MDR3 P-glycoprotein, a Phosphatidylcholine Translocase, Transports Several Cytotoxic Drugs and Directly Interacts with Drugs as Judged by Interference with Nucleotide Trapping*

The human MDR3 gene is a member of the multidrug resistance (MDR) gene family. The MDR3 P-glycoprotein is a transmembrane protein that translocates phosphatidylcholine. The MDR1 P-glycoprotein related transports cytotoxic drugs. Its overexpression can make cells resistant to a variety of drugs. Attempts to show that MDR3 P-glycoprotein can cause MDR have been unsuccessful thus far. Here, we report an increased directional transport of several MDR1 P-glycoprotein substrates, such as digoxin, paclitaxel, and vinblastine, through polarized monolayers of MDR3-transfected cells. Transport of other good MDR1 P-glycoprotein substrates, including cyclosporin A and dexamethasone, was not detectably increased. MDR3 P-glycoprotein-dependent transport of a short-chain phosphatidylcholine analog and drugs was inhibited by several MDR reversal agents and other drugs, indicating an interaction between these compounds and MDR3 P-gp. Insect cell membranes from Sf9 cells overexpressing MDR3 showed specific MgATP binding and a vanadate-dependent,N-ethylmaleimide-sensitive nucleotide trapping activity, visualized by covalent binding with [α-32P]8-azido-ATP. Nucleotide trapping was (nearly) abolished by paclitaxel, vinblastine, and the MDR reversal agents verapamil, cyclosporin A, and PSC 833. We conclude that MDR3 P-glycoprotein can bind and transport a subset of MDR1 P-glycoprotein substrates. The rate of MDR3 P-glycoprotein-mediated transport is low for most drugs, explaining why this protein is not detectably involved in multidrug resistance. It remains possible, however, that drug binding to MDR3 P-glycoprotein could adversely affect phospholipid or toxin secretion under conditions of stress (e.g. in pregnant heterozygotes with oneMDR3 null allele).

P-glycoproteins (P-gps) 1 are 170-kDa glycosylated membrane proteins that actively transport their substrates out of the cell. Two P-gp genes have been identified in humans. The human MDR1 gene encodes a drug transporting P-gp that can actively extrude a range of cytotoxic anticancer drugs from the cytoplasm. Overexpression of this P-gp gene results in a decreased intracellular accumulation of these drugs and renders the cell multidrug resistant (1)(2)(3). The second human P-gp gene is MDR3 (also known as MDR2) (4,5). Attempts to obtain resistance against cytotoxic drugs by transfecting drug-sensitive cell lines with the MDR3 cDNA or its mouse homolog, Mdr2, yielded only negative results (4, 6 -9). Amplification or activation of the MDR3 gene, independent of the closely linked MDR1, has never been found in multidrug-resistant cell lines (10 -12). Nevertheless, two studies suggest a role for MDR3 P-gp in drug transport: 1) analysis of B-cell leukemias showed a correlation between MDR3 overexpression and daunorubicin transport (13,14), and 2) Kino et al. (15) found low level resistance against the antifungal agent aureobasidin A in yeast transformed with the human MDR3 cDNA.
The function of the murine homolog of human MDR3 P-gp, Mdr2 P-gp, became clear when Smit et al. (16,17) generated mice, homozygous for a disruption of the Mdr2 gene (Mdr2 (Ϫ/Ϫ) mice). These mice were unable to excrete phospholipids (PL) into their bile, indicating that Mdr2 P-gp is a PL translocator or flippase, translocating PL from the cytosolic leaflet to the luminal leaflet of the canalicular membrane. The excretion of phosphatidylcholine (PC; the major PL in the bile) into the bile can be restored in these mice by expression of the human MDR3 gene in the hepatocytes (18,19), demonstrating that the closely related Mdr2 and MDR3 P-gps fulfill the same function. Further evidence for the PC translocator function of the MDR3/ Mdr2 P-gps came from studies in yeast and cultured mammalian cells. Gros and co-workers showed that secretory vesicles from yeast transformed with Mdr2 cDNA can accumulate fluorescent short chain PC (20,21). We demonstrated increased translocation of PC through the plasma membrane of cells overproducing the MDR3 P-gp (22,23) and defined the substrate specificity of the translocator for PL (23). Whereas the MDR1 P-gp was found to translocate a variety of PL analogs, the MDR3 P-gp proved highly selective, only translocating PL with a choline head group and a diacyl glycerol backbone. Even a PC with two short acyl chains (C 8 -C 8 -PC) was not transported (23). This selectivity seemed an adequate explanation for the inability of the MDR3/Mdr2 type of P-gp to contribute to MDR.
