The Expression of Decorin in Human Ovarian Tumors

Decorin is a small, leucine-rich proteoglycan component of the normal cell matrix. It consists of an N-glycosylated M_r 40,000 protein core to which a single glycosaminoglycan side chain made up of dermatan sulfate or chondroitin sulfate is attached. The side chain is extended from a common linker of xylose-galactose-galactose-glucuronic acid bound to the protein core at a serine residue. The dermatan sulfate chain is made of repeats of sulfonated dimers of d-galactosamine and d-galactosamine or l-iduronic acid, whereas chondroitin sulfate is composed of repeated dimers of d-galactosamine and d-glucuronic acid. Sulfate linkages in chondroitin sulfate may vary, i.e., O-chondroitin sulfate, 4-chondroitin sulfate, and 6-chondroitin sulfate isomers have all been described (1).

Decorin is an effective inhibitor of tumor cell growth. Ectopic decorin was shown to inhibit the growth of colon carcinoma cells (2), glioma (3), and ovarian cancer cells (4), as well as tumors of other histiogenic origins (5). Growth inhibition was shown to be caused by up-regulation of the cyclin kinase inhibitor p21^Waf1/Cip1 (4, 5, 6) resulting from decorin binding to and activating the epidermal growth factor receptor (7, 8, 9).

Decorin is not expressed in tumor cells; however, decorin is strongly expressed in the stroma of colon (10) and breast tumors (11). Given the tumor-inhibitory activity of decorin, it might appear that the higher levels of decorin in peritumoral stroma represent an effort by the host to contain the growth of the tumor. However, it was also shown in head and neck tumors that stromal decorin expression was reduced at the site of aggressive tumor growth (12), which may mean that not all tumor types express the same form of stromal decorin. Also, decorin in the stroma of colon tumors contained high levels of chondroitin sulfate side chains (10). This finding in patients may be significant because decorin-expressing dermatan sulfate side chains were 20-fold more effective than decorin-expressing chondroitin sulfate side chains in inhibiting the migration of bone tumor cells (13). Also, the stromal decorin of colon tumors showed a higher percentage of O- and 6-sulfate linkages than that found in normal tissue (10).

The peptide portion of decorin can bind to and inhibit the activity of TGF-β² (14). Bound decorin blocks the activity of TGF-β by preventing it from binding with its receptor complex (15). However, side chains, when present, were shown to inhibit this activity to some degree (15). The therapeutic potential gained by the inhibition of TGF-β function by decorin was illustrated by showing that ectopic decorin could block the growth of glioma cells in rats (3) and reduce chemoresistance to syngeneic breast tumors in mice (16). Interestingly, decorin obtained from bone, which contains only chondroitin sulfate side chains (17), appeared to stimulate the activity of TGF-β (18). Thus, the side chains of stromal decorin may not only be important in regulating the migration of tumor cells (13) but may also have important roles in regulating other functions essential for tumor growth as well.

The observation that some tumors produce stromal decorin that is different from decorin found in normal tissue (10) leads us to question whether the tumor actually dictates to the stromal cells what type of decorin side chains to produce, or whether migration of a particular type of cell(s) into the developing tumor (cells preprogrammed to make decorin with high chondroitin sulfate composition and high numbers of O- and 6-sulfate linkages) is a more likely explanation. Fibrosis is a common feature in many tumors, including ovarian cancer. Myofibroblasts are common to many types of fibrosis and have been identified in fibrotic tumors (19). Myofibroblasts infiltrate areas containing high concentrations of GM-CSF or TGF-β (20), two cytokines that are produced by ovarian cancer cells (21, 22, 23). In the lung, myofibroblasts were attracted early in the development of fibrotic lesions by TGF-β that was induced in macrophages by overexpressed GM-CSF (24). Also, both decorin and myofibroblasts were shown to coaccumulate in fibrous crescents in crescentic glomerulonephritis (25). These observations have led us to speculate that myofibroblasts may be the source of the unique decorin found in the stroma of certain types of tumors.

