ABCC2 (MRP2, cMOAT) Can Be Localized in the Nuclear Membrane of Ovarian Carcinomas and Correlates with Resistance to Cisplatin and Clinical Outcome Free

Purpose: Cisplatin resistance is a major obstacle in the treatment of ovarian carcinoma. ABCC2 is commonly localized in apical cell membranes and could confer cisplatin resistance. Here, we show that ABCC2 can be localized in the cytoplasmic membrane as well as in the nuclear membrane of various human tissues including ovarian carcinoma cells.

Experimental Design: For the subcellular detection of ABCC2, immunohistochemistry was done using 41 Federation Internationale des Gynaecologistes et Obstetristes stage III ovarian carcinoma specimens prepared before treatment with cisplatin-based schemes and 35 specimens from the same group after chemotherapy. Furthermore, 11 ovarian carcinoma cell lines as well as tissue microarrays consisting of various human tissues were analyzed.

Results: Nuclear membranous localization of ABCC2 was associated with response to first-line chemotherapy at primary (P = 0.0013) and secondary surgery (P = 0.0060). Cases with relapse showed higher nuclear membrane expression at primary (P = 0.0003) and secondary surgery (P = 0.0024). Kaplan-Meier analyses showed that weak nuclear membrane ABCC2 expression before treatment was associated with significantly longer overall (P = 0.04) and progression-free survival (P = 0.001); following chemotherapy, it correlated with significantly longer progression-free survival (P = 0.038). Tissue microarrays confirmed nuclear membranous localization of ABCC2, in particular, in poorly differentiated cells. In ovarian carcinoma cells, it correlated with resistance against cisplatin, whereas localization in the cytoplasmic membrane did not.

Conclusions: ABCC2 confers resistance to cisplatin of ovarian carcinoma in cell culture systems and in clinics when expressed in the nuclear membrane. Thus, ABCC2 localization can predict platinum therapy outcome. Furthermore, expression of ABCC2 in nuclear membranes in human tissues is specific for poorly differentiated cells including stem cells.

Ovarian carcinoma is the leading cause of death among all gynecologic neoplasms. About 190,000 new cases and 114,000 deaths from ovarian cancer are estimated to occur annually (1). Because the majority of patients are diagnosed in an advanced stage of the disease, first-line therapy commonly consists of surgical resection, followed by platinum drug- and taxane-based chemotherapeutic regiments (2). Despite a high initial response rate, however, most patients will relapse. Unfortunately, second-line chemotherapeutic treatment merely achieves response rates of ∼20%. These clinical observations indicate that long-term prognosis of ovarian carcinoma depends on intrinsic and acquired drug resistance. Knowledge of the biological mechanisms mediating drug resistance may lead to new and more effective treatment strategies.

It is now well established that several members of the superfamily of ATP-binding cassette (ABC) transporters play an important role in drug resistance in tumor cell models as well as in the clinic (3). ABC transporters are found in all cells of all species: from the most primitive microorganism to man, and accordingly, they play central roles in various physiologic systems.

One of the 48 human ABC transporters involves ABCC2, also called the multidrug resistance–associated protein 2 (MRP2) or the canalicular multiple organic anion transporter (cMOAT). The ABCC2 gene is expressed in the apical membranes of canalicular cells in the liver (4). There, it functions as the major exporter of organic anions from the liver into the bile (5). Besides the expression in hepatocytes, ABCC2 is also localized in the apical membranes of kidney proximal tubules, epithelial cells of gall bladder, small intestine, colon, and lung (6).

In vitro experiments showed that overexpression of ABCC2 could confer resistance to platinum-containing anticancer drugs like cisplatin and carboplatin to cancer cell lines, including ovarian carcinoma cells (4, 7–9). Although expression of ABCC2 could be detected in clinical specimens of ovarian carcinoma, an immunohistochemical study using frozen tissue sections of tumors could not show a prognostic value of ABCC2 assessment for response to chemotherapy or progression-free survival (10). These data could be confirmed by results of a real-time reverse transcription-PCR study on mRNA level (11). This study showed that ABCC2 mRNA levels were not associated with clinical outcome after platinum-based chemotherapy in patients suffering from ovarian carcinoma. Another reverse transcription-PCR study showed that in patients with primary Federation Internationale des Gynaecologistes et Obstetristes (FIGO) stage III carcinomas, the absence of ABCC2 mRNA expression showed a tendency to be associated with progression-free survival (12). However, none of these studies considered the cellular localization of the ABC transporter, ABCC2, neither in ovarian carcinoma cell lines nor in tissue sections obtained from patients.

