Tenascin-W Is a Novel Marker for Activated Tumor Stroma in Low-grade Human Breast Cancer and Influences Cell Behavior Free

Epithelial tissues depend on normal stromal cells and the basement membrane for maintenance of tissue homeostasis, cell adhesion to extracellular matrix (ECM) through integrins, and generation and preservation of epithelial cell polarity (1, 2). These interactions continue to be important in epithelial pathologies. During carcinogenesis, the stromal cells get activated by an initial trigger that usually comes from the cancer cell, leading to stromal changes and the formation of a tumor-permissive microenvironment (1). More and more evidence is accumulating that this aberrant tumor stroma influences cancer development and has an effect on the malignancy of a tumor (for reviews, see refs. 3–6). Tumor-associated cells of the stromal compartment express proproliferative paracrine signals for epithelial cells (7), stimulate angiogenesis (8, 9), and can even show loss of tumor-suppressor genes (10). Therefore, understanding the mechanisms of the complex crosstalk between the cancerous epithelial cells and the tumor stroma might lead to novel approaches for cancer therapies that target the functions of the activated stromal cells (for reviews, see refs. 3, 11–13).

Because of the newly recognized importance of the tumor stroma in cancer development, it is necessary to fully characterize this tissue compartment. A prominent ECM protein specifically present in tumor stroma is tenascin-C (ref. 14; for reviews, see refs. 15–17). Interestingly, tenascin-C was shown to be expressed around angiogenic vessels in many tumors (18–20) as well as to promote angiogenesis in cell culture studies (21). Furthermore, tenascin-C addition to a fibronectin substratum stimulated cancer cell growth in in vitro studies (22, 23). Therefore, tenascin-C is one of the potential candidate molecules mediating the protumorigenic effects of tumor stroma (for review, see ref. 15).

Recently, we found in the stroma of mouse mammary tumors the induction of a second member of the tenascin family of ECM proteins, tenascin-W (24). Because the human orthologue has never been analyzed, we cloned the human tenascin-W cDNA, expressed the protein, and raised antibodies against it. We determined the presence of tenascin-W in a large number of breast cancers where it was more prevalent in low-grade tumors. In vitro, tenascin-W did not interfere with cancer cell adhesion to fibronectin, but promoted migration of breast cancer cells toward fibronectin. Furthermore, fibroblasts were able to adhere to a tenascin-W substratum. Our data suggest tenascin-W as a marker for transformation of the normal physiologic stroma to an activated stroma in breast cancer, and that tenascin-W can influence cancer cell behavior.

Cloning, expression, and purification of full-length human tenascin-W and tenascin-C. cDNA was synthesized from total RNA extracted from the human osteosarcoma cell line KRIB with TRIzol (Invitrogen). PCR was done using the KRIB cDNA as template and primers for nested PCR were designed that amplified three separate but overlapping parts of human tenascin-W that we call part A, part B, and part C. The first reaction was done with the primer sets 5′-CAGGCATCCTGGAGGGTCTGCTCCC-3′/5′-GTGAGGGCATTGGTGTCAGCTTTC-3′ (for part A), 5′-GCGCTACACTTCTGCTGATG-3′/5′-CTGTGGAGAGGGTGGTGG-3′ (for part B), and 5′-CGCAGTCTGGTGGCATATTG-3′/5′-CATGATTTGTTCTGCGGGC-3′ (for part C). Primer sets for the second reactions included XhoI and BamHI restriction sites, respectively. Primer pair 5′-GACTCGAGCTTTCCAAGGATGAGTCTCCAGG-3′/5′-GAGGATCCCCTGGTTGCCCCTTTCAGCCC-3′ was used for part A, primer pair 5′-GACTCGAGTGCACAAGGATGAGAGCAG-3′/5′-GAGGATCCACCCTTAAAGGCAACAAGGG-3′ was used for part B, and primer pair 5′-GACTCGAGCGGCTACATTCTGACTTACC-3′/5′-GAGGATCCTCAGTGATGGTGATGGTGATGGAACGTTCGCAGCCTTCCTC-3′ was used for part C. The reverse primer for the second PCR of part C contained the sequence encoding a 6×His-tag followed by a stop codon. The first two fragments (A and B) were ligated using an internal AccI restriction site present in the overlapping region, and the third fragment was added using an internal NarI restriction site located in the overlap of part B and C. This resulted in a full-length human tenascin-W cDNA of 3,885 bp (corresponds to the sequence 16–3,900 of Genbank accession no. NM_022093). This full-length cDNA was cloned into the expression vector pCEP/Pu (provided by J. Engel, Biozentrum, Basel, Switzerland) and transfected into EBNA293 cells [American Type Culture Collection (ATCC)] using FuGENE6 (Roche). Transfected cells were selected with puromycin, and serum-free medium containing the secreted recombinant human tenascin-W was collected. After ammonium sulfate precipitation (0–50%) and dialysis against PBS, the protein was passed over a gelatin-agarose column to remove fibronectin followed by affinity purification on a Ni-NTA matrix (Qiagen) in the presence of 0.5 mol/L urea. After extensive washing of the column, tenascin-W was eluted with 250 mmol/L imidazole (pH 6.9) and finally dialyzed against PBS. The protein appeared as a single band on a Coomassie-stained acrylamide gel, and fibronectin could not be detected in the purified sample by immunoblotting or by ELISA.

