ABCC Drug Efflux Pumps and Organic Anion Uptake Transporters in Human Gliomas and the Blood-Tumor Barrier

The therapy of gliomas, the most common brain neoplasms of adults, is often impeded by their high resistance to chemotherapeutic agents although some entities such as malignant oligodendrogliomas and oligoastrocytomas exhibit response rates of up to 70% (1). Many different mechanisms may account for this chemoresistance, including reduced blood supply to the tumor, up-regulation of antiapoptotic pathways, enhanced DNA repair mechanisms, and increased metabolic inactivation and subsequent elimination of the applied drugs (2). Moreover, in recent years, the expression and function of ATP-driven export pumps of the ATP-binding cassette (ABC) protein superfamily have been extensively studied in cancer cell lines, in in vitro models, and in tissues. Besides transporting a variety of physiologic substrates, several of these transporters also confer resistance to cytotoxic and antiviral drugs (3). Entry of many hydrophobic cationic drugs into the brain is decisively limited by multidrug resistance 1 (MDR1) P-glycoprotein (symbol ABCB1) in the blood-brain barrier (4). The presence of other ABC family members in the blood-brain barrier of human brain, including breast cancer resistance protein, ABCG2 (5), and efflux proteins of the ABCC subfamily (6), which have other substrate specificities than MDR1 P-glycoprotein (7), may add to the resistance of the brain to cytotoxic and antiviral drugs, especially of organic anions conjugated with sulfate, glucuronate, or glutathione, and of nucleoside analogues.

In gliomas, mainly MDR1 P-glycoprotein and ABCC1 have been studied thus far. Both were localized to the tumor vasculature and partially also to the tumor cells, however, with varying results (8–12). In glioma cell lines, MDR1 P-glycoprotein and ABCC1 were shown to confer resistance to various anticancer drugs (9, 12). Because several other ABCC subfamily members confer resistance to cytotoxic drugs as well, such as thioguanine and 6-mercaptopurine in case of ABCC5 (13) or methotrexate and topotecan in case of ABCC4 (14, 15), we were interested in their expression patterns in human gliomas, particularly with respect to the histologic subtype. Recently, we localized ABCC4 and ABCC5 in endothelial cells and astrocytes, but not in oligodendrocytes, of perilesional human brain samples (6). Given the possibility that gliomas originate by dedifferentiation of mature glial cells (16), the question arises whether the tumor cells retain the ABCC expression pattern of the cell type they derive from while undergoing malignant progression, which could account for the differential chemoresistance of the different glioma entities. Another possible mechanism causing insufficient drug delivery to the brain and its neoplasms is a reduced expression of uptake transporters in the blood-brain barrier or blood-tumor barrier or in the glioma cells. Members of the family of organic anion transporting polypeptides (OATP) transport chemotherapeutic agents, such as methotrexate (17) and SN-38, the active metabolite of irinotecan (18). OATPs are sodium-independent uptake transporters for various amphiphilic anionic compounds, including bile salts, prostanoids, steroid conjugates, thyroid hormones, anionic peptides, and xenobiotics (19). Several Oatps have been identified in the blood-brain barrier of rat brain (20). Because many rat Oatps do not have a human orthologue (19), it is of considerable interest to identify OATPs in the human brain. Thus far, only OATP1A2 protein has been localized in human brain (21).

To explore the possible role of ABCC export pumps and OATP uptake transporters in the resistance of gliomas, we investigated the expression and cellular distribution of six members of the ABCC subfamily (ABCC1-ABCC6) in frozen glioma samples of different histologic subtypes and of six members of the OATP family (OATP1A2, OATP1B1, OATP1B3, OATP1C1, OATP2B1, and OATP4A1). ABCC4 and ABCC5 were the major ABCC subfamily members expressed in the blood-tumor barrier and in the glioma cells, notably in astrocytic tumors. This is in good agreement with the overall higher chemoresistance of these tumors compared with oligodendroglial and mixed gliomas. Both ABCC proteins may therefore contribute to the MDR phenotype of gliomas. In addition, OATP1A2 and OATP2B1 proteins were detected in the blood-brain barrier and the blood-tumor barrier.

