Endothelin Receptor Type B Counteracts Tenascin-C–Induced Endothelin Receptor Type A–Dependent Focal Adhesion and Actin Stress Fiber Disorganization Free

Tenascin-C, an extracellular matrix molecule of the tumor-specific microenvironment, counteracts the tumor cell proliferation–suppressing effect of fibronectin by blocking the integrin α₅β₁/syndecan-4 complex. This causes cell rounding and stimulates tumor cell proliferation. Tenascin-C also stimulates endothelin receptor type A (EDNRA) expression. Here, we investigated whether signaling through endothelin receptors affects tenascin-C–induced cell rounding. We observed that endothelin receptor type B (EDNRB) activation inhibited cell rounding by tenascin-C and induced spreading by restoring expression and function of focal adhesion kinase (FAK), paxillin, RhoA, and tropomyosin-1 (TM1) via activation of epidermal growth factor receptor, phospholipase C, c-Jun NH₂-terminal kinase, and the phosphatidylinositol 3-kinase pathway. In contrast to EDNRB, signaling through EDNRA induced cell rounding, which correlated with FAK inhibition and TM1 and RhoA protein destabilization in the presence of tenascin-C. This occurred in a mitogen-activated protein kinase/extracellular signal-regulated kinase kinase–dependent manner. Thus, tumorigenesis might be enhanced by tenascin-C involving EDNRA signaling. Inhibition of tenascin-C in combination with blocking both endothelin receptors could present a strategy for sensitization of cancer and endothelial cells toward anoikis. [Cancer Res 2007;67(13):6163–73]

Cancer is a product of the tumor-host microenvironment, where mutual stimulation of tumor and stromal cells induces tumor formation and progression into malignant metastasizing cancers. The adhesion modulatory extracellular matrix molecule tenascin-C is one factor in the tumor-specific microenvironment that is highly expressed in most solid tumors. Tenascin-C actions promote malignant transformation, uncontrolled proliferation, metastasis, angiogenesis, and escape from tumor immunosurveillance. Tenascin-C acts very early during malignant transformation as well as throughout tumor progression through distinct effects on various cell types within a tumor (1). We and others have previously shown that tenascin-C inhibits the tumorigenesis-suppressing activity of integrin α₅β₁-mediated cell adhesion to fibronectin by blocking syndecan-4 (2, 3) and downstream RhoA (4) and focal adhesion kinase (FAK; refs. 2, 5, 6) activation. This in turn triggers tumor cell proliferation, presumably by overriding the G₀ and G₁-S cell cycle checkpoints (7). Moreover, tenascin-C can activate a variety of oncogenic signaling pathways, such as epidermal growth factor receptor (EGFR; refs. 8, 9), extracellular signal-regulated kinase (ERK)/mitogen-activated protein kinase (MAPK), and Wnt, and can down-regulate the tumor suppressor–like molecule tropomyosin-1 (TM1; ref. 10).

Endothelins (ET1, ET2, and ET3) and their G protein–coupled receptors endothelin receptor (EDNR) subtypes A and B were originally found to be involved in the regulation of blood pressure (11). Endothelin receptor signaling plays an important role during embryonic development and in physiologic processes, such as neurotransmission, renal function, and regulation of cell proliferation (12). EDNRB signaling has a promigratory and proliferative effect on microvascular endothelial cells (13). Signaling through EDNRA also stimulated angiogenesis particularly by induction of vascular endothelial growth factor (14, 15). ET1 activates phospholipase C (PLC) β, thus increasing intracellular calcium ion levels and activation of protein kinase C. It also activates phosphatidylinositol 3-kinase (PI3K), c-Jun NH₂-terminal kinase (JNK), ERK/MAPK, and EGFR signaling. Moreover, ET1-induced signaling leads to activation of FAK and paxillin (16). Because signaling by EDNRA and EDNRB plays an important role in endothelial cell proliferation and survival, blocking these receptors provides a promising approach in clinical treatment of systemic pulmonary hypertension and chronic heart failure (17). Inhibition of endothelin receptor signaling may also be useful in cancer therapy because inhibition of EDNRB with a selective pentapeptidic antagonist (BQ788) inhibited growth of human melanoma cells in cell culture and in the nude mouse (18). In advanced prostate cancer, treatment with an EDNRA-selective inhibitor (ABT-627) delayed disease progression in patients with hormone-refractory prostate cancer (19).