A full-length human tenascin-C cDNA called HxBL was kindly obtained from H.P. Erickson (Department of Cell Biology, Duke University, Durham, NC). It was modified by the addition of a 6×His-tag at the COOH terminus and subcloned into pCEP/Pu for transfection into EBNA293 cells as described above for human tenascin-W (25). The same purification procedure as described above for tenascin-W was used for the isolation of human tenascin-C.

Anti–tenascin-W antibody production. To raise polyclonal antisera in rabbits, a recombinant fragment of human tenascin-W was cloned, bacterially expressed, and purified. To clone the recombinant fragment, specific primers were designed to amplify the sequence encoding the last two fibronectin type III domains (Fig. 1A) with the Expand High Fidelity PCR System (Roche). The cDNA of the full-length human tenascin-W (described above) was used as template and the PCR was done with the primer set 5′-GAGGATCCGAAATTGACGGCCCCAAAAACC-3′/5′-ATAAGCTTATGTGGAGAGGGTGGTGGA-3′. The forward primer included a BamHI restriction site and the reverse primer a stop codon immediately followed by a HindIII restriction site to enable the directional cloning into the bacterial expression vector pQE30 (Qiagen), supplying a COOH-terminal His tag for the purification of the recombinant fragment. The recombinant fragment corresponding to fibronectin type III domains 3F/4 (Fig. 1A) was expressed and purified by affinity chromatography to a Ni-NTA matrix (Qiagen) following the supplier's instructions. Purification was done under native conditions and elution by 250 mmol/L imidazole (pH 6.9). The bacterially expressed fragment of tenascin-W was then used to raise polyclonal antisera in rabbits using standard immunization procedures.

To raise monoclonal antibodies (mAb) in mice, a recombinant fragment of human tenascin-W was cloned containing the last three fibronectin type III domains (Fig. 1A), bacterially expressed, and purified as described above. To clone the recombinant fragment, cDNA was synthesized from total RNA that was extracted from Saos-2 cells (ATCC) by TRIzol reagent (Invitrogen). Primers, amplifying the last three FN type III domains, were used for nested PCR reactions with the Expand High Fidelity System (Roche) using the Saos-2 cDNA as template. The first reaction was done with the primer set 5′-GGGAAGGAGCAGAGTAGCACTG-3′/5′-CCGCCTCTGGAAGACAATCC-3′, the second reaction with the primers 5′-AGGGATCCGACATTGACAGCCCCCAAAACC-3′/5′-CTAAGCTTTCATGTGGAGAGGGTGGTGGATAC-3′. The forward primer for the second PCR included a BamHI restriction site and the reverse primer a stop codon immediately followed by a HindIII restriction site to enable the directional cloning into the bacterial expression vector pQE30 (Qiagen), supplying a COOH-terminal His tag for the purification of the recombinant fragment. mAbs were obtained by immunizing female BALB/c mice with 34.6 μg of the purified recombinant tenascin-W fragment emulsified with STIMUNE (ID-Lelystad, Institute for Animal Sciences and Health). For boosting, mice were injected twice with a 4-week interval with 25 μg tenascin-W fragment. Splenic lymphocytes were fused with the myeloma cell line P3X63Ag8.653 (ATTC) and cultured according to standard protocols. The hybridoma supernatants were analyzed by ELISA and Western blot analysis of a tenascin-W fragment expressed in HEK293 cells using a construct containing the sequence of the last three FN III domains of tenascin-W fused to an NH₂-terminal fragment of chicken tenascin-C containing its secretion signal and the epitope for the well-characterized monoclonal anti-chicken tenascin-C antibody mAb60. IgGs from two mAb hybridoma clones were purified from conditioned medium by protein G Sepharose (Amersham) and clone 56O was used in this study.

