Genetic polymorphism at the 5′ regulatory region of multidrug resistance 1 (MDR1) and its association with interindividual variation of expression level in the colon

Drugs are detoxified and conjugated in vivo and then exported out of the cells. The activity of the detoxification system affects the pharmacokinetics of drugs. Many recent studies have correlated polymorphisms of detoxification-related genes such as cytochrome P450s and glutathione S-transferases with the efficacy and side effects of drugs (1–3). On the other hand, the export of drugs has been shown to involve a group of proteins belonging to ATP binding cassette (ABC) transporters (reviewed in http://nutrigene.4t.com/humanabc.htm; 4–9). A member of ABC transporters, P-glycoprotein (P-gp), affects the pharmacokinetics of drugs by limiting the rate at which they are absorbed (10–14). Thus, interindividual variations in the levels of activity and expression of ABC transporters might be a critical factor in the development of pharmacokinetics.

The multidrug resistance 1 (MDR1) gene encodes a 170-kDa transmembrane protein, P-gp, located at the cytoplasmic surface of the cell. P-gp consists of two membrane-spanning domains and two nucleotide-binding domains. Of the various molecular targets, P-gp expression is responsible for cell resistance to the widest variety of anticancer drugs (15, 16). P-gp overexpression plays an important role in the acquisition of drug resistance in various cancer cells. The enhanced expression of the MDR1 gene in malignant cancer cells has been attributed to various mechanisms, including nuclear translocation of Y-box-binding protein-1 (17, 18), promoter rearrangement (19), and alteration of methylation status at CpG sites on the MDR1 promoter (20–22). On the other hand, P-gp is expressed in normal cells of various organs, such as intestine, liver, kidney, brain, and placenta (23–25). P-gp's wide substrate specificity and apical localization strongly suggest that it plays a critical role in drug disposition in the human body as well as in animal model (10–14). In the intestine, P-gp is thought to participate in drug absorption after drug ingestion (26–28). At the blood-brain barrier, P-gp influences the uptake of substrates into brain (23, 29–31). Because MDR1 expression levels vary widely among individuals (32), these variations may affect the toxicity of drugs and the efficacy of drug treatment from individual to individual through different drug dispositions. Furthermore, these variations may have another clinical relevance as a cancer risk factor because we and others have recently observed the suppression of polyp formation in mdr1a (mouse orthologue of MDR1)-disrupted mice (33, 34). On a molecular basis, however, the mechanism underlying the interindividual variations in the basal expression level is unknown.

Genetic polymorphisms and their association with P-gp level in the MDR1 gene have been reported recently (35, 36). Hoffmeyer et al. reported 15 single nucleotide polymorphisms (SNPs), including 6 in the coding region, in healthy Caucasians. Those authors suggested that one of the SNPs, c.3435C>T (exon 26), is correlated with intestinal P-gp expression and uptake of digoxin, a P-gp substrate, administered p.o. However, the C/T exchange at the nucleotide 3435 of P-gp is a wobble base substitution. Thus, it remains unclear whether this substitution alters P-gp expression level directly. In the present study, we identified genetic polymorphisms in the 5′ regulatory region of MDR1 and examined them for any association with interindividual variation of MDR1 expression level in normal colorectal mucosa and liver.

Blood samples were obtained from 25 healthy Japanese volunteers at Kyushu University. Clinical samples of normal colorectal mucosa were taken from 72 Japanese patients with colorectal cancer who had undergone surgical resection of the cancer at the Department of Surgery II, Kyushu University Hospital (Fukuoka, Japan), the Coloproctology Center, Takano Hospital (Kumamoto, Japan), or the Department of Surgery I, Gunma University Hospital (Maebashi, Japan) between September 1993 and August 1998. These samples of blood and noncancerous mucosa were obtained under an Institutional Review Board-approved protocol, with all subjects providing their informed consent. The samples were frozen in liquid nitrogen and stored at −80°C until RNA and DNA were extracted. None of the patients had received chemotherapy before the surgical resection. Clinical samples of normal liver tissues were taken as a noncancerous tissue surrounding cancer from 43 Japanese patients with hepatocellular carcinoma, who had undergone surgical resection of cancer at the Department of Surgery II, Kyushu University Hospital.

