Transport of Methotrexate (MTX) and Folates by Multidrug Resistance Protein (MRP) 3 and MRP1: Effect of Polyglutamylation on MTX Transport¹

As essential cofactors for the synthesis of nucleotides, reduced folates are required for DNA synthesis in dividing cells. This requirement has been extensively exploited in cancer treatment by the use of FA⁵ analogues, most notably MTX, as chemotherapeutic agents. The cellular pharmacology of MTX is relatively well defined. It has been determined that cellular uptake occurs primarily by carrier-mediated transport via the RFC1 (1, 2, 3, 4) and that once inside the cell, MTX acts as a potent competitive inhibitor of DHFR. Inhibition of DHFR results in a depletion of the intracellular reduced folate pools required for the biosynthesis of purines and thymidine.

A crucial property of MTX is its susceptibility to polyglutamylation by the same enzymes that polyglutamylate physiological folates (5, 6, 7, 8, 9, 10). FPGS catalyzes the condensation of successive glutamate residues to the γ-carboxyl group of MTX, a monoglutamate, to yield MTX-Glu_2–7 derivatives. By comparison with the parent molecule, which is subject to efflux via an energy-dependent process, polyglutamylated MTX is effluxed poorly and exhibits prolonged intracellular retention (11, 12, 13, 14, 15, 16, 17). As a consequence, polyglutamylation results in a massive enhancement of cytotoxicity. The precise mechanism of efflux of this agent is therefore fundamental to its potency as a chemotherapeutic and also has a bearing on the physiology of endogenous folates because their retention is similarly dependent on polyglutamylation (18). However, the long-standing issue of the identity of the cellular component(s) responsible for efflux of this agent remains unresolved (1).

Clues regarding the molecular identity of one or more components of the cellular efflux system have come recently from investigations of the transport properties of members of the MRP family of ATP-binding cassette transporters. Rat strains deficient in MRP2 are impaired in hepatobiliary excretion of MTX (19), and canalicular membrane vesicles prepared from MRP2-deficient rat strains have a diminished capacity for the MgATP-dependent transport of physiological folates (20). Cultured cells transfected with MRP1, MRP2, MRP3, or MRP4 are resistant to and accumulate lower levels of MTX, particularly in short-term drug exposure assays (21, 22, 23, 24). Cloned human MRP3 is not only able to transport glutathione and glucuronate conjugates and monoanionic bile acids but is also able to transport MTX in the low micromolar range (25). A similar substrate selectivity has been reported for rat MRP3 (26, 27). The kinetics have not been defined, but MRP1 and MRP2 have also been found to transport MTX into membrane vesicles (23).

If these findings are indeed to be consistent with a role for at least some MRP family members in the energy-dependent cellular efflux system for MTX, then they must be able to accommodate a critical feature of the system: the capacity of the system to mediate the efflux of MTX but not that of its polyglutamylated metabolites. Here we examine the potential of MRP3 participation in these and related processes by determining the selectivity of MRP3 for MTX versus its metabolites and physiological folates. The results of these analyses show that the addition of only a single glutamate residue to MTX is sufficient to severely attenuate MRP3-mediated transport and that MRP3 is competent in the transport of FA and N⁵-formyltetrahydrofolic acid (leucovorin) as well as MTX. Parallel experiments on another MRP, MRP1, demonstrate similar properties.

Two conclusions derive from these findings: (a) MRP3 and MRP1 satisfy the requirements of components of the MTX efflux system; and (b) given that MRP3 and MRP1 are able to transport physiological folates at high capacity, they may not only indirectly influence antifolate cytotoxicity but may also contribute to cellular resistance in chemotherapeutic regimens in which leucovorin is a component.

