Identification and Characterization of Functionally Important Elements in the Multidrug Resistance Protein 1 COOH-terminal Region*

The ATP binding cassette (ABC) transporter, multidrug resistance protein 1 (MRP1/ABCC1), transports a broad spectrum of conjugated and unconjugated compounds, including natural product chemotherapeutic agents. In this study, we have investigated the importance of the COOH-terminal region of MRP1 for transport activity and basolateral plasma membrane trafficking. The COOH-terminal regions of some ABCC proteins have been implicated in protein trafficking, but the function of this region of MRP1 has not been defined. In contrast to results obtained with other ABCC proteins, we found that the COOH-proximal 30 amino acids of MRP1 can be removed without affecting trafficking to basolateral membranes. However, the truncated protein is inactive. Furthermore, removal of as few as 4 COOH-terminal amino acids profoundly decreases transport activity. Although amino acid sequence conservation of the COOH-terminal regions of ABC proteins is low, secondary structure predictions indicate that they consist of a broadly conserved helix-sheet-sheet-helix-helix structure. Consistent with a conservation of secondary and tertiary structure, MRP1 hybrids containing the COOH-terminal regions of either the homologous MRP2 or the distantly related P-glycoprotein were fully active and trafficked normally. Using mutated proteins, we have identified structural elements containing five conserved hydrophobic amino acids that are required for activity. We show that these are important for binding and hydrolysis of ATP by nucleotide binding domain 2. Based on crystal structures of several ABC proteins, we suggest that the conserved amino acids may stabilize a helical bundle formed by the COOH-terminal three helices and may contribute to interactions between the COOH-terminal region and the protein's two nucleotide binding domains.

Multidrug resistance protein 1 (MRP1/ABCC1) 1 is a member of the C branch of the ATP binding cassette (ABC) transporter family that was discovered by virtue of its overexpression in a multidrug-resistant, doxorubicin-selected human small cell lung cancer cell line, H69AR (1). The protein has been shown to confer multidrug resistance to structurally diverse, natural product cytotoxic drugs and to mediate their efflux in an ATPdependent manner (1,2). MRP1 has also been detected in tumors with many different cellular origins (3). Although the spectrum of drugs to which MRP1 confers resistance is similar to that of P-glycoprotein (P-gp) (4), the two transporters differ functionally in some important respects. For example, MRP1 but not P-gp transports a wide range of conjugated hydrophilic molecules such as the GSH conjugate leukotriene C 4 (LTC 4 ), 17␤-estradiol-17-␤-D-glucuronide (E 2 17␤G), and estrone 3-sulfate, as well as many other substrates, the transport of which is either stimulated by GSH or GSH-dependent (5)(6)(7)(8)(9)(10)(11)(12)(13).
Other members of the ABCC branch of the superfamily include the more recently identified MRP2 to -9 as well as the cystic fibrosis transmembrane conductance regulator (CFTR), the sulfonylurea receptors (SUR1 and SUR2), and ABCC13 (2,14). Structurally, the "core" region of the ABCC proteins is similar to that of many eukaryotic ABC proteins, with the exception of ABCC13, and consists of two membrane-spanning domains (MSD), each containing six transmembrane helices followed by a nucleotide binding domain (NBD). MRPs 1, 2, 3, 6, and 7 and SUR1 and SUR2 are atypical among ABC proteins in that they have a third NH 2 -terminal MSD (MSD1) that, based on topology predictions and, in some cases, strong experimental data, contains five transmembrane helices (15)(16)(17)(18)(19). This domain is required for the trafficking or activity of MRP2 and the SURs but in MRP1 appears to be dispensable for the binding and transport of at least some substrates, and its function is presently not well defined (20 -26). However, a region of the cytoplasmic loop (cytoplasmic loop 3) of MRP1 that connects MSD1 to the remainder of the protein is essential for both substrate binding and appropriate protein trafficking (22,23,25,26). Based on alignment with ABCC proteins lacking the third MSD, the region in cytoplasmic loop 3 begins at a location that aligns closely with the predicted NH 2 terminus of the core structure of MRP1 (27).
In this study, we have examined the importance of the COOH-terminal region of MRP1 for the trafficking and activity of the protein. In some ABCC proteins, such as MRP2, CFTR, and SUR1, mutations in, or deletions of, their COOH-terminal regions have been shown to affect targeting to the plasma membrane (28 -30). Mutations of ABC proteins that result in aberrant trafficking in mammalian cells frequently result in retention and rapid degradation in the endoplasmic reticulum (22,25,28,31). This precludes analysis of their effect on the function of the protein. However, when expressed in insect cells, some of these mutant proteins are stable and active, indicating that the mutations have not caused gross perturbations in the structure of the protein (25,31). These observations are presumed to be attributable to the lower temperatures at which the insect cells are cultured and a less stringent proofreading machinery for monitoring protein folding in the endoplasmic reticulum (31). Consequently, we have carried out parallel studies of the COOH-terminal region of MRP1 using polarized Madin-Darby canine kidney (MDCK)-I cells to screen for mutations and deletions that affect protein trafficking and insect Sf21 cells to examine the effects of these structural alterations on protein function. In contrast to results obtained with other ABCC proteins (28,29), we found that removal of up to 30 amino acids from the COOH terminus had no effect on protein trafficking or binding of a substrate such as LTC 4 but essentially eliminated transport activity. Furthermore, deletion of as few as 4 amino acids from the COOH terminus was sufficient to decrease activity by ϳ75%. We show that this reduction in transport function is accompanied by a decrease in ATP binding and hydrolysis by NBD2 and an impaired ability to shift from a high to low affinity substrate binding state. Despite little amino acid sequence identity, transport activity can be restored by substitution of the COOH-terminal region of MRP1, with the comparable regions of MRP2 and the distantly related P-gp, indicating that the secondary and tertiary structures of these regions must be very similar. Alanine scanning mutagenesis revealed that five hydrophobic residues within the COOH-terminal region that are broadly conserved among ABC proteins are critical for activity of MRP1. Comparison of the predicted tertiary structure of MRP1 in this region with the crystal structures of distantly related ABC proteins suggests that these residues contribute to the stability of a highly conserved ␣-helical bundle, elements of which are in close proximity to functionally important residues in both NBDs.
