Determinants of the Substrate Specificity of Multidrug Resistance Protein 1

Human multidrug resistance protein 1 (MRP1) confers resistance to many natural product chemotherapeutic agents and actively transports structurally diverse organic anion conjugates. We previously demonstrated that two hydrogen-bonding amino acid residues in the predicted transmembrane 17 (TM17) of MRP1, Thr1242 and Trp1246, were important for drug resistance and 17β-estradiol 17-(β-d-glucuronide) (E217βG) transport. To determine whether other residues with hydrogen bonding potential within TM17 influence substrate specificity, we replaced Ser1233, Ser1235, Ser1237, Gln1239, Thr1241, and Asn1245 with Ala and Tyr1236 and Tyr1243 with Phe. Mutations S1233A, S1235A, S1237A, and Q1239A had no effect on any substrate tested. In contrast, mutations Y1236F and T1241A decreased resistance to vincristine but not to VP-16, doxorubicin, and epirubicin. Mutation Y1243F reduced resistance to all drugs tested by 2–3-fold. Replacement of Asn1245 with Ala also decreased resistance to VP-16, doxorubicin, and epirubicin but increased resistance to vincristine. This mutation also decreased E217βG transport ∼5-fold. Only mutation Y1243F altered the ability of MRP1 to transport both leukotriene 4 and E217βG. Together with our previous results, these findings suggest that residues with side chain hydrogen bonding potential, clustered in the cytoplasmic half of TM17, participate in the formation of a substrate binding site.

A major limitation to the successful chemotherapeutic treatment of human tumors is the appearance of multidrug resistance (MDR). 1

MDR in certain tumors in vivo and cultured cell
lines in vitro is characterized by cross-resistance to structurally and functionally dissimilar compounds and is most frequently associated with the overexpression of multidrug resistance protein 1 (MRP1) and/or P-glycoprotein (P-gp) (1)(2)(3)(4)(5). Transfection experiments in mammalian cells have shown that MRP1, like P-gp, can confer resistance to many commonly used, structurally diverse natural product chemotherapeutic agents including anthracyclines, Vinca alkaloids, and epipodophyllotoxins (6 -8). In addition, unlike P-gp, MRP1 is also a primary active transporter of many glutathione-, glucuronide-, and sulfateconjugated organic anions. Some of these compounds are potential physiological substrates, such as cysteinyl leukotriene 4 (LTC 4 ) and 17␤-estradiol 17-(␤-D-glucuronide) (E 2 17␤G) (2-4, 9 -11). Knock-out mice lacking murine MRP (mMRP) have an impaired response to a leukotriene-mediated inflammatory stimulus (12), and mMRP1 has also been shown recently to regulate dendritic cell migration to lymph nodes by effluxing LTC 4 (13).
MRP1 is a member of the ABCC branch of the ATP binding cassette (ABC) superfamily and has been designated ABCC1. This branch also includes the cystic fibrosis conductance regulator (ABCC7) and the two sulfonylurea receptors (ABCC8 and ABCC9) as well as a number of more recently identified MRP1related proteins including MRP2 to -7 and ABCC11 and -12 (4, 14 -23). The predicted topologies of MRP2, MRP3, MRP6, and MRP7 are similar to that of MRP1. In addition to a typical ABC transporter core structure containing two membrane-spanning domains (MSDs) and two nucleotide binding domains, these proteins are characterized by an additional NH 2 -terminal MSD (MSD1) that is composed of five transmembrane (TM) helices. Thus, the proteins are predicted to contain a total of 17 TM helices with an extracellular NH 2 terminus ( Fig. 1) (24). MRP4 and MRP5, as well as the more recently identified ABCC11 and -12 appear to lack MSD1 and to have a 12-TM helix structure more typical of ABC transporters (14, 17, 20 -23). Of all MRP family members, MRP2 and MRP3 are functionally closest to MRP1. Like MRP1, when overexpressed, both MRP2 and MRP3 are able to confer resistance to some cytotoxic drugs (25)(26)(27)(28)(29). In addition, both are active organic anion transporters that display varying substrate specificities and affinities (25, 27, 29 -32).