Cell culture and cell proliferation assay. Human carcinoma cells were grown in modified Leibovitz L-15 medium as described previously (9, 13, 14). The cisplatin-resistant cell line, A2780RCIS, was derived from the ovarian carcinoma cell line, A2780 (9). The human ovarian carcinoma cell lines CAOV-3, EFO 21, EFO 27, ES-2, Mdah 2774, OAW 42, OVCAR-3, PA-1, and SKOV-3 were kindly provided by Dr. Carsten Denkert (Institute of Pathology, Charité, Berlin, Germany). In order to ensure maintenance of cisplatin-resistant phenotype of A2780RCIS cells, the medium was supplemented with 10 μg/mL of cisplatin (33.3 μmol/L; GRY-Pharm, Kirchzarten, Germany). Chemoresistance was tested using a proliferation assay based on sulforhodamine B staining, as described previously (9, 13, 14).

Quantitative real-time reverse transcription-PCR. Quantitative real-time reverse transcription-PCR for ABCC2, and as control, porphobilinogen deaminase (PBGD) mRNA (GenBank accession no. NM 000190) were carried out using a LightCycler instrument and SYBR-Green fluorescent dye (Roche Diagnostics, Mannheim, Germany), as described previously (8, 9, 13). Expression levels are shown as relative values, whereby the ABCC2/PBGD expression ratio in A2780 was set to 1.0. Oligonucleotide primers used for the amplification of ABCC2 were described previously (8, 14); for the amplification of PBGD, oligonucleotide primers were PBGD-forward, 5′-GAGAAGAATGAAGTGGACCTGGTT-3′; and PBGD-reverse, 5′-GCGGGAACTTTCTCTGCAGC-3′. Cycling conditions were as follows: initial enzyme activation at 95°C for 10 minutes; followed by 40 cycles at 95°C for 15 seconds; 55°C (ABCC2) or 60°C (PBGD) for 5 seconds; and 72°C for 10 seconds. Fluorescence was measured at 83°C (ABCC2) or 85°C (PBGD). The specificity of the amplification products was confirmed by the melting curve analysis. Three independent experiments were done in triplicate.

Prediction of nuclear localization signals in ABCC2. The software “PredictNLS Online” (http://cubic.bioc.columbia.edu/cgi/var/nair/resonline.pl; ref. 15) was used for the prediction of potential nuclear localization signals within the ABCC2 amino acid sequence. The complete protein sequence of ABCC2 (GenBank accession number NP 000383) was submitted.

Patients with ovarian carcinoma and tumor samples. Immunohistochemical examination was done retrospectively on tissue samples taken for routine diagnostic purposes. The study included 41 patients with FIGO stage III ovarian carcinoma operated on from 1999 to 2002 at the Department of Gynaecology and Obstetrics, University Medical School in Poznañ, Poland. The cases were selected based on the availability of tissue and were not stratified for known preoperative or pathologic prognostic factors. The study was approved by an institutional review board and the patients gave their informed consent before their inclusion into the study. Following primary laparotomy, all patients were subjected to chemotherapy using platinum-based schemes (Table 1). Thirty-four patients were also subjected to a secondary cytoreduction. In five cases, no tumor cells were detected after cytoreduction. The patients were monitored by periodic medical check-ups, CA-125 serum levels, and ultrasonographic and radiological examinations. Clinical response to first-line chemotherapy was defined according to the criteria suggested by Response Evaluation Criteria in Solid Tumors guidelines (16). During the follow-up period, 21 (51%) patients had a recurrent disease, and 12 (29%) patients died of the malignancy. The mean progression-free survival time was 17.1 months (range, 0-52 months), whereas the mean overall survival time was 25.1 months (range, 6-52 months).

Tumor specimens were obtained during primary laparotomies and secondary cytoreductions. For histologic classification and immunohistochemical staining, tissue samples were fixed in 10% buffered formalin and embedded in paraffin. In each case, H&E-stained preparations were subjected to histopathologic evaluation by two pathologists. The stage of the tumors was assessed according to FIGO guidelines (17). Tumors were graded according to the Silverberg grading system (18).

