pH-Dependent Transport of Pemetrexed by Breast CancerResistance Protein

Li Li, Yuk Yin Sham, Zsolt Bikadi, and William F. Elmquist

Department of Pharmaceutics and Brain Barriers Research Center, University of Minnesota, Minneapolis, Minnesota (L.L.,W.F.E.); Center for Drug Design, Biomedical Informatics and Computational Biology Program, University of Minnesota,

Minneapolis, Minnesota (Y.Y.S.); and Virtua Drug, Ltd., Budapest, Hungary (Z.B.)

Received March 8, 2011; accepted May 31, 2011

ABSTRACT:

Breast cancer resistance protein (BCRP), an ATP-dependent effluxtransporter, confers drug resistance to many chemotherapyagents. BCRP is overexpressed in tumors exposed to an acidicenvironment; therefore, it is important to establish the effect of lowpH on BCRP transport activity. It has recently been reported thatBCRP transports substrates more efficiently in an acidic microen-vironment. In the study presented here, we examine the pH depen-dence of BCRP using methothrexate (MTX), pemetrexed (PMX),and estrone sulfate (ES) as model substrates. Our study revealedan increase of approximately 40-fold in the BCRP-mediated trans-port of PMX and MTX when the pH was decreased from 7.4 to 5.5.In contrast, only a 2-fold increase was observed for ES. Theseresults indicate a mechanism of transport that is directly depen-dent on the effective ionization state of the substrates and BCRP.For ES, which retains a constant ionization state throughout theapplied pH, the observed mild increase in activity is attributable tothe overall changes in the effective ionization state and conforma-tion of BCRP. For MTX and PMX, the marked increase in BCRPtransport activity was likely due to the change in ionization state ofMTX and PMX at lowered pH and their intermolecular interactionswith BCRP. To further rationalize the molecular basis of the pH

dependence, molecular modeling and docking studies were car-ried out using a homology model of BCRP, which has previouslybeen closely examined in structural and site-directed mutagenesisstudies (Am J Physiol Cell Physiol 299:C1100–C1109, 2010). On thebasis of docking studies, all model compounds were found toassociate with arginine 482 (Arg482) by direct salt-bridge interac-tions via their negatively charged carboxylate or sulfate groups.However, at lower pH, protonated MTX and PMX formed an addi-tional salt-bridge interaction between their positively charged moi-eties and the nearby negatively charged aspartic acid 477 (Asp477)carboxylate side chain. The formation of this “salt-bridge triad” isexpected to increase the overall electrostatic interactions betweenMTX and PMX with BCRP, which can form a rational basis for thepH dependence of the observed enhanced binding selectivity andtransport activity. Removal of Arg482 in site-directed mutagenesisstudies eliminated this pH dependence, which lends further sup-port to our binding model. These results shed light on the impor-tance of electrostatic interactions in transport activity and mayhave important implications in the design of ionizable chemother-apeutics intended for tumors in the acidic microenvironment.

Introduction

Multidrug resistance in tumors represents a significant obstacle inchemotherapeutic treatment of cancers. Drug efflux is one major drugresistance mechanism through which chemotherapeutics are actively re-moved from tumor cells. Clinically relevant efflux transporters includeP-glycoprotein, multidrug resistance-related proteins, and breast cancerresistance protein (BCRP) (Diestra et al., 2003). BCRP (ABCG2) is amembrane-bound protein and belongs to the ATP-binding cassette(ABC) family (Doyle et al., 1998; Allen and Schinkel, 2002). BCRP is

present in many solid tumors and normal tissues including placenta, blood-brain barrier, small intestine, testis, liver, adrenal gland, and stem cells (Doyleet al., 1998; Fetsch et al., 2006). BCRP confers drug resistance to manyanticancer agents including mitoxantrone, topotecan, irinotecan, tyrosine ki-nase inhibitors (imatinib, gefitinib, and erlotinib), and antifolates [methotrex-ate, tomudex, and (S)-2-[5-[(1,2-dihydro-3-methyl-1-oxobenzo[f]quinazolin-9-yl)methyl]amino-1-oxo-2-isoindolynl]-glutaric acid (GW 1843)] (Robeyet al., 2001; Volk and Schneider, 2003; Shafran et al., 2005; Bram et al.,2006). The implication of BCRP with various anticancer agents makesBCRP a clinically relevant target for studying drug resistance mechanisms.

One of the intriguing characteristics of tumors is the acidic mi-croenvironment (i.e., an extracellular pH as low as 5.8), comparedwith normal tissues (Tannock and Rotin, 1989; Ojugo et al., 1999).The causes for the acidic pH environment in tumors are not well

This work was supported in part by Eli Lilly and Company; and the Center forDrug Design at the University of Minnesota.

Article, publication date, and citation information can be found athttp://dmd.aspetjournals.org.

doi:10.1124/dmd.111.039370.

ABBREVIATIONS: BCRP, breast cancer resistance protein; ABC, ATP-binding cassette; ABCG2, ATP-binding cassette subfamily G member 2;PMX, pemetrexed; MTX, methotrexate; ES, estrone sulfate; GF120918, N-(4-[2-(1,2,3,4-tetrahydro-6,7-dimethoxy-2-isoquinolinyl)ethyl]-phenyl)-9,10-dihydro-5-methoxy-9-oxo-4-acridine carboxamide; Ko143, 3-(6-isobutyl-9-methoxy-1,4-dioxo-1,2,3,4,6,7,12,12a-octahydropyrazino[1�,2�:1,6]pyrido[3,4-b]indol-3-yl)-propionic acid tert-butyl ester; TM, transmembrane helical; GW1843, (S)-2-[5-[(1,2-dihydro-3-methyl-1-oxobenzo-[f]quinazolin-9-yl)methyl]amino-1-oxo-2-isoindolynl]-glutaric acid.