Genomic DNA from the volunteers' blood samples was isolated using the QiaAmp (Qiagen, Inc., Chatsworth, CA) blood kits, and DNA from tissues of patients was isolated using the Easy DNA Kit (Invitrogen, San Diego, CA) according to the manufacturer's protocol. RNA was isolated using the RNA extraction reagent TRIzol (Invitrogen Life Technologies, Gaithersburg, MD) or RNeasy (Qiagen) according to the respective manufacturers' protocols.

Specific oligonucleotide primers for PCR amplification of MDR1 gene fragments were derived from known sequences (Genbank accession nos. AC002457 for the 5′ upstream regulatory region and exons 1–7 and AC005068 for exons 8–28). The locations of the SNPs in the exons corresponded to positions of the MDR1 cDNA (Genbank accession no. M14758, codon TTC in exon 10, F335, is missing in that sequence); the first base of the ATG start codon was set to +1. The exons were defined by Chen et al. (37), and the Human Genome Organization recommended nomenclature (38) was used for SNP nomenclature. The primers were designed to amplify the regions that include sequences including the SNPs reported previously or to cover about 4 kb of the MDR1 upstream regulatory region as shown in Table 1. The PCR conditions for these primers are available by a request to us. Sequences of purified PCR fragments were obtained by automated DNA sequencing on ABI3700 (capillary) sequencers by using BigDye terminator cycle sequencing reactions (Perkin-Elmer/Applied Biosystems, Foster City, CA).

Haplotypes of individuals who were heterozygous at least in one SNP locus were determined by PCR amplification and sequencing using the forward primer MDR1P5F and the reverse primer MDR1P11R (Table 1). PCR amplification was performed by using high fidelity DNA polymerase, KOD-Plus (Toyobo, Osaka, Japan), according to the manufacturer's protocol. The fragments were inserted into pT7Blue-3 vector (Novagen, Madison, WI) and subcloned. At least six colonies were picked up, and plasmids were purified by the Qiagen DNA kit according to the manufacturer's protocol. SNP sites were analyzed by sequencing, and haplotypes were confirmed.

Quantitative reverse transcription-PCR was performed by real-time Taqman technology and Model 7900 sequence detectors (Perkin-Elmer/Applied Biosystems) as described previously (39). The sequences of the primer pairs and the probe used in this study were described previously (32).

To extract type 3 allele, fragments including −2604 to −570 were amplified from templates corresponding to homozygotes for haplotypes 1 and 2 and to heterozygotes for haplotypes 1 and 3. The forward primer 5′-AAAGCTAGCTGTCAGTGGAGCAAAGAAATG-3′ and the reverse primer 5′-AAAGCTAGCCTCGCGCTCCTTGGAA-3′, each of which included a NheI site, were used. The PCR conditions for these primers are available by a request to us. These amplification products were inserted into the NheI site of a pGL3 basic vector (Promega, Madison, WI). SNP sites in the constructs were confirmed by sequencing.

We used the human hepatocarcinoma cell line (HepG2) in this study. Cells were grown at 37°C in a humidified atmosphere containing 5% carbon dioxide. We transfected 1 μg of pGL3 basic vector DNA or reporter construct and cotransfected 100 ng of phRL-TK Vector DNA (Promega) in all wells as a transfection control by using LipofectAMINE 2000 (Life Technologies, Grand Island, NY) reagent according to the manufacturer's protocol. The plates were incubated at 37°C for 6 h after adding DNA-LipofectAMINE complex, and the growth medium was then changed. The plates were incubated for further 24 h prior to luciferase assay. We measured firefly and Renilla luciferase activities in a luminometer using the Dual-Luciferase Reporter Assay System (Promega). Data were normalized for transfection efficiency by the Renilla luciferase activity. In all cases, transfections were carried out in triplicate, with three wells of a 24-well plate containing identical transfection reactions.

The primary antibodies used were P-gp (JSB-1; mouse monoclonal, Sanbio, Uden, the Netherlands). Immunostaining of P-gp was performed as described previously (32). To assure quantitative detection of P-gp by immunohistochemistry, an additional marker protein that is expressed in enterocytes, villin, was used. For quantification, ImageGauge (Fuji Photo Film Co., Tokyo, Japan) software was used.