It remains puzzling, however, that transport of a PC analogue by the MDR3 P-gp could be inhibited by verapamil, a known substrate and inhibitor of MDR1 P-gp (20,23). To resolve this paradox, we reinvestigated drug transport by the MDR3 P-gp, using monolayers of polarized kidney cells transfected with a MDR3 minigene (23), in which the transporter localizes in the apical membrane. We found that the transfected cells transport digoxin, a substrate of the MDR1 P-gp (24,25) and several other cytotoxic drugs, and that transport is inhibited by several inhibitors (reversal agents) of MDR1 P-gp. Moreover, these drugs appear to interact directly with the MDR3 P-gp, since they are able to inhibit vanadate-dependent nucleotide trapping by MDR3 P-gp produced in insect cells.

Materials-[12␣-
Tissue Culture-The LLC-PK1 pig kidney epithelial cells were obtained from the American Type Culture Collection and cultured in M199 medium supplemented with 50 units of penicillin and 50 g of streptomycin/ml and 10% fetal bovine serum at 37°C in 5% CO 2 . The cells were trypsinized and subcultured every 3-4 days. The MDR1-, MDR3-, and MRP-transfected clones of LLC-PK1 were obtained by calcium phosphate coprecipitation of the corresponding cDNAs as de- The flows from the apical to the basolateral compartment (f) and vice versa (Ⅺ) are measured separately and plotted in a single graph. The horizontal bars indicate the values of two independent measurements. Note that the y axis is not identical for all drugs. The net transport (flow from basolateral to apical minus the flow from apical to basolateral) is significantly higher in the MDR3-transfected cells than in the parental cell line (two-sided p Ͻ 0.00001 for all drugs using one-way analysis of variance).
scribed (23,25,26). The cells were maintained routinely in the absence of drugs and tested regularly for P-gp content by immunoblotting with the monoclonal antibody C219, which recognizes all mammalian P-gps tested (27). This was followed by the enhanced chemiluminescence procedure (Amersham Pharmacia Biotech), as described (25,28).
Digoxin Accumulation Experiments-Digoxin accumulation in the various cell lines was measured by seeding 1 ϫ 10 6 cells in a single well of a six-well plate, followed by incubation of the cells for 3 days. The medium was replaced by 1.5 ml of medium containing 10 M [ 3 H]digoxin (0.25 Ci/ml). After a 2-h incubation period, the medium was removed, and the cells were washed twice with 1 ml of ice-cold phosphate-buffered saline. The cells were scraped and removed from the wells in 0.5 ml of phosphate-buffered saline. Radioactivity was determined by liquid scintillation counting. Accumulation was measured in triplicate in the presence and absence of 5 M PSC 833.
Drug Transport Assays-Drug transport assays were performed as described (25,29) with minor modifications. Cells were seeded on microporous polycarbonate membrane filters (pore size 3.0 m, diameter 24.5 mm, Transwell TM , Costar R ) at a density of 1.5-2 ϫ 10 6 cells/filter. The cells were grown in complete medium for 3 days with a medium replacement the day after seeding. One hour before the start of the experiment, medium at both sides of the monolayer was replaced. The experiment was started by replacing the medium at either the apical or the basolateral side by medium containing 3 H-labeled substrate drugs (0.25 Ci/ml) and inulin[ 14 C]carboxylic acid (0.025 Ci/ml, 4.2 M). The cells were incubated at 37°C in 5% CO 2 , and 50-l aliquots were taken from both compartments at t ϭ 1, 2, 3, and 4 h. The appearance of radioactive drug in the opposite compartment was measured and presented as the fraction of total radioactivity added at the beginning of the experiment. Directional transport was measured in duplicate. The paracellular flow was monitored by the appearance of [ 14 C]inulin in the opposite compartment and was always less than 1.5% of total radioactivity per hour.