We showed that tenascin-C induces expression of EDNRA. This was accompanied by enhanced phosphorylation of ERK1/2 and induction of c-Fos (10), all downstream targets of EDNRA. Here, we investigated the possibility that EDNRA signaling contributes to the tenascin-C–induced cell phenotype. We show that EDNRA signaling induced cell rounding in the presence of tenascin-C. This occurred by blocking FAK and paxillin activation and inhibiting RhoA and TM1 protein stability in a MAPK/ERK kinase (MEK)-dependent manner. In contrast to EDNRA, signaling by EDNRB counteracted cell rounding by tenascin-C and stimulated cell spreading on a fibronectin/tenascin-C (FN/TN-C) substratum in an EGFR-, PLC-, JNK-, and PI3K-dependent but MEK-independent manner. In the absence of syndecan-4 activation, EDNRB signaling restored focal adhesion assembly by activating FAK and paxillin and reorganized the actin cytoskeletal by normalizing RhoA and TM1 expression.

Construction and purification of his-tagged human tenascin-C. The cDNA encoding human tenascin-C (HxBL.pBS) encompassing all alternative fibronectin type III repeats (20) was subcloned into the pCEP-Pu vector (21). A COOH-terminal his-tag (six additional his residues) was introduced by PCR with primer PhTN-C R-2 (CGGGATCCTAATGATGATGATGATGATGATGTGCCCGTTTGCGCCT). A three-piece ligation of NotI/NdeI, a NdeI/6his-BamHI and a NotI/BamHI vector fragments, gave rise to pCEP-huTNC-his. The construct was confirmed by restriction enzyme analysis, PCR, and partial sequencing. After transfection of pCEP-huTNC-his into HEK293-EBNA1 cells [American Type Culture Collection (ATCC)], human tenascin-C expression and secretion were determined by immunoblotting with the monoclonal antibody B28.13 (22). For large-scale production of recombinant human tenascin-C, 293:pCEP-huTNC-his cells were grown to confluency in the presence of 2.5 μg/mL puromycin (Sigma) in DMEM and 10% FCS. For collection of conditioned medium, cells were washed with PBS before growth in serum-free DMEM without puromycin for 2 days. Cells were recovered in DMEM supplemented with 10% FCS and 2.5 μg/mL puromycin for 1 day before transfer to serum-free medium. This cycle was repeated up to eight times with no reduction in yield during prolonged culture. Proteins from conditioned medium were ammonium sulfate precipitated and dialyzed against PBS/0.01%Tween 20 before chromatography on a gelatin-agarose column (Sigma). The eluate was passed over a nickel column (ProBond resin, Invitrogen), and bound tenascin-C was eluted with 300 mmol/L imidazole (Sigma), 250 mmol/L sodium phosphate (pH 7.4), 450 mmol/L NaCl, and 0.01% Tween 20 and dialyzed against PBS/0.01% Tween 20. The purity and absence of contamination by fibronectin was determined by GelCode staining and immunoblotting. The biological activity of recombinant human tenascin-C was compared with the chicken tenascin-C used previously (2). In cell adhesion assays, T98G cells were plated for different times onto pure tenascin-C, fibronectin, or mixed fibronectin and fibronectin/fibronectin type III repeat 13 (FNIII13) substrata containing human or chicken tenascin-C. As with chicken tenascin-C (2), cells did not spread on FN/TN-C unless syndecan-4 was activated with FNIII13. To test whether addition of a his-tag at the COOH terminus of tenascin-C interfered with heparin binding to the fibrinogen globe, heparin binding of chicken and human tenascin-C was compared by ELISA. No differences were detected (data not shown).

Cell plating, inhibitor studies, and preparation of cell lysates. Human T98G glioblastoma, J82 urinary bladder carcinoma, and MDA-MB435 breast carcinoma cells (ATCC) were grown in DMEM supplemented with antibiotics and 10% FCS (Sigma). Cells were transferred into DMEM and 10% FCS 24 h before the experiment and serum starved for 18 h. Cells were trypsinized and replated after inhibition of trypsin with 100 ng/mL trypsin inhibitor (Sigma) in serum-free DMEM onto 10-cm² dishes (Falcon Becton Dickinson) coated with equimolar amounts of fibronectin and tenascin-C (1 μg/cm²). Finally, 1% heat-inactivated bovine serum albumin (BSA; Serva) was used to block the uncoated surface before UV sterilization for 15 min in a sterile bench. Because the inhibitors were dissolved in DMSO and DMSO interfered with cell spreading on fibronectin, cells were plated on the different substrata in serum-free medium 1 or 12 h before incubation with 100 ng/mL EGF, 20 nmol/L ET1, 20 nmol/L ET3, conditioned medium (taken from 48-h serum-free cultures of T98G cells), or inhibitor PKI166 (2 μmol/L; Novartis), UO126 (25 μmol/L; Calbiochem), wortmannin (20 μmol/L), SP600125 (20 mmol/L), U73122 (20 nmol/L), BQ123 (100 nmol/L), and BQ788 (100 nmol/L; Sigma) for 4 h followed by lysis in sample buffer [250 mmol/L Tris-HCl (pH 7.0), 10% SDS, 50% glycerol, bromphenol blue, 100 mmol/L DTT].

