increase in K d with no change in B max . These data are summarized in Table 3.
Effect of ATP Hydrolysis on Estrone Sulfate Binding-Having observed that binding of nucleotide caused a switch from high to low affinity for estrone sulfate binding, the next step was to examine the effect of ATP hydrolysis. The immediately post-hydrolysis state of MRP1 (ADP⅐P i ) can be mimicked by vanadate (V i )-induced trapping of ADP, where V i replaces P i to form the stable intermediate ADP⅐V i (43). The efficiency of ADP⅐V i trapping under the conditions used in this study was assayed by measuring the residual MRP1-mediated LTC 4 transport, and these data indicated that 88 Ϯ 2% of the MRP1 was in the ADP⅐V i -trapped state. As observed in the nucleotide-free and AMP-PNP-bound states, GSH or S-mGSH was required for specific binding of estrone sulfate to MRP1 in the ADP⅐V i -trapped state (data not shown). The effect of ADP⅐V i trapping on the binding of estrone sulfate in the presence of 3 mM S-mGSH is shown in Fig. 7. Thus, the K d for estrone sulfate in the ADP⅐V i -trapped state remained low affinity (K d 3.31 Ϯ 0.31 M) compared with the nucleotide-free conformation (K d 0.59 Ϯ 0.05 M) and similar to the AMPPNP-bound state. The B max for estrone sulfate (Table 3) also appeared to decrease moderately in the ADP⅐V itrapped state (B max 66 Ϯ 9 pmol/mg) compared with the nucleotidefree state (B max 87 Ϯ 7 pmol/mg), but this difference was not statistically significant ( p Ͼ 0.1). Similar results were observed during transition through the catalytic cycle when GSH was used rather than S-mGSH, and these results are summarized in Table 3.
Effect of ATP Binding and Hydrolysis on GSH Binding-The above experiments established that GSH or S-mGSH is required for binding of estrone sulfate to MRP1 in both the nucleotide-bound and the post-ATP hydrolysis states of the transporter. It was therefore of interest to know whether nucleotide binding and/or hydrolysis affects GSH binding itself. As detailed earlier, it is technically not possible to determine the affinity of MRP1 for GSH or S-mGSH directly. Consequently, the potency of S-mGSH to increase estrone sulfate binding was used as an indirect measure of the affinity of MRP1 for this GSH analogue. As shown in Fig. 8A, the level of S-mGSH-stimulated estrone sulfate binding was lower in the presence of AMPPNP (3500 dpm) than in the nucleotide-free state (7300 dpm). This is due to the decreased affinity for estrone sulfate when AMPPNP is bound. However, the EC 50 of S-mGSH in the presence of AMPPNP was 0.80 Ϯ 0.18 mM, which is comparable to the EC 50 in the nucleotide-free state (0.77 Ϯ 0.06 mM), suggesting the affinity for S-mGSH is not altered by nucleotide binding.

TABLE 3 Effect of nucleotide binding and hydrolysis on affinity and maximal binding capacity of MRP1 for estrone sulfate
Binding dissociation constant (K d ) and maximal binding capacity (B max ) for MRP1 binding of estrone sulfate in the presence of 3 mM S-mGSH or GSH were determined in the absence of nucleotide, in the presence of AMPPNP (4 mM), and in the ADP⅐V i -trapped state (4 mM ATP, 1 mM V i , 15 min at 37°C). Data are means Ϯ S.E. from at least three independent experiments. In contrast, in the ADP⅐V i -trapped state the EC 50 was significantly increased to 3.36 Ϯ 0.88 mM ( p Ͻ 0.001). These observations suggest the affinity of MRP1 for S-mGSH is decreased after ATP hydrolysis.

DISCUSSION
In the present study we have developed a radioligand binding assay to characterize directly the binding of estrone sulfate and GSH to MRP1. We have also used this assay to investigate how binding of these organic anions is affected by binding and hydrolysis of ATP. The data presented here, together with that in previous reports, are consistent with the model depicted in Fig. 9. In this model, GSH (or S-mGSH) binding to MRP1 causes a conformational change in MRP1 (in the COOH-terminal region of the protein) that opens up a high affinity binding site for estrone sulfate (Fig. 9, steps I-III). When ATP is bound to both NBDs, a conformation change in MRP1 reorients the estrone sulfate binding site to a low affinity state, presumably on the extracellular face of the membrane thus permitting release of the conjugated estrogen, while the GSH binding site is unaffected (Fig. 9, steps IV-VII). Hydrolysis of ATP at NBD2 then causes a decrease in the affinity of MRP1 for GSH (or S-mGSH) so that the tripeptide is released from its binding site on the intracellular side of the membrane (Fig. 9, step VIII). This model thus provides a mechanistic explanation for how GSH (or S-mGSH) can stimulate the transport of estrone sulfate by MRP1 and yet not be transported itself.