Human tissue extracts and Western blot analysis. The following study was done in accordance with the guidelines of the ethical committee of the University of Basel. Fresh human tissue was frozen on dry ice immediately after surgery. For the processing of the tissue, it was thawed on ice, minced, and homogenized in lysis buffer [100 mmol/L phosphate buffer (pH 8.0), 300 mmol/L NaCl, 8 mol/L urea, 1% Triton X-100, 10 mmol/L β-mercaptoethanol, 50 mmol/L guanidinium hydrochloride, and complete protease inhibitor cocktail (Roche)]. Insoluble material was pelleted, and reducing SDS-PAGE sample buffer was added to the supernatant and boiled for 5 min at 95°C. After electrophoresis on 6% polyacrylamide gels, proteins were electrotransferred onto polyvinyldifluoride membranes (Millipore) using a semidry blotting apparatus (Millipore). After the transfer, membranes were stained with Amido Black to control equal protein loading. After blocking for 1 h at room temperature in TBS containing 0.05% Tween and 5% skim milk powder, membranes were incubated overnight with either the polyclonal tenascin-W antiserum (1:750), the mAb 56O raised against tenascin-W (1:1,000), the mAb B28-13 raised against tenascin-C (1:100; ref. 26), or the monoclonal antivinculin antibody (1:2,000; Sigma) followed by an incubation for 1 h with anti-rabbit IgG or anti-mouse IgG coupled to horseradish peroxidase (1:10,000), respectively. Blots were developed using SuperSignal (Pierce) and exposed to Kodak BioMax MR Films.

For Western blot quantification, the software Gene Tools from SynGene was used. Briefly, the quantity of 25 ng was assigned to the specific band obtained from 25 ng of the purified protein loaded on the same gel. By dividing the densitometric values of the bands of the tissue extracts by the value obtained with 25 ng of the purified protein loaded and developed on the same gel, a quantity for each band could be calculated and further normalized to vinculin.

Frozen tissue microarrays and immunohistochemistry. A frozen tissue microarray (TMA) was constructed from frozen tissue samples of 40 breast carcinomas and 2 fibroadenomas. Pathologic features of these samples are summarized in Table 1 (patients 1–42). Histologic grading of the breast carcinomas was done according to the Bloom, Richardson, Elston grading system. A second TMA was built from 10 frozen tissue samples of normal breast tissue. Both TMAs were constructed in frozen Tissue-Tek optimum cutting temperature compound (Miles Laboratories) as described previously (27). We optimized a commercial microarray device (Beecher Instruments) by using a 0.6-mm drill for recipient whole making instead of the conventional hollow needle.

All immunostainings were done using the Discovery XT automated stainer (Ventana Medical Systems), with 3,3′-diaminobenzidine Map detection kit (Ventana). Frozen TMA slides were dried 1 h at room temperature, fixed for 10 min at 4°C in acetone, and then introduced into the automate. No pretreatment was required for any staining. Slides were first blocked twice for 12 min with the AB Block reagent (Ventana). Afterward, they were incubated for 1 h at 37°C with mouse monoclonal anti–tenascin-C (B28-13; 1:2,500), rabbit polyclonal anti–tenascin-W (1:40), and mouse monoclonal anti–tenascin-W (clone 56O; 1:800). Slides were then treated for 32 min at 37°C with a biotinylated universal secondary antibody (Ventana) and counterstained with hematoxylin and bluing reagent (Ventana).