Immunohistochemistry. Immunohistochemistry was done as described previously (19, 20). For the detection of ABCC2, a monoclonal mouse antibody (clone M₂I-4; Monosan, Uden, the Netherlands) was diluted 1:100 and used in the staining protocol. All reactions were done in triplicate.

Tissue microarrays. In order to examine cellular localization of ABCC2 in various human healthy organs and to check the specificities of antibodies by reproduction of the cellular localization of ABCC2 described previously, immunohistochemical reactions were done on two different tissue microarrays (TMA; kindly provided by Oligene GmbH, Berlin, Germany). One TMA contained 31 spots representing healthy human organs (human normal tissue I), whereas the other contained 70 spots of samples originating from 15 different healthy human organs, 6 spots for each of the organs (human normal tissue II). TMA staining was reproduced twice.

Immunocytochemistry. Immunostaining of ABCC2 was done using the complete panel of ovarian carcinoma cell lines and human hepatocellular carcinoma cell line HepG2, as described previously (13, 14). For controls, the cisplatin-resistant ovarian carcinoma cell line, A2780RCIS, was also subjected to reactions with additional anti-ABCC2 antibodies: (a) mouse monoclonal antibody “clone M₂III-6” (Alexis Biochemicals, Lausen, Switzerland) at a dilution of 1:200, (b) rabbit polyclonal antibody “H-300” (Santa Cruz Biotechnology, Santa Cruz, CA) at a dilution of 1:400, and (c) goat polyclonal antibody “H-17” (Santa Cruz Biotechnology) at a dilution of 1:400.

Scoring of immunostaining results. In the case of ovarian carcinoma specimens, the intensity of the immunohistochemical reactions was appraised using the semiquantitative immunoreactive score scale (21), in which the intensity of the reaction and the percentage of positive cells were considered. The final result represented a product of scores given for individual traits and ranged between 0 and 12. The intensity of the reactions was evaluated independently by two pathologists. In case of divergences, the evaluation was repeated using a double-headed microscope. In the case of immunostained TMAs and cancer cell lines, ABCC2-specific staining reactions were localized by two experienced histologists. Immunostaining reactions were evaluated using a simplified scale, providing merely a score for the intensities of the reactions (0, total absence of staining; 1, only faint staining; 2, moderate staining; and 3, strong, intense staining).

Electron microscopy. The cisplatin-resistant human ovarian carcinoma cell line, A2780RCIS, and human hepatocellular carcinoma, HepG2, were fixed in 4% paraformaldehyde in 0.1 mol/L of cacodylate buffer and contrasted with osmium tetroxide. Subsequently, the material was dehydrated and embedded in Epon 812 (Sigma-Aldrich, St. Louis, MO). Ultrathin sections were incubated in 0.5% bovine serum albumin and 0.05% Tween 20. Immunovisualization reactions were done using monoclonal antibodies (MRP2 clone M₂I-4) directed against ABCC2 (1:50, 24 hours at room temperature). Nuclear membrane localization of ABCC2 in A2780RCIS was detected using anti-mouse antibodies coated on colloidal gold (1:50, 1 hour at room temperature; Sigma). Each of the reactions were done in three independent experiments.

Statistical analyses. Statistical analysis of the results took advantage of Statistica 98 PL software (StatSoft, Tulsa, OK). The employed tests included Mann-Whitney's U test and ANOVA rank test of Kruskal-Wallis. Kaplan-Meier's statistics and log-rank tests were done using SPSS software (release 10.0; SPSS Inc., Chicago, IL) to estimate the significance of differences in survival times. The length of progression-free survival was defined as the time between the primary surgical treatment and diagnosis of a recurrent tumor or death. Because we found no significant relationships between the studied clinicopathologic variables (age, histology, grade) and overall survival and progression-free time of the studied patients using the univariate analysis, we did not perform a multivariate analysis. P < 0.05 were considered to indicate a significant relationship. In order to examine the relationship between ABCC2 expression and the sensitivity of ovarian cancer cells to cisplatin, the χ² test was employed. The effect of cisplatin exposure on the expression of ABCC2 was tested using Student's t test.