In the absence of GSH no specific binding of estrone sulfate to MRP1 was detected (Fig. 1); however, we have previously reported that MRP1mediated uptake of estrone sulfate into inside-out membrane vesicles could occur in the absence of GSH, albeit at a very low level (8,34). Whereas others have been unable to detect this (17), we also showed previously that estrone sulfate could inhibit LTC 4 binding and transport in the absence of GSH, although this was only significant at relatively high estrone sulfate concentrations (Ͼ20 M). Therefore, it seems that MRP1 is capable of binding estrone sulfate in the absence of GSH but with such a low affinity that it is beyond the detection level of the radioligand binding assay described here.
Our present findings also indicate that the low level binding of estrone sulfate by MRP1 in the absence of GSH has no detectable effect on the conformation of the transporter, because estrone sulfate alone did not affect its tryptic digestion profile. In contrast, GSH binding does induce a conformational change in MRP1, as suggested previously (41,44). The same change in the proteolytic sensitivity of the COOH-terminal region of MRP1 was induced by both GSH and its non-reducing analogue S-mGSH (Figs. 5 and 9, step II). We have also demonstrated before that both of these analogues are capable of stimulating estrone sulfate transport and, therefore, that the sulfhydryl group of GSH is not required (8). In fact the cysteine residue of GSH can be replaced with several other non-polar amino acids and still stimulate estrone sulfate transport, in some cases even better than GSH (34). The present ligand binding assays showed that the apparent affinity of MRP1 for both GSH and S-mGSH is the same (Fig. 2), as was previously suggested by transport studies (8). The notably different abilities of GSH and S-mGSH to stimulate estrone sulfate transport and binding are instead due to a difference in the affinity for estrone sulfate itself. Thus, estrone sulfate binds to MRP1 with a significantly higher affinity in the presence of S-mGSH than GSH ( Fig. 3 and Table 1). This could be because of a direct interaction between GSH and estrone sulfate in a common substrate binding pocket of MRP1 which is much stronger with the presence of the additional methyl group of S-mGSH, perhaps due to its larger size or hydrophobicity. Alternatively, GSH/S-mGSH may bind to a site distinct from that where estrone sulfate binds, and the conformational changes induced in MRP1 by S-mGSH and GSH may not be exactly the same. Although involving essentially the same regions of the protein (and hence not affecting trypsin sensitivity), the increased size and/or hydrophobicity of S-mGSH may cause a slightly larger movement or a variation in side-chain orientation that results in a higher affinity binding site for estrone sulfate. The fact that GSH or S-mGSH was required for measurement of estrone sulfate binding at each stage of the catalytic cycle tested, yet there is no evidence that GSH is co-transported with estrone sulfate, would argue against a direct interaction between the two ligands. For this reason, we favor the second explanation.
In the presence of the non-hydrolyzable AMPPNP, the affinity for estrone sulfate decreased significantly (Table 3). This observation may be explained by an ATP-dependent reorientation of the estrone sulfate binding site to its low affinity state, which is presumably accessible to the extracellular face of the membrane to facilitate release of estrone sulfate (Fig. 9, steps IV-VII). It has previously been reported that AMPPNP binding to MRP1 causes a substantial conformational change in the protein (44), as would be required for this substrate reorientation. We have also observed a conformational change in MRP1 induced by AMP-PNP as detected by trypsin sensitivity assays. 4 On the other hand, we recently reported that AMPPNP binding has no effect on the affinity of MRP1 for LTC 4 (42). The reasons for these apparently different responses to AMPPNP are not clear. In the previous study membranes from an insect cell expression system were used, in contrast to the human lung cancer cell membranes used here, and potentially the very different membrane composition of the two cell types may affect MRP1 and its interaction with AMPPNP. Alternatively, it may be that the changes induced by AMPPNP binding do not affect binding of all MRP1 substrates in the same way. However, for both LTC 4 and estrone sulfate, the slowly hydrolyzable ATP␥S reduced substrate binding, and, consequently, the evidence supports the conclusion that ATP binding is sufficient for reconfiguration of the substrate binding site. 4 A. Rothnie and S. P. C. Cole, unpublished observations. FIGURE 9. A model for GSH-stimulated transport of estrone sulfate by human MRP1. Shown is a schematic illustrating eight proposed steps of a cycle of GSH-stimulated estrone sulfate transport by MRP1. I, MRP1 in the absence of nucleotide; II, binding of GSH to MRP1 causes a conformational change, which opens up the high affinity binding site for estrone sulfate (ES); III, estrone sulfate binds; IV, ATP binds to NBD1, which promotes binding (V ) of ATP to NBD2; VI, when ATP is bound to both NBDs, the estrone sulfate binding site is reoriented to low affinity, at the extracellular face of the membrane, allowing for release of ES (VII ); VIII, hydrolysis of ATP at NBD2 causes a decrease in affinity for GSH, which is then released on the intracellular side of the membrane.