Inhibition of transport was measured similarly. One hour before the start of the experiment, the medium at both sides of the monolayer was replaced by complete medium containing the appropriate concentration of reversal agent. At t ϭ 0, medium in one of the compartments was replaced by medium with radioactive drug and the indicated amount of reversal agent.
Lipid Transport Assays-Transport of C 6 -NBD-PC and C 6 -NBD-glucosylceramide (C 6 -NBD-GlcCer) was measured as described (23,30). LLC-PK1-derived cell lines were seeded on microporous membrane filters and grown for 4 days. Lipid precursors, C 6 -NBD-PA or C 6 -NBDceramide, were complexed to 1% (w/v) bovine serum albumin (BSA) in HEPES-buffered Hanks' balanced salt solution (pH 7.35) (HBSS) to the following concentrations: 25 M C 6 -NBD-PA and 5 M C 6 -NBD-ceramide. To discriminate surface synthesized C 6 -NBD-PC (30) from transport of intracellularly synthesized C 6 -NBD-PC, cells were preincubated with 10 Ci of [ 3 H]choline/ml HBSS for 1 h at 37°C, after a 1-h choline depletion in HBSS at 37°C. Incubation with the lipid precursors in HBSS plus BSA followed for 3 h at 15°C in the presence of 10 Ci of [ 3 H]choline. For inhibition studies, the appropriate concentration of reversal agents was present 1 h before and during the 3-h incubation at 15°C. During the incubations, short-chain lipid products appearing on the cell surface were depleted into the medium by the BSA. After 3 h, the apical and basolateral medium were collected, and the cells were washed in HBSS plus BSA for 30 min on ice. The apical and basolateral washes were pooled with the respective incubation media, the filters were cut from the wells, and the lipids from media and cells were analyzed.
Lipid Analysis-Lipids were extracted from the cells and media by a modified Bligh and Dyer extraction (31). The upper phase contained 20 mM acetic acid. The organic (lower) phase was dried under N 2 , and the lipids were applied to TLC plates. Lipid products were separated in two dimensions by borate-TLC as described (32) but over 20 cm in the first (alkaline) dimension. In the case of C 6 -NBD-[ 3 H]PC in cellular lipid extracts, the fluorescent spot was circled, and the plate was exposed to a film to allow accurate separation of C 6 -NBD-[ 3 H]PC and long-chain [ 3 H]PC. Fluorescent spots were quantitatively analyzed in a fluorimeter, radiolabeled spots were detected by fluorography, and the radioactivity was quantified by liquid scintillation counting as described (33).
MDR3 Expression and Membrane Preparation-Sf9 (Spodoptera frugiperda) cells were cultured and infected with baculovirus vectors as described in Ref. 34. Baculotransfer virus for MDR3 was constructed as follows: The plasmid constructs harboring human MDR3 cDNA were pJ3⍀-MDR3 and pJ3⍀-MDR3-NotI. Since the MDR3 cDNA was found to be unstable in several baculovirus transfer vectors in E. coli, an alternative approach has been developed, which does not involve bacterial propagation of the recombinant vector. MDR3 cDNA and pieces of the appropriate transfer vector containing essential segments for homologous recombination with the viral DNA were ligated, and this ligation mix was used for cotransfection of insect cells. The 5Ј region of the MDR3 cDNA (Ϫ33 to 1621) was removed from pJ3⍀-MDR3-NotI by NotI and ApaI digestion, and the 3Ј region (1621-4002) was removed from the pJ3⍀-MDR3 by ApaI and XbaI digestion.
