Mutational Analysis of Ionizable Residues Proximal to the Cytoplasmic Interface of Membrane Spanning Domain 3 of the Multidrug Resistance Protein, MRP1 (ABCC1) AND CATALYTIC ACTIVITY OF THE TRANSPORTER*

The multidrug resistance protein MRP1 is an ATP-de-pendent transporter of organic anions and chemothera-peutic agents. A significant number of ionizable amino acids are found in or proximal to the 17 transmembrane (TM) helices of MRP1, and we have investigated 6 of these at the cytoplasmic interface of TM13–17 for their role in MRP1 expression and transport activity. Opposite charge substitutions of TM13 Arg 1046 and TM15 Arg 1131 did not alter MRP1 expression nor did they substantially affect activity. In contrast, opposite charge substitutions of TM16 Arg 1202 and Glu 1204 reduced protein expression by > 80%; however, MRP1 expression was not affected when Arg 1202 and Glu 1204 were replaced with neutral or same-charge residues. In addition, organic anion transport levels of the R1202L, R1202G, and R1202K mutants were comparable with wild-type MRP1. In contrast, organic anion transport by E1204L was substantially reduced, whereas transport by E1204D was comparable with wild-type MRP1, with the notable exception of GSH. Opposite charge substitutions of TM16

ATP-binding cassette (ABC) 1 membrane proteins make up one of the largest protein superfamilies documented and are found in every organism, from the simplest bacteria to the most complex multicellular organisms (1). Of the 48 human ABC proteins identified to date, three of them, P-glycoprotein (gene symbol ABCB1), multidrug resistance protein 1 (MRP1) (gene symbol ABCC1), and breast cancer resistance protein (gene symbol ABCG2), appear to account for nearly all instances of multidrug resistance in tumor cells that display a decrease in drug accumulation (2)(3)(4).
The functional form of most mammalian ABC transporters is comprised of four domains: two membrane spanning domains (MSDs) each containing six transmembrane (TM) ␣-helices that alternate with two nucleotide binding domains (NBDs). However, MRP1 and six of its closest homologs in the ABC "C" family have 17 TMs distributed among three MSDs, in addition to their two NBDs. The NH 2 -terminally located third MSD (MSD1) is ϳ200 amino acids and contains just five TMs. Consequently, the NH 2 terminus of MRP1 is extracellular (5).
MRP1 has a very broad substrate spectrum relative to many other ABC drug transporters (3). Consistently, drug-selected and -transfected tumor cell lines overexpressing MRP1 exhibit increased resistance to natural product drugs such as anthracyclines (e.g. doxorubicin), Vinca alkaloids (e.g. vincristine), and epipodophyllotoxins (VP-16), as well as folic acid analogs (e.g. methotrexate (MTX)) and arsenic and antimony-centered oxyanions (3,4). The profile of potential endogenous substrates of MRP1 is equally diverse. Thus, in vitro studies have shown that MRP1 is a primary active transporter of a wide variety of organic anions, many of which are conjugated to the anionic moieties GSH, glucuronide, and sulfate. Substrates include the inflammatory mediator leukotriene C 4 (LTC 4 ) and the cholestatic estrogen, 17␤-estradiol 17␤(D-glucuronide) (E 2 17␤G), as well as GSH and GSSG (3,4). Unmodified substrates often require co-transport with GSH (6 -10). In addition, GSH and certain of its analogs can stimulate the transport of some conjugated compounds by MRP1, such as estrone 3-sulfate (E 1 SO 4 ). Furthermore, GSH transport itself can be markedly enhanced by a variety of xenobiotics, including bioflavonoids such as apigenin (3,(11)(12)(13)(14)(15)(16). Thus the mechanism of MRP1mediated transport is complex and still far from completely understood.