Antibodies and materials. The monoclonal antibodies (mAb) QCRL-1 and MRPr1 against ABCC1, M2III-6 and M2II-12 against ABCC2, M3II-9 against ABCC3, M5I-1 against ABCC5, M6II-31 against ABCC6, and BXP-21 against ABCG2 and the polyclonal antibody against ABCC4 (MRP4 polyclonal antibody) were purchased from Alexis (Lausen, Switzerland) and used according to the instructions of the manufacturer. Mouse mAbs against glial fibrillary acidic protein (GFAP) and against the endothelial cell marker CD31 were from Chemicon (Temecula, CA) and diluted 1:300 and 1:10, respectively. The use and affinity purification of the polyclonal antibodies EAG5, SNG, and AMF against human ABCC2, ABCC4, and ABCC5, respectively, have been described elsewhere (6). The polyclonal antibodies CKD, ENY, SPA, and LPS were raised in rabbits against the COOH-terminal sequences of OATP1A2 (CKDIYQKSTVLKDDELKTKL; National Center for Biotechnology Information accession no. NP_602307), OATP1C1 (ENYTTSDHLLQPNYWPGKETQL; accession no. NP_059131), OATP2B1 (SPAVEQQLLVSGPGKKPEDSRV; accession no. NP_009187), and OATP4A1 (LPSQSSAPDSATDSQLQSSV; accession no. BAA89288), respectively. The peptides were synthesized and coupled to keyhole limpet hemocyanin (Peptide Specialty Laboratories, Heidelberg, Germany). The antisera were affinity-purified as described earlier (22) using membrane vesicles prepared from HEK cells expressing the respective recombinant OATP. The affinity-purified antibodies were used at final concentrations of 0.2 to 0.4 mg/mL for the immunofluorescence microscopy. Alexa Fluor488–conjugated goat anti-rabbit, anti-mouse, and anti-rat immunoglobulin G (IgG) were from Molecular Probes (Eugene, OR). Cy3-conjugated anti-mouse IgG was from Jackson ImmunoResearch (West Grove, PA). Horseradish peroxidase–conjugated goat anti-rabbit and anti-mouse IgGs were from Bio-Rad (Munich, Germany) and goat anti-rat IgG was from Promega (Mannheim, Germany). All other chemicals were of analytical grade and obtained from Merck (Darmstadt, Germany) or Sigma (St. Louis, MO).

Human tissue samples. The present study was undertaken after informed consent had been obtained from each patient, in accordance with the regulations of the Ethics Committee, University of Heidelberg, and with the Declaration of Helsinki. Glioma samples obtained at the Neurosurgery Hospital, University of Heidelberg (Heidelberg, Germany) were 10 diffuse astrocytomas (WHO grade 2), 11 malignant astrocytomas (WHO grade 3), 10 glioblastomas (WHO grade 4), 7 oligodendrogliomas (WHO grade 2), 9 malignant oligodendrogliomas (WHO grade 3), 6 oligoastrocytomas (WHO grade 2), and 8 malignant oligoastrocytomas (WHO grade 3). The tumors were classified by a neuropathologist according to the WHO classification of brain tumors. A clinical follow-up was assessed for all patients. None of the patients had received radiotherapy and four had received chemotherapy before surgery (Table 1). After resection, each sample was immediately snap-frozen in liquid nitrogen and stored at −80°C until further use. In addition to immunofluorescence microscopy, material from some of the samples was used for mRNA analysis or homogenized and used for immunoblot analysis. Perilesional brain samples from the temporal lobe were from patients undergoing surgical resection of gliomas or undergoing neurosurgery because of cerebral hemorrhage (6). The perilesional tissue was removed and immediately frozen in liquid nitrogen before use. None of the patients had received chemotherapy before neurosurgery.

Immunofluorescence microscopy. Cryostat sections (5 μm) were prepared, air-dried for 16 hours at room temperature, and fixed for 10 minutes in acetone (−20°C), or in phosphate-buffered formaldehyde (10%, room temperature) for SPA antibody only. Sections were then incubated with the primary antibodies diluted in PBS for 60 minutes at room temperature, washed thoroughly in PBS, and incubated with the respective secondary antibody (1:300) for 60 minutes. After washing in PBS, sections were mounted in Moviol (Calbiochem, La Jolla, CA). Fluorescence micrographs were taken with a confocal LSM510 laser scanning microscope (Carl Zeiss, Jena, Germany). For peptide competition experiments, affinity-purified antibodies were incubated at 4°C for 16 hours with the respective peptide (100-200 μmol/L) used to generate the antibody, and then applied to the cryosections. In some experiments, nuclei were stained with propidium iodide added to the secondary antibody solution at a final concentration of 2 μg/mL.

