Differential Gene Expression Analysis Reveals Activation of Growth Promoting Signaling Pathways by Tenascin-C Free

The extracellular matrix surrounding tumor cells is known to be different from the extracellular matrix in normal tissues, and it was shown in numerous studies that tenascin-C is an extracellular matrix protein highly up-regulated in many different cancers (reviewed in ref. 1). We have previously shown that adhesion of tumor cells to fibronectin in the presence of tenascin-C results in compromised cell spreading and enhanced proliferation of tumor cells in comparison to cells grown on fibronectin alone (reviewed in ref. 1). The mechanism of how tenascin-C inhibits cell spreading on fibronectin includes competition of tenascin-C with syndecan-4, an important coreceptor of integrin α₅β₁ in binding to fibronectin (2).

Cell shape and actin polymerization determine intracellular signaling, as for example, nonpolymerized actin can induce expression of c-Fos (3). Upon adhesion to fibronectin, cells spread by polymerizing actin and forming actin stress fibers (reviewed in ref. 4). Establishment of actin stress fibers is a complex process that involves, among other mechanisms, RhoA-dependent polymerization of G-actin and actin fiber stabilization by tropomyosins (reviewed in ref. 5). On a fibrin-based fibronectin/tenascin-C substratum, RhoA activation was found to be inhibited, which prevented fibroblasts from spreading and from establishing actin stress fibers (6). Although cells also remain rounded on a fibronectin/tenascin-C substratum, RhoA activation failed to restore actin stress fiber formation (7), suggesting that a distinct mechanism is responsible for the inhibition of cell spreading on the mixed fibronectin/tenascin-C substratum.

To discover novel mechanisms of tenascin-C action, we describe in this article global changes of transcript levels in tumor cells grown on fibronectin in the presence or absence of tenascin-C to determine substratum-specific signaling pathways involved in the regulation of cell adhesion and proliferation. We found that tenascin-C down-regulates tropomyosin 1 (TM1) expression and that forced expression of TM1 restores cell spreading with the formation of actin stress fibers and focal adhesions. Furthermore, we discovered changes in several transcripts that have an impact on mitogen-activated protein kinase (MAPK) signaling and on Wnt signaling and confirmed that transcript levels corresponded with increased protein expression and the induction of the respective signaling pathways.

T98G glioblastoma cells (CRL-1690; American Type Culture Collection, Manassas, VA) were cultured in DMEM (Sigma, Bucks, Switzerland) containing 10% FCS (Sigma) at 37°C and 5% CO₂. For measuring DNA replication, preparation of RNA, and protein, cells were grown in medium supplemented with 1% FCS. Cells were seeded onto substrata containing fibronectin or fibronectin and tenascin-C, at 1 μg/cm² (10 μg/mL) each (2). In comparison to fibronectin, proliferation of T98G cells as determined by [³H]thymidine incorporation was similarly enhanced on the fibronectin/tenascin-C substratum in 1% FCS-containing medium (data not shown) as in medium supplemented with 40 μg of platelet-derived growth factor BB (2).

Cells were serum starved for 19 hours and transferred to plates (100 mm; Corning Labware & Equipment, Corning, NY) coated with fibronectin or fibronectin and tenascin-C as described above in medium containing 1% FCS for 12 hours. RNA was extracted using Trizol Reagent (Invitrogen, Groningen, the Netherlands) according to the manufacturer’s protocol. RNA was additionally purified by the RNeasy protocol (Qiagen, Basel, Switzerland) until an A_{260 nm}:A_{280 nm} absorbance ratio of >1.9 in 10 mmol/L Tris-HCl (pH 7.5) was reached. The integrity of total RNA was checked according to the manufacturer’s protocol. A total of 10 μg of RNA was processed according to the protocol from Affymetrix and hybridized to the Affymetrix chip HG-U133A. In particular, RNA was reversed transcribed and in vitro transcribed into cRNA in the presence of biotinylated nucleotides. The cRNA was then hybridized to the oligonucleotides on the Affymetrix chip for 16 hours at 45°C, stained with streptavidin phycoerythrin conjugate, and scanned with the Affymetrix GeneChip 2500 scanner. To enforce reproducibility, the same experiment was repeated with a second set of RNA samples that was independently prepared from a different batch of cells giving rise to a total of eight different chip hybridizations.

