Reconstitution of ATP-dependent leukotriene C4 transport by Co-expression of both half-molecules of human multidrug resistance protein in insect cells.

Multidrug resistance protein (MRP) confers a multidrug resistance phenotype similar to that associated with overexpression of P-glycoprotein. Unlike P-glycoprotein, MRP has also been shown to be a primary active ATP-dependent transporter of conjugated organic anions. The mechanism(s) by which MRP transports these compounds and increases resistance to natural product drugs is unknown. To facilitate studies on the structure and function of MRP, we have determined whether a baculovirus expression system can be used to produce active protein. Full-length MRP as well as molecules corresponding to either the NH2- or COOH-proximal halves of the protein were expressed individually and in combination in Spodoptera frugiperda Sf21 cells. High levels of intact and half-length proteins were detected in membrane vesicles from infected cells. Although underglycosylated, the full-length protein transported leukotriene C4 (LTC4) with kinetic parameters very similar to those of MRP produced in transfected HeLa cells. Neither half-molecule was able to transport LTC4. However, a functional transporter with characteristics similar to those of intact protein could be reconstituted when both half-molecules were co-expressed. Transport of LTC4 by Sf21 membrane vesicles containing either intact or reconstituted MRP was competitively inhibited by both S-decylglutathione and 17β-estradiol 17-(β-D-glucuronide), with Ki values similar to those reported previously for MRP expressed in HeLa cells (Loe, D. W., Almquist, K. C., Deeley, R. G., and Cole, S. P. C. (1996) J. Biol. Chem. 271, 9675-9682; Loe, D. W., Almquist, K. C., Cole, S. P. C., and Deeley, R. G. (1996) J. Biol. Chem. 271, 9683-9689). These studies demonstrate that human MRP produced in insect cells can function as an active transporter of LTC4 and that the NH2- and COOH-proximal halves of the protein can assemble efficiently to form a transporter with functional characteristics similar to those of the intact protein.

of drugs that includes many natural products and their derivatives. The most extensively characterized mechanism known to cause resistance to natural product drugs in vitro is overexpression of the ATP-dependent transmembrane transporter P-glycoprotein (P-gp) 1 (1). However, recently, a second member of the ATP binding cassette (ABC) superfamily, multidrug resistant protein (MRP), has been shown to confer a multidrug resistance phenotype similar to that associated with P-gp (2)(3)(4). Since its discovery in the human small cell lung cancer cell line H69AR, MRP has been identified in non-P-gp multidrugresistant cell lines from a variety of tumor types (2,5). Drugresistant cell lines that overexpress both MRP and P-gp have also been described (6 -8).
MRP was predicted from its cDNA sequence to be 1531 amino acids long with a polypeptide molecular weight of 171,000. MRP contains multiple potential transmembrane helices, two nucleotide binding domains (NBDs), and 14 potential N-linked glycosylation sites (2). Biochemical analyses confirmed that MRP is synthesized as a M r 170,000 precursor that is processed to a mature N-glycosylated form with an apparent molecular weight of 190,000 (9,10).
Most drug-selected cell lines and transfected cells overexpressing MRP display reductions in drug accumulation and increases in the rate of ATP-dependent drug efflux (4,(11)(12)(13)(14)(15)(16). However, there is no direct evidence that MRP can bind unmodified forms of the drugs to which it confers resistance, and there are conflicting reports of its ability to transport these compounds (17)(18)(19). MRP can function as a primary active ATP-dependent transporter of cysteinyl leukotrienes, other organic glutathione conjugates including oxidized glutathione, as well as etoposide glucuronide, certain steroid glucuronides, and bile salt derivatives (18 -24). In addition, the protein has been shown to transport the Vinca alkaloid vincristine, but only in the presence of GSH (19). The mechanism by which this occurs is not known.