ABCC immunostaining of tumor cells was assessed semiquantitatively based on the percentage of tumor cells showing specific immunoreactivity and was scored as follows: −, no positive cells detectable; +, 1% to 25% of tumor cells positive; ++, 26% to 75% of tumor cells positive; +++, >75% of tumor cells positive. ABCC immunostaining of vessels was scored as either positive (+) or negative (−).

Preparation of crude membranes, immunoblot analysis, and deglycosylation. Human glioma tissue (∼100 mg) was homogenized during thawing in 1 mL of lysis buffer (10 mmol/L Tris pH 7.4, 250 mmol/L sucrose, supplemented with 0.1 mmol/L phenylmethylsulfonyl fluoride, 1 μmol/L pepstatin). Membrane fractions were collected by ultracentrifugation (100,000 × g, 4°C, 60 minutes) and suspended in 10 mmol/L Tris buffer (pH 7.4). Crude membrane fractions were diluted with sample buffer and incubated at 37°C for 30 minutes. Proteins were separated by SDS-PAGE (7.5% separating gels) and blotted onto nitrocellulose membranes using a tank blotting system (Bio-Rad). For detection of ABCC4, OATP1A2, and OATP1C1, antisera were diluted in PBS/Tween 20 (1%) containing 5% milk powder (1:1,000, 1:5,000, and 1:8,000, respectively). For detection of OATP2B1, the affinity-purified SPA antibody was used in the same manner at a dilution of 1:60. For peptide competition experiments, the antisera were incubated at 4°C for 16 hours with 200 μmol/L of the synthetic peptide used to generate the respective antiserum and then added to the nitrocellulose membrane. Horseradish peroxidase–conjugated secondary antibodies were used at a dilution of 1:1,000 in PBS/Tween 20 (1%) containing 5% milk powder. Deglycosylation was done as described (23).

RNA isolation and quantification of ABCC and SLCO mRNAs. To initially assess which transporter genes may be expressed in human gliomas, total RNA was isolated from glioma samples and reverse transcribed with an oligo(dT)18 primer. ABCC and SLCO mRNAs were quantified using the LightCycler system (Roche Diagnostics, Mannheim, Germany) and ABCC primer pairs as described (6). The primer pair used for β-actin mRNA amplification was from Stratagene (La Jolla, CA). SLCO sense and antisense primer pairs were as follows: 5′-TGCCATACCTGGATATATGGTT-3′ and 5′-CAATTTAGTTTTCAATTCATCATCTT-3′ for SLCO1A2, 5′-CACCTCACATGTCATGCTGATT-3′ and 5′-AACAATGTGTTTCACTATCTGCC-3′ for SLCO1B1, 5′-TCATAAACTCTTTGTTCTCTGCAA-3′ and 5′-GTTGGCAGCAGCATTGTCTTG-3′ for SLCO1B3, 5′-CCTGGATACATATTACTTCTGAG-3′ and 5′-CATGTTTCTAAAGTTGAGTTTCCT-3′ for SLCO1C1, 5′-CGACTCAACGTGCAGCCATC-3′ and 5′-CCGACACTAGCAATTGCTGCT-3′ for SLCO2B1, and 5′-GAGACTGTAGCTGTATCCCTC-3′ and 5′-GCGGTGGTCAGACGCTGCT-3′ for SLCO4A1. The amount of the respective cDNA was determined by serial plasmid dilution (human ABCC2 cDNA or SLCO1A2 cDNA in the expression vector pcDNA3.1, from 1 × 10⁶ to 1 × 10² fg), and ABCC and SLCO mRNA quantities were given as percentage of the β-actin mRNA amount in the respective glioma sample. Finally, amplified PCR products were separated by agarose gel electrophoresis, stained with ethidium bromide, and visualized under UV light. Plasmids containing the respective ABCC and SLCO cDNA served as positive controls.

