Increased Expression of an ATP-binding Cassette Superfamily Transporter, Multidrug Resistance Protein 2, in Human Colorectal Carcinomas¹

Two ABC³superfamily transporters, Pgp and MRP1, confer MDR on cancer cells through enhanced drug efflux (1, 2, 3, 4). Treatment of cancer cells with many anticancer drugs, including Vinca alkaloids(vincristine and vinblastine), anthracyclines (doxorubicin and daunomycin), taxanes (Taxol and taxotere), and epipodophyllotoxins(etoposide and teniposide), can result in overexpression of Pgp and MRP1 (1, 2). Overexpression of human Pgp and MRP1 in cancer cells leads to drug resistance against anthracyclines, Vinca alkaloids, and epipodophyllotoxins (3, 4).

Patients with colorectal cancers have been treated primarily by surgical resection, and these cancers often have decreasing sensitivities to chemotherapeutic agents (5). Intrinsic drug resistance in untreated colon cancers is thought to be due in part to Pgp because normal colon tissues themselves express Pgp (5, 6). The expression of Pgp increases in colon cancers compared with findings in noncancerous regions (7), whereas no significant difference exists in Pgp levels between cancerous and noncancerous regions (8). Many other studies have noted a decreased expression of Pgp in cancerous regions compared with noncancerous regions in colon cancers (9, 10, 11, 12). Thus, Pgp expression does not seem to be a prognostic marker for colorectal cancer (13, 14).

The MRP1 gene is expressed in various tumor types,and MRP1 expression is associated with drug resistance in or prognosis of breast cancer, gastric cancer, neuroblastomas, retinoblastomas, and lung cancers (15, 16, 17, 18, 19). In a study of the expression of MRP1 in colorectal carcinomas, 7 of 30 cases showed strong MRP1 staining in tumors, whereas normal mucosal tissues showed weak MRP1 staining (12). MRP1 expression does not increase in cancerous regions compared with their noncancerous counterparts in colon cancers (20). Because expression of both Pgp and MRP1 occurs in normal colon mucosa, both ABC transporters may be involved in intrinsic drug resistance in colorectal cancers.

In addition to MDR1 and MRP1, five human MRP subfamily members (MRP2–6) that show structural similarity to MRP1 have been identified (21). The genes encoding MRP1–6 are on different chromosomes, and MRP1–6 mRNAs are expressed in a variety of normal tissues (21). Of these MRPfamily genes, MRP2 appears to mediate ATP-dependent transport of various hydrophobic anionic compounds, including camptothecins and methotrexate in liver canalicular membranes and other tissues (22). MRP1 and MRP3,but not MRP2, are expressed in normal colorectal mucosa (23, 24). Introducing MRP2 antisense cDNA into human hepatic cancer cell lines results in increased sensitivity to cisplatin, vincristine, doxorubicin, and camptothecin derivatives (25).

The complete cDNA sequence of human MRP3 (23)has a 56% amino acid identity to MRP1 and a 45% identity to MRP2. MRP3 was expressed mainly in the liver and was expressed to a lesser extent in the colon, small intestine, and prostate. Transfection of MRP3 cDNA confers drug resistance against etoposide and methotrexate (26). MRP3,like MRP1, may contribute to drug resistance against etoposide, vincristine, and cisplatin in human lung cancer cells (27). In this study, we investigated which transporter of MDR1, MRP1, MRP2, and MRP3is specifically expressed and is responsible for the responses to the anticancer agents in colorectal carcinomas. We determined the mRNA levels of MDR1, MRP1, MRP2, and MRP3 in tissues from 45 patients with colorectal cancer. The enhanced expression of MRP2 in cancerous regions compared with noncancerous regions is discussed in association with resistance to anticancer agents.

Between June 1994 and January 1999, 45 Japanese men and women with colorectal cancer underwent colorectal resection at the Department of Surgery II, Kyushu University Hospital (Fukuoka,Japan). Table 1 shows the clinicopathogenic characteristics of these 45 patients. No patient had received chemotherapy before surgery. Tumor and normal mucosal tissue samples were obtained after subjects provided informed consent, frozen in liquid nitrogen, and stored at −80°C until RNA extraction.