Elucidation of the molecular mechanism by which MRP transports conjugated organic anions will require identification of the specific protein domains and amino acid residues that participate in substrate binding and transport. As an initial step, we have expressed intact human MRP as well as the NH 2and COOH-proximal halves of the protein, both independently and in combination, using a baculovirus expression system. The ability of these molecules to transport leukotriene C 4 (LTC 4 ) into inside-out membrane vesicles has been determined. We found that intact MRP expressed in Spodoptera frugiperda Sf21 cells, although underglycosylated, exhibited a high level of ATP-dependent LTC 4 transport with kinetic parameters similar to those obtained with vesicles from MRPtransfected human cells. We also found that neither half-molecule alone would support LTC 4 transport, but a high level of ATP-dependent transport was obtained when both half-molecules were co-expressed. These results indicate that the two half-molecules of MRP can assemble to form an active transporter despite the lack of covalent linkage, and that transport of LTC 4 requires interaction of both halves of MRP.

EXPERIMENTAL PROCEDURES
Generation of Constructs-The MRP expression cassette (MRP1) in pBluescript II KS ϩ (Stratagene) used previously to generate a vector for transfection of mammalian cell lines included 86 nucleotides of the 5Ј-untranslated region of MRP mRNA (3). To eliminate the potential of this very GC-rich region to decrease the translational efficiency in insect cells, a second MRP construct lacking this sequence was constructed. The 5Ј-end of the MRP coding sequence was amplified by polymerase chain reaction (PCR) using primers 5Ј-TCCCCGGGGC-CGCCATGGCGCTCCGGGGCTTC-3Ј (forward primer), which includes a SmaI site (underline) and consensus Kozak sequence (double underline), and 5Ј-GAAGTAGCCCTGCCAGTC (reverse primer). A PCR product of approximately 1.1 kilobases was generated and subsequently digested at the SmaI site in the primer and at a BamHI site in the MRP coding sequence to yield an 840-base pair fragment. This fragment was inserted in the MRP1 expression cassette subsequent to removal of the existing SmaI-BamHI fragment to produce the new expression cassette pBSMRP-fc-ATG. A 4.7-kilobase SacI-KpnI fragment containing the full-length coding region flanked by an untranslated sequence of 31 and 77 nucleotides at the 5Ј-and 3Ј-ends, respectively, was isolated from pBSMRP-fc-ATG and inserted into the donor plasmid pFASTBAC1 (Life Technologies, Inc.) downstream from the polyhedrin promoter.
To generate the constructs in which only one-half of the MRP was expressed, two fragments corresponding to nucleotides 2355-2808 and 2782-3279 were amplified by PCR. The primers used for the 2355-2808 fragment were primer 10.2F4, 5Ј-CGCTGACATTTACCTCTTCG-3Ј (forward primer), and primer 1013 5Ј-GCGGGTACCTCATGCGGTGCTGT-TGTGGTG-3Ј (reverse primer). The primer 1013 contained an in-frame stop codon TGA (double underline) and a KpnI site (underline). The primers used for the 2782-3279 fragment were primer 1014, 5Ј-CAC-GAGCTCATGGCAGAACTGCAGAAAGCTG-3Ј (forward primer), and primer 10.2R3, 5Ј-CATCTTGATGACCTCCG-3Ј (reverse primer). In primer 1014, a SacI site (underline) and a start codon (double underline) were introduced. The full-length vector was then digested with XhoI and KpnI, leaving the sequence extending to nucleotide 2570 attached to the vector backbone. The digested vector was purified, ligated to the XhoI-KpnI fragment of the 2355-2782 PCR product, and transfected into DH5␣ FЈ cells. Fidelity of the PCR fragments was verified by dideoxy sequencing of the modified construct. Similarly, the construct expressing the COOH-proximal half of MRP was produced by digesting the full-length vector with HindII and SacI, leaving a 3Јproximal sequence beginning at nucleotide 3015 attached to the vector backbone. The purified digested vector was ligated to the SacI-HindII fragment of the 2782-3279 PCR product. Digestion of the modified vectors with SacI and KpnI yielded fragments of 2.8 and 1.9 kilobases, encoding amino acids 1-932 and 932-1531, respectively, which were cloned individually into the donor plasmid pFASTBAC1.