What determines the ability of MRP1 to transport structurally unrelated cytotoxic drugs and conjugated organic anions while retaining the ability to distinguish between structurally similar conjugates of the same parental compound remains largely unknown (10). Structure/function studies have only recently begun to identify domains and individual amino acid residues in MRP1 that are involved in determining substrate specificity. Photoaffinity labeling studies have shown that LTC 4 labels sites in both the NH 2 -and COOHproximal halves of MRP1 and that labeling of the COOHproximal half of the protein is confined to a region encompassing TM14 to -17 of MSD3 (33). Using photoreactive drug analogs, N-(hydrocinchonidin-8Ј-yl)-4-azido-2-hydroxybenzamide (IACI) and [ 125 I]iodoaryl azidorhodamine 123 (IAARh123), Daoud et al. (34) have mapped major photoaffinity-labeled sites to MRP1 sequences encoding TM10 and -11 of MSD2 and TM16 and -17 of MSD3. In contrast, 125 Ilabeled 11-azidophenyl agosterol A, a derivative of the polyhydroxylated sterol acetate, agosterol A, photolabels only a site in the COOH-proximal region of MRP1 (amino acids 932-1531), and it does so in a GSH-dependent manner (35). Our original functional analyses of hybrid mMRP1 and human MRP1 proteins localized regions that are important for both anthracycline resistance and efficient transport of E 2 17␤G to the COOH-terminal third of MRP1 (36). More detailed analyses of individual nonconserved amino acids within these regions subsequently revealed that Glu 1089 in the predicted TM14 is critical for the ability of the protein to confer anthracycline resistance (37). We have also shown that mutations of a nonconserved hydrophilic residue, Thr 1242 , within the putative TM17 of MRP1 dramatically decreases the ability of the protein to confer drug resistance and to transport E 2 17␤G without significant effect on LTC 4 transport (38). In addition, substitutions of the nearby highly conserved aromatic polar residue in TM17, Trp 1246 , eliminates the ability of MRP1 to transport E 2 17␤G and to confer drug resistance but also has a relatively minor effect on LTC 4 transport (39).
It has been proposed that hydrogen bond formation plays an important role in the interaction of MRP1 with its substrates (38 -40). Interestingly, of all putative TM helices of MRP1, TM17 is the most amphipathic, containing ϳ50% polar and polar-aromatic residues (40). As mentioned previously, we have shown that two such amino acid residues in TM17, Thr 1242 and Trp 1246 , are essential for MRP1 to confer drug resistance and to transport E 2 17␤G efficiently (38,39). We have now investigated the importance of all other residues with hydrogen bonding capability within TM17 in determining the substrate specificity of MRP1. We have systematically mutated the amino acids Ser 1233 , Ser 1235 , Ser 1237 , Gln 1239 , Thr 1241 , and Asn 1245 to alanine and each of two tyrosines, Tyr 1236 and Tyr 1243 , to Phe. These mutant proteins were then stably expressed in human embryonic kidney (HEK293) cells, and the transfectants were characterized with respect to their drug resistance profiles and their ability to transport LTC 4 and E 2 17␤G. The results of these studies indicate that residues affecting substrate specificity cluster in the two cytoplasmic proximal turns of the predicted TM helix.  16), and vincristine sulfate were obtained from Sigma, and epirubicin was from ICN Biomedicals (Irvine, CA).
All mutations were confirmed by sequencing using DNA Thermo Sequenase and Cy5.5 and Cy5.0 dye terminator/primer (Amersham Biosciences). DNA fragments containing the desired mutations were transferred into pCEBV7-MRP1, after which the entire mutated inserts and the cloning sites were verified by DNA sequencing.
Cell Lines and Tissue Culture-Stable transfection of HEK293 cells with the pCEBV7 vector containing the wild type MRP1 cDNAs has been described previously (37,41). All of the mutated MRP1 constructs were analyzed as stably transfected HEK293 cells grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 100 g/ml hygromycin B (Roche Molecular Biochemicals). Briefly, HEK293 cells were transfected with pCEBV7 vectors containing mutant MRP1 using Fugene6 (Roche Molecular Biochemical) according to the manufacturer's instructions. After ϳ48 h, the transfected cells were supplemented with fresh medium containing 100 g/ml hygromycin B. Approximately 3 weeks post-transfection, the hygromycin B-resistant cells were cloned by limiting dilution, and the resulting cell lines were tested for expression of the mutant proteins.