The data generated by the scanner for the eight chips was analyzed to calculate the P values of the 16-way comparison. Candidates that were similarly regulated and with P < 0.003 in at least 11 of 16 comparisons were additionally analyzed. The replicate concordance analysis of the P values was performed using the Friedrich Miescher Institute (FMI) data-mining wizard to remote control the Affymetrix Microarray Suite analysis and to automate the transfer of results into GeneSpring (E. J. Oakeley, software available on request). The expression values were calculated using the Robust Multi-Array Analysis package (8), which is part of the BioConductor suite of R. Genes were also required to have a minimum expression of 50 expression units and a detection P of <0.05 in at least one condition and with a fold change of at least 1.5 between the control and the comparative condition. An ANOVA [GeneSpring 6.1 (Silicon Genetics, Redwood City, CA)] with a cutoff of 0.05 was performed on the 399 genes that remained, using a Benjamini and Hochberg false discovery rate multiple testing correction. A total of 373 genes passed this test, and the origins of these statistical differences were investigated using a Tukey posthoc test (Supplemental Table S1). They were classified with the Gene Ontology tool of GeneSpring (Supplemental Table S2).

For SYBR Green real-time RT-PCR (Applied Biosystems, Foster, CA), a total of 1.5 μg of RNA was reverse transcribed into cDNA using the Omniscript Reverse Transcriptase (Qiagen) following the manufacturer’s protocol. The following primer sets were used: glyceraldehyde-3-phosphate dehydrogenase, 5′-TCCTCTGACTTCAACAGCGACA-3′ and 5′-CGTTGAGGGCAATGCCA-3′; endothelin receptor type A, 5′-GCCATATTTTAGGACAGGTAAAATAACA-3′ and 5′-AACACACAAAAGGGCAGTACTTCTT-3′; c-Fos, 5′-GTCCTTACCTCTTCCGGAGATGT-3′ and 5′-ACTAACTACCAGCTCTCTGAAGTGTCA-3′; and Dickkopf 1 (DKK1): 5′-CCAAAAACCTGGAGTGTAAGAGCT-3′ and 5′-TGCCACACTGAGAATTTACAATACAGT-3′. A single amplification product was obtained per primer set as visualized by ethidium bromide staining after gel electrophoresis (data not shown). Analysis was performed using the QPCR software from Applied Biosystems. The number of cycles necessary to produce a product above background (C_t value) was recorded and, after normalization to the C_t value for glyceraldehyde-3-phosphate dehydrogenase, the relative expression was determined with the formula: relative expression = 10 ^(ΔCt/3.3).

Total RNA was prepared with the RNeasy kit (Qiagen) according to the manufacturer’s protocol from 4/5-confluent T98G and T98G:DKK1 cells that were grown for 20 hours in DMEM containing 10% FCS. The first-strand cDNA synthesis was done with the SuperScript II RT kit (Life Technologies, Inc., Rockville, MD), and random primer N6. PCR (35 cycles) was done at 57°C with specific primers 5′-GGATATCCCAGAAGAACCACACTGACTT-3′ and 5′-CACTGAAGATTCCTACATCCTTGGGATT-3′ (DKK1) and 5′-ACCACAGTCCATGCCATCAC-3′ and 5′-TCCACCACCCTGTTGCTGTA-3′ (glyceraldehyde-3-phosphate dehydrogenase), respectively, on 1 of 10 of 2.6 μg of in vitro-transcribed RNA. As a positive control served 185 ng of DKK1 plasmid.