COOH-terminal hybrid constructs were generated by recombinant PCR. The MRP1/MRP2 COOH-terminal chimeric fragments were generated by first amplifying MRP1 nt 3152-4395 with the forward primer D1241 and the reverse primer 5Ј-GGTCGTCTGAATGAGGTTGTC-CGTTTCCAGGTCCACGGC-3Ј (MRP2 nt 4434 -4417 underlined) with pFB-MRP1 as the template. The MRP2 COOH-terminal 219 base pairs were then amplified from pCDNA3.1-MRP2 (34) with the recombinant forward primer 5Ј-GCCGTGGACCTGGAAACGGACAACCTCATTC-AGACGACC-3Ј (MRP1 nt 4375-4394 underlined) and the reverse primer 5Ј-CACGTGGACTAGAATTTTGTGCTGTTCACATTC-3Ј that contained a stop codon (TAG) after MRP2 nt 4680 incorporated into a SalI site (underlined). The MRP1/P-gp COOH-terminal fragments were generated by first amplifying MRP1 nt 3152-4527 with the forward primer D1241 and the recombinant reverse primer 5Ј-CTGCTGAT-GCGTGCCGTACTCCTGGATTTCTCC-3Ј (P-gp nt 3776 -2766 underlined) with pFB-MRP1 as a template. Full-length P-gp cDNA was generated by PCR with reverse transcribed A549 mRNA as a template and subsequently cloned into pFB (pFB-P-gp). The fidelity of the COOH-terminal region of P-gp was confirmed by sequencing. Amplification of the COOH-terminal 75 nt from P-gp using pFB-P-gp as a template was achieved with the forward recombinant primer 5Ј-CAG-GAGTACGGCACGCATCAGCAGCTG C-3Ј (MRP1 nt 4516 -4527 underlined) and the reverse primer 5Ј-TATGTCGACTCACTGGCG CTT-GTTCC-3Ј that contained a stop codon (TAG) after the last coding sequence of P-gp followed by a SalI site (underlined). Amplified COOHterminal DNA products for the MRP1/MRP2 and the MRP1/P-gp hybrids were denatured and permitted to anneal to the respective recombinant PCR fragments prior to the addition of deoxyribonucleotides and the Klenow fragment to polymerize the hybridized PCR fragments. The recombinant DNA was then amplified by PCR using the D1241 forward primer and the MRP2 or P-gp COOH-terminal reverse primers. Subsequently, the MRP1/MRP2 and MRP1/P-gp COOH-terminal DNA fragments were digested with ClaI and SalI and cloned into the pBluescript-KSϩ (pBS) (Stratagene, La Jolla, CA) vector after ClaI/SalI digestion. The fidelity of the coding sequence of pBS-MRP1 1353-1465 /MRP2 1473-1545 and pBS-MRP1 1353-1509 /P-gp 1255-1280 hybrid DNA were confirmed by sequencing. pBS-MRP1 1353-1465 /MRP2 1473-1545 was digested with ClaI and KpnI and cloned into pFB-MRP1 as described above for the MRP1 COOH-terminal truncation constructs. MRP1 1-1509 /P-gp 1255-1280 was assembled in pBS by excising the MRP1 NH 2 -terminal region from pFB-MRP1 with SacI and ClaI and ligating the 4.0-kb fragment into pBS-MRP1 1353-1509 /P-gp 1255-1280 digested with SacI/ClaI. The MRP1/ P-gp COOH-terminal hybrid construct in pBS was then digested with EagI/KpnI and inserted between the NotI and KpnI sites in pFB. Chimeric constructs were subsequently moved into NotI-KpnI digested pCDNA3.1(Ϫ) via an EagI-KpnI digest of the pFB DNA.
All COOH-terminal mutations were generated using the QuikChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA). MRP1 template DNA was prepared by digesting pFB-MRP1 with EcoRI and KpnI and cloning the COOH-terminal 713 base pairs into pBS. Oligonucleotides bearing mismatched bases at the COOH-terminal residues to be mutated were generated, and mutagenesis was performed according to the instructions from the manufacturer. Mutations were confirmed by sequencing. DNA fragments containing the desired mutations were cloned into the pFB-MRP1 as described above for the MRP1 COOH-terminal truncation constructs, and the fidelity of constructs was confirmed by sequencing.
Expression and Localization of MRP1 in MDCK-I Cells-Generation of MDCK-I cells stably expressing COOH-terminal truncation constructs cloned into the pCDNA3.1(Ϫ) vector has been previously described (25). For localization studies, stable expressing MDCK-I cells were seeded on glass coverslips and polycarbonate filters and grown for 10 -14 days at 37°C to allow polarization. In some cases, after cells had polarized on glass coverslips and polycarbonate filters they were incubated at 27°C for 48 h. Ethanol fixation of MDCK-I cells and immunofluorescent detection with the mAb MRPr1 and either the mAb anticalnexin or a polyclonal antibody Na ϩ /K ϩ -ATPase have been previously described (25). mAb MRPr1 was immunodetected with Alexa 488 goat anti-rat, whereas Alexa 546 goat anti-rabbit and Alexa 546 goat antichicken were used to detect antibody interactions with the cellular markers calnexin and Na ϩ /K ϩ ATPase, respectively. The fluorescent secondary antibodies and the 7-aminoactinomycin D nuclear stain were excited with a Leica TCS SP2 dual photon confocal microscope.