[³H]MTX (26.8 Ci/mmol), [³H]MTX-Glu₂ (19.7 Ci/mmol), [³H]MTX-Glu₃ (23.2 Ci/mmol), [³H]FA (20.2 Ci/mmol), [³H]N⁵-formyltetrahydrofolic acid (26 Ci/mmol), and unlabeled MTX-Glu₂ were purchased from Moravek Biochemicals (Brea, CA). Creatine kinase and creatine phosphate were purchased from Boehringer Mannheim (Indianapolis, IN). Glycocholate, taurocholate, verapamil, CsA, probenecid, etoposide, vincristine, doxorubicin, ATP, and AMP were purchased from Sigma Chemical Co. (St. Louis, MO). ZD1694, MK571, ONO-1078, PAK-104P, AG-A, and PSC833 were kindly provided by AstraZeneca (Wilmington, DE), Dr. A. W. Ford-Huchinson (Merck-Frosst Center for Therapeutic Research, Pointe Claire-Dorval, Canada), Ono Pharmaceutical Co. Ltd. (Osaka, Japan), Nissan Chemical Industries (Chiba, Japan), Drs. Motomasa Kobayashi and Shunji Aoki (Osaka University, Osaka, Japan), and Sandoz (Tsukuba, Japan), respectively. The MRP3-transfected human embryonic kidney HEK293 cell line (HEK/MRP3-5), parental vector-transfected HEK293 cell line (HEK/pcDNA3), MRP1-transfected NIH3T3 cell line (pSRα-MRP1-32), and parental vector-transfected NIH3T3 cell line (pSRα) were as described previously (21, 28).

Membrane vesicles were prepared as described previously (29), except that cell disruption was accomplished using a motor-driven Dounce homogenizer. The rates of MTX uptake were increased by approximately 25-fold over the values reported previously for this material (25) as a result of this modification of the preparation protocol. Vesicular uptake of [³H]MTX and its derivatives into inside-out membrane vesicles was measured by the rapid filtration technique as described previously (25, 29). For concentration dependence determinations, initial rates were measured at 5 min.

Kinetic parameters were computed by nonlinear least squares analysis (30) using the Ultrafit computer software (BioSoft, Ferguson, MO).

MRP3-dependent transport was assayed on density-fractionated membrane vesicles prepared from HEK293 cells transfected with MRP3 expression vector [HEK/MRP3-5 cells (21, 25)]. These membranes were previously determined to be a rich source of MRP3 protein as assessed by immunoblot analysis (25).

Previous studies on membrane vesicles purified from HEK/MRP3-5 cells have established that MRP3 is competent in the high-capacity MgATP-energized transport of MTX (25). However, whereas these studies demonstrated transport of MTX, they did not assess transport of MTX polyglutamates (structures are shown in Fig. 1). Thus, to explore the implications of polyglutamylation, the kinetics of uptake of MTX, MTX-Glu₂, and MTX-Glu₃ into MRP3-enriched membrane vesicles were compared. To assess the relative contribution of MRP3 to overall uptake, parallel experiments were also performed on membrane vesicles purified from HEK293 cells transfected with parental plasmid (HEK/pcDNA3 cells).

Of the three MTX species assayed, the parent compound was subject to the highest rates of MRP3-mediated, MgATP-dependent transport (Fig. 2,A). By comparison with the negligible rates of uptake of 1.0 μm [³H]MTX by HEK/MRP3-5 membrane vesicles in media containing MgAMP or HEK/pcDNA3 vesicles in media containing either MgATP or MgAMP, membrane vesicles purified from HEK/MRP3-5 cells mediated MgATP-dependent uptake at an initial rate of 5.7 pmol/mg/min (Fig. 2,A). By contrast, uptake of MTX-Glu₂, when measured by the same procedure on the same membrane preparations, was slow or undetectable. Whereas an increase in MgATP-dependent [³H]MTX-Glu₂ uptake was evident in HEK/MRP3-5 vesicles versus the equivalent membrane fraction from HEK/pcDNA3 cells, the rate (0.2 pmol/mg/min) was <5% of that measured with [³H]MTX (Fig. 2,B; note that in Fig. 2, the ordinates of B and C are ≥10 times more sensitive than those of A). The degree to which the transport of MTX was attenuated by glutamylation increased with increase in the number of glutamyl repeats because MgATP-dependent uptake of 1.0 μm [³H]MTX-Glu₃ was not detectable in any of the membrane vesicle preparations assayed (Fig. 2 C).

