Single nucleotide polymorphisms result in impaired membrane localization and reduced atpase activity in multidrug transporter ABCG2


  • Shinji Mizuarai,

    1. Banyu Tsukuba Research Institute in collaboration with Merck Research Laboratories, Tsukuba, Ibaraki, Japan
    Search for more papers by this author
  • Naohiko Aozasa,

    1. Banyu Tsukuba Research Institute in collaboration with Merck Research Laboratories, Tsukuba, Ibaraki, Japan
    Search for more papers by this author
  • Hidehito Kotani

    Corresponding author
    1. Banyu Tsukuba Research Institute in collaboration with Merck Research Laboratories, Tsukuba, Ibaraki, Japan
    • Banyu Tsukuba Research Institute in collaboration with Merck Research Laboratories, Tsukuba, Ibaraki 300-2611, Japan
    Search for more papers by this author
    • Fax: +81-29-877-2027


ABCG2/MXR/ABCP1/BCRP is a member of the ATP-binding cassette membrane transporter, which consists of six transmembrane regions and one ATP-binding cassette. The transporter is known to be involved in the efflux of various anticancer compounds such as mitoxantrone, doxorubicin and topoisomerase I inhibitor. In this study, we analyzed the effects of polymorphisms in ABCG2, V12M and Q141K on transporter function. When polarized LLC-PK1 cells were transfected with variant ABCG2, drug-resistance to topoisomerase I inhibitor of cells expressing V12M or Q141K was less than 1/10 that of wild-type ABCG2 transfected cells, and was accompanied by increased drug accumulation and decreased drug efflux in the variant ABCG2-expressing cells. We further elucidated the molecular mechanisms of the transport dysfunction by investigating membrane localization and ATPase activity. Confocal microscopic analysis revealed that apical plasma membrane localization of V12M was disturbed, while the localization of wild-type transporters occurred specifically in the apical plasma membrane of polarized LLC-PK1 cells. Also, ATPase activities measured in the membrane of SF9 cells infected with variant ABCG2 showed that Q141K decreased activity by 1.3 below that of wild-type ABCG2. In addition, kinetic analysis of ATPase activity showed that the Km value in Q141K was 1.4-fold higher than that of wild-type ABCG2. These results indicated that naturally occurring SNPs alter transport functions of ABCG2 transporter and analysis of SNPs in ABCG2 may hold great importance in understanding the response/metabolism of chemotherapy compounds that act as substrates for ABCG2. © 2003 Wiley-Liss, Inc.

ABCG2 (ABCG2/MXR/ABCP1/BCRP) belongs to a subfamily (white) of ABC half transporter, which is expressed in cancer cell lines selected at high concentrations of mitoxantrone1 or verapamil and doxorubicin.2 ABCG2 possesses a half-transporter-structure that consists of 1 transmembrane domain and one ABC region, while other drug-resistant transporters such as MDR and MRP are composed of two transmembrane domains and two ABC regions. The ABCG2 transporter is highly expressed in the placenta and moderately expressed in the liver, small intestine, colon, ovary, kidney and heart. This pattern of tissue distribution suggests that ABCG2 is a xenobiotic clearance transporter that effluxes chemicals from organs to protect them from toxic effects and may also function as a maternal-fetal barrier.3 Recently, reports have indicated that ABCG2 is expressed in a wide variety of pluriopotent stem cells, suggesting that ABCG2 maintains stem cells in a quiescent state4, 5, 6 and mice lacking ABCG2 become sensitive to dietary chlorophyll-derived phototoxin.7

Several groups have reported that ABCG2, like P-gp8 or MRP,9 promotes resistance to anticancer drugs such as mitoxantrone, topotecan, SN-38 and doxorubicin.1, 2, 10, 11 To overcome this drug resistance, ABCG2 inhibitors or camptothecin derivatives with poor affinity for ABCG2 were investigated.12, 13, 14, 15 Previously, we reported that ABCG2 expression also confers resistance to indolocarbazole-topoisomerase I inhibitor.16 Mouse fibroblast LY cells were exposed to indolocarbazole topoisomerase I inhibitor to select a cell line resistant to the anticancer drug. Of 30,000 genes expressions examined by oligonucleotide microarray to investigate the changes between the resistant and parental cells, the gene with the greatest expression change was ABCG2 transporter.