Viral Infection, Membrane Vesicle Preparation, and Immunoblotting-Baculoviruses generated from recombinant bacmids were used to infect Sf21 insect cells as described previously (32). To generate membrane vesicles, insect cells were disrupted by nitrogen cavitation, and the vesicles were isolated by sucrose density centrifugation (7). Fulllength wild-type and COOH-terminal truncated, mutated, and hybrid membrane proteins were subjected to SDS-PAGE (7.5%), whereas dual half-expressed proteins were resolved using SDS-PAGE gradient gels (7.5-15%). Immunoblots of intact proteins were probed with mAb MRPr1 (35,36), whereas the dual expressed NH 2 -half was immunodetected using mAb MRPr1, and the COOH-half was immunodetected with mAb MRPm5 (37). Antibody-protein interactions were detected using horseradish peroxidase secondary antibodies utilizing an enhanced chemiluminescence, and the relative expression of MRP1 fragments was determined by densitometry.
LTC 4 and E 2 17␤G Transport into Sf21 Membrane Vesicles-Uptake of [ 3 H]LTC 4 (50 nM) and [ 3 H]E 2 17␤G (400 nM) was measured at 23 and 37°C, respectively, as described previously (6,32). Briefly, 4 g (LTC 4 uptake) or 8 g (E 2 17␤G uptake) of Sf21 membrane vesicles were incubated with tritiated substrates in the presence of 4 mM ATP or AMP and 10 mM MgCl 2 in TB (50 mM Tris, pH 7.4, 250 mM sucrose). ATPdependent transport was determined by subtracting uptake in the presence of ATP from uptake in the presence of AMP. For ATP-dependent kinetic determinations, [ 3 H]LTC 4 (150 nM, 182 Ci mmol Ϫ1 ) uptake after 1 min was measured at different ATP or AMP concentrations (25-2000 M) and plotted by the Hanes-Woolf method.

Photoaffinity Labeling of MRP1 Constructs with [ 3 H]LTC 4 and [ 125 I]LY475776 -Labeling of MRP fragments by [ 3 H]LTC 4 and
[ 125 I]LY4775776 was carried out as described previously (24,38). Briefly, Sf21 membrane vesicles (75 g of protein) were incubated with 200 nM [ 3 H]LTC 4 (0.25 Ci) for 30 min at room temperature. Samples were frozen in liquid nitrogen and immediately UV-irradiated at 302 nm for 30 s and repeated 10 times. For [ 125 I]LY475776 labeling, membrane vesicles (75 g of protein) were incubated with the MRP1 inhibitor (0.5 nM) at 37°C for 10 min before UV irradiation for 5 min at 302 nm on ice. Vesicle proteins labeled with radioactive LTC 4 and LY475776 were then resolved on SDS-PAGE (7.5%). Gels were dried for 2 h at 80°C and exposed to film.
Photolabeling of MRP1 NBDs Using 8-Azido-[␣-32 P]ATP-8-Azido-[␥-32 P]ATP and 8-azido-[␣-32 P]ATP photolabeling was performed as described previously (33). Briefly, Sf21 membrane vesicles (20 or 30 g of protein) were incubated with 5 M 8-azido-[␣-32 P]ATP in TB containing 5 mM MgCl 2 for 5 min on ice, followed by UV irradiation at 302 nm for 8 min on ice. The reaction was stopped, and free radioactive nucleotide was removed by the addition of 500 l of cold wash buffer (50 mM Tris-HCl, pH 7.4, 0.1 mM EGTA, 5 mM MgCl 2 ) before centrifugation at 14,000 rpm. The wash was repeated, and the pelleted vesicles were resuspended in Laemmli buffer (4ϫ) containing dithiothreitol (100 mM). Under conditions where MRP1 wild-type and COOH-terminal truncated proteins were proteolytically digested, UV-irradiated radioactive nucleotide-labeled vesicle proteins were resuspended in TB containing 20 ng of trypsin (1:1000) and incubated for 30 min at 37°C. Proteins were resolved on gradient SDS-PAGE (7.5-15%), and the gels were dried for 2 h at 80°C before exposure to film. For K d determinations, vesicle proteins were labeled with various concentrations of 8-azido-[␣-32 P]ATP (1-128 M). Densitometry was performed to determine the radioactivity incorporated into the NBDs. Because of the higher concentrations of 8-azido-[␣-32 P]ATP that were required to saturate the COOH-terminal NBD of the truncated proteins in K d determinations, binding was corrected for the background observed in a parallel experiment using membrane vesicles prepared from Sf21 cells infected with a ␤-gus expression vector.
Vanadate and Beryllium Fluoride Trapping of 8-Azido-[␣-32 P]ATP at 37°C-ADP orthovanadate and beryllium fluoride trapping was performed as described previously (33,39). Briefly, Sf21 membrane vesi-cles (20 g of protein) were incubated with of 8-azido-[␣-32 P]ATP (5-45 M) for 15 min in TB containing 5 mM MgCl 2 for 15 min at 37°C in the presence or absence of 1 mM orthovanadate or 200 M beryllium fluoride. Wash buffer was added to the samples, followed by centrifugation, and the wash was repeated. The pelleted vesicles were resuspended in 15 l of wash buffer and transferred to a 96-well plate, after which the samples were UV-irradiated at 302 nm for 8 min on ice. Vesicle proteins were then diluted in Laemmli buffer (4ϫ) containing dithiothreitol (100 mM) and resolved on gradient SDS-PAGE (7.5-15%). The gels were dried for 2 h at 80°C before audioradiography.
