GSH-dependent Photolabeling of Multidrug Resistance Protein MRP1 (ABCC1) by [125I]LY475776

Substrates transported by the 190-kDa multidrug resistance protein 1 (MRP1) (ABCC1) include endogenous organic anions such as the cysteinyl leukotriene C4. In addition, MRP1 confers resistance against various anticancer drugs by reducing intracellular accumulation by co-export of drug with reduced GSH. We have examined the properties of LY475776, an intrinsically photoactivable MRP1-specific tricyclic isoxazole modulator that inhibits leukotriene C4 transport by this protein in a GSH-dependent manner. We show that [125I]LY475776 photolabeling of MRP1 requires GSH but is also supported by several non-reducing GSH derivatives and peptide analogs. Limited proteolysis revealed that [125I]LY475776 labeling was confined to the 75-kDa COOH-proximal half of MRP1. More extensive proteolysis generated two major125I-labeled fragments of ∼56 and ∼41 kDa, and immunoblotting with regionally directed antibodies showed that these fragments correspond to amino acids ∼1045–1531 and ∼1150–1531, respectively. However, an ∼33-kDa COOH-terminal immunoreactive fragment was not labeled, inferring that the major [125I]LY475776-labeling site resides approximately between amino acids 1150–1250. This region encompasses transmembrane (TM) segments 16 and 17 at the COOH-proximal end of the third membrane spanning domain of the protein. [125I]LY475776 labeling of mutant MRP1 molecules with substitutions of Trp1246 in TM17 were reduced >80% compared with wild-type MRP1, confirming that TM17 is important for LY475776 binding. Finally, vanadate-induced trapping of ADP inhibited [125I]LY475776 labeling, suggesting that ATP hydrolysis causes a conformational change in MRP1 that reduces the affinity of the protein for this inhibitor.

MRP1 1 (ABCC1) belongs to subfamily "C" of the ATP-binding cassette (ABC) superfamily of membrane transporters that is composed of ϳ50 human proteins (1)(2)(3)(4). Many ABC transport proteins, such as the drug efflux pump P-glycoprotein that belongs to ABC subfamily "B" (ABCB1), have a four-domain structure with two hydrophobic membrane spanning domains (MSDs), each of which is followed by a nucleotide binding domain (NBD). Most evidence indicates that each MSD contains six transmembrane (TM) segments, and both the NH 2 and COOH termini of the protein reside on the cytoplasmic side of the membrane. In contrast, MRP1 and several other members of subfamily C contain a third NH 2 -proximal MSD (MSD1) of ϳ200 amino acids with five TM segments and an extracytosolic NH 2 terminus (5-9). Thus, MRP1 and its related proteins MRP2 (ABCC2), MRP3 (ABCC3), MRP6 (ABBC6), MRP7 (ABCC10), and the sulfonylurea receptors, SUR1 (ABCC9) and SUR2 (ABCC8), are five-domain proteins with two NBDs and three MSDs containing up to 17 TM segments (6,7,10). The role that the NH 2 -proximal MSD1 plays in the function of these proteins is currently not well understood, but it does not appear to be essential for at least some transport functions of MRP1 (11). On the other hand, certain mutations or truncations of MSD1 and the intracellular loop connecting it to MSD2 can eliminate the transport activity or alter membrane trafficking of MRP1 (11)(12)(13). Thus MSD1 may participate in the regulation of some aspects of the biogenesis, structure, and/or function of the protein.
Recently, a screen of potential MRP1 inhibitors has identified a series of tricyclic isoxazoles as novel compounds that specifically and potently modulate MRP1-mediated drug resistance in vivo and LTC 4 transport in vitro (31). In the present study, we have investigated the photolabeling properties of LY475776, an intrinsically photoactivable iodinated azidotricyclic isoxazole analog. We show that inhibition of LTC 4 transport and photolabeling of MRP1 by [ 125 I]LY475776 are dependent on GSH, although photolabeling is also supported by certain non-reducing GSH derivatives. Mapping of tryptic fragments of [ 125 I]LY475776-labeled membrane proteins revealed that this compound bound exclusively to the COOH-proximal half of MRP1 with a major labeling site in a region containing the last two TM segments (TM16 and TM17) in MSD3. Reduced labeling of mutant MRP1 proteins with substitutions of Trp 1246 in TM17 confirmed the importance of this region for LY475776 binding. Finally, GSH-dependent [ 125 I]LY475776 labeling of MRP1 was significantly reduced after vanadate-induced trapping of ADP, suggesting that hydrolysis of ATP causes a conformational change in MSD3 that reduces the affinity of MRP1 for this reversing agent.
