High cyclooxygenase activity is a common feature of human epithelial malignancies,1, 2 however, the biological significance of this metabolic activity has never been established. Recent epidemiologic studies have indicated that prolonged use of nonsteroidal antiinflammatory drugs (NSAIDs) is associated with a decreased risk of several malignancies, most notably colorectal cancers.3 More recently, similar relationships have been observed in lung, breast and other cancers as well.4–6 Because NSAIDs have, as a principle action, inhibition of cyclooxygenases, these findings suggest that prostaglandin synthesis contributes to the risk of developing primary malignancies.
We sought evidence for a role of prostaglandin synthesis in late tumor progression, i.e., tumor metastasis. Using a murine model of human breast cancer, we showed that higher prostaglandin E2 (PGE2) levels were observed in malignant mammary tissue in comparison to normal or premalignant gland. Furthermore, the highest PGE2 levels were observed in the most malignant and metastatic tumors.7 Two isoforms of cyclooxygenase (cox) are responsible for prostaglandin synthesis. In many human tumors the cox-2 isoform is overexpressed. Overexpression of cox-1 is observed less frequently. In human breast cancer, expression of both isoforms has been observed.8, 9 We have now examined the contribution of both cox isoforms in a murine model system and show a positive correlation of both cox-2 expression and metabolic activity to tumorigenic and metastatic potential. Furthermore, we provide evidence that the tissue milieu differentially regulates cox expression in metastatic vs. nonmetastatic tumors.
MATERIAL AND METHODS
The C4 HAN line was isolated and described by Medina.10 It is a benign tissue that will not grow outside the mammary fat-pad, but it is preneoplastic and will form tumors after prolonged growth in vivo. Murine mammary tumor lines, all derived from a single spontaneously arising tumor of a Balb/cfC3H mouse, have been described previously.11 Lines 410, 67 and 68H are nonmetastatic. Line 168 does not metastasize spontaneously from a subcutaneous site, but will form lung colonies after intravenous injection. Lines 66.1, 410.4, 4501 and 4526 are highly metastatic from subcutaneous transplants. All tumor cell lines, with the exception of line 168, are maintained in Dulbecco's MEM medium supplemented with 10% fetal calf serum (Gemini Bio-Products, Inc., Calabasas, CA), 2 mM glutamine, penicillin (100 U/ml), streptomycin (100 ug/ml) and 0.1 mM nonessential amino acids. Line 168 is maintained in Waymouth's medium supplemented with 5% calf serum, 5% fetal calf serum, glutamine and antibiotics. For determination of cox levels in tumors, 1–3 × 106 viable cells of each tumor line was injected subcutaneously into syngeneic Balb/cByJ female mice (Jackson Laboratories, Bar Harbor, ME). When tumors achieved an average diameter of 8 mm, mice were sacrificed by cervical dislocation, tumors removed and portions were prepared for immunohistochemistry, protein analysis, or RNA studies.
Isolation and measurement of tissue prostaglandins was described previously and reproduced in Table I.7 Determination of PGE2 levels in cell- conditioned medium is by ELISA kit following the manufacturer's instructions (Cayman Chemicals, Ann Arbor, MI). In some cultures, the cox-1 and cox-2 inhibitor indomethacin (Sigma Chem. Co., St. Louis, MO), dissolved in absolute ethanol or the selective cox- 2 inhibitor NS398 (Cayman Chem.), dissolved in DMSO was added to achieve final concentrations of 1–300 μM. Control dishes contained vehicle at the highest concentration used with cox inhibitors. Forty-eight hours after the addition of these agents, conditioned medium was harvested for PGE2 determinations by ELISA. Growth assays were carried out in a similar fashion where viable cell counts were determined at Days 1,2 and 3 after addition of indomethacin or NS398 to cultured cells. Data expressed as percent of cells in drug-treated cultures vs. number of cells in control (solvent-containing) dishes.
Table I. Prostaglandin Levels in Mouse Mammary Neoplasms Correlate with Tumor Behavior
Cell lysates from cultured cells treated as for prostaglandin assays were prepared in M-Per (mammalian protein extraction reagent, Pierce, Rockford, IL), tumor tissue was homogenized in T-Per (tissue protein extraction reagent, Pierce), both containing 1 mM PMSF. Lysate was centrifuged and 20–40 ug protein of supernatant was denatured in Laemmli sample buffer and resolved on 10% tris-HCl ready gels (BioRad, Hercules, CA), electrophoretically transferred to a nitrocellulose membrane (Amersham Pharmacia Biotech, Piscataway, NJ) and immunoblotted with cox-1 antibody (mouse anti-ovine, monoclonal, Cayman Chemicals) or cox-2 antibody (rabbit anti-mouse polyclonal, Cayman) followed by HRP-conjugated second antibodies (Transduction Laboratories, Lexington, KY). Specific bands were visualized by Super Signal West Pico chemiluminescent substrate (Pierce). Relative protein expression was then quantitated by densitometry.