Several studies using photoaffinity drug analogs and substrates and site-directed mutagenesis have localized discrete domains and, in some cases, individual amino acids in both the cytosolic and membrane spanning regions of both halves of MRP1 that are required for the binding and/or transport of various substrates and inhibitors of this transporter (3,(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29). Of the individual residues identified, a significant number are ionizable ("charged") and located in MSD2 (TM6 -11) and MSD3 (TM12-17) (19, 22-24, 27, 28). For example, we demonstrated previously (22,24) that both conservative and nonconservative mutations of Lys 332 in the inner leaflet of TM6 of MSD2 cause a selective loss of LTC 4 and GSH transport, whereas mutation of the nearby Asp 336 as well as Arg 593 in TM11 essentially eliminates organic anion transport activity. In other studies, we showed that various substitutions of ionizable residues located in or proximal to the membrane-cytosol interface of TM14 in MSD3 could also differentially affect both MRP1 transport activity and substrate specificity (23,27). Because many organic anions transported by MRP1 are relatively hydrophilic and are formed inside the cell, it seems likely that they would first come into contact with MRP1 from the cytoplasm rather than from the membrane, as postulated for the hydrophobic substrates of the P-glycoprotein transporter (29). Thus, it is reasonable to suppose that amino acids located at the membrane-cytosol interface of MRP1 might form initial contacts for these substrates. Furthermore, ionizable residues in this region may also serve to anchor the positions of some TM helices and thereby promote stable expression of the transporter in the plasma membrane.
In this paper, we have extended our studies of MSD3 by investigating the role of six additional ionizable residues predicted by hydropathy analyses to be in or proximal to the cytoplasmic interface of TM13 and -15-17 in the expression and transport activity of MRP1 (Fig. 1). Thus, Arg 1046 , Arg 1131 , Arg 1197 , Arg 1202 , Glu 1204 , and Arg 1249 were targeted for substitution by site-directed mutagenesis, and the expression, organic anion transport properties, and, in some cases, the substrate and nucleotide binding properties of the resulting mutant proteins were examined.
Transfection of MRP1 Expression Vectors into HEK293T Cells, Preparation of Membrane Vesicles, and Quantitation of MRP1 Protein Levels-SV40-transformed human embryonic kidney cells (HEK293T) were seeded at 10 7 cells per 150-mm plate. Twenty four to 48 h later, cells were transfected with either mutant or wild-type MRP1 pcDNA3.1(Ϫ)MRP1 DNA (16 g) using FuGENE 6 (Roche Applied Science) (48 l) at a ratio of 1:3, as suggested by the manufacturer. Seventy two hours later, cells were harvested, and membrane vesicles were prepared as described (6). Protein concentrations were determined and then the vesicles were aliquoted and stored at Ϫ70°C until needed.
To quantitate levels of MRP1 expression, membrane vesicle proteins (1 and 2 g) were resolved on a 7% SDS-polyacrylamide gel and then transferred onto an Immobilon-P membrane (Millipore). mAb QCRL-1 was used to detect MRP1 (30), and the signal was enhanced using Renaissance chemiluminescence reagent (PerkinElmer Life Sciences). Relative levels of MRP1 expression were determined by densitometry of exposed films.
Membrane was cross-linked to the protein by UV irradiation (1100 microwatts, 302 nm) for 10 1-min exposures. The samples were then resolved on a 7% SDS-polyacrylamide gel and processed for autoradiography. After drying, the gel was exposed to film at Ϫ70°C.
To measure 8-azido-ATP labeling of MRP1, membrane vesicles (20 g of protein) were dispersed in 20 l of transport buffer (50 mM Tris-HCl, pH 7.4, 250 mM sucrose) containing 5 mM MgCl 2 and 5 M 8-azido-[␣-32 P]ATP (31,32). After incubation on ice for 5 min, the samples were exposed to UV light at 302 nm on ice for 8 min. The reactions were stopped by the addition of ice-cold Tris-EGTA buffer, and the membrane proteins were collected by centrifugation and then subjected to SDS-PAGE and exposed to film as before.
To determine whether orthovanadate-induced trapping of ADP influences photolabeling of MRP1 by LTC 4 , membrane proteins (50 g) were incubated with 5 mM MgCl 2 in the presence or absence of 5 mM ATP and 1 mM orthovanadate for 20 min at room temperature followed by addition of [ 3 H]LTC 4 (200 nM, 110 nCi) and a further incubation for 30 min (20). Samples were UV cross-linked, subjected to SDS-PAGE, and exposed to film as before.

