Abstract
In this study we compared the in silico predictions of the effect of ABCC1 nonsynonymous single nucleotide polymorphisms (nsSNPs) with experimental data on MRP1 transport function and response to chemotherapeutics and multidrug resistance protein 1 (MRP1) inhibitors. Vectors encoding seven ABCC1 nsSNPs were stably expressed in human embryonic kidney (HEK) cells, and levels and localization of the mutant MRP1 proteins were determined by confocal microscopy and immunoblotting. The function of five of the mutant proteins was determined using cell-based drug and inhibitor sensitivity and efflux assays, and membrane-based organic anion transport assays. Predicted consequences of the mutations were determined by multiple bioinformatic methods. Mutants C43S and S92F were correctly routed to the HEK cell plasma membrane, but the levels were too low to permit functional characterization. In contrast, levels and membrane trafficking of R633Q, G671V, R723Q, A989T, and C1047S were similar to wild-type MRP1. In cell-based assays, all five mutants were equally effective at effluxing calcein, but only two exhibited reduced resistance to etoposide (C1047S) and vincristine (A989T; C1047S). The GSH-dependent inhibitor LY465803 (LY465803 [N-[3-(9-chloro-3-methyl-4-oxo-4H-isoxazolo-[4,3-c]quinolin-5-yl)-cyclohexylmethyl]-benzamide)] was less effective at blocking calcein efflux by A989T, but in a membrane-based assay, organic anion transport by A989T and C1047S was inhibited by MRP1 modulators as well as wild-type MRP1. GSH accumulation assays suggest cellular GSH efflux by A989T and C1047S may be impaired. In conclusion, although six in silico analyses consistently predict deleterious consequences of ABCC1 nsSNPs G671V, changes in drug resistance and inhibitor sensitivity were only observed for A989T and C1047S, which may relate to GSH transport differences.
Introduction
The human 190-kDa multidrug resistance protein (MRP) 1 is a member of the C subfamily of the ATP-binding cassette (ABC) superfamily of proteins that mediate the cellular efflux of small molecules (Cole et al., 1992; Cole, 2013). MRP1 (encoded by ABCC1) and six other ABCC proteins have a distinct five-domain structure with two nucleotide-binding domains (NBDs) and 17 transmembrane (TM) α-helices distributed among three membrane-spanning domains (MSD0, MSD1, and MSD2) (Slot et al., 2011) (Fig. 1). The “pore” in the membrane through which substrates are effluxed in an ATP-dependent manner is formed by MSD1 and MSD2.
MRP1 effluxes many hydrophobic (e.g., doxorubicin) and hydrophilic (e.g., methotrexate) chemotherapeutic agents, causing tumor cells to become multidrug resistant (Cole et al., 1994; Rappa et al., 1997; Cole, 2013). High ABCC1 expression has been associated with unfavorable clinical outcomes in several malignancies, most notably in neuroblastoma where it correlates with MYCN oncogene expression for which ABCC1 is a downstream target (Pajic et al., 2005; Henderson et al., 2011).
MRP1 can also efflux many other therapeutic agents and physiologic molecules, including organic anion metabolites conjugated to glutathione (GSH) (e.g., leukotriene C4 [LTC4]), and glucuronide (e.g., estradiol glucuronide [E217βG]) (Cole, 2013). Thus, MRP1 is involved in the disposition and elimination of drugs and their metabolites. It also transports GSH in both its antioxidant and prooxidant (glutathione disulfide) states. Physiologic roles for MRP1 have been established, not only by its ability to modulate GSH homeostasis (Ballatori et al., 2009; Cole, 2013) but also by the characterization of Abcc1−/− mice which exhibit impaired inflammatory responses attributed to disrupted leukotriene homeostasis (Wijnholds et al., 1997; Yoshioka et al., 2009).
Single-nucleotide polymorphisms (SNPs) are common in genes encoding proteins involved in drug metabolism and transport, and interindividual differences can contribute to differences in absorption, distribution, and elimination that ultimately affect drug efficacy and toxicity (Kroetz et al., 2010; Ieiri, 2012). The polymorphic ABCC1 on chromosome 16p13 is no exception (Conseil et al., 2005; http://www.ncbi.nlm.nih.gov/snp). Of the many human ABCC1 polymorphisms, >80 are nonsynonymous SNPs (nsSNPs), and the encoded amino acid substitutions occur in all five domains of MRP1 (Fig. 1). Previously, we generated and partially characterized recombinant forms of ABCC1 nsSNPs: rs45511401 (2012G>T; G671V), rs60782127 (1299G>T; R433S), and rs41395947 (128G>C; C43S) (Conrad et al., 2001, 2002; Leslie et al., 2003). We found that the G671V mutation had no deleterious effects on the levels or organic anion (i.e., LTC4, E217βG) transport function of MRP1 when expressed in human embryonic kidney (HEK) cells, despite the fact this mutation is located near NBD1 (Conrad et al., 2001). In contrast, the R433S mutant exhibited reduced LTC4 and estrone sulfate transport levels (Conrad et al., 2002). Cells expressing MPR1-R433S also showed increased doxorubicin resistance, suggesting R433S can better efflux this drug despite its lower organic anion transport activity. Yet another phenotype was observed for the C43S nsSNP in the first TM helix of MSD0. Not only did C43S increase organic anion transport activity and reduce resistance to vincristine and arsenite (Leslie et al., 2003), it also mildly disrupted MRP1 membrane trafficking. In a subsequent characterization of the organic anion transport activity of 10 additional ABCC1 nsSNPs, only one (rs35529209; 2965G>A; A989T) showed a significant reduction in E217βG transport (Létourneau et al., 2005).