Photoaffinity Labeling with [ 3 H]LTC 4 in the presence of ATP, ATP and Orthovanadate, or ATP␥S-Photoaffinity labeling of MRP fragments by LTC 4 in the presence of nucleotides has been described previously (24,39). Briefly, membrane vesicles (50 g) were incubated with either ATP (4 mM), ATP (4 mM), and orthovanadate (1 mM) or with ATP␥S (4 mM) with 5 mM MgCl 2 in TB for 30 min at room temperature, followed by incubation with 200 nM [ 3 H]LTC 4 for 30 min at room temperature. Subsequently, proteins were UV-irradiated and resolved on SDS-PAGE (7.5%), and gels were dried before exposure to film.

Plasma Membrane Trafficking by COOH-terminal MRP1
Truncated Proteins in MDCK-I Cells-To determine whether or not the MRP1 COOH-terminal region is required for plasma membrane localization, we prepared stable MDCK-I transfectants expressing various MRP1 COOH-terminal truncated proteins (Fig. 1A). Deletion of the COOH-terminal 4, 9, 22 (not shown), or 30 amino acids of the protein did not prevent trafficking to the basolateral membranes in MDCK-I cells (Fig.  1B). However, further truncation to Ile 1498 (Fig. 1B) or Asp 1493 (not shown) resulted in accumulation of the mutant proteins in intracellular membranes that stained positively for calnexin, a well established marker for the endoplasmic reticulum (40).
Previously, it has been demonstrated that short term growth at 27°C can promote surface expression of CFTR mutant proteins that display trafficking defects (31). Growth under such conditions did not correct the trafficking defect of MRP1 1-1498 , as assessed by its lack of colocalization with the Na ϩ /K ϩ -ATPase (Fig. 1C), a known marker for the basolateral membrane in this cell line (25). It has recently been reported that Ala substitution of Val 1497 -Ile 1498 -Val 1499 -Leu 1500 in the COOH-terminal region of MRP1 impaired protein maturation in BHK-21 cells (41). The analogous tetramer of hydrophobic residues within the COOH-terminal region of CFTR, Phe 1413 -Leu 1414 -Val 1415 -Ile 1416 , has also been shown to be important for protein processing (41). Our findings support the importance of this motif and demonstrate that elimination of the COOH-terminal region to include Val 1499 and Leu 1500 is sufficient to disrupt normal processing in MDCK-I cells.
LTC 4 and E 2 17␤G Transport Activity by COOH-terminal Truncated MRP1-To assess the effect of COOH-terminal truncations on the activity of MRP1, the mutant proteins were expressed in Sf21 cells using a baculovirus expression system. Densitometry performed on immunoblots of membrane proteins from the infected Sf21 cells indicated that the MRP1 truncations were expressed at levels similar to the full-length protein, with the exception of MRP1 1-1493 , the levels of which were ϳ60% of wild-type ( Fig. 2A). In all data shown, ATP-dependent LTC 4 uptake was corrected by subtracting the low levels of endogenous activity present in membranes from control cells infected with a ␤-gus vector and then normalized for differences in expression levels (Fig. 2B). The transport activity of MRP1 1-1527 and MRP1 1-1522 was ϳ30% of that observed for wild-type MRP1. Further truncation to Gly 1509 reduced transport to ϳ15%, whereas MRP1 1-1501 , MRP1 1-1498 (not shown), and MRP1 1-1493 had less than 10% of the transport activity of the full-length protein (Fig. 2B). The relative rates of E 2 17␤G transport by the various truncated proteins were similar to those observed for LTC 4 , with the exception of MRP1 1-1522 , which, for reasons that are presently not known, consistently (C) and immunodetected with mAb MRPr1 and Alexa 488 goat anti-rat (green signal) using confocal microscopy as described under "Experimental retained 50% activity (Fig. 2C). Thus, although the COOHterminal 30 amino acids of MRP1 are dispensable as far as trafficking of the protein is concerned, elimination of only the COOH-terminal 4 amino acids has a major effect on transport activity. However, the addition of Ala 1528 -Val 1531 to the COOH terminus of partially active MRP1 1-1522 had no effect on transport of either LTC 4 or E 2 17␤G (not shown), indicating that the COOH-terminal 4 residues are not sufficient to restore transport activity to the more extensively deleted protein.
Transport Activity of MRP1/MRP2 and MRP1/P-glycoprotein COOH-terminal Hybrid Proteins-To determine whether MRP1-specific COOH-terminal elements were required for transport activity, we exchanged the residues immediately following the Walker B in NBD2 for the analogous but longer region of MRP2 (Fig. 3A). ATP-dependent LTC 4 uptake studies showed that the activity of the MRP1 1-1465 /MRP2 1473-1545 hybrid was similar to that of wild-type MRP1 (Fig. 3B), and trafficking was unaffected (not shown).
To determine whether primary sequence conservation between MRP1 and MRP2 contributed to the retention of activity by the MRP1 1-1465 /MRP2 1473-1545 hybrid, we made a similar construct in which the COOH-terminal region was exchanged for that of the more distantly related P-gp (MRP1 1-1509 /Pgp 1255-1280 ). After normalizing for protein expression levels (Fig. 3A), the rate of ATP-dependent LTC 4 uptake by the hy-brid was similar to that of wild-type MRP1 (Fig. 3B), and trafficking was unaffected (not shown). Thus, despite the lack of amino acid sequence conservation, the COOH-terminal region of P-gp, as observed with MRP2, can replace the tail region of MRP1 with little or no loss of activity.
Substrate and Inhibitor Binding by COOH-terminal Truncated MRP1 Proteins-To determine whether the effect of the COOH-terminal deletions on transport activity was a consequence of altered substrate binding, we carried out photoaffinity labeling experiments using LTC 4 and the recently developed, high affinity MRP1 inhibitor, LY475776 (38,42,43). We have shown previously that LTC 4 preferentially labels a site in MSD2 and to a lesser extent a site in MSD3 (24). In contrast, LY475776 displays GSH-dependent binding and almost exclusively labels a site in MSD3 (38,43). The results of these studies indicated that binding of substrate and inhibitor was not affected by any of the COOH-terminal truncations examined that extended to Asp 1501 (not shown).
