Breast cancer resistance protein (BCRP), also called mitoxantrone-resistant protein, is the second member of the G family of ATP-binding cassette transporters (ABCG2) (Allikmets et al., 1998; Doyle et al., 1998; Miyake et al., 1999; Doyle and Ross, 2003). The BCRP gene is located at 4q22 and encodes a 72-kDa membrane protein composed of 655 amino acids (Allikmets et al., 1998; Doyle et al., 1998; Allen et al., 1999; Bailey-Dell et al., 2001). In contrast to many other ABC transporters, BCRP has only one ATP-binding region and one transmembrane domain. Therefore, BCRP is referred to as a half-transporter, and its homodimerization may be necessary to transport substrates (Kage et al., 2002).

In normal human tissues, BCRP is highly expressed in the placenta, colon, small intestine, and liver (Maliepaard et al., 2001). On the basis of its tissue distribution and findings in knockout mice, BCRP is speculated to have a major influence on the pharmacokinetic and pharmacodynamic profiles of certain xenobiotics and endogenous substrates. For example, inhibition of mouse Bcrp 1 by GF120918, a dual inhibitor for BCRP and P-glycoprotein, has been demonstrated to increase the bioavailability of topotecan when GF120918 was administered orally to mdr1a/1b(-/-) mice (Jonker et al., 2000). In a clinical study, coadministration of GF120918 was also associated with a marked increase in the bioavailability of and systemic exposure to topotecan (Kruijtzer et al., 2002).

Recent clinical studies indicate that the large interindividual variability in drug response occurs as a result of molecular alterations to various proteins such as drug-metabolizing enzymes, drug targets and receptors, and drug transporters. Most studies on molecular alterations have focused on the impact of single-nucleotide polymorphisms (SNPs) on the expression and function of these proteins (Evans and Relling, 1999; Evans and Johnson, 2001). Several groups have reported naturally occurring SNPs in the BCRP gene. G34A and C421A occur at relatively high frequency in most ethnic populations (Imai et al., 2002; Bäckström et al., 2003; Zamber et al., 2003; Mizuarai et al., 2004). Although effects of these SNPs on the drug resistance mechanism have not been well documented, the mutation C421A has been reported to reduce BCRP protein levels in PA317 cells (Imai et al., 2002). In contrast, no significant difference was observed in human intestinal samples (Zamber et al., 2003). Since BCRP may function as a maternal-fetal barrier, effects of these SNPs on expression levels in the placenta are of interest.

Recently, it has been suggested that various gene expression mechanisms, such as genomic imprinting (Constāncia et al., 1998; Brannan and Bartolomei, 1999) and expression imbalance (Yan et al., 2002; Bray et al., 2003), contribute to human diversity. Genomic imprinting is an epigenetic modification leading to a functional inequality of paternal and maternal genomes in somatic cells (McGrath and Solter, 1984; Surani et al., 1984). In addition to epigenetic modifications, cis-acting SNP(s) has been reported to be a causative factor for mRNA expression imbalance. In that case, the cis-acting SNP(s) alters the expression of the gene transcript from the chromosome carrying it, resulting in an imbalance of expression between the two alleles. For example, Wojnowski and Brockmöller (2004) have recently indicated a hepatic transcriptional imbalance of the CYP3A5 gene in CYP3A5^*1A/^*3 heterozygous samples. In the CYP3A5 gene, cis-acting ^*1A variant, which increases the expression of the CYP3A5 gene transcript from the allele carrying the variant, is a possible mechanism for the imbalance. Thus, allelic expression imbalance is useful for assessing the functional characterization of SNPs.

The initial aim of this study was to determine polymorphisms of the BCRP gene using 100 Japanese placentas and 420 genomic DNA samples obtained from Japanese, Caucasian, and African-American populations and then to compare the allelic frequency among these three populations. The second and major aim of this study was to evaluate the substantial variability in placental BCRP expression from two perspectives: genetic polymorphisms and expression imbalance.