RESULTS
Expression and Transport Properties of MRP1 Mutants Containing Opposite and Same Charge Substitutions of Arg 1046 , Asp 1084 , and Arg 1131 in or Near TM13-15-In the first series of experiments, the importance of ionizable amino acids located in or near predicted TM13-15 for MRP1 protein expression and function were investigated by replacing Arg 1046 and Arg 1131 with Asp and Glu, respectively. Levels of expression of the R1046D and R1131E mutant proteins were determined by the immunoblotting of membrane vesicles prepared from the transfected HEK293T cells with mAb QCRL-1. The TM14-associated D1084R mutant described previously (27) was included for comparison. A single 190-kDa band was detected in both the wild-type and mutant MRP1 preparations, whereas no immunoreactive bands were detected in membrane vesicles prepared from control transfected cells ( Fig. 2A). Densitometry was performed on immunoblots of vesicle protein from three independent transfections, and the relative expression levels of the mutant MRP1 proteins were determined to be comparable (70 -130%) to wild-type MRP1 levels, indicating that these mutations did not substantially affect the biogenesis of MRP1.
To determine the functional consequences of replacing Arg 1046 , Asp 1084 , and Arg 1131 with an oppositely charged amino acid, the ability of the R1046D, D1084R, and R1131E mutants to transport different organic anions was tested. Fig. 2, B and C, shows that uptake levels of E 2 17␤G and LTC 4 , respectively, by the R1046D and R1131E mutants were moderately reduced (by 30 -50%) compared with wildtype MRP1, whereas uptake by the D1084R mutant was substantially reduced by Ͼ90%. Similarly, levels of E 1 SO 4 and MTX uptake by the R1046D and R1131E mutants were comparable with or moderately different from wild-type MRP1, whereas uptake of these organic anions by D1084R was substantially diminished (Յ25% of wild-type MRP1) (Fig. 2, D and E). These data are summarized in Table I, and overall, they indicate that mutation of Arg 1046 and Arg 1131 caused only a moderate effect or no effect on MRP1 transport activities. In contrast, replacement of Asp 1084 with an oppo- sitely charged amino acid caused a global and almost complete loss of transport activity. We reported previously (27) that substitution of Asp 1084 with Asn, Ala, or Val also eliminated LTC 4 and E 2 17␤G transport by MRP1. In the present study, we found that these nonconservative substitutions also resulted in a substantial loss (Ͼ70%) of E 1 SO 4 and MTX transport as well (results not shown).
Because the overall transport activity of the oppositely charged Asp 1084 mutant was greatly diminished (Fig. 2, B-E), a like-charge substituted D1084E mutant was created to clarify whether it was the loss of the acidic character or a change in size of the amino acid side chain that was responsible for the diminished activity. Like the D1084R mutant, the D1084E mutant showed substantially reduced LTC 4 uptake levels (Ͻ20% of wild-type MRP1) (27) (Fig. 2C). In contrast to D1084R, however, the D1084E mutant exhibited E 2 17␤G (Fig.  2B), E 1 SO 4 (Fig. 2D), and MTX (Fig. 2E) uptake levels that were similar or only moderately reduced compared with those of wild-type MRP1. Because of the apparent selectively greater loss of LTC 4 transport activity by the D1084E mutant, GSH uptake by this mutant was also examined and compared with that of D1084R. As shown in Fig. 2F, apigenin-stimulated GSH uptake by the D1084R and D1084E mutants was reduced by Ͼ90 and 80%, respectively. These data are summarized in Table I, and overall, our results indicate that both the acidic character and the size of the Asp 1084 side chain are critically important for MRP1-mediated LTC 4 and GSH transport activities. On the other hand, moderate to wild-type E 2 17␤G, E 1 SO 4 , and MTX transport activity can be supported by Asp 1084 mutants provided there is an acidic side chain at position 1084.