In contrast to our experimental data, the predictive algorithms SIFT (Sorting Tolerant From Intolerant) and PolyPhen indicated that G671V would adversely affect MRP1 function but C43S and A989T were less likely to do so, whereas predictions for R433S were mixed (Létourneau et al., 2005). Thus, although our in vitro studies indicated that these four nsSNPs can cause distinct phenotypic changes in MRP1 levels and/or activity, these effects were not well predicted by SIFT and PolyPhen (Hao et al., 2011).
Point mutations introduced for the purpose of investigating MRP1 structure and function relationships can also have variable effects on transporter function, ranging from a very substrate-selective effect on transport activity to disrupting interactions with ATP (Ito et al., 2001; Haimeur et al., 2002; Létourneau et al., 2008). Still others lead to misfolding and lower levels of MRP1 at the plasma membrane (Conseil et al., 2009; Iram and Cole, 2012). Consequently, we wondered if the discordance between the effects of the ABCC1 nsSNPs predicted in silico and our experimental data were due to limitations of the SIFT and PolyPhen algorithms and/or because the functional assays used were not sufficiently comprehensive to detect all possible phenotypic changes.
In the present study, we further examined the phenotypes of seven ABCC1 nsSNPs in vitro. Four (C43S, S92F, G671V, A989T) were mutants that our earlier studies showed had a phenotype discordant from that predicted by Polyphen and/or SIFT (Létourneau et al., 2005). Two others (R633Q, R723Q) were selected because of their location in NBD1. The last, C1047S, was selected because the predicted probability of this mutation having a deleterious effect by SIFT approached statistical significance.
Materials and Methods
Doxorubicin, vincristine, etoposide (VP-16), MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide), GSH, S-methylGSH (S-MeGSH), AMP, ATP, apigenin, E217βG, Triton-X100, and BAYu9773 (6(R)-(4-carboxyphenylthio)-5(S)-hydroxy-7(E),9(E),11(Z),14(Z)-eicosatetraenoic acid) were from Sigma-Aldrich (Oakville, ON, Canada). Sodium arsenite (Na2HAsO4.7H2O) and potassium antimony tartrate (KSbO3.3H2O) were from J.T. Baker Chemical Co. (Phillipsburg, NJ). Calcein-AM (calcein acetoxymethylester) was from Alexis Biochemicals (Cedarlane Laboratories, Burlington, ON, Canada). Dulbecco’s modified Eagle’s medium, OptiMEM, fetal bovine serum (FBS) and geneticin (G418) were from GIBCO (Invitrogen, Scarborough, ON) and diphenylcarbamyl chloride-treated trypsin was from ICN MP Biomedical (Aurora, OH). LY171883 (1-[2-hydroxy-3-propyl-4-[4-(1H-tetrazol-5-yl)butoxy]phenyl]-ethanone) was from Cayman Chemical (Ann Arbor, MI). LY465803 (N-[3-(9-chloro-3-methyl-4-oxo-4H-isoxazolo-[4,3-c]quinolin-5-yl)-cyclohexylmethyl]-benzamide) was a kind gift from Eli Lilly (Indianapolis, IN) (Dantzig et al., 2004). We purchased [6,7-3H]E217βG (55 Ci mmol−1), and [glycine-2-3H]GSH (40-44.8 Ci mmol−1) from Perkin Elmer Life Sciences (Boston, MA).The MRP1-specific mAbs 897.2 and 42.4 were kind gifts from Dr. X.B. Chang (Mayo Clinic, Scottsdale, AZ) (Hou et al., 2000). The mAbs MRPm5, MRPm6, and MRPr1 were kind gifts from Drs. R.J. Scheper and G.L. Scheffer (VU University Medical Center, Amsterdam, the Netherlands) (Hipfner et al., 1998); and mAbs QCRL-1 and QCRL-3 were generated in this laboratory (Hipfner et al., 1999).
Bioinformatic Analyses.
The computer algorithms PolyPhen2 (Polymorphism Phenotyping) (http://genetics.bwh.harvard.edu/pph2/index.shtml), SIFT (Sorting Tolerant From Intolerant) version 2, and SIFTBLink (http://blocks.fhcrc.org/sift/SIFT.html) (Ng and Henikoff, 2001, 2002) were used to analyze the nsSNPs of ABCC1. In addition, Grantham value differences and values from the Blosum50 (BLOcks SUbstitution Matrix) and PAM250 (Point Accepted Mutation) matrices were determined as described elsewhere (Grantham 1974; Dayhoff et al., 1978; Henikoff and Henikoff, 1992).
Generation of HEK293 Cell Lines Stably Expressing MRP1 Mutants.
To generate stably transfected cell lines, HEK293 cells were seeded in a six-well plate (0.5 × 106 cells per well), and 24 hours later transfected with pcDNA3.1(−) expression vectors containing wild-type or mutant cDNA (Conrad et al., 2001; Leslie et al., 2003; Létourneau et al., 2005) using Lipofectamine 2000 (Invitrogen) according to manufacturer’s instructions. Forty-eight hours later, the cells were transferred to a 15-cm plate and grown in the presence of G418 (1 mg ml−1) for 1 week. Individual colonies were picked and expanded in six-well plates. Cells were then cloned by limiting dilution and assessed for MRP1 protein levels by immunoblotting using mAb QCRL-1. The homogeneity of MRP1 expression in the cell lines was confirmed by flow cytometry using MAb QCRL-3 on an Epics Altra HSS flow cytometer (Beckman Coulter, Brea, CA), after fixing the cells with paraformaldehyde (0.5%) and permeabilizing with Triton X-100 (0.1%) (Hipfner et al., 1996, 1999). The stable cell lines were maintained in Dulbecco’s modified Eagle’s medium supplemented with 7.5% FBS and G418 (0.5 mg ml−1).
