Transport of Cyclic Nucleotides and Estradiol 17-β-d-Glucuronide by Multidrug Resistance Protein 4

Human multidrug resistance protein 4 (MRP4) has recently been determined to confer resistance to the antiviral purine analog 9-(2-phosphonylmethoxyethyl)adenine and methotrexate. However, neither its substrate selectivity nor physiological functions have been determined. Here we report the results of investigations of thein vitro transport properties of MRP4 using membrane vesicles prepared from insect cells infected with MRP4 baculovirus. It is shown that expression of MRP4 is specifically associated with the MgATP-dependent transport of cGMP, cAMP, and estradiol 17-β-d-glucuronide (E217βG). cGMP, cAMP, and E217βG are transported withK m and V max values of 9.7 ± 2.3 μm and 2.0 ± 0.3 pmol/mg/min, 44.5 ± 5.8 μm and 4.1 ± 0.4 pmol/mg/min, and 30.3 ± 6.2 μm and 102 ± 16 pmol/mg/min, respectively. Consistent with its ability to transport cyclic nucleotides, it is demonstrated that the MRP4 drug resistance profile extends to 6-mercaptopurine and 6-thioguanine, two anticancer purine analogs that are converted in the cell to nucleotide analogs. On the basis of its capacity to transport cyclic nucleotides and E217βG, it is concluded that MRP4 may influence diverse cellular processes regulated by cAMP and cGMP and that its substrate range is distinct from that of any other characterized MRP family member.

The multidrug resistance protein (MRP) 1 family of ATPbinding cassette transporters first came to light as a result of the identification in drug-resistant cell lines of the M r 190,000 protein product and cDNA of its founding member, MRP1 (1)(2)(3). Based upon the determination of complete coding sequences and putative topologies, this family is now known to consist of at least seven members (4). The substrate selectivities and drug resistance profiles of several of these pumps have been determined. MRP1, MRP2 (cMOAT), and MRP3 (MOAT-D), which confer resistance to certain natural product agents and methotrexate (5)(6)(7)(8)(9)(10)(11)(12)(13)(14), are the best characterized family members. These three transporters are lipophilic anion pumps whose substrates include glutathione S-conjugates, such as leukotriene C 4 (LTC 4 ) and S- (2,4-dinitrophenyl)glutathione (DNP-SG), and glucuronate conjugates such as estradiol 17-␤-D-glucuronide (⌭ 2 17␤G) (10,11,(15)(16)(17)(18)(19)(20)(21)(22). However, whereas MRPs 1-3 have similar substrate ranges, they subserve distinct physiological functions. MRP1 is distinguished from MRP2 and MRP3 by its higher affinity for LTC 4 , a feature that is reflected in the specific role it plays in mediating immune responses involving cellular export of this cysteinyl leukotriene (23,24). By contrast with MRP1, which is ubiquitously expressed and localized at basolateral surfaces of polarized cells (25)(26)(27), MRP2 is primarily expressed in the hepatocyte canaliculus where it functions as an apical efflux pump for organic anions such as bilirubin glucuronide and in provision of the biliary fluid constituent glutathione (28). MRP3 is also a glutathione and glucuronate conjugate pump but has the additional capability of mediating the transport of monoanionic bile acids (22,29). This substrate selectivity, together with its induction at basolateral surfaces of hepatocytes under cholestatic conditions (30 -32), has led to the notion that it functions as a compensatory backup mechanism to eliminate from hepatocytes potentially toxic compounds that are ordinarily excreted into the bile.
Two members of the MRP family, MRP4 (MOAT-B) and MRP5 (MOAT-C), are no more related to each other than they are to MRPs 1-3 in terms of degree of amino acid identity, but they are structurally distinct from the latter proteins in that MRPs 4 and 5 do not possess a third (N-terminal) membrane spanning domain (4,33,34). The topological dissimilarity of MRP5 is reflected in its distinct drug resistance capabilities and substrate selectivity. By contrast with MRPs 1-3, MRP5 is not known to confer resistance to natural product anticancer agents or methotrexate, but instead it has the facility for conferring resistance to purine analogs (35,36). Similarly, membrane vesicle transport assays suggest that glutathione and glucuronate conjugates are not substrates of MRP5. Instead it is able to transport cyclic nucleotides (37).