Correlation of cisplatin resistance and ABCC2 mRNA expression in ovarian carcinoma cells. The sensitivity of various human ovarian carcinoma cell lines against treatment with cisplatin was determined by the assessment of IC₅₀ values (Table 2). Between the most sensitive ovarian carcinoma cell lines, PA-1, exhibiting an IC₅₀ of 0.2 μmol/L, and the most cisplatin-resistant cell line, A2780RCIS, with an IC₅₀ of 16.5 μmol/L, an 82.5-fold difference in cisplatin sensitivity could be observed.

The ABCC2-specific mRNA expression level of 11 different human ovarian carcinoma cell lines was determined by quantitative real-time reverse transcription-PCR analysis (Table 2; Supplementary Fig. S1). The ovarian carcinoma cell line A2780RCIS, displaying the highest IC₅₀ value for cisplatin and a strong nuclear membrane staining for ABCC2 protein, also showed the highest expression level of the ABCC2 encoding mRNA. The human ovarian cancer cell lines PA-1, OAW 42, Mdah 2774, ES-2, A2780, and A2780RCIS showed a good correlation between cisplatin sensitivity, expressed by the IC₅₀ value, and ABCC2 mRNA expression level. Nevertheless, some cell lines, e.g., SKOV-3, displayed a higher IC₅₀ value but no elevated ABCC2 mRNA expression, suggesting the involvement of additional mechanisms contributing to the cisplatin-resistant phenotype.

Correlation between cisplatin resistance and subcellular ABCC2 protein localization in ovarian carcinoma cells. For the detection of ABCC2 transporter proteins in ovarian carcinoma cells, immunocytochemistry was used. As expected, in the cases of the cell lines CAOV-3, EFO 27, ES-2, and Mdah 2774, ABCC2 was found to be expressed in the plasma membrane (Table 2). Most interestingly, the immunocytochemical experiments showed that ABCC2 could also be detected in nuclear membranes of the A2780, A2780RCIS, EFO 21, and SKOV-3 cell lines (Table 2).

In particular, in A2780RCIS cells, exhibiting the highest level of cisplatin resistance, light microscopy analyses showed that ABCC2 was predominantly localized in the nuclear membrane and at a much lower level in the cytoplasmic membrane (Fig. 1B). Control reactions using other antibodies confirmed the localization of ABCC2 in the nuclear membrane (Supplementary Fig. S2A-C). Furthermore, the localization of ABCC2 in the nuclear membrane was confirmed by electron microscopy (Fig. 1C; Supplementary Fig. S2D). For controls, the cytoplasmic membrane localization of ABCC2 in the human hepatocellular carcinoma cell line HepG2 was confirmed by electron microscopy (Supplementary Fig. S2E and F).

The subcellular localization of ABCC2 was correlated with the cisplatin sensitivity of ovarian carcinoma cells (Table 2). The expression of ABCC2 in the nuclear membrane significantly correlated with cisplatin sensitivity (χ² test, P = 0.023), whereas ABCC2 localization in the cytoplasmic membrane did not show a statistically significant correlation with the cisplatin-resistant phenotype (χ² test, P = 0.15).

In silico analysis of nuclear localization signals within the ABCC2 sequence. Analysis of the ABCC2 protein sequence using the PredictNLS Online software showed that the first 293 amino acids were followed by the nuclear localization signal (KKKKKSGTKK), starting at amino acid 294. Sequences with the general KKKKKx{3,6}KK motif are recognized to represent a nuclear localization signal. The analysis showed that a nuclear localization of ABCC2 could be derived from the amino acid sequence.

Effect of cisplatin on ABCC2 expression in the cisplatin-resistant cell line, A2780RCIS. To evaluate whether ABCC2 expression was induced by cisplatin, ABCC2 was analyzed on mRNA and protein levels in the cisplatin-resistant cell line A2780RCIS by quantitative real-time reverse transcription-PCR and immunocytochemistry after treatment for 3 days with 20 μg/mL of cisplatin (66.6 μmol/L). In A2780RCIS cells without cisplatin treatment, the ABCC2 mRNA expression level was 5-fold higher, as compared with parental, drug-sensitive A2780 cells (Fig. 1A). Exposure to cisplatin enhanced the ABCC2 mRNA level 7-fold (P = 0.015), indicating that ABCC2 could be induced by cisplatin treatment. Immunocytochemical analysis of ABCC2 protein expression showed a weak staining reaction in A2780 cells, a moderate ABCC2 expression in nuclear membranes in untreated A2780RCIS cells as well as a strong staining signal in nuclear membranes following cisplatin treatment of A2780RCIS cells (Fig. 1B).