Selective Loss of MRP1 Expression in Mutants Containing Opposite Charge Substitutions of Arg 1197 , Arg 1202 , Glu 1204 , and Arg 1249 in or Near TM16 -17-In the next series of experiments, the importance of ionizable amino acids located in or near predicted TM16 and TM17 for MRP1 protein expression was investigated by initially replacing Arg 1197 , Arg 1202 , Glu 1204 , and Arg 1249 with oppositely charged residues, and levels of expression were determined by immunoblotting of membrane vesicle proteins as before (Fig. 3A). The relative mean expression levels of the R1197E and R1249D mutants from 2 to 3 independent transfections were comparable with wild-type MRP1 (110 and 70%, respectively). In contrast, expression levels of the R1202D and E1204K mutants were substantially reduced (by Ͼ75%). A Northern blot analysis performed on total RNA isolated from the transfected cells indicated that the R1202D and E1204K mRNA levels were comparable with wild-type MRP1 mRNA levels, thus excluding the possibility that the low R1202D and E1204K protein levels could be caused by differences in transfection efficiency (results not shown).   TM13  R1046D  115  70  80  120  ND b  TM14  D1084R  Ͻ10  Ͻ10  15  25  Ͻ10  D1084E  80  20  65  90  20  TM15  R1131E  70  50  80  60  ND  TM16  R1197E  Ͻ10  Ͻ10  Ͻ15  Ͻ10  ND  R1197K  20  Ͻ25  Ͻ20  Ͻ10  ND  R1202G  115  115  75  70  ND  R1202L  115  120  50  110  ND  E1204L  Ͻ10  50  10  110  Ͻ25  E1204D  100  115  100  115  Ͻ25  TM17 R1249D

MRP1 Expression Is Re-established in Mutants Containing
The values shown are means of duplicate or triplicate determinations and are derived from Fig. 2, 4, and 5 (see figure legends for details). For clarity, numbers have been rounded to the nearest 5%, and S.D. values (which were typically Ͻ10%) were omitted. b ND, not done.
sion of MRP1 in HEK cell membranes was further explored by replacing Arg 1202 with the hydrophobic Leu, and Glu 1204 with Leu and Asp. Arg 1202 is highly conserved in MRP1 orthologs and homologs except MRP3 (Table II). Consequently, Arg 1202 was also mutated to Gly because this latter amino acid is present in the corresponding position in MRP3. Immunoblots showed that expression levels of the R1202G and R1202L mutants (Fig. 3B) and the E1204L and E1204D mutants (Fig. 3C) ranged from 80 to 225% of wild-type MRP1. These results indicate that unlike substitutions with oppositely charged residues, neutral and same-charge substitutions of Arg 1202 and Glu 1204 did not adversely affect MRP1 biogenesis. Transport Activities of the Expressed Arg 1202 and Glu 1204 Mutants-In the next series of experiments, the organic anion transport activities of the neutrally substituted, expressed mutants of Arg 1202 (Fig. 4, A-D) and Glu 1204 (Fig. 4, E-I) were assessed. At 1 min, E 2 17␤G and LTC 4 uptake by the R1202G and R1202L mutants was comparable with wild-type levels (Fig. 4, A and B). Levels of E 1 SO 4 uptake by the R1202G mutant were reduced by ϳ25%, whereas uptake by the R1202L mutant was reduced by ϳ50% (Fig. 4C). Uptake levels of MTX by the R1202G mutant was moderately reduced (by ϳ30%), whereas uptake by R1202L was comparable with wild-type MRP1 (Fig. 4D). In contrast to the neutrally substituted Arg 1202 mutants, E 2 17␤G uptake by the neutrally substituted Glu 1204 mutant E1204L was Ͻ10% of wild-type MRP1 levels (Fig. 4E). In addition, LTC 4 uptake by E1204L was reduced by 50% (Fig. 4F), and E 1 SO 4 uptake was reduced by 90% (Fig. 4G). On the other hand, MTX uptake was comparable with wild-type MRP1 (Fig. 4H). These results are summarized in Table I. To determine whether the substrate-selective loss of transport function observed in the E1204L mutant was because of the loss of the acidic character or the change in the size of the side chain, organic anion uptake by the same-charge mutant, E1204D, was also assessed. As shown in Fig. 4, E-H, and Table  I, the E1204D mutant exhibited transport levels comparable with wild-type MRP1 for all substrates tested.