MgATP-dependent retention of [³H]MTX was largely attributable to its transport into the intravesicular compartment rather than nonspecific binding to the membrane vesicles or filters. MgATP-dependent uptake of 1.0 μm [³H]MTX into HEK/MRP3-5 vesicles increased as a linear function of the reciprocal of the sucrose concentration of the uptake medium, demonstrating that the vesicles were osmotically responsive and that the transported species was delivered into an osmotically active compartment (Fig. 3). By contrast, osmoticum exerted only a moderate effect on substrate retention measured in media containing MgAMP, suggesting that an appreciable fraction of the apparent uptake measured under nonenergized conditions represented binding to the membrane vesicles and/or membrane filters (Fig. 3).

The facility of MRP3 for the transport of MTX suggested the possibility that this pump might also transport physiological folates. To determine whether this was the case, two folate species were selected for analysis: the folate that MTX most closely resembles, FA, and N⁵-formyltetrahydrofolic acid (leucovorin), a reduced 1-carbon-bearing folate, which, in addition to being a physiological species, is also used as a cancer therapeutic (Fig. 1).

As shown in Fig. 4, MRP3 is indeed able to transport FA and leucovorin. When measured at an initial concentration of 1.0 μm, [³H]FA and [³H]leucovorin were taken up by HEK/MRP3-5 vesicles at rates of 2.1 and 2.7 pmol/mg/min, respectively, from media containing MgATP and at rates of <0.6 pmol/mg/min from media containing MgAMP (Fig. 4). The corresponding values for HEK/pcDNA3 vesicles were consistently <0.4 pmol/mg/min.

The substrate concentration dependence of MgATP-energized [³H]MTX, [³H]FA, and [³H]leucovorin uptake by membrane vesicles prepared from HEK/MRP3-5 cells approximated Michaelis-Menten kinetics. When measured over a broad range of substrate concentrations, the initial rates of MgATP-dependent uptake of all three compounds, enumerated as the difference between uptake in media containing MgATP and uptake in media containing MgAMP, exhibited saturation kinetics (Fig. 5). Nonlinear least squares fitting of the data to the Michaelis-Menten equation (30) yielded K_m and V_max values for the transport of MTX, FA, and leucovorin of 0.62 ± 0.23 mm and 2.93 ± 0.32 nmol/mg/min, 1.96 ± 0.13 mm and 1.71 ± 0.05 nmol/mg/min, and 1.74 ± 0.65 mm and 3.63 ± 1.20 nmol/mg/min, respectively (Table 1). The efficiencies of transport fell in the rank order MTX (V_max/K_m = 4.8) > leucovorin (V_max/K_m = 2.1) ≫ FA (V_max/K_m = 0.9; Table 1). Because membrane vesicles purified from HEK/pcDNA3 cells catalyzed negligible MgATP-dependent uptake of all three folates at all of the concentrations examined (Fig. 5), the transport measured in the membranes from HEK/MRP3-5 cells was solely attributable to MRP3.

To gain insight into the interactions of MRP3 with MTX and other compounds, a range of natural product agents to which MRP3 does (etoposide and vincristine) or does not confer resistance (doxorubicin; Ref. 21) and compounds that MRP3 does (E₂17βG and glycocholate) or does not transport (taurocholate; Ref. 25) were screened for their capacity to inhibit MTX transport.

With the exception of taurocholate, all of the compounds tested in this category inhibited MRP3-mediated MTX transport in a manner consistent with the results from previous resistance or transport studies. One hundred μm concentrations of the natural product anticancer agents inhibited MTX transport in the rank order etoposide > vincristine ≫ doxorubicin (89.4%, 28.3%, and 8.5% inhibition, respectively; Table 2). One hundred μm concentrations of E₂17βG and glycocholate inhibited MTX transport to degrees (91.7% and 73.2%; Table 2) commensurate with their relative affinities for the transporter (K_m(E217βG) = 25.6 μm, K_{m(glycocholate)} = 248 μm; Ref. 25). The exception, taurocholate, inhibited MTX transport by 56.1% (Table 2), although this bile acid, unlike glycocholate, is not transported by human MRP3. In addition, although human MRP3 is able to transport glutathione conjugates, glutathione itself was a poor inhibitor (13% inhibition at 1 mm).