RNA was extracted from cultured cells using TRIZAL reagent (Gibco BRL, Gaithersburg, MD), 20 ug RNA resolved in 1% formaldehyde/MOPS denatured gels and transferred to Zeta-Probe membranes (BioRad). Blots were hybridized at 42°C in 5× SSC, 50% formamide, 0.5% SDS, 5 × Denhardt's solution containing 100 ug/ml ssDNA and 32P-labeled probe specific for murine cox-2 (Cayman) or elongation factor-1α.
Tumors were removed, placed in Baker's formol calcium fixative, routine processing was carried out and immunohistochemical staining carried out using cox-2 antibody and biotin-labeled goat anti-rabbit IgG followed by strep avidin-HRP and DAB.
We have examined arachidonate metabolism leading to the production of PGE2 in a murine model of breast cancer that reflects the heterogenous nature of human breast malignancies. We used tumor cell lines that originate from one spontaneous mammary carcinoma, however, they differ in many properties including tumorigenicity, immunogenicity and metastatic potential. We had reported previously that high PGE2 levels are detected in tumors arising from transplantation of these tumor cell lines to syngeneic mice (Table I, reproduced in part from Fulton and Heppner7). Lesions are listed in the table in order of increasing tumorigenic and metastatic potential. Metastatic potential was positively correlated with higher levels of PGE2 extractable from tumors. We have now examined PGE2 levels in these same tumor cells grown in vitro. These data show a weaker but still positive trend for increased PGE2 accumulation in vitro and more aggressive lesions in vivo (Table I).
Two isoforms of the cyclooxygenase enzyme, responsible for prostaglandin synthesis, have been identified. Therefore, we asked which isoform was expressed by murine mammary tumors and whether quantitative or qualitative differences in cox expression correlate with metastatic capacity. We prepared lysates of the same tumor types analyzed previously for PGE2. Tumor lysates were analyzed by western blotting employing antibodies specific for either the cox-1 or cox-2 isoform. Figure 1 shows significant quantities of the 70 kD cox-1 protein present in all tumor samples. Although there are some differences in the quantity of cox-1 protein detected in different tumors, there is no correlation with phenotype. In contrast, expression of the 72 kD cox-2 protein is very high in the metastatic tumors (4526, 410.4, 66.1 and 168), but no cox-2 protein is detectable in nonmetastatic tumors 67 and 410. After prolonged exposure, a faint band is visible in these latter lanes. We also examined cox-2 expression in lung metastases from mice bearing s.c. implants of tumor 410.4. Figure 2 shows that, like the subcutaneous implant, metastatic lesions of 410.4 are also cox-2 positive. Thus, cox-2 protein, like PGE2 levels, is positively correlated with metastatic potential.