Apigenin-stimulated GSH uptake by the expressed Glu 1204 mutants was also investigated. As shown in Fig. 4I, both E1204L and E1204D exhibited a similar and substantial decrease in GSH transport levels (Ͼ75%). Overall, these results show that although MRP1 is still expressed when Arg 1202 and Glu 1204 are replaced with neutral or same-charge amino acids, the phenotypes of the resulting mutants are distinctly different. In general, the neutral mutants of Arg 1202 showed only moderate and substrate-specific decreases in their transport activities. On the other hand, a neutral substitution of Glu 1204 reduced or eliminated transport of all organic anions except MTX, whereas substitution with Asp had no effect with the exception that E1204D no longer transported GSH (see Table I for summary). These findings indicate that the longer side chain of Glu 1204 is particularly critical for GSH transport by MRP1.
Transport Activities of TM16/17 Arg 1197 and Arg 1249 Mutants-Unlike the R1202D and E1204K mutants, oppositely charged substitutions of Arg 1197 (R1197E) and Arg 1249 (R1249D) did not adversely affect expression of MRP1 (Fig. 3A). Nevertheless, these mutants displayed a global and substantial loss of transport activity (Fig. 5, Table I). Thus, levels of E 2 17␤G uptake by both mutants were Ͻ10% of wild-type MRP1 levels (Fig. 5A), whereas LTC 4 uptake by the R1197E and R1249D mutants was Ͻ15% of wild-type MRP1 (Fig. 5B). Uptake of E 1 SO 4 by the R1197E mutant was Ͻ15% of wild-type MRP1, whereas uptake of this sulfated estrogen by the R1249D mutant was reduced by Ͼ90% (Fig. 5C). Finally, the R1197E and R1249D mutants showed MTX transport activity that was not significantly different from the empty vector control (Fig. 5D).
To determine whether the charge or size of the Arg 1197 and Arg 1249 side chains were important for MRP1 transport function, the like-charge substituted mutants R1197K and R1249K were created. Expression levels of these mutants were comparable or somewhat greater than wild-type MRP1 (results not shown). Nevertheless, ATP-dependent E 2 17␤G uptake by the R1197K and R1249K mutants was 20 and Ͻ10% of wild-type MRP1, respectively (Fig. 5A). Similarly, uptake levels of LTC 4 (Fig. 5B), E 1 SO 4 (Fig. 5C), and MTX (Fig. 5D) by the R1197K and R1249K mutants were reduced to Յ10% of wild-type MRP1 uptake levels. These results are summarized in Table I, and overall, our data indicate that neither the basic character nor the size of the Arg 1197 and Arg 1249 side chains is important for MRP1 expression, but both are critical for organic anion transport activity.

Effect of Glu 1204 , Arg 1197 , and Arg 1249 Mutations on Photolabeling with [ 3 H]LTC 4 and 8-Azido-[␣-32 P]ATP-
In the next series of experiments, those same-charge or neutrally substituted mutants that showed substantially reduced transport activities (R1197K, E1204L, and R1249K) were further examined to determine whether their loss of transport activity was accompanied by a decrease in substrate binding. As shown in Fig. 6A, [ 3 H]LTC 4 photolabeling of the R1197K and R1249K mutants was completely abrogated, indicating that the reduced LTC 4 transport activity of these mutants was associated with decreased binding of this substrate. In contrast, [ 3 H]LTC 4 labeling of the E1204L mutant was comparable with wild-type MRP1, despite the fact that transport of this organic anion by this mutant was substantially reduced.
To determine whether the mutations of Arg 1197 , Glu 1204 , and Arg 1249 that altered the transport properties of MRP1 also affected the interaction of the transporter with nucleotide, the ability of the R1197K, E1204L, and R1249K mutants to be photolabeled with 8-azido-[␣-32 P]ATP, both at 4°C to minimize hydrolysis and at 37°C in the presence of sodium vanadate to trap azido-ADP after hydrolysis, was examined (31, 32). As shown in Fig. 6B, 8-azido-[␣-32 P]ATP labeling of the R1197K and R1249K mutants was comparable with wild-type MRP1.