Proteins were lysed in radioimmunoprecipitation assay buffer, and equal amounts (10 μg) as determined by Bradford assay (Bio-Rad Laboratories, Hercules, CA) were separated in 4 to 12% precasted Tris-Bis gels (Invitrogen), transferred onto Immobilon-P (Millipore, Bedford, MA) polyvinylidene difluoride membrane. Equal loading was visualized by Ponceau-S (Sigma) staining and confirmed by Western blotting for extracellular signal-regulated kinase (ERK) or α-tubulin. Membranes were blocked with 5% horse serum (Fluka, Milwaukee, WI) in PBS and incubated with antibodies against high molecular weight tropomyosins TM1-3 (mouse; Sigma), α-tubulin (mouse; Calbiochem, San Diego, CA), endothelin receptor type A (rabbit; Calbiochem), β-catenin [mouse; Transduction Laboratories (Lexington, KY) and BD Biosciences, Basel, Switzerland], non-phospho-β-catenin (clone 8E4, mouse; Upstate Biotechnology, Lake Placid, NY), ERK/MAPK (rabbit; Calbiochem), phospho-ERK/MAPK (mouse; Calbiochem), Id2 (mouse; Transduction Laboratories, BD Biosciences) and the secondary horseradish peroxidase-conjugated antimouse and antirabbit antibodies (Amersham, Piscataway, NJ). The signals from Western blotting for β-catenin were scanned with an Agfa Arcus Scanner (Agfa Gevaert N.V., Kontich, Belgium). Densitometric analysis was performed with the Scion Image Software (Scion Corporation, Frederick, MD).

Serum-starved cells were transferred onto substrata-coated dishes (4-well Cellstar; Greiner, Frickenhausen, Germany) in serum-free or 1% FCS containing medium, fixed with 4% paraformaldehyde, blocked with 5% horse serum, 0.5% Tween-20 in PBS, and incubated with the indicated antibodies (see above) or with tetramethylrhodamine isothiocyanate-phalloidin (Sigma), antivinculin (mouse; Sigma) and followed by incubation with tetramethylrhodamine isothiocyanate- and FITC-coupled goat antimouse and goat antirabbit antibodies (Alexa, Molecular Probes, Eugene, OR).

To generate stable transfectants T98G:TM1 and T98G:DKK1, cells were transfected with 1 μg of mouse DKK1 cDNA in pcDNA3.1 (gift from Cathrin Brisken; ISREC, Lausanne, Switzerland) and 1 μg of mouse TM1 cDNA in pCGE (9) together with 0.01 μg of the pCIneo plasmid (Invitrogen), respectively. Cells were selected with G418 and overexpression of DKK1, and TM1 was determined by semiquantitative RT-PCR, Western blotting, and immunofluorescence, respectively. Subclones of T98G:TM1 were obtained by limited dilution.

Tissue arrays have been constructed from archived paraffin blocks at the University Hospital in Lausanne, Switzerland, as described previously (10). The tissue arrays are composed of 190 glioblastoma multiforme (WHO grade 4) and 158 lower grade gliomas (WHO grades 1 to 3). The histopathology of all cases was reviewed by Robert C. Janzer, neuropathologist (Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland). Immunohistochemical determination for tenascin-C (monoclonal, B28.13, dilution 1:2500; ref. 11) and Id2 (1:10,000) 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 minutes) and overnight incubation with the primary antibody. The immunostaining was scored semiquantitatively in a range of 0 to 3, independently by two researchers. Low (score 0 to 1) versus high expression score (2 to 3) was retained for statistical analysis by Fisher’s exact test. In this study, we excluded gliomas WHO grade 3 and samples that were differently scored.