Among known MRP family members, MRP3 bears the closest resemblance to MRP1 (31, 32). MRP3 and MRP1 are 58% sequence identical, and both transport glutathione and glucuronate conjugates (25, 33, 34, 35). To assess whether these similarities might be reflected in the inhibitor sensitivity of MRP3-mediated MTX transport, a number of compounds capable of inhibiting MRP1 were screened. These included the leukotriene D4 receptor antagonists MK571 and ONO-1078 (36, 37), AG-A, a natural product isolated from sponge (38), PAK-104P, a pyridine analogue (39), and probenecid, an amphipathic anion (40, 41). Of the MRP1 reversing agents tested, ONO-1078, MK571, and PAK-104P were found to be the most efficacious inhibitors of MRP3-mediated transport, inhibiting uptake by 100%, 98%, and 97%, respectively, at 100 μm concentrations (Table 2). The inhibitions exerted by 111 μm concentrations of probenecid and AG-A (36.5% and 15.5%, respectively) were moderate by comparison (Table 2).

Neither MTX-Glu₂ nor the quinazoline-based antifolate ZD1694, a monoglutamate that inhibits thymidylate synthase, was a potent inhibitor of MRP3-mediated MTX transport. Concentrations of 100 μm or more inhibited MTX uptake by only 12.7% and 26.6%, respectively (Table 2).

The inhibitory activities of three agents capable of modulating the drug resistance activity of Pgp, an ATP-binding cassette transporter that shares with MRP3 the facility for conferring resistance to natural product agents but has only a weak structural resemblance to MRP3, were examined to test for the effects of Pgp antagonists on MRP3-mediated MTX transport. Of these agents, CsA was a surprisingly potent inhibitor, whereas PSC833 exhibited moderate activity (73.7% and 38.6% inhibition at 10 μm, respectively). By comparison with the latter bulky lipophilic molecules, the calcium channel blocker verapamil was a weak inhibitor (13.1% inhibition at 111 μm). The effects of Pgp inhibitors on MRP3-mediated transport were therefore similar to their effects on MRP2, for which PSC833 is a more potent inhibitor than CsA (42), and distinct from their effects on MRP1, for which these agents are weak inhibitors (43).

In combination, these results showing that either amphipathic (glycocholate, E₂17βG, PAK-104P, and MK571) or lipophilic (etoposide and CsA) substrates and antagonists are potent inhibitors of MRP3-mediated MTX transport add support to the possibility inferred previously from the finding that MRP3 transports glutathione and glucuronate conjugates and glycocholate (25) that the binding pocket of MRP3 has specific sites for both lipophilic and negatively charged ligands.

Having established that the transport characteristics of MRP3 are consistent with those of the MTX efflux system, the potential participation of other MRPs was explored. Knowing that MTX efflux systems are widely distributed in tissues, the involvement of MRP1, the most ubiquitously expressed of the MRPs (44, 45), was examined. For this purpose, uptake measurements were performed on membrane vesicles prepared from MRP1-transfected NIH3T3 cells (pSRα-MRP1-32) and NIH3T3 cells transfected with parental vector (pSRα; Ref. 28). As judged by immunoblot analyses, membrane vesicles purified from pSRα-MRP1-32 cells are a rich source of the M_r 190,000 MRP1 protein (data not shown).

Regardless of the property examined, MRP1-dependent MTX transport closely resembled that mediated by MRP3. Whereas pSRα-MRP1-32 membrane vesicles catalyzed the MgATP-dependent uptake of 1.0 μm [³H]MTX at an initial rate of 1.4 pmol/mg/min, pSRα membranes mediated little or no uptake (Fig. 6,A). In strict correspondence with the results obtained with MRP3, glutamylation dramatically attenuated the amenability of MTX to transport because neither 1.0 μm [³H]MTX-Glu₂ nor 1 μm [³H]MTX-Glu₃ was transported by either pSRα-MRP1-32 or pSRα membranes (Fig. 6, B and C; note that the ordinates in Fig. 6, B and C, are >4-fold more sensitive than those in Fig. 6 A).