Subcellular localization of ABCC2 in various human tissues. For examination of subcellular localization of ABCC2 in various healthy human tissues, two different TMAs were stained in triplicate for ABCC2 expression. In polarized cells (e.g., hepatocytes, cells of renal tubules, or exocrine part of pancreas), a cytoplasmic membranous reaction could be detected (Fig. 2); whereas in many other cells, the staining reaction was predominantly localized in the nuclear membrane (Fig. 2). In particular, there was a tendency for the less differentiated cells to show a more pronounced nuclear membranous localization of ABCC2 (Table 3).

Subcellular localization of ABCC2 in patients with ovarian carcinoma. To investigate where ABCC2 is expressed in ovarian cancer tissues, tumor specimens were analyzed by immunohistochemistry. The ABCC2-specific staining reactions showed a subcellular localization of ABCC2 in the cytoplasmic membrane as well as in the nuclear membrane (Fig. 2K and L). The localization and expression levels of ABCC2 varied in individual cases (Supplementary Table S1).

Differences in ABCC2 expression levels between primary laparotomies and secondary cytoreductions. Mann-Whitney's U test was used to compare the staining intensities of ABCC2 expression in different membranes at primary laparotomies and secondary cytoreductions. The analysis showed no statistically significant differences in the distribution of ABCC2 expression of cytoplasmic membrane localization (P = 0.74) and nuclear membrane localization (P = 0.87) at both time points.

Relation between immunostaining for ABCC2 and grade, response to first-line chemotherapy, relapse, and death. Using the ANOVA rank test of Kruskal-Wallis, relations were analyzed between the intensities of ABCC2 expression in different membranes on one hand, and tumor grade, response to first-line chemotherapy, relapses, and cases of death on the other. It could be shown that none of the examined variables were related to the grade of the tumor (Supplementary Table S2). At both time points, nuclear membrane expression of ABCC2 was most pronounced in cases of progressive disease.

In the case of cytoplasmic membrane localization of ABCC2, no relationship between clinical response to the first-line chemotherapy and immunostaining intensity could be shown (Supplementary Table S2). In contrast, nuclear membrane localization of ABCC2 was highly significantly associated with adverse response to first-line chemotherapy at primary laparotomy (P = 0.0013), and at secondary surgery (P = 0.006; Supplementary Table S2).

Ovarian carcinoma cases with relapse showed higher nuclear membrane expression of ABCC2 at primary laparotomy (P < 0.001), and at secondary surgery (P = 0.0024), compared with cases without relapse. No such relationship could be found in the case of cytoplasmic membrane localization of ABCC2 at both postsurgery time points (Supplementary Table S2).

In specimens obtained from patients after their death, nuclear membrane expression of ABCC2 was significantly higher at primary laparotomy (P = 0.0087) and at secondary cytoreduction (P = 0.011) than in surviving patients. In the case of cytoplasmic membrane staining, ABCC2 expression at primary laparotomy was specific for defunct patients (P = 0.04; Supplementary Table S2).

Survival analyses. For Kaplan-Meier analyses, a cutoff value of immunostaining score 3 was used for all calculations. Accordingly, the overall survival and the progression-free survival times were examined for groups manifesting the expression of immunostaining score 0 to 2 (no or very low expression) and those manifesting an immunostaining score of 3 to 12 (higher expression intensities). The calculations showed that cases of no or low nuclear membrane ABCC2 expression at primary laparotomy showed a significantly longer overall survival time (P = 0.01) and a significantly longer progression-free survival time (P < 0.001; Fig. 3). For nuclear membrane ABCC2 expression at secondary cytoreductions (P = 0.09 and P = 0.06) and for cytoplasmic membrane expression of ABCC2 at primary laparotomies (P = 0.12 and P = 0.37) and secondary cytoreductions (P = 0.61 and P = 0.75), there was no significant correlation with overall survival time and progression-free time (Fig. 3).

