Tenascin-C Is a Novel RBPJκ-Induced Target Gene for Notch Signaling in Gliomas

Tenascin-C (TNC) is highly expressed in all major solid cancer types (for reviews, see refs. 1, 2). In most carcinoma cases, it is produced and localized in the tumor stroma, but the cancer cells themselves are mostly devoid of TNC (3, 4). This is, however, different in glioblastoma (GBM) where the highly invasive cancer cells are the source of TNC themselves. The high expression of TNC in certain brain cancers was the basis for the original discovery of TNC as a glioma-mesenchymal extracellular matrix antigen (5). In the meantime, the presence of TNC in gliomas was found to correlate significantly with increased malignancy and poor clinical outcome of the disease (for review, see ref. 6). Immunohistochemical studies of gliomas have shown that TNC expression increases with tumor malignancy and function-blocking antibodies inhibited TNC-mediated proliferation and migration of GBM cells in culture (7). Furthermore, GBM patients with TNC immunopositivity in the tumor extracellular matrix had a significantly shorter survival compared with patients with GBM lesions lacking stromal TNC (8). It is, therefore, of interest to investigate the factors regulating TNC expression in GBM.

Notch signaling is an evolutionary conserved pathway that controls cell fate and growth. Ligand-dependent cleavage of the Notch transmembrane receptor releases the Notch intracellular domain into the cytoplasm. The intracellular domain translocates to the nucleus and binds and converts the transcriptional repressor RBPJκ/CBF1/Su(H)/CSL to an activator by recruiting and binding coactivators such as mastermind-like proteins (MAML) to induce expression of Notch targets such as hairy/enhancer of split HES-1 and HES-5 in humans (9, 10). Notch2 is expressed during brain development in the cerebellar external granule layer and subventricular zones, where it maintains proliferation and prevents neuronal precursor differentiation (11). Notch2 expression in postnatal brain is restricted to ventricular germinal zones and dividing immature glial cells (12). Conversely, Notch2-deficient mice show increased apoptosis in neural tissues, leading to embryonic lethality (13). Thus far, oncogenic Notch2 functions have been reported in human B-cell leukemia (14) as well as in medulloblastoma (15). Notch1 oncogenic activity has been shown in gliomas (16) and other malignancies (17); however, it can also act as a tumor suppressor, depending on the cellular context (18).

Malignant gliomas are invasive tumors that can only transiently be controlled by surgery, radiotherapy, and chemotherapy (19–21). From a clinical point of view, conventional therapeutic interventions based on tumor resection and radiotherapy and chemotherapy have only moderately improved glioma patient survival over the past decades (22). Depending on histologic subtype, median survival of glioma patients varies between 10 months in the prevalent GBM and 10 years in low-grade oligodendroglioma (23). This is reflected in distinct pathogenetic pathways operative in these different glioma entities associated with variable degrees of aggressiveness (21). Because we found in our previous study that loss of heterozygosity of Notch2 is associated with prolonged survival of patients with oligodendroglioma or GBM with 1p loss (24), we investigated here whether Notch2 expression differs in oligodendroglioma versus GBM and whether Notch signaling could be linked to TNC expression.

Cell lines and antibodies. Cells were grown in DMEM with 10% FCS and 1% glutamine. Immunoreactions were performed with the monoclonal antibodies anti-Notch2 C651.6dbHN and anti-Notch1 bTAN20 (Developmental Studies Hybridoma Bank; ref. 25), anti-TNC B28-13 (26), anti-RBPJκ (Institute of Immunology), anti-actin (Sigma), and antiserum against fibronectin (27).

Glioma biopsies and immunohistochemistry. Paraffin sections of formalin-fixed tissue samples were obtained from the University of Basel Department of Neuropathology in accordance with the guidelines of the ethical committee of the University of Basel. Tumors were diagnosed and graded according to the WHO Classification of Tumors of the Nervous System. Immunohistochemistry was performed with anti-Notch2 and anti-TNC using an overnight incubation at 4°C of a 1:50 dilution. Bound antigens were detected using avidin-biotin-peroxidase (Vectastain, Elite kit; Vector Laboratories), and sections were weakly counterstained with hematoxylin.