Immunoblotting. Equal amounts of protein were separated in 8% self-made or 4% to 12% precast Bis-Tris-glycine gels (Invitrogen, Novex), transferred onto polyvinylidene difluoride nylon membrane (Immobilon-P, Millipore), and stained with 0.2% Ponceau-S (Sigma) in 7.5% trichloroacetic acid for confirmation of equal protein loading before blocking the membrane in 10% horse serum (or 1% BSA), TBS, and 1% Tween 20. For immunoblotting, the following mouse monoclonal antibodies against TM1, TM2, and TM3 (TM311, 1:1,000; Sigma); vinculin (hVIN-1, 1:1,000; Sigma); α-tubulin (Ab-1, 1:5,000; Oncogene); RhoA (26C4, 1:1,000; Santa Cruz Biotechnology, Inc.); phosphorylated ERK1/2 (P-ERK1/2; 1:1,000; Cell Signaling); FAK and P-Y397-FAK (1:1,000; BD Biosciences); rabbit polyclonal antibodies, paxillin, and P-Y118-paxillin (1:1,000; Abcam); and ERK1/2 (1:1,000; Cell Signaling), P-S910-FAK (1:1,000; Biosource), and secondary horseradish peroxidase–coupled antibodies (Amersham) were used. Binding of antibodies was detected with enhanced chemiluminescence plus (Amersham).

Real-time reverse transcription-PCR. Total RNA was isolated from three independent plates using the RNeasy Mini kit (Qiagen) and RNase-free DNase set (Qiagen) following the manufacturer's instructions. From 0.5 to 1 μg of total RNA, single-strand cDNA was generated by using the SuperScript III First-Strand Synthesis Super Mix (Invitrogen) with random hexamers. Expression of the respective gene was detected by quantitative reverse transcription-PCR (RT-PCR) on an ABI Prism 7000 Taqman using SYBR Green PCR MasterMix (Applied Biosystems) with the following primers (5′ to 3′): glyceraldehyde-3-phosphate dehydrogenase (GAPDH), ATCTTCTTTTGCGTCGCCAG (forward) and AATCCGTTGACTCCGACCTTC (reverse); tenascin-C, TGCCCATATCTCAGGGCTAC (forward) and GATGCCATCCAGGAAACTGT (reverse); EDNRA, GCCATATTTTAGGACAGGTAAAATAACA (forward) and AACACACAAAAGGGCAGTACTTCTT (reverse; ref. 10); EDNRB, TCACCTAAAGCAGAGACGGGAA (forward) and AGGACCAGGCAAAAGACGG (reverse); αTM (TPM1), GCACCGAAGATGAACTGGACAA (forward) and CATCGGTGGCCTTTTTCTCTG (reverse); βTM (TPM2), CCAACAACTTGAAATCCCTGG (forward) and CTTTGGTGGAATACTTGTCCGC (reverse); and RhoA, GCAGGTAGAGTTGGCTTTATGG (forward) and CTTGTGTGCTCATCATTCCGA (reverse; ref. 23). Primers were designed with the ABI Primer Express software (Applied Biosystems). Relative expression of the respective gene was determined after normalization to GAPDH and calculated with the following formula: relative expression level = 2^−ΔΔCT.

Immunofluorescence. Cells were serum starved before plating on fibronectin and FN/TN-C for the indicated time points in serum-free medium plus ET1 or ET3. Cells were fixed in 4% paraformaldehyde and stained with the indicated primary and secondary FITC-labeled antibodies or with TRITC-labeled phalloidin (Sigma). Images were captured using a Nikon Diaphot300 (Nikon video microscope with OpenLab program) with 40× and 100× objectives for immunofluorescence and a Leica Leitz DMIL microscope with a 10× objective for phase contrast.

Tissue microarrays and immunohistochemistry. Tissue microarrays (TMA) with 190 glioblastoma and 158 gliomas WHO grade I to III have been constructed from archived paraffin blocks as described (24). Expression of tropomyosin and tenascin-C in gliomas grades I, II, and IV was determined by immunohistochemistry with antibody TM311 (dilution, 1:3,500) and B28.13 (22), respectively, as described before (10). The staining was scored semiquantitatively in a range of 0 to 3, independently by two researchers.