Materials and Methods

Placentas and Lymphoblast Samples. Human full-term placentas (highly enriched placental trophoblast populations) were obtained from 100 Japanese patients at Tottori University Hospital. These tissues were immediately frozen in liquid nitrogen and stored at -80°C for the preparation of DNA and RNA (Tanabe et al., 2001). We also obtained 18 lymphoblast samples for which the parental origin of the BCRP alleles was determined (Mitsuya et al., 1997). EB virus-transformed lymphoblast cultures were obtained using standard procedures. This study was approved by the Ethical Board of the Faculty of Medicine, Tottori University, and informed consent was obtained from all individuals.

RNA Extraction and cDNA Synthesis. Total RNA extraction and RT-PCR procedures for placental samples were previously described (Tanabe et al., 2001). For EB virus-transformed lymphoblasts, the RNeasy Kit (QIAGEN GmbH, Hilden, Germany) was used to extract total RNA. Prior to RT, total RNA samples were first treated with RNase-free DNase I and digested with BsrI (New England Biolabs, Beverly, MA). BsrI digests the potential DNA template, which would lead to the amplification of both alleles and thus mask allelic imbalance. RT from total RNA was performed in a 20-μl reaction mixture containing 5 μg of total RNA in 1× First-strand Buffer, 25 mM DTT, 0.5 μg of the random primers (Promega, Madison, WI), a 2 mM concentration of each deoxynucleoside-5′-triphosphate, and SuperScript II RNase H^- reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were incubated at 42°C for 1 h. As a negative control, template RNA was processed without reverse transcriptase.

Screening of Variants in the BCRP Gene (PCR-SSCP). Genomic DNA was isolated from placental samples using the QIAamp DNA Mini Kit (QIAGEN). The primer design was based on published sequences (GenBank accession number AC084732) of BCRP to avoid the amplification of sequences from homologous genes. PCR was carried out in a total volume of 25 μl in the presence of 50 ng of DNA, a 0.25 μM concentration of each of the forward and reverse primers (totally, 25 sets of primers were designed and presented in the Appendix), 1× PCR buffer II, 1.5 mM MgCl₂, a 0.2 mM concentration of each deoxynucleoside-5′-triphosphate, and 0.625 U of AmpliTaq Gold DNA polymerase (Applied Biosystems, Foster City, CA). After an initial denaturation at 94°C for 5 min, 40 to 45 cycles of 0.5 to 1 min at 94°C, 0.5 to 1 min at 50-68°C, and 1 min at 72°C, as well as a final extension period of 5 min at 72°C, were carried out. PCR products were analyzed on 3% agarose gels to check both the size and specificity of the products. To screen variants of the BCRP gene, SSCP analysis was performed. The details of the procedures were described elsewhere (Tanabe et al., 2001).

DNA Sequence. All PCR products were sequenced either directly or after subcloning on an ABI 3100 automatic sequencer (Applied Biosystems) using a Big-Dye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems). If the direct sequencing was incomplete, each amplified PCR product was subcloned into the vector pGEM T-easy (Promega) and transformed into competent JM109 cells (Promega). Before the sequencing, reaction mixtures were purified with Sephadex G-50 (Amersham Biosciences AB, Uppsala, Sweden). The sequencing primers were those used in the PCR amplifications. The sequences of both strands were analyzed for products from at least two independent PCR amplifications to ensure that the identified mutations were not PCR-based artifacts.

Genotyping for BCRP in Three Ethnic Populations. We examined the allelic frequencies of the G34A, C376T, and C421A mutations using genomic DNA samples from unrelated Japanese (n = 120), Caucasian (n = 150), and African-American (n = 150) volunteers. Genomic DNA was isolated from blood samples with use of the Toyobo blood kit on a Toyobo HMX-2000 robot (Toyobo, Osaka, Japan). The three polymorphisms were genotyped by the PCR-RFLP method. The PCR conditions were the same as for PCR-SSCP, but mismatch primers were designed for the genotyping of G34A and C421A. Primer sequences were as follows: BseMI RFLP for G34A, forward, 5′-CAGTAATGTCGAAGTTTTTATCGCA-3′ and reverse, 5′-AAATGTTCATAGCCAGTTTCTTGGA-3′; AfaI RFLP for C376T, forward, 5′-ATAGCATGTGTTGGAGGGAAAAA-3′ and reverse, 5′-ATTGGTATCACTGTCCTTACAAG-3′; TaaI RFLP for C421A, forward, 5′-GTTGTGATGGGCACTCTGATGGT-3′ and reverse, 5′-CAAGCCACTTTTCTCATTGTT-3′. After digestion with an appropriate restriction enzyme, all RFLP products were analyzed on a 3% agarose gel.