The six MRP1-specific murine mAbs used in this study have been described previously (32,33). mAb QCRL-1 has been mapped to a linear epitope corresponding to amino acids 918 -924 located in the cytoplasmic linker region of the protein that connects NBD1 to MSD3, whereas MRPm6 maps to amino acids 1511-1520 in the COOH terminus of the protein (34,35). mAb MRPm5 also recognizes a linear epitope that lies within a region of MSD3 between amino acids 986 and 1096, but because the epitope is accessible only in permeabilized cells, it likely is within the predicted intracellular loop (ICL6) between TM13 and TM14 (approximately amino acids 1040 -1096) (6,36). mAbs QCRL-2, -3, and -4 recognize conformation-dependent intracellular epitopes that map to NBD1 (mAbs QCRL-2 and -3) and NBD2 (mAb QCRL-4) (36). A schematic diagram illustrating the domain organization of MRP1 and approximate location of the epitopes recognized by these mAbs is shown in Fig. 1. The schematic diagram shows a proposed topology of MRP1, its domain organization, and the approximate location of its 17 predicted transmembrane segments (TM1-17). Also shown are the known N-glycosylation sites at Asn 19 , Asn 23 , and Asn 1006 , the locations of the epitopes recognized by the mAbs used in this study (QCRL-1, -2, -3, -4, MRPm5 and MRPm6), and the protease-hypersensitive linker region of MRP1 connecting the NH 2 -proximal ("N1") and COOH-proximal ("C1") halves of the protein. The approximate assignments of the tryptic fragments described in the text are indicated as is the chemical structure of [ 125 I]LY475776.
Cell Culture and Membrane Protein Preparation-The doxorubicinselected, multidrug-resistant H69AR small cell lung cancer cell line that expresses high levels of MRP1, and the transfected HeLa cell lines that express recombinant wild-type MRP1 (T5 or WT-MRP1), and mutant MRP1 bearing substitutions of Trp 1246 (W1246F-MRP1, W1246Y-MRP1, W1245C-MRP1, and W1246A-MRP1) were maintained as described previously (15,37). The preparation of unglycosylated "sugarfree" N19/23/1006Q-MRP1 (SF-MRP1) is described elsewhere (6). The LTC 4 transport and drug resistance conferring activities of SF-MRP1 are comparable with those of wild-type MRP1. Crude membranes or membrane vesicles enriched in wild-type, mutant MRP1, and SF-MRP1 were prepared from H69AR cells or transfected HeLa cells as described (15) and stored at Ϫ70°C. LTC 4 Transport Assay-Uptake of LTC 4 into inside-out membrane vesicles prepared from T5 HeLa cells was measured as described previously (38). Briefly, membrane vesicles were incubated with [ 3 H]LTC 4 (50 nM) and 4 mM ATP or AMP-PNP at 37°C for 45 s. Various concentrations of LY475776 were added in the presence or absence of 1 mM GSH. ATP-dependent uptake was calculated by subtraction of uptake measured in the presence of AMP-PNP from that measured in the presence of ATP and expressed as a percentage of control.
Photolabeling of MRP1 with [ 125 I]LY475776 -Crude membranes (150 g of protein) or membrane vesicles (50 -100 g of protein) were gently mixed with [ 125 I]LY475776 (110 -300 nCi) and GSH (1 mM) in a final volume of 30 l and incubated for 5 min at 37°C and then irradiated at 254 nm for 5 min. Proteins were immediately solubilized in sample buffer and subjected to SDS-PAGE using 7.5 or 15% acrylamide gels or 3-12% gradient gels. Gels were dried and exposed to film, and autoradiograms were developed after overnight exposure at Ϫ70°C. Alternatively, proteins were transferred to a nylon membrane by electroblotting and then the membrane was exposed to film overnight at Ϫ70°C. Relative levels of photolabeling were determined by densitometric analysis of the autoradiograms with a ChemiImager (Alpha Innotech) and ChemiImager 4000 software.
In some labeling experiments, GSH analogs or derivatives were added instead of GSH at a concentration of 1 mM. To investigate the effect of various mAbs on photolabeling, purified mAbs QCRL-1, -2, -3, -4 were added to the photolabeling mixture at a concentration of 100 g/ml and incubated for 1 h at room temperature prior to addition of GSH and [ 125 I]LY475776 and UV irradiation. In other experiments, drugs and bioflavonoids were added 30 min prior to cross-linking at the concentrations indicated in the text and figure legends. To assess the effect of sodium orthovanadate-induced trapping of ADP, membrane proteins were incubated with [ 125 I]LY475776 (300 nCi), 1 mM GSH, 5 mM MgCl 2 , and various concentrations of orthovanadate in the presence or absence of 2.5 mM ATP for 15 min at 37°C prior to UV cross-linking. Experiments were also carried out in which the non-hydrolyzable analog ATP␥S (0.5 and 2.0 mM) or the covalent modifiers NEM (2 mM) and NBD-Cl (200 M) were incubated with membranes for 30 min at 37°C prior to labeling with [ 125 I]LY475776 in the presence of GSH.