Kinetic Analysis of ATP-dependent LTC 4 Uptake and 8-Azido-[␣-32 P]ATP Photolabeling-Given that no major changes in substrate binding were caused by the COOH-terminal truncations, we examined the possibility that loss of transport activity was attributable to decreased nucleotide binding and/or hydrolysis. Kinetic analysis of LTC 4 uptake by MRP1  revealed that the K m for ATP (495 M) was 3-fold higher than   4A). Furthermore, after normalizing for differences in protein expression levels, the V max for LTC 4 transport by the truncated protein was ϳ2.5-fold lower (75 pmol mg Ϫ1 min Ϫ1 ) than that of wild-type MRP1 (190 pmol mg Ϫ1 min Ϫ1 ). These results strongly suggest that the 70% reduction in LTC 4 and E 2 17␤G transport observed after removal of the COOH-terminal 4 amino acids is attributable to impaired ATP binding and/or hydrolysis. Similar analyses of the more extensively deleted proteins was precluded by the very low levels of residual transport activity.
To determine more directly whether or not COOH-terminal truncation altered ATP binding, partially active MRP1 1-1527 and inactive MRP1 1-1501 proteins were photolabeled with 8-azido-[␣-32 P]ATP at 4°C followed by digestion with trypsin. We and others have shown that under conditions of limiting digestion, it is possible to generate two major proteolytic fragments that contain either NBD1 or NBD2 (15,44,45). Based on immunoblotting with two MRP1-specific mAbs, limited trypsinolysis resulted in the generation of similar amounts of NH 2and COOH-proximal halves of the wild type and proteins COOH-terminally truncated by 4 or 30 amino acids (Fig. 4B). 8-Azido-[␣-32 P]ATP binding to NBD1 present in the larger of the two major tryptic fragments appeared unaltered by removal of the COOH-terminal 4 or 30 residues (Fig. 4C), whereas nucleotide binding to the smaller of the two major MRP1 1-1527 and MRP1 1-1501 fragments appeared weaker and barely detectable, respectively, when compared with labeling of the wild-type protein. Thus, COOH-terminal truncation appears to selectively impair nucleotide binding by NBD2.

Photolabeling of Dual Expressed MRP1 1-932 /MRP1 932-1527 with 8-Azido-[␣-32 P]ATP and 8-Azido-[␣-32 P]ADP under Nonhydrolytic
Conditions-To confirm that nucleotide interaction with NBD2 was altered by COOH-terminal truncation, we co-expressed the NH 2 -and COOH-proximal halves of both the wild-type protein and MRP1  . We have shown that it is possible to use a baculovirus dual expression system to produce stoichiometrically equivalent amounts of NH 2 -and COOHproximal fragments of MRP1 that reassociate to form a functional LTC 4 transporter with greater than 90% efficiency (24,32,33,39). Densitometry of immunoblots (Fig. 5A) indicated that both fragments of MRP1 1-1527 (MRP1 1-932 and MRP1 932-1527 ) were consistently expressed at 2-fold lower levels than those obtained with the vector encoding the two wild-type halves, MRP1 1-932 and MRP1 932-1531 . To compensate for this difference in expression levels, vesicles containing the wild-type half molecules were diluted with an equivalent amount, based on total protein, of vesicles from control cells infected with a ␤-gusϩCAT vector prior to analysis. LTC 4 transport assays were used to confirm that the addition of the control vesicles resulted in the predicted decrease in specific uptake activity (Fig. 5B).
Consistent with the activity observed for full-length MRP1 1-1527 , the LTC 4 uptake for the dual half-expressed mutant protein was ϳ30% of the 2-fold diluted dhMRP1 1-1531 vesicles (dh1-1531) (Fig. 5B). The effect of COOH-terminal truncation on the affinity for ATP was determined by examining 8-azido-[␣-32 P]ATP binding at 4°C as a function of the concentration of nucleotide (Fig. 5C). The K d values for NBD1 of dhMRP1 1-1531 and dhMRP1 1-1527 NBD1 were 18.3 and 16.4 M, respectively, in agreement with a previously reported value (46). In contrast, removal of only 4 amino acids caused a 2-fold reduction in nucleotide affinity at NBD2 (dhMRP1 1-1527 K d(NBD2) ϭ 61.3 M) in comparison with the wild-type protein (K d(NBD2) ϭ 28.7 M).
To determine whether the reduced photolabeling of NBD2 of MRP1 1-1527 was the result of a decreased ability to stabilize binding via interactions with the ␥-phosphate of ATP, we carried out similar experiments with 8-azido-[␣-32 P]ADP. Previously, we have shown that 8-azido-[␣-32 P]ADP strongly photolabels NBD2 of dhMRP1  in the presence of vanadate at 4°C (39). ADP binding at NBD2 of MRP1 1-1527 was weak in comparison with the strong binding of the wild-type NBD2 (Fig. 5D). Thus, removal of the COOH-terminal 4 amino acids decreases binding of both ATP and ADP to NBD2.