Tissue microarray. The tissue microarray comprising 190 GBM has been constructed from archived paraffin blocks at the University Hospital in Lausanne (28). Immunohistochemical determination for TNC (dilution, 1:2,500), anti-RBPJκ (1:120), Notch2 (1:100), and Notch1 (Abcam; 1:200) was performed according to standard procedures for paraffin sections using a high-temperature epitope retrieval technique in citrate buffer (pH 6.0; pressure cooker, 3–5 min) and overnight incubation with the primary antibody. The immunostaining was scored semiquantitatively by M-F. Hamou, who was blinded to the working hypothesis. In brief, the scores for TNC were as follows: 0, no staining; 1, partial or weak staining; 2, positive staining; 3, strong staining. Scores for Notch1 and Notch2: 0, no staining; 1, weak cytoplasmic and/or nuclear stain; 2, up to 50% positive tumor cells; 3, >50% positive tumor cells. Scores for RBPJκ: 0, no staining; 1, few tumor cells with positive nuclei; 2, 15% to 40% of positive nuclei; 3, >40% of tumor cells with positive nuclei. Furthermore, scores included “ni,” which means not interpretable due to damaged or absent GBM tissue or unsuitable histologic area such as necrosis. Therefore, the number of interpretable GBM varied between the analyses. For pairwise comparisons (Fisher's exact test, 2-sided), the scores were collapsed to low (score, 0–1) versus high (score, 2–3) expression, and excluding not interpretable samples. Tumors were ordered by similarity of expression profiles (scores, 0–3) using the SPIN software (29).

Nucleic acid extraction and analysis. Genomic DNA from glioma cell lines was extracted using the genomic DNA purification kit (Qiagen). RNA was isolated using Trizol (Invitrogen) reverse transcribed with the ThermoScript RT-PCR system (Invitrogen). Real-time quantitative PCR was performed on an ABI Prism sequence 7700 detector (PE Applied Biosystems) at microsatellite marker D1S2696 using primers GAATTACATCCCAGGCAATCTGA and CACACAACAGGCCCCTAATCA and probe FAM-AGCCCATGCTCATTCCCACTACACTGG-TAMRA. GAPDH primers are AATGGGACTGAGGCTCCCAC and TTATGGGAAAGCCAGTCCCC and probe FAM-ATCCAAGACTGGCTCCTCCCTGCTG-TAMRA. Notch2 cDNAs from glioma cells were sequenced at mutation hotspots (i.e., epidermal growth factor repeats 11–14, 24–25, 29 and 32; Lin-12 domains; and N2-IC). The Notch2 mutation L1711M found in Hs683 cDNA was confirmed by sequencing genomic DNA.

Western blot analysis. Cell extracts and conditioned media were separated by SDS-PAGE and electroblotted to polyvinylidene difluoride membranes (Millipore). Proteins were detected with primary and secondary antibodies and blots were developed using Super Signal (Pierce) on Kodak BioMax MR Films. Protein bands were quantified using the software GeneTools (SynGene).

Plasmids and lentiviruses. Activated Notch2 (N2-IC, from nucleotides 5107–7425 of Notch2 cDNA sequence AF308601) was cloned into plasmid pcDNA3 (Invitrogen). Mutations in the Notch2-IC (L1711M) and in the RBPJκ binding site (30) were obtained by site-directed mutagenesis (Stratagene). The Notch1 construct (31) was kindly provided by Dr. F. Radtke (Ludwig Institute for Cancer Research, Epalinges, Switzerland), the dominant negative RBPJκ plasmid (32) was provided by Dr. T. Hanjo (Kyoto University Graduate School of Medicine, Kyoto, Japan), and the dominant negative MAML1 plasmid (33) was provided by Dr. J.C. Aster (Brigham and Women's Hospital, Boston, MA). The TNC promoter construct included 102-bp 5′ flanking sequence from the transcription start and 97 bp of the first exon.