To our knowledge, there has been no previous investigation on the expression of decorin in ovarian cancer. Decorin has been reported to be absent in other epithelial cancers, such as breast and colon cancer, whereas enhanced decorin was reported in the stroma of these tumors (10, 11). The experiments described in this report were performed to determine whether the expression of decorin in the microenvironment of ovarian tumors was different from or similar to its expression in other epithelial tumors such as breast and colon cancer, i.e., whether decorin is expressed by ovarian cancer cells, and what type of decorin was expressed in the stroma of ovarian tumors. Also, we sought to see whether myofibroblasts were involved in the synthesis of the chondroitin sulfate decorin side chains found in tumor stroma.

Ovarian tissues used in these experiments were from a tissue bank maintained by the Ovarian Cancer Research Project, The University of Texas M. D. Anderson Cancer. All specimens were obtained using institutional review board-approved consent procedures.

The 2774 ovarian cancer cell line was obtained originally from Dr. J. Sinkovics. The cell lines MCR-5 (human embryonic fibroblast), SKOV3 (ovarian cancer), and SW480 (colon cancer) were obtained from the American Type Culture Collection. All cell lines were grown in RPMI 1640 (Life Technologies, Inc., Grand Island, NY) supplemented with 10% fetal bovine serum. The proteasome inhibitor MG132 was obtained from Calbiochem (La Jolla, CA) and stored as a 25 mm stock in DMSO at −25°C.

Mouse IgG raised against 6-sulfate chondroitin and mouse monoclonal IgG raised against human decorin isolated from ovarian fibroma were purchased from Seikagaku Corp./Associates of Cape Cod, Inc. (Falmouth MA). Unlike some antibodies raised against decorin peptides, this antibody equally recognizes core proteins with or without side chains. Mouse monoclonal IgG raised against human smooth muscle α-actin conjugated to HRP was obtained from DAKO Corp. (Carpinteria, CA). Mouse IgG prepared against GAPDH was obtained from Chemicon International, Inc. (Temecula, CA), and HRP-conjugated goat antimouse IgGs were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).

Nearly confluent cell cultures grown in 25-cm² culture flasks were rinsed with PBS and lysed with guanidinium thiocyanate-phenol-chloroform, and the phases were separated by centrifugation (26). The upper phase was precipitated with isopropanol, and the pelleted RNA was washed with 75% ethanol, suspended in diethylpyrocarbonate-treated water, and stored at −80°C. Protein was extracted from cells cultured in 25-cm² culture flasks by scraping the cells into prewarmed SDS sample buffer (27), boiling the samples for 5 min, and storing them at −25°C.

RNA was reverse-transcribed overnight into DNA as described previously (28). The cDNAs were amplified in a GeneAmp 9700 thermocycler (PE Applied Biosystems, Foster City, CA) in 50-μl reaction mixtures containing 0.25 unit of Taq polymerase (Roche), as recommended by the supplier. Primers specific for decorin (forward primer, 5′-AGGGCTCCTGTGGCAATT-3′; reverse primer, 5′-TCAGATGACCGCTGTTGG-3′; Ref. 29) were used in these experiments. These primers were designed to cross intron boundaries and thus prevent amplification of genomic DNA. Amplification was carried out for 35 cycles of 0.5 min at 94°C, 0.5 min at 55°C, and 1 min at 72°C, followed by a final incubation for 7 min at 72°C. Reaction products and DNA molecular weight marker VI (Roche Molecular Biochemicals, Indianapolis, IN) were loaded onto a 1.8% agarose gel in Tris-acetate-EDTA buffer (27) containing 0.25 μg/ml ethidium bromide. The gel was photographed under UV light.