Cell culture. EBNA293 (ATCC), T47D (ATCC), MCF-7 (ATCC), MDA-MB-435 (ATCC), and Detroit 551 cells (ATCC) were grown in DMEM supplemented with 10% FCS, 100 units/mL penicillin, and 100 μg/mL streptomycin and cultured under standard conditions. CHOB2 (28) and sublines CHOB2α27 (ref. 29; subclone 2C8), CHOB2α4β1 (30), and CHOB2v7 (31), expressing α₅α₁, α₄β₁, and α_vβ₁, respectively, were grown in MEMα supplemented with 10% FCS, 100 units/mL penicillin, and 100 μg/mL streptomycin and cultured under standard conditions.

Cell adhesion assays and cell morphology. Cell adhesion assays were done with the human mammary carcinoma cell lines T47D, MCF-7, MDA-MB-435; the integrin-deficient CHOB2 cells; CHOB2 cells expressing single integrins; and with human dermal fibroblasts Detroit 551. Sixty-well microtiter plates (Nunc) were coated for 1 h at room temperature with 5 μL per well PBS containing 0.01% Tween and 10 μg/mL of the indicated ECM protein and blocked for 30 min with PBS containing 1% bovine serum albumin (BSA). Cells for the adhesion assays were detached using trypsin-EDTA, resuspended in serum-free medium, and counted. Cells (1,500 per well) were plated on the different substrata for the indicated time at 37°C. Adherent cells were then fixed by the addition of 4% formaldehyde and stained with 0.1% crystal violet. Pictures of the entire wells were taken and cells were counted in triplicate wells in at least three independent experiments. Mixed substrata were prepared as follows: After coating the wells with a first ECM protein, they were washed with PBS and coated for another hour with the second ECM protein. The order of coating for the mixed substrata was not important as tested by ELISA.

To analyze cell morphology, 3.5-cm dishes containing four separate wells (Greiner) were coated as described above with the different ECM proteins. After 30 min of blocking in PBS/1% BSA, 10³ Detroit 551 fibroblasts, resuspended in serum-free medium, were plated on the individual wells and incubated for 90 min at 37°C. To block β₁ integrin function, fibroblasts were plated in the presence of a function-blocking anti-integrin β₁ antibody (Chemicon, clone P5D2) at a concentration of 10 μg/mL. Fibroblasts were fixed by 4% paraformaldehyde in PBS for 30 min and permeabilized with PBS/0.1% Triton X-100 for 5 min followed by incubation with a monoclonal antivinculin antibody for 1 h at room temperature. Cells were washed thrice with PBS and incubated with a fluorescein-coupled anti-mouse IgG (Cappel, 1:1,000) and rhodamine-labeled phalloidin (Sigma; 1:500) for 1 h. After incubation with the secondary antibody, cells were washed thrice with PBS, once with H₂O, mounted using Mowiol (Calbiochem), examined, and photographed using an Axiophot microscope (Carl Zeiss MicroImaging) connected to a DFC480 camera (Leica).

Transwell migration assays. The transwell migration assay has been described previously (24). Briefly, the lower side of the membrane (Costar; porosity: 8 μm) was coated with 10 μg/mL ECM proteins for 2 h at 37°C. Serum-starved cells were trypsinized and resuspended in serum-free DMEM and a total of 10⁵ cells were added to the upper chamber of each well coated on the lower side with 10 μg/mL ECM proteins. DMEM containing 10 μg/mL purified tenascin-W or tenascin-C was added to the bottom chamber and cells were allowed to migrate across the filter for 24 h at 37°C. Cells remaining on the upper side of the membrane were removed with a cotton swab and the cells that had migrated to the underside of the filter were fixed with 3.7% formaldehyde in PBS. Fixed cells were then stained with 0.1% crystal violet solution. Migration was quantified by counting cells per eight randomly selected fields of view using ×10 magnifications in triplicates in at least three independent experiments.