Ionizable Amino Acids in MRP1 TM13-17; Expression/Activity
Furthermore, orthovanadate-induced trapping of 8-azido-[␣-32 P]ADP of the R1197K and R1249K mutants was also comparable with wild-type MRP1 (Fig. 6C). These results indicate that major changes in nucleotide interactions are not associated with these transport-inactivating mutations of Arg 1197 and Arg 1249 . 8-Azido-[␣-32 P]ATP labeling of the transport-compromised E1204L mutant was also comparable with wild-type MRP1 (Fig. 6B). However, vanadate-induced trapping of 8-azido-[␣-32 P]ADP was substantially increased (ϳ4-fold), specifically in the COOH-proximal NBD2 (Fig. 6C) (27). Because E1204L could still be photolabeled with LTC 4 despite substantially reduced transport of this organic anion, [ 3 H]LTC 4 photolabeling of the mutant protein after prior incubation with ATP and vanadate was examined to determine whether the increased trapping of azido-ADP by E1204L altered the substrate binding properties of MRP1. As shown in Fig. 6D

DISCUSSION
In the present study, we have extended our previous investigations on the importance of ionizable residues for the expression and activity of MRP1 by examining the consequences of mutating Arg 1046 , Arg 1131 , Arg 1197 , Arg 1202 , Glu 1204 , and Arg 1249 . These residues are predicted by hydropathy analyses (33) to be located in or proximal to the cytoplasmic interface of the TM helices of MSD3 (Fig. 1), and all six of them (or at least their charge) are quite highly conserved among mammalian ABC C family members with the exception of Arg 1131 (Table II). Substitutions of two of the residues (TM13 Arg 1046 and TM15 Arg 1131 ) with oppositely charged residues had no substantial effect on MRP1 protein expression levels and caused either no reduction or only a moderate reduction in transport activity. On the other hand, opposite charge substitutions of the other four MSD3 residues, viz. Arg 1197 , Arg 1202 , Glu 1204 , and Arg 1249 , like the Asp 1084 described previously (27), revealed them to be critically important for MRP1 expression and/or function (Fig. 7A).
We reported previously (27) that nonconservative substitutions of Asp 1084 proximal to TM14 caused a substantial reduction in E 2 17␤G, LTC 4 , and GSH transport and drug resistance, and we have now shown that these mutations also reduce MTX and E 1 SO 4 transport activity, demonstrating a global disruption of MRP1 activity. We have also found that the same-charge mutant, D1084E, has significant transport activity with respect to E 2 17␤G, MTX, and E 1 SO 4 whereas transport of GSH and the glutathione conjugate LTC 4 remains quite low. Thus both the acidic character and smaller volume of the Asp 1084 side chain (compared with Glu) appear to be particularly crucial for GSH and LTC 4 transport, whereas only a negative charge at this position is important for transport of other organic anions. However, neither physical property of Asp 1084 appears necessary for substrate binding, because GSH is still able to stimulate E 1 SO 4 transport indicating that GSH binding to D1084E is intact. Similarly, D1084E and D1084R can still be photolabeled with LTC 4 as well as wild-type MRP1. The altered transport properties of the Asp 1084 mutants have been attrib-uted to their inability to convert from a high to low affinity binding state (27).
The highly conserved Arg 1202 and Glu 1204 residues in TM16 are predicted to be well embedded in the membrane bilayer, which is usually considered to be an energetically unfavorable environment for ionizable amino acids (34). Mutants containing opposite charge substitutions of either of these residues were very poorly expressed, which suggests that the introduction of an opposite charge at either position 1202 or 1204 impairs insertion of MRP1 into the membrane bilayer or causes aberrant TM helix packing and misfolding, which presumably then targets the mutant protein for degradation. This idea is supported by the observation that membrane expression of the R1202D and E1204K mutants in mammalian cells can be substantially increased when transfected cells are grown at lower temperatures (30°C) where the stringency of the proofreading machinery for monitoring protein folding is diminished. 2 Because neutral (Leu, Gly) substitutions of TM16 Arg 1202 and Glu 1204 did not affect MRP1 expression in any significant way, it may be concluded that it is the opposite charge of the side chains of the substituting amino acids in the nonexpress- ing R1202D and E1204K mutants that is key to MRP1 protein destabilization. In this regard, it is worth noting that replacing Arg 1202 with an Asp (or Glu 1204 with a Lys) results in a potential net gain of two charges in the cytoplasmic half of TM16 as well as placing two same-charge ionizable residues in relatively close proximity to one another which could well perturb the ␣-helical geometry of TM16 and contribute to misfolding or aberrant TM helix-packing.