ABCC4 is predominantly expressed in endothelial and astrocytic glioma cells. Real-time PCR yielded ABCC4 mRNA quantities between 0.1% and 1.0% of β-actin mRNA (Table 2). ABCC4 protein expression was analyzed in cryosections of the surgical specimens by immunostaining with the affinity-purified SNG antibody (Table 1; Fig. 1). ABCC4 was present in three different cell types: in capillary endothelial cells, in astrocytes, and in glioma cells. In 53 samples (87%), ABCC4 was detectable in brain capillaries. This localization was confirmed by costaining with an antibody specific for the endothelial marker protein CD31 as shown in the representative images of Fig. 1A, to C. Besides a cytoplasmic staining in the endothelial cells, ABCC4 was localized to the luminal membrane, which was identified by costaining for ABCG2 known to be expressed on the luminal side of brain tumor endothelial cells (ref. 5; Fig. 1D). In a minority of tumors (8 of 61; 13%), brain capillaries did not label for ABCC4, which was particularly observed for the proliferating endothelial cells in malignant gliomas and glioblastomas (Fig. 1F and G). ABCC4 was present in the glioma cells of 23 of 31 (74%) astrocytic tumors, 6 of 16 (38%) oligodendroglial tumors, and 7 of 14 (50%) mixed gliomas. In the latter ones, especially the astrocytic components expressed ABCC4, in either diffusely or focally distributed astrocytic cells, the astrocytic nature of which was confirmed by simultaneous staining of ABCC4 and GFAP. In astrocytic tumors, both gemistocytic (Fig. 1H) and fibrillary tumor cells showed ABCC4 labeling. ABCC4 was also present in reactive astrocytes (Fig. 1D and E), particularly in their perivascular end feet as shown by costaining for GFAP (Fig. 1I). This expression pattern was especially visible in those tumors, in which the astrocytes were entrapped within a mass of ABCC4-negative tumor cells (e.g., in malignant oligodendrogliomas; Fig. 1G).

The specificity of the SNG antibody was shown by using the SNG antibody preincubated with the synthetic SNG peptide for immunostaining experiments. All staining patterns described above were abolished in this experimental approach (not shown). Moreover, the Alexis ABCC4 polyclonal antibody showed the same labeling patterns as the affinity-purified SNG antibody (not shown). To further rule out any cross-reactions, immunoblot analyses of glioma homogenates from all histologic types were done and showed a strong band at ∼165 kDa with the SNG antibody (Fig. 1J). As a positive control, membrane vesicles prepared from a clonal cell line stably expressing ABCC4 (V79-ABCC4) were used (24) and revealed a broad band at ∼170 to 200 kDa, slightly above the one observed in the glioma samples (Fig. 1J). However, after deglycosylation using peptide N-glycosidase F (PNGaseF), the relative masses of the glycoproteins detected in the V79-ABCC4 membrane vesicles and in the glioma samples were reduced to the same size of ∼150 kDa, the expected size for the ABCC4 protein (not shown). All bands disappeared after preincubation of the SNG antiserum with the synthetic SNG peptide (Fig. 1J).

ABCC5 expression in human gliomas resembles that of ABCC4. Quantitative real-time PCR analysis of ABCC5 mRNA in human glioma samples showed higher expression levels than for ABCC4, ranging from 0.1% to 3.1% β-actin mRNA (Table 2). Table 1 summarizes the results obtained from immunostaining experiments on cryostat sections of 59 glioma samples using the affinity-purified AMF antibody, which has previously been used to characterize ABCC5 expression in tissues of the genitourinary system (25). ABCC5 expression was abundant in the endothelial cells of almost all tumors (51 of 59; 86%), particularly in the luminal membrane (Fig. 2A-C). In glioma cells, staining patterns were similar to those of ABCC4 in most samples (representative images in Fig. 2D-F) and ABCC5 labeling was detected in the glioma cells of 22 of 30 (73%) astrocytic tumors, in 7 of 15 (47%) oligodendroglial tumors, and in 6 of 13 (46%) mixed gliomas. In contrast to ABCC4, ABCC5 was seen more frequently in the glioma cells of malignant oligodendrogliomas (Fig. 2G). Reactive astrocytes were also intensely labeled in the border of tumor tissue and within the tumorous mass (Fig. 2H). Neurons found in the tumor tissue also revealed ABCC5 expression (Fig. 2I). Except for some perinuclear artifacts, all stainings were entirely abolished by preincubation of the AMF antibody with the synthetic AMF peptide (not shown). The specificity of staining was further verified by exemplary immunostaining with the M5I-1 antibody, which showed weaker but identical staining patterns as the AMF antibody (not shown).