Proteolysis of Membrane Proteins and Immunoblotting-Crude membranes (150 g of protein for H69AR cell membranes and 900 g of protein for SF-MRP1 HeLa cell membranes) or [ 125 I]LY475776-labeled H69AR cell membranes (150 g of protein) were mixed with diphenylcarbamyl chloride-treated trypsin at trypsin-to-protein ratios ranging from 1:800 to 10:1 in phosphate-buffered saline in a final volume of 60 l. After incubating the mixtures at 37°C for 30 min, reactions were stopped by the addition of phenylmethylsulfonyl fluoride and leupeptin as described (6). After 15 min at room temperature, samples were solubilized in SDS sample buffer and subjected to SDS-PAGE as well as immunoblotting with enhanced chemiluminescence detection as described (6). For immunoblotting, mAbs QCRL-1, MRPm5 and MRPm6 were used at dilutions of 1:10,000, 1:100, and 1:2,000, respectively.
Electrophoresis was performed using SDS-polyacrylamide minigels (7.3 ϫ 10.2 cm) in a Bio-Rad Mini Protein II electrophoresis cell or 16 ϫ 18-cm gels in a vertical slab gel electrophoresis unit (Hoefer Scientific Instruments, San Francisco, CA). Protein concentrations in membrane preparations were determined by the Bio-Rad Protein Assay with bovine serum albumin as standard. 4 Uptake by LY475776 -The effect of LY475776 on MRP1-mediated uptake of [ 3 H]LTC 4 into inside-out membrane vesicles was measured in the absence and presence of 1 mM GSH. In the absence of GSH, LY475776 had no detectable effect on LTC 4 uptake up to the highest concentration tested (0.8 M) (Fig. 2). However, in the presence of GSH, LY475776 caused a potent, concentrationdependent inhibition of LTC 4 transport. Based on results from three independent experiments, the EC 50 value for LY475776 was determined to be 0.049 Ϯ 0.010 M. Thus LY475776 is an extremely potent inhibitor of MRP1-mediated LTC 4 uptake but only in the presence of reduced GSH.

GSH-dependent Inhibition of [ 3 H]LTC
Effect of GSH Derivatives and Analogs on Photolabeling of MRP1 with [ 125 I]LY475776 -To investigate further the role of GSH in the interaction of LY475776 with MRP1, the molecule was radioiodinated and determined to photolabel specifically the MRP1 in H69AR membranes but only in the presence of reduced GSH (Fig. 3). S-Alkyl derivatives of GSH also stimulated labeling of MRP1 by [ 125 I]LY475776, but their ability to do so diminished when the length of the S-alkyl substituent exceeded three carbon atoms. Thus, S-methyl-GSH, S-ethyl-GSH, and S-propyl-GSH supported [ 125 I]LY475776 labeling but S-hexyl-GSH and S-decyl-GSH did not. The GSH analog, ophthalmic acid, in which the Cys side chain contains a methyl group instead of a sulfhydryl group, stimulated [ 125 I]LY475776 labeling of MRP1 as well as GSH. Similarly, when the glycine residue of GSH was replaced by ␤-alanine (␥Glu-Cys-␤Ala; homo-GSH), a moderate degree of labeling was observed. On the other hand, [ 125 I]LY475776 labeling in the presence of analogs in which the ␥-glutamate or the cysteine residue of GSH had been replaced with glycine (Gly-Cys-Gly or ␥Glu-Gly-Gly) was very low and comparable with photolabeling in the absence of GSH. (22)(23)(24)(25)(26)36) have shown that the conformation-dependent MRP1-specific mAbs QCRL-2, -3, and -4 can inhibit the MRP1-mediated transport of numerous substrates including LTC 4 and vincristine in the presence of GSH. In addition, we have shown that one of these mAbs, QCRL-3, which detects an epitope contained within NBD1 of MRP1, can inhibit photolabeling of the protein by [ 3 H]LTC 4 . In contrast, mAb QCRL-2, which also maps to NBD1, and mAb QCRL-4, which maps to NBD2, do not (15,36).

MRP1-specific mAb QCRL-3 Partially Inhibits Labeling of MRP1 by [ 125 I]LY475776 -Previously, we and others
To determine the effect of these antibodies on binding of LY475776 to MRP1, H69AR membranes were preincubated with mAbs QCRL-2, -3, and -4 (100 g/ml) followed by photolabeling with [ 125 I]LY475776 in the presence of 1 mM GSH. As a negative control, membranes were preincubated with mAb QCRL-1 which detects a linear epitope in the cytoplasmic region of MRP1 that links NBD1 to MSD3 and does not inhibit transport activity (34,36). The results shown in Fig. 4A indicate that of the four MRP1-specific mAbs, only mAb QCRL-3 inhibited the photolabeling of MRP1 by [ 125 I]LY475776 and the effect was modest (ϳ30%).