Characterization of human tenascin-W. The gene encoding human tenascin-W is located on chromosome 1 in a tail-to-tail configuration next to the tenascin-R gene (32). The derived cDNA and amino acid sequences can be found under accession no. NM_022093. Because no data existed about the human tenascin-W protein, we decided to clone parts of tenascin-W for antibody production as well as the full-length protein for cell biological studies (described in Materials and Methods). The primary sequence of human tenascin-W encodes a protein of 1,294 amino acids. The structural motifs and their arrangement are shown in Fig. 1A. Tenascin-W contains an NH₂-terminal oligomerization domain, including heptad repeats followed by 3.5 epidermal growth factor (EGF)–like repeats, 9 fibronectin type III (FNIII) domains, and a COOH-terminal fibrinogen globe. Human tenascin-W shows high sequence conservation with mouse tenascin-W. There are, however, some noticeable differences. First, the mouse tenascin-W gene contains an RGD sequence located in the second FNIII domain, whereas the human tenascin-W does not. Interestingly, it is the opposite in the case of tenascin-C where the human protein contains an RGD sequence and the mouse orthologue does not. Second, the two tenascin-W orthologues differ in the number of FNIII domains (32). To analyze the structure and function of tenascin-W, we raised antibodies against bacterially expressed tenascin-W fragments consisting of FNIII domains (Fig. 1A). Full-length tenascin-W was purified from transfected EBNA293 cells. SDS-PAGE analysis revealed a molecular weight of 160 kDa per subunit under reducing conditions, which corresponds to the calculated size deduced from the cDNA sequence. Tenascin-W monomers are smaller than the monomeric forms of human tenascin-C and fibronectin (Fig. 1B). Under nonreducing conditions, tenascin-W migrates as a hexamer slightly faster than hexameric tenascin-C (Fig. 1B). To prove specificity of the tenascin-W antibodies, we did immunoblots of purified full-length tenascin-W and tenascin-C (Fig. 1C). The polyclonal as well as the monoclonal anti–tenascin-W antibodies specifically react with tenascin-W and show no cross-reactivity with tenascin-C. Furthermore, the immunoreactivity of both antibodies can be abolished by preincubation with either the bacterially expressed tenascin-W fragment or the full-length human tenascin-W protein (Fig. 1C).

Tenascin-W is present in extracts of human breast tumors but not in the corresponding normal tissue. To test whether tenascin-C shares its prominent expression in neoplasms with tenascin-W, we analyzed extracts from cancer tissues of 63 breast tumor patients (carcinomas as well as benign tumors; see Table 1 for details) compared with normal breast tissue. As shown by the immunoblots in Fig. 2, the majority of breast tumor tissue samples contained detectable levels of tenascin-C (54 of 63; 86%). Also, 81% of breast cancer samples (51 of 63) were positive for tenascin-W. For tenascin-C, we detected different isoforms due to alternative splicing as previously reported (33). In the case of tenascin-W, we only observed a single major band for all extracts. Thus, we do not have evidence for the existence of prominent alternatively spliced tenascin-W isoforms in breast cancer (Fig. 2). However, in some cases, the detected tenascin-W band migrated slightly differently between patients, possibly representing differences in posttranslational modifications and in rare cases a very faint second lower band was detectable. We assume that the lower bands represent degradation products of tenascin-W, but we cannot exclude that alternative splicing might occur at very low levels. For patients 54 to 63, extracts from corresponding normal breast tissue were also available. In these normal tissues, neither tenascin-C nor tenascin-W could be detected, although most of the tumor extracts of the same patients revealed expression of tenascin proteins (Fig. 2). In summary, the majority of human breast tumors express both tenascins, but their relative amount varies largely between patients. In addition, some cancers express either tenascin-W or tenascin-C alone. This observation suggests that expression and/or deposition and degradation of tenascin-C and tenascin-W can be regulated differentially.

To confirm these results by immunohistochemistry and to localize the proteins within the tissues, we made use of frozen TMAs. The breast cancer TMA contained spots from the tumors of patients 1 to 42 (see Table 1; Fig. 2) and a second TMA contained spots from 10 normal mammary tissues. In 36 of the 42 cases, the staining revealed a very strong expression of tenascin-C in the tumor stroma surrounding the transformed epithelial cells but no staining in the normal tissues. In four cases, tenascin-C staining was also present within the epithelial compartment of the tumors (Table 1). These tumors all belong to high-grade G3 breast cancers. In contrast, tenascin-W was exclusively detected in the tumor stroma in 34 of the 42 patients and normal tissue was negative. Examples of the different staining patterns are shown in Fig. 3. Both monoclonal and polyclonal anti–tenascin-W antibodies gave the same staining patterns (Fig. 3A) and in most cases stromal staining for tenascin-C overlapped with tenascin-W staining (Fig. 3B). In patient 24, only very faint stromal staining is seen for either tenascin antibody consistent with the corresponding Western blots that were weakly positive for tenascin-C and negative for tenascin-W. In a rare case of ductal carcinoma (patient 27), tenascin-C staining was observed throughout the tumor, including the transformed epithelial cells, whereas tenascin-W is only expressed in the stromal compartment. Also, benign tumors were rich in tenascin-W and tenascin-C. Patient 33 gives an example of a fibroadenoma with strong staining by all antibodies throughout the tumor mass. The relative amounts of tenascin-W and tenascin-C in these benign tumors varied greatly (cf. Table 1) with mean tenascin-W levels of 74.2 ± 62 (Fig. 4A).

