Malignant astrocytomas, the most common primary brain tumors in children and adults, are characterized by an extremely short postdiagnosis survival time. This bad prognosis rests on both local aggressiveness and a distinct resistance to diverse therapeutic options.1 Surgery is almost never curative due to the diffuse invasion of astrocytoma cells into the surrounding brain tissue, and application of radiotherapy only modestly increases mean survival time. Chemotherapy with either single agents or drug combinations has not resulted in appreciable progress in malignant astrocytoma treatment. Penetration of many drugs into the brain is limited by the blood-brain barrier (BBB). This protective shield for the brain tissue is at least partly based on overexpression of P-glycoprotein (P-gp) and several members of the multidrug-resistance protein family (MRP) in the endothelium of brain capillaries.2, 3 The proper functioning of the BBB within tumor tissues, however, might be impaired, and radiotherapy is known to further disrupt the BBB function.1 A combination of radio- and chemotherapy only marginally increased therapeutic efficiency in glioma patients, pointing toward resistance at the tumor cell level as a major cause for therapy failure.
The mechanisms causing the MDR phenotype of malignant astrocytes have not been entirely clarified. P-gp, encoded by the mdr-1 gene, is a transmembrane efflux pump capable of conferring broad-spectrum cellular resistance against diverse lipophilic drugs. Although the important role of P-gp in the BBB function is beyond all doubt, conflicting data exist about its expression in brain tumor cells (reviewed in ref. 4). Most of the authors agree that P-gp overexpression cannot completely explain the widespread MDR phenotype of glioma cells.4 MRP1, the second well-defined multidrug transporter, has been found to be expressed in a portion of astrocytic tumors and to be functionally active in transporting drugs.5
Another protein that is overexpressed in diverse MDR cells and predicts chemotherapy resistance in several malignancies is the lung resistance protein (LRP). It is identical to the human major vault protein (MVP) and represents ≥70% of the vault particle mass. Vaults are the largest ribonucleoprotein particles known so far (13 MDa); they are almost ubiquitously expressed at the highest levels in potentially toxin-exposed epithelia of the gastrointestinal tract and in macrophages.6 The cellular function of vaults and their definite role in chemotherapy resistance have not been completely clarified.
We have investigated MVP expression and cellular localization in cell lines, primary cultures and patient material derived from brain tumors and other tumors of neuroectodermal origin including mainly astrocytomas and glioblastomas but also meningiomas and primitive neuroectodermal tumor cell lines as well as brain metastases. We show that tumors of astrocytic origin and meningiomas generally express high levels of vault particles, in contrast to medulloblastomas and neuroblastomas and that MVP might be 1 factor responsible for the intrinsic and/or acquired MDR phenotype of astrocytoma cells.
MATERIAL AND METHODS
During neurosurgery, tissue samples from 7 glioblastoma and 3 meningioma patients, as well as from 3 patients with brain metastases (colon carcinoma, nonsmall cell lung cancer and kidney carcinoma), were taken from the central area of the tumor. Additionally, 1 sample was obtained from a necrotic brain area of a patient with gliosis. Tissue specimens were immediately snap-frozen in liquid nitrogen. For protein extraction, samples were minced with an Ultra-Turrax T8 (IKA, Staufen, Germany) in cell lysis buffer containing several protease inhibitors.7 Lysates were incubated on ice for 20 min with frequent vortexing and then centrifuged at 1,000g. Supernatants were collected and protein content analyzed by the Micro BCA protein assay reagent kit (Pierce, Rockford, IL) using BSA as standard.
Cell lines and primary cultures
The study included 22 cell lines derived from the following tumors: glioblastoma (n = 13), gliosarcoma (n = 1), anaplastic astrocytoma (n = 1), medulloblastoma (n = 2) and neuroblastoma (n = 5). CCF-STTG1, DBTRG, Hs 683, T98G, U373, SW 1088, Daoy, D341 Med, SK-N-DZ, SK-N-FI and SK-N-MC were obtained from ATCC (Rockville, MD). LN40 and LN140 were kindly donated by Dr. Tribolet (Lausanne, Switzerland), MGC, MR-1 and GOTO by Dr. Kurata (Tokyo, Japan) and STA-BT-1 and STA-BT–3 by Dr. Ambros (Vienna, Austria). All other cell lines were established at the Institute for Cancer Research, Vienna or the Department of Neurosurgery, Wagner-Jauregg-Hospital, Linz, Austria.