GSH-stimulated Photolabeling by [ 125 I]LY475776 Is Attenuated by Compounds That Stimulate GSH Transport by
MRP1-It has been demonstrated previously by our group and others (19, 22, 29, 39 -44) that MRP1 is a low affinity transporter of GSH and that transport of this tripeptide is dramatically increased in the presence of certain xenobiotics and bioflavonoids. To determine whether compounds that stimulate MRP1-mediated GSH binding and transport also affect GSHdependent interaction of [ 125 I]LY475776 with MRP1, photolabeling was carried out with membrane proteins that had been preincubated for 30 min with verapamil (100 M), vincristine (60 M), quercetin (50 M), naringenin (50 M), or apigenin (50 M). As shown in Fig. 4B, verapamil inhibited labeling of MRP1 by ϳ80%, whereas vincristine inhibited labeling by 60%. GSHstimulated [ 125 I]LY475776 labeling of MRP1 was also inhibited by the three flavonoids but to different degrees; thus photolabeling was inhibited 80% by apigenin, 60% by quercetin, and 35% by naringenin.
[ 125 I]LY475776 Photolabels the COOH-proximal Half But Not the NH 2 -proximal Half of MRP1-We and others have shown previously (6,34,45,46) that limited trypsin digestion of membrane proteins from mammalian cells expressing MRP1 generates two N-glycosylated fragments of ϳ110 and 75 kDa. By using antibodies against defined epitopes in different regions of the protein, these two tryptic fragments were shown to correspond to the NH 2 -proximal (designated N1) and COOHproximal (designated C1) halves of MRP1, respectively (Fig. 1). Thus the cytoplasmic region of MRP1 that links NBD1 to MSD3 is hypersensitive to cleavage by proteases (6,34). To begin to identify the regions involved in LY475776 binding to MRP1, H69AR membranes were treated with various concentrations of trypsin under mild conditions and then photoaffinity-labeled with [ 125 I]LY475776 in the presence of 1 mM GSH and processed as before. As shown in Fig. 5, increasing concentrations of trypsin resulted in the appearance of increasing amounts of a single 125 I-labeled fragment of ϳ75 kDa. The size of this tryptic fragment indicated that it probably corresponded to the COOH-proximal half of MRP1 (C1). In contrast, there was no evidence of binding to any larger tryptic fragment that would correspond to the NH 2 -proximal half of MRP1.
To verify that [ 125 I]LY475776 photolabeled the ϳ75-kDa COOH-proximal half of MRP1, the trypsin-digested 125 I-labeled membrane preparations were transferred to a nylon membrane after SDS-PAGE which was then subjected to both autoradiography and immunoblotting with mAbs QCRL-1 and MRPm6. mAb QCRL-1 has been mapped to amino acids 918 -924 and under certain mild proteolysis conditions, it can detect H69AR crude membrane proteins (150 g) were incubated with 3.5 nM [ 125 I]LY475776 (220 nCi) and 1 mM GSH or GSH derivative or analog for 5 min before UV cross-linking at 254 nm for 5 min on ice. The proteins were then resolved by SDS-PAGE on a 15% vertical slab gel, and the gel was dried and exposed to film at Ϫ70°C overnight.  Fig. 2 and then resolved by SDS-PAGE on a 3-12% gradient gel. The gel was dried and then exposed to film at Ϫ70°C overnight.
both the NH 2 -proximal (ϳ110 kDa; N1) and COOH-proximal (ϳ75 kDa; C1) halves of MRP1 because there are several enzyme cleavage sites on both sides of its epitope (Fig. 6A, right panel) (6,34). However, only the smaller of the two QCRL-1-immunoreactive tryptic fragments was labeled by [ 125 I]LY-475776 (Fig. 6A, left panel). On the other hand, mAb MRPm6, which has been mapped to amino acids 1511-1520 at the extreme COOH terminus of the molecule (34), detected only a single tryptic fragment of ϳ75 kDa as expected under conditions of mild proteolysis (Fig. 6B, right panel). The electrophoretic mobility of this MRPm6-immunoreactive fragment corresponded exactly to that of the single 75-kDa 125 I-labeled tryptic fragment of MRP1 (Fig. 6B, left panel). Moreover, treatment of the tryptic fragments with peptide N-glycosidase F showed that, as expected, the 75-kDa fragment was glycosylated (data not shown). Taken together, these observations confirm that the [ 125 I]LY475776-labeled 75-kDa tryptic fragment corresponds to the COOH-proximal half of MRP1 (C1).