Determination of Protein Levels in Transfected Cells-Plasma membrane vesicles were prepared by centrifugation through sucrose, as described previously (9,10). After determination of protein levels by Bradford assay (Bio-Rad), total membrane protein (0.5, 1.0, and 1.25 g) from transfectants expressing wild type MRP1 and various mutant proteins was analyzed by SDS-PAGE using a 7.5% gel. Proteins were subsequently transferred to Immobilon-P polyvinylidene difluoride membranes (Millipore Corp., Bedford, MA) by electroblotting. The proteins were detected with mAb, QCRL-1 (42). Antibody binding was detected with horseradish peroxidase-conjugated goat anti-mouse IgG (Pierce), followed by enhanced chemiluminescence detection and X-Omat™ Blue XB-1 films (PerkinElmer Life Sciences). Densitometry of the film images was performed using a ChemiImager™ 4000 (Alpha Innotech Corp., San Leandro, CA). The relative protein expression levels were calculated by dividing the integrated densitometry values obtained for 0.5, 1.0, and 1.25 g of total membrane protein from transfectants expressing the mutant proteins by the integrated densitometry values obtained for the comparable amounts of total membrane proteins from transfectants expressing the wild type protein. Each comparison was performed at least three times in independent experiments. The results were then pooled, and the mean values were used for normalization purposes.
Chemosensitivity Testing-Drug resistance was determined using the colorimetric 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay as described previously (7,37,43). Briefly, cells were seeded at 5 ϫ 10 3 cells/well in 100 l of culture medium in 96-well tissue culture plates. The following day, various concentrations of drug diluted in culture medium were added to cells (100 l/well). After incubation for an additional 96 h, 100 l of medium were removed from each well, and the MTT reagent (25 l/well, 2 mg/ml phosphate-buffered saline) (Sigma) was added. After 3 h, the formazan was solubilized by mixing with 1 N HCl/isopropyl alcohol (1:24) (100 l/well). Color density was determined using the ELX 800 UV spectrophotometer (570 nm). Mean values of quadruplicate determinations Ϯ S.D. were plotted using GraphPad software. IC 50 values were obtained from the best fit of the data to a sigmoidal curve. Relative resistance is expressed as the ratio of the IC 50 value of cells transfected with MRP1 expression vectors to that of cells transfected with empty vector. Resistance factors were determined in three or more independent experiments. The significance of the difference between relative resistance factors of wild type and mutant MRP1 transfectants was determined using an unpaired Student's t test.
Confocal Microscopy-Confocal microscopy was carried out as described previously (37,38). Briefly, ϳ5 ϫ 10 5 stably transfected HEK293 cells were seeded in each well of a six-well tissue culture dish on coverslips. When the cells had grown to confluence, they were washed once in phosphate-buffered saline and then fixed with 2% paraformaldehyde in phosphate-buffered saline, followed by permeabilization using digitonin (0.25 mg/ml in phosphate-buffered saline). MRP1 proteins were detected with the monoclonal antibody MRPm6, which reacts with an epitope close to the COOH terminus of MRP1 (amino acids 1511-1520) (44). Antibody binding was detected with Alexa Fluor 488 anti-mouse IgG (H ϩ L) (FabЈ) 2 fragment. Nuclei were stained with propidium iodide. Localization of MRP1 in the transfected cells was determined using a Meridian Insight confocal microscope (filter, 620/40 nm for propidium iodide; 530/30 nm for Fluor 488).