Haplotype Analysis. In addition to the unphased SNP analysis, we performed a haplotype analysis for the three major polymorphisms, G34A, C376T, and C421A, using the following two approaches. 1) Haplotypes in individuals who were homozygous at all SNP sites or heterozygous at no more than one of the variable sites were assigned directly from the result of the unphased SNP analysis; 2) haplotypes in the remaining multiheterozygous subjects were determined by a combination of allele-specific PCR and RFLP methods. For the allele-specific PCR, the BseMI RFLP forward primer and the allele-specific reverse primers for the C421 allele or A421 allele were used: the reverse primer for the C421 allele was 5′-GAAGAGCTGCTGAGAACTG-3′ and for the A421 allele was 5′-GAAGAGCTGCTGAGAACTT-3′. All allele-specific PCR products were digested with both BseMI (G34A) and AfaI (C376T).

Quantitative Real-Time PCR Analysis. The mRNA levels were measured by TaqMan quantitative real-time PCR with an ABI PRISM 7000 sequence detection system (Applied Biosystems). The following primers and TaqMan probe were used for determining the BCRP mRNA: forward primer, 5′-TTCTGCCCAGGACTCAATGC-3′; reverse, 5′-GCCACGTGATTCTTCCACAA-3′; TaqMan probe, 5′-FAM-CCAAATATTCTTCGCCAGTACATGTTGC-TAMRA-3′. The endogenous reference gene was determined using the commercially available human GAPDH TaqMan PreDevelopped Assay Reagent (Applied Biosystems). The quantitative PCR was carried out in a total volume of 25 μl in the presence of 1 μl of cDNA, 200 nM each of the forward and reverse primers, 100 nM of probe and 1×TaqMan universal PCR master mix (Applied Biosystems).

Assessment of Expression Imbalance. To assess the allelic imbalance of BCRP, the BseMI RFLP (G34A, for lymphoblast and placental samples) and TaaI RFLP (C421A, for placental samples) were used. Primer sequences for lymphoblast samples were as follows: forward and reverse primers for genomic DNA were 5′-CAGTAATGTCGAAGTTTTTATCGCA-3′ and 3′-AAATGTTCATAGCCAGTTTCTTGGA-3′, respectively; whereas those for cDNA were 5′-CAGTAATGTCGAAGTTTTTATCGCA-3′ and 5′-TAACGAAGATTTGCCTCCACCTGTG-3′, respectively. A 291-bp and a 259-bp PCR product were obtained from genomic DNA and cDNA samples, respectively. After digestion of the PCR products from genomic DNA samples with BseMI, homozygotes of the G allele yielded a 291-bp fragment, whereas heterozygotes of the A allele yielded 291-, 261-, and 30-bp fragments. In contrast to the genomic DNA, homozygotes of the G allele yielded a 259-bp fragment, and heterozygotes of the A allele yielded 259-, 229-, and 30-bp fragments in the PCR products from cDNA samples.

In placental samples, the primer sets for genomic DNA and cDNA were the same: BseMI RFLP, forward, 5′-CAGTAATGTCGAAGTTTTTATCGCA-3′ and reverse, 5′-TTCGACAAGGTAGAAAGCCACTCTT-3′; and TaaI RFLP, forward, 5′-GTTGTGATGGGCACTCTGATGGT-3′ and reverse, 5′-CCTAACTCTTGAATGACCCTGTT-3′. PCR was carried out under the same conditions for the PCR-SSCP analysis, but only for 24 to 31 cycles. RFLP products were electrophoresed on a 3% agarose gel, then stained with SYBR Green I (Takara, Kyoto, Japan). Allelic expression was quantified on a fluorescence image analyzer (Hitachi, Tokyo, Japan) using Analysis Version 7.5 software. As a control, genomic DNA PCR-RFLP products were included and ratios of the allele-specific band intensities were taken as a 1:1 allelic representation. To eliminate sampling or measurement error, we conducted the experiment for each sample with three replicates.