Single nucleotide polymorphisms (SNPs) in drug metabolism and drug transporter genes play significant roles in the differences of patients' responses to medication.17, 18, 19, 20 In ABC transporters, polymorphisms that alter expression or protein function have been reported. Among these, C3435T (Wobble) of MDR1 correlates with intestinal expression level in vivo. Homozygous CC patients expressed lower P-gp in the intestine compared to TT patients, resulting in increased digoxin plasma concentration after orally administered digoxin.21 Another report showed that a naturally occurring mutation of R433S in MRP1 caused increased organic anion transport and decreased doxorubicin resistance.22

Several groups have reported naturally occurring ABCG2 SNPs in various ethnic populations, including Caucasian, Asian and African.23, 24, 25, 26, 27 In those reports, polymorphisms at V12M and Q141K occurred at high frequency in most of the ethnic populations. However, effects of naturally occurring ABCG2 polymorphisms on the drug resistance mechanism of ABCG2 are not well characterized. For example, Q141K polymorphism has been reported to reduce protein expression level compared to wild-type ABCG2 in vitro;24 however, this observation was not confirmed by in vivo analysis using intestinal samples.25 Mutations at R482 have been well characterized by several groups including our previous studies,17, 28, 29, 30 demonstrating that they affect drug resistance and are considered mutations acquired during drug selection and not naturally occurring polymorphisms.

In our study, we confirmed the positions and frequencies of SNPs in ABCG2, and examined whether these polymorphisms influence function of the transporter. We further elucidated the molecular mechanism of these changes by using in vitro cell culture systems. Analysis of 150 Caucasian samples and 30 cancer cell lines uncovered 12 polymorphisms, including novel 5 positions. The previously reported polymorphisms,V12M and Q141K, had high frequencies of 10.3% and 9.0%, respectively. Drug resistance to the ABCG2 substrate, indolocarbazole topoisomerase I inhibitor, was reduced more than 10-fold in polarized cells that expressed variant ABCG2 with either V12M or Q141K compared to wild-type ABCG2. We concluded that the functional impairment of these 2 variants were due to disturbance of apical plasma membrane localization for V12M and reduced ATPase activity for Q141K, indicating ABCG2 gene SNPs may greatly influence resistance to ABCG2 substrate.


ABC, ATP-binding cassette; ABCP1, placenta-specific ABC transporter; BCRP, breast cancer-resistant protein; GFP, green fluorescent protein; in; HRP, horse radish peroxidase; MDR, multidrug resistance protein; MRP, multidrug resistance-associated prote; MXR, mitoxantrone resistance protein; P-gp, P-glycoprotein; PVDF, polyvinylidene difluoride; Sf9 cells, Spodoptera frugiperda ovarian cells; SNPs, single nucleotide polymorphisms.


Cell culture

LLC-PK1 cells were cultured in Medium 199 (Invitrogen, La Jolla, CA) supplemented with 1 mM L-glutamine, penicillin (50 units/ml), streptomycin (50 mg/mL) and 10% (v/v) fetal calf serum. All cultures were incubated at 37°C in a humidified atmosphere of 5% CO2/95% air.

Identification of single nucleotide polymorphisms (SNPs) in ABCG2

Genomic DNA was extracted from the following 30 cancer cell lines with Trizol reagent (Invitrogen): A-427, DLD-1, NCI-H69, HelaS3, PC-13, MKN-45, UM-UC-3, HCT116, PA-1, RT4, MKN1, SK-OV-3, MADH, KATO III, U118, HS746, T24, MSTO-211H, OVCR3, Lu135, Lx-1, SCC25, Cal27, MKN-74, SCaBER, BxPC-3, HeLa, J82, NCI-H187 and ES-2. Human clinical DNA samples extracted from blood were obtained from IMPATH-BCP. Blood samples were obtained from 150 normal healthy Caucasians (75 males and 75 females), ranging in age from 18 to 83 years. The volunteers had no evidence of disease (e.g., HIV, HCV or HBV) or any known risk factors (injection drug abuse, sexual contract with infected persons, high risk sexual behavior etc.). DNA samples were extracted with QIAmp DNA Blood Maxi Kit (Qiagen, Chatsworth, CA) and their quality confirmed by measuring A260/280 ratio within a range of 1.5 and 2.3. SNPs in ABCG2 16 exons and peripheral intron sequences were determined by direct sequencing. First, 16 exons were amplified from genomic DNA by PCR with primers and thermostable DNA polymerase (LA Taq, Takara, Japan). Next, amplified DNA fragments were treated with ExoSAP-IT (Usb corporation) to digest primers and remove unwanted dNTPs. Then, fragments were subjected to cycle sequencing via the dye-terminator method (Dynamic ET Dye Terminator Cycle Sequencing Kit; Amersham, Arlington Heights, IL) using sense primers. After the removal of the dye-terminator by G-50 gel filtration, the sequences and SNPs were determined by a capillary sequencer MegaBACE1000 (Molecular Dynamics, Sunnyvale, CA). The identified SNPs were reconfirmed by sequencing a complementary strand using antisense primers. The utilization of human clinical DNA samples was approved by the Institutional Review Board (IRB).