Confocal Fluorescence Microscopy.
To localize wild-type and mutant MRP1 in the HEK293 cell lines, cells were seeded on gelatinized coverslips, and 24 hours later were incubated with mAb QCRL-3 (1:2,500) for 1 hour, followed by 30 minutes with Alexa Fluor488 anti-mouse IgG (H+L) (Fab′)2 fragment (1:500) (Molecular Probes/Life Technologies, Eugene, OR) (Conseil et al., 2006). Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich), and then the cells were examined using a Leica TCS SP2 MS multiphoton system confocal microscope (Leica Microsystems, Heidelberg, DE).
Chemosensitivity Testing.
The drug sensitivity of the MRP1 HEK293 cell lines was tested over a range of concentrations using the colorimetric MTT assay with determinations done in quadruplicate in each experiment (Leslie et al., 2003). The IC50 was calculated from the best fit of the data to a sigmoidal curve using GraphPad prism software (GraphPad Software, San Diego, CA). The mean IC50 (± S.D.) was determined from three independent experiments. The relative resistance factor was determined as the ratio of the IC50 for the MRP1 transfected cell line to the IC50 of the sensitive HEK293 parental cell line. Data were analyzed using a t test, and P < 0.05 was considered statistically significant.
Calcein Efflux Activity.
The effect of the nsSNPs on MRP1-mediated calcein efflux from the HEK cell lines was measured using a modification of the protocol of van Zanden et al. (2005) (Myette et al., 2013). Briefly, HEK293 cells were seeded at 5 × 104 per well in a 96-well plate (black side, clear bottom) (Corning Costar, Corning, NY) coated with poly-l-lysine, in OptiMEM supplemented with 7.5% FBS. Twenty-four hours later, calcein-AM (dissolved in dimethylsulfoxide [DMSO]) and DMSO (vehicle control; 1% final concentration) were added. Media were removed 1 hour later and replaced with OptiMEM/7.5% FBS, and then the cells were incubated for a further 3 hours to allow for MRP1-dependent calcein efflux. In experiments with MRP1 modulators, the modulators (BAYu9773, LY171883, and LY465803) were added for the entire 1-hour incubation with calcein-AM as well as the subsequent 3-hour incubation period. The medium was then replaced by phosphate-buffered saline, and the fluorescence inside the cells was measured using a Synergy HT Multi-Mode microplate reader (BioTek Instruments, Inc., Winooski, VT) (λexc 485/20 nm and λem 528/20 nm, sensitivity = 25) in an end-point mode. In experiments with the LY465803 inhibitor, various concentrations of the inhibitor (range: 0–10 µM) were added before the addition of a single concentration of calcein-AM (6 µM); conversely, a single concentration of LY465308 (2 µM) was added before the addition of different concentrations of calcein-AM (range: 0–6 µM). The amount of calcein accumulated in the cells was measured as arbitrary fluorescence units, and the mean (± S.D.) of measurements performed in triplicate in each experiment was calculated. All experiments were repeated at least once.
[3H]E217βG and [3H]GSH Uptake by Inside-out Membrane Vesicles.
Membrane vesicles from the HEK293 cell lines were prepared, and MRP1 levels were determined by immunoblotting with MAb QCRL-1 as described previously elsewhere (Conseil et al., 2009). The ATP-dependent uptake of [3H]E217βG by the inside-out membrane vesicles was measured using a modified rapid filtration method adapted to a 96-well microtiter plate format (Létourneau et al., 2005). Transport assays were performed at 37°C in a 30-µl reaction volume containing 10 mM MgCl2 and 4 mM ATP (or 4 mM AMP) in transport buffer (250 mM sucrose/50 mM Tris–HCl, pH 7.5) and 1) 2 µg of membrane vesicle protein and [3H]E217βG (400 nM; 20 nCi per reaction); or 2) 20 µg of membrane vesicle protein, [3H]GSH (100 µM;120 nCi per reaction) and 30 µM of apigenin. The modulators (BAYu9773, LY465803 [plus 1 mM GSH/10 mM DTT], and LY171883) were prepared as stock solutions in DMSO and then diluted at least 100-fold in transport buffer before preincubating (on ice) with the membrane vesicles.
After the times indicated in the figure legends, uptake was stopped by rapid dilution in ice-cold transport buffer, and the reactions were filtered through a Perkin Elmer unifilter GF/B plate. Tritium accumulated in the vesicles was counted on a Top Count NXT Microplate Scintillation and Luminescence Counter (Perkin Elmer Life Sciences), and the ATP-dependent E217βG or GSH uptake was calculated by subtracting the uptake in the presence of AMP from the uptake measured in the presence of ATP. All transport assays were performed in duplicate, and the results were expressed as a percentage of control uptake (no modulator); data were fitted to a sigmoidal curve by nonlinear regression analysis using GraphPad Prism.
Determination of Cellular GSH Levels.