The reporter construct was cloned with the Expand High Fidelity PCR System (Roche) using the primer pair GAGCTAGCCTTTAATTCGCCAACTG/GACTCGAGCTGCCTTTGTGCCCACGGAG with genomic DNA from HEK293 cells as template and cloned into the pGL3 luciferase reporter vector (Promega).

The lentiviral vectors pLKO.1-puro, pLKO.1-scrambled-shRNA (Addgene), and pLKO.1-shRNA that targets RBPJκ at 3-UTR (CCGGGCTGGAATACAAGTTGAACAACTCGAGTTGTTCAACTTGTATTCCAGCTTTTT) were transfected into HEK293 cells along with packaging plasmid (pCMV_dr8_91) and plasmid coding for envelope protein (pMD2-VSV-G). Concentration of infectious particles in the supernatant was titrated on HeLa cells.

Transfections and reporter gene assays. Transfections were performed using Fugene (Roche). For the reporter gene assays, Hs683 cells were seeded in 6-well plates and transiently transfected with 1 μg TNC luciferase reporter vector, 1 μg expression plasmid, and 0.1 μg Renilla Luciferase reporter vector phRL-TK for standardization (Promega). Cells were harvested after 24 h and assayed for luciferase activity using the Dual-Luciferase Reporter Assay System (Promega). Luciferase activity was measured from at least three independent experiments in a Mithras LB940 Luminometer (Berthold Technologies). Western blots showing similar levels between N2-ICWT and N2-ICL1711M in transfected cells ruled out instability of N2-ICL1711M.

LN319 cells were transfected with the MAML1DN construct using CaCl₂ precipitation for 8 h. Stably transfected clones were selected for resistance to 100 μg/mL Geneticin for 20 d. LN319 stably transduced with lentivirus constructs were selected for resistance to 0.5 μg/mL puromycin.

Transwell migration assays. Transwell migration assays were performed using modified Boyden chamber units with polycarbonate filters of 8-μm porosity (Costar). The lower side of the filter was coated with 10 μg/mL fibronectin for 2 h at room temperature. The bottom chamber was filled with serum-free DMEM containing 0.1% BSA plus/minus increasing amounts of purified human TNC. Cells (10⁴ per well in serum-free DMEM) were plated in the upper chamber and incubated for 16 h at 37°C. After removal of the remaining cells from the upper surface of the filter, migrated cells at the bottom of the filter were fixed with 3.7% formaldehyde in PBS and stained with 0.1% crystal violet. For each condition, cells in five fields of three independent experiments were counted.

Statistical analyses. Kruskal-Wallis test was used for significance of genomic copy dosage. Significances of reporter gene expression and endogenous TNC expression were established with the Student's t test available at the Web server of St. John's University.