LTC 4 and E 2 17␤G Transport by Membrane Vesicles-Plasma membrane vesicles were prepared as described previously, and ATP-dependent transport of [ 3 H]LTC 4 into the inside-out membrane vesicles was measured by a rapid filtration technique (9, 10). Briefly, vesicles (10 g of protein) were incubated at 23°C in 100 l of transport buffer (50 mM Tris-HCl, 250 mM sucrose, 0.02% sodium azide, pH 7.4) containing ATP or AMP (4 mM), 10 mM MgCl 2 , and [ 3 H]LTC 4 (50 nM, 100 nCi). At the indicated times, 20-l aliquots were removed and added to 1 ml of ice-cold transport buffer, followed by filtration under vacuum through glass fiber filters (type A/E; Gelman Sciences, Dorval, Quebec, Canada). Filters were immediately washed twice with 4 ml of cold transport buffer and then dried before the bound radioactivity was determined by scintillation counting. All data were corrected for the amount of [ 3 H]LTC 4 that remained bound to the filter in the absence of vesicle protein (usually Ͻ5% of the total radioactivity). [ 3 H]LTC 4 uptake was expressed relative to the total protein concentration in each reaction. ATP-dependent [ 3 H]E 2 17␤G (400 nM, 120 nCi) uptake was measured as described for [ 3 H]LTC 4 except that the reaction was carried out at 37°C.
K m and V max values of ATP-dependent [ 3 H]LTC 4 uptake by membrane vesicles (2.5 g of protein) were measured at various LTC 4 concentrations (0.01-1 M) for 1 min at 23°C in 25 l of transport buffer containing 4 mM ATP and 10 mM MgCl 2 , followed by nonlinear regression analyses. Kinetic parameters of ATP-dependent [ 3 H]E 2 17␤G (0.1-16 M) uptake were determined as described for [ 3 H]LTC 4 except that the reaction was carried out at 37°C.

Expression of Mutant MRP1 in Stably Transfected HEK293
Cells-We have previously shown that single mutations at either Thr 1242 or Trp 1246 within predicted TM17 of MRP1 drastically affect the ability of MRP1 to confer drug resistance. They also cause a marked decrease in the ability of the protein to transport E 2 17␤G while having little or no effect on LTC 4 transport (38,39). Predicted TM17 of MRP1 contains an unusually large number of amino acids with side chains capable of hydrogen bonding, in addition to Thr 1242 and Trp 1246 (Fig. 1). To examine their possible importance in determining the substrate specificity of MRP1, we generated a series of eight mutant proteins in which Ser 1233 , Ser 1235 , Ser 1237 , Gln 1239 , Thr 1241 , and Asn 1245 were replaced with Ala, and Tyr 1236 and Tyr 1243 were substituted with Phe (Fig. 1).
The episomal expression vector, pCEBV7, containing mutated forms of MRP1 cDNAs was used to stably transfect HEK293 cells, and populations of transfected cells were se-lected in hygromycin B. The resultant stably transfected cell populations were cloned by limiting dilution, and subpopulations expressing high levels of MRP1 mutant proteins were used in subsequent studies. The levels of mutant proteins relative to wild type MRP1 in previously characterized HEK transfectants were determined by immunoblotting and densitometry (Fig. 2). Endogenous MRP1 in HEK293 cells transfected with the empty vector was undetectable under the conditions used (data not shown). As shown in Fig. 2, the expression levels of seven of the mutant proteins in stably transfected HEK293 cells were similar to that of wild type MRP1. The exception was MRP1 S1235A , which was expressed at a level approximately half that of the other mutant and wild type proteins (Fig. 2).
Resistance Profiles of Wild Type and Mutant Human Proteins-The drug resistance profiles of transfectants expressing mutant proteins were determined using MTT assays. The results are presented as relative drug resistance factors in Table  I. Based on the effects of the mutations on resistance to one or more drugs, they were classified into three groups. Substitution of four hydrophilic amino acid residues within TM17 with Ala (S1233A, S1235A, S1237A, Q1239A) had no significant effect on the ability of MRP1 to confer resistance to any drug tested. Mutation of one polar-aromatic residue, Y1243F, caused an approximately 2-3-fold reduction of resistance to all four drugs, whereas three mutations, Y1236F, T1241A, and N1245A, resulted in a 2-3-fold loss of resistance to only certain drugs. Interestingly, both mutation Y1236F and mutation T1241A decreased vincristine resistance ϳ3-fold without a significant effect on the resistance to VP-16, doxorubicin, and epirubicin. In contrast, conversion of Asn 1245 to Ala resulted in a mutant protein with enhanced resistance to vincristine (1.6fold) but a 2-3-fold partial decrease in the ability to confer resistance to VP- 16  resistance profiles of MRP1 might be attributable to changes in trafficking of the protein, we compared the subcellular localization of wild type and mutant human proteins by confocal microscopy. As shown in Fig. 3, the subcellular distribution of the mutated proteins assessed by immunoreactivity with the MRP1-specific mAb MRPm6 was indistinguishable from that of cells expressing wild type protein. In all cases, strong plasma membrane staining was observed, indicating that trafficking was unaffected.