Kinetics of EDNRA induction by tenascin-C. We previously described that a tenascin-C–containing fibronectin substratum triggered EDNRA expression in T98G cells 12 h after contact with the substratum both at RNA and protein level (10). Here, we investigated the kinetics of EDNRA induction by tenascin-C with real-time RT-PCR. EDNRA RNA levels did not vary between cells plated on fibronectin or FN/TN-C at 1 and 3 h (Fig. 1A). However, a slight and 2.7-fold increase was observed after 5 and 19 h, respectively, in cells plated on FN/TN-C. We also determined expression of EDNRB and found that, in contrast to EDNRA, RNA levels of EDNRB were low and did not change at any of the time points mentioned above (data not shown). Thus, in contrast to EDNRB, expression of EDNRA is strongly induced by tenascin-C at RNA and protein level.

ET1 signaling restores cell adhesion on a FN/TN-C substratum. To examine whether EDNR signaling affects tenascin-C–induced cell rounding, we determined cell spreading of T98G cells on fibronectin in the presence and absence of tenascin-C on stimulation with the EDNR ligand ET1 in serum-free medium. As previously observed, we confirmed that cells did not spread on FN/TN-C. But stimulation with ET1 restored cell spreading on the FN/TN-C substratum (Fig. 1B). This was accompanied by actin stress fiber and focal adhesion formation 1, 4, and 12 h after plating of the cells (Fig. 1C; Supplementary Fig. S1).

To determine how ET1 inhibited tenascin-C–induced cell rounding, the expression and function of tenascin-C target molecules was assayed. T98G cells were grown for 4 h on fibronectin or FN/TN-C in the presence or absence of ET1. Cell extracts were then analyzed by immunoblotting for FAK, active FAK (P-Y397-FAK), RhoA, and TM1. Autophosphorylation of FAK at Y397 and expression of RhoA and TM1 were largely reduced in serum-free medium in the presence of tenascin-C. However, on stimulation with ET1, expression of P-Y397-FAK, RhoA, and TM1 was restored to levels observed on fibronectin alone. On fibronectin, ET1 also accelerated spreading and enhanced phosphorylation of FAK (Fig. 1D).

ET1 inhibits tenascin-C signaling through activation of EDNRB. T98G cells express two ET1 receptors: EDNRA (Fig. 1A) and EDNRB (25). To examine EDNR-specific effects, we analyzed cell adhesion on FN/TN-C in the presence of specific inhibitors for EDNRA (BQ123) and EDNRB (BQ788) on stimulation with ET1. T98G cells were plated for 1 or 12 h on fibronectin or FN/TN-C before addition of ET1 and the respective EDNR inhibitor. Cells were then analyzed for adhesion followed by lysis and immunoblotting. As shown in Fig. 2A, inhibition of EDNRB with BQ788 caused cell rounding, thus blocking ET1-induced cell spreading on FN/TN-C 5 h (data not shown) and 16 h after plating. Inhibition of EDNRB by BQ788 also blocked restoration of autophosphorylation of FAK and reexpression of RhoA and TM1 on FN/TN-C at 5 and 16 h (Fig. 2B). In contrast, the EDNRA-specific inhibitor BQ123 did not block restoration of cell spreading by ET1. Inhibition of EDNRA through BQ123 did also not affect FAK phosphorylation and RhoA and TM1 reexpression induced by ET1.

To confirm that signaling specifically through EDNRB inhibits tenascin-C–induced cell rounding, T98G cells were stimulated with ET3, an EDNRB ligand that has several fold higher affinity for EDNRB than for EDNRA (11). Subsequently, cell adhesion was determined before lysis and immunoblotting. ET3 also stimulated cell spreading and formation of focal adhesions and actin stress fibers in the presence of tenascin-C (Fig. 1C). Moreover, spreading was EDNRB dependent because it was blocked with BQ788 (Supplementary Fig. S2A). As for ET1, activation of EDNRB with ET3 restored TM1 and RhoA expression and autophosphorylation of FAK in the presence of tenascin-C (Supplementary Fig. S2B).

ET1 restores cell spreading by mediating focal adhesion assembly. EDNRB signaling leads to cell matrix adhesion and appearance of numerous focal contacts even in the presence of tenascin-C (Fig. 1C). To investigate in more detail how EDNRB signaling could enhance focal adhesion assembly, we examined whether FAK was phosphorylated at S910, a site that is involved in binding of FAK to paxillin (26, 27). Whereas P-S910-FAK levels were low on FN/TN-C, phosphorylation of FAK at S910 returned to levels as on fibronectin on stimulation with ET1 (Fig. 2C). Next, we examined whether FAK was functional in phosphorylating paxillin at Y118, one of its cognate phosphorylation sites (28), after stimulation with ET1. In cells plated on FN/TN-C in the absence of ET1, phosphorylation at Y118 in paxillin was abolished (Fig. 2C). In contrast, cells plated on FN/TN-C in the presence of ET1 showed high levels of phosphorylated paxillin (Fig. 2C). Altogether, these data show that ET1 signaling through EDNRB restored FAK and paxillin activities in the presence of tenascin-C.