Proliferation of several tumor cell lines including T98G glioblastoma cells is stimulated by tenascin-C in the presence of a fibronectin substratum (2). Except for an involvement of syndecan-4 (2) and 14-3-3τ (12) little is known about how adhesion to tenascin-C alters intracellular signaling and causes enhanced tumor cell proliferation. To analyze global changes of gene expression by exposure of cells to tenascin-C, we performed an RNA profiling experiment. T98G cells were seeded onto fibronectin in the presence or absence of tenascin-C. RNA was prepared 12 hours after growth on the two substrata. After reverse and subsequent in vitro transcription, the cRNA was hybridized to the Affymetrix chip HG-U133A, and differentially expressed genes were identified using the Affymetrix Microarray Suite 5, FMI data analysis wizard and GeneSpring 5.1 programs. A total of 373 of the >24,000 genes passed the statistical analysis (see Materials and Methods) and were considered to be significantly up- or down-regulated on the fibronectin/tenascin-C substratum in comparison to fibronectin alone (Fig. 1; Table 1 and Supplemental Table 1). Some genes are represented by more than one probeset on the chip, usually representing alternate polyadenylated forms for the transcript. The hybridization results showed similar regulation of the alternative transcripts, providing additional evidence for the substratum-specific expression of the corresponding genes (Table 1). The tenascin-C target genes could be grouped according to functions in signaling, adhesion, cytoskeleton, transcription/DNA binding, cell cycle, and DNA repair/chromatin (Supplemental Table 2). We verified the substratum-specific expression of five key tenascin-C target genes by independent methods such as quantitative real-time RT-PCR, Western blotting, immunofluorescence, and functional analysis upon ectopic expression (Table 2).

Because tenascin-C is known to affect cell shape and to modulate the actin cytoskeleton, the down-regulation of mRNA encoding molecules involved in actin cytoskeletal dynamics such as TM1, zyxin, dystrophin, and the cdc42 effector protein 3 on the mixed fibronectin/tenascin-C substratum in comparison to fibronectin alone was of particular interest (Table 1). Especially intriguing was the down-regulation of the high M_r TM1, a protein involved in stabilizing actin stress fibers. Therefore, we investigated the expression of TM1 in more detail by Western blotting and immunofluorescence and by performing functional studies. In T98G cells grown for 4 and 12 hours on fibronectin/tenascin-C, the expression of TM1 was several fold reduced in comparison to cells grown on fibronectin (Fig. 2,A). As determined by Western blotting with an antibody that recognizes high molecular weight tropomyosins 1, 2, and 3, T98G cells express little to no TM2 and TM3 (Fig. 2 A).

To address whether reduced TM1 expression levels were responsible for the inhibition of cell spreading in the presence of tenascin-C, we attempted to rescue cell spreading of T98G cells by overexpression of TM1. We first confirmed overexpression of TM1 in the T98G:TM1 cells by immunofluorescence (Fig. 2,B) and Western blotting (data not shown). We then analyzed cell spreading of parental T98G and T98G:TM1 (pool) cells on fibronectin in the presence or absence of tenascin-C by immunofluorescence microscopy (Fig. 2,C). Cells were fixed and stained with tetramethylrhodamine isothiocyanate-phalloidin and an antivinculin-specific antibody, respectively, and actin stress fibers and focal adhesions were visualized by immunofluorescence microscopy. Whereas parental T98G cells did not form actin stress fibers and focal adhesions on fibronectin/tenascin-C, the majority of T98G:TM1 cells were found to spread on this substratum and generated actin stress fibers and vinculin-containing focal adhesions (Fig. 2,C). We also generated subclones A4, C6, and D10 of TM1-overexpressing cells. Overexpression was confirmed by Western blotting (Fig. 2,D). The subclones were plated onto fibronectin and fibronectin/tenascin-C for 4 hours, and their spreading behavior was analyzed. All of the subclones showed largely improved cell spreading over the parental cells transfected with empty vector (Table 3). In summary, our experiments revealed that tenascin-C down-regulates the expression of TM1 and that increased TM1 expression restores cell spreading.