Transport of [ 3 H]LTC 4 and [ 3 H]E 2 17␤G by Wild
Type and Mutant MRP1-In addition to the effects on drug resistance profiles, we found previously that replacement of Thr 1242 with the nonpolar Ala decreased the efficiency of E 2 17␤G transport by ϳ50%, whereas substitutions of Trp 1246 with conserved (Phe, Tyr) or nonconserved (Cys, Ala) amino acids essentially eliminated transport of this substrate. In contrast, the T1242A mutation had no significant effect on LTC 4 transport, and the effect of the Trp 1246 mutations was relatively minor (a 2-fold increase in K m ) (38,39). 2 To determine whether any of the other mutations in TM17 of MRP1 altered the efficiency with which the protein transported either LTC 4 or E 2 17␤G, we examined ATP-dependent uptake of these compounds by membrane vesicles prepared from HEK transfectants expressing each of these mutant proteins (Figs. 4 and 5).
The levels of LTC 4 uptake by vesicles prepared from HEK transfectants expressing either wild type MRP1 or mutants MRP1 S1233A , MRP1 S1235A , MRP1 Y1236F , MRP1 S1237A , MRP1 Q1239A , MRP1 T1241A , and MRP1 N1245A were proportional to the relative expression levels of the wild type and mutant proteins. Thus, none of these mutations had any effect on the transport of this substrate. The only mutation that affected LTC 4 transport was conversion of Tyr 1243 to Phe, which decreased transport of LTC 4 by ϳ30% (Fig. 4).

Kinetic Parameters of [ 3 H]LTC 4 and [ 3 H]E 2 17␤G
Transport-We have shown that mutation Y1243F affected the ability of the protein to transport both LTC 4 and E 2 17␤G and that mutation N1245A decreased the transport of only E 2 17␤G. To determine the influence of these mutations on the kinetic parameters of transport, we compared K m and V max values for the wild type protein with those of mutants MRP1 Y1243F and 2 Ito, K., Deeley, R. G., and Cole, S. P. C., unpublished results.  3. Confocal microscopy of cells expressing wild type and mutant MRP1. Cells were grown and stained for immunofluorescence detection of MRP1 as described. MRP1 was detected using mAb MRPm6. The location of MRP1 is indicated in green. Nuclei were stained with propidium iodide and are shown in red. Transfectants tested were expressing wild type or mutant MRP1 as indicated. A-E, an x-y optical section of the cells is shown to illustrate the distribution of the wild type and mutant proteins between plasma and intracellular membranes. MRP1 N1245A (Fig. 6). For wild type MRP1 and mutant MRP1N1245A, the K m values for LTC 4 uptake were identical (90 nM), and the normalized V max value for N1245A was also very similar to that of wild type MRP1 (V max ϭ 100 pmol/mg/ min for mutation N1245A; 117 pmol/mg/min for wild type MRP1) (Table II). However, replacement of Tyr 1243 with Phe decreased the normalized V max value for LTC 4 transport ϳ30% relative to wild type MRP1 (V max ϭ 75 pmol/mg/min for mutation Y1243F). The apparent K m value for mutation Y1243F was 112 nM (Fig. 6 and Table II). Thus, the Y1243F mutation decreased the V max /K m ratio for LTC 4 ϳ2-fold.