Two segments of the pVL 1392 baculovirus transfer plasmid (Invitrogen) were isolated; the first one extends from 4154 to 8361 (obtained by XbaI and BglI digestion), and the second one covers the region from 9479 to 4138 (obtained by BglI and NotI digestion).
The two MDR3 fragments were ligated with the two pVL1392 fragments, and this "ligation mix" was used for generating recombinant viruses with the Baculogold Transfection Kit (Pharmingen, San Diego, CA).
Sf9 insect cells were infected and cultured as described previously (34). The virus-infected Sf9 cells were suspended in a low ionic strength medium (50 mM Tris-HCl, pH 7.0, 50 mM mannitol, 2 mM EGTA, 10 . Depicted is the difference in the absolute amounts of drug transported from the basolateral to the apical compartment and vice versa in 4 h (mean Ϯ range in two independent experiments). Each compartment contained 2 ml of tissue culture medium (B). Protein blot analysis is shown of total cell lysates of the corresponding subclones. Detection of P-gp content was with the monoclonal antibody C219, which recognizes all mammalian P-gps tested (27). g/ml leupeptin, 8 g/ml aprotinin, 0.5 mM phenylmethylsulfonyl fluoride, 2 mM dithiothreitol) and disrupted using a glass-Teflon homogenizer. Membrane fractions were isolated by repeated centrifugations and homogenizations, and the membrane protein concentrations determined as described in Ref. 35. After cloning the baculoviruses producing high level MDR3 expression, the sequences of the cDNA inserts were checked by polymerase chain reaction amplification.
ATP Binding to MDR3 P-gp-Binding of ATP to MDR3 P-gp was measured by incubation of isolated Sf9 cell membranes (100 g of protein) for 5 min at 4°C in a reaction buffer containing 50 mM Tris-HCl through the four time points of each well was determined, resulting in two estimations of the slope per experiment. Because we assumed that the random variation of the slope was equal in all experiments, we used the differences between the two independent estimations of all experiments to determine this random variation. All p values are two-sided. Differences are considered significant if p is Ͻ0.05. Fig. 1 shows the transport of several MDR1 P-gp substrate drugs through monolayers of polarized pig kidney epithelial cell lines. LLC-PK1 is the untransfected parent cell line, LLC-M3.4.44 (23) is a subclone of the MDR3 transfectant LLC-M3.4, and LLC-1.1 is the MDR1-transfected LLC-PK1 clone (25). Directional transport of the cardiac glycoside digoxin; of the cytostatic drugs paclitaxel, daunorubicin, and vinblastine; and of the antihelmintic drug ivermectin to the apical compartment is significantly increased in the MDR3-transfected cells compared with the parental cell line (two-sided p Ͻ 0.00001 for all drugs using one-way analysis of variance). With the exception of paclitaxel, the net transport rate (flow from basolateral to apical minus the flow from apical to basolateral) of these drugs is lower in the MDR3-transfected cells than in the MDR1 transfectant (two-sided p Ͻ 0.00001 for daunorubicin, vinblastine, and ivermectin; two-sided p ϭ 0.001 for digoxin), whereas the P-gp expression levels are comparable (see Fig. 2). The transport rate of paclitaxel is extremely high in both cell lines, and this makes it difficult to compare the activities of the transporters. The transport experiments are performed twice for ivermectin and three times or more for the other drugs with various batches of the cell lines. The statistically significant differences were comparable in the various experiments showing that the increased transport of drugs in the MDR3-transfected cell line is reproducible (results not shown). It should be noted that the expression of MDR3 is rather heterogeneous (23), in contrast to the expression of MDR1 (25); this might influence the transport efficiency through the monolayers. These results were reproduced in an independent MDR3 transfectant clone LLC-M3.1 (e.g. see Fig. 2A). The transport rate of digoxin in several subclones of LLC-M3.4 with various MDR3 expression levels correlated with the amount of MDR3 P-gp determined on immunoblot (Fig. 2).