The functional characteristics of MRP4, the other MRP family member that lacks an N-terminal membrane spanning domain, have yet to be defined in any detail. The drug resistance capabilities of MRP4 have been assessed to some degree in transfected NIH3T3 cells and in a drug-selected cell line in * This work was supported in part by National Institutes of Health Grant CA73728 (to G. D. K.) and by an appropriation from the Commonwealth of Pennsylvania. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Materials and Cell
Immunoblot Analysis-Membrane vesicles preparations were analyzed by 7.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis, as described previously (41). Proteins were transferred to nitrocellulose filters using a wet transfer system as described previously (42). MRP4 was detected using monoclonal MRP4 antibody (1:2000) and alkaline phosphatase-conjugated secondary antibody.
Preparation of Membrane Vesicles and Transport Experiments-Membrane vesicles were prepared by the nitrogen cavitation method as described previously (43). Transport experiments were performed using the rapid filtration method essentially as described previously (16). Transport experiments were carried out in medium containing membrane vesicles (10 g), 0.25 M sucrose, 10 mM Tris-HCl, pH 7.4, 10 mM MgCl 2 , 4 mM ATP, 10 mM phosphocreatine, 100 g/ml creatine phosphokinase, and radiolabeled substrate Ϯ unlabeled substrate, in a total volume of 50 l. Reactions were carried out at 37°C and stopped by the addition of 3 ml of ice-cold stop solution (0.25 M sucrose, 100 mM NaCl, 10 mM Tris-HCl, pH 7.4). Samples were passed through 0.22-m Durapore membrane filters (Millipore, Bedford, MA) under vacuum. The filters were washed 3 times with 3 ml of ice-cold stop solution and dried at room temperature for 30 min. Radioactivity was measured by the use of a liquid scintillation counter. Rates of net ATP-dependent transport were determined by subtracting the values obtained in the presence of 4 mM AMP from those obtained in the presence of 4 mM ATP. Uptake rates were linear for up to 5 min, and rates for concentration dependence experiments were measured at 5 min.
Drug Accumulation and Efflux-For accumulation experiments sub-confluent 3T3/pSR␣ and 3T3/MRP4-3 cells seeded in triplicate in 100-mm plastic dishes were incubated overnight in DMEM growth medium. After growth overnight the cells were incubated at 37°C with 10 M [ 14 C]6-MP for 10, 30, and 60 min. Cells were washed 3 times with cold PBS and immediately harvested by trypsinization. The cells were washed 2 times with cold PBS, and an aliquot was used to count cell number. Radioactivity was measured by the use of a liquid scintillation counter. For efflux experiments, subconfluent 3T3/pSR␣ and 3T3/ MRP4-3 cells seeded in triplicate in 100-mm plastic dishes were incubated overnight in DMEM growth medium. The next day the cells were washed and incubated for 2 h in energy depletion medium (glucose-free, pyruvate-free DMEM containing 10% dialyzed calf serum, 10 mM deoxyglucose, and 10 mM sodium azide) containing 10 M [ 14 C]6-MP. The cells were then washed 3 times with a total volume of 20 ml of PBS, and the medium was replaced with ordinary growth medium without radiolabeled drug. The cells were incubated at 37°C, and at various time points medium was collected for measuring radioactivity. Cell-associated radioactivity was counted at the end of the 2-h incubation in energy depletion medium containing radiolabeled drug and after 2 h of efflux. Data Analysis-Kinetic parameters were computed by nonlinear least squares analysis (44) using the Ultrafit computer software (Bio-Soft, Ferguson, MO). For drug sensitivity experiments the nonparametric two-tailed Wilcoxon test was used to make inferences about the significance of differences in the IC 50 values of MRP4-transfected and control cells.