Production of Recombinant Baculovirus-The recombinant donor plasmids were used to transform Escherichia coli-competent DH10BAC cells (Life Technologies), which contain the parent bacmid and a helper plasmid. Recombinant bacmids were selected on LB plates containing 50 g/ml kanamycin, 7 g/ml gentamicin, 10 g/ml tetracycline, 300 g/ml 5-bromo-4-chloro-3-indolyl-␤-D-galactoside, and 40 g/ml isopropylthio-␤-D-galactoside. The purified recombinant bacmids were transfected into Sf21 cells using CELLFECTIN (Life Technologies) reagent. Three days later, cells were harvested, and protein expression was determined by immunoblotting. The culture supernatant was used to infect fresh Sf21 cells, and stocks of the amplified virus were prepared. Their titers were determined, and the virus was stored at 4°C. Viral Infection and Membrane Vesicle Preparation-One hundred million Sf21 cells were infected with the recombinant virus using a multiplicity of infection of 0.2 plaque-forming units/cell. Sixty hours after infection, cells were collected, and membrane vesicles were prepared as described previously (19).
Immunoblotting and Glycosylation Studies-Membrane proteins were solubilized in Laemmli buffer (25) and subjected to SDS-polyacrylamide gel electrophoresis (PAGE), electrotransferred to a polyvinylidene difluoride membrane (Millipore), and blotted as described (26). Blots were incubated with MRP-specific mAb QCRL-1 (purified, diluted 1:10,000; Centocor, Malvern, PA) (26,27), or mAb MRPm6 (ascites, diluted 1:250; kindly provided by Dr. R. Scheper, Free University Hospital, Amsterdam, The Netherlands) (28), followed by horseradish peroxidase-conjugated goat anti-mouse IgG and IgM (HϩL) and F(abЈ) 2 (Pierce) (10,26). Antibody binding was determined by enhanced chemiluminescence detection (Amersham Corp.). Membrane proteins were deglycosylated by incubation with peptide N-glycosidase F (PNGase F; New England Biolabs, Mississauga, Ontario, Canada) as described (10). Quantification of MRP Protein by Immunoblotting-Since the epitopes for MRP-specific mAbs QCRL-1 and MRPm6 remained unaltered in the various MRP constructs used to infect Sf21 cells, immunoblotting with one or the other of these mAbs was used to estimate the amount of MRP in different membrane preparations. To reduce sizedependent differences in transfer efficiency, membrane proteins were resolved on a 5-15% gradient gel, and aliquots of membranes from MRP-transfected HeLa T14 cells (14) were loaded on both sides of the gel to serve as internal standards. The relative amount of MRP or MRP fragments in various samples was quantified by laser densitometry of appropriately exposed films following enhanced chemiluminescence detection using a Molecular Dynamics (Sunnyvale, CA) computing densitometer.
Vesicle Transport of LTC 4 -Membrane vesicles were prepared from Sf21 cells infected with virus encoding intact MRP or MRP fragments, as well as from control cells transfected with virus encoding ␤-glucuronidase, as described previously for the preparation of vesicles from MRP-transfected HeLa cells (19). Uptake of [ 3 H]LTC 4 (50 nM, 132 Ci mmol Ϫ1 ; DuPont NEN) into membrane vesicles (10 g of total membrane protein in 120 l) was measured at room temperature in the presence of ATP or AMP (4 mM) using a rapid filtration technique, as described (29) with modifications (19). Initial rates of LTC 4 uptake were determined at various concentrations of ATP and the leukotriene itself, and double reciprocal plots of the data were used to determine K m values for both LTC 4 and ATP (19). The effects of S-decylglutathione and 17␤-estradiol 17-(␤-D-glucuronide) on LTC 4 uptake by Sf21 vesicles prepared from cells infected with vectors expressing either intact MRP or both half-molecules were determined at various LTC 4 concentrations in the presence of a fixed concentration of inhibitor. Double reciprocal plots of the data were used to determine K i values for both compounds, as described (19). nated ⌬C, includes most of the relatively nonconserved connector region, which contains the linear epitope for mAb QCRL-1 (27). The COOH-proximal half-molecule, designated ⌬N, contains the MRP sequence beginning at amino acid 932, preceded by an initiator methionine introduced during construction of the vector. It also contains the epitope for mAb MRPm6 (28). Recombinant bacmids coding for the full-length and two halfmolecules of MRP, as well as for ␤-glucuronidase, were used to transfect Sf21 cells. Expression of full-length and both halfmolecules was confirmed by immunoblotting of total cell lysates (data not shown). Recombinant viruses were amplified, and titers were determined, which usually ranged between 2-8 ϫ 10 7 plaque forming units/ml. These viral stocks were used to infect Sf21 cells from which membrane vesicles were prepared.