Total cellular glutathione in HEK293 wild-type, A989T and C1047S cells was measured using the enzymatic recycling method of Tietze (1969) and Brehe and Burch (1976). Briefly, 5–20 × 106 cells were suspended in 5% (w/v) sulfosalicylic acid (Sigma-Aldrich), sonicated for 45 seconds, and the disrupted cells were left on ice for 20 minutes. After centrifugation at 12,000g, 10 µl of supernatant from different samples were distributed in triplicate in a 96-well plate. Multiple concentrations of GSSG were also plated in triplicate to generate a standard curve. Reaction mix (200 µl) consisting of EDTA (1 mM), Na2H2PO4 (100 mM), DTNB (0.15 mM), NADPH (0.2 mM), and GSH reductase (1 U ml−1) (Sigma-Aldrich) were added, and absorbance at 405 nm was measured using a Synergy HT microplate reader (Bio-Tek Instruments Inc.). The data were analyzed using the KC4 v3.3 program.
Limited Trypsin Digestion of Membrane Proteins.
Tryptic fragmentation patterns of wild-type and A989T mutant MRP1 were determined as described previously elsewhere (Rothnie et al., 2006; Iram and Cole, 2012). Briefly, membrane vesicles (2 µg of protein per lane) enriched for wild-type or A989T mutant MRP1 were incubated in hypotonic buffer (50 mM HEPES, pH 7.4) in the absence or presence of S-methylGSH (10 mM) for 30 minutes on ice. Trypsin (at trypsin/protein ratios ranging from 1:5,000 to 5:1) were then added, and the samples were incubated for 15 minutes at 37°C. Digestions were stopped by placing samples on ice and adding Laemmli buffer containing protease inhibitors phenylmethylsulfonyl fluoride (PMSF) and leupeptin. Samples were resolved by SDS-PAGE on 4%–20% tris-glycine gradient gels (BioRad Laboratories, Hercules, CA) and immunoblotted. Full-length and tryptic fragments of MRP1 were detected using MRP1-specific mAbs MRPr1 (1:5,000), 42.4 (1:5,000), MRPm5 (1:5,000), 897.2 (1:5,000), and MRPm6 (1:5,000), whose epitopes have been mapped to amino acids 238–247, 723–732, 1063–1072; 1318–1388, and 1511–1520, respectively (Hipfner et al., 1998; Hou et al., 2000; Iram and Cole, 2012).
Results
Bioinformatic Analyses of Predicted Consequences of MRP1 Mutations.
The seven ABCC1 nsSNPs were analyzed by the most up-to-date versions of PolyPhen2 (Adzhubei et al., 2010), SIFT v.2, and SIFT/BLink (Kumar et al., 2009) available to determine their predicted effects on MRP1 function; the results are shown in Table 1. Also included in Table 1 are the Grantham value differences, stability values according to I-mutant Suite, as well as the scores from the Blosum50 and PAM250 matrices for the seven nsSNPs (Dayhoff et al., 1978; Henikoff and Henikoff, 1992; Capriotti et al., 2005).
All the algorithms and matrices are in agreement that the G671V mutation located close to the Walker A motif in NBD1 is the most likely to affect MRP1 activity. Thus, Val substitution of this highly conserved Gly residue (Supplemental Table 1) is expected to probably be damaging according to PolyPhen2 and because a large decrease in stability (I-mutant suite) and a dramatic change in physicochemical properties are predicted (Grantham value). The two matrices Blosum50 and PAM250 as well as SIFT also predict that this substitution is unlikely to be tolerated without adverse phenotypic consequences.
The analyses of the two other nsSNPs located in NBD1, namely, R633Q and R723Q, yield results very similar to one another (low probability of adverse effect), except that Arg633 is considered to be more critical to MRP1 stability by I-mutant Suite (Table 1), probably because Arg633 is more strictly conserved than Arg723 (Supplemental Table 1). The predicted effects of the nsSNPs C43S and S92F in MSD0 are mixed (Table 1). They are predicted to be possibly damaging by Polyphen2 but tolerated by SIFT (except for SIFTBLink which places S92F at the threshold), and to cause substantial physicochemical changes and probably low occurrence (Grantham and matrices), yet the I-mutant Suite predicts C43S to be more destabilizing than S92F (Table 1).
The predicted effects of the nsSNPs A989T and C1047S located in TM12 and TM13 of MSD2, respectively, are also mixed. Although they are predicted to be benign and tolerated substitutions by Polyphen2 and SIFT, the I-mutant Suite predicts that these substitutions would cause a large decrease in MRP1 stability, with a significant change in physicochemical properties (Grantham) but a low probability of occurrence (Blosum and PAM250 matrices).
Given that the predicted consequences of the various MRP1 nsSNPs varied widely among the different in silico methods, we attempted to provide a simpler overview by assigning each nsSNP a score (0, no predicted effect; 1, mild effect; 2, severe effect) according to the analysis by each of the six predictive methods, and then summing the individual scores to obtain a ranking (Supplemental Table 2). According to this method, the mutants ranked (from highest to lowest likelihood of an adverse effect) as follows: G671V > S92F > C43S > C1047S > R633Q > R723Q = A989T.
nsSNPs R633Q, G671V, R723Q, A989T, and C1047S Have No Effect on Total on Plasma Membrane MRP1 Levels in HEK293 Cells.