Tumors are a mixture of host and malignant cells and analysis of tumor lysates does not identify the origin of cox proteins. Others have reported that in animal tumor models, cox-2 is observed in tumor-associated vasculature, but not cancer cells per se.12 Likewise, in polyps arising in APC mutant mice, cox-2 is detected in the interstitial cells rather than the intestinal epithelium.13 To determine if mammary epithelial tumor cells express cox proteins, we examined expression of cox proteins in lysates of mammary tumor cells grown in vitro in complete, serum-containing medium. Figure 3 shows that, unlike tumor tissue lysates, all cultured cells expressed both the cox-1 and cox-2 isoforms. Thus, tumor cell lines 67 and 410 readily expressed cox-2 protein in vitro, but this species was absent from tumors derived from these cell lines. These data indicate that cox-2 expression is regulated in the tumor environment and that the inducers or suppressors differ in metastatic vs nonmetastatic tumors. Immunohistochemical examination of tumors confirms the western data. Thus, cox-2 protein is expressed in the epithelial tumor cells of metastatic 66.1 tumors, but not in nonmetastatic line 67 tumors (Fig. 4). Some weaker staining of host elements (fibroblasts, macrophages) at the periphery of lesions is also observed (data not shown). Thus, these mammary tumors resemble human malignancies in which cox-2 is prominently expressed in the malignant cells and differ from other rodent lesions that have been examined in which the malignant cells are not cox-2 positive.12, 13
We had shown previously that indomethacin inhibits more than 95% of the PGE2 produced by these cells in culture.14 Because indomethacin inhibits both isoforms, it was not possible from those studies to determine the relative contribution of either isoform to PGE2 metabolism. Therefore, we compared the PGE2 inhibition obtained with indomethacin to that observed with a selective cox-2 inhibitor, NS398. Figure 5 shows that, like indomethacin, NS398 (1 μM) also inhibited >95% of PGE2 accumulation in line 66.1 cells, suggesting that cox-2, rather than cox-1 is the major source of prostaglandin synthesis in these cells in vitro. Interestingly, examination of concentrations of both drugs ranging from 1.0–200 μM revealed an unexpected dose-response. Although all concentrations of both drugs markedly inhibited PGE2 synthesis, there was slightly less inhibition observed at the higher concentrations of both inhibitors. Thus, for indomethacin, the percent inhibition ranged from 97–74% as the drug concentration increased from 1–300 uM. Likewise, NS398 inhibited 95% of PGE2 at the lowest concentration examined, but achieved 86% inhibition at the highest concentration tested (200 uM).
These unexpected findings prompted us to determine if either drug affected the levels of cox protein or mRNA in these cells. Cox protein levels were determined in lysates of cells treated with a range of concentrations of indomethacin or NS398. Western blots show that, even under conditions where PGE synthesis was maximally inhibited, the enzyme protein levels were increased in NS398 treated cells in comparison to vehicle-treated cells and, again, the higher concentrations of NS398 resulted in greater increases in cox-2 protein expression (Fig. 6). Densitometric analysis indicates that NS398 increased cox-2 by 1.3–2.1-fold over levels observed in vehicle-treated cells. Like NS398, indomethacin treatment resulted in increased cox-2 protein (1.3–1.7-fold). Levels of cox-1 protein were not changed by drug treatment of these cells (data not shown).
Cox-2 protein is transcriptionally regulated by most inducers. To determine if drug treatments also altered cox-2 mRNA levels, we isolated RNA from NS398 or indomethacin-treated line 66.1 cells and performed northern analyses with cox-2-specific probe. Figure 7 shows that, consistent with the protein data, cox-2 mRNA is also increased by indomethacin or NS398 in a dose-dependent manner.
We determined the effect of cox inhibitors on growth of metastatic line 66.1 cells. Both indomethacin and NS398 lead to comparable dose-dependent inhibition of cell growth (Fig. 8). It is interesting to note, however, that marked growth inhibition is observed only at drug concentrations exceeding those required to ablate PGE2 synthesis. These data suggest that growth inhibitory effects of cox inhibitors are not solely the result of cyclooxygenase enzyme inhibition.
The overexpression of cox proteins in human epithelial tumors is now well documented.1, 2 It has also been known for many years that tumors often contain high levels of one or more prostaglandin, most notably PGE2. The physiologic role of this metabolic activity was uncertain, however, recent epidemiologic studies have shown that use of nonsteroidal antiinflammatory drugs reduces the incidence of cancers of the gastrointestinal system.3 Prospective studies in populations at high risk for the development of colorectal adenocarcinoma also support a protective role for NSAIDs. Several recent studies also indicate a potential protective role for NSAIDs in cancers of the breast as well as other tissue sites.4–6 Furthermore, examination of several NSAIDs in chemoprevention models of breast cancer in rodents indicate that NSAIDs can prevent the development of breast cancer if given at the time of carcinogen administration.15, 16
NSAIDs have, as a principle action, inhibition of prostaglandin synthesis, suggesting that prostaglandin metabolism contributes to early cancer development. More recently, inhibition of carcinogenesis has been achieved with NSAID derivatives that no longer inhibit cyclooxygenase, suggesting that protection may not be solely attributable to inhibited prostaglandin synthesis.
Less attention has been focused on the role of cox metabolism in later steps in tumor progression. Two early studies reported high levels of PGE-like material (by bioassay) present in breast tumors that had metastasized to bone, and found that post-surgical survival time is shorter in women with high PGE tumors.17, 18 Data showing that breast tumors with high PGE2 activity tend to have lymphatic, nodal and vascular invasion also indicated that high PGE2 is a marker of high metastatic potential.19–21 More recent studies have attempted to relate changes in prostaglandin synthesis activity and metastatic potential to differences in cox protein expression. A comparison of 2 human breast cancer cell lines showed higher PGE2 and cox-2, but not cox-1 in the more metastatic cell line.22 Using a unique model system that allows us to compare many closely related lesions that vary widely in phenotype, we have now shown that cox-2 protein expression in vivo correlates absolutely with metastatic potential.