Western blots were performed as described previously (4). Briefly, aliquots of protein and prestained molecular weight markers (Life Technologies, Inc., Gathersburg, MD) were separated on a 10% polyacrylamide mini-gel (Bio-Rad Laboratories, Hercules, CA) and transferred onto Hybond enhanced chemiluminescence nitrocellulose (Amersham Life Sciences, Arlington Heights, IL) with a semidry transfer unit (Bio-Rad Laboratories). The blots were blocked with 5% nonfat dry milk and probed with either antidecorin or anti-GAPDH monoclonal IgG in blocking solution. The blots were treated with HRP-conjugated antimouse IgG in blocking solution, soaked in luminol (Kirkegaard and Perry Laboratory, Gaithersburg, MD), and exposed on Hyperfilm enhanced chemiluminescence X-ray film (Amersham).

Cryopreserved tumors were sectioned as described previously (21). The sections were fixed with room temperature acetone and blocked first with 0.3% hydrogen peroxide in methanol and then with 1% of the appropriate normal serum in PBS. The sections were incubated for 2.5 h with a primary antibody and for 1 h with a secondary (biotin-conjugated) antibody at room temperature. The sections were stained with the avidin-biotin-peroxidase complex vector staining kit (Vector Laboratories, Burlingame, CA) following the manufacturer’s directions. Color was developed with NovaRed substrate (Vector Laboratories). The sections were counterstained with Mayer’s modified hematoxylin (Polyscientific, Bayshore, NY) and Scott’s bluing reagent (STAT Lab Medical Products, Inc., Lewisville, TX).

Also, cultured ovarian cancer cells were grown on an 8-chambered Lab-Tek microscope slide (Nalge Nunc International, Naperville, IL). The cells were washed with PBS, fixed with room temperature acetone, and stored at −25°C until use. The slides were stained as described above, except that no counterstain was used.

Paraffin-embedded sections of ovarian tissue were renatured in a microwave oven for 10 min and stained with the appropriate antibody as described above.

When we amplified cDNA made from human fibroblasts, we observed a predominant transcript with the expected size of 881 bp (MRC-5, Fig. 1). A second, minor band of ∼553 bp was also present. As reported previously, the decorin gene may be expressed in the form of two species of RNA ∼0.5 kb apart in size (30). When we examined two ovarian cancer cell lines for decorin expression (SKOV3 and 2774), we found expression of the decorin gene in both cell lines, shown again by the presence of two bands of product (Fig. 1). These same two transcription products were also seen in cDNA made from extracts of fresh ovarian carcinoma (epithelial ovarian cancer, Fig. 1). We also saw decorin expression in the SW480 colon cancer cell line. Colon cancer cells were previously reported to be negative for decorin transcripts when the less sensitive Northern blot assay was used (10). Interestingly, immortalized normal ovarian epithelial cells (kindly provided by Rosemarie Schmandt, The University of Texas M. D. Anderson Cancer) also expressed both decorin transcripts (immortalized ovarian epithelial cells, Fig. 1). It is obvious that the decorin transcripts from the fibroblasts were in much greater abundance because we used up to five times more RNA (1.1–4.35 μg) in the cDNA reactions from the epithelial cells than was used in the cDNA reactions from the fibroblasts (0.85 μg). Also, the relationship between the amount of product and input RNA in RT-PCR reactions is logarithmic, not linear (28, 31). However, it is clear that epithelial cells, including tumor cells, do produce decorin transcripts. Also, in contrast to fibroblasts, the epithelial cells show an equal or greater expression of the smaller decorin transcript, perhaps by using an alternate splicing site that had been described previously (32).

Because we could not find any previous work describing decorin expression in ovarian cancer cells, and we had found transcripts for decorin in both ovarian cancer cell lines and tumors, we thought it possible that this type of tumor may actually express decorin. To examine this possibility, frozen sections of papillary serous ovarian carcinomas were stained for decorin as described in “Materials and Methods.” Eighteen tumor specimens were examined using a peptide-specific antidecorin polyclonal antibody (kindly provided by Telios/Integra), and all of the specimens showed intense staining of the tumor stroma. An example is shown in Fig. 2,A. Similar results were obtained when a decorin-specific monoclonal antibody was used to stain a representative section shown in Fig. 2 B. Both examples clearly show that whereas the peritumoral stroma stained heavily for decorin, the tumor cells did not. This result is in agreement with previous findings in other tumors (10, 33). Thus, although ovarian cancer cells express decorin transcripts, they did not appear to produce significant amounts of protein.