Resistance to platinum-containing anticancer drugs is a major limitation for the successful chemotherapeutic treatment of ovarian carcinoma. An improved understanding of the biological mechanisms mediating platinum resistance will provide molecular targets for therapeutic intervention, may facilitate prediction of response, and provide the basis for individually designed therapy. In the last few years, various ovarian carcinoma–derived in vitro models have been analyzed to identify the molecular mechanisms mediating cisplatin resistance (4, 7, 9). These studies showed that platinum resistance of ovarian carcinoma cell lines could be associated with overexpression of the ABC transporter ABCC2.

In the cisplatin-resistant ovarian carcinoma cell line, A2780RCIS, inhibition of nuclear membranous protein expression by treatment with different anti-ABCC2 hammerhead ribozymes could increase the cisplatin-induced formation of platinum-DNA adducts. In this way, DNA adduct–dependent apoptotic signals were triggered, and the cisplatin-resistant phenotype was reversed (9, 14). Furthermore, RNA interference technology was applied for the specific inhibition of nuclear membranous ABCC2 expression in A2780RCIS cells (13). Likewise, the disappearance of the nuclear membranous ABCC2 staining signal and the reversal of the cisplatin-resistant phenotype was confirmed. On the other hand, transfection of the drug-sensitive ovarian carcinoma cell line, A2780, with the ABCC2-encoding cDNA could protect ovarian cancer cells from cisplatin-induced formation of platinum-DNA adducts and therewith increased the level of cisplatin resistance.⁵

In accordance with these data, this study could show that in various ovarian carcinoma–derived cell lines, the level of ABCC2 expression showed a tendency to correlate with the cisplatin-resistant phenotype. However, it has to be considered that in these cell models, alternative biological mechanisms may contribute to the cellular sensitivity against platinum drugs.

Because there was no significant variation of the ABCC2 localization in the tumors at the two time points of surgery (secondary cytoreductions were done after chemotherapy), but there was a clear induction of nuclear membrane expression of ABCC2 following cisplatin treatment of the drug-resistant ovarian carcinoma cell line, A2780RCIS, apparently, there is a discrepancy between the in vitro data and the clinical situation. This phenomenon may be explained by the much higher drug concentration used in the in vitro situation for treatment of already cisplatin-resistant ABCC2-positive cells. Furthermore, in the last few years, different new models have been discussed for describing treatment failure in clinics, e.g., the model of quiescent cells (22), taking into account that anticancer drugs primarily attack proliferating cells. The tumor is much more heterogeneous than the in vitro model and consists of cycling cells and a significant contingent of cells in a quiescent state. The cancer cells in phase G₀ exhibit a higher level of drug resistance relative to proliferating cells. Thus, the tumor consists of more or less different cisplatin-resistant or -sensitive cells and it is difficult to show the strict effects of cisplatin exposure. However, the experimental data show that nuclear membranous expression of ABCC2 can be induced in ovarian carcinoma cells by cisplatin treatment.

For confirmation of ABCC2's expression patterns described in other studies, immunostainings were done on different TMAs containing various healthy human tissues. By this approach, the typical ABCC2 expression pattern in the apical cytoplasmic membranes of hepatocytes, gall bladder epithelium, kidney proximal tubules, and epithelial cells of small intestine and colon could be confirmed. However, it should be taken into consideration that in those tissues, ABCC2 is occasionally also localized in nuclear membranes; e.g., the nuclear membrane staining intensity in epithelial cells of the gall bladder spotted on the TMAs is similar to that shown in an alternative study (23). Although that study showed, similar to the TMA staining in this study, a predominantly plasma membrane localization of ABCC2, a weak but unequivocal ABCC2-specific signal was also detected in the nuclear membranes.