The largest and most diverse group of molecules specifically regulated by the respective substratum is that involved in signal transduction (Table 1). The RNA profiling experiment revealed that the endothelin receptor type A, a G-protein coupled receptor with a potential function in glioma cells (13), was up-regulated 2-fold in T98G cells on fibronectin in the presence of tenascin-C. We confirmed this substratum-specific expression of endothelin receptor type A by Western blotting (Fig. 3,A) and by real-time RT-PCR (Table 4). A 2-fold increase in expression of endothelin receptor type A is potentially significant because signals emanating from the activated receptor are amplified before activation of downstream protein kinase C and ERK-MAPK signaling pathways (reviewed in ref. 14). To test whether ERK was activated by tenascin-C, we determined the phosphorylation levels of ERK1/2, representative for ERK1/2 activity, in T98G cells using a phospho-tyrosine–specific ERK antibody in Western blotting experiments. We found elevated phosphorylation of ERK1/2 in cells plated for 12 and 18 hours on the tenascin-C containing substratum compared with cells grown on fibronectin alone (Fig. 3,B), whereas overall levels of ERK1/2 remained the same (Fig. 3,B). Equal protein loading was also visualized by Ponceau-S staining (data not shown). In general, ERK1/2 expression levels were similar to that of α-tubulin, vimentin, and vinculin (data not shown), which confirms ERK1/2 as suitable loading control. The difference in phosphorylation was apparent only after 12 hours but was not present after 4 and 8 hours (data not shown), suggesting that ERK1/2 activation triggered by tenascin-C is an event that occurs later after cell adhesion to the fibronectin/tenascin-C substratum. A known downstream target of the endothelin receptor type A/MAPK pathway is c-Fos (reviewed in ref. 14). The RNA profiling experiment revealed an up-regulation of this molecule in dependence of tenascin-C (Table 1). By real-time RT-PCR, we confirmed that c-Fos transcription was indeed increased ∼2.5-fold in T98G cells grown on fibronectin/tenascin-C for 12 hours in comparison to fibronectin (Table 4). Thus, increased c-Fos transcription in the presence of tenascin-C may be due to enhanced signaling through endothelin receptor type A and associated ERK-MAPK activation.

Activation of the canonical Wnt signaling pathway is intimately linked to early events of tumorigenesis in many human cancers (reviewed in ref. 15). We found that this pathway was up-regulated in T98G cells that were grown on fibronectin in the presence of tenascin-C. In particular, the RNA profiling (Table 1 and Supplemental Table 1) and real-time RT-PCR experiment (Fig. 3 B) showed a 3- and 5-fold reduction of the transcript encoding the Wnt inhibitor DKK1 (16, 17) in cells plated on fibronectin/tenascin-C, respectively, in comparison to cells grown on fibronectin.

Because DKK1 expression was reduced by tenascin-C we wanted to know whether β-catenin was stabilized in the presence of tenascin-C. To investigate β-catenin stability, we used Western blotting and found that the protein levels of total β-catenin (Fig. 4,A) and of stabilized non-phospho-β-catenin encompassing Ser³⁷, whose unphosphorylated state correlates with enhanced in vitro transactivation activity (refs. 18, 19, 20; Fig. 4,B) were enhanced several fold in cells grown in the presence of tenascin-C for 4 and 12 hours in comparison to those cells grown in the absence of tenascin-C. This is consistent with the stronger β-catenin immunostaining of cells attached to fibronectin/tenascin-C than of cells attached to fibronectin (Fig. 4,C). Moreover, in cells plated on fibronectin, β-catenin was accumulated at the cell membrane, whereas β-catenin appears to be localized to nuclei of cells grown on the tenascin-C containing substratum (Fig. 4,C, arrow). These observations suggests that β-catenin expression and localization is regulated in T98G cells. Because expression of β-catenin was enhanced in cells grown on the fibronectin/tenascin-C substratum, we next sought to determine whether β-catenin activated transcription of TCF/LEF target genes in T89G cells. This possibility was supported by the enhanced expression of three known β-catenin/TCF target genes ubitorax (21), TCF/LEF1 (22), and Id2 (ref. 23; Table 1). We were able to confirm a very marked, tenascin-C–specific Id2 induction by Western blotting (Fig. 4,D). ERK1/2 expression levels, which were used as a loading control, were similar on the two substrata (Fig. 4 D). In contrast to Id2, ubitorax, and TCF/LEF1, other Wnt signaling target genes such as c-myc and cyclin D1 were not induced by tenascin-C. The differential activation of Wnt target genes by tenascin-C suggests that several signaling events are triggered by tenascin-C that may influence each other.