Transport of Drugs by MDR3-transfected Cells-
Transport rates of 0.1 M PSC 833 and 2 M cyclosporin A (CsA), dexamethasone, ondansetron, and morphine did not differ significantly between MDR3 transfectants and parental cells (not shown). Directional transport of these drugs was readily detectable in the MDR1-expressing cells at these concentrations (not shown; c.f. Refs. 25, 29, and 38).
Inhibition of Drug Transport by Reversal Agents-The active transport of drugs by MDR1 P-gp through the monolayers of LLC-PK1-derived cell lines can be inhibited by the addition of P-gp substrates and inhibitors, also called reversal agents. We have tested the effect of several of these inhibitors on MDR3 P-gp-mediated digoxin transport. Fig. 3 shows that the increased transport of 2 M digoxin through the LLC-M3.4.44 monolayers is abolished by 10 M cyclosporin A, 5 M PSC 833, 20 M verapamil, 10 M vinblastine. We also found inhibition by 10 M paclitaxel (not shown). The low level of digoxin transport through the parental cells, due to the presence of endogenous porcine transporters (25,29), is inhibited only partially by verapamil, whereas this drug completely inhibits the increased MDR3 P-gp-mediated transport. In the MDR1 transfectant, inhibition of digoxin transport by PSC 833 is complete, and the other reversal agents are not able to inhibit directional transport of digoxin by MDR1 P-gp completely at the concentrations used here. Fig. 4 shows that the MDR3 transfectant accumulated far less digoxin than the parental cells as compared with conditions where active digoxin transport was in-hibited by 5 M PSC 833. This suggests that the MDR3 P-gp in the transfected cells extrudes digoxin or prevents its influx by removal of the drug from the inner leaflet of the plasma membrane. However, indirect effects of MDR3 P-gp on drug transport (e.g. by influencing the activity of another transporter) are not excluded by these experiments.
Inhibition of Lipid Translocation by MDR3 P-gp-An established physiological function of MDR3 P-gp is the translocation of PC (16,22). We therefore tested whether the substrates transported by MDR3 transfectants and the inhibitors blocking this process had an effect on the translocation of PC to the extracellular leaflet of the apical membrane. Instead of PC, we used C 6 -NBD-PC, a fluorescent short chain analog of PC, as the translocation substrate (23). In order to measure specifically the translocation across the plasma membrane, the intracellular pool of newly synthesized C 6 -NBD-PC was double-labeled with [ 3 H]choline, and vesicular transport mechanisms were inhibited by performing the experiment for 3 h at 15°C. Fig. 5A shows that reversal agents that inhibit the increased drug transport in the LLC-M3.4.44 cells also decrease the rate of C 6 -NBD-PC translocation by MDR3 P-gp in these cells. Lipid translocation is also reduced in the presence of a 10 M concentration of the putative substrate drugs paclitaxel and vinblastine, suggesting that a direct interaction exists between MDR3 P-gp and the compounds tested. To exclude the possibility that the inhibition of lipid translocation was due to a nonspecific toxic effect (e.g. on the cytoskeleton) rather than a direct interaction with the transporter, the effect of paclitaxel was measured on the lipid translocation by the multidrug resistance-associated protein MRP1 (26). We have shown that MRP1 translocates C 6 -NBD-GlcCer in MRP1-transfected LLC-PK1 cells (39), whereas MRP1 is known to be a poor transporter of paclitaxel (40,41). We found that 10 M paclitaxel did not inhibit the translocation of C 6 -NBD-GlcCer to the outer leaflet of the membrane, whereas 2 mM sulfinpyrazone, a known inhibitor of MRP1 (26), completely abolished transport of the lipid (Fig. 5B). The results in Fig. 5, A and B, indicate that paclitaxel has an inhibitory effect on MDR3 P-gp itself rather than on lipid metabolism or lipid trafficking in general.