Western Blotting. Human placental trophoblast samples were homogenized in a lysis buffer containing 50 mM Tris-HCl (pH 7.6), 150 mM NaCl, 0.1% SDS, 1 mM dithiothreitol, and 1×Complete Protease Inhibitor Cocktail (Roche Diagnostics, Germany). The lysate was centrifuged at 15,000g for 30 min at 4°C, and the supernatant was separated. Protein concentrations of the supernatants were determined by the Bio-Rad DC Protein Assay (Bio-Rad, Hercules, CA) using bovine serum globulin as a standard. The supernatants (80 μg of protein) were loaded onto SDS 4 to 20% (w/v) gradient polyacrylamide gels (Tefco, Tokyo, Japan) and transferred to Sequi-Blot polyvinylidene difluoride membranes (Bio-Rad) at 180 mA for 1 h. Thereafter, the membranes were blocked with 5% skim milk in TPBS (1× phosphate-buffered saline, 0.1% Tween 20) for 2 h at room temperature, then incubated overnight at 4°C with anti-BCRP, clone BXP-21 (1:50; Kamiya Biomedical, Thousand Oaks, CA) in skim milk in TPBS. The membrane was washed five times with TPBS and then incubated for 1 h at room temperature with horseradish peroxidase-conjugated goat anti-mouse IgG (1:1000; DakoCytomation Denmark A/S, Glostrup, Denmark). Polyvinylidene difluoride membranes were rinsed five times for 10 min with TPBS and then evenly coated using the ECL Western blotting detection system (Amersham Biosciences Inc.) for 1 min. The membrane was immediately exposed to Kodak X-OMAT AR film (Kodak, Tokyo, Japan) at room temperature. To assure the quantitative expression of BCRP, an additional marker protein expressed in placenta, alkaline phosphatase (ALP), was measured according to the same protocol except that different primary (polyclonal rabbit anti-human placental ALP; Biomeda, Foster City, CA) and secondary (horseradish peroxidase-conjugated goat anti-rabbit IgG; MP Biomedicals, Irvine, CA) antibodies were used. The immunoblots were quantitated using a public domain NIH Image program.

Statistical Analysis. Data are shown as the mean ± S.D. The statistical differences between various groups were determined with either the Mann-Whitney U test or the one-way analysis of variance (with the Tukey-Kramer multiple comparisons test), as appropriate. A χ² test was used to compare the allele frequency of each variant with that expected for a population in Hardy-Weinberg equilibrium. p < 0.05 was taken to be the minimum level of statistical significance.

Results

Identification of Variants in the BCRP Gene. For the identification of polymorphisms in all 16 exons and the 5′-flanking region of the BCRP gene, PCR-SSCP analysis was performed with genomic DNA obtained from 100 placental samples. Twenty polymorphisms were detected by SSCP analysis and identified by subsequent sequencing (Table 1). Of these, five SNPs resulted in the following amino acid substitutions: G34A (Val12Met), C376T (Gln126stop), C421A (Gln141Lys), G1322A (Ser441Asn), and T1465C (Phe489Leu). G34A (18.0%) and C421A (35.5%) variants were found at a relatively high incidence in this study. C376T, which is associated with an amino acid substitution from Gln to a stop codon at codon 126 (Gln126stop), was detected in only two placental samples (1.0%) as heterozygosity. Another polymorphism, a C1515 deletion, which results in a frame shift (Phe506Ser, Phe507Ser, Val508Leu, and Met509stop), was extremely rare (0.5%) in our samples. In the 5′-flanking and 3′ untranslated regions, four and three polymorphisms were identified, respectively: C-20445T, A-20296G, A-19781G, and a CTCA deletion at -19572 to -19569 in the 5′-flanking region; and A2332TA, A2364C, and C2512T in the 3′ untranslated region.

Haplotype Assessment. On the basis of the haplotype analysis for the three major variants (i.e., G34A, C376T, and C421A), four haplotypes were identified: G-C-C, G-C-A, A-C-C, and G-T-C. Their corresponding allelic frequencies were 46.0, 35.0, 18.0, and 1.0%, respectively.