Establishment of LLC-PK1 cells expressing wild-type and mutant ABCG2

Wild-type ABCG2 and mutated ABCG2 genes (G34A for V12M and C421A for Q141K), the point mutations of which were introduced by PCR mutagenesis, were cloned into HindIII and XhoI sites of pcDNA3.1(+) (Invitrogen). The ABCG2 expression vectors and control vector were transfected to LLC-PK1 cells by the lipofection method (Lipofectamine; Invitrogen). Stable cells were selected with 1,600 μg/ml Geneticin (Invitrogen) for 2 weeks. To determine the expression level of ABCG2, total RNA was extracted from the transfectant cells and HeLa cells with Trizol (Invitrogen), and 7 μg of RNA from each clone was used for Northern hybridization. Full length ABCG2 cDNA (2.2 kb) was labeled with 32P and used as a probe. Several transfectants, which expressed equal amounts of ABCG2 mRNA, were selected to eliminate the variable of expression level.

Drug sensitivity to anticancer drugs

Indolocarbazole topoisomerase inhibitor was synthesized as described previously.31 Mitoxantrone, camptothecin, vincristine and doxorubicin were purchased from Sigma Chemical Co. (St. Louis, MO).Topotecan was purchased from LKT Laboratories, Inc. (St. Paul, MN). The cytotoxicity of anticancer drugs was determined by the sulforhodamine B dye-staining method.32 LLC-PK1 cells were seeded at 1 × 103 per well in a 96-well tissue culture plate. After 24 hr, various concentrations of anticancer drugs were added to the culture medium and incubated for 72 hr. Then, cultured cells were fixed with trichloroacetic acid and stained for 30 min with 0.4% sulforhodamine B dissolved in 1% acetic acid. Unbound dye was removed by 4 washes with 1% acetic acid, and protein-bound dye was extracted with 10 mM unbuffered Tris base. Then, optical density was measured in a plate reader at 564 nm, and IC50 was determined.

Drug Accumulation Assay

Cells were seeded at a density of 2 X 106 in a 25 cm2 culture flask. After 24 hr, growth medium was replaced with opti-MEM (Invitrogen) containing 50 or 0.5 μM [14C]topoisomerase inhibitor. After the drug accumulated for 3 hr, the cells were washed with PBS and treated with trypsin. The cells were collected and solubilized in PBS containing 0.2% Triton X-100. The lysates were centrifuged at 2,000g for 10 min, the supernatants were collected and then the remaining pellets were solubilized in 2 N NaOH to lyse the nuclei. Radioactivity of supernatant fractions and nuclear lysates was determined in a liquid scintillation counter.

Drug efflux assay

LLC-PK1 cells were incubated with the indicated concentration of indolocarbazole compound for 120 min under energy-depleted conditions. After washing the cells 3 times with PBS, they were incubated with growth medium for 10 min. The cell lysates, which were prepared as described in the drug accumulation assay, were collected and radioactivity counted to calculate the initial velocity of drug efflux. The membrane vesicle transport assay of indolocarbazole compound was performed as previously reported.33


Cells grown on tissue culture chamber slides (Lab TekII Chamber Slide; Nalge Nunc International, Rochester, NY) were fixed in acetone for 8 min. The slides were then incubated with PBS containing 5% goat serum for 30 min, followed by incubation with a 1:30 dilution of a monoclonal antibody BXP-34 (Kamiya Biochemicals, Thousand Oaks, CA) for 60 min. For detection of ABCG2 to establish stable LLC-PK1 clones, biotinylated goat-antimouse IgG (1:200 dilution) and HRP-conjugated streptavidin were used as secondary antibodies. Color development was achieved using 0.4 mg/ml AEC. Counterstaining was performed with hematoxylin. For detection of ABCG2-subcellular localization, goat-antimouse IgG (1:200 dilution) and FITC-conjugated streptavidin were used as secondary antibodies. Nuclei or F-actin were stained with propidium iodide or F-actin-specific probe (Alexa Fluor 568 phalloidin; Molecular Probes, Eugene, OR) according to the manufacturer's instructions. Then, confocal microscopic analysis was performed using an MRC-1024 laser-scanning confocal microscope (Bio-Rad, Richmond, CA). Cross-sections were generated with a 0.4 μm motor step.

ATPase activity measurement

Recombinant baculoviruses were generated with a BakPAK expression system (ClonTech, Palo Alto, CA) according to the manufacturer's instruction. Wild-type and variant ABCG2 cDNA fragments were cloned into the EcoRI and XhoI sits of transfer vector pBacPAK9, and the sequences were confirmed. Sf9 cells were incubated in BacPAK Complete Medium (ClonTech) at 27°C for the experiments. Recombinant baculoviruses were produced by cotransfecting Sf9 cells with the respective transfer vector and BacPAK6 viral DNA. After amplifying the recombinant viruses, an approximate titer of 108 cfu/ml virus was used for scale-up protein production (1.5 × 107 cells, moi=100). Sf9 cells that expressed wild-type and variant ABCG2 transporters were harvested 4 days after virus infection, and membrane fractions were prepared as described previously.34 ATPase activity of ABCG2 was measured by determining the level of sodium vanadate-sensitive release of inorganic phosphate from ATP with colorimetry.34 Membrane proteins (5 μg) were first incubated at 37°C in Tris-HCl (pH 7.5), 2 mM EGTA, 50 mM KCl, 2 mM dithiothreitol, 5 mM Na-azide and 1 mM ouabain. The reactions (a total volume of 50 μl) were started by adding 5 mM ATP to the assay mixtures. The mixtures were incubated for 20 min at 37°C. Vanadate-sensitive ATPase activities were calculated as the difference between the activities obtained in the absence and presence of 300 μM vanadate.