Attempts to inhibit MDR3 P-gp lipid translocation activity with 10 M digoxin were unsuccessful (Fig. 5A). However, translocation of C 6 -NBD-GlcCer and C 6 -NBD-PC by MDR1 P-gp was not inhibited by 10 M digoxin either (Fig. 5C), showing that digoxin is unable to inhibit P-gps in this assay system. This is probably due to the inability of digoxin to penetrate the cells at 15°C. To study translocation of PL from the inner to the outer leaflet of the plasma membrane, all assays have to be performed at 15°C to minimize the contribution of vesicular transport to the phospholipid reaching the outer leaflet of the plasma membrane. At 15°C, however, we find no diffusion of digoxin through the monolayer, suggesting that digoxin is unable to pass the lipid bilayer and reach P-gp at this temperature.
Expression of MDR3 in Sf9 Cells-To verify that the MDR3 P-gp can interact directly with drugs, we analyzed whether the drugs transported by this P-gp also affect enzymatic reactions associated with P-glycoprotein, such as ATPase activity and vanadate-dependent nucleotide trapping. MDR3 expression levels in mammalian cells were not high enough for studying these partial reactions, and we therefore cloned MDR3 cDNA into baculovirus and expressed the protein in insect (Sf9) cells. Since baculovirus vectors containing this cDNA could not be amplified in any bacterial strain, we had to clone it directly into baculovirus by co-transfection of Sf9 cells (see "Experimental Procedures"). As shown in Fig. 6A Nucleotide Trapping in MDR3 P-gp-Low concentrations of vanadate can inhibit ATP-hydrolysis by MDR1 P-gp, probably by replacing inorganic phosphate bound to the protein. This results in the formation of a complex between ADP and the protein that cannot be dissolved by washing with high MgATP concentrations (42,43). Vanadate-dependent nucleotide trapping by MDR3 P-gp was studied in isolated Sf9 cells mem- This labeling is also NEM-sensitive (not shown) but is greatly stimulated by verapamil (lane 7) (44). No nucleotide trapping was observed in either P-gp if no vanadate was added to the incubation media. Equal amounts of P-gp were present in each lane as verified in each experiment by immunoblotting of the same membrane filters (Fig. 7B).
In experiments not documented here, we studied the time course and ATP dependence of nucleotide trapping in MDR3 P-gp. Labeling increased up to 20-min incubation time at 37°C and up to 8-azido-ATP concentrations of 50 M. Since modulating effects of substrates or inhibitors are expected to be more pronounced at relatively lower levels of nucleotide trapping (44), we have performed such experiments at conditions similar to those in Fig. 7. Table I summarizes the effects of several compounds on the nucleotide trapping by MDR3 P-gp in the presence of 5 M ATP during 2-min incubations at 37°C. A strong inhibition of nucleotide trapping in MDR3 P-gp was observed after the addition of several compounds that inhibit drug transport by this protein. Phospholipids (phosphatidylcholine, phosphatidylserine, or phosphatidylinositol) added as sonicated liposomes up to 50 g/ml did not modulate nucleotide trapping of MDR3 P-gp or MDR1 P-gp (not shown). This may be due to saturation of P-gp with endogenous lipids or to a low spontaneous exchange of PL between the liposomes and the membrane vesicles.

DISCUSSION
Previous attempts to demonstrate drug transport by the human MDR3 P-gp and its murine homolog Mdr2 P-gp have met with little success (4, 6 -9). It came therefore as a surprise that expression of MDR3 in kidney cell monolayers results in the directional transport of digoxin, paclitaxel, daunorubicin, vinblastine, and ivermectin. The transport of these drugs is specific, since the transport of several other drugs and reversal agents is not increased in the MDR3 transfected cell lines. The rate of digoxin transport correlates with the amount of MDR3 P-gp in various subclones of MDR3 transfectant LLC-M3.4 (Fig. 2). Transport is inhibited by the MDR1 P-gp reversal agents, CsA, PSC 833, and verapamil, agents that also inhibit the translocation of short-chain PC analogs by the MDR3 P-gp. Conversely, C 6 -NBD-PC translocation is also inhibited by the MDR3 P-gp substrates, paclitaxel and vinblastine. PSC 833 causes an increase in the accumulation of digoxin in the LLC-M3.4.44 cells, suggesting that in the absence of PSC 833 digoxin is extruded from these cells against a concentration gradient.