RESULTS
Transport of cGMP, cAMP, and E 2 17␤G by MRP4 -MRP4dependent transport activity was assayed on density-fractionated membrane vesicles prepared from insect (Sf9) cells infected with MRP4 baculovirus. As determined by immunoblot analysis, these membranes are a rich source of MRP4 protein, which migrates as an M r 150,000 electrophoretic species (Fig.  1). As observed previously (38) MRP4 expressed in insect cells, which are unable to synthesize complex N-linked oligosaccharide chains, migrates with a lower apparent molecular weight than the same recombinant protein expressed in NIH3T3 cells (M r 170,000).
Previous studies employing MRP4-transfected NIH3T3 cells and PMEA-selected cells indicate that the pump confers resistance to the antiviral purine analog PMEA (38,39). The topological resemblance of MRP4 to MRP5 (4, 34), and the observation that MRP5, which also confers resistance to purine analogs including PMEA (36), is competent in the transport of cyclic nucleotides (37), suggested that cyclic nucleotides might also be transport substrates of MRP4. To explore this possibility, uptake of cGMP and cAMP into inside-out membrane vesicles prepared from insect cells infected with MRP4 baculovirus was examined. To assess the relative contribution of MRP4 to overall uptake, parallel experiments were also performed on membrane vesicles purified from uninfected insect cells.
Both cGMP and cAMP were indeed subject to MRP4-mediated, MgATP-dependent transport (Fig. 2, A and B). When measured at initial concentrations of 1.0 M and at the 5-min time point of the assay, [ 3 H]cGMP and [ 3 H]cAMP were taken up by MRP4-enriched vesicles at rates of 1.42 and 0.49 pmol/ mg/min, respectively, from media containing MgATP and at rates of less than 0.39 and 0.17 pmol/mg/min, respectively, from media containing MgAMP. By contrast, the rates of uptake of cGMP and cAMP, respectively, for membranes prepared from uninfected insect cells were consistently less than 0.39 and 0.17 pmol/mg/min, under either energized or nonenergized conditions.
Glucuronate and glutathione conjugates are established substrates of MRPs 1-3, but not of MRP5 (37). To determine whether conjugates are substrates of MRP4, transport of E 2 17␤G, DNP-SG, and LTC 4 , prototypical glucuronate and glutathione conjugates, were selected as model test compounds. Of these three compounds, robust uptake was observed only for E 2 17␤G (Fig. 2C). When measured at initial concentrations of 1.0 M and at the 5-min time point of the assay, [ 3 H]E 2 17␤G was taken up by MRP4-enriched membranes at a rate of 4.4 pmol/mg/min from media containing MgATP and at a rate of only 1.0 pmol/mg/min from media containing MgAMP. Uptake rates of less than 1.1 pmol/mg/min from media containing either MgATP or MgAMP were observed for membranes prepared from uninfected insect cells. Transport of [ 3 H]LTC 4 and [ 3 H]DNP-SG uptake was not consistently observed in that low levels of uptake of these compounds were detected in some but not all membrane vesicle preparations (data not shown).
Osmotic Sensitivity of [ 3 H]cGMP Transport by MRP4 -The osmotic sensitivity of [ 3 H]cGMP uptake was examined to confirm that radiolabel retained by MRP4-enriched membrane vesicles was largely attributable to transport of the substrate into the intravesicular compartment as opposed to nonspecific binding to the vesicles and/or filters. MgATP-dependent uptake of 1.0 M cGMP increased as a linear function of the reciprocal of the sucrose concentration of the uptake medium, indicating that the transport substrate was delivered into an osmotically active compartment (Fig. 3). By contrast, the sucrose concentration exerted a moderate effect on substrate retention measured in medium containing MgAMP, suggesting that under nonenergized conditions an appreciable fraction of the apparent uptake measured represented binding to the membranes and/or filters. The magnitude of the ordinate intercepts indicated that nonspecific substrate binding constituted 18% of the radiolabel retained by MRP4-enriched membranes in media containing MgATP, but as much as 62% of the radioactivity retained in media containing MgAMP. MgATP-dependent uptake of [   vesicles prepared from insect cells infected with MRP4 baculovirus approximated Michaelis-Menten kinetics. When measured over a broad range of substrate concentrations, the initial rates of MgATP-dependent uptake of both compounds, enumerated as the difference between uptake rates in media containing MgATP and uptake in media containing MgAMP, exhibited saturation kinetics (Fig. 4) (Table I). The efficiencies of transport fell in the rank order E 2 17␤G (V max /K m ϭ 3.4) Ͼ cGMP (0.21) Ͼ cAMP (0.09) ( Table I).