Activation of EDNRB induces cell spreading in an EGFR-, PLC-, PI3K-, and JNK-dependent manner. We next investigated by which pathway EDNRB inhibits tenascin-C–mediated cell rounding. Because EDNR signaling activates EGFR, we addressed whether EDNRB-induced cell spreading on FN/TN-C involves EGFR function. Cells were plated on FN/TN-C for 5 h together with ET1 and the inhibitor PKI166, which prevents autophosphorylation of EGFR (29). Cell adhesion was documented before lysis and immunoblotting. Inhibition of EGFR by PKI166 completely blocked cell spreading by ET1/EDNRB (Fig. 3A). Moreover, as shown in Fig. 3B, PKI166 efficiently blocked ERK1/2 phosphorylation and ET1-induced TM1 and RhoA expression in the presence of tenascin-C. However, ET1-induced autophosphorylation of FAK was not affected by PKI166 (Fig. 3B). Thus, ET1 restores TM1 and RhoA protein expression in the presence of tenascin-C through activation of EGFR. In contrast, activation of FAK by EDNRB is independent of EGFR signaling, but FAK activation alone is not sufficient to mediate cell spreading.

ET1-treated cells were allowed to adhere on FN/TN-C in the presence of inhibitors for PLC, MEK, PI3K, and JNK. Inhibition of PLC with U73122 blocked ET1-induced cell spreading on FN/TN-C (Fig. 3A), which correlated with low levels of phosphorylated FAK, TM1, and RhoA (Fig. 3C). Examining MEK and PI3K downstream of EGFR and PLC revealed that the MEK inhibitor UO126 did not affect ET1-induced cell spreading on the tenascin-C substratum. This was in contrast to the PI3K inhibitor wortmannin that diminished cell spreading (Fig. 3A). Cell rounding by wortmannin correlated with largely reduced RhoA and TM1 levels and lack of FAK autophosphorylation in the presence of ET1 on FN/TN-C (Fig. 3D). These data suggest that EDNRB-induced cell spreading on FN/TN-C depends on EGFR, PLC, and PI3K but does not involve the MEK pathway.

Next, we examined whether JNK, a downstream effector of EDNR and EGFR signaling, is involved in cell spreading on FN/TN-C on activation of EDNRB. Inhibition of JNK with SP600125 not only prevented ET1-induced cell spreading (Fig. 3A) but also blocked FAK autophosphorylation and TM1 reexpression in the presence of tenascin-C, which was in contrast to RhoA, the expression of which was not altered on JNK inhibition (Fig. 3E). Altogether, these data suggest that EDNRB activation counteracts tenascin-C–induced cell rounding through EGFR and downstream PLC, JNK, and PI3K and that all three molecules, FAK, RhoA, and TM1, need to be active to allow cell spreading on FN/TN-C.

Inhibition of EDNRA induces cell spreading in the presence of tenascin-C. Because plating cells on a tenascin-C–containing substratum leads to increased expression of EDNRA, we asked whether EDNRA signaling could induce cell rounding in cells plated on FN/TN-C. We determined cell adhesion in the presence of both EDNR inhibitors at a time point when EDNRA expression was induced by tenascin-C. T98G cells were plated in serum-free medium for 12 h before addition of ET1 and the respective inhibitor for an additional 30 min (Supplementary Fig. S3) or 4 h (Fig. 4). We observed that, although cells were round 12 h after plating on FN/TN-C, they spread with ET1 in the presence of both inhibitors (Fig. 4A). Because inhibition of EDNRB by BQ788 prevented cell spreading (Fig. 2A), restoration of cell spreading under these conditions was due to inhibition of EDNRA. Already 30 min on addition of exogenous ET1, T98G cells were spread on FN/TN-C (Supplementary Fig. S3A). Thus, inhibition of EDNRA rapidly overcomes cell rounding on FN/TN-C. Simultaneous inhibition of both EDNRs in the absence of exogenously added ET1 also induced cell spreading, suggesting that EDNRA is active in the absence of exogenously provided ET1 (Fig. 4A; Supplementary Fig. S3A). EDNRA is presumably activated in T98G cells on FN/TN-C by secreted endothelins (25).⁷

To prove this possibility, cells on FN/TN-C were stimulated with conditioned medium derived from 48-h serum-free cultures of T98G cells. Cells remained round on the FN/TN-C substratum with conditioned medium but spread under these conditions on inhibition of EDNRA with BQ123 (Fig. 4A). Immunoblotting revealed that FAK autophosphorylation and expression of RhoA and TM1 were restored at 30 min (Supplementary Fig. S3B) and 4 h (Fig. 4B) on blocking of EDNRA. In addition, inhibition of both endothelin receptors enhanced ERK1/2 phosphorylation on FN/TN-C, which was in contrast to the control with very low P-ERK1/2 levels. To examine whether activation of EDNRA by tenascin-C causes rounding in MDA-MB435 breast carcinoma and J82 urinary bladder carcinoma cells that express EDNRA as determined by real-time RT-PCR,⁶ we determined adhesion of these cells on fibronectin and FN/TN-C on treatment with BQ123. As for T98G, these cells did not spread on FN/TN-C and inhibition of EDNRA with BQ123 restored cell spreading on FN/TN-C (Supplementary Fig. S4A). Thus, activation of EDNRA mediates rounding of all three cell lines by tenascin-C.