After we had isolated stably transfected HEK293 cell lines by G418 selection, cell lines expressing R633Q, G671V, R723Q, A989T, and C1047S and wild-type MRP1 were cloned to >90% homogeneity for MRP1 expression. Immunoblots of cell lysates showed that the levels of the five mutant proteins were comparable to (R633Q, A989T, C1047S) or somewhat (<50%) higher than (G671V, R723Q) wild-type MRP1, indicating that the mutations do not cause any major misfolding of MRP1 that would result in its degradation (Fig. 1B). However, despite our repeated attempts, HEK cell lines homogeneously expressing the two MSD0 mutants, C43S and S92F, could not be isolated. The maximal proportion of cells positive for these mutants that could be obtained was 55% and 65%, respectively (data not shown). Immunoblots of lysates prepared from the S92F and C43S cell lines showed MRP1 protein levels that were much lower than wild-type MRP1(Supplemental Fig. 1), as might be expected from nonclonal cell lines.
Confocal fluorescence microscopy experiments showed that the R633Q, G671V, R723Q, A989T, and C1047S mutant proteins in the five clonal HEK cell lines were also routed correctly to the plasma membrane in a manner indistinguishable from wild-type MRP1 (Fig. 2). The C43S and S92F mutants in the nonclonal cell populations were also properly routed to the plasma membrane although, as expected because of the heterogeneity of MRP1 expression in these cell lines, significantly fewer cells expressed these proteins (Supplemental Fig. 1). Because of the confounding effects of mixed cell populations on the interpretation of subsequent functional assays, the C43S and S92F mutants were not characterized further.
nsSNPs Selectively Affect Anticancer Drug Resistance of HEK Cell Lines.
The HEK cell lines expressing R633Q, G671V, R723Q, A989T, and C1047S were tested for their levels of resistance to five xenobiotics for which human MRP1 is known to confer resistance, including the antineoplastic agents vincristine, etoposide (VP-16), doxorubicin, and the heavy metal oxyanions arsenite and antimony tartrate (Cole et al., 1994). As summarized in Table 2, the relative resistance levels of the cell lines expressing mutant MRP1 to doxorubicin and heavy metal anions were very similar in most cases to the levels of resistance conferred by wild-type MRP1. Levels of vincristine resistance were the most variable, but the differences were only statistically significant (approximately 2.5-fold lower than wild-type MRP1, P < 0.05) in the cell lines expressing A989T and C1047S. Etoposide resistance of the C1047S cell line was also lower than the wild-type MRP1 cell line, but the difference was more modest (1.6-fold; P < 0.05); doxorubicin resistance was also decreased, but the difference was not statistically significant. Resistance of the A989T and C1047S cell lines to arsenite was comparable to the wild-type MRP1 cell line while resistance to antimony tartrate was reduced (1.7-fold and 3.0-fold, respectively).
nsSNPs Do Not Affect Calcein Efflux from HEK293 Cell Lines.
To determine whether the five nsSNPs, R633Q, G671V, R723Q, A989T, and C1047S, affected the ability of MRP1 to mediate efflux of calcein, HEK293 cells stably expressing wild-type and mutant MRP1 as well as untransfected HEK cells were incubated with several concentrations of the cell permeable acetoxymethyl ester of calcein (calcein-AM) at 37°C; 3 hours later, the intracellular hydrolyzed calcein that had not been effluxed by MRP1 was measured. As shown in Fig. 3, the calcein levels in nontransfected HEK293 control cells increased proportionally with the increase in concentration of calcein-AM loaded. The calcein levels in the cell lines expressing both wild-type and mutant MRP1 also increased as the concentration of calcein-AM increased; however, the levels were >80% lower than in the nontransfected cells, confirming the role of MRP1 in active efflux of calcein. Calcein levels in the five mutant cell lines were similar to those in the cell line expressing wild-type MRP1, indicating that none of the five nsSNPs significantly affected cellular calcein efflux. Thus, unlike the chemosensitivity assay, which showed that the drug-resistance patterns of A989T and C1047S were different from wild-type MRP1 (Table 2), the calcein efflux assay did not detect any differences (Fig. 3).
nsSNPs A989T, But Not C1047S, Affects the Inhibition of MRP1-Mediated Calcein Efflux by LY465803.
To further explore the selective effects of nsSNPs A989T and C1047S on their ability to recognize xenobiotics, we determined whether the two mutations affect the efficacy of MRP1 modulators. Consequently, the GSH-dependent tricyclic isoxazole MRP1 inhibitor LY465803 was tested for its ability to inhibit calcein efflux from the HEK cell lines (Mao et al., 2002; Dantzig et al., 2004). As shown in Fig. 4A, similar fluorescence intensities were recorded in control nontransfected HEK cells and in cells expressing wild-type MRP1 and exposed to 2 µM LY465803, showing the wild-type protein was completely inhibited by LY465803 as expected (Maeno et al., 2009). Similarly, the calcein efflux activity of the C1047S mutant was completely inhibited. However, the fluorescence levels in cells expressing A989T were decreased by 50%, indicating that the mutation had diminished the ability of LY465803 to inhibit MRP1. Similarly, in the converse experiment, when cells were preincubated with increasing concentrations of LY465803 before adding a single concentration of calcein-AM (6 µM), the fluorescence levels observed for the cells expressing the A989T mutant were consistently ∼50% lower than the fluorescence levels in the control untransfected HEK293 cells as well as the cells expressing wild-type and C1047S mutant MRP1 at all LY465803 concentrations tested (Fig. 4B).
nsSNPs A989T and C1047S Do Not Affect Inhibition of [3H]E217βG Vesicular Transport by LY465803 and the Leukotriene Modifiers LY171883 and BAYu9773.