The DNA sequences of the sense strand of each oligonucleotide were 5′-AAATGAAAGGTGAGATAAAGCAACAA-3′ (−2352G), 5′-AAATGAAAGGTGAAATAAAGCAACAA-3′ (−2352A), 5′-GAGCTCATTCGAGTAGCGGCTCTTCC-3′ (−692T), and 5′-GAGCTCATTCGAGCAGCGGCTCTTCC-3′ (−692C). Nuclear extracts (2 μg/μl of protein) were prepared from HepG2 cells as described previously (40, 41). They were then incubated for 30 min at room temperature in a final volume of 10 μl of reaction mixture containing 2 μl of 5× binding buffer, 5-mm DTT, 10-ng poly(deoxyinosinic-deoxycytidylic acid), and 1 × 10⁴ counts/min of [³²P]-labeled oligonucleotide probe in the absence or presence of various competitors. We tried five different kinds of binding buffer, determined the one that generated the clearest retarded band, and used them for further analyses. The composition of the 5× binding buffers used for the detailed analyses were as follows: Buffer A, 60-mm HEPES, 300-mm KCl, 20-mm MgCl₂, 5-mm EDTA, 60% (v/v) glycerol; Buffer B, 50-mm Tris-HCl (pH 7.5), 250-mm NaCl, 12.5-mm CaCl₂, 5-mm EDTA, 40% (v/v) glycerol. Next, the samples were electrophoresed on 4% polyacrylamide gel (polyacrylamide/bisacrylamide ratio, 79:1) in a Tris-borate-EDTA buffer (0.089-m Tris, 0.089-m boric acid, and 0.002-m EDTA). The gel was exposed to an imaging plate and analyzed using a Fujix BAS 2000 Bioimage Analyzer (Fuji Photo Film).

Statview 5.0 software (SAS Institute, Cary, NC) was used for statistical analysis. Results of MDR1 mRNA levels versus diplotypes were analyzed by the Kruskal-Wallis test. Significance was defined as P < 0.05. The correlations between MDR1 mRNA level and P-gp level were determined using Spearman's test. This test is usually used for nonparametric analysis when it is unclear whether the variables show normal distribution. P values of <0.05 were considered significant. The Spearman's coefficient (r) and associated P were calculated. Unpaired t tests were performed to compare relative luciferase activities of reporter constructs containing haplotype 1, 2, or 3 at the 5′ upstream regulatory region of MDR1 gene in transfection experiments.

To find MDR1 polymorphisms at the 5′ upstream regulatory region, we analyzed genomic DNA isolated from peripheral blood of 25 healthy Japanese volunteers. The upstream region, spanning about 4 kb from the transcriptional start site, was amplified by PCR and analyzed by direct sequencing. Eight SNPs were identified at the 5′ upstream regulatory region, and six of them had not been reported before. The allele frequencies of these SNPs observed in the 50 chromosomes are presented in Table 2. The ATG start codon locates in exon 2, and the transcription start site corresponds to −699 from the ATG in the genomic DNA. SNPs −692T>C and −934A>G were identical to the previously reported −129T>C and −41aA>G (42), respectively, by this numbering system (see “Materials and Methods” for details). The sequences were inspected for deviations from the original MDR1 sequences (Genbank accession nos. AC002457 and AC005068), which we defined as the major type. We also analyzed the other six SNPs in exons and introns: these six SNPs were reported previously to have allele frequencies of >0.05 in Caucasians as well as in Japanese (35, 36, 42). The allele frequencies of these SNPs obtained from our analysis are shown in Table 2. The allele frequencies were within the range expected from sample size as those reported before. A strong association between c.3435C>T and c.2677G>T, A was observed as reported previously (42–44), whereas there was no linkage between the polymorphisms at the 5′ upstream regulatory region and those of coding region including c.1236C>T, c.2677G>T, A, and c.3435C>T. No association between −692T>C and c.2677G>T, A or c.3435C>T is consistent with previous reports (43, 44).