Western blotting

Membrane proteins prepared from the infected cells were utilized for Western blotting. Proteins (9 μg/lane) were separated on a 7.5% polyacrylamide gel and then transferred on a PVDF membrane. Proteins were hybridized with a 1:125 dilution of monoclonal antibody BXP-21 (Kamiya Biochemicals) and HRP-conjugated goat antimouse IgG (1:40,000). Subsequently, ABCG2 was visualized with ECL-Plus (Amersham, the Netherlands).

Dot blotting

The expression level of ABCG2 mRNA was analyzed using commercially available dot blotting membrane (Cancer Profiling Array; ClonTech) according to the manufacturer's instructions. Full length ABCG2 cDNA (2.2 kb) was labeled with 32P for use as a probe. Subsequently, an ubiquitin probe was hybridized and confirmed the detection of nearly equal signal levels in each dot.


Determination of SNPs and their frequencies in ABCG2 gene

Positions and frequencies of SNPs in ABCG2 exons were determined in 30 cancer cell lines and 150 clinical samples from Caucasian patients. DNA sequencing was performed on 16 exons and adjacent intronic regions. We detected 9 sequence variations in the exon sequences and 3 in the intron regions (Table I) when the sequences were compared to the original ABCG2 GenBank entry sequence (gb; AB051855). Two polymorphisms, V12M and Q141K, had high frequency rates of 10.3% and 9.0%, respectively, in the 150 Caucasian subjects. V12M was located at the N-terminal intracellular region and Q141K at the ATP-binding cassette region. Another variant, C376T, which was only detected in 1 case in cancer cell lines (MKN45), was detected as a termination mutation in the ABC region. Although several groups have reported the positions and frequencies of polymorphisms in ABCG2, 5 of the detected 12 polymorphisms were novel: C458T accompanying the amino acid substitution of T153M, C369T and C474T with silent mutations, and G2237T and G2393T in 3′UTR of mRNA.

Table I. Single Nucleotide Polymorphisms in ABCG21
SNPEffectRegionDomainFrequency in 30 cell linesFrequency in 150 clinical samples
HeteroHomeHeteroHomoAllele (%)
  1. The positions of the polymorphisms correspond to that of the ABCG2 cDNA (GenBank accession number AB051855) with the first base of the ATG start codon set to 1. TM, transmembrane; EC, extracellular; UTR, untranslated region; ND, not determined.

G34AV12MExon2N-terminal5 (16.7%)027 (18.0%)2 (1.3%)10.3
A+10G Intron3 NDND21 (14.0%)4 (2.7%)9.7
C369TWobbleExon4ABC001 (0.67%)00.3
C376TQ126TermExon4ABC01 (3.3%)000.0
C421AQ141KExon5ABC5 (16.7%)1 (3.3%)22 (15.3%)2 (1.3%)9.0
C458TT153MExon5ABC1 (3.3%)0000.0
C474TWobbleExon5ABC001 (0.67%)00.3
A+20G Intron11 NDND34 (22.7%)10 (6.7%)18.0
C-21T Intron13 NDND32 (21.3%)4 (2.7%)13.3
A1768TN590YExon15EC3001 (0.67%)00.3
G2237T Exon163′UTR1 (3.3%)0000.0
G2393T Exon163′UTR1 (3.3%)0000.0

Resistance profile of variant ABCG2 transporters to anticancer compounds

Among the polymorphisms detected, V12M and Q141K had a high frequency of amino acid substitutions. Therefore, we decided to analyze the effects of the mutations on drug-resistance profiles with established stable cell lines expressing wild-type and variant ABCG2. Porcine kidney LLC-PK1 cells were transfected with the ABCG2 expression vectors, and stable transfectants were selected. Northern blotting and immunocytochemical analysis with a monoclonal antibody revealed that approximately equal levels of ABCG2 were expressed in V12M and Q141K clones compared to the expression in the wild-type (WT) clone (Fig. 1). To assess the effects of these ABCG2 polymorphisms on drug transport activities, we used a surrogate drug resistance assay that tests drug sensitivity to anticancer compounds by sulforhodamine B assay (Table II). Wild-type cells conferred greater than 420-fold higher resistance to an indolocarbazole I topoisomerase inhibitor compared to that of control C4 cells as previously reported.10 In contrast, IC50 values of variant ABCG2 clones, V12M and Q141K, were less than 1/10 that of WT cells. These results were confirmed using mitoxantrone and topotecan, both of which are ABCG2 substrates. The IC50 values of WT cells to mitoxantrone and topotecan increased 19- and 22-fold, respectively, over C4 cells, whereas in the variant V12M and Q141K cells, IC50 values to mitoxantrone and topotecan decreased as observed with the indolocarbazole compound. Two anticancer drugs, camptothecin and vincristine, which are not ABCG2-substrates, had similar drug resistance values for the wild-type and variants. Similar drug resistant profile results were obtained using another clone set expressing almost equal amounts of recombinant protein (data not shown); however, we could not totally rule out possible effect of using cloned cell lines.