Proteolysis of [ 125 I]LY475776-labeled MRP1 Reveals a Major Substrate-binding Site in MSD3-To localize further the regions of MRP1 involved in GSH-stimulated binding of [ 125 I]LY475776, photolabeled H69AR membranes were subjected to more extensive proteolysis at increasing trypsin-toprotein ratios, and the resulting fragments were resolved on polyacrylamide gels. Fig. 7A shows that, in addition to the ϳ75-kDa C1 fragment corresponding to the COOH-proximal half of MRP1, two major (ϳ56 and ϳ41-kDa) and two minor (ϳ25 and ϳ15-kDa) 125 I-labeled tryptic fragments were detected under these conditions. Labeling of the larger fragments became more evident after transfer to a nylon membrane followed by autoradiography that reduced some of the nonspecific background (Fig. 7B).
Immunoblot analysis of trypsin-digested H69AR membrane proteins with mAb MRPm6 identified two fragments with exactly the same electrophoretic mobility as the strongly 125 Ilabeled 56-and 41-kDa tryptic fragments (Fig. 7C, filled arrowheads). Because the MRPm6 epitope is located at the COOH-terminal end of MRP1 (amino acids 1511-1520) (35), it follows that the 56-and 41-kDa fragments must be derived from the 75-kDa tryptic fragment and, furthermore, based on their estimated molecular masses, that the 56-kDa tryptic fragment corresponds to a region from approximately amino acids 1045 to 1531, and the 41-kDa tryptic fragment to a region from approximately amino acids 1150 to 1531. Consistent with these assignments, the 56-and 41-kDa fragments were not N-glycosylated (data not shown) as expected, based on the estimation that their NH 2 termini would be downstream of the only known glycosylation site in the COOH-proximal half of MRP1 at Asn 1006 (6). On the other hand, mAb MRPm6 reacted with several additional tryptic fragments that were not labeled by [ 125 I]LY475776 (Fig. 7C, open arrowheads). The largest of these was ϳ33 kDa, and because the COOH terminus of MRP1 must be intact in order for this fragment to be detected by mAb MRPm6, it can be deduced that its NH 2 terminus is located at approximately amino acid 1250. This is consistent with trypsin-mediated cleavage occurring in the cytoplasm in close proximity to COOH-terminal end of TM17. This in turn infers that the non-photolabeled MRPm6-reactive 33-kDa tryptic fragment corresponds to the cytosolic COOH-proximal region of MRP1 that includes NBD2. Taken together, these data indicate that the major [ 125 I]LY475776-labeling site in the COOH-proximal half of MRP1 is located between approximately amino acids 1150 and 1250. This region corresponds to a region of MSD3 that includes part of the intracellular loop connecting TM15 to TM16 (ICL7), the last two predicted TM helices (TM16 and TM17), and the extracellular loop connecting TM16 to TM17 (Fig. 1).
The presence of two smaller tryptic fragments (ϳ25 and ϳ15 kDa) that were weakly labeled with [ 125 I]LY475776 (Fig. 7, A  and B) suggests that other, more NH 2 -proximal regions of MSD3 may also be involved in binding of LY475776, albeit with a lower efficiency than the region containing TM16 and TM17. Immunoreactive tryptic fragments of comparable sizes to these smaller weakly 125 I-labeled fragments were detected by immunoblot analysis of H69AR membrane digests with mAbs QCRL-1 and MRPm5, respectively (Fig. 7, D and E, filled  arrowheads). Additional immunoblots of tryptic digests of un-glycosylated SF-MRP1 HeLa cell membranes further showed that the electrophoretic mobility of the ϳ25-kDa fragment of wild-type MRP1 migrated as an unglycosylated fragment of ϳ18 kDa, whereas the mobility of the 15-kDa fragment was unchanged, indicating that the former but not the latter tryptic fragment was glycosylated (not shown). Thus, the estimated molecular masses of these tryptic fragments and their glycosylation status, together with the location of the epitopes recognized by mAbs QCRL-1 and MRPm5, suggest that the two weakly 125 I-labeled tryptic fragments of ϳ25 and ϳ15 kDa could correspond to approximately amino acids 860 -1020 (includes TM12 and TM13) and 1020 -1150 (includes TM14 and TM15), respectively. However, conclusive identification of

FIG. 7. Trypsin digestion profiles of MRP1 in H69AR membrane proteins after GSH-stimulated photolabeling with [ 125 I]LY475776.