To unequivocally determine the frequency of haplotypes at the regulatory region, we amplified by PCR the 2-kb fragment containing these polymorphic sites at the −2410, −2352, −1910, −934, and −692. Of 25 blood samples, we omitted analysis of homozygous samples at all these sites and used heterozygous samples at least in one of those sites. After subcloning the amplified fragments into the pT7Blue-3 vector, we determined their nucleotide sequences. Because the frequencies of the minor alleles at −2903, −1717, and −1325 were too low (0.02) for statistical analysis, we omitted these three alleles from the analysis. As shown in Table 3, −2410T(C), −1910T(C), and −692T(C) were detected together in each clone, but −2352G(A) was independent. We then defined the haplotypes as follows: haplotype 1 (−2410T, −2352G, −1910T, −934A, and −692T), haplotype 2 (−2410T, −2352A, −1910T, −934A, and −692T), and haplotype 3 (−2410C, −2352G, −1910C, −934G, and −692C). Three haplotypes (haplotypes 1–3) accounted for >95% of the population. The promoter haplotypes were not associated with any SNPs examined in coding and intron regions in Japanese (data not shown).

We tested each of the five polymorphisms (−2410, −2352, −1910, −934, and −692) at the 5′ upstream regulatory region for any association with MDR1 mRNA levels in 72 normal colorectal mucosa. The 72 samples were divided according to diplotype into four groups: diplotype A (haplotypes 1/1), diplotype B (haplotypes 1/2), diplotype C (haplotypes 1/3), and diplotype D (haplotypes 2/2). Then, we found that the mean MDR1 mRNA level of diplotype A, which had two haplotype 1, was higher than that of diplotype D, which did not have haplotype 1 (P = 0.04; Fig. 1A). The mean MDR1 mRNA levels of diplotypes B and C, each of which had one haplotype 1, were intermediate between those of diplotypes A and D. MDR1 mRNA level was normalized with glyceraldehyde-3-phosphate dehydrogenase mRNA level. When we analyzed MDR1 mRNA levels in the samples from normal liver, we obtained results similar to those from colon (Fig. 1B), although the association was statistically much lower compared with that in colon (P = 0.2).

We also examined each SNP for any individual associations with mRNA level. The association between −692T>C and mRNA level in colon was not statistically significant in the present study, although there was a tendency toward lower mRNA levels in T/C heterozygotes than in T/T homozygotes (P = 0.2; data not shown). For −2352G>A, A/A homozygotes showed lower mRNA levels than in G/G homozygotes (P = 0.07).

We then confirmed a correlation of mRNA expression levels of MDR1 and the expression levels of P-gp. Because we could not get a sufficient volume of each sample for Western blotting, we measured P-gp levels by immunohistochemistry method using antibody JSB-1. Therefore, the measurements were semiquantitative rather than quantitative. Fourteen of the 72 samples were examined and we found that MDR1 mRNA level showed a significant correlation with P-gp level by Spearman's test (r = 0.428, P = 0.01). Representative data are presented in Fig. 2.

c.3435C>T, which was reported to affect P-gp level in intestine and/or function (35, 45), was not associated with the MDR1 mRNA level in colon in this study (P = 0.7; data not shown), consistent with the previous report analyzing intestine (46) or placenta (42). c.2677G>T, A, which was also reported to correlate with the MDR1/P-gp expression level in placenta (42), was not associated with the MDR1 mRNA level in colon in the present analysis (P = 0.89, data not shown); consistent with the results analyzing mRNA level in intestine (46). None of other SNPs examined in this study was associated with MDR1 mRNA level in colorectal epithelium (data not shown).

To examine the direct association between the haplotypes and MDR1 promoter activity, we cloned the 5′ upstream regulatory region between −2604 and −570 of genomic DNA from volunteers carrying three naturally occurring haplotypes (haplotypes 1–3). We then ligated the fragments to the reporter gene in the pGL3 basic vector. Because of the low frequencies of the polymorphisms at −1717 and −1325, we used genomic DNA with T monomorphic at −1717 and with A monomorphic at −1325 for reporter plasmid construction. These three constructs were then subjected to transient transfection in a human hepatoma cell line, HepG2. The promoter activity was analyzed after 48 h of transfection and normalized with the cotransfected phRL-TK activity. As shown in Table 4, the minor-type construct carrying haplotypes 2 and 3 showed expression of 85.3 ± 4.65% and 87.1 ± 1.64%, respectively, of the major-type construct carrying haplotype 1. Together, these experiments suggest that polymorphisms at the 5′ upstream regulatory region affect the basal promoter activity of reporter constructs containing the human MDR1 gene upstream promoter region.