Figure 1.

Establishment of ABCG2-transfected LLC-PK1 cells. (A) Northern blot analysis of ABCG2-transfected LLC-PK1 cells. Samples of 7 μg RNA extracted from HeLa, control C4, WT, V12M and Q141K cells (lanes 1, 2, 3, 4 and 5 respectively) were subjected to Northern hybridization and the blots were probed with a cDNA fragment of ABCG2. The same blots reprobed with GAPDH. (B) Immunostaining of ABCG2-transfected LLC-PK1 cells. Vector transformant C4 (a) and stable clones expressing wild-type (b), V12M (c) and Q141K (d) were stained with monoclonal antibody BXP-34. Counter staining was performed with hematoxylin. (C), Quantification of ABCG2 protein expression. Integrated optical densities of stained ABCG2 in each transfectant were quantified by densitometer (normalized to WT=1.0).

Table II. Resistant Profile (IC50) of ABCG2 to Anti-Cancer Compounds
Anti-cancer compoundIC50 (μM)1
Control C4Wild typeV12MQ141K
  • 1

    Relative resistances to control cells are described in parentheses.

  • 2

    TopoI inhibitor, Indolocarbazole topoisomerase I inhibitor.

TopoI inhibitor20.12>50 (420)0.94 (7.8)5.9 (49)
Mitoxantrone0.00150.029 (19)0.00093 (0.62)0.0053 (3.5)
Topotecan0.0982.1 (22)0.16 (1.7)0.48 (4.9)
Doxorubicin0.0100.039 (3.9)0.0073 (0.73)0.014 (1.4)
Vincristine0.00340.0053 (1.6)0.0021 (0.62)0.0058 (1.7)
Camptothecin0.00870.021 (2.4)0.012 (1.4)0.027 (3.1)

Accumulation and efflux assay of topoisomerase I inhibitor in ABCG2 variant-expressing cells

Increased sensitivities of V12M and Q141K cells are thought to arise from changes in the intracellular drug concentration of topoisomerase I inhibitor. A drug accumulation assay was conducted using [14C] radio-labeled indolocarbazole topoisomerase I inhibitor.17 Cells were incubated for 3 hr with 0.5 or 50 μM indolocarbazole topoisomerase I inhibitor. Next, the cells were washed, and the intracellular concentration of compound was measured (Fig. 2A,B). WT cells exhibited a 4.5-fold lower accumulation compared to C4 cells. However, significantly higher indolocarbazole topoisomerase I inhibitor accumulation was observed in both V12M and Q141K cells compared to that in WT cells (2.7- and 1.8-fold higher, respectively). A drug accumulation assay performed at a low dose (0.5 μM) of the compound confirmed that compound accumulations increased in variant cell lines (3.1- and 2.8-fold increase for V12M and Q141K, respectively). This accumulation level was consistent with the rate of decreased drug resistance in variant-expressing cells.

Figure 2.

Accumulation and efflux rate of indolocarbazole compound in ABCG2-transfected cell lines. (A,B) Drug accumulation assay. LLC-PK1 cells expressing WT, V12M and Q141K were incubated with opti-MEM with 50 μM (A) or 0.5 μM (B) of radiolabeled topoisomerase inhibitor for 180 min. Accumulated drug in the cells was determined by scintillation counter. (C,D) Drug efflux assay. The indicated cells were incubated with radiolabeled topoisomerase inhibitor at various concentrations under energy-depleted conditions. After changing the drug solution for growth medium, the drug efflux velocity was measured. The Michaelis-Menten plot (C) and Line weaver-Burk plot (D) are shown. (E,F) Vesicle transport assay. Indolocarbazole compound uptake by membrane vesicles prepared from LLC-PK1 cells expressing wild-type or variant ABCG2 was measured at various indolocarbazole compound concentrations for 5 min. Michaelis-Menten plots C,E; line weaver-Burk plots, D,F. Data are expressed as mean ± SE from 3 independent experiments. *, **p < 0.05, 0.01 vs. WT.