Besides consistent ABCC2-specific staining data, some apparent discrepancies in ABCC2 expression could be found, comparing the results of this study with the data reported in an alternative TMA study (24). For example, in the latter study, no expression of ABCC2 could be detected in the endometrium. In contrast, in the present study, ABCC2 expression was found in the endometrium, both in the nuclear membrane and in the cytoplasmic membrane. However, the authors of the former study remarked that “the tumor cell plasma was distinctly stained with all antibodies (directed against ABCC2). Occasionally the nucleus was stained with both antibodies.” Likewise, apparently different observations were reported on the pancreatic tissue. In this study, we could show unambiguous ABCC2-specific staining reactions in nuclear and plasma membranes of various pancreatic cells. In pancreatic islands, which are rich in stem cells, the presence of the antigen was shown in cytoplasmic membrane, and in exocrine (polar) cells both in cytoplasmic and in nuclear membranes. Another study showed a clear expression of the ABCC2 encoding mRNA in normal human pancreatic tissue samples as well as in different specimens of pancreatic carcinoma (25), but failed to detect ABCC2 in cryosections of normal human pancreas. However, apart from different tissue preparations, both studies used different antibodies to detect ABCC2, i.e., polyclonal EAG5 antisera or the monoclonal mouse antibody, M₂I-4. Taken together, the discrepancies are merely apparent and may be due to different experimental techniques and interpretations of the data. However, such differences in the interpretation of expression data of ABC transporters are not novel, and they have already been discussed extensively in the case of detection of the multidrug resistance–associated ABC transporter ABCB1 (MDR1/P-glycoprotein; ref. 26).

A further interesting observation is that ABCC2 is localized in cytoplasmic membranes of keratinocytes in the granular layer of the skin but is localized in the nuclear membranes of poorly differentiated cells of the basal layer. Moreover, ABCC2 is expressed in the plasma membranes of vascular endothelial cells in the testis, forming a putative essential part of the blood/testis barrier, but in the nuclear membranes of poorly differentiated testicular cells important for spermatogenesis. Likewise, in the blood/brain barrier, neurons show a predominant ABCC2-specific staining intensity in nuclear membranes, whereas in astrocytes, ABCC2 is localized in the plasma membrane. Thus, ABCC2 is predominantly localized in the plasma membrane in cells involved in the transport of metabolites or in barrier cells. In poorly differentiated, more intensely dividing cells, ABCC2 is predominantly localized in the nuclear membrane. This interpretation is in line with the observation that alternative ABC transporters involved in drug resistance are components of poorly differentiated stem cells (27). Thus, our studies showed that ABCC2 expression in plasma membrane is typical of highly differentiated polar cells, whereas its expression in the nuclear membrane is typical for poorly differentiated cells, most probably including stem cells. Expression of ABC transporters, i.e., ABCB1 and ABCG2, has already been described in stem cells (27, 28). The transporters are supposed to protect stem cells and to inhibit apoptosis until the cells receive differentiation-promoting signals. Expression of the ABC transporters in cancer stem cells is also thought to provide reasons for drug resistance of the relapsed tumor.

Using various antibodies and fixation techniques, we have shown by light microscopy and electron microscopy that ABCC2 may be present in the nuclear membrane of poorly differentiated cells, including stem cells. In cells of surface epithelium of normal ovaries, ABCC2 was detected in nuclear membranes. Thus, the presence of ABCC2 in nuclear membranes of ovarian cancer might represent a natural sequel of its expression in the normal epithelium. Manifestation of ABCC2 in most ovarian cancers in the nuclear membrane may also indicate that ovarian cancers develop from poorly differentiated cells of the surface epithelium. The effect of nuclear membrane localization of ABCC2 for cisplatin resistance has been shown in vitro and in specimens of ovarian cancer. The Kaplan-Meier analyses have also shown that plasma membrane localization of ABCC2 has no statistically significant influence on overall or progression-free survival time. These findings are in line with an alternative study investigating the potential effects of ABCC2 in ovarian cancer (10). In contrast, nuclear membranous localization of ABCC2 has shown a significant effect on overall and progression-free survival times before chemotherapy using cisplatin. After platinum therapy, nuclear ABCC2 expression has a significant influence on progression-free survival and a nearly significant effect on overall survival time. These observations support the importance of nuclear ABCC2 for cellular protection against cytotoxic agents and have relevance for designing treatment regimens in patients suffering from ovarian carcinoma.

In conclusion, nuclear membrane localization can predict platinum drug therapy outcome and survival of patients with ovarian carcinoma. Thus, the determination of ABCC2 expression and subcellular localization should be considered in clinical practice. Furthermore, the expression of ABCC2 in nuclear membranes in healthy human tissues is specific for poorly differentiated cells, including stem cells, and its expression in plasma membrane is typical for differentiated polar cells.