0090-9556/11/3909-1478–1485$25.00DRUG METABOLISM AND DISPOSITION Vol. 39, No. 9Copyright © 2011 by The American Society for Pharmacology and Experimental Therapeutics 39370/3707414DMD 39:1478–1485, 2011 Printed in U.S.A.

1478

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from

understood. It has been linked to the increased use of the glycolyticpathways and the compromised vasculature of tumors, which maylead to the poor removal of lactic acid (Vaupel, 2004). On the basis ofthe observation that the activities of many transporters are affected bythe environmental pH, it is likely that BCRP acts differently underacidic pH, which further affects its drug resistance profile.

Breedveld et al. (2007) recently demonstrated that BCRP transportssubstrates more efficiently under acidic pH using intact cell-based andmembrane vesicle-based assays. In the study presented here, weexplored the possible mechanisms for the pH-dependent BCRP trans-port using methothexate (MTX), pemetrexed (PMX), and estronesulfate (ES) as model compounds (Fig. 1). In addition, the molecularbasis of the observed pH dependence was examined using dockingstudies with the recent inward-facing conformation homology modelof BCRP (Rosenberg et al., 2010).

Materials and Methods

Chemicals. [3H]Pemetrexed, [3H]methotrexate, and [3H]estrone sulfatewere obtained from Moravek Biochemicals (Brea, CA). GF120918 (N-[4-[2-(6,7-dimethoxy-3,4-dihydro-1H-isoquinolin-2-yl)ethyl]phenyl]-5-methoxy-9-oxo-10H-acridine-4-carboxamide) was a gift from GlaxoSmithKline (ResearchTriangle, NC). 3-(6-Isobutyl-9-methoxy-1,4-dioxo-1,2,3,4,6,7,12,12a-octahydropyrazino[1�,2�:1,6]pyrido[3,4-b]indol-3-yl)-propionic acid tert-butylester [(Ko143) a fumitremorgin C analog] was kindly provided by Dr. AlfredSchinkel (The Netherlands Cancer Institute, Amsterdam, The Netherlands).GF120918 and Ko143 were dissolved in dimethyl sulfoxide and diluted to thedesired concentration with assay buffer (25 mM NaHCO3, 122 mM NaCl, 10mM glucose, 10 mM HEPES, 1.2 mM MgSO4, 3 mM KCl, 1.4 mM CaCl2, and0.4 mM K2HPO4, pH 7.4). The final concentration of dimethyl sulfoxide in all

reaction solutions was less than 0.1%. All other chemicals used were ofhigh-pressure liquid chromatography or reagent grade.

Cell Lines. HEK293 cells transfected with ABCG2-Arg482, ABCG2-R482G, and ABCG2-R482T and plasmid vectors were obtained from Dr.Susan E. Bates (National Cancer Institute, Bethesda, MD). Stable transfectantswere maintained in Dulbecco’s modified Eagle’s medium (Mediatech, Inc.,Herndon, VA) fortified with 10% heat-deactivated fetal bovine serum (Sera-Care Life Sciences, Inc., Oceanside, CA), 100 U/ml penicillin, and 100 �g/mlstreptomycin (Sigma-Aldrich, St. Louis, MO) at 37°C under humidity and 5%CO2 tension. The medium for ABCG2-transfected cells was supplementedwith geneticin (G418) at a concentration of 2 mg/ml to maintain positiveselection pressure for BCRP expression. Comparable protein expression in thewild-type (Arg482) and mutant (R482G, R482T) ABCG2-transfected cells hasbeen demonstrated by the Bates group (Robey et al., 2003).

Membrane Vesicle Preparation. Vesicle preparation was adapted fromTabas and Dantzig (2002). Vector control or BCRP-transfected HEK293 cellswere cultured in 500-cm2 plates (Corning Life Sciences, Lowell, MA). When75% confluence was reached, the cells were washed twice with phosphate-buffered saline and scraped into a residue of phosphate-buffered saline. Aftercentrifugation at 1700 rpm at 4°C for 5 min, the cell pellet was diluted with ahypotonic buffer containing 1 mM sodium bicarbonate at pH 9 and 1%EDTA-free protease inhibitors. The cells were allowed to sit and swell on icefor 5 min, followed by vigorous shaking 20 times. The resultant cell lysate wascentrifuged at 1700 rpm at 4°C for 5 min to remove nuclei, mitochondria, andwhole cells. The supernatant (crude membrane fraction) was layered over 40%(w/v) sucrose solution and centrifuged at 25,000 rpm at 4°C for 30 min. Theturbid layer at the interface was collected and suspended in Tris-sucrose buffer(250 mM sucrose containing 50 mM Tris/HCl, pH 7.4) and centrifuged at25,000 rpm for 40 min. The membrane fraction was collected and resuspendedin a small volume (150–250 �l) of Tris-sucrose buffer. The membrane vesicleswere made by slowly passing the suspension through a 27-gauge needle 20times. The vesicle containing solution was aliquoted into 100-�l fractions andfrozen at �80°C. The amount of membrane vesicles was quantified by mea-suring membrane protein using the BCA protein assay (Thermo Fisher Scien-tific, Waltham, MA). Our procedure for the preparation of inverted membranevesicles is similar to that of Volk and Schneider (2003), and the percentage ofinside-out vesicles was reported to be on average 70% (range �50–96%).