A-C, MRP1-enriched crude membranes from H69AR cells (150 g of protein) were incubated with 3.5 nM [ 125 I]LY475776 (220 nCi) for 5 min at 37°C in the presence of 1 mM GSH, followed by digestion with trypsin, at different trypsin:protein ratios (range 1:500 to 10:1 (w/w)) at 37°C for 30 min. The reactions were stopped by addition of protease inhibitors, and the tryptic fragments were separated on a 15% polyacrylamide vertical slab gel (A) or transferred to a nylon membrane (B) and exposed to film at Ϫ70°C overnight. The membrane was also probed with mAb MRPm6 (diluted 1:2000), and MRP1 and its fragments containing the MRPm6 epitope (amino acids 1511-1520) were detected by enhanced chemiluminescence (C). D and E, H69AR membranes (150 g of protein) were incubated with trypsin at different trypsin:protein ratios ranging from 1:500 to 5:1 at 37°C for 15 min and processed for immunoblotting. MRP1 and its tryptic fragments containing either the QCRL-1 epitope (amino acids 918 -924) (D) or the MRPm5 epitope (between amino acids 1040 -1096) (E) were detected with mAb QCRL-1 (diluted 1:10,000) or with mAb MRPm5 (diluted 1:100), respectively. these weakly 125 I-labeled fragments will require further investigation.
Mutation in the presence of GSH as before (Fig. 8). The results show that mutation of Trp 1246 significantly reduces photolabeling of MRP1 (Fig. 8A). After normalizing expression levels of the mutant MRP1 molecules relative to wild-type MRP1 (Fig. 8B), it was estimated that [ 125 I]LY475776 labeling of HeLa cell membrane proteins containing the W1246F-MRP1 mutant was ϳ23% of wild-type MRP1 membrane proteins, whereas labeling of membranes containing the W1246Y-MRP1, W1246C-MRP1, and W1246A-MRP1 mutants was less than 10% of wild-type MRP1 levels.
Orthovanadate-induced Trapping of ADP Inhibits Photolabeling of MRP1 by [ 125 I]LY475776 -Sodium orthovanadate inhibits ATPase activity by stable trapping of the MgADP⅐V i species in place of MgADP⅐P i , thereby locking the protein in a catalytic transition state intermediate (48,49). Previous studies (50,51) have shown that binding of several substrates of P-glycoprotein is attenuated in the presence of ATP and orthovanadate, and it has been suggested that this reflects a reduced affinity of the protein in a MgADP⅐P i transition state which facilitates the release of bound substrates. To determine whether this also held true for LY475776 binding to MRP1, we examined the effect of orthovanadate-induced trapping of ADP on [ 125 I]LY475776 photolabeling of intact MRP1 as well as its proteolytic fragments. Fig. 9A shows that, under conditions used for vanadate trapping but in the absence of the metal oxyanion, no inhibition or stimulation of photolabeling by ATP and MgCl 2 was evident. In the absence of ATP, GSH-stimulated [ 125 I]LY475776 labeling was unaffected by sodium orthovanadate, indicating that vanadate itself had no direct effect on binding of this ligand to MRP1. However, in the presence of 2.5 mM ATP, labeling of intact MRP1 was significantly decreased by sodium orthovanadate in a concentration-dependent manner. In addition, the decrease in [ 125 I]LY475776 labeling of intact MRP1 occurred in parallel with a decrease in labeling of the 75-kDa COOHproximal half of the protein. These results are consistent with the idea that trapping of an MgADP⅐V i complex results in a structural change in MRP1 that significantly attenuates [ 125 I]LY475776 binding.

Effect of ATP␥S and Sulfhydryl-modifying Agents on GSHstimulated Photolabeling of MRP1 by [ 125 I]LY475776 -When
labeling was carried out in the presence of the poorly hydrolyzable nucleotide, ATP␥S, at a concentration of either 0.5 or 2 mM, a modest decrease (ϳ35%) in labeling was observed (Fig.  9B). GSH-stimulated [ 125 I]LY475776 labeling of MRP1 was also completely eliminated by the covalent inhibitors NEM (2 mM) and NBD-Cl (200 M) (Fig. 9B). These sulfhydryl-modifying compounds have been shown previously (52,53) to inhibit the ATPase activity of purified reconstituted MRP1 as well as its transport of conjugated organic anions. The attenuation of photolabeling caused by NEM and NBD-Cl may result from covalent modification of a critical cysteine residue in MRP1. Alternatively, or additionally, these inhibitors may reduce the levels of free GSH such that the efficiency of [ 125 I]LY475776 labeling of MRP1 was greatly diminished.