T98G cells express Frz1, 2, 6 and 7 but not 3, 4, 5, 9, and 10 under both culture conditions as indicated by a hybridization signal on the corresponding chip (Table 5). Analysis of the expression of Wnt ligands revealed that T98G cells expressed Wnt-5A and Wnt-5B but not Wnt-1, Wnt-2, Wnt-2B, Wnt-3, Wnt-4, Wnt-6, Wnt-7A, Wnt-8B, Wnt-10B, and Wnt-11 under both culture conditions (Table 5). Wnt-5A has been demonstrated to activate Wnt signaling, including β-catenin–dependent gene transcription through Frz7 (24), and to stimulate proliferation of hematopoietic stem cells through a not further identified Frz receptor (25). Because both Wnt5A and Frz7 are expressed in T98G cells, we suggest that this ligand/receptor pair activates Wnt signaling in T98G cells.

Next, we wanted to know whether wnt signaling was sensitive to DKK1 in T98G cells. Therefore, we transfected DKK1 into T98G cells, selected DKK1 overexpressors with G418, and confirmed DKK1 overexpression by RT-PCR (Fig. 5,A). To analyze the effect of DKK1 on β-catenin expression, lysates of T98G and T98G:DKK1 cells were first immunoblotted for β-catenin (Fig. 5,B). Secondly, conditioned medium of these cells was added to serum-starved T98G cells for 4 hours before lysis and Western blotting (Fig. 5,C). These experiments revealed that β-catenin levels were reduced 1.8- and 1.45-fold, respectively, in dependence of enhanced DKK1 expression, whereas α-tubulin expression was similar (Fig. 5, B and C). Kremen is required for an optimal effect of DKK1 on β-catenin levels (17). If its expression is limiting in T98G:DKK1 cells, the total reduction of β-catenin by overexpressed DKK1 may not be very pronounced. Experiments by Liu et al. (26) suggest that increased DKK1 expression can reduce β-catenin–specific transactivation. Taken together with our results, which show that DKK1 expression lowers β-catenin expression, we deduce that decreased expression of DKK1 in T98G cells by tenascin-C leads to the induction of the Wnt signaling pathway.

To analyze whether expression of tenascin-C and Id2 are potentially linked in gliomas, we determined the expression of tenascin-C and Id2 in lower grade gliomas of WHO grade 1 to 2 (lower grade gliomas) and glioblastomas (WHO grade 4) by immunohistochemical staining of tissue arrays composed of 158 lower grade gliomas and 190 glioblastomas (10). We observed that the majority of glioblastomas expressed high levels of extracellularly localized tenascin-C (67%) and high levels of nuclear Id2 (54%; Table 6 and Fig. 6, A and B). This was in contrast to lower grade gliomas where only a minor fraction of the samples expressed high levels of Id2 (4%) and high levels of tenascin-C (21%; Table 6). Moreover, 89% of glioblastomas that expressed high levels of tenascin-C also expressed high levels of Id2 (Table 7). In conclusion, high expression of tenascin-C appears to be linked to high Id2 expression and correlates with a more malignant phenotype in gliomas.

Tenascin-C is highly expressed in tumor stroma and its expression negatively correlates with the survival prognosis of patients with glioma, breast cancer, and gastric tumors. Tenascin-C appears to enhance tumor cell proliferation and migration in vivo and might play a role in angiogenesis and immune suppression (reviewed in ref. 1). Therefore, it is important to understand the biological responses toward tenascin-C and their underlying mechanisms for designing a strategy to counteract the function of tenascin-C in cancer development. Apart from a role for tenascin-C in inducing the expression of MMP-1, MMP-3, and MMP-9 in synovial fibroblasts (27) and the adaptor protein 14-3-3-τ in MCF7 breast carcinoma cells (12), nothing is known about how tenascin-C alters the expression of intracellular molecules involved in signaling pathways associated with proliferation. We addressed this question in T98G glioblastoma cells for which we recently showed that tenascin-C stimulates tumor cell proliferation by inhibiting syndecan-4 as coreceptor in integrin signaling (2). Here, we determined the RNA expression profile by Affymetrix chip hybridization of RNA prepared from T98G glioblastoma cells that were growth stimulated on a fibronectin/tenascin-C substratum and compared it to cells with a lower proliferation index grown on fibronectin. Our analysis revealed reproducible and profound changes in the expression profile of genes that are linked to proliferation and tumorigenesis and that are involved in signal transduction, transcription, actin cytoskeleton remodeling, cell cycle regulation, chromatin organization, and maintenance of genomic integrity. On the other hand, tenascin-C did not alter transcriptional regulation of genes linked to apoptosis. A RNA profiling approach reveals differential expression of many candidates genes. It is obvious that one needs to focus on a few potentially important candidates. Here, we addressed tenascin-C-specific expression of five candidate genes and confirmed their expression deduced from the Affymetrix chip hybridization data, which supports reliability of the RNA profiling results.