Vesicular Uptake Study. The vesicular uptake studies were performedusing the rapid filtration method according to Pratt et al. (2006) and Tabas andDantzig (2002) with minor modifications. The uptake was started by additionof 20 �g of inside-out membrane vesicles (BCRP-expressing vesicles orpc-DNA vector control vesicles) to 20 �l of Tris/HCl sucrose buffer containing[3H]PMX, 20 mM MgCl2, 4 mM ATP, 10 mM phosphocreatine, and 100�g/ml creatine kinase. Reactions were carried out at 37°C for 60 min and werestopped by the addition of 200 �l of ice-cold Tris/HCl sucrose buffer (pH 7.4).Vesicles were separated from free drug by passing through 0.22-�m Duraporemembrane filters with a 96-channel cell harvester (PerkinElmer Life andAnalytical Sciences, Waltham, MA). The filters were washed three times withice-cold Tris/HCl sucrose buffer and dried at room temperature overnight.Radioactivity was measured by a liquid scintillation counter (TopCount; Lifeand Analytical Sciences).

Inhibitor Assays. PMX uptake into BCRP or control vesicles was mea-sured in the presence or absence of BCRP inhibitors at maximal inhibitoryconcentrations (200 nM Ko143).

Kinetic Studies. PMX uptake into BCRP or control vesicles was measuredafter a 15-min exposure to 0, 100, 250, 500, 1000, 2000, and 4000 �M PMX.The same batch of membrane vesicles was used in all experiments to reducevariability, and the vesicular transport assay buffer was adjusted to pH 5.5 or6.5. Kinetic parameters (Km and Vmax) were determined by nonlinear regres-sion analysis using SigmaPlot, version 9.0.1 (Systat Software, Inc., SanJose, CA).

Calculation. ATP-dependent uptake was calculated by subtracting uptakemeasured in the absence of ATP from uptake measured in the presence of ATP.BCRP-mediated transport was calculated by subtracting ATP-dependent up-take into vector control vesicles from the ATP-dependent uptake into BCRPoverexpressing vesicles.

Molecular Modeling. All molecular modeling studies were carried outusing the Schrodinger modeling suite package (Schrodinger, LLC, New York,FIG. 1. Structures of PMX, MTX, and ES at pH 7.0.

1479pH-DEPENDENT TRANSPORT OF PEMETREXED BY BCRP

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from

NY). The homology model of BCRP in the inward-facing closed conformationwas previously reported (Rosenberg et al., 2010) and was based on the recentlysolved X-ray structure of mouse P-glycoprotein (Protein Data Bank code3G5U) (Aller et al., 2009). This model has been previously shown to beconsistent with experimentally observed conformation change upon substratebinding (Rosenberg et al., 2010) and for the identification of important resi-dues for drug transport (Ni et al., 2010). The homology-modeled BCRPstructure was refined by restraint energy minimization using the OPLS 2005force field (Jorgensen et al., 1996) with the generalized born solvent accessibleimplicit solvent model (Still et al., 1990). Docking of all compounds wascarried out using Glide at Standard Precision (Friesner et al., 2004). For eachligand, the top-ranked docking geometry based on G-score scoring function isreported. The electrostatic potential surfaces of MTX and PMX were evaluatedbased on the protonated and deprotonated forms of both compounds at pH 5.5and physiological pH. The pKa values of PMX, MTX, and ES were calculatedby ACD pKa predictor software (Advanced Chemistry Development, Inc.,Toronto, Canada; http://www.acdlabs.com/products/phys_chem_lab/pka/), areliable knowledge-based pKa calculator using Hammett equations derivedfrom a database of over 30,000 experimental pKa values (Balogh et al., 2009).

Statistical Analysis. Statistical comparisons between the two groups weremade using the two-sample t test. Differences were considered to be statisti-cally significant when P � 0.05. Multiple groups were compared usinganalysis of variance with Holm–Sidak post hoc test at P � 0.05 level ofsignificance.

Results

Effect of pH on ATP-Dependent BCRP Transport Activity ofPMX, MTX, and ES. At pH 7.4, PMX uptake in BCRP membranevesicles was 2-fold higher than that of control vesicles (P � 0.05). Incontrast, PMX uptake in BCRP vesicles was approximately 20-foldhigher than that in the control vesicles at pH 5.5 (Fig. 2A). Bysubtracting the uptake in BCRP vesicles from that in control vesicles,BCRP-mediated transport of PMX increased by 43-fold at pH 5.5compared withthe physiological pH. MTX was used here as a positivecontrol for method validation. Consistent with a previous report(Breedveld et al., 2007), a marked increase in BCRP transport ofMTX was observed when the pH decreased from 7.4 to 5.5 (Fig. 2B).The observed pH-sensitive BCRP-mediated transport may be a con-sequence of changed intrinsic protein activity at different pH and/oraltered ionization state of the substrate molecules and thus differentsubstrate-transporter protein interaction. To elucidate the mechanismof BCRP pH dependence, we investigated the BCRP-mediated trans-port of estrone sulfate, a compound with constant ionization status atthe applied pH range (Fig. 3). As the pH decreased from 7.4 to 5.5, weobserved a 1.7-fold increase in the BCRP uptake of estrone sulfate, afinding that was quite subtle compared with the 40-fold increaseobserved for PMX and MTX. No optimal pH was observed formaximal transport activity for BCRP.