For E 2 17␤G transport, a nonlinear regression analysis of the data generated a K m value of 1.4 M for wild type MRP1, consistent with previous estimates (10), compared with 5.4 and 10.9 M for mutations Y1243F and N1245A, respectively ( Fig.  6 and Table II). The normalized V max values for mutations Y1243F and N1245A were lower than that for wild type MRP1 (V max ϭ 403 pmol/mg/min for wild type MRP1 versus 316 pmol/ mg/min for mutation Y1243F and 288 pmol/mg/min for mutation N1245A) ( Fig. 6 and Table II). Thus mutations Y1243F and N1245A decreased the V max /K m ratio for E 2 17␤G ϳ5and 11-fold, respectively.
Effect of Mutations Y1243F and N1245A on the Inhibition of MRP1-mediated E 2 17␤G/LTC 4 Transport by LTC 4 / E 2 17␤G-As an alternative means of assessing the effects of the TM17 mutations on the interaction between LTC 4 and the human protein, we examined the ability of LTC 4 to inhibit transport of E 2 17␤G. IC 50 values for wild type and mutant human proteins were obtained from the best fit of the inhibition data to a sigmoidal curve (Fig. 7A). For the wild type protein, the IC 50 value for LTC 4 was 357 nM. Converting Asn 1245 to Ala reproducibly decreased the IC 50 value (207 nM), whereas mutation of Tyr 1243 to Phe resulted in a slight, reproducible increase (407 nM).
In a reciprocal set of experiments, we also examined the ability of the conjugated estrogen to inhibit transport of LTC 4 (Fig. 7B). For the wild type protein, the IC 50 value for E 2 17␤G was 3.6 M compared with 11.8 M for mutation Y1243F and 15.1 M for mutation N1245A. Since these results are independent of protein expression levels, they provide additional evidence that the observed decrease in transport of mutations Y1243F and N1245A at nonsaturating concentrations of E 2 17␤G is primarily attributable to changes in the affinity of the proteins for this substrate. DISCUSSION The mechanism by which MRP1 binds and transports such structurally unrelated cytotoxic drugs and conjugated organic anions is presently not fully understood. It is apparent from previous mutagenesis studies of MRP1 orthologs and homologs that it is impossible to predict, based on conservation of function and primary structure, which amino acids may be important for determining substrate specificity or overall activity of the protein. For example, both MRP1 and mMRP1 are equally effective at conferring resistance to vincristine and VP-16. However, mutation of a nonconserved hydrophilic amino acid residue, Thr 1242 , within TM17 to Ala, the corresponding amino acid in mMRP1, markedly decreases resistance to both drugs. Furthermore, resistance to both vincristine and VP-16 can be restored by mutating another nonconserved amino acid in TM helix 14, Glu 1089 , to Gln as in the murine protein (38). Thus, these two pairs of nonconserved amino acids in MRP1 and mMRP1 interact functionally to maintain the common ability to confer resistance to vincristine and VP-16 (38). In contrast, Trp 1246 is conserved in most members of the ABCC branch including MRP2, -3, -4, and -6 and cystic fibrosis conductance regulator. Both MRP1 and MRP2 transport LTC 4 and E 2 17␤G. In MRP1, mutation of Trp 1246 essentially eliminates E 2 17␤G transport with little effect on transport of LTC 4 (39). However, when the corresponding residue, Trp 1254 , is mutated in MRP2, only nonconserved substitutions eliminate E 2 17␤G transport, and in contrast to MRP1, only the most conservatively substituted MRP2 Trp 1254 mutant, W1254Y, transports LTC 4 (45).
In addition to the mutagenesis studies of Thr 1242 and Trp 1246 , photolabeling of MRP1 with IACI and IAARh123 followed by partial proteolysis also indicates that TM17 and possibly TM16 may be involved directly in drug binding (34). One of the two sites in MRP1 photolabeled by LTC 4 is also located in a region that encompasses TM14 to -17 (33), but whether the region is completely coincident with that labeled by IACI and IAARh123 has not been established. TM17 is predicted to span from approximately Gly 1228 to Val 1248 . Its precise positioning in the membrane has not been determined experimentally, but based on studies of other membrane proteins and model peptides, it appears likely that the highly conserved Trp 1246 may contribute to anchoring the helix at the membrane/cytoplasm interface (39, 45-47) (Fig. 1).