Nucleotide Trapping by Wild-type dhMRP1 and dhMRP1 1-1527 at 37°C-Since ATP and ADP binding by NBD2 was impaired at 4°C following removal of the COOH-terminal 4 amino acids, we next determined whether ATP hydrolysis was also affected. Comparison of the dhMRP1 1-1531 and dhMRP1 1-1527 proteins following vanadate trapping studies at 37°C showed that ADP trapping was almost completely abolished at NBD2 of the COOH-terminal truncated protein (Fig. 6A). Similar experiments were also performed with beryllium fluoride, since ADP trapped in the presence of beryllium is believed to mimic an earlier step in catalysis and to correspond to an ATP binding ground state, as opposed to the posthydrolytic transition state formed in the presence of vanadate (47). Despite the fact that trapping was carried out at concentrations of 8-azido-[␣-32 P]ATP as high as 45 M, no trapping at NBD2 of dhMRP1 1-1527 was detected (Fig. 6B). Comparison of the approximate 2-fold reduction in ATP binding at NBD2 of MRP1 1-1527 with the almost complete abolition of ADP trapping in the presence of either beryllium and vanadate suggests that hydrolysis is also compromised. This observation is consistent with the ϳ2.5-fold reduction in V max for LTC 4 transport by the truncated protein (Fig. 4A).
LTC 4 Photolabeling in the Presence of Nucleotides-Previously, we have shown that the occupancy of NBD2 by either ATP or ADP is sufficient to convert MRP1 from a high to low affinity LTC 4 binding state (24, 39). As observed before (39), preincubation of wild-type MRP1 1-1531 vesicles with ATP alone or ATP plus vanadate had marked effects on LTC 4 photolabel-ing (Fig. 7, A and B). Although ATP in the absence or presence of vanadate decreased LTC 4 binding by mutant MRP1 1-1527 , the reduction was less pronounced than observed either with the full-length wild-type protein or with the fully active hybrid protein containing the COOH tail of P-gp (MRP1 1-1509 /Pgp 1255-1280 ). We have also shown previously that the poorly hydrolyzable nucleotide analogue, ATP␥S, can decrease the binding of LTC 4 by MRP1 although not as effectively as ATP  (39). As observed with ATP, the effect of ATP␥S on LTC 4 binding by MRP1 1-1527 was diminished when compared with the wild-type or MRP1/P-gp hybrid protein (Fig. 7A). Overall, these results also suggest that removal of the COOH-terminal 4 amino acids decreased but did not completely eliminate the ability of NBD2 to bind ATP, consistent with the increased K m observed in ATP-dependent LTC 4 transport studies. Further-more, preincubation of MRP1 1-1501 with ATP and vanadate or with ATP␥S did not decrease LTC 4 photolabeling when compared with that observed in the absence of nucleotides (Fig.  7B). Together with the nucleotide photolabeling and transport activity studies, these results suggest that truncation of the COOH terminus to Asp 1501 results in a loss of the ability of NBD2 to bind ATP.
Effect of Point Mutations in the COOH-terminal Region on the Function of MRP1-To examine the extent to which individual amino acids within the COOH-terminal region are functionally important, we generated an extensive bank of point mutations and assessed their effect on the ability of the protein to transport LTC 4 ( Table I). The relatively few residues conserved in distantly related ABC proteins as well other amino acids located within predicted ␣-helical and ␤-sheet structures were mutated to Ala (Fig. 8, Table I).
Many of the single mutations had no effect on function. The only mutations that significantly decreased LTC 4 transport activity involved two conserved residues in helix C3, Leu 1488 and Ile 1491 ; the COOH-terminal residue of the conserved "hydrophobic patch" motif in sheet C2, Leu 1500 ; and the highly conserved aromatic amino acid, Phe 1521 , predicted to be at the start of helix C1. The L1488A, I1491A, L1500A, and F1521A mutations reduced uptake by ϳ50, 40, 80, and 60%, respectively (Table I). Mutation of one residue in the middle of the predicted helix C1, Met 1524 , increased LTC 4 transport by ϳ30%.
In addition to single mutations, Ala substitutions were made of the conserved hydrophobic tetramer Leu 1497 -Ile 1498 -Val 1499 -Leu 1500 and the dileucine motif in predicted helix C2. The L1497A/I1498A/V1499A/L1500A mutation essentially inactivated the protein, and the double L1514A/L1515A mutation reduced transport by 80%. Similar to the MRP1 1-1527 , the L1514A/L1515A mutant had a 4-fold increase in the K m for ATP and an approximate 2.5-fold decrease in the V max for LTC 4 transport. In contrast, the D1527A/G1529A double mutation increased LTC 4 uptake by ϳ25% (Table I). Comparison with more distantly related ABC proteins indicated that Val and Ile are also found at one or the other of the positions of the highly conserved helix C2 residues, Leu 1514 and Leu 1515 (Fig. 8). Consequently, we created two additional double mutations, L1514V/L1515V and L1514I/L1515I. The former decreased LTC 4 transport activity by ϳ25% and the latter by only ϳ5% (Table I). These results confirm the requirement for hydrophobic amino acids with side chains larger than that of Ala at this location. Similarly, the residue present at the analogous position of MRP1 Phe 1521 in other ABC proteins is typically aromatic. Consistent with a requirement for an aromatic side chain, mutation of Phe 1521 to Tyr had no effect on LTC 4 transport, whereas substitution with Gly, Gln, or Glu reduced LTC 4 transport activity by ϳ60, ϳ70, and ϳ75%, respectively (Table I).
Since mutation of Met 1524 in helix C1 to Ala increased LTC 4 transport activity, we also mutated this residue to Gly, which, unlike Ala, would be expected to decrease the stability of the helix. In contrast to Ala substitution, the M1524G mutation decreased LTC 4 transport by ϳ35%. Furthermore, although a single mutation of the adjacent Ala 1524 to Gly did not affect activity, the double mutation M1524G/A1525G reduced LTC 4 transport by ϳ75% (Table I). Kinetic analysis of this mutant showed that the K m for ATP was increased by 4.5-fold, whereas the V max for LTC 4 transport was 3-fold lower (Table I). Photoaffinity binding studies indicate that this double glycine mu-  tation impaired nucleotide binding and trapping at NBD2 to an extent comparable with that resulting from removal of the COOH-terminal 4 amino acids (not shown). These observations strongly suggest that the integrity of the ␣-helix at the COOH terminus of MRP1 is important for efficient nucleotide binding and possibly hydrolysis by NBD2.