In summary, the immunohistochemical staining of the TMAs confirmed our results from the immunoblotting experiments and showed that tenascin-W is a specific marker for breast tumor stroma of benign as well as malignant lesions.

Tenascin-W is enriched in low-grade breast cancer. In most of the published work on breast cancer, no clear correlation of tenascin-C with malignancy or any other poor diagnostic or prognostic factors was observed (reviewed in ref. 34). We wanted to know whether this could be different for tenascin-W and whether expression of tenascin-W does correlate with any known diagnostic factor. When we compared the amount of tenascin-W in tumor lysates, as determined by immunoblotting (Table 1), to the histologic tumor grade, we realized that tenascin-W is significantly (P < 0.03) enriched in low-grade tumors (G1 and G2; 49.0 ± 47) compared with high-grade tumors (G3; 21.6 ± 30). The mean relative amount of tenascin-W in G1/G2 tumors is >2-fold higher than in G3 tumors (Fig. 4A). In contrast, there is no significant difference between high-grade (24.2 ± 23) and low-grade tumors (25.9 ± 22) with respect to tenascin-C expression (Fig. 4B). This correlation between higher tenascin-W expression and low-grade breast tumors could be confirmed in an independent RNA profiling study of different breast cancer patients done by van't Veer et al. (35), accessible in the ONCOMINE database (36). The heat maps presented in Fig. 4C show enrichment in tenascin-W transcripts in G1/G2 tumors compared with G3 tumors, whereas there is no obvious tendency visible in the same samples for the tenascin-C transcripts. In another study by Farmer et al. (37), transcript profiling was done of basal versus luminal breast carcinomas. Interestingly, tenascin-W is highly elevated in luminal compared with basal carcinomas, indicating that it may correlate with the estrogen receptor (ER) status as well (Fig. 4D).

Tenascin-W promotes fibroblast adhesion and stimulates cancer cell migration toward fibronectin. Because tenascin-W is present in the stroma surrounding tumors, we investigated its effects on stromal cells such as fibroblasts as well as on the neighboring cancer cells. Tenascin-C is an antiadhesive protein with adhesion-modulating effects leading to the expression of growth-promoting proteins (22, 23). To elucidate whether tenascin-W might act in a similar way, we first did cell adhesion assays using different breast cancer cell lines and normal human fibroblasts (Fig. 5A). The breast cancer cells T47D neither spread nor attached to a tenascin-W substratum, whereas they adhered to fibronectin or type I collagen. The same was the case for two other cell lines tested (MCF-7 and MDA-MB-435; not shown). Furthermore, we could confirm the adhesion-modulating effect by tenascin-C on breast cancer cell adhesion to fibronectin. In this respect, tenascin-W differed from tenascin-C because it did not affect tumor cell adhesion when it was offered as a mixed substratum with fibronectin or type I collagen. In contrast to tumor cells, fibroblasts did attach to a tenascin-W substratum and partially spread (Fig. 5A). This property is unique for tenascin-W, because on tenascin-C fibroblasts remained round. When a mixed substratum of the two tenascin proteins was used, the antiadhesive effect of tenascin-C was slightly counteracted by tenascin-W. However, there is a distinct morphology of the spread fibroblasts on tenascin-W as revealed by phalloidin staining when compared with fibroblasts plated on fibronectin, type I collagen, or tenascin-C (Fig. 5B). Fibroblasts plated on fibronectin and type I collagen form actin stress fibers and a lot of focal contacts visualized by staining for vinculin (Fig. 5B,, arrowhead). The same cells plated on tenascin-W appeared much more compact and irregularly shaped. They fail to form long actin cables, but instead produce many long actin-rich protrusions (Fig. 5B,, arrows), which are also rich in vinculin (Fig. 5B,, arrows). In contrast, the few fibroblasts that attached to tenascin-C did not spread and remained with a round morphology (not shown). Adhesion to tenascin-W was integrin mediated, because addition of anti-β₁ integrin antibodies inhibited adhesion to tenascin-W (Fig. 5B). To find out which β₁ integrin(s) is able to mediate adhesion to tenascin-W, we made use of CHOB2 cells expressing single integrin α chains (29–31) and compared their morphology upon adhesion to tenascin-W, tenascin-C, or fibronectin (Fig. 5C). Integrin-deficient CHOB2 cells did not adhere to any substratum. Integrin α₅β₁–expressing cells adhered and spread preferentially on fibronectin, whereas α_vβ₁- and α₄β₁-expressing cells adhered to fibronectin and tenascin-W substrata with similar efficiency. Therefore, α_vβ₁ and α₄β₁ can serve as receptors for cell adhesion to tenascin-W. None of the CHOB2 integrin-expressing cells tested adhered to tenascin-C– or BSA-coated plates (Fig. 5C).