Frequencies of G34A, C376T, and C421A in Different Ethnic Populations. We compared frequencies of the polymorphisms G34A, C376T, and C421A among three ethnic populations (Table 2). The frequency distributions of these three variants in all populations were in Hardy-Weinberg equilibrium. Japanese subjects had significantly higher frequencies of G34A and C421A than were found in the other two ethnic populations (p < 0.05). In addition, C376T was only detected in Japanese, but its frequency was extremely low (1.0%), as has been observed in placental samples. These results indicate that genetic frequencies of BCRP variants appeared to be dependent on ethnicity.

Polymorphisms and Placental BCRP Expression. The expression levels of the BCRP protein in 99 human placentas were determined by Western blotting using an anti-human BCRP monoclonal antibody, BXP-21 (Fig. 1). The membrane vesicles isolated from wild-type human BCRP-transfected HEK293 cells were used as the positive control (Kondo et al., 2004). The BCRP protein expression level was normalized to the placental ALP level, and the normalized value was compared with the control sample. The BCRP protein levels in various genotypic groups are shown in Fig. 2. The mean BCRP protein level was significantly lower in homozygotes for the A421 allele than in those for the C421 allele (0.37 ± 0.21 versus 0.75 ± 0.50, p < 0.05), and heterozygous samples displayed an intermediate value. In contrast to the C421A polymorphism, the mean of the BCRP level for the G/G, G/A and A/A genotypes at nucleotide position 34 was 0.58 ± 0.37, 0.70 ± 0.50, and 0.65 ± 0.51, respectively, and did not differ among these genotypes.

The relationship between various genotypic patterns and mRNA levels is also shown in Fig. 2. In contrast to the findings at the protein level, significant genotype-dependent changes in the mRNA expression were not observed.

Allelic Imbalance of the BCRP Gene in Human Placentas. We estimated the allelic expression bias, a measure of the expression of the reference allele divided by the variant allele, using placental cDNA samples to confirm the interindividual variation in the allelic imbalance of the BCRP gene and to characterize the contribution of the SNPs to the allele-specific mRNA expression (Bray et al., 2003; Yan et al., 2002).

After the screening of the genomic DNA from all 100 placental samples, it was possible to identify 56 individuals who were heterozygous for either the BseMI (G34A) or TaaI (C421A) site. These two SNPs were used as markers in the present study. Some of the 56 individuals had fractional allelic expression values outside the 95% confidence interval for the mean (1.16 ± 1.21; 95% confidence interval, 0.85∼1.48) (Fig. 3A). Notably, the values in subjects 19 (9.94) and 62 (0.00) were extremely high and low, respectively, being well outside the intervals, indicating monoallelic expression (i.e., preferential expression of one of the two alleles) (Fig. 3B). Although the difference in expression between the two alleles varied among samples, the expression bias in most samples (43 of 56) was within the approximated 95% confidence interval. The allelic expression bias ranged from 0.60 (65) to 1.47 (82), excluding these two outside samples, suggesting that neither variant is associated with large changes in transcription from the allele carrying each variant.

Allelic Expression Pattern in Informative Lymphoblasts. Among the 56 placental samples, 2 indicated a preferential monoallelic expression. To determine whether the two alleles of the BCRP gene were differentially expressed according to parental origin, we used RT-PCR of total RNA extracted from EB virus-transformed lymphoblasts, followed by PCR-RFLP. The parental origin of alleles expressed in children was identified by RFLP analysis. Lymphoblasts were obtained from a panel of 18 healthy Japanese individuals who were members of five distinct families. These samples allowed the precise determination of the parental origin of alleles in the heterozygous children. Of all the subjects, two siblings were heterozygous for a polymorphism at the BseMI site (G34A). All RT reactions in the present study included a negative control to ensure that genomic DNA did not contaminate the subsequent PCR. We first determined the parents' genotypes using genomic DNA samples. Whereas the paternal genotype was homozygous for the G34 allele, the maternal genotype was heterozygous for the A34 allele (Fig. 4). Since the two siblings were heterozygous carriers, their G and A alleles were of paternal and maternal origin, respectively. In contrast to the genomic DNA-based genotypes, both of the siblings showed a monoallelic paternal expression of the RT-PCR products (i.e., cDNA samples) (Fig. 4). These results suggested that the expression pattern of the BCRP gene in lymphoblasts is regulated by parental imprinting.