Cytoplasmic signaling is induced upon cell adhesion to adhesive extracellular matrix molecules and involves activation of small GTPases, c-Jun NH₂-terminal kinase, phosphatidylinositol 3′-kinase, and ERK/MAPK (reviewed in ref. 28). The associated pathways are often found to be deregulated in tumor cells. Our study reveals novel links of tenascin-C to gene expression and tumor growth promoting signaling events. In particular, tenascin-C stimulates activation of ERK1/2, which phosphorylates and activates ELK-1 and c-Fos (AP1), two transcription factors that are directly involved in induction of genes driving proliferation (reviewed in ref. 29), and indeed, c-Fos expression is elevated in the presence of tenascin-C. Moreover, our study showed that Sox4 and CASK were induced by tenascin-C (Table 1). Both molecules form nuclear complexes, Sox4 with syntenin (30) and CASK with Tbr1 (31), respectively, and function as transcriptional regulators of target genes. Because both complexes are implied in syndecan signaling (32) and fibronectin-induced syndecan-4 adhesion signaling is blocked by tenascin-C (2), it will be interesting to see whether these complexes have an altered transactivation activity in cells grown on a fibronectin/tenascin-C substratum in comparison to cells grown on fibronectin. The fact that expression of two molecules involved in syndecan signaling is regulated by tenascin-C is intriguing and supports an important role of tenascin-C in affecting the coreceptor function of syndecan-4 in integrin signaling.

Cell adhesion to fibronectin in the presence of tenascin-C profoundly changes cell shape in comparison to cells attached to fibronectin: cells are impaired in cell spreading and do not form actin stress fibers on the fibronectin/tenascin-C substratum (reviewed in ref. 1). One underlying mechanism might include the down-regulation of TM1 by tenascin-C. Tropomyosins are a family of actin-binding proteins that stabilize actin microfilaments (reviewed in ref. 33). They can be grouped into a high and a low molecular weight (M_r) species. In nonmuscle cells such as fibroblasts and epithelial cells, multiple high M_r TMs are expressed which are referred to as TM1, TM2, and TM3. TMs are rod-shaped molecules that bind to actin, thereby stiffening actin filaments and preventing their depolymerization (reviewed in ref. 34). TM1 can be phosphorylated (35) and acetylated (36) and both modifications enhance actin filament binding. Upon experimental cellular transformation, TM1 expression is often found to be down-regulated and correlates with the lack of actin stress fiber formation (37). TM1 expression in src-transformed cells restored cell spreading with the formation of actin stress fibers and focal adhesions. Moreover, these cells lost their anchorage-independent growth upon TM1 expression (38). These results together with the observation that TM1 is down-regulated in prostate carcinoma (39), breast carcinoma (40), and in highly malignant central nervous system tumors (41) suggest that TM1 behaves as tumor suppressor (42).

The importance of TM1 in actin fiber stabilization is supported by our experiments showing that overexpression of TM1 rescued cell spreading on fibronectin/tenascin-C, including the formation of actin stress fibers and focal adhesions. This is in contrast to TM2 and TM3, which failed to rescue focal adhesion formation on the fibronectin/tenascin-C substratum.⁴ High M_r TMs, including TM1, were found to be down-regulated or absent in 77% of high-grade pediatric astrocytomas and adult glioblastomas in contrast to lower grade astrocytomas (41). Because of the well-described tumor suppressor activity of TM1, the high expression of tenascin-C in glioblastoma (10, 43), and our observation that tenascin-C down-regulates TM1, it will be interesting to see whether reduced TM1 levels contribute to tumorigenesis of gliomas.