DISCUSSION
Previous studies have shown that the COOH-terminal regions of the ABCC proteins MRP2 and SUR1 are important for protein trafficking and/or maturation (28,29). In contrast, up to 60 amino acids can be eliminated from the COOH terminus of CFTR without affecting protein maturation (48). However, it has also been reported that the terminal PDZ motif, which is in this region, is important for apical membrane localization (30). We found that the COOH-terminal region of MRP1 was not required for trafficking, but unlike CFTR, elimination of as few as 4 amino acids markedly reduced activity (41). Despite this observation, exchange of the COOH-terminal region of MRP1 for comparable regions of MRP2 and P-gp did not affect either activity or basolateral trafficking of the hybrid proteins. This result, particularly in the case of the hybrid involving the distantly related P-gp, strongly suggests that the higher order structure of the COOH-terminal regions of these proteins are sufficiently conserved to permit retention of function, despite their relatively low amino acid sequence conservation.
Sequence alignments of the COOH-terminal regions of a number of prokaryotic and human ABC transporters are shown in Fig. 8. Secondary structure algorithms predict a conserved helix-sheet-sheet-helix-helix motif in the region COOH-proximal to the highly conserved "switch" histidine in NBD2 (Fig. 8).
Comparison with analogous regions of the crystal structures of several distantly related ABC proteins reveals broad conservation of these secondary structure elements. Based on the model depicted in Fig. 8, the studies of MRP2 and SUR1 suggest that elimination of part of helix C1 together with the downstream COOH-terminal region of the protein is sufficient to impair localization to the plasma membrane (28,29). In contrast, elimination of helix C1 of MRP1 by truncation to amino acid 1522 or even more extensive truncation to Asp 1501 did not prevent targeting to the basolateral plasma membrane in polarized cells, indicating that necessary trafficking signals must reside elsewhere in the protein. Thus, despite the predicted conservation of secondary structure, the requirement for various conserved elements in the COOH-terminal region for protein processing and/or plasma membrane trafficking differs among the ABCC proteins. Further truncation of MRP1 to Ile 1498 did result in retention in the endoplasmic reticulum, and it has been shown recently that mutation of the hydrophobic tetramer Val 1497 -Ile 1498 -Val 1499 -Leu 1500 in MRP1 as well as the corresponding Phe-Leu-Val-Ile sequence in CFTR impaired maturation of both proteins (41). Together, these results indicate that the integrity of sheet C2 is critical for trafficking of MRP1, whereas the COOH-terminal sheet C1 and ␣-helices C2 and C1 are not essential (Fig. 8).
Studies with MRP1 and other ABC proteins, notably CFTR, have shown that mutant proteins that fail to traffic in mammalian cells may be functional when expressed in insect cells (25,31). At present, it is not known whether COOH-terminal truncations of MRP2 and SUR1 that prevent trafficking also affect the activity of the protein, since comparable studies in insect cells have not been reported (28,29). Nevertheless, the results described here represent the converse situation in which truncations of the COOH-terminal region of MRP1 that do not affect trafficking profoundly decrease activity. Notably, removal of only the 4 (MRP1 1-1527 ) or 9 (MRP 1-1522 ) COOHterminal amino acids reduced transport activity by ϳ70%.
Since elimination of either 4 or 9 amino acids had a similar effect, we examined whether the COOH-terminal 4 amino acids might comprise a discrete element required for activity. However, their addition to MRP1 1-1522 failed to restore function (not shown). Consequently, it appears likely that the decrease in activity of the protein following elimination of the COOHterminal 4 or 9 amino acids is a result of destabilizing helix C1. This suggestion is further supported by the effect of introducing destabilizing Gly mutations into this helix, as exemplified by the double M1524G/A1525G mutation, which had the same effect on activity as deletion of the terminal 4 or 9 amino acids.
Truncation of MRP1 to Asp 1501 , which removes sheet C1, helix C2, and helix C1 and eliminates transport activity, did not affect substrate binding, as evidenced by photolabeling with LTC 4 and the MRP1 reversing agent, LY475776 (not shown). Rather, the loss of activity following COOH-terminal truncation appears to be the consequence of a reduced ability of NBD2 to bind nucleotide and hydrolyze ATP. Removal of the terminal 4 amino acids increased the K m(ATP) for LTC 4 uptake 3-fold and decreased the V max 2.5-fold. Photolabeling studies of wild-type and truncated proteins with 8-azido-[␣-32 P]ATP, followed by limited tryptic digestion, suggested that nucleotide binding to NBD1 was unaffected, whereas binding to NBD2 of MRP1 1-1527 appeared to be impaired relative to the wild-type protein and to be even lower for the more severely compromised MRP1 1-1501 . More quantitative analyses using a dual vector to express the NH 2 -and COOH-halves of MRP1 1-1527 indicated that the K d of NBD2 for ATP was increased by 2-fold, whereas the binding of nucleotide to NBD1 was unaffected (Fig. 5C). Furthermore, the trapping of ADP in the presence of vanadate or beryllium was barely detectable even at relatively high concentrations of ATP, suggesting that nucleotide hydrolysis as well as binding is impaired in the truncated protein (Fig. 6, A  and B). Similarly, truncation of CFTR to Lys 1420 (which aligns with MRP1 Glu 1504 ) reduces channel activity 4 -5-fold, and it has been suggested that this is attributable to reduced ATP binding and hydrolysis by NBD2 (41). This CFTR truncation would be predicted to eliminate the analogous sheet C1 and helices C2 and C1 (Fig. 8).