We were also unable to detect decorin in ethanol/potassium acetate precipitates of conditioned medium from cultured ovarian cancer cells (data not shown). We next decided to see whether decorin was produced in tumor cells but destroyed before being released from the cell. Thus, we examined cells treated with a proteasome inhibitor to see whether we would then detect decorin protein in treated cells. This prevents the movement of unbiquitinated protein into protostomes and results in the buildup of ubiquitin molecules on protein, thus forming distinctive ladders when examined by Western blot analysis.

Ovarian cancer cells were cultured on an 8-well chambered microscope slide and treated overnight with increasing concentrations (25–100 mm) of MG132. The cells were then stained for decorin. As shown in Fig. 3, cells treated with 25 μm (and 50 μm, data not shown) MG132 exhibited a pronounced increase in staining for decorin. The pattern of staining was almost exclusively perinuclear. For this reason, the nuclear counterstain ordinarily used was not applied. The MG132-induced increase in staining was much greater in the 2774 cell line. The dose of 100 μm MG132 was toxic to both cell lines but was better tolerated by the 2774 cell line (data not shown).

MG132-treated cells were also analyzed by Western blot analysis. As seen in Fig. 4, extracts from both cell lines exhibited a protein ladder detected with decorin-specific monoclonal antibody. Treating cells with MG132 obviously resulted in considerably increased production of these antidecorin-stained proteins, which is easily seen when the blot of the housekeeping gene GAPDH is used to compare sample loading. Such ladders are commonly produced by ubiquitination. Also, ubiquitin complexes have been shown by immunofluorescent staining to be perinuclear aggregates (34), which tallies with the perinuclear staining for decorin we saw in Fig. 3.

The stroma of colon tumors contains high levels of 6-chondroitin sulfate (10), and peritoneal tumors of colon and ovarian cancer commonly are highly fibrotic. Thus, it occurred to us that cells common to fibrotic tissue may be the source of the type of decorin found in tumor stroma. To investigate this possibility, serial sections of two ovarian tumors were stained with antibody for 6-chondroitin sulfate and for smooth muscle α-actin, a molecule characteristic of myofibroblasts (20), cells commonly associated with fibrosis (24). As demonstrated in Fig. 5, A and B, heavy staining by both antibodies was colocalized in the tumor stroma of both specimens. We next sought to determine the specificity of the role played by the tumor in expression of 6-chondroitin sulfate by stromal myofibroblasts.

Endometriosis is a noncancerous gynecological condition that may be associated with fibrosis. Serial sections were made of tissue obtained from two ovaries taken from patients who had undergone oophorectomy for the treatment of their endometriosis and stained as described in Fig. 5, A and B. As in the tumor specimens shown above, we saw overlapping of staining by smooth muscle α-actin and by 6-sulfate chondroitin (Fig. 5, C and D). These data support our hypothesis that 6-sulfate chondroitin is a natural product of myofibroblasts and that the presence of this form of chondroitin sulfate in tumor stroma could be independent of tumor cell regulation of its manufacture.

It seems likely that tumor cells would acquire protective mechanisms against the effects of decorin because of the many ways in which decorin inhibits tumor growth. These include binding to thrombospondin-1 and blocking the spread and growth of metastatic tumor cells (17, 35, 36), inhibiting cell proliferation via up-regulation of expression of the p21^Waf1/Cip1 molecule in various types of tumor cells (5) including ovarian cancer cells (4), and binding to and inhibiting activated TGF-β molecules (15). It was shown previously that decorin was not expressed in colon cancer (10). Similarly, decorin production was seen to be inhibited in guinea pig bile duct carcinoma (33). As we have shown, tumor cells appear to be inhibited from producing decorin through a posttranscriptional mechanism. Although we detected the presence of decorin transcripts in ovarian tissue and cultured cells (and colon tumor cells), we were unable to detect any exported decorin. We have developed, in this study, preliminary evidence showing that decorin may be cotranslationally ubiquitinated and destroyed in ovarian cancer cells. This same fate was shown to occur for the core protein of mutant forms of aggrecan, another proteoglycan (37).