DISCUSSION
There is compelling evidence that the mechanisms by which MRP1 and P-glycoprotein bind and transport their substrates differ, and one of the most remarkable differences is that MRP1-mediated transport of unmodified natural product drugs requires the presence of physiological concentrations of GSH, whereas transport of the same substrates by P-glycoprotein does not. Earlier studies (15,22,25) have shown that it is some property other than its capacity to reduce sulfhydryl groups that is responsible for the effect of GSH on MRP1 transport activity. Because the transport inhibiting activity of LY475776 as well as its ability to label MRP1 is GSH-dependent, we explored further the structural features required in GSH for the binding of this intrinsically photoactivable reversing agent to the protein. We found that a free thiol group was not required to stimulate LY475776 labeling of MRP1, and that increasing the length of the S-alkyl side chain in GSH derivatives diminished binding to MRP1, results that are consistent with those reported previously (22) for GSH dependent transport of vincristine. On the other hand, because labeling in the presence of ␥Glu-Gly-Gly was very low, some space-filling substituent on the amino acid in the Cys position appears important for GSH interaction with MRP1. Thus some limits may be placed on the minimum and maximum size of the side chain of the central amino acid in GSH that can be accommodated by the "GSH"-binding site on MRP1. Additional features important for stimulating LY475776 binding appear to be contributed by the COOH-terminal Gly and the NH 2 -terminal ␥Glu residues of GSH because ␥Glu-Cys-␤Ala did not support labeling as well as GSH and Gly-Cys-Gly did not support labeling at all.
It has been demonstrated previously (15, 19, 22, 25, 26, 39 -43, 54) that MRP1 is a low affinity transporter of GSH and that transport of this tripeptide by MRP1 is dramatically increased in the presence of a variety of xenobiotics as well as certain bioflavonoids. It has been proposed that these compounds stimulate GSH transport by enhancing release of the tripeptide from MRP1 once it has been translocated across the membrane. In this way, these compounds might reduce binding of other GSH-dependent substrates or modulators to MRP1. We found that the ability of verapamil to inhibit GSH-dependent [ 125 I]LY475776 labeling of MRP1 is significantly greater than that of vincristine, which is in keeping with the relative ability of these two drugs to stimulate GSH transport (22,41,43). Similarly, of the three bioflavonoids tested, apigenin was the most effective inhibitor of [ 125 I]LY475776 labeling, and it is also the most potent stimulator of GSH transport (41,44). Naringenin is considerably less effective than apigenin in all aspects of modulating MRP1 function (41) including inhibiting labeling by [ 125 I]LY475776, which is remarkable given that the chemical structures of naringenin and apigenin differ by only a single CϭC double bond, which nevertheless enhances the lipophilicity of the latter bioflavonoid quite significantly (41).
Binding of the high affinity substrate LTC 4 to MRP1 has been demonstrated directly by protein labeling experiments with this intrinsically photoactivable arachidonic acid derivative (14,15,55). More recently, we have shown (45) that [ 3 H]LTC 4 photolabels sites in both the NH 2 -and COOH-proximal halves of the protein with an apparently greater efficiency for the NH 2 -proximal site. Both halves of MRP1 are also labeled with the photoactive 125 I-labeled quinoline-based azido derivative, IACI, as well as the structurally dissimilar iodoaryl azidorhodamine 123 (IAARh123), although in both cases, labeling of the two halves appears equivalent (46,56,57). In contrast, not only does photolabeling of MRP1 by [ 125 I]LY475776 occur only in the COOH-proximal half of the protein, it also requires the presence of GSH. However, this does not necessarily exclude the participation of the NH 2 -proximal half of MRP1 in forming the binding pocket for LY475776 because, by using a baculovirus system to co-express both halves of the protein, we have found that the NH 2 -proximal half of MRP1 is necessary for efficient labeling by [ 125 I]LY475776. 2 Whether this interaction is one between a GSH-binding site in the NH 2proximal half of MRP1 and a binding site(s) for the chemosensitizer in the COOH-proximal half, or between other parts in the two halves of MRP1, is not yet known.
In contrast to LTC 4 , IAARh123, and IACI, the labeling properties of [ 125 I]LY475776 display some similarities with those of the 125 I-labeled azido analog of Agosterol-A, a marine sponge sterol that restores drug sensitivity in MRP1-overexpressing cells (58,59). The radioiodinated azido derivative of Agosterol-A has also been shown to label exclusively the COOHproximal half of MRP1 in a GSH-dependent manner and, like LY475776, requires the presence of the NH 2 -proximal half of the protein (59). It has been proposed that the COOH-proximal site of MRP1 labeled by 125 I-azido Agosterol-A has a high affinity for this drug and that GSH is binding to a different site in the NH 2 -proximal half of the protein. However, direct evidence localizing the GSH-binding site(s) in MRP1 is still lacking.
Like the results reported here for [ 125 I]LY475776, the findings of Ren et al. (59) are consistent with the involvement of TM16 and/or TM17 in binding of Agosterol-A, although the involvement of NBD2 in binding of the sterol analog was not formally excluded. Nevertheless, the binding sites for LY475776 and Agosterol-A are highly unlikely to be exactly the same because these two chemosensitizers display such differences in their ABC transporter specificities. Thus, LY475776 is highly specific for MRP1 and is not recognized by either Pglycoprotein or even the MRP1 homologs MRP2 (ABCC2) and MRP3 (ABCC3). 3 In contrast, the photoaffinity analog of Agosterol-A is also capable of labeling P-glycoprotein even though this transporter shares less than 15% amino acid sequence identity with MRP1 (4,59). Similarly, the photoactive analog IAARh123 also binds to P-glycoprotein (56).