It is interesting to know whether a monoallelic parental expression of the BCRP gene is inherited. To address this issue, we further analyzed expression pattern using maternal RT-PCT product, because the maternal genotype was heterozygous for the A34 allele. As shown in Fig. 4, the maternal inactive A34 allele was inherited by both siblings. These results suggest that a monoallelic parental expression is inherited, at least in EB virus-transformed lymphoblasts.

Discussion

Before the functional characterization of the variants, we analyzed genetic polymorphisms in Japanese placental samples and compared allelic frequencies among different ethnic groups. Similar to other ABC-transporters (Ieiri et al., 2004; Marzolini et al., 2004), various synonymous and nonsynonymous polymorphisms were observed (Table 1). Among the nonsynonymous polymorphisms, G34A (Val12Met) and C421A (Gln141Lys) appeared commonly in Japanese subjects, and allelic frequencies of these polymorphisms were in keeping with those of a previous report (Imai et al., 2002). As shown in Table 2, the C421A variant is widespread not only in Japanese subjects but also in Caucasian subjects, with a frequency of between 10% and 30%. However, this variant is rare in African Americans, suggesting an ethnic difference in the frequency of polymorphisms in the BCRP gene. These trends were well consistent with the findings reported by de Jong et al. (2004), who indicated that the frequency of C421A variant was the most common (i.e., 34%) in Han Chinese among various ethnic populations they evaluated. It is difficult to assess a plausible explanation for the difference; the C421A allele appears to be very common in Asian populations.

Another polymorphism, C376T in exon 4, which substitutes a stop codon for Gln126, was detected in only four of our study samples as heterozygosity. Although the frequency of the C376T allele was extremely low, C376T may have an impact on protein expression because of the premature stop codon.

Recent studies indicate that haplotype-based approaches, which take into consideration the combination of SNPs present in one allele, offer greater ability to predict changes in phenotype than do SNP-based approaches (Judson et al., 2000; Stephens et al., 2001). Thus, we determined haplotypes for the three major SNPs (i.e., G34A, C376T, and C421A) in Japanese subjects. In the present study, four allelic patterns were identified. Interestingly, the C421A variant existed only as a G-C-A allele, suggesting that it does not coexist with the other two variants.

In addition to the cording haplotypes, we determined haplotypes for four SNPs in the 5′-flanking (C-20445T, A-20296G, A-19781G, and CTCA deletion at -19572 to -19569) and for three SNPs in the 3′ untranslated (A2332TA, A2364C, and C2512T) regions. Five major haplotypes were identified: C-A-A-CTCA, C-A-A-(CTCA)deletion, and C-G-A-(CTCA)deletion in the 5′-flanking region; and A-A-C and TA-A-C in the 3′ untranslated region. Although we evaluated the relationship between various haplotypes and both mRNA and protein levels, significant haplotype-dependent changes in both expression levels were not observed (data not shown).

Our findings indicate that the C421A variant may affect the placental expression of BCRP protein; samples with an A421 allele had a reduced protein level, as compared with those without the variant. These findings are consistent with the report that C421A BCRP-transfected PA317 cells and KB-3-1 human epidermoid carcinoma cells showed a markedly decreased protein expression and low-level drug resistance (Imai et al., 2002). Similar findings were made in a different cell line, HEK293 (Kondo et al., 2004). Although one study demonstrated no significant correlation between the C421A variant and expression of intestinal BCRP mRNA or protein (Zamber et al., 2003), the present study and collective evidence suggest that C421A is potentially of functional significance.

C421A is located in the functionally important ATP-binding region between Walker A and B motifs and is associated with a differentially charged amino acid substitution. These characteristics suggest that the C421A variant alters the tertiary structure of the BCRP protein, leading to greater susceptibility to degradation (Imai et al., 2002). In addition, Mizuarai et al. (2004) reported that ATPase activity in the membrane of SF9 cells infected with the C421A variant showed a decrease of 1.3 below that of wild-type cells. They also indicated that the C421A variant exhibited reduced drug resistance in polarized LLC-PK1 cells along with increased intracellular drug accumulation.