Finally, we investigated the effect of tenascin-W on cell migration of breast cancer cells using transfilter migration chambers with the underside of the filters coated with fibronectin. Addition of soluble tenascin-W to the lower chamber stimulated T47D cell migration across the filters toward the fibronectin substratum (Fig. 5D). Migration toward fibronectin was enhanced 2-fold by the presence of tenascin-W in the culture medium, whereas cells did not migrate toward BSA. A similar migration-stimulatory effect was observed by the addition of soluble tenascin-C to the lower chamber of the transwell system (Fig. 5D). Therefore, the presence of tenascin-C as well as tenascin-W in tumor stroma may stimulate migration of cancer and cancer-associated cells.

Here, we present the first report about human tenascin-W, the final member of the human tenascin family. We cloned the human tenascin-W full-length cDNA, expressed it in mammalian cell culture, and purified the protein. Human tenascin-W is built up from the same domain types as the other three known tenascins, tenascin-C, tenascin-R, and tenascin-X, namely heptad repeats, 3.5 EGF-like repeats, 9 FN III domains, and a COOH-terminal fibrinogen globe (see review in ref. 32). Although there are a lot of alternatively spliced variants in the other tenascins (38–41), which show different expression pattern and functions (42–44), there is no evidence for the existence of splice variants in tenascin-W, because we detected a single protein on immunoblots of tumor extracts.

Tenascin-W was originally identified in zebrafish, where it is coexpressed with tenascin-C in somites and by neural crest cells (45). More recently, murine (46) and chicken (47) tenascin-W have been described. In both species, tenascin-W is expressed in smooth muscle cells and bone. Chicken tenascin-W modulated adhesion and spreading of calvarial cells in vitro (47). In certain murine mammary tumors, tenascin-W was up-regulated and its presence appeared to correlate with the metastatic potential of the tumors (24).

In this report, we investigated the expression of tenascin-W in human breast cancer and found that 81% of the tumors tested expressed tenascin-W and 86% were positive for tenascin-C. However, the amount of tenascin-C and tenascin-W differed between samples, indicating independent mechanisms that regulate their expression. In our previous study on mouse tenascin-W in primary mouse embryo fibroblasts, we found that BMP-2 was a potent inducer of tenascin-W but not tenascin-C expression, and transforming growth factor-β showed the opposite effect and induced tenascin-C expression much more than that of tenascin-W (24). On the other hand, tumor necrosis factor-α strongly induced both proteins. It is possible that the relative amounts of these cytokines could account at least partially for the differential expression of the two tenascins in the tumor stroma of breast cancer.