ABCC1 was not detected in human gliomas on the protein level. ABCC1 cDNA could be amplified by real-time PCR and mRNA levels were in the range of 0.1% to 0.8% of β-actin mRNA (Table 2). However, immunofluorescence microscopy of cryosections immunostained either with the MRPr1 or the QCRL-1 antibody did not show any specific staining of endothelial or glioma cells in any of the 50 investigated gliomas (Fig. 3A), except for some needlelike artifacts caused by erythrocyte membranes within and around large blood vessels. Strong staining was also observed in large hemorrhagic areas of some tumors. Human testis tissue served as positive control for the MRPr1 antibody, yielding an intense signal in Leydig cells and in the basal membrane of Sertoli cells (Fig. 3A), which is consistent with previous findings (26).

ABCC2, ABCC3, and ABCC6 were not detected in human gliomas. Real-time PCR experiments yielded no significant amount of cDNA for ABCC2 and no amplification product at all for ABCC6 (Table 2). ABCC3 mRNA, however, could be detected in all investigated samples, with varying expression levels ranging from 0.1% (sample OAII-5) to 1.8% (sample GB-7) of β-actin mRNA. Particularly in glioblastoma samples, comparatively high ABCC3 mRNA amounts were amplified. To investigate protein expression, cryosections were immunostained using the mAbs M2III-6 and M2II-12 and the affinity-purified polyclonal EAG5 antibody directed against ABCC2, M3II-9 against ABCC3, and M6II-31 against ABCC6. In none of the 46 samples tested we observed a specific staining for ABCC2, ABCC3, or ABCC6 (Fig. 3B-D). However, liver samples, serving as positive controls as described before (6), revealed an intense labeling of the sinusoidal hepatocyte membrane in the case of ABCC3- and ABCC6-specific antibodies (Fig. 3C and D) and of the canalicular hepatocyte membrane in the case of the three antibodies directed against ABCC2 (Fig. 3B). Taken together, these results indicate that ABCC2, ABCC3, and ABCC6 are expressed in human gliomas at detectable amounts.

OATP1A2 and OATP2B1 are localized at the blood-brain barrier and the blood-tumor barrier. The mRNA expression levels of SLCO1A2, SLCO1C1, and SLCO4A1 were in the same range as those of ABCC4 and ABCC5 (Table 2). No amplification products were obtained for SLCO1B1 and SLCO1B3. Within the group of transporter mRNAs investigated in this study, SLCO2B1 emerged as the most abundant one with values in the range from 0.2% to 11.6% of β-actin mRNA (Table 2).

OATP1A2 protein was analyzed in two perilesional brain samples and in seven gliomas of different histologic entities (AII-2, AII-5, AII-6, GB-1, OAIII-1, OAIII-4, and OIII-1) and detected in the endothelial cells as shown by colocalization with CD31 (Fig. 4A-C). To verify the specificity of the immunostaining with the affinity-purified CKD antibody, crude membrane homogenates from HEK cells stably expressing OATP1A2 and from three gliomas were blotted onto a nitrocellulose membrane and incubated either with the CKD antiserum or with antiserum that had been preincubated with CDK synthetic the peptide before. Three different bands in the HEK cells and two different bands in the gliomas were detected, all between 60 kDa (expected size of deglycosylated OATP1A2; ref. 21) and 110 kDa (Fig. 4D). These bands were abolished in the peptide competition experiments (Fig. 4D) and, therefore, probably represent different glycosylation forms of OATP1A2 as discussed before (21). In the immunofluorescence microscopy experiments, the staining was also completely eliminated after incubation with the synthetic CKD peptide (not shown).