EDNRA-specific cell rounding by tenascin-C is MEK dependent. We wanted to know whether the MAPK pathway, which blocks expression of TM1 (30, 31), is involved in EDNRA-stimulated inhibition of TM1 expression by tenascin-C. Therefore, cells plated onto FN/TN-C were stimulated with ET1 in the presence of the EDNRB inhibitor BQ788, which allows to determine signaling in response to EDNRA activation. Inhibition of the MEK pathway by UO126 restored cell spreading in combination with BQ788 (Fig. 4C). Thus, MEK is downstream of EDNRA. UO126 also normalized TM1 and RhoA expression and FAK autophosphorylation on FN/TN-C to levels as on fibronectin on treatment with ET1 and BQ788 (Fig. 4D). Thus, activation of MEK through EDNRA contributes to tenascin-C–induced cell rounding on fibronectin by blocking FAK activation and RhoA and TM1 expression.

Regulation of TM1 and RhoA RNA and protein levels on FN/TN-C. To determine whether tenascin-C affects RNA levels of αTM (coding for TM2 and TM3), βTM (coding for TM1), and RhoA, a real-time RT-PCR experiment was done on RNA prepared from cells that were grown in the presence or absence of tenascin-C for different time points. In T98G cells, RNA levels of αTM started to drop at 5 h and remained about 40% to 50% below those on fibronectin after 19 h on FN/TN-C (Fig. 5A). A similar observation was made for J82 cells (Supplementary Fig. S5B). This was in contrast to earlier time points (1 and 3 h) where no substratum-specific differences in αTM expression were observed in T98G cells (Fig. 5A). Similarly, βTM and RhoA RNA levels were not down-regulated on FN/TN-C in comparison with fibronectin at any time point (Fig. 5B and C). These results show that RNA levels of αTM, βTM, and RhoA are not reduced in the presence of tenascin-C at early time points.

Next, we examined TM1 and RhoA protein stability on FN/TN-C and observed that TM1 and RhoA protein levels were restored on FN/TN-C to levels as observed on fibronectin on treatment with the proteasome and calpain inhibitor ALLN (Fig. 5D). In contrast, phosphorylation of FAK was not restored with ALLN and this observation correlated with the inability of cells to spread on FN/TN-C under these conditions (Table 1).

Because αTM levels dropped significantly after 19 h, and BQ123 restored TM1 expression on FN/TN-C, we examined whether EDNRA signaling contributed to reduced αTM RNA levels in T98G and J82 cells. But inhibition of EDNRA by BQ123 only slightly increased αTM levels on FN/TN-C in both cell lines (Supplementary Fig. S5). In summary, our data suggest that enhanced proteolysis of TM1 and RhoA is the major mechanism by which EDNRA negatively regulates expression of RhoA and TM1 on FN/TN-C.

Expression of tenascin-C and EDNRA in cancer tissue. Next, we compared gene expression levels of tenascin-C and EDNRA in 80 glioblastoma and 4 nontumoral brain tissues (GeneChip U133Plus, Affymetrix;⁸

ref. 32) and by systematically evaluating gene expression changes using the Oncomine 3.1 platform (33). Tenascin-C and EDNRA mRNA expression revealed some correlation in glioblastoma (Pearson correlation = 0.47) with high expression in gliomas and low expression in nontumoral brain tissue (Fig. 6A; Supplementary Table S1). This is compatible with the notion that tenascin-C might have some role in EDNRA regulation in glioblastoma. Moreover, the majority of glioblastoma exhibited also a high expression of EDNRB (Fig. 6A; Supplementary Table S1) with EDNRA and EDNRB showing a negative correlation (Pearson correlation = −0.25; Fig. 6A). On protein level, as determined by immunohistochemistry on TMAs, tropomyosin expression displayed different staining patterns. Tumor tissue was homogenously positive or negative for TM1-3 whereas staining of aberrant blood vessels was prominent in most tumors (Fig. 6B). There was neither a correlation of TM1-3 protein expression with tumor grade (P = 0.09, Fisher's exact test) nor TM1-3 expression with tenascin-C protein expression (P = 0.09, Fisher's exact test) in glioblastoma. These observations suggest that other factors than tenascin-C influence tropomyosin expression in gliomas under steady-state conditions in a cell type–specific manner.