Effect of BCRP Inhibitor on PMX and MTX Uptake intoBCRP-Expressing and Control Vesicles at Low pH. To confirmthat the markedly increased PMX and MTX transport at low pH wasa BCRP-mediated process, we determined the effect of BCRP-specificinhibitor on the pH-dependent BCRP transport of PMX and MTX. Asshown in Fig. 4, 200 nM Ko143 significantly decreased the uptake ofPMX and MTX in BCRP-expressing membrane vesicles at pH 5.5,suggesting that the enhanced uptake at the low pH was mainly a resultof stimulated transport by BCRP.

pH-Dependent BCRP Transport Kinetics of PMX. To charac-terize the influence of pH on the BCRP-mediated transport kinetics ofPMX, concentration-dependent uptake studies were performed at pH5.2 and 6.5. From the time course of PMX uptake at pH 5.5, a 15-mintime point was within the linear uptake range and was used todetermine the concentration-dependent uptake of PMX (data notshown). As shown in Fig. 5, the BCRP-mediated uptake of PMX was

saturable and followed Michaelis–Menten kinetics, with Km of 0.39 �0.08 mM and Vmax of 4725 � 279 pmol � mg�1 � min�1 at pH 5.2 anda Km of 1.47 � 0.35 mM and Vmax of 1993 � 213 � mg�1 � min�1 atpH 6.5.

pKa and Ionization State of BCRP Transport Substrates. ThepKa values of PMX, MTX, and ES were evaluated and are shown inFig. 6. The predicted pKa values of MTX were within 0.5 pKa units of itsexperimentally determined pKa values of 3.2, 4.5, and 5.6 (Szakacs andNoszal, 2006). The predicted protonated site for the base was on the N5atom for PMX and on the N5 atom for MTX. The predicted pKa of ESwas 2.0 for the sulfate group. On the basis of the predicted pKa values, itis expected that the acidic group of all three model compounds will bepermanently charged at the applied experimental pH of 7.4 and 5.5.However, at the acidic pH of 5.5, the basic group of PMX and MTX wasexpected to undergo protonation of its pteridine ring.

FIG. 2. Effect of pH on ATP-dependent uptake of [3H]PMX (A) and [3H]MTX (B)into BCRP-expressing vesicles and control vesicles. HEK293-BCRP membranevesicles and pc-DNA membrane vesicles were incubated with 0.3 �M [3H]PMX or[3H]MTX at pH 5.5 at 37°C for 60 min in the presence or absence of 4 mM ATP.Values shown are mean � S.D. of each experiment (n � 4) (***, P � 0.001compared with the BCRP uptake at pH 7.4; *, P � 0.05 compared with uptake intocontrol vesicles at pH 7.4).

1480 LI ET AL.

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from

BCRP Arg482 Binding Site. The amino acid residue 482 is areported hot spot for BCRP substrate specificity, and it plays a crucialrole in the transport of MTX (Honjo et al., 2001; Allen et al., 2002).Studies from our laboratory showed that Arg482 was also importantfor the recognition of PMX and estrone sulfate. To understand themolecular basis for the increased binding selectivity of PMX andMTX at low pH, the BCRP binding site consisting of Arg482 wasclosely examined for possible key residues that can provide signifi-cant interaction with the charged carboxylate and pteridinium groupsof PMX and MTX in the recently published homology model.

BCRP is a homodimeric transport membrane protein consisting ofan extracellular nucleotide binding domain and a transmembranehelical (TM) domain in each of its monomeric subunits. The TMdomains (Wang et al., 2008) consist of six transmembrane heliceswith Arg482 located in the helix III region along the transport channelof BCRP (Fig. 7A). In the model, residues Asp477, Arg465, andHis630 were calculated to be the only ionizable residues within an18-Å radius of Arg482 located within the transport channel of BCRPthat were capable of interacting simultaneously with the chargedmoiety (Fig. 7C). The nearest interatomic distance between theArg482 and Arg465 guanidinium groups was 6.9 Å, which is similarto the interatomic distance of 6.0 Å of the carboxylate atoms in theenergy-minimized conformation of PMX (Fig. 7B). His630, whichwas located in the helix VI region further inside along the BCRPtransport channel, has the nearest interatomic side-chain distances toArg482 and Arg465 of 6.0 and 9.5 Å, respectively. Asp477, whichwas located in the transmembrane helix III region at the interfacebetween the TM and nucleotide binding domains, has the nearestinteratomic side-chain distances to Arg482 and Arg465 of 6.9 and 8.7Å, respectively. The nearest interatomic side-chain distances betweenthe arginines to His630 and Asp477 were found to be similar to theinteratomic distances of pteridine and carboxylate groups in the en-ergy-minimized structure of PMX. The nearest interatomic side-chaindistance of the guanidinium group between the two Arg482s on each

of the monomers was 11.2 Å. With the interatomic distance of 6.0 Åbetween the dicarboxylic acids in MTX and PMX, the homologymodel also suggests a possible mode of binding involving the directinteraction between each of the negatively charged carboxylate groupsin PMX and MTX with the positively charged guanidinium group ofArg482 in each of the monomeric units.

Potential Mode of Binding. Docking studies were carried outwithin the Arg482 binding site to identify the potential mode ofbinding. Under acidic pH, His630 is expected to become positivelyionized. Our docking study did not identify any reasonable mode ofbinding involving His630 and Arg482. At physiological pH whensubstrate and His630 are neutralized, the mode of binding involvingAsp477 was consistently ranked higher than His630. As a result,His630’s involvement was excluded as a potential contributor towardthe observed pH-dependent activity of BCRP. With the defined con-straints of Arg482 and Asp477 side chains as potential hydrogen bond

FIG. 4. Effect of BCRP-specific inhibitor Ko143 (200 nM) on ATP-dependentuptake of [3H]PMX (A) and [3H]MTX (B) at pH 5.5. HEK293-BCRP membranevesicles and pc-DNA membrane vesicles were incubated with 0.11 �M [3H]PMXand 0.02 �M [3H]MTX at pH 5.5 at 37°C for 60 min in the presence or absence of4 mM ATP. Values shown are mean � S.D. of each experiment (n � 4) (***, P �0.001 compared with the BCRP uptake without Ko143).