Inhibition of MRP4-mediated Transport of E 2 17␤G-These membrane vesicle experiments indicated that MRP4 is able to transport cGMP, cAMP, and E 2 17␤G, and in a previous report (38) we inferred from drug sensitivity studies that MRP4 can

FIG. 4. Concentration dependence of [ 3 H]cGMP, [ 3 H]cAMP, and [ 3 H]E 2 17␤G uptake by MRP4. The rates of MgATP-dependent uptake of [ 3 H]cGMP (A), [ 3 H]cAMP (B), and [ 3 H]E 2 17␤G (C) into membrane vesicles (10 g) prepared from insect cells infected with MRP4
baculovirus were measured at 37°C. Values shown (means Ϯ S.E.) are rates measured in the presence of MgATP minus rates measured in the presence of MgAMP for triplicate determinations. Uptake rates were measured at 5 min. The lines of best fit and kinetic parameters were computed by nonlinear least squares analysis (44). Representative experiments are shown.   transport methotrexate. Taken together these studies suggest that MRP4 is able to transport diverse amphipathic anions. To gain further insight into the substrate selectivity of MRP4, the ability of cGMP, cAMP, and methotrexate to inhibit transport was examined in experiments in which E 2 17␤G was employed as the test substrate. As would be expected if these substrates were transported by a common mechanism, all three were capable of inhibiting E 2 17␤G transport (Table II). cGMP was the most potent inhibitor (83.8% inhibition at 300 M), consistent with its higher affinity by comparison with cAMP. The degree of inhibition exerted by methotrexate (59.7%) was comparable to that of cAMP (60.4%) at 300 M concentrations but slightly higher at concentrations of 30 and 100 M. Sensitivity of MRP4-transfected NIH3T3 Cells to Anticancer Purine Analogs-By having determined that the pump has the facility for MgATP-dependent transport of cyclic purine nucleotides, and knowing from previous studies (38,39) that MRP4 confers resistance to the antiviral purine analog PMEA, an acyclic phosphonate that cannot be converted in the cell to a nucleoside phosphate (45), the involvement of MRP4 in resistance to anticancer purine analogs that are known to be metabolized in the cell to nucleotides was examined. For this purpose growth assays were performed on MRP4-transfected NIH3T3 cells (3T3/MRP4-3) and parental vector-transfected control cells (3T3/pSR␣) in the presence and absence of several agents. From these experiments it was determined that MRP4 is not only able to confer resistance to antiviral agents but also to anticancer purine analogs.
Three purine analogs, 6-MP, 2-MP, and 6-TG, and two purine nucleoside analogs, CDA and DCF, were selected for analysis. Of the agents examined, 3T3/MRP4-3 exhibited significantly higher levels of resistance by comparison with 3T3/pSR␣ for two of the three purine analogs (Fig. 5 and Table III). The MRP4-transfected cells exhibited 4.6-fold resistance for 6-MP and 2.7-fold resistance for 6-TG. A difference in sensitivity was also observed for 2-MP, but this value did not reach statistical significance. By contrast, MRP4 did not confer resistance to the two anticancer nucleoside analogs tested.
Analysis of [ 14 C]6-MP Accumulation and Efflux in MRP4transfected NIH3T3 Cells-To gain insight into the mechanism by which MRP4 confers resistance to purine analogs, accumulation and efflux of [ 14 C]6-MP were analyzed in 3T3/MRP4-3 and 3T3/pSR␣ cells. From these experiments it was determined that expression of MRP4 is associated with reduced drug accumulation and enhanced drug efflux, as would be expected if resistance were based upon the operation of a plasma membrane efflux pump.