The substrate concentration dependence of MgATP-energized [³H]MTX uptake by membrane vesicles prepared from pSRα-MRP1-32 cells approximated Michaelis-Menten kinetics (Fig. 7 and Table 1) to yield K_m and V_max values of 2.15 ± 0.79 mm and 2.05 ± 0.91 nmol/mg/min, respectively. Membrane vesicles prepared from pSRα cells catalyzed negligible MgATP-dependent uptake at all of the concentrations examined (Fig. 7), indicating that the transport measured was largely attributable to MRP1.

MRP1, like MRP3, catalyzed transport of the physiological folates FA and leucovorin, albeit at rates (0.7 and 0.6 pmol/mg/min, respectively) markedly lower than those obtained with MTX (Fig. 8).

Total intracellular MTX levels reflect a balance between competing reactions. At the plasma membrane, influx via RFC1 is opposed by an energy-dependent efflux system. Within the cell, FPGS catalyzes the sequential addition of 1–6 glutamyl residues, whereas γ-glutamyl hydrolase serves to cleave the poly-γ-glutamate chain to regenerate the parental monoglutamate (46, 47). Of these opposing reactions, the reaction catalyzed by γ-glutamyl hydrolase is considered to contribute least to intracellular MTX pool sizes. As a result, and by virtue of the inability of the energy-dependent efflux system to export polyglutamates, polyglutamylation likely plays a determinative role in modulating overall intracellular MTX levels.

In the experiments described here, we explored the possibility that certain members of the MRP family, which had previously been implicated in MTX transport, might satisfy the requirements predicted for components of the cellular MTX efflux system. In so doing, it has been determined that the two MRPs (MRP1 and MRP3) whose wide distributions (31, 44, 48, 49, 50) coincide with that of the MTX efflux system do indeed satisfy the predicted requirements (Fig. 9). Both are high capacity, low affinity MTX transporters exhibiting little or no activity toward polyglutamates. We demonstrate that the addition of a single glutamyl residue is sufficient to abrogate transport by >95% in the case of MRP3 and to totally abrogate transport in the case of MRP1 and that the addition of further glutamyl residues augments this effect.

Several biochemical mechanisms for antifolate resistance have been determined, including increased expression of DHFR and mutations in this target enzyme that render it less susceptible to MTX binding, reduced expression or mutations in RFC1, reduced expression of FPGS, and increased expression of γ-glutamyl hydrolase (2, 51, 52, 53, 54, 55, 56, 57). Although enhanced efflux mediated by MRP1–MRP4 represents an additional MTX resistance mechanism, it remains to be determined whether this activity contributes to clinical resistance. The basic problem in understanding the relative contribution of cellular efflux to MTX resistance is the seeming strict time dependence of the phenomenon. The degree of resistance associated with the overexpression of MRPs depends on how the cells are exposed to MTX (21, 22, 23, 24). MRP1-MRP4 are highly efficacious resistance factors when drug exposure is limited to only 1–4 h of a 3-day growth assay, but they are poor resistance factors when a standard continuous exposure assay is used.

The properties of MRP3 and MRP1 reported here are capable of explaining this unusual time dependence. If, as indicated from the results of measurements of cellular free and polyglutamylated MTX levels, this resistance phenotype is attributable to enhanced polyglutamylation consequent to continuous drug exposure (22, 23), then it might in fact be expected, given that MRP3 and MRP1 are severely limited in their capacity to transport MTX polyglutamates by comparison with the parent molecule, that the resistance conferred by these transporters will decrease with increase in time of exposure. If correct, a potential clinical implication follows from this interpretation; namely, that the capacity of MRP3 and MRP1 to confer resistance to MTX may be critically dependent on the schedule of drug administration in a way distinct from the resistance conferred by these pumps to natural product agents. Specifically, MRP-mediated MTX resistance may be less likely to be a factor when this agent is administered by prolonged continuous infusion rather than by short duration i.v. infusions. We speculate that these considerations may also apply to two other MRPs, MRP2 and MRP4, which have also been demonstrated to confer a time-dependent MTX resistance phenotype (23, 24).