OATP2B1 was expressed in the blood vessels of all eight glioma samples analyzed (AII-6, AIII-3, GB-1, OAIII-3, OAIII-5, OII-1, OII-7, and OIII-1) and in two perilesional brain samples, mainly in the luminal membrane of the endothelial cells (Fig. 4E-H). This staining pattern disappeared in the peptide competition experiments (not shown). Immunoblot analyses using liver tissue as a positive control (27) revealed two bands at ∼84 and 95 kDa that were selectively abolished in the peptide competition experiments (Fig. 4I), indicating the presence of two differently glycosylated OATP2B1 forms in human gliomas and liver. No specific immunostaining was observed for OATP1B1, OATP1B3, OATP1C1, or OATP4A1.

The investigation of ATP-dependent efflux pumps conferring MDR has become a major topic in the exploration of tumor chemoresistance (28). Regarding gliomas, the focus has been on ABCC1 and MDR1 P-glycoprotein, but other ABC transporters also facilitate the cellular export of cytotoxic substances (7, 20). Thus far, none of the OATP family members has been investigated in human gliomas although a reduced expression of these organic anion uptake transporters may also contribute to the chemoresistance of brain tumors. We here describe the localization of ABCC4 and ABCC5 proteins in human gliomas, primarily in the blood-tumor barrier and in astrocytic glioma cells. The presence of OATP1A2 and OATP2B1 proteins in the blood-tumor barrier, but not in the glioma cells, indicates that uptake of organic anions into the endothelial cells may proceed, but entry into the glioma cells may be impaired.

The localization of ABCC4 and ABCC5 in the luminal membrane of brain endothelial cells (6, 15, 29) seems to be conserved also in the blood-tumor barrier (Figs. 1 and 2; Table 1). Only in some proliferating endothelial cells, especially in malignant oligodendrogliomas (Fig. 1H) and glioblastomas, ABCC4 and ABCC5 expression were largely absent, indicating a disturbance of the integrity of the barrier with respect to transporter expression. Additionally, ABCC4 and ABCC5 were present in glioma cells, mainly of astrocytic neoplasms (Table 1). ABCC4 and ABCC5 mRNAs have been detected in glioma biopsies (5) and in glioma cell lines (30). The ABCC5 protein could not be detected in those studies, probably due to a low affinity of the mAb M5I-1. As observed before (25), as well as in the present study, the polyclonal AMF antibody seems to have a higher affinity for ABCC5 so that ABCC5 could be detected in the glioma samples (Fig. 2). In many tumor samples, particularly those of oligodendroglial descent, ABCC4 and ABCC5 were expressed in reactive astrocytes (Figs. 1E and 2H), including their perivascular end feet (Fig. 1G), being in line with their localization in astrocytes of perilesional human brain samples (6). The expressions of ABCC4 and ABCC5 in the glioma cells did not correlate with the grade of malignancy but rather with the histologic subtype; although some oligodendrogliomas showed a low expression of ABCC4 and ABCC5, both transporters were predominantly found in astrocytic neoplasms (Table 1). Assuming that gliomas originate by dedifferentiation of mature glial cells (16), these tumor cells seem to retain the ABCC expression pattern of the original cell. Otherwise, if gliomas emerge from the transformation of an immature precursor (31), the expression of ABCC4 and ABCC5 may accompany a differentiation towards an astrocytic phenotype. This may also account for their expression primarily in the astrocytic portions of mixed oligoastrocytomas, of which oligodendroglial and astrocytic portions are thought to derive from a single common progenitor cell (32). The low expressions of ABCC4 and ABCC5 also in some oligodendroglial tumor cells (Table 1) may then be explained as an earlier stage of differentiation of these cells, as shown for the expression of other astrocytic proteins in oligodendroglioma cells (33).