To determine whether reduced accumulation was consequent upon enhanced drug efflux, extrusion of radiolabeled drug into the growth medium was measured over a 2-h time course. In order to perform this experiment under conditions in which intracellular drug levels were comparable in the two cell lines at the beginning of the assay, 3T3/MRP4-3 and 3T3/pSR␣ cells were first incubated in the presence of 10 M [ 14 C]6-MP under energy-depletion conditions for 2 h. Following this incubation period, the growth medium was replaced with complete medium lacking drug, and efflux of radiolabeled drug into the medium was measured (Fig. 6B). As anticipated, 3T3/MRP4-3 cells exhibited markedly enhanced drug efflux by comparison with the control cells. After 30 min the MRP4-transfected cells effluxed 1.8-fold more drug than the control cells, and this ratio  Examination of intracellular drug content at the beginning and ending of the efflux assay indicated that by contrast with the ability of MRP4 to diminish accumulation under ordinary growth conditions (Fig. 6A), under energy-depletion conditions drug accumulation in the two cell lines differed by no more than 6% (Fig. 6C). However, after 2 h of efflux in complete media intracellular drug in 3T3/MRP4-3 cells was 0.48-fold less than the 3T3/pSR␣ control cells (Fig. 6C). This value is in reasonably good agreement with the 1.8-fold increased level of drug effluxed into the medium by 3T3/MRP4-3 cells (Fig. 6B). DISCUSSION In the present study the in vitro transport properties of human MRP4 were investigated to gain insight into its substrate selectivity and potential physiological functions. Cyclic nucleotides were selected as one class of target compounds because MRP4 has been determined previously to confer resistance to the antiviral nucleotide analog PMEA (38,39) and because cyclic nucleotides have been established recently as transport substrates of an MRP family member (MRP5) whose protein topology resembles that of MRP4 (37). In agreement with these structural and drug resistance features, it was determined that MRP4 can indeed transport cGMP and cAMP. In addition, it is shown that cGMP is a higher affinity substrate of MRP4 than is cAMP, as is also the case for MRP5. However, whereas both transporters are able to mediate transport of cyclic nucleotides there are significant differences in their kinetic parameters. The affinity of MRP4 for cGMP (K m ϭ 9.7 M) is ϳ5-fold lower than that of MRP5 (K m ϭ 2.1 M) (37). By contrast, the affinity of MRP4 for cAMP (K m ϭ 44.5 M) is ϳ9-fold higher than that reported for MRP5 (K m ϭ 379 M). The markedly higher affinity of MRP4 for cAMP may be of considerable significance in view of the involvement of this signaling molecule in diverse regulatory processes.
Cellular efflux of cyclic nucleotides has been described in both prokaryotes and eukaryotes (46 -49). Analyses employing a variety of cultured cells and membrane vesicle preparations have established that cyclic nucleotide efflux in mammals is energy-dependent and mediated by amphipathic anion transporters in that it can be blocked by inhibitors of organic anion pumps (46, 47, 50 -63). The present study and that by Jedlitschky et al. (37) indicate that MRP4 and MRP5, respectively, are components of the previously described membrane efflux systems for these critical signaling molecules (Fig. 7). However, while these studies have identified molecular components of the export systems, the precise physiological roles of cyclic nucleotide efflux are not completely understood. A well defined role for cellular export of cyclic nucleotides is best established for the slime mold Dictyostelium discoideum, for which cAMP effluxed by solitary amoebae under low nutrient conditions acts both as a chemoattractant that mediates formation of multicellular aggregates and as a differentiation agent (64). In mammals it is thought that efflux of cyclic nucleotides subserves two functions. One proposed function is that it contributes to the modulation of cyclic nucleotide signaling by reducing intracellular levels of these second messengers. In support of this notion is the consistent observation that triggered elevations in intracellular cyclic nucleotide levels are associated with enhanced cellular efflux (52)(53)(54)(55)57). A second proposed function for efflux is in provision of extracellular cAMP involved in intercellular signaling. This idea is consistent with the detection of cAMP in a variety of extracellular fluids (65)(66)(67) and is also supported by characterizations of cellular activities attributed to extracellular cAMP and presumably mediated by proteins located in the plasma membranes of target cells (see for example Refs. 68 -71). It might be expected that MRP4, as a result of its higher affinity for cAMP by comparison with MRP5, plays a more prominent role in modulating intracellular cAMP levels and in efflux of cAMP involved in intercellular signaling. On the other hand MRP5 might be a more potent factor in the modulation of intracellular cGMP levels. Detailed studies concerning the tissue-specific expression patterns of MRP4 and MRP5, which are currently understood primarily at the transcript level (33,34,38,72), should provide further insights as to which of these pumps are deployed in specific situations.