A second issue pertaining to whether MRP3 and MRP1 are capable of influencing clinical resistance to MTX concerns their low affinity for this antimetabolite. The K_m values we determined for MRP3 and MRP1 (0.62 and 2.15 mm, respectively) are in the range of plasma levels (0.1–1 mm) achieved after high-dose MTX infusions (e.g., 1.5 g/m²) but are significantly higher than the plasma levels (1–10 μm) associated with routinely used dosages of this agent (25–100 mg/m²; Ref. 58). Nevertheless, the possibility that the activities of MRP3 and MRP1 may be pharmacologically relevant is supported by the observation that cell lines transfected with these two transporters (as well as those transfected with MRP2 and MRP4) exhibit enhanced resistance at low micromolar MTX concentrations (22, 23, 24).

Another property of MRP3 and MRP1, their facility for the transport of physiological folates, may also have clinical implications. Because sensitivity to antifolates is not determined solely by the intracellular level of drug but also by the size of the intracellular folate pool (physiological folates compete with antifolates for binding to DHFR as well as FPGS), any factor that influences physiological folate pool size could alter drug efficacy. This being the case, enhanced expression levels of MRPs could paradoxically diminish antifolate resistance by decreasing the size of the intracellular folate pool. Reciprocally, reduced expression levels of MRPs could enhance antifolate resistance. The observation that cell lines resistant to lipid-soluble antifolates have increased folate levels (59) is at least consistent with such a scheme. However, similar to the case with MTX, the high K_m value of MRP3 for FA (1.7 mm), in contrast with the low folate concentrations in plasma (∼10–20 nm), is a consideration in understanding whether expression of the pump can influence intracellular folate homeostasis in vivo.

Analogous reasoning and kinetic considerations may apply to another MRP3 and MRP1 transport substrate, the reduced folate leucovorin (K_m = 1.74 mm for MRP3). Leucovorin is used in cancer chemotherapy in two clinical situations. It is administered as a source of reduced folates after high-dose MTX treatment to reverse the MTX-induced metabolic block in normal tissues and thereby rescue bone marrow and mucosal surfaces from the massive cytotoxicity that would otherwise occur, and it is coadministered with 5-fluorouracil to increase binding of 5-fluoro-dUMP to its target, thymidylate synthetase, and thus increase its cytotoxicity (leucovorin serum concentrations are ∼5–100 μm as a rescue agent and ∼5–10 μm as an adjuvant). In the former case, increased MRP3 and MRP1 expression might diminish the rescue effect by decreasing the net uptake of reduced folate. In the latter case, increased expression of MRPs might diminish the cytotoxicity of the chemotherapeutic regimen by decreasing the intracellular concentration of adjuvant.

The susceptibility of MRP3-mediated MTX transport to inhibition by a number of natural product agents provides an explanation for the interesting observation made nearly three decades ago that vincristine and etoposide enhance intracellular MTX levels when these drugs are administered simultaneously (60, 61). The demonstration that etoposide, which is presumed to be a substrate of MRP3 because resistance to this agent is conferred by the pump (21, 22), is a potent inhibitor of MRP3-mediated MTX transport suggests that natural product agents, which are also known substrates for MRP1 and MRP2, may impair cellular efflux of MTX by acting as competitive inhibitors (Fig. 9). This finding is of potential clinical relevance in that MTX is sometimes combined with natural product agents in chemotherapy regimens. By the same token, the determination that several MRP1 and Pgp inhibitors are potent in vitro inhibitors of MRP3-mediated transport raises the possibility that these agents, when used in clinical trials with the intention of augmenting cellular levels of natural product agents, might concomitantly augment intracellular MTX levels. Additional studies should help to determine whether these and other speculative considerations concerning the potential involvement of MRPs in clinical MTX resistance are valid.

We thank Dr. S. Akiyama (Institute for Cancer Research, Kagoshima University, Kagoshima, Japan) for generously providing AG-A and PAK-104P.