To determine whether increases in drug accumulation in the variant cells were associated with a decrease in drug efflux rate by the transporter, kinetic analyses of drug efflux were performed. LLC-PK1 cells expressing WT and variant ABCG2 were incubated with [14C] indolocarbazole compound at various concentrations under energy-depleted conditions. After removal of the radio-labeled compounds, initial velocity of drug efflux was measured. Figure 2C,D shows the Michaelis-Menten plot and Line weaver-Burk plot of drug efflux velocity, respectively, in AGCG2-expressing cells as a function of indolocarbazole compound concentration. Compared to WT cells, Vmax of V12M and Q141K cells decreased 2.5- and 1.8-fold, respectively, indicating that the increased drug accumulation and consequent reduction in drug resistance were due to the decreased efflux velocity. Km values were not significantly different among the clones. Rates of drug efflux in variant ABCG2 transporters also were analyzed by membrane vesicle transport assay (Fig. 2E,F). Membrane vesicles purified from the stable cell lines were incubated with indolocarbazole compound, and compound uptake was measured. Vmax values of V12M and Q141K decreased 2.2- and 1.7-fold, respectively, which agrees well with the results of drug efflux assays using cell lines.

Subcellular localization of wild-type and variant ABCG2

Experimental data described above indicate that the V12M polymorphism impaired function of the transporter, leading to increased drug accumulation and subsequent decreased drug resistance to anticancer compounds. To decipher the molecular mechanism underlying this reduced transporter activity, we examined the subcellular localization of variant ABCG2 transporters by confocal laser-scanning microscopic analysis. ABCG2 stable clones were grown to confluency on chamber slides and immunocytostaining was performed with the anti-ABCG2 antibody BXP-34.35 We collected a series of pictures from the top of the cells to the glass slide (0.4 μm dissections) to construct 3-dimensional pictures of the cells. X-Z cross sections of each stable clone are shown in Figure 3. In WT cells, ABCG2 protein was localized to the apical cytoplasmic membrane, judging from the nuclear positions stained with propidium iodide (Fig. 3b), while no ABCG2 staining was observed in the C4 clone control (Fig. 3a). Despite speculation that ABCG2 protein localizes to the apical side in LLC-PK1 cells because LLC-PK1 cells expressing mouse Bcrp1 transport topotecan from the basal to the apical side,36 this is the first direct evidence that shows apical membrane localization of ABCG2 in LLC-PK1 cells. In the V12M clone, however, ABCG2 staining was not specifically localized to the apical membrane and showed sparse staining in all basolateral, apical and cytosolic stains (Fig. 3c). In Q141K cells, ABCG2 staining was detected in the apical side, similar to WT cells (Fig. 3d). These data indicate that the polymorphism at V12M impairs the specific apical membrane localization of this transporter.

Figure 3.

Immunocytochemical localization of variant ABCG2 transporter in polarized LLC-PK1 cells. Staining of cells with ABCG2 antibody and propidium iodide (AD). Cells were stained with monoclonal antibody BXP-34 recognizing ABCG2. Other antibodies included biotinylated goat antimouse IgG and streptavidin conjugated FITC. Counter staining was performed with propidium iodide. Staining of cells with ABCG2 and F-actin (EG). After staining with ABCG2 antibody as described above, the cells were stained with F-actin specific probe, Alexa Fluor 568 phalloidin. The samples were examined by laser-scanning confocal microscopy, and X-Z cross sections shown. (A) Control C4 cells; (B) and E, WT cells; (C,F), V12M cells; (D,G), Q141K cells.

Impairment of ABCG2 localization in V12M also was confirmed by double staining of cells with ABCG2 and F-actin (Fig. 3EG). F-actin was used to outline LLC-PK1 cells to understand the position of ABCG2 expression relative to the cell membrane. After staining, confocal images were collected in the same manner as ABCG2/nuclei staining. In WT and Q141K cells, apical membrane staining of ABCG2 was detected (Fig. 3E,G). In V12M cells, specific apical localization was disturbed, and the staining of cytosolic and other membrane regions was observed (Fig. 3F). To further confirm disturbed membrane localization of ABCG2 in V12M stable cell lines, cells co-stained with E-cadherin (localized to the lateral-membrane) also were analyzed with confocal microscopy. The expected pattern of apical staining with anti-ABCG2 antibody was observed in wild-type and Q141K variant expressing cell lines, whereas dispersed staining was confirmed in V12M cells (data not shown). Presence of substrate did not restore the localization of V12M variant (data not shown).