Two earlier studies have suggested that the MDR3 P-gp may be able to transport some drugs. Nooter et al. (13) and Herweijer et al. (14) found overexpression of MDR3 (and not of MDR1) in certain drug-resistant leukemias and an increase in daunorubicin accumulation after treatment of the cells with CsA. This is in line with our results on daunorubicin transport by MDR3 P-gp and its inhibition by CsA. Like our assays, their accumulation experiments had a sensitive read-out that allowed the detection of small changes in transport efficiency.
Drug resistance caused by MDR3 P-gp has been shown by Kino and co-workers (15) in an MDR3 transformed yeast strain. In their experiments, resistance to aureobasidin A, a fungicide, correlated with the expression of MDR3 and was inhibited by high concentrations of CsA, verapamil, and vinblastine. The transport of aureobasidin A by MDR1 P-gp was more efficient than by MDR3 P-gp, comparable with what we find for the drugs tested in our study.
To exclude the possibility that the drug transport found in MDR3-transfected cells is only an indirect effect of the presence of MDR3 P-gp (e.g. because it activates an endogenous transporter), we sought direct evidence of binding of drug to MDR3 P-gp. Experiments with the MDR1 P-gp have shown that this transporter catalyzes two reactions that can be used to study protein-drug interaction, hydrolysis of ATP (ATPase activity) and trapping of nucleotide in the ATP binding site in the presence of vanadate (nucleotide trapping). We have studied both reactions for MDR3 P-gp, expressed to high levels in Sf9 cells. We were unable to find any ATPase activity of MDR3 P-gp above the background present in uninfected Sf9 cells with or without a range of phospholipids or drugs, under conditions that allow ready detection of the ATPase activity of MDR1 P-gp (35) and the multidrug resistance protein 1 (results not shown). This negative result could be due to a low turnover number of MDR3 P-gp, but other explanations cannot be excluded at this stage.
We were able to demonstrate vanadate-dependent nucleotide trapping by MDR3 P-gp, however, and substantial drug effects on this process. Whereas most drugs at moderate concentrations increase the rate of nucleotide trapping by MDR1 P-gp and only decrease this rate at much higher concentrations, we only observed inhibitory effects on MDR3 P-gp. The drug concentrations required for inhibition of trapping (Table I) were similar to those needed to inhibit transport of NBD-PC (Fig. 5A) or drugs (Fig. 3) by MDR3 (and MDR1 P-gp). We conclude that the MDR3 P-gp can bind a range of cytotoxic drugs and P-gp inhibitors with relatively high affinity.
If drug is bound with rather high affinity to MDR3 P-gp, why is this P-gp such a poor drug transporter that it is unable to cause MDR in transfected cells? We consider three explanations for this paradox. 1) PC is the preferred substrate of MDR3 P-gp. Since membranes are full of PC, drugs cannot compete with PC. This explanation is in line with our inability to block PC transport with PSC 833 in mice in vivo. Mayer et al. (45) have shown that PSC 833 has no effect at all on secretion of PC into bile at drug concentrations that strongly inhibit the secretion of digoxin or paclitaxel from the liver into bile. Since the MDR3 P-gp concentration has been shown to be a controlling step for PC secretion into bile (18,46), it is unlikely that a  partial inhibition by PSC 833 would have been missed in these experiments. The fact that we can readily inhibit C 6 -NBD-PC transport with reversal agents in vitro (Fig. 5) could be due to a lower affinity of MDR3 P-gp for short chain PC than for natural PC; in vitro inhibition of natural PC transport by reversal agents has not been tested. 