For quantitative RT-PCR, we used real-time TaqMan technology and a Model 7700 sequence detector (Perkin-Elmer Applied Biosystems,Foster City, CA) as described previously (28). Four primer pairs and four TaqMan probes for MDR1, MRP1, MRP2, and MRP3 were designed using the primer design software Primer Express (Perkin-Elmer Applied Biosystems). To avoid amplifying contaminated genomic DNA, primer pairs were placed in a different exon, and the probe was placed at the junction between two exons. Primers for GAPDH (TaqMan GAPDH control reagent kit)were purchased from Perkin-Elmer Applied Biosystems. Sequences for the TaqMan probes and primers were as follows: (a) MDR1, sense primer 5′-TGCTCAGACAGGATG TGAGTTG-3′, antisense primer 5′-TAGCCCCTTTAACTTGAGCAGC-3′, and probe 5′-AA AACACCACTGGAGCATTGACTACCAGGC-3′; (b) MRP1, sense primer 5′-TACCTCCTGT GGCTGAATCTGG-3′, antisense primer 5′-CCGATTGTCTTTGCTCTTCATG-3′, and probe 5′-ATGGCGATGAAGACCAAGACGTATCAGGTG-3′; (c) MRP2,sense primer 5′-CAAACTCTATCTTGCTAAGCAGG-3′, antisense primer 5′-TGAGTACAAGGGCCAGCTCTA-3′, and probe 5′-TTCGTTGGTTTTCTTCTTATTCTAGCAGCC-3′; and (d) MRP3, sense primer 5′-CTTAAGACTTCCCCTCAACATGC-3′, antisense primer 5′-GGTCAAGTTCCTCTTGGCTC A-3′, and probe 5′-AGTGTGTCTCTGAAACGGATCCAGCAATTC-3′. A hybridization probe specific for each PCR product was labeled with a reporter fluorescent dye(6-carboxyl-fluorescein or 2,7-dimethoxy-4,5-dichloro-6-carboxy-fluorescein) at the 5′ end and a quencher fluorescent dye (6-carboxy-tetramethyl-rhodamine) at the 3′end. Reaction solutions (50 μl) contained 5 μl of 10× reverse transcription sample prepared as described previously (29), 5 μl of each 3 μm primer pair, 5 μl of 2 μl TaqMan probe, 5 μl of dH₂O,and 25 μl of 1× PCR mix (1.25 units of Ampli-Taq DNA polymerase, 1×PCR reaction buffer, and 0.5 unit of amperase; Perkin-Elmer Applied Biosystems). The C_Tvalues corresponding to the cycle number at which the fluorescent emission monitored in real time reached a threshold of 10 SDs above the mean baseline from cycles 1–15 were measured. Serial 1:10 dilutions of plasmid DNA were analyzed for each target cDNA. These served as standard curves from which we determined the rate of change of threshold cycle values. Cycling parameters were as follows: 2 min at 50°C and 10 min at 95°C followed by 40 cycles of 30 s at 95°C and 2 min at 60°C.

Resected specimens of colorectal cancer were fixed in 10%formaldehyde, processed, and embedded in paraffin (30). Five-μm-thick sections were cut and stained immunohistochemically using the avidin-biotin-peroxidase complex method with mouse monoclonal antibodies for Pgp (C219) and MRP1 (QCRL-1) and rabbit polyclonal antibodies for MRP2 (25) as described previously.

The SDI test was performed using the methods described previously (31). In brief, tumor tissues were cut with scissors into fragments that were put into a sterile flask containing a mixture of pronase (protease type XXV; Sigma, St. Louis, MO) and collagenase (type 1; Sigma) in McCoy’s 5A (Life Technologies, Inc.) solution with antibiotics. Enzymatic disaggregation was performed for 20 min at 37°C with gentle stirring and was ended by adding sufficient amounts of MEM. Aliquots (100 ml) of this single cell suspension (3.0 ×10⁵ cells/ml) were dispensed into 96-well microtiter plates and incubated at 37°C in a humidified 5%CO₂ atmosphere for 3 days in the presence of anticancer drugs. Each anticancer drug was tested at 10× the peak plasma concentration. The chemosensitivity to a certain drug is given as a percentage of the succinate dehydrogenase activity in drug-treated cells compared with control cells. Chemosensitivity was defined as sensitive or resistant to the drug when the succinate dehydrogenase activity of the drug-treated cells decreased to less than 50% or more than 76% of the control, respectively.