The whole of TM17 is exceptionally well conserved among members of the C branch of ABC transporters and is highly amphipathic (38,39). It contains no charged residues but is unusually rich in amino acids with hydroxyl group substituents and other residues with side chains capable of hydrogen bonding ( 1228 GLVGLSVSYSLQVTTYLNWLV 1248 ) (Fig. 1). In addition to the previous mutations of Thr 1242 and Trp 1246 , we have now mutated the remaining hydrophilic and polar aromatic amino acid residues within TM17 (Fig. 1) to further assess the possible influence of side chain hydrogen bonding on the substrate specificity and overall activity of the protein.
With the exception of Tyr 1236 , mutation of the amino acids with polar substituents predicted to be most distant from the membrane/cytoplasm interface (S1233A, S1235A, S1237A, and Q1239A) to Ala had no effect on the ability of MRP1 to confer resistance to any of the drugs tested. In contrast, mutation of Thr 1241 , Tyr 1243 , and Asn 1245 affected either the overall drug resistance activity or drug specificity of the protein. Together with Thr 1242 and Trp 1246 , these five residues create a zone with significant polar character that would be predicted to cross the inner leaflet of the membrane (Fig. 8). As can be seen from the figure, the contiguity of the polar residues is interrupted by only a single nonpolar amino acid, Leu 1244 . Thus, assuming a pitch of 3.6 residues per turn, this region would be predicted to occupy almost two complete turns of TM17. Tyr 1236 , which is the only polar residue in the NH 2 -proximal half of TM17 that affects drug resistance, is predicted to be aligned "above" Tyr 1243 and to be separated from it by Gln 1239 . Based on the substrates we have analyzed, mutation of Gln 1239 to Ala had no detectable effect on MRP1 function. It is notable that mutation of all of the residues examined affected resistance to vincristine, the largest of all substrates tested, whereas mutation of Tyr 1236 and Thr 1241 , the two residues closest to the outer leaflet of the membrane, had no effect on resistance to VP-16 and the anthracyclines or on the transport of E 2 17␤G.
The exact mechanism by which the individual mutations affect substrate specificity is not known. The mutations we have made in TM17 of MRP1 had no detectable effect on protein trafficking or, in most cases, the ability of MRP1 to transport LTC 4 . Consequently, they appear not to have altered protein folding to any significant extent. We previously proposed that the hydrogen bonding capability of Thr 1242 and Trp 1246 played an important role in the interaction of substrates with the protein (38,39). The similar effect of the mutation, T1241A, on the resistance to vincristine might be also due to the elimination of the hydrogen bonding capability. Results obtained from mutation Trp 1246 suggested that aromatic stacking interactions and possibly -bonding might contribute to the role played by this residue in determining substrate specificity (39). We noted that a conservative substitution of each of the two tyrosine residues in TM17 with a nonpolar aromatic residue (Phe) caused a 2-3-fold decrease in vincristine resistance. However, in the case of Tyr 1236 , the decrease was restricted to vincristine, whereas mutation of Tyr 1243 also decreased resistance to VP-16 and the two anthracyclines by ϳ50%. These results support the importance of hydrogen bonding by these two residues, particularly with respect to vincristine resistance. Mutation to Phe was chosen to minimize the probability of introducing structural changes that might nonspecifically affect substrate binding. Consequently, we presently cannot assess the importance of hydrophobic and aromatic stacking interactions with these two residues.
Substitution of Asn 1245 with Ala decreased the resistance to VP-16 and the two anthracyclines tested, consistent with the involvement of hydrogen bonding in the interaction with these drugs. In contrast, the mutation increased the ability of the  4 and E 2 17␤G transport were determined from nonlinear regression analysis of the combined data and are shown in Table II. protein to confer vincristine resistance. This effect is similar to that observed during mutagenesis studies of His 61 within TM1 of P-gp in which the bulkiness of the side chain of the amino acid at this position was found to be an important factor in determining substrate specificity. Replacement of His 61 by amino acids with small side chains increased the relative resistance to vinblastine, whereas replacement by amino acids with large side chains generally increased the relative resistance to smaller substrates such as colchicine, VP-16, and doxorubicin, presumably by decreasing the ability of larger substrates to engage other residues in the drug binding pocket (48). By analogy, conversion of Asn 1245 to Ala in MRP1 may increase accessibility to a relatively large substrate, such as vincristine, whereas the loss of hydrogen bonding may contribute to the decrease in resistance to other drugs, such as VP-16, doxorubicin, and epirubicin.