Primary cell cultures were established from glioblastoma (n = 15), gliosarcoma (n = 1), anaplastic astrocytoma (n = 6) and meningioma (n = 5) specimens as well as from brain metastases (n = 3). Cell lines were grown in respective culture media (GIBCO, Life Technologies, Austria), all containing 10% FCS. For establishing primary cultures, tumor specimens were transferred during surgery into culture medium (RPMI-1640 with 20% FCS, 1% glutamine, 1% penicillin/streptomycin). The tissue was minced with fine scissors and washed several times with physiologic NaCl solution to remove blood cells. After retransferring the fragments into a 25 cm culture flask containing cell culture medium, cells were incubated for 24 hr (37°C; 5%CO2) followed by another growth medium change. Up to the first passage, which generally was performed about 20–30 days after setting up the culture, medium was changed twice a week. After passage 5, cells were grown in a culture medium containing 10% FCS without antibiotics. The presence of malignant cells in primary cultures was proved by comparative genomic hybridization (data not shown). Cytotoxicity experiments were performed between passage 6 and 14. All cells were periodically checked for Mycoplasma contamination. Normal human astrocytes were obtained from Clonetics, BioWitthaker (Verviers, Belgium) and cultured following the instructions of the manufacturer using the astrocyte growth medium BulletKit.
Monoclonal antibodies and chemicals
Two monoclonal antibodies (mAbs) directed against different epitopes of MVP were used: LRP56 (IgG2b, Sanbio, Uden, The Netherlands) and LRP (IgG1, PharMingen/Transduction Laboratories, San Diego, CA). The isotype control antibody and all chemicals used were obtained from Sigma (St. Louis, MO). Complete protease inhibitor mix (Roche, Mannheim, Germany) was diluted in a. d. at concentrations suggested by the manufacturer. Drugs were dissolved freshly for each experiment and diluted in culture medium before addition to the cell cultures.
MVP mRNA expression was determined by RT-PCR analysis. Total cellular RNA was prepared and RT-PCR performed as described previously,7 using an oligonucleotide primer set (sense 5′-TTCTGGATTTGGTGGACGC-3′; antisense 5′-ACTTCTCTCCCTTGACCAC-3′) resulting in the amplification of a 284 bp product (bp 740–1024 of MVP mRNA) specific for the human MVP gene sequence (Genbank accession number X79882). GAPDH was amplified (358 bp)7 as a housekeeping gene (sense 5′-CGGGAAGCTTGTGATCAATGG-3′; antisense 5′-GGCAGTGATGGCATGGACTG-3′). Dynamics of PCR amplification were evaluated at different PCR cycle numbers. Twenty-four cycles were chosen for MVP and 18 cycles for GAPDH, resulting in an analysis within the exponential portion of the amplification. Several negative controls were included in each experiment. Expression levels (arbitrary units) were calculated relative to GAPDH mRNA amplified concomitantly.
Cell fractionation and immunoblot analysis
Cell fractionation, sucrose gradient centrifugation and MVP detection by Western blot were performed as described.7 Data were expressed relative (arbitrary units) to extracts of strongly MVP-expressing embryonic fibroblasts F2000 (Flow, Scotland) included in each blot (means of 3 experiments, SD < 15%).
MVP expression was assessed on cells grown in chamber slides (Becton Dickinson, Mountain View, CA) by indirect immunofluorescence staining using the mAb LRP56 (1:25) and a fluorescein isothiocyanate (FITC)-labeled rabbit anti-mouse IgG mAb (1:1,000; Sigma). Slides were mounted in Vectashield mounting medium with DAPI (Vector, Burlingame, CA). Staining was analyzed with a DMRXA Leica with epifluorescence equipment (Leica, Cambridge, UK) and captured with a COHU high-performance CCD camera.