Searching the Oncomine database for a potential linked expression of tenascin-C and EDNRA in other human cancers revealed a high tenascin-C expression and increased EDNRA expression in malignant pancreatic ductal carcinoma, seminoma, bladder carcinoma, and breast carcinoma and in metastatic ovarian carcinoma (Supplementary Fig. S6; Supplementary Table S1). On the contrary, a lowered tenascin-C expression in metastatic prostate and a subset of colon carcinoma correlated with reduced EDNRA expression. These observations are compatible with a potential regulation of EDNRA by tenascin-C in these cancers. EDNRB and αTM were also increased in glioma, ovarian carcinoma, breast carcinoma, seminoma, and pancreatic ductal carcinoma (Supplementary Table S1), which also matches with a potential regulation of αTM by EDNRB in these cancers.

Previously, we and others showed that the function of FAK (2, 5), RhoA (4), and TM1 (10) is compromised in cells grown on FN/TN-C. This occurred in a syndecan-4–dependent manner because activation of syndecan-4 restored cell spreading and FAK activation in the presence of tenascin-C (Fig. 7A; refs. 2, 5). Here, we observed that tenascin-C does not only interfere with RhoA activation (4) but also with RhoA protein stability. Moreover, we showed a complex repression of TM1 by tenascin-C. Although TM1 protein levels were reduced by tenascin-C, RNA levels were not lowered by tenascin-C at any time point up to 19 h. In contrast, TM1 protein stability is largely reduced by tenascin-C because inhibition with ALLN restored TM1 protein levels on FN/TN-C. After 19 h, RNA levels of αTM coding for TM2 and TM3, which are less expressed in T98G cells, significantly dropped. Reduced αTM levels might contribute to reduced TM1 expression because TM1 forms heterodimers with TM2 and TM3.

Our results show that inhibition of FAK, RhoA, and TM1 is responsible for cell rounding by tenascin-C. To override the cell spreading block by tenascin-C, the function of all three molecules, FAK, RhoA, and TM1, needs to be restored (Fig. 7B; Table 1). This can be accomplished by activation of EDNRB, which induces cell spreading on FN/TN-C presumably due to FAK and paxillin activation and RhoA and TM1 protein stabilization. EDNRB-induced cell spreading on FN/TN-C was specific because it could be blocked with BQ788, a specific inhibitor of EDNRB (18) that is used in treatment of pulmonary hypertension (17). Restoration of expression and function of FAK, RhoA, and TM1 by EDNRB signaling involved EGFR, PLC, PI3K, and JNK because specific inhibitors for these enzymes blocked EDNRB-specific cell spreading and reexpression of the tenascin-C target molecules.

We observed that activation of paxillin, an important component of integrin (28), syndecan-4 (34), and ET1 (11) signaling, plays a key role in EDNRB-induced cell spreading on FN/TN-C because EDNRB signaling induced phosphorylation of paxillin at Y118, one of its cognate phosphorylation sites for FAK (28), in the presence of tenascin-C. Whereas phosphorylation of FAK at S910 and of paxillin at Y118 was absent or very low in the presence of tenascin-C in serum-free medium, EDNRB signaling restored phosphorylation in FAK and paxillin. We conclude that phosphorylation of FAK at S910 is involved in restoration of cell spreading on FN/TN-C presumably by phosphorylating paxillin. Localization of paxillin into focal adhesions is dependent on syndecan-4, which involves binding of syndesmos to syndecan-4 and paxillin (34). Tenascin-C impairs syndecan-4 activation by fibronectin (2) and prevents localization of paxillin into focal adhesions (5). Now, we showed that EDNRB activation restores focal adhesion formation on the FN/TN-C substratum. This suggests that EDNRB signaling restores paxillin function on a FN/TN-C substratum through bypassing the requirement for syndecan-4 for the recruitment of paxillin into focal adhesions (Fig. 7B).

Tenascin-C stimulated EDNRA expression (10). By using the clinically relevant EDNRA-specific inhibitor BQ123 (17), we observed that inhibition of EDNRA restored cell spreading, which suggests that signaling through EDNRA mediated cell rounding on FN/TN-C. Similar to T98G glioblastoma cells, activation of EDNRA in MDA-MB435 breast carcinoma and J82 urinary bladder carcinoma cells also supported cell rounding by tenascin-C that could be blocked with BQ123. Whereas endogenously expressed ET1 activated EDNRA, exogenously added ET1 at a high concentration of 20 nmol/L also activated EDNRB. Thus, on activation of both receptors, signaling by EDNRB dominated over signaling by EDNRA in T98G and in MDA-MB435 and J82 cells. Different affinities of EDNRA and EDNRB for the ligand, as has been reported by others (35), can explain our observation that activation of EDNRB as well as inhibition of EDNRA restored cell spreading on FN/TN-C.