TNC in gliomas correlates with Notch2 and RBPJκ expression. In an initial immunohistochemical survey of glioma patients, we noticed strong Notch2 immunostaining in 2 of 2 astrocytomas grade II and 3 of 4 GBMs. In contrast, Notch2 protein was neither detectable in 6 of 6 oligodendrogliomas grade II and 2 of 2 oligodendrogliomas grade III, nor in normal brain tissue. We observed similar results for TNC in the same glioma biopsies and could correlate strong Notch2 staining in GBM sections with strong TNC immunopositivity. Examples are shown in Fig. 1A. This observation was confirmed by immunoblot analysis of TNC in brain tumor extracts. TNC expression was below detection in normal brain extract, low in oligodendrogliomas, whereas many high-grade astrocytomas and GBMs were rich in TNC (Fig. 1B). The median level of TNC in oligodendroglioma was 3.75 units compared with 36.5 in astrocytoma and 56.2 units in GBM. Out of 19 GBM extracts analyzed, 17 (89.5%) showed higher TNC levels than any of the oligodendrogliomas; in 14 (82.4%) of these TNC-positive samples, Notch2 was detectable as well. Because canonical Notch2 signaling requires the presence of RBPJκ, we determined whether this transcriptional regulator was coexpressed with TNC and Notch2. Indeed, there was a good correlation for the presence of all three proteins in GBM samples but not in extracts of oligodendrogliomas (Fig. 1B). Judging from the GBM immunoblots, it seems that the levels of TNC expression correlated better with RBPJk than with Notch2 levels (Fig. 1B,, left). To extend our analysis, we made use of a GBM tissue microarray that allowed us to study a larger number of tumor samples; we also included Notch1 in our expression study (Fig. 2A). We obtained labeling frequencies of 83% (118 of 143) for Notch2, 59% (84 of 142) for Notch1, 37% (54 of 146) for RBPJκ, and 55% (70 of 127) for TNC. Thus, Notch2 was more frequently detectable than Notch1. We found a significant association between RBPJκ and TNC expression (P = 0.02) and Notch1 and TNC expression (P = 0.02). We also found a trend for an association between Notch2 and TNC (P = 0.15; Fig. 2B).

In a previous study, we have found that loss of a 1p centromeric marker within intron 12 of the Notch2 gene was associated with favorable prognosis in oligodendroglioma as well as in GBM patients (24). Here, we show now that differing Notch2 gene dosages are also found in glioma cell lines. Real-time PCR-based genomic DNA dosage analysis revealed that 4 of 8 GBM lines had amplification at the Notch2 marker D1S2696, whereas LN18 cells exhibited local haploidy (Fig. 3A). Immunoblot analysis of Notch2 in these lines revealed concordance between genomic status and protein levels (Fig. 3B), suggesting that the genomic status at this locus may influence Notch2 protein levels. The glioma line expressing oligodendroglioma markers Hs683 contained only one Notch2 gene copy per diploid genome. In all cell lines tested, we detected a striking coincidence between Notch2 and TNC protein levels, particularly visible between low Notch2 expression and absence of detectable TNC in Hs683 and LN18 cells (Fig. 3B). Finally, Notch1, Notch2, and RBPJκ protein levels were compared with Notch signaling activity as assessed by HES-1 transcript levels (Fig. 3C). RBPJκ was present in all GBM lines, and decreased amounts of Notch1 or Notch2 were accompanied by lower HES-1 expression, as can be seen in Hs683, LN18, and LN401 cells, indicating that there is a direct link between Notch protein levels and Notch signaling activities in the glioma cells analyzed.

The TNC promoter contains an RBPJκ-responsive element. The coexpression of TNC with Notch2 and RBPJκ suggested the possibility of a causal relationship; therefore, trans activation of TNC by Notch/RBPJκ signaling was considered. Alignment of the promoter sequences of the human and murine Notch target genes HES-1 and HES-5 with those of TNC revealed conservation of a perfect RBPJκ binding motif GTGGGAA at the same distance from the TATA box in each of these genes (Fig. 4A). We therefore tested whether the intracellular domains of Notch2 (N2-IC) and Notch 1 (N1-IC) were able to induce luciferase reporter gene expression driven by a 102-bp fragment of a human TNC minimal promoter containing the RBPJκ binding motif. Cotransfection of Hs683 cells with a plasmid expressing either N2-IC or N1-IC both led to a 2-fold increase compared with the basal luciferase activity in the presence of a control plasmid (Fig. 4B). To test whether this induction is dependent on RBPJκ function, point mutations previously described to prevent RBPJκ binding to target genes (30) were introduced into the potential RBPJκ binding site of the TNC promoter. Transfection of the mutant TNC promoter constructs M1 and M2 resulted in similar basal luciferase levels as the wild-type construct but no longer allowed induction by N2-IC (Fig. 4B). Consistently, cotransfection of a plasmid expressing a dominant negative RBPJκ mutant (32) impaired N2-IC–mediated induction of the TNC promoter (Fig. 4B).