We found previously that mutation of either Thr 1242 or Trp 1246 had a pronounced effect on the ability of MRP1 to transport E 2 17␤G but no or little effect on the transport of LTC 4 (38,39). Among the mutations analyzed here, only mutation of Tyr 1243 affected LTC 4 transport, and, like Trp 1246 , the effect was minor. In contrast, mutation of both Tyr 1243 and Asn 1245 increased the K m for E 2 17␤G approximately 4-and 8-fold, respectively. Thus, to date, no residue has been identified in TM17 that is critical for LTC 4 transport, despite the fact that LTC 4 competitively inhibits E 2 17␤G transport. Given that many MRP1 and MRP2 substrates are anionic, it has been suggested that positively charged amino acids may be particularly important for substrate binding (40,49). However, TM17 contains no charged residues of any kind. This observation combined with the lack of effect of mutations in TM17 that might influence hydrogen bonding suggests that the high affinity binding of LTC 4 is primarily attributable to interaction with residues in other TM helices.
We have proposed previously that competition for transport between structurally diverse MRP1 substrates, involving both conjugated and unconjugated compounds, may be explained if they interact with partially shared, but individually distinct, sets of determinants in a common binding pocket. Recent crystallographic studies of independently evolved multidrug-binding proteins (including QacR, a regulator of expression of the multidrug transporter from Staphylococcus aureus, QacA (50); the human nuclear xenobiotic receptor, PXR (51); and the ABC E. coli lipid transporter, MsbA (52)) suggest that these proteins all contain a relatively large drug binding pocket lined with residues that interact with substrate in different combinations to determine specificity. Two features of these binding pockets are of particular note with respect to the data presented here. In the case of both QacR and PXR, polar residues appear to play key roles in determining substrate specificity, and in QacR, the protein appears to have two separate but potentially overlapping drug binding sites in the same drug binding pocket. The existence of a similar structure in MRP1 may explain why the mutations created to date have very different effects on the transport of two reciprocally competitive sub-  4 and E 2 17␤G uptake by vesicles from HEK cells transfected with vectors encoding wild type and mutant proteins The kinetic parameters of LTC 4 and E 2 17␤G uptake were determined as described in the legend to Fig. 6. The normalized V max values were obtained by adjusting determined V max values to compensate for differences in the relative levels of the wild type and mutant proteins and shown in parentheses.  Shown is a predicted three-dimensional structure for TM17, obtained using WebLab View-erPro. For clarity of presentation, residues found in this and earlier studies (38,39) to influence substrate specificity are indicated in color at predicted positions on the helix. The qualitative effects of mutating each of the residues with respect to the ability to confer resistance to various drugs and to transport E 2 17␤G are also shown. For comparative purposes, a three-dimensional model of vincristine obtained by WebLab ViewerPro is also illustrated at the same scale as the model of TM17. strates, LTC 4 and E 2 17␤G. The residues in TM17 that we have shown to affect substrate specificity clearly cannot all lie on a single exposed face of a typical ␣-helix. However, the spatial relationships between the TM helices of MRP1 and their precise positions and tilts in the membrane are not known. Without this information, it is difficult to predict the exposure that individual residues may have to relatively hydrophilic cytoplasmic substrates or more hydrophobic compounds that may bind from within the membrane. While it is impossible at present to describe for the entire binding pocket how each substrate establishes distinct but mutually exclusive interactions, the data presented here illustrate how this may occur in a limited region of the pocket contributed by a single TM helix. The data confirm the existence of multiple overlapping interactions with amino acid side chains capable of hydrogen bonding that can contribute to the binding of structurally diverse hydrophobic drugs and an anionic conjugate, such as E 2 17␤G.