Cytotoxicity of several drugs was analyzed by MTT tests as described previously.7 Dose-response curves were evaluated for daunomycin (DM; 1–100 ng/ml), adriamycin (ADR; 5–250 nM), etoposide (VP-16; 0.5–25 μM), cisplatin (CDDP; 0.1–5 μM), vinblastine (VBL; 0–25 nM), bleomycin (BM; 0.1–10 μg/ml) and 1,3-bis(2-chloroethyl)-1-nitrosourea (BCNU; 10–400 μM) using a 72 hr continuous drug exposure setting. Only proliferating cell cultures that doubled at least twice during the incubation period were included in the analysis. Experiments were repeated 2–4 times in triplicate. IC50 values were calculated from whole dose-response curves as described.7 Due to the high amount of data, definite IC50 values for all drugs and cell lines were not included in the manuscript but may be obtained from the authors on request. Several cell lines were completely insensitive to the evaluated drug concentrations. To allow statistical analyses including those cell lines, the respective IC50 values were arbitrarily set at the highest drug concentration used.
Correlation analysis regarding data obtained by different MVP detection methods and chemosensitivity data were set up by either linear regression or 2-tailed Spearman rank correlation analysis (GraphPad Prism). Correlations were considered significant at p < 0.05. If not stated otherwise, results from linear regression analysis are given. Data regarding MVP expression of histologic subtypes were compared by 2-tailed Mann-Whitney U tests and considered significant at p < 0.05.
MVP gene expression in normal astrocytes, tumor cells and surgery material
MVP mRNA and MVP expression were analyzed by semiquantitative RT-PCR and Western blot, respectively (Fig. 1). MVP mRNA and protein expression data correlated significantly (Spearman r = 0.425; p < 0.005). At least trace amounts of MVP mRNA were detectable in all RNA preparations analyzed (Fig. 1a), whereas in several cell extracts no MVP protein was detectable by the methods used (Fig. 1b). Significant differences between the investigated tumor types with regard to expression of the MVP gene were detected (Figs.1a,b, 2). Cells derived from astrocytic brain tumors, namely, glioblastomas and anaplastic astrocytomas, almost always expressed high levels of MVP mRNA and protein, in contrast to the medulloblastoma- and neuroblastoma-derived cells, which contained low or undetectable amounts of MVP (Fig. 2). Meningioma-derived cell lines expressed MVP at levels comparable with tumor cells of astrocytic origin. Interestingly, normal human astrocytes grown on cell culture plastic also expressed MVP, but in low amounts compared with most astrocytoma- and glioblastoma-derived cell cultures (Fig. 1b).
Within tumor cells of astrocytic origin, glioblastomas expressed significantly higher MVP mRNA compared with astrocytomas (p < 0.01); however, the trend was opposite at the protein level (p < 0.05; Fig. 2). This paradoxical situation was mainly based on the fact that several fast growing, dedifferentiated glioblastoma cell lines (analyzed at higher passage numbers) were characterized by high MVP mRNA but attenuated MVP expression (e.g., DBTRG, U373).
In protein extracts derived from surgery material of brain tumor patients, MVP was consistently detectable in all glioma (n = 7) and meningioma (n = 3) specimens analyzed (Fig. 1c). In some samples, MVP was partly degraded, resulting in an additional band at ∼50 kDa, which has also been detected in purified rat vault preparations and initially was misinterpreted as a fourth vault protein.8 The MVP content in the tumor tissues varied distinctly within a range comparable to one seen in primary cultures in vitro (Fig. 1b,c). The highest amount of MVP was found in a metastasis derived from a colon carcinoma; another sample from an NSCLC primary tumor was MVP-negative. Interestingly, an extract prepared from a nonmalignant but necrotic brain area of a patient with gliosis also contained considerable amounts of MVP.