Generation of Recombinant
Synthesis of Recombinant MRP in Sf21 Cells-Immunoblotting using the MRP-specific mAb QCRL-1 revealed an abundant protein of approximately 170 kDa in membranes from Sf21 cells infected with the full-length MRP vector (Sf21/wt) that was not detectable in membranes from Sf21 cells infected with virus coding for ␤-glucuronidase (Sf21/c; Fig. 2A). The apparent molecular weight of MRP produced in Sf21/wt cells was approximately 20,000 less than that of the mature protein produced in transfected HeLa cells (data not shown) and was similar to that of its unglycosylated precursor (10). To determine whether the difference in molecular weight was attributable to underglycosylation of the baculovirus-encoded MRP, membrane proteins from T14 and Sf21/wt cells were treated with PNGase F, resolved by SDS-PAGE, and immunoblotted with mAb QCRL-1 (Fig. 2B). PNGase F treatment reduced the apparent molecular weight of MRP in T14 membranes from 190,000 to 170,000, as observed previously (10). Deglycosylation reduced the apparent molecular weight of MRP produced in insect cells to a much lesser extent, so that following PNGase F treatment the proteins produced in Sf21 and HeLa cells co-migrated. The relative levels of MRP expressed in Sf21/wt cells and HeLa T14 transfectants were compared by immunoblotting of a serial dilution of total membrane protein isolated from each cell type (Fig. 2C). Densitometry of the immunoblot indicated that the levels of baculovirus-encoded MRP were 50 -75% of those present in T14 membranes. Thus full-length MRP can be efficiently expressed and integrated into the membranes of Sf21 cells, but the protein is clearly underglycosylated.
MRP Expressed in Sf21 Cells Functions as an ATP-dependent LTC 4 Transporter-The cysteinyl leukotriene LTC 4 has been shown to be a high affinity substrate for human MRP. Its transport characteristics have been established in several laboratories using membrane vesicles from both drug-selected MRP-overexpressing cells and MRP transfectants such as the T14 HeLa cell population (19 -22). We used this substrate to determine whether MRP produced by Sf21 cells was functional and, if so, whether its transport characteristics were similar to those of the protein produced in human cells. Fig. 3, A and B  Expression of Half-molecules of MRP in Sf21 Cells-MRP half-molecules were expressed either individually or together in Sf21 cells, and the relative levels of the NH 2 -and COOHproximal halves of the molecule were determined by immunoblotting with the MRP-specific mAbs QCRL-1 and MRPm6, respectively (Fig. 4). Both the NH 2 -and COOH-proximal halfmolecules were efficiently expressed and recovered in the membrane preparations as polypeptides of the anticipated molecular weights of 104,000 and 67,000, respectively. Minor amounts of higher molecular weight species were also detected. Based on their apparent molecular weights and the observation that their levels increased when protein samples were subjected to less rigorous denaturation prior to electrophoresis (data not shown), the minor species appear to be homodimers and possibly trimers of the half-molecules. Using the same multiplicities of infection of baculoviruses, the expression levels of the halfmolecules were similar to each other and approximately the same as those obtained with full-length MRP. In addition, when co-expressed, the levels of both half-molecules were comparable with those obtained when they were expressed individually.