2) PC is not a preferred substrate for MDR3 P-gp, and the protein transports drugs at least as well (or better) than PC. Transport of drugs does not lead to drug resistance, because the catalytic efficiency is lower than of MDR1 P-gp. Our inability to detect ATPase activity associated with MDR3 P-gp is in line with this explanation. It can also account for the PSC 833-sensitive biliary excretion of digoxin and paclitaxel in mice that lack both the Mdr1a and the Mdr1b P-gp in the bile canalicular membrane (45,47). Mayer et al. (45) conclude that another digoxin transporter must be present in the canalicular membrane, in addition to the Mdr1a/b P-gps. Since human MDR3 P-gp can transport both digoxin and paclitaxel and is sensitive to inhibition by PSC 833, the unidentified drug transporter could be the Mdr2 P-gp. 3) A compromise between explanations 1 and 2 is that PC is the preferred substrate but that the transport of natural PC by MDR3 P-gp might be so efficient that the inner leaflet of the membrane around the transporter runs out of substrate. If the (bile-type) PC molecules in the proximity of the transporter are transported to the outer leaflet of the membrane fast, this will leave only (non-bile-type PC molecules) aminophospholipids and drugs. For the lack of better substrates, the latter may then be transported as well. This hypothesis implies that the transport of PC by MDR3 P-gp is faster than the lateral diffusion of PC molecules in the inner leaflet of the membrane or backflipping from the outer leaflet. The very high concentration of MDR3 P-gp in the canalicular membrane (18) would therefore create the conditions in which this P-gp would be able to contribute to drug transport. It should be obvious that none of these explanations is satisfactory. Explanations 1 and 3 do not explain high affinity binding of drugs under conditions where PC transport is low (i.e. the nucleotide-trapping experiments at low ATP concentration). Explanation 2 cannot account for lack of inhibition of PC secretion by PSC 833 in vivo. Clearly, it is important to test the effect of PSC 833 and other inhibitors on the transport of PC by MDR3 P-gp in cell lines. Unfortunately, this is technically difficult.
The MDR3 and MDR1 genes are very similar. They both contain 27 introns, inserted at identical positions in the coding sequence (48). The proteins encoded by these genes have virtually identical hydropathy plots; they are 77% identical and 82% similar in amino acid sequence (5,49). It does not require a bold leap of the imagination to infer that the PC translocators encoded by the MDR3 and Mdr2 genes evolved from drug transporting P-gps by gene duplication during vertebrate evolution when the need arose to package the aggressive bile salts required for fat digestion in mixed micelles with phospholipids. It is unlikely, although still untested, that MDR1-type P-gps transport natural membrane phospholipids. This would interfere with their physiological function, protection against toxins, and could lead to futile flipping of, for example, phosphatidylethanolamine, resulting in useless ATP hydrolysis.
Since bile appeared in evolution more than 100 million years ago, it is remarkable that the MDR3 P-gp is still able to transport/bind a range of drugs if its only function is PC transport. We briefly consider three explanations for this paradox. 1) it is possible that drug transport by MDR3 P-gp is an accidental side product of evolution. There is evidence for gene conversion between the MDR1 and MDR3 genes (4), and such gene conversion may have counteracted the genetic drift accompanying the evolutionary optimization of the protein as a PC transporter. 2) optimal transport of PC may require a protein that transports some drugs as an unavoidable consequence. We find that unlikely. ABC transporters have been adapted in evolution to a variety of highly specialized tasks, and, in fact, the substrate specificity of the MDR3 P-gp toward membrane PLs is exquisite (23). 3) MDR3 P-gp might be a dual function protein, available to transport PC but also some toxins that are especially threatening to the liver, the only organ where the protein is present in high concentrations.
At present, both explanations 1 and 3 are compatible with the experimental evidence. More work on the effect of inhibitors on the transport of long-chain PC and on binding of drugs to MDR3 P-gp reconstituted in PC-free membranes is required to fully define the physiological functions of this interesting transporter. The importance of a better definition is reinforced by the discovery that the complete absence of the MDR3 P-gp in humans causes a severe liver disease requiring liver transplantation (50) and that women heterozygous for a MDR3 null allele are at risk for developing intrahepatic cholestasis of pregnancy (51). It is therefore conceivable that the type of drug interactions with MDR3 P-gp described here could result in disease of heterozygotes under conditions of stress requiring maximal phospholipid secretion into bile.