To determine whether the nsSNP A989T also affected the sensitivity of MRP1 to the GSH-dependent LY465803 inhibitor in a vesicular transport assay, the ability of LY465803 to inhibit uptake of [3H]E217βG into membrane vesicles prepared from the HEK cell line expressing A989T was examined; cell lines expressing the C1047S mutant and wild-type MRP1 were included as controls. Preliminary experiments confirmed that E217βG vesicular uptake by A989T was partially (50%) reduced in vesicles from the A989T HEK cell line, as it had been shown previously in vesicles prepared from transiently transfected A989T HEK cells (Létourneau et al., 2005) (Supplemental Fig. 2). Representative concentration-response curves are shown in Fig. 5. As can be seen in Table 3 from the data summary of multiple experiments, the IC50 values of LY465803 for A989T (0.05 ± 0.01 µM) and C1047S (0.03 ± 0.03 µM) were not significantly different from the IC50 for wild-type MRP1 (0.07 ± 0.01 µM) although the difference approached statistical significance for C1047S (P = 0.07).
The IC50 values were also determined for two non-GSH-dependent MRP1 inhibitors, BAYu9773 and LY171883 (Fig. 5; Table 3). BAYu9773, a nonselective antagonist of the CysLT receptors 1 and 2 and LY171883, a selective antagonist of CysLT receptor 1, inhibited E217βG transport by A989T with IC50 values of 0.32 ± 0.19 µM and 19.58 ± 6.15 µM, respectively. These values were not significantly different from the IC50 values obtained for BAYu9773-mediated and LY171883-mediated inhibition of E217βG transport by either wild-type or C1047S MRP1 (Table 3) (0.69 ± 0.18 µM and 0.62 ± 0.16 µM [for BAYu9773] and 16.10 ± 1.28 µM and 23.90 ± 9.85 µM [for LY171883] for wild-type and C1047S, respectively; P > 0.05).
Effect of nsSNPs A989T and C1047S on [3H]GSH Transport and Cellular GSH Levels.
Because the A989T nsSNP was associated with a decreased inhibitory potency of the GSH-dependent LY465803 inhibitor in a cell-based (where its effects are dependent on intracellular GSH concentrations) but not in a membrane-based transport assay (where exogenous GSH is provided and is not limiting), it was of interest to determine whether vesicular transport of GSH and/or cellular levels of GSH were affected by this nsSNP. The C1047S mutant was included for comparative purposes. As shown in Fig. 6A, levels of apigenin-stimulated GSH uptake by A989T and C1047S were 16% and 26% lower than for wild-type MRP1, but in neither case were these differences statistically significant (P = 0.2 and 0.16, respectively). Thus, the difference in results obtained with LY465803 and the nsSNP A989T in the cell-based versus membrane-based assays do not appear explainable by differences in transport and hence binding of GSH.
The GSH content of the HEK cell lines was also quantified using a colorimetric assay. As shown in Fig. 6B, GSH levels were approximately 50% lower in cells expressing wild-type MRP1 than control untransfected cells (26.0 ± 6.0 µM versus 46.3 ± 12.9 µM), a difference that approached statistical significance (P = 0.07). Reduced levels of GSH are expected in MRP1-overexpressing cells because of the active efflux of GSH-mediated by MRP1 (Rappa et al., 1997; Cole, 2013). In contrast to the wild-type MRP1 cell line, the GSH levels present in the A989T and C1047S cell lines were comparable (106% and 94%, respectively) to those of the control untransfected cell line, suggesting that both nsSNPs adversely affect the ability of MRP1 to efflux GSH from intact cells (P = 0.08 and 0.02, respectively, for A989T and C1047S cells compared with wild-type MRP1 cells).
nsSNPs A989T and C1047S Have No Major Impact on the Protease Susceptibility of MRP1.
Because the functional changes detected for A989T and C1047S were the most significant among all the nsSNPs tested, we determined whether any of these changes were associated with any differences in protein conformation as reflected by a change in protease susceptibility. Thus, membrane vesicles enriched for wild-type and mutant MRP1 were digested with a range of trypsin/protein ratios, and tryptic fragments were detected by immunoblotting with a panel of antibodies that detect epitopes in different domains of the transporter (Iram and Cole, 2012). Tryptic digests were performed in the presence and absence of S-methyl GSH because previous studies have shown that this tripeptide (and the less stable GSH) can stimulate substrate transport (and modulate inhibitor activity) as well as alter the conformation of MRP1 (Loe et al., 1998; Mao et al., 2002; Peklak-Scott et al., 2005; Ren et al., 2005; Rothnie et al., 2006; Cole, 2013).
As shown in Supplemental Fig. 3A, when blots were probed with mAbs MRPr1 and 42.4 that detect epitopes in the NH2-half of MRP1, no differences in the tryptic fragment patterns obtained were observed between A989T (and C1047S, data not shown) and wild-type MRP1. Similarly, no differences were detected when the blots were probed with mAbs MRPm5, 897.2, and MRPm6, which detect epitopes located in the COOH-half of MRP1 (Supplemental Fig. 3B). When the experiments were performed in the presence of S-methyl GSH (10 mM), the fragment patterns were similar to those obtained in the absence of S-methyl GSH (Supplemental Fig. 3, A and B, and data not shown), suggesting that the nsSNPs did not affect the change of conformation that can be observed upon binding of GSH (Rothnie et al., 2006).