LTC 4 Transport by Individual and Co-expressed MRP Halfmolecules-Membrane vesicles were prepared from cells infected with vectors encoding each MRP half-molecule as well as from cells that were co-infected with both. LTC 4 transport by vesicles from cells producing either half-molecule alone did not differ significantly from control cells in the presence of either ATP or AMP. In contrast, vesicles prepared from co-infected cells displayed initial rates of ATP-dependent LTC 4 uptake that were approximately 40% of those obtained with vesicles from Sf21 cells producing intact MRP (Fig. 5). The rates of LTC 4 uptake by vesicles from co-infected cells were determined at various concentrations of leukotriene and ATP to yield K m and V max values for LTC 4 of 56 nM and 168 pmol mg Ϫ1 min Ϫ1 , respectively and a K m for ATP of 44 M. To further compare the transport characteristics of the co-expressed half-molecules with those of the intact protein, we examined the ability of S-decylglutathione and 17␤-estradiol 17-(␤-D-glucuronide) to inhibit LTC 4 transport (Fig. 6). Despite their marked structural differences, these two compounds have been shown to be competitive inhibitors of LTC 4 transport by vesicles from MRPtransfected HeLa cells (19). S-Decylglutathione is one of the most potent competitive inhibitors identified to date, and 17␤estradiol 17-(␤-D-glucuronide) has been shown by direct transport studies to be a substrate for MRP (24). Both of these compounds acted as competitive inhibitors of LTC 4 transport by vesicles from Sf21/wt and the co-infected cells. The K i values for S-decylglutathione and 17␤-estradiol 17-(␤-D-glucuronide) with the intact protein were 64 nM and 55 M, respectively and 50 nM and 33 M with the co-expressed half-molecules. DISCUSSION Intact MRP has been expressed efficiently in several types of transfected mammalian cells (3,4,16). However, we have found that a number of MRP mutant proteins fail to accumulate in the plasma membrane despite high levels of expression of the cognate mRNAs. 2  some mutated forms of these proteins to the plasma membrane is very inefficient (31)(32)(33), probably as the consequence of aberrant protein folding and association with chaperones (34 -36). Although the well characterized naturally occurring mutation of CFTR, ⌬F508, fails to fold correctly in the endoplasmic reticulum when expressed in mammalian cells growing at 37°C (32,37), in insect cells grown at 27°C (38), CFTR-⌬F508 is efficiently synthesized and is functional. Truncated forms of both CFTR and P-gp have also been synthesized efficiently in baculovirus-infected insect cells (39,40). Consequently, we have determined whether this system may be a useful alternative to transfected mammalian cells for structure-function studies of MRP.
Our results demonstrate that full-length MRP can be expressed at relatively high levels in Sf21 cells. Under the conditions described, the protein accumulates to a level that is approximately one-half of that obtained with mammalian cells transfected with episomal, multicopy vectors containing the Epstein-Barr virus origin of replication (14). MRP produced in insect cells is significantly underglycosylated when compared with the protein produced in drug-selected and transfected human cells, which is N-glycosylated at sites in both the NH 2and COOH-proximal halves of the molecule (5). 3 However, previous studies with tunicamycin-treated mammalian cell MRP transfectants and cells expressing mutant P-gps lacking glycosylation sites suggested that lack of this modification would not be critical for function (10,41,42).
CFTR produced in insect cells forms a Cl Ϫ channel with characteristics similar to those found in mammalian cells (43). In addition, baculovirus-encoded P-gp is functional with respect to both its drug-binding and ATPase activities (44,45). However, there has been no demonstration of direct ATP-dependent drug transport by membrane vesicles from P-gp-infected insect cells, and since the infected cells do not replicate, drug sensitivity assays are not possible. To assess the functional integrity of MRP produced in Sf21 cells, we have taken advantage of the fact that its ability to transport LTC 4 and the inhibition of this process by various organic anions have been characterized in several mammalian cell transfectants (19,20,22,23). No ATP-dependent LTC 4 uptake could be detected by vesicles from cells infected with the control virus, and ATPindependent uptake was considerably lower than observed with membrane vesicles from HeLa cells. In contrast, ATP-dependent LTC 4 transport was readily demonstrable using vesicles from cells infected with virus encoding intact MRP. We confirmed previously that the rates of LTC 4 uptake by vesicles from T14 HeLa transfectants and drug-selected H69AR cells were approximately proportional to their relative levels of MRP expression (14,19). Based on immunoblotting data, the levels of MRP in Sf21 membranes were 50 -75% of those in T14 cells, and the V max values for LTC 4 transport were 80 -90%. The K m values for LTC 4 and ATP and the K i values for the competitive inhibitors S-decylglutathione and 17␤-estradiol 17-(␤-D-glucuronide) were also similar to those obtained previously with T14 vesicles, indicating that there are no major differences between the transport activity and substrate binding characteristics of MRP expressed in either HeLa or Sf21 cells.