ABCC4 and ABCC5 are the major ABCC subfamily members in human gliomas identified in this study (Figs. 1 and 2; Table 1). Their presence in the blood-tumor barrier and in glioma cells renders them candidate transporters that could contribute to the intrinsic MDR phenotype (13, 34, 35) of gliomas. Methotrexate transport by ABCC4 and ABCC4-mediated resistance to ganciclovir and topotecan has been shown (15, 36, 37). A recent study (15) strongly supports the importance of Abcc4 in the blood-brain barrier as well as in the blood-cerebrospinal fluid barrier by showing an enhanced accumulation of topotecan in the brain and the cerebrospinal fluid of Abcc4-deficient mice. In contrast, ABCC1, ABCC2, ABCC3, and ABCC6 are apparently not expressed in relevant amounts (Fig. 3) so that their role in the intrinsic chemoresistance of gliomas may be limited. We did not observe any specific ABCC1 immunoreactivity in gliomas, being in line with other studies (5, 38). However, some studies showed ABCC1 expression in human gliomas with a varying number of positive tumor cells, few of them also in the blood-tumor barrier (9, 11, 12, 39–41). Several other studies did not detect ABCC1 in brain at all (5, 10, 41, 42). Our recent demonstration of ABCC1 in endothelial cells of perilesional human brain (6), but not in the blood-tumor barrier (Fig. 3), indicates a lower expression of ABCC1 in the blood-tumor barrier than in the blood-brain barrier. This may be due to the loss of modulating expression factors released from astrocytes or from other cells of non-neoplastic brain. By using three different antibodies, we could not detect ABCC2 in the human glioma samples. Our results are in agreement with previous studies in which ABCC2 mRNA or ABCC2 protein was not detected in significant amounts in normal or perilesional brain tissue (6, 10) in cultured endothelial or parenchymal cells of different species (29, 43, 44) or in glioma tissue and glioma cells (45, 46). Similarly, ABCC3 was not detectable in any of the samples, which is in accordance with earlier studies (5, 38). Presence of ABCC3 in glioma cell lines (30, 46) indicates a discrepancy between protein detectability in glioma biopsies and glioma cell lines, which might be caused by long-term culturing conditions as proposed for ABCC1 (38).

Besides ABCC4 and ABCC5 (Figs. 1 and 2), other ABC transporters such as MDR1 P-glycoprotein (47) and ABCG2 (5) are present in the blood-tumor barrier. Although not systematically examined in the present study, we also observed staining for ABCG2 exclusively in the blood-tumor barrier, but not in the glioma cells, and therefore we used it as a marker for the luminal membrane of endothelial cells (Figs. 1 and 4). Most recently, Aronica et al. (48) reported similar findings. Presence of ABCG2 in the blood-tumor barrier may also contribute to the chemoresistance of gliomas.

In addition to ATP-dependent export pumps, the blood-brain barrier and the blood-tumor barrier comprise the uptake transporters OATP1A2 and OATP2B1 (Fig. 4) whereas OATP1C1, OATP1B1, OATP1B3, and OATP4A1 were not detectable by immunofluorescence microscopy. Only the localization of OATP1A2 in the blood-brain barrier had been described before (21). The observed expression pattern, with OATP1A2 and OATP2B1 present in the blood-tumor barrier and all of the six investigated OATP family members absent in the glioma cells, does not fully rule out the existence of other pathways of transporter-mediated drug uptake [e.g., by additional members of the OATP family (19) or by members of the SLC22 family (49)].

In conclusion, localization of OATP1A2 and OATP2B1 in the blood-tumor barrier indicates that organic anions can be taken up into the endothelial cells, however, other transporters for the uptake of organic anions into endothelial and glioma cells may exist. The presence of ABCC4 and ABCC5 proteins in the endothelial and in the astrocytic glioma cells suggests that expression of both proteins is a characteristic of the astrocytic phenotype in human gliomas and that both could participate in the drug resistance of these tumors. In view of the divergent results between the in vivo and in vitro expressions of drug transporters, this study provides a basis for the assessment of results previously obtained from cell culture experiments in comparison with our results from frozen human glioma tissue.

Grant support: Tumorzentrum Heidelberg/Mannheim (A.T. Nies and H.-H. Steiner), Wilhelm-Sander-Stiftung München (A.T. Nies, H.-H. Steiner, and C. Herold-Mende), research collaboration between the German Cancer Research Center and Pfizer Research Laboratories, Groton, Connecticut (D. Keppler), Deutsche Forschungsgemeinschaft (J. König and D. Keppler).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Horst-Peter Schmitt (Institute of Pathology, University of Heidelberg) for helpful discussions, Dr. Herbert Spring (Division of Cell Biology, German Cancer Research Center) for expert help in confocal laser scanning microscopy, Dr. Maria Rius (Division of Tumor Biochemistry, German Cancer Research Center) for providing ABCC4-transfected V79 cells, and Elke Herrmann and Jessica Longin for excellent technical assistance.