Another growth promoting signaling pathway that is affected by tenascin-C is the Wnt signaling pathway. Besides the involvement of integrin-linked kinase ILK (reviewed in ref. 44), little is known about how cell adhesion to the extracellular matrix affects the canonical Wnt/Frz signaling pathway, which is intimately involved in cellular transformation and cancer (reviewed in ref. 15). Upon binding of Wnt to its cognate receptor complex Frz/LRP5/6, a phosphorylation signaling cascade is induced that leads to suppression of β-catenin degradation by the proteasome. Thus, β-catenin can translocate to the nucleus where it changes the TCF/LEF/groucho repressor complex into a transcriptional activator and increases expression of target genes (reviewed in ref. 15). The Wnt/Frz signaling pathway is tightly regulated by molecules that prevent receptor/ligand complex formation. These include DKK1, which, together with Kremen, sequesters LRP5/6, an important coreceptor of Frz (17).

Our data suggest that tenascin-C–specific down-regulation of DKK1 is responsible for stabilization of β-catenin, which leads to enhanced transcription of TCF/LEF target genes ubitorax, TCF/LEF1, and Id2. Interestingly, Id2 is a negative transcriptional modulator, which, due to the lack of a DNA binding domain, inhibits other basic helix-loop-helix transcription factors through heterodimerization (reviewed in ref. 45). Id2 can induce cellular transformation and proliferation. Among other mechanisms, enhanced proliferation by Id2 seems to be caused by binding of Id2 to members of the E2F family, thereby deregulating their effect on the cell cycle machinery (reviewed in ref. 46). Expression of Id1–3 is significantly higher in high-grade compared with lower grade astrocytic tumors and correlates with higher proliferation indices (47). Our immunohistochemical study, using tissue arrays of a large number of glioblastomas and lower grade gliomas, confirms and extends this observation and shows a link between high tenascin-C and high Id2 expression in the majority of glioblastomas, whereas the majority of lower grade gliomas expressed little or no tenascin-C and little or no Id2. Taken together, these observations raise the possibility that activation of the wnt signaling pathway and especially increased expression of the Wnt target Id2 might play an important role in tumorigenesis of astrocytes leading to gliomas.

Thus far, it was unclear how binding of tumor cells to fibronectin attenuates their growth in cell culture and in vivo (48). Our data raise the interesting possibility that adhesion of tumor cells to fibronectin represses Wnt signaling, thus attenuating tumor cell proliferation. Cell binding to fibronectin occurs mainly through integrin α₅β₁ and requires syndecan-4 (49). Whether inhibition of the cell adhesion function of syndecan-4 by tenascin-C is responsible for DKK1 down-regulation and induction of wnt signaling remains to be determined.

In summary, our results suggests that tenascin-C enhances tumor cell proliferation by several mechanisms that include disruption of the actin cytoskeleton through reduced TM1 expression and alteration of gene expression through derepression of Wnt signaling and activation of MAPK signaling. Whether altered signaling by tenascin-C involving these pathways contributes to in vivo tumorigenesis remains to be determined.

We thank Claudia Bagutti for help with the real-time RT-PCR experiments, Richard Janssen (University of Nijmegen, Nijmegen, the Netherlands) for the tropomoyosin constructs, and Cathrin Brisken (ISREC, Lausanne, Switzerland) for the DKK1 expression plasmid. We also thank Herbert Angliker from the genomics facility of the FMI for assistance with the RNA profiling experiments and Robert C. Janzer (Centre Hospitalier Universitaire Vaudois) for reviewing the tissue arrays. We thank Monilola Olayioye (Department of Pharmacology and Toxicology, University of Ulm, Ulm, Germany) and Miguel Cabrita (Institute of Biochemistry, University of Basel, Basel, Switzerland) for critically reading the manuscript.