FIG. 3. Effect of pH on BCRP-mediated uptake of [3H]ES. HEK293-BCRP mem-brane vesicles and pc-DNA membrane vesicles were incubated with 0.03 �M[3H]ES at pH 5.5, 6.0, 6.5, and 7.0 at 37°C for 10 min in the presence or absenceof 4 mM ATP. The BCRP-mediated uptake was calculated by subtracting ATP-dependent uptake into vector control vesicles from the ATP-dependent uptake intoBCRP-overexpressing vesicles. Values shown are mean � S.D. of each experiment(n � 4) (*, P � 0.05 compared with the BCRP uptake at pH 7).

1481pH-DEPENDENT TRANSPORT OF PEMETREXED BY BCRP

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from

donor and acceptor, two common modes of binding were observed forMTX and PMX (Fig. 8). With the farthest interatomic distance of 6 Åbetween the oxygen atoms of the dicarboxylic acid group, each of thecarboxylate groups of PMX and MTX has the potential to form directsalt bridges with the guanidinium side chains of Arg482 and Arg465within each of the monomeric subunits or with the guanidinium sidechains of Arg482 between the two monomeric units. With the dicar-boxylic acid anchored in place, the positively charged base of PMXand MTX at low pH was found to be within the distal constraint toform a third salt bridge with the negatively charged side chain of

Asp477. On the basis of the calculated pKa, ES was expected to retainits negatively charged state at the applied pH. The only mode ofbinding observed for ES involved a single salt-bridge interactionbetween the negatively charged sulfate and the positively chargedguanidinium group of Arg482.

The Role of Arg482 in pH-Dependent BCRP Transport. Theresults from docking studies indicate that the formation of a “salt-bridge triad” between PMX/MTX molecules and BCRP at amino acidresidues Arg482, Arg465, and Asp477 may contribute to the greatlyenhanced transport activity of BCRP for PMX and MTX at low pH.To confirm the role of the salt-bridge triad in the observed pH-dependent BCRP transport, we determined the influence of polymor-phic BCRP mutation at 482 on the pH-dependent transport of PMX,MTX, and estrone sulfate. As shown in Fig. 9A, the extent of pHdependence, calculated as the ratio of uptake at pH 5.5 to the uptakeat pH 7.4, was approximately 61, 12, and 6 for Arg482 (wild-typeBCRP), R482G (mutant BCRP), and R482T (mutant BCRP), respec-tively. A similar trend was observed for MTX; that is, pH-dependenttransport was significantly disturbed when positively charged arginineat 482 (Arg482) was replaced by nonionized amino acids (R482G andR482T) at lower pH (Fig. 9B). In contrast, the transport of estronesulfate by wild-type and mutant BCRP were similarly enhanced(1.5-fold) when the pH changed from 7.4 to 5.5 (Fig. 10).

Discussion

It is well known that pH plays an important role in modulating thefunctional activity of proteins. This phenomenon also applies to manytransporters. Studies by Nozawa et al. (2004) have shown that humanorganic anion transporting polypeptide-B (SLC21A9) exhibits pH-sensitive transport activity for various organic anions. Likewise, hu-man oligopeptide transporter 1 exhibits a bell-shaped transport activ-ity with an optimal pH of 5.5 (Fujisawa et al., 2006). pH-dependenttransport has also been reported for proton-coupled folate transporter(SLC46A1) and organic cation transporters 1 (SLC22A1) and 2(SLC22A2) (Urakami et al., 1998; Fujita et al., 2006; Zhao andGoldman, 2007). Various mechanisms are responsible for the pH-sensitive transport activity, including proton-coupled transport, mem-brane potential-dependent transport, and the pH-dependent ionizationof key amino acid residues at active binding sites. In particular, it has

FIG. 5. Concentration-dependent study of PMX. BCRP-expressing vesicles andcontrol vesicles were exposed for 15 min to various concentrations of [3H]PMX atpH 5.2 (F) and 6.5 (E), ranging from 0 to 4 mM, in the presence or absence of ATP.Values shown are mean � S.D. of each experiment (n � 4). The solid linerepresents a best-fit Michaelis–Menten plot of the net initial velocity relative toincreasing substrate concentrations.

FIG. 6. Electrostatic potential surface of PMX, MTX, and ES at pH 7.4 and 5.5 onthe basis of the predicted pKa values of each ionization functional group (yellow).

FIG. 7. A, homology model of BCRP with its transmembrane helical domainsembedded within the phospholipid bilayers (illustrated by the yellow dotted lines).B, an energy-minimized folded conformation of PMX. C, BCRP molecular recog-nition site involving Arg482. The ionization residues within 18 Å of Arg482 locatedwithin the transport channel are shown.

1482 LI ET AL.

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from

been documented that histidine plays an important role in pH-sensi-tive transporters such as proton-coupled folate transporter and peptidetransporter 1 because histidine can interact differently with substratemolecules depending on its protonation state at various pH values(Said and Mohammadkhani, 1993; Fei et al., 1997; Metzner et al.,2008; Unal et al., 2009).

In the study presented here, we examined the influence of pH onBCRP transport activity using BCRP substrates PMX and estronesulfate and characterized the possible mechanism for the observed pHdependence. Vesicular uptake assays revealed that BCRP-mediatedtransport of PMX was 43-fold higher at acidic pH (5.5) compared withthe transport at physiological pH (7.4). In addition, the increaseduptake at lower pH associated with an increased affinity of PMX forBCRP (as indicated by decreased Km values) and a markedly in-creased BCRP transport capacity (Vmax). The pH-dependent BCRP

transport was also observed for estrone sulfate. However, extent of pHdependence was mild, with a pH alteration from 7.4 to 5.5 onlyleading to a 2-fold increase in BCRP-mediated transport of estronesulfate. This result is consistent with an observation from Breedveldet al. (2007) that BCRP transports substrate drugs more efficiently atlow pH. However, our data further suggest that the extent of pH-dependent transport activity can vary significantly depending on theionization status of the substrates.