ATPase activity of variant ABCG2

Since Q141K was mapped in the ATP-binding cassette region, ATPase activity of the variant transporters was measured. The baculovirus-expression system in Sf9 cells was used for the production of ABCG2 transporters, and the ATPase activity of the transporters was measured with the membrane fraction prepared from virus-infected Sf9 cells. The baculovirus titer was adjusted to yield approximately equal levels of ABCG2 expression in all 3 forms, and protein levels were confirmed by Western blotting after 2 and 4 days of infection (Fig. 4A) with loading control adjusted to total protein at 4 days (Fig. 4A, bottom). Vanadate-sensitive ATPase activities of ABCG2 transporters in the presence of 5 mM of ATP are shown in Figure 4B. As previously reported, expression of wild-type ABCG2 increased ATPase activity compared to the membrane prepared from vector virus-infected control cells. However, the ATPase activity of Q141K was reduced by about 1.3-fold compared to that of wild-type ABCG2. In the case of V12M, no significant difference in ATPase activity was observed. Since the ATPase activity of other ABC transporters is stimulated by their substrates, we investigated the effect of substrate on ATPase activity of ABCG2 transporters (Fig. 4C). In all WT- and variant ABCG2 transporters, no stimulation of ATPase activity by the indolocarbazole compound was observed. Other known substrates of ABCG2, such as mitoxantrone and topotecan, did not stimulate ATPase activity (data not shown). We concluded that Q141K and V12M polymorphisms did not restore drug stimulation, which was observed in R482 mutant. Figure 4D,E shows data obtained from further kinetic analysis of ATPase activity as a function of ATP concentration in ABCG2-expressing membranes. In the Q141K membrane, the Vmax value of ATPase activity decreased 1.3-fold in accordance with the values of Figure 4A. Unexpectedly, the Km value of Q141K also increased 1.4-fold compared to that of WT (0.64 mM; WT, 0.64 mM; V12M, 0.91 mM; Q141K). In the V12M membrane, the ATPase activity increased in almost same manner as in the WT membrane. No significant difference in either Vmax or Km between WT and V12M membranes was found, indicating that V12M does not affect any changes in ATPase activity.

Figure 4.

Vanadate-sensitive ATPase activity of variant ABCG2. (A) Western blotting analysis of ABCG2 expressed in Sf9 cells. Membrane fractions isolated from Sf9 cells expressing wild-type, V12M and Q141K ABCG2 transporters were subjected to Western blotting with BXP-21 monoclonal antibody. Integrated optical density of the bands is shown under the blots (normalized to WT=1.00). Duplicate samples obtained 4 days after transfection also were subjected to SDS-PAGE analysis and CBB staining as an internal control. (B) ATPase activity of Sf9 membrane containing ABCG2 transporters. The membrane fractions were incubated with 5.0 mM ATP, and inorganic phosphate liberation was determined by colorimetry. (C) Effect of indolocarbazole concentration on ATPase activity of ABCG2. Various concentrations of indolocarbazole compound were added to the reaction mixture and ATPase activity was measured as described in B. (D,E) Effect of ATP concentration on ATPase activity of ABCG2. Various concentrations of MgATP were added to the reaction mixture and ATPase activity was measured as described in B. Michaelis-Menten plot, D; line weaver-Burk plot, E. Data are expressed as mean ± SE from 3 independent experiments. *, **p < 0.05, 0.01 vs. WT.


In our study, we confirmed the locations and frequencies of SNPs in ABCG2 using 30 cancer cell lines and 150 Caucasian clinical samples, and then characterized the functional effects of the major SNPs, V12M and Q141K (Fig. 5). These polymorphisms exhibited reduced drug resistance activity in polarized LLC-PK1 cells that accompanied increased drug accumulation. Furthermore, we were able to show these dysfunctions were directly coupled to the impaired function of ABCG2, membrane localization of transporter for V12M and ATPase activity for Q141K.

Figure 5.

Schematic diagram of SNPs in ABCG2. Rectangles indicate the predicted transmembrane helices. Three identified variants, which affect transporter function, were designated as V12M (1), Q126Term (2) and Q141K (3).

The polymorphism at V12M impaired apical membrane localization of ABCG2 in polarized LLC-PK1 cells. Although this region is not recognized as a signal sequence for membrane localization in ABC transporters, several studies have suggested that the terminal regions of the ABC transporter influenced membrane localization efficiency. For example, a 15 C-terminal amino acid truncation of the GFP-MRP2 fusion protein changed apical localization of MRP2, indicating that the C-terminal intracellular region determines the subcellular localization.37 Since ABCG2 is a half transporter, the ATP-binding domain and transmembrane segments are oriented opposite to those of full transmembrane ABC transporters such as MDR1 or MRP. Therefore, it is quite likely that the N-terminal intracellular region of ABCG2 corresponds to the C-terminal intracellular region of MRP2. Our observation of the deficiency of apical membrane localization of V12M implies that the N-terminal intracellular region may be critical for apical membrane localization of ABCG2 protein. It would be interesting to study the effect of V12M on vectorial transport because the transporters are expressed in polarized LLC-PK1 cells.

The other variant, Q141K, also reduced drug resistance against previously known ABCG2 substrates when it was expressed in LLC-PK1 cells. The mutation locus indicated that the reduction in drug resistance was due to the effect on the ATPase activity of the ABC domain. Several mutagenesis studies on ABC transporters have shown that introduced mutations in the ABC domain completely disrupted ATPase activity.30, 38, 39 For example, an induced mutation in ABCG2 prevented K86M from transporting ABCG2 substrates due to the loss of ATPase activity.30 When measured in the Sf9 membrane containing ABCG2 transporters, ATPase activity of Q141K was 1.3-fold lower than that of wild-type ABCG2, although the effect on ATPase activity was relatively moderate compared to that of the introduced mutation at K86M. While K86M was located at the walker catalytic region in the ABC domain, Q141K was located between the walkerA and walkerB regions. In addition to the changes in ATPase activity of Q141K, the Km value of Q141K was different from that of WT by a factor of 1.4. The change in Km value infers that the difference in ATPase activity of Q141K and WT might be larger than 1.3-fold at the physiological concentration of ATP.