Discussion
MRP1 was first cloned in 1992 from a drug-selected lung cancer cell line based on the amplification of its cognate ABCC1 gene that was associated with a 100-fold increase in mRNA relative to its drug-sensitive parental cell line (Cole et al., 1992). Since then, elevated MRP1 levels have been reported in many drug-resistant tumors, both in vitro and in patient samples (Deeley et al., 2006). MRP1 is also involved in the normal tissue distribution of many therapeutic agents in addition to antineoplastic drugs; consequently, it can influence their efficacy and toxicity (Knauer et al., 2010; Cole, 2013).
Many factors contribute to therapeutic and adverse drug responses in cancer and other human diseases, including polymorphisms in drug transporters such as ABCC1. In this study, we have further investigated the phenotypic consequences of several ABCC1 nsSNPs and compared the experimental data obtained with stable HEK cell lines with the consequences of the amino acid substitutions predicted by various in silico methods. The phenotypes of the mutant MRP1 proteins observed may be best considered as belonging to one of three categories.
The first category includes the two nsSNPs in TM1 (C43S) and TM2 (S92F) that could not be functionally characterized because stable homogeneously expressing clonal HEK cell lines for these mutants could not be isolated. However, in the heterogeneous cell populations, the mutant proteins appeared properly routed to the plasma membrane. Cys43 and Ser92 are both located NH2-proximal to the core four-domain structure of MRP1 in MSD0; and, although a role in dimerization has been proposed, MSD0 appears dispensable for organic anion transport activity (Bakos et al., 2000; Deeley et al., 2006; Yang et al., 2010). However, most in silico prediction programs based on chemical properties, amino acid conservation, and structural data classify the C43S and S92F nsSNPs as very likely to be deleterious to MRP1 function.
Previously, we showed that C43S expressed in HeLa cells was associated with decreased arsenite and vincristine resistance, and moderately impaired MRP1 trafficking to the plasma membrane (Leslie et al., 2003). This was somewhat surprising because replacing a Cys residue with a Ser is a relatively conservative substitution. On the other hand, if the side chain of Cys43 in TM1 is accessible, a Ser substitution would prevent it from forming a potentially functional disulfide bridge with another Cys residue in the same or different molecule of MRP1. To date, however, no disulfide bonds have been demonstrated to exist for any membrane-embedded Cys residue in MRP1 or other ABC transporter. The inability of the S92F mutant to be established as a stable homogenously expressing HEK cell line is also somewhat unexpected because S92F was readily expressed in transiently transfected HEK cells (Létourneau et al., 2005). In this latter system, S92F transported organic anions LTC4, E217βG, and methotrexate at levels only modestly below those of wild-type MRP1.
The second category of mutants examined includes the three nsSNPs located in NBD1 (R633Q, G671V, R723Q), which exhibited little, if any, change in transport levels or sensitivity to MRP1 substrates or inhibitors. For R633Q, this is not surprising because Arg633 is not located in or close to any functionally important motifs in NBD1. Furthermore, only one of the in silico methods (I-mutant Suite) suggested this substitution would reduce MRP1 stability, and all others predicted no significant change. Thus, our results here showing little or no effect of R633Q on MRP1 expression or function are consistent with in silico predictions and with our earlier study on transiently transfected cells (Létourneau et al., 2005).
This contrasts with G671V, which all prediction algorithms, physicochemical parameters, and structure analyses concur should be deleterious (Table 2). Indeed, the G671V mutation ranked as the most damaging of all seven nsSNPs examined here. This is consistent with the strict conservation of Gly671 in the large majority of MRP1 homologs and orthologs (Supplemental Table 1) as well as other ABC transporters, and with its location just seven amino acids away from the functionally critical Walker A motif (Walker et al., 1982). Furthermore, the 2012G>T (G671V) nsSNP has been associated with an altered response to certain drugs, including adverse reactions, in several studies. For example, the 2012T allele is associated with the acute anthracycline-induced cardiotoxicity in patients with non-Hodgkin’s lymphoma (Wojnowski et al., 2005). More recently, Vulsteke et al. (2013) reported an association of the 2012T allele with febrile neutropenia in breast cancer patients receiving a chemotherapy regimen that included epirubicin. Pajic et al. (2011) also found that the 2012T allele had the potential to be a significant prognostic indicator of a better outcome for children older than 1 year with neuroblastoma. They further showed that the variant 2012T mRNA transcript levels were reduced in cell lines heterozygous for this nsSNP, a change associated with reduced ABCC1 mRNA stability. These findings are consistent with our earlier observation that lymphocytes from normal individuals bearing this G671V nsSNP expressed ABCC1 mRNA at relatively low levels (Conrad et al., 2001), and they suggest a change in mRNA structure is more important than the amino acid change caused by this nSNP.
Thus, none of the cell-based or membrane-based assays in our present and previous studies with stably and transiently transfected HEK or HeLa cells revealed any changes in levels, organic anion transport activity, drug/inhibitor sensitivity, or plasma membrane localization relative to wild-type MRP1 (Conrad et al., 2001). In contrast, however, Jungsuwadee et al. (2012) reported an approximately 3-fold reduction in doxorubicin resistance in stably transfected HEK293 cells. Because redox cycling of doxorubicin causes lipid peroxidation, they suggested that this might be related to the reduced transport of the GSH-conjugated metabolite of 4-hydroxy-2-trans-nonenal also observed, which they largely attribute to a change in apparent Vmax for this substrate. The reasons for the discordance between most of the experimental observations with the G671V mutant protein versus the uniformly adverse effects predicted for the Val substitution of Gly671 by the in silico analytical methods and the genetic association studies are unclear but may well be related to the cellular context in which the mutant allele is expressed (Pajic et al., 2011). It is also possible that analyses in a cell culture setting may not be adequate to discern all aspects of G671V function.