In our model shown in Fig. 9, we have suggested that the reorientation of the estrone sulfate binding site occurs only when ATP binds to both NBDs. The reasons for this are severalfold: (i) at the concentration used in this study, AMPPNP binds to both NBDs (42); (ii) it has been reported that ATP binding to NBD1 promotes the subsequent binding of ATP to NBD2 (45); (iii) mutations in NBD2 of MRP1, which increase the affinity for ATP, potentiate the reorientation of the LTC 4 binding site to low affinity (46); and (iv) the related ABCC protein, the cystic fibrosis transmembrane conductance regulator (ABCC7) like MRP1 has non-equivalent NBDs, and ATP binding at both NBDs of cystic fibrosis transmembrane conductance regulator is required for opening of this chloride channel (47,48). Thus, when both cystic fibrosis transmembrane conductance regulator NBDs bind ATP, a tight dimerization of the NBDs occurs, which leads to channel opening (49), and it has been proposed that such a tight dimerization of the NBDs may be a common feature of the translocation mechanism for all ABC proteins.
Unlike estrone sulfate, the apparent affinity of MRP1 for GSH/S-mGSH was not altered by the binding of AMPPNP (Fig. 8A). This suggests that, although the estrone sulfate binding site was reoriented to a low affinity state, the GSH binding site was not significantly altered ( Fig.  9, steps VI-VII). As previously mentioned, this supports the idea that GSH does not simply play a "space-filling" role in a common substrate binding pocket alongside estrone sulfate. Instead it seems more likely that GSH affects estrone sulfate binding by an allosteric mechanism. In addition, because MRP1 can also transport GSH alone or together with substrates such as vincristine (14), and assuming that nucleotides play a common mechanistic role in the transport of different substrates, then these findings could be construed to indicate the presence of two distinct GSH binding sites: one where GSH undergoes transport, and a second where GSH elicits allosteric effects on MRP1 without being translocated across the membrane. Such an idea has been proposed previously on the basis of mutagenesis experiments where the mutation of a single residue prevents GSH transport, yet GSH-stimulated estrone sulfate transport is only mildly affected (50,51), and vice versa (52), 5 but this remains to be directly proven.
Following hydrolysis of ATP, as mimicked experimentally by vanadate-induced trapping of ADP by MRP1, the estrone sulfate binding site on the transporter remained in a low affinity state (Fig. 7). However, at this stage of the catalytic cycle, unlike ATP binding, the apparent affinity of MRP1 for GSH/S-mGSH was also decreased (Fig. 8B). Because we can find no evidence that GSH is co-transported with estrone sulfate, it is quite conceivable that the GSH is released back on the cytoplasmic side of the membrane (Fig. 9, step VIII). However, even though estrone sulfate does not cause an associated increase in GSH transport in the manner that vincristine does (14), there is the formal possibility that the "basal" level of GSH transport (i.e. the rate of transport of GSH in the absence of other substrates) might be sufficient to support co-transport of estrone sulfate (8). However, the potency with which GSH mediates its effects on estrone sulfate binding and transport (EC 50 0.5-0.8 mM) is substantially higher than the estimated affinity for GSH undergoing basal transport (K m ϳ 4 mM (35)), suggesting that these are two quite different processes. In addition, release of GSH on the cytoplasmic side of the membrane as we suggest would further reduce the affinity for estrone sulfate at the external face, thus accelerating the dissociation of estrone sulfate and elevating the rate of transport.
Clearly the model presented in Fig. 9 is not complete, because the steps that result in the restoration of high affinity estrone sulfate binding after transport are not described. Thus far the only information available on this aspect of the MRP1 transport cycle has been provided by Yang et al. (53). In their study they showed that nucleotide dissociation from NBD1 is required for resetting of MRP1 and allowing a new cycle of transport to commence. However, it is not known if this is simply the release of the bound ATP, or whether the ATP must first be hydrolyzed, and it is the subsequent release of ADP that is required.
How the mechanism proposed here for GSH-stimulated transport of estrone sulfate relates to the GSH-dependent transport of other MRP1 substrates remains to be tested. However, it seems reasonable to postulate that substrates, such as the nicotine metabolite 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol-O-glucuronide (16) and glutathionylquinolineoxide (17), may utilize a similar mechanism. Like estrone sulfate, ATP-dependent vesicular transport of these molecules is undetectable in the absence of GSH. GSH stimulates their transport, yet these metabolites either inhibit or have no effect on the basal level of GSH transport by MRP1. It also remains to be determined how the mechanism of vincristine and GSH co-transport, and drug-stimulated GSH efflux, compare with the mechanism proposed here.