We did not detect tenascin-W in normal human mammary tissue but could correlate tenascin-W expression levels with tumor grade. There is a statistically significant higher mean expression of tenascin-W in low-grade tumors (G1/G2) compared with high-grade tumors (G3). In contrast, in the present study, tenascin-C could not be correlated with tumor grade in mammary cancer. The differentially expressed tenascin-C isoforms could not be correlated with tumor grade either, although there is one report showing that some specific tenascin-C isoforms are only expressed in invasive breast carcinomas (48). However, in the transcript profiling study by Farmer et al. (37) extracted from ONCOMINE (36), only cases of basal cancers showed very high levels of tenascin-C, whereas tenascin-W was almost exclusively found to be elevated in luminal cancers. This indicates that tenascin-W might be elevated in ER-positive cancers, because ER-positive cancer cells tend to be enriched in luminal cancers. In contrast, high tenascin-C expression in basal cancers suggests a correlation with ER-negative cells enriched in basal cancers known to have a worse prognosis (49). In the literature, contradictory studies have been published on the value of tenascin-C as a prognostic marker in breast cancer, indicating that the correlations may not be very strong and may depend on the sampling (reviewed in ref. 34). In the case of tenascin-W, the data seem to be more consistent. In support of our findings, tenascin-W enrichment in low-grade breast cancers was independently confirmed in a different patient cohort by RNA profiling studies of breast cancer patients (35). These data are available from the ONCOMINE database. A good correlation between tenascin-W but not tenascin-C transcript levels and tumor grade could be found in this data set (cf. Fig. 4). Tenascin-C is, however, a useful prognostic marker for other types of tumors than breast cancer such as gliomas and lung cancer and in these cases seems to play a role in tumorigenesis (for review, see ref. 34). In many tumors, tenascin-C expression correlates with invasion and angiogenesis, whereas tenascin-C–deficient mice exhibit impaired angiogenesis (50, 51).

In the majority of breast cancers analyzed by immunohistochemistry, there was an almost perfect overlap between tenascin-C and tenascin-W expression in the tumor stroma, suggesting that both tenascins may serve similar but not identical functions. In contrast to tenascin-C, tenascin-W is an adhesive substratum for fibroblasts. They attached to tenascin-W and partially spread with an irregular cell shape that differed from fibroblasts on fibronectin or collagen type I where cells formed actin stress fibers and focal adhesions. The adhesion to tenascin-W was dependent on β₁ integrins and CHOB2 cells expressing either α_vβ₁ or α₄β₁ adhered and spread on tenascin-W to a similar extent as to their known ECM ligand fibronectin. Thus, tenascin-W is a novel ligand for these two integrins.

Although both tenascins do not support breast cancer cell adhesion when they are used as a single substratum, they differ in their action when used as mixed substrata with fibronectin. Although tenascin-C inhibited cancer cell spreading on fibronectin, tenascin-W did not interfere with cell binding to fibronectin or type I collagen. Interestingly, both tenascins were able to induce cancer cell migration toward fibronectin. Because in vivo, the stromal ECM surrounding cancer cells contains a mixture of both tenascins together with many other ECM proteins, they may be part of an activated tumor stroma promoting cancer cell migration and invasion. Because benign tumors can also exhibit high levels of stromal tenascin-W and tenascin-C, the presence of these two proteins is not sufficient to induce invasion. In fibroadenomas, the intact basement membrane separating the epithelial cells from the stromal compartment may protect them from the promigratory effect of the tenascins. The situation is different in carcinomas where additional factors lead to the breakdown of basement membranes and expose the epithelial cancer cells to the stromal environment.

For a long time, cancer research was mainly focused on the cancer cells alone. An amazing wealth of information about oncogenes and tumor suppressors was obtained, which led to the improvement of our understanding of the molecular events occurring in cancer cells and the proteins and signaling pathways affected represent promising new therapeutic targets. However, in recent years, it became clear that a cancer cell requires a permissive environment for progression and that carcinogenesis is accompanied by several changes in the stroma, which finally leads to an aberrant microenvironment that facilitates tumor growth and invasion (reviewed in refs. 4, 5). Because tenascin-W is expressed in lower-grade breast cancers and enhances cell migration, it might be an early marker of activated tumor stroma and thus a good antitumor target.

Grant support: Swiss National Science Foundation SNF 3100A0-114103/1 and Oncosuisse OCS-01419-08-2003 (G. Orend).

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 Peter Hans Schraml for supplying us with breast cancer lysates, Susanne Schenk for the generation of mAbs, Sandrine Bichet for help with immunohistochemistry, Erika Fluri for generating the subclone 2C8 of CHOB2α27 cells, and Erkki Ruoslahti (Burnham Institute, La Jolla, CA) and Jean Schwarzbauer (Department of Melecular Biology, Princeton University, Princeton, NJ) for the gifts of CHOB2v7 and CHOa4b1 cells.