The correlations between groups were determined using Spearman’s test. Spearman’s test is usually used for nonparametric analysis when it is unclear whether the variables show normal distribution. Probability values of less than 0.05 were significant. The Spearman’s correlation coefficient (r) and associated probability(P) were calculated for each combination of mRNA and SDI data sets. The relationships between the MRP2 mRNA expression level and the drug sensitivity data sets were calculated using the same method. Four groups of T:N (tumor:normal) mRNA, the mRNA expression level of the cancerous region divided by the expression level of the noncancerous region, were compared using repeated-measures ANOVA with Bonferroni’s correction for multiple comparison. Probability values of less than 0.05 were statistically significant. We determined that a strong correlation would have a r value of 0.7 or above and that a weak correlation would have a rvalue of less than or equal to 0.5.

To compare the mRNA levels of MDR1, MRP1, MRP2, and MRP3 in human colorectal carcinomas and adjacent noncancerous tissues, we examined surgically removed colorectal samples from 45 patients using quantitative RT-PCR with specific primers and probes. The data were standardized against GAPDH mRNA levels of both cancerous and noncancerous regions. Fig. 1 shows the expression of ABC transporters MDR1, MRP1, MRP2, and MRP3 in both noncancerous regions and cancerous regions. In noncancerous regions, MRP3 mRNA expression was the highest,and MDR1 and MRP1 mRNA expressions were moderate among the four transporter genes. Very low or little expression of MRP2 was seen in noncancerous regions (Fig. 1). The expression of MDR1 and MRP3 mRNA decreased,whereas MRP1 mRNA expression was unchanged in cancerous regions compared with noncancerous regions. However, the expression of the MRP2 gene was significantly increased in the cancerous tissues compared with noncancerous regions (Fig. 1).

In the noncancerous tissues, MRP1, MRP2, and MRP3 mRNA expression showed a significant correlation with each other, and MRP1 mRNA expression had the strongest correlation with MRP3 mRNA expression (Table 2). In the cancerous tissues, all pairwise comparisons of mRNA expression showed significant correlations, and the MRP1 mRNA level also showed the strongest correlation with the MRP3 mRNA level (Table 2).

We next determined the relative expression of these four ABC transporter genes in cancerous tissue compared with noncancerous tissue. Table 3 summarizes the data of Fig. 1. Of the four ABC transporters, only MRP2 showed a significant (P < 0.05) increase in the cancerous regions compared with their counterparts. Moreover, 19 of 45 patients(42.2%) showed a >3-fold increase in MRP2 expression in cancerous regions compared with noncancerous regions. However, only 1–7 of 45 patients (2.2–15.6%) showed a >3-fold increase in expression of the other ABC transporters, MDR1, MRP1, and MRP3 (Table 3).

Fig. 2 shows the results of immunostaining analysis with anti-Pgp and anti-MRP2antibodies. Positive staining for Pgp and MRP2 was found primarily in the cytoplasm and cytoplasmic membrane of cancer cells. Each case also had increased expression of Pgp and MRP2in the cancerous region compared with adjacent noncancerous tissue(Fig. 2). Immunostaining with antibodies against Pgp and MRP2in the other five samples showed that MDR1 and MRP2 mRNA levels were comparable to Pgp and MRP2 protein levels in these cases (data not shown).

To investigate the effects of up-regulation of MRP2expression in response to anticancer agents, we searched for correlations between MRP2 mRNA expression and drug response to various anticancer agents, including cisplatin, 5-fluorouracil,mitomycin C, doxorubicin, etoposide, and a camptothecin derivative(CPT-11). The averages of succinate dehydrogenase activities were 67.5 ± 20.8% for cisplatin, 77.2 ± 20.1% for 5-fluorouracil, 71.6 ± 22.4% for mitomycin C, 83.9 ±18.2% for doxorubicin, 66.1 ± 30.0% for etoposide, and 88.0 ± 11.0% for CPT-11. We previously proposed that succinate dehydrogenase activities of <50% indicated drug sensitivity, whereas succinate dehydrogenase activities of >76% indicated drug resistance (31). According to these criteria, the colorectal carcinomas appeared to be partially sensitive to cisplatin,5-fluorouracil, mitomycin C, and etoposide but was resistant to CPT-11 and doxorubicin. Spearman’s correlation coefficients (r)and associated probabilities (P) were calculated to determine whether expression of these MRP-related genes was correlated with drug response of the 45 studied samples to cisplatin,5-fluorouracil, mitomycin C, and doxorubicin (Fig. 3). The tissue samples showed a wide range of drug sensitivities. The relationships among MRP2mRNAs and the drug response to the six drugs were determined in six paired comparisons, and the probabilities were calculated to correct for multiple comparisons. Taking into account the number of comparisons and the variability inherent in PCR and SDI testing, we used a level of significance of 0.05 for each comparison, recognizing that any relationship identified as significant would require subsequent confirmation in independent studies. MRP2 mRNA expression showed a significant correlation with drug response to cisplatin (Fig. 3). We could not recognize significant correlations among the mRNA expressions of the other three ABC transporters and the drug responses to the six anticancer agents (data not shown).