In an in vivo human study, Sparreboom et al. (2004) recently evaluated the effects of the C421A variant on the pharmacokinetics of diflomotecan, a synthetic derivative of camptothecin, in 22 cancer patients, and provided the first evidence linking variant BCRP alleles to altered drug exposure. Patients with this variant as heterozygosity indicated about 3-fold higher plasma levels than did patients with wild-type alleles. These results suggest that interindividual variability in substrate drug effects might be influenced, in part, by BCRP genotype.

Other nonsynonymous variants, Arg482Gly and Arg482Thr, have been reported to have a crucial role in protein function and in altering the multidrug resistance phenotype by changing substrate specificity (Honjo et al., 2001; Allen et al., 2002). However, these variants ware not detected in the present study, indicating that mutations at position 482 may occur due to drug selection.

In the present study, we assessed the molecular mechanisms responsible for the low protein levels in A421-placental samples from two perspectives, polymorphism and allelic imbalance. Allelic imbalance is a differential expression which can be determined by the measurement of the relative expression level of two alleles of one gene (Yan et al., 2002; Bray et al., 2003; Knight, 2004). One example of this phenomenon is provided by a recent study by Wojnowski and Brockmöller (2004), which demonstrated a hepatic transcriptional imbalance of the CYP3A5 gene in heterozygous CYP3A5^*1A/^*3 samples. In that case, CYP3A5^*1A may act as a cis-acting SNP, which increases the expression of the CYP3A5 transcript from the allele carrying it, resulting in an imbalance of mRNA expression. Thus, allelic imbalance can be used for the functional evaluation of SNPs that affect mRNA expression. If the C421A variant is a functional cis-acting polymorphism differentially affecting the expression level of each copy of a gene, a strong over- or under-shift in allelic expression bias will be expected. However, as shown in Fig. 3, although only two samples (19 and 62) exhibited significant allelic imbalance (i.e., preferential expression of one of the two alleles), the allelic expression bias in most samples was within the approximated 95% confidence interval (0.85∼1.48). These results suggest that the predominant expression pattern of BCRP in placental samples is biallelic, and C421A is not a cis-acting SNP. The C421A variant was associated with a reduction in placental protein levels but not in mRNA levels (Fig. 2). These observations are in line with a finding by Imai et al. (2002), who demonstrated that murine PA317 cells expressing C421A had increased intracellular drug accumulation, coupled with reduced protein levels and a similar mRNA expression. Therefore, taking these observations into consideration, the reduction in protein expression levels in the A421-placental samples is considered to be a result of the translation efficiency of BCRP. Nevertheless, we cannot exclude the possibility that the cis-acting regulatory polymorphisms, which may form haplotypes with the C421A variant, responsible for the change in BCRP expression reside far up- and down-stream of the gene of the affected allele.

Since protein expression is known to be regulated not only by mRNA and/or DNA sequences but also by epigenetic modifications such as genomic imprinting, chromatin alteration, methylation, and acetylation, we introduced an epigenetic concept in this study. As shown in Fig. 4, we have demonstrated that human lymphoblasts showed a paternal expression of BCRP, in contrast with the findings in placental samples, in which a monoallelic expression was observed in only a minority of samples, suggesting a unique allele-specific expression profile of BCRP. Among known imprinted genes, the Wilms' tumor suppressor gene (WT1) has been reported to exhibit a unique expression profile (Mitsuya et al., 1997); cultured human fibroblasts and lymphocytes showed a paternal or biallelic expression of WT1 in some cases, whereas a maternal or biallelic expression was observed in human placental villi and fetal brain tissue (Jinno et al., 1994; Nishiwaki et al., 1997). These results suggest that the allele-specific expression profile of certain genes depends on the tissue source. Thus, expression patterns in other BCRP-enriched human tissues such as liver and small intestine are of interest.

The data presented here suggest that the C421A variant in the BCRP gene, a common SNP in both Japanese and Caucasian populations, alters protein levels in the human placenta. In addition, a tissue-dependent allele-specific expression profile of BCRP is suspected. In vivo human studies with regard to these findings are required for the establishment of individualized anticancer therapy.

Appendix

Primer sequences used for the PCR-SSCP analysis of the human BCRP gene are shown in Table 3.