Decorin is also modified in tumor stromal cells (10). This modification may act to protect the tumor from the growth-inhibitory effects of decorin. The high proportion of chondroitin sulfate side chains found in tumor-associated decorin (10, 33) may allow the tumor to grow more aggressively because chondroitin sulfate side chains are 20-fold less effective in retarding tumor cell migration than dermatan sulfate side chains (13). A question we set out to answer in this study was how the tumor directs the stroma to produce a decorin component of the cell matrix so favorable to tumor cell growth. It was shown in colon cancer that the decorin gene of stromal cells was hypomethylated, thus facilitating the production of decorin (10). However, it was not shown how the tumor could regulate this activity. We questioned whether the tumor acted, instead, to attract tumor site cells predisposed to produce a stromal matrix favorable to tumor growth. Because many tumors have highly fibrotic stroma, we were particularly interested in myofibroblasts, which are a common feature in many forms of fibrosis. We showed that tumor stroma contains very high expression of smooth muscle α-actin, a signature protein for myofibroblasts, and that staining for smooth muscle α-actin largely overlapped with staining for 6-chondroitin sulfate, a form of chondroitin sulfate found in high concentrations in tumor stroma (10). We showed that fibrotic ovarian tissue from patients with endometriosis also stained with smooth muscle α-actin and 6-chondroitin sulfate IgGs. Thus, we concluded that myofibroblasts are the probable source of the 6-chondroitin sulfate associated with tumor stroma and that they may produce it independently of tumor coordination. This seems like a reasonable proposal because many types of tumors, including tumors of the ovary (21, 22), produce significant levels of TGF-β, which could behave as a chemotactic agent for myofibroblasts (24). GM-CSF, also produced by ovarian cancer cells (23), has also been implicated in myofibroblast accumulation (20, 24). However, it also remains possible that epithelial cells, including tumor cells, interact with myofibroblasts after they have accumulated to induce them to produce this form of chondroitin sulfate. This might be affected by the high levels of TGF-β involved in causing them to accumulate (20, 24). Interestingly, another protective role, i.e., to sequester the tumor and thus protect it from immune cells, has also been proposed recently for stromal myofibroblasts (38).

Because we see similar transcript expression in epithelial cells and tumors, it is interesting to speculate that, compared with decorin transcripts found in fibroblasts, the severely reduced level of transcription and the shift in emphasis toward the smaller transcript may reflect a genomic event that occurs during epithelial cell development. The decorin gene is located at 12q23 (32), which, coincidentally, is a common site for LOH in several types of tumors, including ovarian, gastric, and pancreatic cancers (39). LOH at this locus has been observed in 30% of the ovarian tumors examined (40). A LOH could suggest the presence of a tumor suppressor gene, and although two other potential tumor suppressor genes (i.e., TEL and p27^Kip) are both located at that site, neither was shown to be altered in ovarian tumors with LOH at 12q23 (40). On the other hand, a competent decorin gene, when transferred into tumor cells, was shown to reduce or abrogate tumorigenicity by the transformed cells (2, 3). Recent microarray studies have reported that decorin gene expression is significantly reduced in endometrial (41) and ovarian (42) tumor tissues. Our RT-PCR data showed that epithelial cells almost exclusively express the transcript that probably results from the use of an alternative splicing site that fails to make functional decorin. Thus, decorin may represent an example of a potential tumor suppressor gene that becomes inactivated in epithelial cells as a result of normal cell differentiation.

We thank Rebecca Patenia for her histochemical staining contribution to this work.