With a pKa of 2, the ionization state of estrone sulfate is constantthrough the applied pH (pH 7.4 and 5.5) (Fig. 6). Therefore, anychanges in estrone sulfate uptake from pH 7.4 to 5.5 should be a resultof the pH-induced conformational and/or ionization changes associ-ated with the transporter protein. The mild pH effect observed for

FIG. 8. The two common modes of binding observed are shown in A and B withPMX binding to BCRP. The models suggest the importance of Asp477 in theformation of the third salt bridge with the protonated aromatic base of PMX andMTX at low pH. The binding of ES to BCRP is shown in C involving a single saltbridge.

FIG. 9. ATP-dependent transport of [3H]PMX (A) and [3H]MTX (B) into BCRP-expressing vesicles (wild-type BCRP: ABCG2-R and two mutant BCRP: ABCG2-Gand ABCG2-T) and control vesicles. HEK293-BCRP membrane vesicles and pc-DNA membrane vesicles were incubated with [3H]PMX or [3H]MTX at pH 7.4 and5.5 at 37°C for 60 min in the presence or absence of 4 mM ATP. Values shown aremean � S.D. of each experiment (n � 4) (***, P � 0.001, compared with theBCRP-mediated uptake at pH 7.4).

1483pH-DEPENDENT TRANSPORT OF PEMETREXED BY BCRP

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from

estrone sulfate indicates that the intrinsic activity of BCRP increasedat lower pH, but only to a moderate extent. In contrast, the significantpH-dependent transport observed for MTX and PMX may be mainlydue to the change in the ionization state of the substrate and thesubsequent interaction with BCRP. PMX is a polyelectrolyte carryingtwo carboxyl groups, with pKa of 3.46 (�-carboxyl) and 4.77 (�-carboxyl), and the guanidinic N-1 on the pterine ring (pKa 5.27)(predicted using ACD software) (Fig. 6). At physiological pH, PMXexists in a predominantly negatively charged form because of thedeprotonation of the two carboxyl groups. When the pH decreasedfrom 7.4 to 5.5, there is a negligible change in the ionization ofcarboxyl groups (i.e., from 99 to 97%) but a marked increase in thecalculated ionization of the nitrogen group (i.e., from 0 to 30%) inwater. It is possible that the enhanced transport of PMX at low pH isdue to the increased positive charge in the PMX molecule, which inturn leads to the stronger electrostatic interaction with BCRP.

To understand molecular basis for the increased BCRP transport ofPMX and MTX at low pH, molecular modeling and docking studieswere carried out. On the basis of the homology model of BCRP,residues Asp477, Arg465, and His630 were identified to be the onlyionizable residues within an 18-Å radius of Arg482 located within thetransport channel of BCRP. In the presence of two positively chargedArg482 and Arg465 residues, the pKa values of Asp477 and His630were expected to be significantly perturbed and will remain predom-inantly negatively and neutrally charged, respectively, at the appliedpH of 7.4 and 5.5. This will result in an overall hydrophilic environ-ment for the recognition of ionizable substrates. With the presence ofthe doubly negatively charged carboxylate groups and a positivelycharged aromatic group, electrostatic interactions were expected toplay a crucial role in the molecular recognition of PMX and MTXwithin the binding site. The decrease in Km for PMX from 1.47 to 0.39mM with pH changing from 6.5 to 5 further supports an overallincrease in binding selectivity as PMX undergoes protonation at lowerpH. The two observed modes of binding along the BCRP transport

channel suggested two potential molecular recognition mechanisms ofthe doubly negatively charged MTX and PMX substrates that werenot observed in ES. Although Asp477 and His630 locate at similardistance to that of the folded PMX conformation, the docking studyshowed consistently that the mode of binding favors the formation ofa third salt bridge between the protonated base of PMX and MTX withthe negatively charged side chain of Asp477, resulting in a salt-bridgetriad for molecular recognition. The formation of the third salt bridgeis expected to increase the electrostatic interaction between PMX/MTX with BCRP, which would lead to an enhancement in the bindingselectivity. This was confirmed by site-directed mutagenesis study inwhich Arg482 was replaced with nonionized amino acid residuals toabolish the salt-bridge triad. The results showed that the enhancementin PMX and MTX transport at acidic pH was much less in mutantBCRP compared with that in the wild-type BCRP, suggesting theimportance of Arg482 in the pH-dependent transport of PMX. Forestrone sulfate, mutant (R482G, R482T) and wild-type (Arg482)BCRP exhibited a similar extent of increase in the transport activity,suggesting that Arg482 was not involved in the pH-dependent trans-port of estrone sulfate. This is reasonable because the intrinsic pKa ofarginine is approximately 12. Therefore, with a constant ionizationstate throughout the applied pH, Arg482 should not produce anychange in the electrostatic properties of BCRP and subsequent inter-action with estrone sulfate.

In summary, the BCRP-mediated transport of PMX was markedlyenhanced under acidic pH conditions. For estrone sulfate, a mildincrease in the BCRP transport was observed. On the basis of dockingstudies with the BCRP homology model, at low pH, PMX formed anadditional salt-bridge interaction between their positively chargedmoieties and the negatively charged Asp477 carboxylate side chain.The formation of a salt-bridge triad is expected to increase theelectrostatic interaction of PMX with BCRP and lead to a dramaticenhancement in the binding selectivity at acidic pH. The proposedmodel provided insights into the charge interactions between thesubstrate and the transporter proteins and furthermore demonstratedthe importance of electrostatic interactions in the BCRP transportprocess. This study may have important implications in the handlingof PMX and other relevant chemotherapeutic drugs that are exposedto the acidic environment of tumors.