The localization of the major [ 125 I]LY475776 labeling site to the COOH-proximal end of MSD3 is of interest in view of previous mutagenesis studies that have shown that both con-served and non-conserved amino acids in the highly amphipathic TM17 that spans approximately from Ala 1227 to Val 1248 are important determinants of MRP1 substrate specificity (37,60). Thus the diminished ability of the MRP1 Trp 1246 mutants to be photolabeled by [ 125 I]LY475776 is consistent with the involvement of TM17 in the action of tricyclic isoxazole modulators that block transport of MRP1 substrates (31). Although it is clear that the major [ 125 I]LY475776 labeling site is localized to a region in and around TM16 and TM17, fragments of ϳ25 and ϳ15 kDa corresponding approximately to TM12-13 and TM14 -15 of MSD3, respectively, also appeared weakly photolabeled by this ligand. Whether these TM segments all come together to form a single binding site for LY475776 or whether they represent two independent drug-interaction domains requires further investigation.
Sodium orthovanadate inhibits the ATPase activity of Pglycoprotein, MRP1, and other ATPases, and this has been attributed to the stable trapping of the MgADP⅐V i species in place of MgADP⅐P i , thereby locking the protein in a catalytic transition state intermediate (48,49,61). Under these conditions, it has been demonstrated that binding of photoaffinity analogs of several of its substrates to P-glycoprotein is reduced, suggesting that ATP hydrolysis is essential for vanadate-induced inhibition of substrate binding (49 -51, 62). Trapping of ADP also dramatically reduced [ 125 I]LY475776 labeling of the COOH-proximal half of MRP1. This suggests that ATP hydrolysis induces a major conformational change in MRP1, likely through repacking of the TM segments including TM17, that reduces the affinity of the LY475776-binding site in the COOHterminal half of the protein. These findings are consistent with the previously described (63) alternating site model put forward for the coupling of substrate transport with ATP hydrolysis by the homodimeric bacterial drug transporter LmrA and mammalian P-glycoprotein. This model proposes that the alternate trapping of ADP at each NBD reduces or eliminates substrate binding by the same subunit of LmrA (or half of P-glycoprotein) where the trapping has occurred and that the protein in the MgADP⅐P i transition state of the catalytic cycle has a reduced binding affinity for substrates to ensure the release of bound substrates after translocation across the membrane. Recently, however, we observed that trapping of ADP preferentially attenuated [ 3 H]LTC 4 labeling of the NH 2 -proximal half of MRP1 with little or no effect on photolabeling of the COOH-proximal half of the protein (45). In contrast, ADP trapping by MRP1 clearly attenuates [ 125 I]LY475776 labeling of the COOH-half of the protein. The basis for these apparently different effects of vanadate-induced trapping on photolabeling of MRP1 domains by [ 3 H]LTC 4 and [ 125 I]LY475776 is unclear but may simply reflect differences in the preferred binding sites for these molecules. Alternatively, vanadate-induced trapping of ADP could reduce binding of GSH to a site in the NH 2 -proximal half of the protein which, becauseLY475776 labeling of MRP1 requires GSH, would in turn reduce [ 125 I]LY475776 labeling of the COOH-proximal half of the protein. Finally, the reduction in LY475776 binding to MRP1 in the presence of the poorly hydrolyzable ATP␥S is reminiscent of the moderate reduction in vinblastine binding to P-glycoprotein observed under the same conditions (51). However, it differs from the lack of effect of this nucleotide analog on [ 3 H]LTC 4 binding to MRP1 (45). Martin et al. (51) proposed that binding of nucleotide, rather than hydrolysis, causes an initial conformational shift in the vinblastine-binding site of P-glycoprotein during a transport cycle. Our data suggest that nucleotide binding may cause a similar shift in the LY475776-binding site of MRP1.
In summary, we have characterized the GSH dependence of the potent MRP1-specific reversing agent LY475776 and iden-tified a potential binding site(s) for this azido tricyclic isoxazole derivative located in MSD3 of MRP1 that is disrupted by substitution of Trp 1246 at the membrane-cytosol interface of TM17. We have also presented evidence that ATP hydrolysis, as well as possibly binding of nucleotide, can reduce the affinity of this site(s) for LY475776. Knowledge gained from these and other similar studies will be useful for understanding the molecular mechanism of GSH-dependent and independent inhibition of MRP1-mediated drug resistance.