In addition to being readily expressed, the neutrally substituted mutants of TM16 Arg 1202 (R1202G and R1202L) exhibited transport activities that were, in the case of most substrates, similar to those of wild-type MRP1. Our findings are consistent with a recent report that a neutrally substituted Arg 1202 mutant (R1202G) could still transport LTC 4 (28). Most interestingly, when the analogous residue in human MRP2 (Arg 1210 ) was mutated to Ala, the ability of MRP2 to mediate cellular efflux of a fluorescent GSH conjugate was substantially reduced (35). Whether or not this reflects a different role for this TM16 Arg residue in GSH conjugate binding and transport by MRP1 and MRP2 remains to be determined.
Unlike the neutrally substituted Arg 1202 mutants, transport of organic anions by the neutrally substituted Glu 1204 mutant E1204L was substantially reduced or eliminated with the exception of MTX. Nevertheless, the substrate (LTC 4 )-binding site of E1204L remained intact. Furthermore, GSH transport remained very low, although other MRP1 transport activities of the same-charge E1204D mutant were comparable with wildtype MRP1. Thus, on the one hand, MRP1 can accommodate significant changes (except an opposite charge) at position 1202 and still retain some of its transport activities. On the other hand, both the acidic character and the volume of Glu 1204 need to be preserved for full MRP1 function. These findings suggest that the TM16 Glu 1204 side chain is critical for establishing interhelical hydrogen bonding or ion pair interactions that are important for substrate transport as well as stable membrane expression of the MRP1 protein.
Previous studies have indicated that TM16 may form part of a drug-binding site on MRP1 (18,20). Our present data indicate that Glu 1204 is not an essential component of this site but instead is more likely to be involved in a step of the transport process subsequent to initial substrate binding. This was shown to be the case when the nucleotide interactions of the transport-compromised E1204L mutant were examined. Previous studies have shown that the NBDs of MRP1, unlike Pglycoprotein, do not contribute equally to the activity of the  (39). The location of the ionizable residues in or proximal to the cytoplasmic face of MSD3 shown in this study and shown previously (27) to be important for MRP1 expression and/or function is indicated. B, three-dimensional surface representations of the atomic model of the core membrane spanning regions of MRP1 (MSD2-TM6 -11 (magenta flat ribbon); MSD3-TM12-17 (deep purple flat ribbon)) (39) showing a side view (left) and a view onto the cytoplasmic face (right). The surface (translucent gray) was generated using a probe radius of 1.40 Å (ViewerPro 4.2 software, www.accelrys. com/) rolling over the TM residue side chains. Left, the distance between the ␣-carbons of the amino acids predicted to be at the ends of the TMs is ϳ35 Å, consistent with the thickness of the hydrophobic core (30 -40 Å) of the lipid bilayer of a mammalian cell plasma membrane. Right, the cytoplasmic view shows how the mutation-sensitive TM16 Arg 1197 (orange), TM16 Glu 1204 (yellow), and TM17 Arg 1249 (green) residues could all face into the substrate translocation pathway of MRP1. In the view shown, the cytoplasmic domains and parts of the TM helices in the outer leaflet of the lipid bilayer have been "hidden" to enhance the visibility of the mutation-sensitive ionizable residues.