Acknowledgments

We thank Rob Robey and Susan E. Bates (National Cancer Institute,National Institutes of Health) for kindly providing HEK293 cells, Dr. Alfred H.Schinkel (Netherlands Cancer Institute) for generously providing Kol43, andGlaxoSmithKline for their gift of GF120918. We also appreciate the help fromthe scientists at Eli Lilly (Victor Chen in particular) for the tremendous supportand scientific input on this project and Kate Hillgren’s laboratory (YoungeenAnne Pak) and Anne Dantzig’s laboratory (Susan Pratt) for technical assis-tance on membrane vesicle studies. We thank Tate Winter for his input on themanuscript.

Authorship Contributions

Participated in research design: Li, Sham, and Elmquist.Conducted experiments: Li.Contributed new reagents or analytic tools: Sham and Bikadi.Performed data analysis: Li, Sham, and Elmquist.Wrote or contributed to the writing of the manuscript: Li, Sham, and

Elmquist.Other: Sham conducted computational studies.

References

Allen JD, Jackson SC, and Schinkel AH (2002) A mutation hot spot in the Bcrp1 (Abcg2)multidrug transporter in mouse cell lines selected for Doxorubicin resistance. Cancer Res62:2294–2299.

FIG. 10. ATP-dependent [3H]ES accumulation into BCRP-expressing vesicles(wild-type BCRP: ABCG2-R and two BCRP mutants: ABCG2-G and ABCG2-T)and control vesicles. HEK293-BCRP membrane vesicles and pc-DNA membranevesicles were incubated with 0.03 �M [3H]ES at pH 7.4 and 5.2 at 37°C for 10 minin the presence or absence of 4 mM ATP. Values shown are mean � S.D. of eachexperiment (n � 4) (*, P � 0.05, **, P � 0.01, ***, P � 0.001 compared with theBCRP-mediated uptake at pH 7.4).

1484 LI ET AL.

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from

Allen JD and Schinkel AH (2002) Multidrug resistance and pharmacological protection mediatedby the breast cancer resistance protein (BCRP/ABCG2). Mol Cancer Ther 1:427–434.

Aller SG, Yu J, Ward A, Weng Y, Chittaboina S, Zhuo R, Harrell PM, Trinh YT, Zhang Q,Urbatsch IL, et al. (2009) Structure of P-glycoprotein reveals a molecular basis for poly-specific drug binding. Science 323:1718–1722.

Balogh GT, Gyarmati B, Nagy B, Molnar L, and Keseru GM (2009) Comparative evaluation ofin silico pKa prediction tools on the gold standard dataset. QSAR Combin Sci 28:1148–1155.

Bram E, Ifergan I, Shafran A, Berman B, Jansen G, and Assaraf YG (2006) Mutant Gly482 andThr482 ABCG2 mediate high-level resistance to lipophilic antifolates. Cancer ChemotherPharmacol 58:826–834.

Breedveld P, Pluim D, Cipriani G, Dahlhaus F, van Eijndhoven MA, de Wolf CJ, Kuil A, BeijnenJH, Scheffer GL, Jansen G, et al. (2007) The effect of low pH on breast cancer resistanceprotein (ABCG2)-mediated transport of methotrexate, 7-hydroxymethotrexate, methotrexatediglutamate, folic acid, mitoxantrone, topotecan, and resveratrol in in vitro drug transportmodels. Mol Pharmacol 71:240–249.

Diestra JE, Condom E, Del Muro XG, Scheffer GL, Perez J, Zurita AJ, Munoz-Seguí J, ViguesF, Scheper RJ, Capella G, et al. (2003) Expression of multidrug resistance proteins P-glyco-protein, multidrug resistance protein 1, breast cancer resistance protein and lung resistancerelated protein in locally advanced bladder cancer treated with neoadjuvant chemotherapy:biological and clinical implications. J Urol 170:1383–1387.

Doyle LA, Yang W, Abruzzo LV, Krogmann T, Gao Y, Rishi AK, and Ross DD (1998) Amultidrug resistance transporter from human MCF-7 breast cancer cells. Proc Natl Acad SciUSA 95:15665–15670.

Fei YJ, Liu W, Prasad PD, Kekuda R, Oblak TG, Ganapathy V, and Leibach FH (1997)Identification of the histidyl residue obligatory for the catalytic activity of the humanH�/peptide cotransporters PEPT1 and PEPT2. Biochemistry 36:452–460.

Fetsch PA, Abati A, Litman T, Morisaki K, Honjo Y, Mittal K, and Bates SE (2006) Localizationof the ABCG2 mitoxantrone resistance-associated protein in normal tissues. Cancer Lett235:84–92.

Friesner RA, Banks JL, Murphy RB, Halgren TA, Klicic JJ, Mainz DT, Repasky MP, Knoll EH,Shelley M, Perry JK, et al. (2004) Glide: a new approach for rapid, accurate docking andscoring. 1. Method and assessment of docking accuracy. J Med Chem 47:1739–1749.

Fujisawa Y, Tateoka R, Nara T, Kamo N, Taira T, and Miyauchi S (2006) The extracellular pHdependency of transport activity by human oligopeptide transporter 1 (hPEPT1) expressedstably in Chinese hamster ovary (CHO) cells: a reason for the bell-shaped activity versus pH.Biol Pharm Bull 29:997–1005.

Fujita T, Urban TJ, Leabman MK, Fujita K, and Giacomini KM (2006) Transport of drugs in thekidney by the human organic cation transporter, OCT2 and its genetic variants. J Pharm Sci95:25–36.