Previously, the effects of Q141K on protein and mRNA expression were investigated in an ABCG2-overexpressing murine cell line24 or human clinical intestinal samples.25 The initial report showed that murine PA317 cells expressing Q141K increased intracellular drug accumulation compared to cells expressing WT ABCG2, coupled with reduced protein levels with similar mRNA expression, suggesting that the translation efficiency of ABCG2 was the cause of this phenomenon. In contrast, a second study involving 32 human intestinal samples demonstrated no correlation between the amount of ABCG2 protein and Q141K polymorphism. Our study, using polarized LLC-PK1, agreed with the latter study in that Q141K did not affect ABCG2 protein expression; instead, it showed impaired Q141K transporting function. We also confirmed this observation using Sf9 membranes containing a similar amount of ABCG2 protein, in which WT- and Q141K-expressing cells exhibited altered ATPase activity. Differences between our observation and previous reports may arise from differences in cell lines because regulation of protein expression differs among organisms, tissues and cell lines. Since V12M affected apical membrane localization of ABCG2, the variant V12M could have a more significant effect on polarized LLC-PK1 cells than on PA317 cells that are not polarized. Since the ABCG2 protein localizes to the apical side of cells in several normal tissues, such as intestinal villi or colon glandular epithelium,40 we believe that the localization of ABCG2 in LLC-PK1 may mimic the physiological state of the transporter more accurately.

Small changes in drug accumulation are known to result in large changes in drug resistant profiles from transporter studies. For example, the increase in vincristine accumulation in MRP1 transfected cells was approximately 3-fold, while the resistance value increased 30-fold.41 In our previous study using an indolocarbazole compound and ABCG2 transporter, a 2-fold reduction of drug accumulation resulted in a 12-fold increase in drug resistance.17 These relationships also were observed in nontransfected variants of cell lines, in which a 0.7-fold difference in mitoxantrone accumulation between human 8226 cell line and its derivative caused a 37-fold difference in mitoxantrone resistance.42 These observations agree well with those of our present study in that a 2- to 3-fold increase in drug accumulation values resulted in a 10-fold decrease in drug resistance.

In our study, we found 5 novel polymorphisms, C458T accompanying the amino acid substitution of T153M in the ABC region: C369T and C474T in the ABC region with silent mutations and G2237T and G2393T in 3′UTR of mRNA. Three of the mutations were detected only in cancer cell lines: C458T in HCT116, G2393T in BxPC and G2393T in MADH. We cannot conclude that the 3 polymorphisms detected only in cell lines exist in naturally occurring polymorphisms or that they are acquired mutations after the cell lines were established. Since SNPs are defined as the mutations with a frequency greater than 1%, we did not analyze the 5 novel polymorphisms with a lower frequency, as shown in Table I. However, it might be of interest to analyze the effects of the polymorphisms on expression level or transport function of ABCG2.

ABCG2 protein is involved in cellular resistance mechanisms of several anticancer compounds.1, 2, 10, 11 Even though the in vivo role of this transporter activity in the resistance mechanism against chemotherapy agents remains to be investigated, our results imply that genetic variations in ABCG2 must be considered together with expression levels of ABCG2. We have conducted expression analysis of ABCG2 using 241 cancer and corresponding normal tissue samples, which showed that ABCG2 is generally suppressed in most types of cancer tissues (Fig. 6). This observation is unique to the ABCG2 gene among other drug-resistant ABC transporters. MRP1, for example, was downregulated in bronchial epithelium cancer tissues, upregulated in lung and esophagus cancer tissues and not changed in breast, colorectal or kidney cancer.43, 44, 45, 46 Hence, information on the SNPs in ABCG2 as well as its expression levels of ABCG2 would be necessary to understand the mechanism of cancer cell resistance.

Figure 6.

Dot blotting analysis of ABCG2 in normal and tumor tissues. In total, 241 samples of normal (N) and corresponding tumor (T) tissues were analyzed. The name of the tissue is indicated above each sample.

Data presented here suggest that 2 polymorphisms in the multidrug-resistant transporter ABCG2 alter function through changes in membrane localization or ATPase activity. However, several more SNPs in the controlling region of the ABCG2 gene, both in intron and promoter regions (data not shown), possess functions related to the ABCG2 proteins that are unknown. Investigation of the SNPs and their roles in the expression/function of ABCG2 would provide a comprehensive picture of the genetic regulation of this protein, and may be required for understanding pharmacological activities and pharmacokinetic profiles of ABCG2 transporting-anticancer compounds.