Many tumors overexpress only the cox-2 isoform, however, the data in breast cancer is more controversial. Thus, 1 study revealed a predominance of cox-2 mRNA;8 the other study rarely detected cox-2 protein, but commonly observed cox-1 protein.9 Fewer studies have examined cox levels in relation to metastatic potential. A small study of human gastric carcinoma reported that cox-2 overexpression was positively correlated with tumor invasion into lymphatic vessels and metastasis to lymph nodes.23 A comparison of cox-2 expression in human lung adenocarcinomas revealed higher cox-2 in metastatic cells in the lymph node than in the primary tumor.24 Although these studies suggest a linkage between cox-2 expression and increased metastatic potential, other studies report no linkage for these 2 properties in colorectal carcinomas.25 Thus, further studies are needed to examine the relationship between cox expression, cox activity and tumor progression.
Studies in vitro using cultured murine mammary tumor cells indicate that both cox-1 and cox-2 proteins can be expressed by mammary epithelial tumor cells regardless of phenotype. Thus, it is quite striking that cox-2 protein is not detected in nonmetastatic tumors derived from lines 67 or 410. These data suggest several possibilities: (i) the tissue milieu in metastatic tumors provides the correct inducers of cox-2 activity that are lacking in nonmetastatic tumors, or (ii) the tissue environment of nonmetastatic tumors actively downregulates cox-2 expression. We are exploring these possibilities. Despite the absence of detectable cox-2 protein in tumor 410, PGE2 is present in these tumors (Table I). Thus, the cox-1 isoform, expressed by either tumor or host cells, is likely to contribute to PGE2 synthesis in vivo and may still be an important target for therapy. Indeed, earlier studies by us show that indomethacin inhibits growth of tumor 410.26
Despite the fact that both isoforms are detected in vitro, nearly complete inhibition of metabolic activity (PGE2 accumulation) was achieved with a selective cox-2 inhibitor, indicating that most PGE2 synthesis in vitro is due to cox-2 activity. Although low concentrations of either drug are able to inhibit most of the PGE2, it was surprising to find that higher drug concentrations were somewhat less inhibitory. Although these differences were not dramatic, we examined this unusual dose-response effect further. Examination of cox protein levels revealed an apparent paradox: that treatment with either NS398 or indomethacin inhibited the activity of the cox-2 enzyme, but increased the levels of the target enzyme. Thus, the diminished PGE inhibition, detectable at higher concentrations of cox inhibitors, was associated with increased cox expression at both the RNA and protein level.
A number of NSAIDs have now been shown to activate transcription of their target enzyme. This occurs via activation of peroxisome proliferator-activated receptors (PPARs).27–29 PPARs are members of the steroid hormone receptor superfamily that alter transcription of genes containing peroxisome proliferator response elements. Meade et al.28 have shown that NSAIDs activate cox-2 but not cox-1 through a peroxisome proliferator response element within the cox-2 promoter.
These studies describe a murine model of metastatic breast cancer that suggests a role for heightened cox-2 protein activity in the metastatic dissemination of tumor cells. Although previous studies using more limited comparisons have suggested that such a relationship exists, these studies examine the question in a model system that allows comparison of a number of tumor cell lines having a common origin. Previous studies focused on protein expression, but did not examine enzymatic activity. The present studies show that both parameters are associated with aggressive behavior in vivo. We have also shown that, like human breast cancer, cox-2 protein expression is observed primarily in the malignant epithelial cells of murine mammary lesions. Thus, the murine model described here may be an excellent model of human breast cancer. We have provided evidence suggesting, for the first time, that cox-2 expression and activity may be regulated differently in the milieu of tumors with variable phenotypes. In preliminary studies, we have shown that higher PGE2 levels are associated with poor long term survival in women with breast cancer.30 We showed previously that indomethacin is effective at blocking tumor metastasis in the murine model system.31 The current studies indicate that selective cox-2 inhibitors should be evaluated in this preclinical model to determine if they may have future applications to prevent tumor progression in the clinical setting.
We are grateful to Dr. S. Rao, University of Maryland, for help with photomicroscopy.