The TNC-negative Hs683 cells contain a Notch2 mutation. We sequenced Notch2 fragments encoding the N2-IC and covering mutation hotspots for gain-of-function or loss-of-function described in Drosophila Notch (34), Notch1 (35), and Notch2 (36) in our glioma lines. Interestingly, the oligodendroglioma line Hs683 had a C-to-A mutation in Notch2 encoding leucine 1711, resulting in the substitution of this leucine by methionine (Fig. 4C). Leucine 1711 is conserved throughout vertebrate Notch2 proteins and is located within the RBPJκ-interacting RAM domain. Coexpression of the N2-IC L1711M mutant with the TNC promoter construct no longer increased luciferase activity (Fig. 4D). In parallel, we performed immunoblots of cell extracts to verify that both the wild-type as well as the mutated N2-IC constructs resulted in similar expression levels (Fig. 4D). Therefore, Notch2, although present in the HS683 cells, seems to be nonfunctional.

Notch signaling regulates endogenous TNC expression. Next, we tested whether Notch signaling was able to induce the expression of endogenous TNC protein. Because the abundant TNC levels produced by GBM cell lines may mask any further increases, we focused on fibroblast lines Detroit 551 and MRC-5 that secrete moderate amounts of TNC. Supernatants of fibroblasts transfected with N2-IC showed increased accumulation of TNC in the medium compared with the controls, whereas secreted fibronectin levels were similar (Fig. 5A). The same stimulation of TNC expression was detected after transfection of N1-IC but not with the mutant N2-IC L1711M construct (Fig. 5B). To further prove that TNC production was dependent on Notch signaling, we inhibited this pathway in LN319 cells, which express high levels of Notch1/2, RBPJκ, and TNC (Fig. 3), by overexpression of a dominant negative form of MAML1 (33) or by down-regulation of RBPJκ by shRNA (Fig. 5C). In both cases, a significant reduction of secreted TNC, with a more dramatic effect through MAML1 inactivation, was obtained (Fig. 5C). Thus, stimulating or blocking Notch signaling is accompanied by either up-regulation or down-regulation of TNC, respectively.

TNC enhances glioma cell migration. To test whether TNC expression plays a role in glioma cell migration, we performed transfilter migration assays. Cell migration of the TNC-negative Hs683 cells could be stimulated by the addition of purified TNC to the bottom chamber (Fig. 6A). Because inhibition of Notch signaling by overexpression of dominant negative MAML1 in LN19 cells led to concomitant reduction of TNC expression, we compared cell migration of the parental versus the MAML1DN transfected LN319 cells. We found that cell migration was reduced in the MAML1DN transfectants to <50% of the parental cells (Fig. 6B). Thus, Notch signaling directly regulates TNC expression, which in turn affects glioma cell migration. Because the cell migration capacity is required for the invasive behavior of glioma cells, our results provide a potential mechanism for oncogenic Notch signaling in gliomagenesis.

Differential Notch2 expression in GBM and oligodendroglioma. In this study, we found that Notch2 protein expression was much higher in GBM than in oligodendroglioma. The Notch2 loss-of-function mutation in Hs683 cells described in this article provides an inactivation mechanism of Notch2 in oligodendroglioma, in addition to homozygous deletions detected in primary oligodendroglioma with 1p loss described previously (24). Thus, oligodendroglioma development seems to depend on Notch2 reduction or loss.

In normal brain, Notch2 is expressed in immature glial cells of ventricular germinal zones (12, 37) and nestin, a marker for neural precursors, is regulated by Notch signaling (38). This implies that Notch2 may be involved in maintaining glioma cells in an undifferentiated state. The Notch target HES-1 drives an astrocytic over oligodendrocytic cell fate arguing for a role of Notch signaling in astrocytic differentiation (39). It has been shown that primary GBM shows a high HES-1 content, whereas astrocytoma progression coincides with decreasing HES-1 levels (40). Because HES-1 is under direct control of both Notch proteins, the question of whether Notch1 plays redundant or complementary roles to Notch2 in glioma progression remains open.