We first compared the mRNA levels of four ABC transporter family genes, MDR1, MRP1, MRP2, and MRP3, in carcinoma regions and adjacent noncancerous tissue in 45 patients with colon cancer, using a quantitative RT-PCR assay. Consistent with previous studies (23, 26, 32, 33), MDR1, MRP1, and MRP3 were intrinsically expressed in normal colon mucosa tissue (Fig. 1). However, in colorectal cancers,either a decrease or no change in MDR1 mRNA levels occurred(Table 3), findings consistent with previous studies (8, 34). We further observed a decreased expression of MRP3in colon cancer regions compared with noncancerous regions (Table 3). Decreased expression of both MDR1 and MRP3 in cancerous regions may perhaps lead to a sensitivity to anticancer agents targeted by both transporters.

MRP1 and MDR1 are not coexpressed in colon cancers (12, 20), as we also observed (Table 2). In contrast, MRP1 and MRP3 were coexpressed,suggesting coordinate regulatory mechanisms for both ABC transporter genes in colorectal tissues and colorectal cancers. MRP2 is not intrinsically expressed in normal colorectal mucosa (24). Unlike MDR1, MRP1, and MRP3,we observed up-regulation of MRP2 in cancerous regions compared with noncancerous regions in colorectal carcinoma tissues(Table 3). MRP2 might be involved in the membrane transport of endogenous substrates, such as glutathione and glutathione conjugates, in cancerous regions of colorectal cancers. The manner in which MRP2 is specifically up-regulated in malignant colorectal tumors remains to be determined.

Of the six anticancer agents prescribed for treatment of colorectal carcinomas, we observed a significant correlation between MRP2 mRNA levels and drug response to cisplatin (r = 0.3236; P = 0.0321). MRP2 levels thus appeared to be more closely correlated with response to cisplatin than with responses to 5-fluorouracil, mitomycin C, doxorubicin, etoposide,and CPT-11. Drug sensitivities to these agents were assayed using the SDI test. SDI tests can often predict the drug sensitivity of cancer cells in patients (35), but the fact can also be argued that SDI testing does not always reflect in vivo drug sensitivity (31). This study suggests that MRP2could be a sensitivity marker for cisplatin in patients with colorectal carcinoma. Expression of MRP2 increased in cisplatin-resistant cell lines from various human tumor types,including colon cancer (21). Drug sensitivity to cisplatin, as well as vincristine and camptothecins, is also increased in stable transfectants of antisense MRP2 cDNA in hepatic cancer cell lines (25). Sensitivity to vinblastine changes in MRP2 cDNA-transfected canine kidney cell lines (36), transfection of MRP2 cDNA confers drug resistance against methotrexate and camptothecin derivatives (37, 38, 39), and stable transfection of MRP2 cDNA into both polarized and nonpolarized cells results in acquisition of drug resistance against cisplatin as well as vincristine,camptothecins, and methotrexate (36, 40, 41). However, it is unclear whether the expression of MRP2 is directly associated with drug sensitivity to cisplatin, and whether MRP2 can increase the transport of cisplatin itself, the cisplatin-glutathione conjugate, and both glutathione and cisplatin is unclear. Further study of the molecular basis of the GS-X pump for cisplatin, as originally proposed by Ishikawa et al.(42), is required.

In human lung cancer cell lines, MRP mRNA levels correlate with resistance to doxorubicin (43). A strong correlation of MRP3 mRNA levels with drug resistance against doxorubicin exists in lung cancer cell lines (27). Moreover, mRNA levels of both MRP1 and MRP3correlate with resistance against vincristine, etoposide, and cisplatin in human lung cancer cell lines (27). In this study, MDR1, MRP1, and MRP3 were not correlated with sensitivity to anticancer agents in colorectal cancers.

In conclusion, among MDR1, MRP1, MRP2,and MRP3, only MRP2 was up-regulated in malignant colorectal tumors and correlated with resistance to cisplatin when surgically resected clinical samples from 45 patients were analyzed. Increased expression of MRP2 might cause resistance to cisplatin in patients with human colorectal cancers.

We thank Morimasa Wada, Sei Haga, and Takanori Nakamura for helpful discussion.