Cellular localization of MVP
Figure 3 shows representative immunofluorescence staining of MVP in normal and malignant cells of the central nervous system. Correlating with Western blot detection, intense MVP staining was present in tumor cells of astrocytic origin (Fig. 3a,b) and to a lower extent in normal human astrocytes (Fig. 3d). Generally, MVP was detectable in a dotted cytoplasmic texture that is characteristic of the cellular distribution of single vaults.9 As also shown for rat fibroblasts grown in vitro,9 MVP localization was rearranged in human glioma cells during diverse cellular processes, predominantly during mitosis (arrow in Fig. 3a) and in case of cell movement and formation of cell protrusions (arrowhead in Fig. 3b). Nuclei remained widely unstained, with the exception of occasional spots within the nucleoli. Also, meningioma-derived cells displayed intense staining (not shown), in contrast to medulloblastoma- (not shown) and neuroblastoma-derived cells (Fig. 3c), which were widely MVP-negative.
Detection of MVP in cell fractions via immunoblotting (Fig. 4a) proved the dominant localization of MVP in the cytoplasm. Most of the MVP could be pelletted from the postnuclear supernatant of the high and low MVP expression glioma cell line MR-1 and DBTRG at 100,000g. A small portion of MVP was, however, always present in the respective nucleic extracts. MVP in the glioma cell line MR-1 was shown to be assembled in intact vaults, as was proved by sucrose gradient centrifugation experiments (Fig. 4b). MVP was solely detectable in those fractions of the sucrose gradient (45 and 50%) known to accumulate intact vaults.10
Chemosensitivity and relation to MVP expression
Twenty cell lines and 18 primary cell cultures met the requirements for inclusion in the cytotoxicity analysis (see Material and Methods). Generally, strong variations in the chemosensitivity of particular cell lines and tumor types were detectable (data not shown). In particular, relatively slow growing primary cultures from astrocytic brain tumors and meningiomas were highly chemoresistant. However, most of the fast growing glioma cell lines were also markedly more chemoresistant compared with cell lines of medulloblastoma and neuroblastoma origin.
A surprisingly broad correlation between chemoresistance of the tested cell lines and their MVP gene expression at both the mRNA and the protein level was detectable (Table I, upper part). Only the resistance against the antibiotic drug BM and the spindle poison VBL did not, at least weakly, correlate with MVP expression. The significant correlations reflect in part the very low MVP expression in some highly chemosensitive neuroblastoma cell lines (data not shown). However, when only cells derived from astrocytic brain tumors were included in the analyses (Table I, lower part), resistances to DM, ADR, VP-16 and CDDP were still significantly correlated with MVP expression. Only the resistance to BCNU had completely lost significance compared with the analysis including all tumor types.
Table I. MVP Gene Expression and Chemoresistance of Brain Tumor-Derived Cells and Other Malignant Cells of Neuroecdodermal Origin: Correlation of Results From Molecular and Cytotoxicity Assays1
Statistical comparison of MVP gene expression and chemosensitivity (IC50 values may be obtained from the authors on request) by Spearman rank correlation analysis (GraphPad Prism). In total 38 cell lines were included in the cytotoxicity tests (GB+GS, AA, NB, MB and MG: n = 25, 3, 5, 1, and 4, respectively). Significant correlations are indicated by bold face. ns, not significant.
All cell lines (upper part) or GB and AA only (lower part) were included in the statistical comparison.
The prognosis of patients with malignant astrocytomas is poor. Unfortunately, and despite valiant efforts, systemic chemotherapy has been the least effective conventional treatment in terms of prolonged patient survival time. These disappointing results may be caused by a specific dilemma of glioma: the agents that are effective in vitro, as demonstrated by a meta-analysis of all data published between 1966 and 1995,11 are unable to cross the BBB in sufficient amounts. Generally, the in vitro analyses uncovered distinct differences in the chemosensitivity of glioma cell lines and primary cultures from patient material, suggesting that glioma cells might frequently exhibit an MDR phenotype.
In this study we demonstrate that malignant astrocytes in vitro, either as stable cell lines or as primary cultures, generally express high amounts of the multidrug resistance protein MVP. Gradient centrifugations suggest that all MVP in glioma cells is assembled in particles behaving like intact vaults. MVP expression level correlated with chemosensitivity against several antineoplastic drugs including anthracyclins, CDDP and VP-16. These observations suggest that MVP, together with MRP1, which is also frequently expressed in glioma cells (manuscript in preparation), contributes to the intrinsic MDR phenotype of malignant astrocytes.