Whereas the substrate selectivity of MRP4 is similar to that of MRP5 with regard to transport of cyclic nucleotides, our experiments indicate that there are also significant differences. By contrast with MRP5 (37), MRP4 is able to transport the glucuronide E 2 17␤G. In this regard MRP4 is similar to MRPs 1-3, for which this compound is an established substrate. The affinity of E 2 17␤G transport by MRP4 (K m ϭ 30.3 M) is comparable to the K m value we previously reported for human MRP3 (25.6 M) (22). However, both MRP1 (K m ϭ 1.5-2.5 M) and MRP2 (K m ϭ 7.2 M) are higher affinity transporters of this substrate (10,15,17). In addition to the transport of E 2 17␤G, the substrate range of MRP4 is distinct from MRP5 with regard to at least one other compound, namely methotrexate. Transport of this anionic antimetabolite was inferred previously from studies demonstrating that MRP4-transfected cells are resistant to and accumulate reduced amounts of this agent (38). In further support of the notion that methotrexate is an MRP4 substrate, it is demonstrated here that this agent can inhibit MRP4-mediated transport of E 2 17␤G. As with FIG. 7. Schematic diagram depicting the role played by MRPs in the cellular physiology of cyclic nucleotides. cGMP is synthesized by guanylyl cyclases located in the cytoplasm (sGC) and plasma membrane (GC) when triggered by nitric oxide (NO) or peptide ligands (L), respectively. cAMP is synthesized by adenylyl cyclases (AC) located in the cytoplasm (sAC) or associated with G protein-coupled receptors (GPCR) in the plasma membrane, when trigged by bicarbonate or peptide ligands (L), respectively. Cyclic nucleotides are enzymatically degraded by specific phosphodiesterases (PDEs) or extruded from the cell by an efflux system that includes MRP4 and MRP5. E 2 17␤G transport, MRP4-mediated transport of methotrexate is similar to MRPs 1-3 which are also able to confer resistance to and transport methotrexate (13,14,22).
In the present study we demonstrate that the drug resistance profile of MRP4 extends beyond the antiviral purine analog PMEA, an acyclic phosphonate, to include the commonly used anticancer purine analog 6-MP. It is further demonstrated that the pump functions to reduce intracellular levels of this agent by an energy-dependent efflux mechanism. However, we have not detected transport of 6-MP in membrane vesicle assays. 2 It is therefore unlikely that 6-MP or 6-TG, both of which are uncharged purine base analogs, are direct substrates of MRP4. Rather, the facility of MRP4 for transporting cyclic nucleotides and E 2 17␤G, both of which are amphipathic anions, suggests that the nucleotide metabolites of 6-MP and 6-TG, which are the toxic forms of these agents, are likely to be the anionic species effluxed by the pump. By contrast with 6-MP and 6-TG, PMEA is an amphipathic anion (45). Hence, in this case it is likely that either PMEA and/or its di-or triphosphorylated metabolites are direct substrates of MRP4. These notions concerning how MRP4 confers resistance to antiviral and anticancer purine analogs are supported by analyses of species effluxed from MRP4-overexpressing CEMr-1 cells treated with PMEA and MRP5-transduced cells treated with PMEA and 6-MP (13,73).
In view of the fact that 6-MP and methotrexate are significant components of chemotherapeutic regimens used in the treatment of childhood leukemias, the ability of MRP4 to confer resistance to both of these antimetabolites is noteworthy. In this regard MRP4 is unique among characterized MRP family members that confer resistance to either methotrexate (MRPs 1-3) or 6-MP (MRP5) but not to both agents. Whether MRP4 or MRP5 contributes to clinical resistance associated with either of these agents remains to be determined.