We used electrophoretic mobility shift assays to investigate whether the SNPs of MDR1 5′ regulatory region altered binding of nuclear proteins. We first examined −2352G>A (Fig. 3A). A retarded band was observed when the probe −2352G incubated with nuclear extracts of liver cells. This band was thrice weaker when −2352A incubated. The specificity of the DNA-protein interaction was demonstrated by appropriate competition assays (i.e., the upper band almost completely disappeared under a 10-fold excess of the unlabeled oligonucleotide −2352G, while the addition of excess amounts of minor-type oligonucleotide −2352A did not inhibit the protein from binding to probe −2352G). We then examined −692T>C (Fig. 3B). The allele-specific appearance of retarded band was also observed when the probe −692T incubated with nuclear extracts. In competition assays, the upper band almost completely disappeared under a 10-fold excess of the unlabeled oligonucleotide −692T, and much weaker (nine times compared with −692T) inhibition of the binding was observed with the competitor −692C. Other oligonucleotide probes (−2410T>C, −1910T>C, and −934A>G) showed no difference of protein-binding property between major and minor types under the present conditions.

Concerning the polymorphic MDR1 genotypes, Hoffmeyer et al. (35) demonstrated that in Caucasians, a SNP in exon 26 of the MDR1 gene (c.3435C>T) was correlated with lower intestinal P-gp levels and subsequently a higher rate of intestinal absorption of P-gp substrate. In Japanese subjects, however, c.3435C>T was reported not to be related to placental expression of P-gp (42). Furthermore, in contrast with the report by Hoffmeyer et al., a T allele of c.3435C>T increased the expression level of MDR1 mRNA in duodenal enterocytes of healthy Japanese subjects (45). c.3435C>T is a silent mutation that does not cause amino acid substitution. Kim et al. (43) reported that P-gp function could be affected by c.2677G>T, A, a SNP at exon 21 producing Ala893Ser and Ala893Thr, respectively, which is partially linked to c.3435C > T. Tanabe et al. (42) reported that P-gp expression levels in placenta were affected by c.2677G>T, A. Thus, the relationship between c.3435C>T genotype and biochemical phenotypic P-gp activity appears not to be clearly established as pointed recently (47).

We could not find any association between c.3435C>T and MDR1/P-gp expression level; this finding was inconsistent with that of Hoffmeyer et al. (35). Although we analyzed the association between MDR1 mRNA level and the haplotypes consisting of the SNPs at both the 5′ upstream regulatory region and the coding region including c.2677G>T, A and c.3435C>T, we could not have better association than that between the MDR1 mRNA level and the haplotypes consisting of only the regulatory SNPs (data not shown). Whereas we examined the mRNA level of MDR1, they examined P-gp protein level by immunohistochemical analysis. Although MDR1 mRNA levels could possibly be different from P-gp levels in human samples, such a difference is unlikely: When 14 samples were compared, mRNA levels detected by real-time reverse transcription-PCR were found to be proportional to protein levels detected by immunostaining. Another possibility would be an ethnic specificity in the association between SNPs and MDR1/P-gp expression level.

In this paper, we showed that polymorphisms at the 5′ upstream regulatory region were associated with MDR1 mRNA level in the Japanese population. Although we did not find any novel binding sequences for known transcriptional factors around the SNPs that we found, we confirmed a direct association between the haplotypes carrying these SNPs and transcriptional activity in vitro. We employed three types of reporter constructs, carrying haplotypes 1–3, respectively, and then assessed transcriptional activity by transient transfection assay. We observed that the reporter constructs carrying the human MDR1 gene to the upstream promoter region of haplotype 1 showed a statistically significant increase in promoter activity compared with the reporter constructs carrying haplotype 2 or 3. Concerning the only small reduction of the reporter activities by the minor haplotypes, it may be due to the limitation of transient transfection assay system. It is generally believed that transcription activity is affected by chromatin structure, and expression plasmid transiently introduced into cells sometimes is unable to maintain the proper chromatin structure for the appropriate expression. Esterbauer et al. reported that a promoter polymorphism in UCP2 gene is associated with decreased risk of obesity in middle-aged humans, while transcriptional activities of variant promoter was only 22% higher than wild-type promoter when they were transiently transfected into adipocytes (48). Furthermore, Ince et al. reported that stable transfection of a MDR1 promoter/luciferase construct reproduced the overexpression phenotype in a drug-resistant cell line despite the failure in transient transfection assays (49).