The mechanism by which vaults may contribute to chemoresistance is unclear so far. Several cell lines selected for resistance to drugs (predominantly anthracyclins) distinctly overexpressed MVP and vaults.10 Moreover, when a panel of malignant cell lines of different histologic origins was used, MVP expression correlated with a resistance phenotype even broader than the 1 characterized by overexpression of the classical MDR transporter proteins P-gp and MRP-1.12
Also, in the present study, a comparably broad resistance phenotype correlated with MVP expression in tumor cells of neuroectodermal origin. This finding corroborates the hypothesis that vaults are involved in a protection mechanism active on a basic cellular level, which is widely independent of the administered drug. In vivo, MVP levels have been found to indicate chemotherapy resistance in several malignancies including AML13 and multiple myeloma.14 In case of ovarian cancer, conflicting data about a correlation between MVP expression and chemoresistance have been published.15, 16 MVP overexpression by gene transfection was surprisingly not sufficient to induce MDR.17 This has been explained by the multicomponent structure of vaults containing 2 other proteins in addition to MVP and an untranslated RNA molecule (vRNA). Indeed, an overexpression of other vault components was demonstrated in drug-selected MDR cell lines.10, 18
Some investigations have suggested that vaults might represent the central plug of the nuclear pore complexes, as indicated by their dimension and their 8-fold symmetry.19 This hypothesis has been corroborated by recent data regarding an MVP-dependent MDR phenotype induced by differentiation-inducing agents in colon cancer cells that was based on nuclear drug extrusion.20 However, only a very few studies were able to localize considerable amounts of MVP at the nuclear membrane or in the nucleus. In most cell types, MVP was described to be localized in the cytoplasm or the perinuclear region.7, 9 This is also the case in the neuroectodermal cell types investigated in this study. Only a small amount of MVP was present in extracts from pure nuclei, suggesting that some vaults could be localized either within the nucleus or tightly connected to the nuclear membrane. The vast majority of MVP immunostaining was present in the typical dotted cytoplasmic texture. Expression was frequently enhanced in cells with extending processes and was often located next to the intense, diffuse actin staining (not shown) near the ruffling edges of lamellipodia. This is in agreement with observations made in rat fibroblasts grown on cell culture plastic.9
Tumor extracts obtained from glioma and meningioma patient material contained MVP comparable to that found in tumor cells in vitro. This suggests activation of MVP expression during the process of malignant transformation. In a recent study, activation of MVP (LRP) expression in glioblastomas has also been shown by immunohistochemistry.21 Generally, the brain in humans6 and rats22 contains very low numbers of vaults. Only during late embryonic and early postnatal development of rats is intense vault immunostaining present in the microglia, which represents macrophage-like cells.22 The adult astrocytes of the human brain in situ lack any detectable MVP expression.6 Thus, activation of MVP expression in astrocytic brain tumors might indicate dedifferentiation with concurrent activation of an embryonic marker protein, as is known, for example, for the carcinoembryonic antigen.23
Interestingly, we have also detected a comparably low but substantial expression of vaults in normal human astrocytes when they were cultured sparsely on cell culture plastic. The vault content attenuated after reaching confluence of the culture (manuscript in preparation). Thus, high vault expression in malignant astrocytes may also reflect the manifestation of a defensive competence of normal astrocytes turned on in stress situations or by activating stimuli. This has also been shown for P-gp, another protection protein, which was upregulated following activation of astrocytes due to neuronal injury.24
Summing up, we demonstrate that expression of MVP is a general feature of malignant astrocytic and meningioma cells. Basal MVP levels correlated with chemotherapy resistance against several antineoplastic agents in vitro. Thus vault overexpression might represent a major component of the drug-resistance phenotype of glioma and meningioma cells or at least might indicate a cellular state characterized by a very broad chemoresistance. Substantial MVP expression in normal astrocytes in vitro opens the hypothesis that vaults, although almost absent in the normal adult brain,6 might serve a protective function in the brain after activation of astrocytes by diverse stimuli including injury but also malignant transformation.
We thank Mrs. V. Bachinger, Mrs. J. Krejsa and Mrs. M. Spannberger for skilful technical assistance.