Because activation of EDNRA (our data) and MEK down-regulates TM1 expression (30, 31, 36), we examined whether EDNRA-associated cell rounding on FN/TN-C was dependent on MAPK signaling. Indeed, inhibition of MEK with UO126 blocked EDNRA-specific cell rounding on FN/TN-C and caused restoration of FAK phosphorylation and stabilization of RhoA and TM1. Our data link TM1 and RhoA protein instability to EDNRA (Table 1).

Tenascin-C induced EDNRA expression later than 5 h after plating, suggesting that EDNRA signaling is important for tenascin-C–induced cell rounding on fibronectin at later time points. This could present a mechanism to enhance or maintain tenascin-C–induced cell rounding. Whether this involves syndecan-4 is unknown. A potential link of EDNRA to syndecan-4 in causing cell rounding by tenascin-C is supported by our observation that ligation of syndecan-4 is sufficient to restore cell spreading on fibronectin and to attenuate tumor cell proliferation in the presence of tenascin-C (2). Moreover, activation of EDNRA did not induce cell rounding in cells that were spread on fibronectin, where syndecan-4 is engaged as coreceptor in integrin signaling. We speculate that, in situations where syndecan-4 is not working as coreceptor for integrin signaling, syndecan-4 might acquire other functions (e.g., promoting EDNRA signaling), resulting in cell rounding in the presence of tenascin-C.

Tenascin-C and endothelin receptors are instrumental in neural crest cell (NCC) migration during embryonic development because NCC migration was blocked in chicken embryos that were treated with tenascin-C antisense morpholinos (37). The homozygous knockouts of EDNRA and EDNRB are embryonic lethal (38) and EDNRA and EDNRB signaling is crucial in NCC migration (39). Our data provide a functional link between tenascin-C, EDNRA, and NCC migration. Tenascin-C may promote NCC migration along tenascin-C–rich cues by stimulating EDNRA signaling.

Our results suggest that responses toward tenascin-C largely depend on the growth factor receptor status of a cell. EDNRA supports cell rounding by tenascin-C through concomitant repression of RhoA and TM1. Our report is the first to link EDNRA with MEK signaling and cell rounding on a FN/TN-C substratum and enhanced proteolysis of RhoA and TM1. Moreover, similar expression of tenascin-C and EDNRA in eight human cancers is compatible with tenascin-C playing a role in regulation of EDNRA in vivo. In contrast to EDNRA, we found that EDNRB blocks cell rounding by tenascin-C through activation of FAK and normalizing expression of RhoA and TM1. We linked EDNRB, EGFR, PLC, PI3K, and JNK to cell spreading in the presence of tenascin-C. Moreover, EDNRB signaling seems to antagonize signaling through EDNRA. A potential interdependence of the two signaling pathways has been reported; inhibition of EDNRA with BQ123 significantly enhanced EDNRB-induced proliferation of endothelial cells (13). An increased EDNRB and αTM expression found in five human cancers supports the possibility that EDNRB signaling could be involved in αTM regulation in a steady-state situation. More detailed information is required to unravel potential interrelations between tenascin-C, EDNRA, EDNRB, and tropomyosin in invasive human cancer tissue.

In summary, we presented a detailed mechanism by which tenascin-C causes cell rounding through EDNRA and how cells modulate the antiadhesive properties of tenascin-C through EDNRB and presumably other signaling. This is the first study to reveal that cell responses toward tenascin-C are modulated by growth factor signaling. More information is required about signaling pathways that support or counteract tenascin-C actions in vivo to use this knowledge for prediction of tumor malignancy.

Grant support: Swiss National Science Foundation grant 3100A0-102145/1 (G. Orend), Novartis Stiftung für Medizinisch-Biologische Forschung (G. Orend), Association for International Cancer Research (G. Orend), Swiss Cancer League grant OCS-01419-08-2003 (G. Orend), Swiss National Science Foundation grant 3100A0-108266/1 (M.E. Hegi), Münster University Hospital grant IMFGÖ120415 (M. Götte), and National Medical Research Council, Singapore (G.W. Yip). Work from A. Dittmann was part of her bachelor thesis in biotechnology at the University of Mannheim, Germany.

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 Harold Erickson and Pia Wülfing for providing human tenascin-C cDNA and breast cancer material, respectively; Marie-France Hamou and E.S. Yong for technical assistance; Eytan Domany for sharing the SPIN software; and Francois Lehembre and Matthias Chiquet for critically reading the manuscript.