TNC is regulated by Notch signaling. Although Notch1 as well as Notch2 induced expression of TNC, a much larger number of GBM cases expressed high levels of Notch2 than Notch1. This high Notch2 expression may be based on the frequent Notch2 gene amplification and high protein expression we found in GBM cells. The significant association between RBPJκ and TNC expression on a GBM tissue microarray together with the identification of an RBPJκ-responsive element in a minimal TNC promoter provide a novel mechanism of TNC transactivation. Molecular cooperation between Notch2 and RBPJκ has been suggested in B-cell development, where Notch2- and RBPJκ-deficient mice present a common phenotype both lacking a B-cell subset (41), and in B-cell leukemia, where Notch2 and RBPJκ up-regulate CD23a transcription (14).

Nevertheless, in half (31 of 62) of the TNC-positive GBM on the tissue microarray, RBPJκ was not detectable, suggesting alternative RBPJκ-independent regulatory pathways for TNC induction. In human fibroblasts, the TNC promoter is activated by transforming growth factor-β (TGF-β; ref. 42). Because TGF-β signaling promotes PDGF-β–dependent cell proliferation in glioma (43), activation of the TNC promoter by TGF-β signaling may be an alternative pathway for TNC induction in glioma.

The Notch-TNC connection and invasion. TNC-deficient mice show compromised proliferation/migration of neural precursors and accelerated oligodendrocyte differentiation (44). Consistently, GBM lines and biopsies showed strong TNC expression, whereas oligodendrogliomas did not, except for an oligodendroglioma subset that has been described as moderately positive (45). Strong TNC expression is associated with the invasive front in many tumor types and is a diagnostic marker for glioma progression, implying a role for TNC in tumor promotion (for review, see ref. 2). These findings support the hypothesis that Notch2/RBPJκ/TNC signaling is an important pathway operational in GBM but not oligodendroglioma development.

Our study presents the first description of an effect of Notch signaling on migration and invasion of glioma cells. Only very recently, reports started to appear describing a role of Notch signaling in invasion of mammary breast cancer cells to brain in a xenograft model (46), induction of snail and EMT in cultured cells (47), or osteosarcoma invasion and metastasis to the lung in an orthotopic mouse model (48). Thus, a link between Notch signaling and invasion may become more generally recognized, and it will be interesting to investigate whether TNC is also involved as a mediator of the Notch effect in these other situations.

From a clinical point of view, conventional therapeutic interventions based on tumor resection and radiotherapy and chemotherapy have only moderately improved glioma patient survival over the past decades (22). In addition to direct targeting of TNC (49, 50), our data suggest the use of drugs blocking Notch signaling.

We propose that the Notch/RBPJκ/TNC pathway regulates tumor cell migration, a hallmark of invasive GBM. This molecular cascade provides a novel mechanism through which Notch acts in neoplastic transformation and possibly in normal development of the neuronal and glial cell lineages.

No potential conflicts of interest were disclosed.

Grant support: Krebsliga Beider Basel (7-2004; A. Merlo); OncoSuisse [OCS-01613-12-2004 (A. Merlo) and OCS01680-2-2005 (M.E. Hegi)]; Swiss National Science Foundation (3100AO-108266/1; M.E. Hegi); and Novartis Research Foundation (M. Degen and R.C. Ehrismann).

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 Beatrice Dolder and Elisabeth Taylor for outstanding technical assistance, Eytan Domany for the SPIN software, Juliane Hänggi for manuscript preparation, Freddy Radtke for the Notch1 construct, Tasuku Honjo for the dominant negative RBPJκ plasmid, and Jon Aster for the dominant negative MAML1 construct.