Many ABC transporters contain four structural domains that include two polytopic membrane-spanning regions and two cytoplasmic NBDs (46). However, the membrane-spanning regions and NBDs of some prokaryotic ABC transporters are separate polypeptides (47,48). In addition, some ABC proteins (49,50), appear to be "half-molecules" with only a single set of six transmembrane segments and one NBD that may function as homodimers. These observations have been interpreted to suggest that many members of the ABC superfamily have evolved by duplication and/or fusion of previously autonomous "half-molecules" or individual domains (46). The fusion or separation of functional domains has, in some cases, been mimicked experimentally (46,51). Notably, co-expression of both halves of P-gp in Sf9 cells was required for drug stimulation of ATPase activity, indicating the ability of the two fragments in association to bind substrate (40). However, direct drug transport by the co-expressed halves of P-gp has not been demonstrated, and when co-expressed in HEK 293 cells, no drugresistant clones were obtained. Consequently, it remains to be established that the two halves of P-gp can associate to form an active drug pump (40). In contrast, the NH 2 -proximal 836 amino acid residues of CFTR, encompassing only the first six transmembrane helices, the first NBD, and the regulatory domain, can form Cl Ϫ channels with conductive properties identical to those of full-length CFTR (43).
The predicted topologies of MRP and, more recently identified, related ABC transporters such as the sulfonylurea receptor, the canalicular multispecific organic action transporter, yeast-cadmium factor 1, and epithelial basolateral conductance regulator, are not consistent with two similarly organized halves, largely because of the extremely hydrophobic NH 2terminal extension present in these proteins (30,(52)(53)(54). Consequently, the NH 2 -and COOH-truncated MRP fragments we have generated differ considerably from each other. The NH 2proximal "half-molecule" may contain as many as 12 transmembrane helices, in addition to the first NBD and most of the connector region, whereas the COOH-proximal fragment contains only 4 -6 membrane-spanning helices and the second NBD (Fig. 1). Despite their structural differences, each of the two MRP half-molecules was able to integrate into the membranes of infected cells equally well when expressed either alone or together. Levels of expression were also comparable to that of the intact protein, suggesting that interaction between the two halves of the protein has little effect on the efficiency of integration into the membrane or on protein trafficking. Although no ATP-dependent LTC 4  either half-molecule alone, co-expression restored transport, with a K m for LTC 4 and K i values for S-decylglutathione and 17␤-estradiol 17-(␤-D-glucuronide) similar to those of the intact protein. The major difference detected between vesicles containing the reconstituted and intact proteins was that the V max obtained with the former was only 30 -35% that of the latter, despite similar levels of protein expression. This may indicate that transport by the reconstituted protein is less efficient than with the intact molecule. However, it appears more likely to be a consequence of the heterogeneity of the infected cell population. Immunocytochemistry indicates that with a multiplicity of infection of 0.2 there is a continuum of MRP expression, with approximately 40% of the infected cells being strongly positive at the time of membrane preparation, either for full-length MRP or for each half-molecule (data not shown). At present, we are unable to determine the stoichiometry with which the halfmolecules are expressed in individual cells. Thus the difference in V max values may simply reflect the proportion of half-molecules that have the opportunity to heterodimerize.
It is presently not known to what extent substrate binding and/or ATP hydrolysis is dependent on association of the two halves of MRP. However, the ability to reconstitute active transport clearly demonstrates that any conformational changes that may occur as a consequence of ATP hydrolysis or during transfer of LTC 4 across the lipid bilayer do not require the covalent linkage of the two halves of MRP. This observation, combined with the structural differences between the NH 2 -and COOH-proximal halves of the protein and the relatively low level of sequence conservation between the first and second NBDs (2), is consistent with the possibility that MRP evolved by the fusion of genes encoding ABC proteins that were different components of a functional transporter. The availability of a convenient vesicle transport assay combined with the ability to reconstitute a functional protein from its component domains should permit identification of interacting regions of MRP, as well as structural elements necessary for substrate binding and/or transport.