transporter (36,37). Thus, inactivation of NBD2 abolishes transport by MRP1, but inactivation of NBD1 results in only a partial loss of activity. Our demonstration that vanadate-induced trapping of azido-ADP by the mutant E1204L protein (and specifically by NBD2) was substantially increased suggests that the mutation may impair the ability of NBD2 to release ADP after hydrolysis of ATP, which could in turn impair substrate translocation and/or release. Alternatively, the E1204L mutant may hydrolyze ATP and release ADP very rapidly in the absence of vanadate but be unable to proceed through a second catalytic cycle (36). In effect, the mutation may diminish the coupling of the catalytic activity of MRP1 to transport in a way that affects some substrates more than others. Further studies are underway to distinguish among these possibilities. As mentioned previously, Glu 1204 is located well within the membrane bilayer in the putative substrate translocation pathway of MRP1. Thus, the altered catalytic activity and impaired transport of the E1204L mutant suggests that Glu 1204 (or at least the region in which it resides) could play a role in the signaling between the substrate translocation pathway and NBD2.
Unlike Arg 1202 and Glu 1204 , substitution of Arg 1197 and Arg 1249 with oppositely charged residues did not adversely affect expression of MRP1 but instead caused a substantial reduction in transport activity. Our observations with respect to the R1249D mutant are consistent with those of Ren et al. (19) who reported that Ala substitution of Arg 1249 impaired MRP1-mediated LTC 4 transport and reduced vincristine resistance. However, we also found that the same-charge mutants of Arg 1197 and Arg 1249 were essentially transport-inactive and furthermore could no longer be photolabeled with LTC 4 . These findings were unexpected because same-charge mutations of other basic and acidic residues in MSD2 and MSD3 have generally been reported to retain substrate binding and transport activities of the protein (24,27). The lack of LTC 4 labeling of the R1197K and R1249K mutants indicates that their binding site for LTC 4 (and likely other organic anions) has been disrupted. Thus, Arg 1197 and Arg 1249 , which are predicted to be at the membrane-cytosol interface of TM16 and TM17, respectively, act as more than just topological determinants and may well be critical for maintaining the architecture of the substrate binding site(s) of MRP1 (38). The bulkier, less ionizable Lys side chains in the R1197K and R1249K mutants presumably either cannot form or are prevented from forming the interhelical and/or intrahelical interactions that are established by Arg in wild-type MRP1 for proper folding into a functional transporter. However, our data show that disruption of these interactions does not destabilize MRP1 to the point where its expression in the membrane is impaired, nor does it appear to affect the catalytic activity of the transporter.
The precise arrangement and boundaries of the 17 TM ␣-helices of MRP1 are not yet known, although several models have been proposed. We have recently derived an atomic model of the 12 TM helices that form MSD2 and MSD3 by computational simulations using an energy-minimized P-glycoprotein simulation in combination with homology modeling based on the crystal structure of the Vibrio cholerae lipid transporter MsbA as a template (39,40). According to our model, TM16 and TM17 are expected to be significantly closer to each other at the cytoplasmic face of the membrane than at the extracellular face because of the tilt angle of the TM helices (Fig. 7B). Our model also predicts that the mutation-sensitive residues Glu 1204 , Arg 1197 , and Arg 1249 identified in the present study may all face in toward a cavity that has been proposed as the substrate translocation pathway of MRP1 (Fig. 7B) (39). Thus, these residues would be well positioned to participate in hydrogenbonding and other electrostatic interactions with polar and/or ionizable residues in other nearby TM helices of MSD2 and MSD3, such as TM7 and TM12, and possibly directly with MRP1 substrates as well (38 -43).
As mentioned previously, the ionizable residues targeted in this study were originally selected because of their predicted location in or close to the cytosol-membrane interface of MSD3 according to the algorithm of Eisenberg et al. (33). Our recently generated atomic model of MSD2 and MSD3 places three additional basic residues (Arg 1138 , Lys 1141 , and Arg 1142 ) close to the NH 2 -proximal end of the cytoplasmic loop (CL7) near the cytosol-membrane interface of TM15 (39) (Fig. 7A). In addition, we have noted that CL7, which is extensively ␣-helical, contains several clusters of ionizable amino acids. Initial characterization of MRP1 mutants containing substitutions of these residues shows that at least some of them are critical for expression and/or organic anion transport. 3 Ongoing studies are aimed at a better understanding of the role that ionizable residues in CL7 and other cytoplasmic loops play in the transport mechanism and substrate specificity of MRP1.