With respect to the R723Q nsSNP, a conserved positive charge is lost (Supplemental Table 1), and the side chain has been shortened. However, Arg723 is at some distance from the functionally important Walker A and B motifs as well as the ABC transport signature motif LSGGQ in NBD1 (Hyde et al., 1990). Consequently, in this case, the prediction that R723Q would be a benign substitution by five out of six in silico methods used is consistent with the experimental data reported here in a stable HEK cell line as well as in our earlier study in transiently transfected HEK cells (Létourneau et al., 2005). However, in contrast with our observations, Yin et al. (2009) reported that HEK cells stably expressing R723Q exhibited varying degrees of reduced resistance to natural product drugs and that the plasma membrane trafficking of the R723Q protein was mildly impaired. The reason for the apparent discrepancy is not clear. It is of interest that in a small phase III study of Caucasian patients with advanced multiple myeloma, the presence of the R723Q nsSNP in 5 of 279 patients being treated with bortezomib and pegylated liposomal doxorubicin was associated with a different overall response rate and durability (Buda et al., 2010). Although bortezomib appears to be a P-glycoprotein substrate (O’Connor et al., 2013), there is no convincing evidence that this proteasome inhibitor is transported by MRP1.
The third and final category consists of A989T and C1047S which, despite the fact that most predictive methods indicated they were unlikely to affect MRP1 activity, exhibited altered phenotypes that distinguished them not only from wild-type MRP1 but to some degree from each other. Thus, HEK cells expressing A989T or C1047S differed in their drug-resistance profiles and their sensitivity to LY465803-mediated inhibition of calcein efflux, as well as their cellular GSH levels as detected in cell-based assays. The C1047S affected both vincristine and etoposide resistance while the A989T mutation affected only vincristine resistance. It is noteworthy that cellular efflux of these two drugs by MRP1 occurs in a GSH-dependent manner (Rappa et al., 1997; Loe et al., 1998; Sakamoto et al., 1999). Inhibition of calcein efflux by the GSH-dependent LY465803 was also reduced in A989T cells but not C1047S cells. In contrast, membrane-based vesicular transport assays indicated that A989T and C1047S did not affect the ability of LY465803 to inhibit MRP1-mediated E217βG uptake. Vesicular transport of GSH was also not affected nor was GSH-stimulated transport of estrone sulfate (results not shown). The latter vesicular transport assays depend on exogenous GSH, which suggests that GSH binding itself is not adversely affected by the mutations. In contrast, the cell-based vincristine and etoposide cytotoxicity assays depend on the coefflux of endogenous GSH, which our GSH accumulation experiments suggest may be impaired. At present, there are no fully satisfactory mechanistic explanations for the differences observed between the cell-based and membrane-based assays.
In conclusion, our studies confirm the continued value of experimentally evaluating ABCC1 nsSNP variants and, together with the results from other investigators, indicate that a broad biochemical and pharmacologic characterization of the corresponding mutant MRP1 proteins may be needed to detect phenotypic consequences of ABCC1 nsSNPs because they may be quite selective. Typically, either a cell-based or membrane-based functional assay is used to measure MRP1 function, but experience with the 2012C>T (G671V) allele suggests that additional nucleic acid-based methods should also be included. Our studies further support the idea that in silico methods for predicting deleterious alleles in ABCC1 (and possibly other membrane transporter genes) are currently limited in their accuracy, so further refinements are needed (Aranyi et al., 2011).
Acknowledgments
The authors thank Kathy Sparks for technical assistance and Maureen Hobbs for help in the preparation of the manuscript and figures.
Authorship Contributions
Participated in research design: Conseil, Cole.
Conducted experiments: Conseil.
Contributed new reagents or analytic tools: Conseil, Cole.
Performed data analysis: Conseil, Cole.
Wrote or contributed to the writing of the manuscript: Conseil, Cole.
Footnotes
- Received August 19, 2013.
- Accepted September 30, 2013.
This work was supported by a grant from the Canadian Institutes of Health [MOP-10519].
↵This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- ABC transporter
- ATP-binding cassette transporter
- BAYu9773
- 6(R)-(4-carboxyphenylthio)-5(S)-hydroxy-7(E),9(E),11(Z),14(Z)-eicosatetraenoic acid
- calcein-AM
- calcein acetoxymethylester
- DMSO
- dimethylsulfoxide
- E217βG
- β-estradiol 17-(β-d-glucuronide)
- FBS
- fetal bovine serum
- G418
- geneticin
- GSH
- glutathione
- HEK
- human embryonic kidney
- LTC4
- leukotriene C4
- LY171883
- 1-[2-hydroxy-3-propyl-4-[4-(1H-tetrazol-5-yl)butoxy]phenyl]-ethanone
- LY465803
- N-[3-(9-chloro-3-methyl-4-oxo-4H-isoxazolo-[4,3-c]quinolin-5-yl)-cyclohexylmethyl]-benzamide
- MRP1
- multidrug resistance protein 1
- MSD
- membrane-spanning domain
- MTT
- 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide
- NBD
- nucleotide-binding domain
- nsSNP
- nonsynonymous single nucleotide polymorphism
- SIFT
- Sorting Tolerant From Intolerant
- TM
- transmembrane
- Copyright © 2013 by The American Society for Pharmacology and Experimental Therapeutics