Honjo Y, Hrycyna CA, Yan QW, Medina-Perez WY, Robey RW, van de Laar A, Litman T, DeanM, and Bates SE (2001) Acquired mutations in the MXR/BCRP/ABCP gene alter substratespecificity in MXR/BCRP/ABCP-overexpressing cells. Cancer Res 61:6635–6639.

Jorgensen WL, Maxwell DS, and Tirado-Rives J (1996) Development and testing of the OPLSall-atom force field on conformational energetics and properties of organic liquids. J Am ChemSoc 118:11225–11236.

Metzner L, Natho K, Zebisch K, Dorn M, Bosse-Doenecke E, Ganapathy V, and Brandsch M(2008) Mutational analysis of histidine residues in the human proton-coupled amino acidtransporter PAT1. Biochim Biophys Acta 1778:1042–1050.

Ni Z, Bikadi Z, Cai X, Rosenberg MF, and Mao Q (2010) Transmembrane helices 1 and 6 of thehuman breast cancer resistance protein (BCRP/ABCG2): identification of polar residuesimportant for drug transport. Am J Physiol Cell Physiol 299:C1100–C1109.

Nozawa T, Imai K, Nezu J, Tsuji A, and Tamai I (2004) Functional characterization ofpH-sensitive organic anion transporting polypeptide OATP-B in human. J Pharmacol ExpTher 308:438–445.

Ojugo AS, McSheehy PM, McIntyre DJ, McCoy C, Stubbs M, Leach MO, Judson IR, and

Griffiths JR (1999) Measurement of the extracellular pH of solid tumours in mice by magneticresonance spectroscopy: a comparison of exogenous (19)F and (31)P probes. NMR Biomed12:495–504.

Pratt S, Chen V, Perry WI 3rd, Starling JJ, and Dantzig AH (2006) Kinetic validation of the useof carboxydichlorofluorescein as a drug surrogate for MRP5-mediated transport. Eur J PharmSci 27:524–532.

Robey RW, Honjo Y, Morisaki K, Nadjem TA, Runge S, Risbood M, Poruchynsky MS, andBates SE (2003) Mutations at amino-acid 482 in the ABCG2 gene affect substrate andantagonist specificity. Br J Cancer 89:1971–1978.

Robey RW, Medina-Perez WY, Nishiyama K, Lahusen T, Miyake K, Litman T, SenderowiczAM, Ross DD, and Bates SE (2001) Overexpression of the ATP-binding cassette half-transporter, ABCG2 (Mxr/BCrp/ABCP1), in flavopiridol-resistant human breast cancer cells.Clin Cancer Res 7:145–152.

Rosenberg MF, Bikadi Z, Chan J, Liu X, Ni Z, Cai X, Ford RC, and Mao Q (2010) The humanbreast cancer resistance protein (BCRP/ABCG2) shows conformational changes with mitox-antrone. Structure 18:482–493.

Said HM and Mohammadkhani R (1993) Folate transport in intestinal brush border membrane:involvement of essential histidine residue(s). Biochem J 290:237–240.

Shafran A, Ifergan I, Bram E, Jansen G, Kathmann I, Peters GJ, Robey RW, Bates SE, andAssaraf YG (2005) ABCG2 harboring the Gly482 mutation confers high-level resistance tovarious hydrophilic antifolates. Cancer Res 65:8414–8422.

Still WC, Tempczyk A, Hawley RC, and Hendrickson T (1990) Semianalytical treatment ofsolvation for molecular mechanics and dynamics. J Am Chem Soc 112:6127–6129.

Szakacs Z and Noszal B (2006) Determination of dissociation constants of folic acid, metho-trexate, and other photolabile pteridines by pressure-assisted capillary electrophoresis. Elec-trophoresis 27:3399–3409.

Tabas LB and Dantzig AH (2002) A high-throughput assay for measurement of multidrugresistance protein-mediated transport of leukotriene C4 into membrane vesicles. Anal Biochem310:61–66.

Tannock IF and Rotin D (1989) Acid pH in tumors and its potential for therapeutic exploitation.Cancer Res 49:4373–4384.

Unal ES, Zhao R, Chang MH, Fiser A, Romero MF, and Goldman ID (2009) The functional rolesof the His247 and His281 residues in folate and proton translocation mediated by the humanproton-coupled folate transporter SLC46A1. J Biol Chem 284:17846–17857.

Urakami Y, Okuda M, Masuda S, Saito H, and Inui KI (1998) Functional characteristics andmembrane localization of rat multispecific organic cation transporters, OCT1 and OCT2,mediating tubular secretion of cationic drugs. J Pharmacol Exp Ther 287:800–805.

Vaupel P (2004) Tumor microenvironmental physiology and its implications for radiationoncology. Semin Radiat Oncol 14:198–206.

Volk EL and Schneider E (2003) Wild-type breast cancer resistance protein (BCRP/ABCG2) isa methotrexate polyglutamate transporter. Cancer Res 63:5538–5543.

Wang H, Lee EW, Cai X, Ni Z, Zhou L, and Mao Q (2008) Membrane topology of the humanbreast cancer resistance protein (BCRP/ABCG2) determined by epitope insertion and immu-nofluorescence. Biochemistry 47:13778–13787.

Zhao R and Goldman ID (2007) The molecular identity and characterization of a Proton-coupledFolate Transporter—PCFT; biological ramifications and impact on the activity of pemetrexed.Cancer Metastasis Rev 26:129–139.

Address correspondence to: William F. Elmquist, University of Minnesota,308 Harvard Street SE, 9-125 Weaver-Densford Hall, Minneapolis, MN 55455.E-mail: elmqu011@umn.edu

1485pH-DEPENDENT TRANSPORT OF PEMETREXED BY BCRP

at ASPE

T Journals on M

ay 7, 2018dm

d.aspetjournals.orgD

ownloaded from