As confirmation that the decreased nucleotide binding by NBD2 of the COOH-terminal truncations was of functional significance, we examined the effect of ATP and ATP␥S on substrate binding by the truncated proteins. We have shown that binding of nucleotide to NBD2 of the wild-type protein results in transition from a high to low affinity substrate binding state (24,39). The transition can be readily demonstrated as a marked nucleotide-dependent decrease in photolabeling with a substrate such as LTC 4 or the inhibitor LY475776 (24,39,43). In the absence of nucleotide, the binding of substrate by MRP 1-1501 was similar to that of wild-type protein. However, unlike full-length MRP1, preincubation of MRP1 1-1501 with ATP and vanadate completely failed to decrease LTC 4 binding. This is consistent with severely impaired ATP binding by NBD2 in the truncated protein. Comparison of LTC 4 photolabeling of wild-type MRP1 and the partially active MRP1 1-1527 demonstrated that ATP alone failed to shift the mutant protein from a high to a low affinity binding state as efficiently as the wild-type protein. Furthermore, no detectable decrease in LTC 4 labeling was observed for MRP1  in the presence of the poorly hydrolyzable ATP analog, ATP␥S. These results support the conclusion that the integrity of the COOH-terminal region of MRP1 is important for ATP binding and hydrolysis at NBD2 and the consequential transition from a high to a low affinity LTC 4 binding state.
To identify specific amino acids that might be critical for MRP1 activity, we carried out Ala scanning mutagenesis of the COOH-terminal region. None of the single Ala mutations were predicted to alter the secondary structure of the region, and, as might be expected from the low level of sequence conservation, most had no effect on activity (Table I). However, alignment of the COOH-terminal regions of a number of distantly related ABC proteins reveals several broadly conserved amino acids, notably the dileucine motif in helix C2 and a Phe or Tyr at the start of helix C1 (Fig. 8). In SUR1, both the highly conserved dileucine motif and the COOH-proximal Phe have been implicated as trafficking signals (29). In contrast, substitution of the comparable dileucine residues in MRP1 with Ala had no affect on plasma membrane localization (not shown). Instead, mutation of these helix C2 residues reduced LTC 4 transport activity (Table I) and nucleotide binding to NBD2 (not shown) to an extent similar to that caused by removal of the COOH-terminal 4 amino acids. Consistent with the presence of Leu-Val and Leu-Ile, rather than di-Leu, in some ABC proteins, substitution of Leu 1514 and Leu 1515 with Ile or Val rather than Ala had minimal effect on LTC 4 transport activity. Similarly, mutation of the conserved Phe 1521 in helix C1 to Tyr had no effect on activity, whereas mutation to any of several nonaromatic amino acids diminished LTC 4 transport to an extent comparable with that resulting from Ala substitution of the dileucine motif. In addition, two residues in helix C3, Leu 1488 and Ile 1491 are also relatively highly conserved, and mutation of either one to Ala decreased LTC 4 transport activity by ϳ40% (Fig. 8).
The crystal structures of hemolysin B (49), LmrA (Protein Data Bank number 1MV5), and Tap1 (50) indicate that the COOH-terminal regions of these proteins contain an ␣-helical bundle composed of helices C3, C2, and C1 (following the nomenclature in Fig. 8) that appears to be stabilized by hydrophobic interactions between certain residues that project into the core of the bundle. These residues correspond to the conserved Leu 1514 -Leu 1515 , Phe 1521 , Leu 1488 , and Ile 1491 in MRP1, mutation of which decreased LTC 4 transport (Table I). Thus, the truncations and mutations we have described that decrease the activity of MRP1 may do so by destabilizing this COOHterminal helical bundle. However, it should be noted that although the COOH-terminal helical bundle is widely conserved among ABC proteins, some prokaryotic NBDs, such as MJ0796, lack helices analogous to helices C3 and C2 but still bind ATP with relatively high affinity (K m ϭ 50 M) (51,52).
Based on the known structures of ABC proteins, there is no evidence to suggest that COOH-terminal helical elements interact directly with bound nucleotide (49,50,(53)(54)(55). However, examination of the known ABC crystal structures supports the possibility of interactions between residues in sheet C2 and the Walker A motif in NBD2. Furthermore, a naturally occurring Val mutation of a Leu in SUR1 (Leu 1551 ) that is analogous to Leu 1500 in MRP1, impairs MgADP activation of K ATP channels and results in congenital hyperinsulinism (56,57). The published model of SUR1 NBD suggests that mutation of Leu 1551 may influence nucleotide binding to NBD2 because of its proximity (Ͻ3.5 Å) to Ser 1382 and Lys 1384 , which correspond to Ala 1331 and Lys 1333 in the Walker A motif of MRP1 (58). In addition, nonconserved COOH-terminal residues have been identified recently in SUR2A and SUR2B splice variants that contribute to differences in ADP sensitivity (59).
Finally, the COOH-terminal region may be important for transduction of conformational changes from NBD1 to NBD2, since ATP binding to MRP1 NBD1 is essential for binding and hydrolysis of ATP at NBD2 (33,44). The requirement for binding to NBD1 is believed to reflect a nucleotide-dependent alteration in the interaction between the two NBDs that enables nucleotide binding by NBD2. The crystal structure of the BtuD NBD dimer indicates that the COOH-terminal residues Tyr 230 and Met 232 near the end of what we have termed helix C1 are within 4 -5 Å of Val 166 from the adjoining NBD (54). Val 166 is adjacent to the highly conserved Asp residue of the D loop. Helix C1 in MRP1 contains Tyr (Tyr 1522 ) and Met (Met 1524 ) at corresponding positions, as do the analogous helices in MRP2 and P-gp (Fig. 8). We are currently investigating the possibility that interactions between the COOH-terminal helical structures in MRP1 and elements near the D loop of NBD1 may be involved in transmitting ATP binding-induced conformational changes from one NBD to the other.