The mechanism underlying the possible association between the SNPs we found at the 5′ regulatory region and the transcriptional activity of MDR1 is not known. SNPs are not located at the typical binding sites for the known transcription factors (see Fig. 4; 50). The SNPs may affect the binding of unknown trans-acting factors. We identified a putative binding protein for the promoter fragment including −692T or −2352G, which could not or weakly bind to the fragment carrying −692C or −2352A, respectively, by gel mobility shift assay. Alternatively, the SNPs may be associated with different epigenetic states that might modulate transcriptional activity. An inverse relationship has been shown between the CpG methylation status at the MDR1 promoter region and the MDR1 expression level in cancer cell lines (20), acute myelogenous leukemia (21), and bladder cancer (22). It is interesting to note from this point of view that conversion from major to minor alleles at −692, −1717, −1910, −2410, and −2903 generates either CG or GC sites, which could be intrinsically methylated. Fryxell et al. reported that the methylation status of the CpG-rich domains around −692 and −1910 was not similar among hematopoietic tumor cell lines and among acute myeloid leukemia patient samples (51). Further study is required to understand whether SNPs at the 5′ flanking region might affect MDR1 expression level through altered CpG methylation status. Epigenetic status may also explain the huge differences of the MDR1 expression within the group of diplotype A compared with other group as seen in Fig. 1. As described above, the expression of the MDR1 might be determined not only by regulatory SNPs but also by other factors such as epigenetic status and expression and/or activity of the transacting factors. Each factor (e.g., SNPs of transcription factors or methyl-binding proteins) may affect the expression level in each individuals with different extent. In this regard, the SNPs at regulatory region could be a necessary condition but not a sufficient condition.

Another interesting question is whether SNP-dependent interindividual variation of expression levels could be tissue specific or not. Tanabe et al. observed that a significant correlation between −692T>C, one of the promoter polymorphisms, and placental P-gp expression (42). We found a tendency between −692T>C and the levels of both MDR1 mRNA and P-gp in colorectal mucosa although not significant statistically. We also analyzed MDR1 mRNA level in normal liver specimens isolated from 43 hepatocellular carcinoma patients and found that the polymorphisms at −692, −934, −1910, −2352, and −2410 tend to be associated with MDR1 mRNA level like colorectal mucosa. The SNPs found at the 5′ regulatory region are likely associated with the levels of MDR1 mRNA and P-gp in multiple tissues rather than in only a specific tissue.

Because many drugs are substrates of P-gp as described in “Introduction,” degree of expression and activity of P-gp can directly affect the therapeutic effectiveness of such agents. Besides pharmacological relevance, interindividual variety of P-gp SNPs and expression level may have another clinical impact as follows. Recently, we and others found the role of P-gp in colorectal carcinogenesis in mice (33, 34). We found that DNA damage was significantly increased in mice disrupted in mdr1a, orthologue of human MDR1, compared with wild-type mice. Surprisingly, we and others also found that statistically smaller numbers of polyps was generated in mdr1a-disrupted mice compared with wild-type mice under Apc^Min background. Interindividual variety of P-gp expression in colon could then be associated with colorectal carcinogenesis in human. The possibility whether polymorphisms at the 5′ upstream regulatory region of MDR1 gene would be correlated with colorectal carcinogenesis is currently under investigation in our laboratory.

In conclusion, the SNPs at the 5′ regulatory region of the human MDR1 gene are associated with the expression of MDR1 mRNA and P-gp in colorectal mucosa in the Japanese population. The results would provide a framework for further analysis of the relationship between the SNPs of MDR1 and drug response and as well as for further assessment of the importance of P-gp in interindividual variability of drug response and cancer risk.

We thank Kimiko Tomita (Coloproctology Center, Takano Hospital) for help in the investigation of clinical data. We also thank Shingo Baba, Ken-ichi Taguchi, Hiroko Baba, Takao Fukuda, Keiji Hisaeda, Takuya Ebihara, and Hidenori Tachida (Kyushu University) for technical support and helpful discussion.