Malignant Transformation and Antineoplastic Actions of Nonsteroidal Antiinflammatory Drugs (Nsaids) on Cyclooxygenase-Null Embryo Fibroblasts

COX-1^+/− or COX-2^+/− pregnant heterozygous female mice 15,16 were killed by CO₂ asphyxiation on day 18 after fertilization. Each embryo was taken from the uterus, separated, minced, and trypsinized twice with 0.1% trypsin-EDTA-DMEM for 15 min. Tissue debris were discarded, and cells were pooled by brief centrifugation and then suspended in DMEM containing 10% fetal bovine serum. Cells from each of the embryos were grown in the same medium at 37°C with 5% CO₂. PCR analysis was used to determine the genotypes of each embryo. Primers used are the following: for COX-1 genotyping, 5′-AGGAGATGGCTGCTGAGTTGG-3′ and 5′-AATCTGACTTTCTGAGTTGCC-3′ were used to detect the intact COX-1 exon 11; 5′-GCAGCCTCTGTTCCACATACAC-3′ and 5′-AATCTGACTTTCTGAGTTGCC-3′ were used to detect the targeted disruption of COX-1 exon 11 containing the neomycin gene. For COX-2 genotyping, primers 5′-ACACACTCTATCACTGGCAC-3′ and 5′-AGATTGTTGTCAGTATCTGCC-3′ were used to detect the endogenous COX-2 gene (the PCR product extending from exon 8 to exon 10); 5′-ACGCGTCACCTTAATATGCG-3′ and 5′-AGATTGTTGTCAGTATCTGCC-3′ were used to detect the targeted disruption of COX-2 exon 8 containing the neomycin gene. COX-1^−/−/COX-2^−/− deficient cell lines were developed by breeding male COX-1^−/−/COX-2^+/− mice and female COX-1^+/−/COX-2^+/− mice; cells from these animals were isolated by an identical protocol as that described above.

For Western blot analysis, proteins were collected and dissolved in protein lysis buffers (10 mM Tris, pH 7.6, 1% Triton X-100, 100 mM NaCl, and 2 mM PMSF). 20 μg of protein was loaded and separated by 10% SDS-PAGE. Proteins were then transferred to polyvinylidene difluoride membranes and blotted with polyclonal antibody specific to murine COX-1 (provided by Dr. William Smith, Michigan State University, East Lansing, MI) and mAb specific to COX-2 (Transduction Labs). Membranes were also probed with anti–β-actin antibody (Santa Cruz Biotechnology) to normalize sample differences between the gel lanes. Immunodetection was done using the Enhanced Chemiluminescence Western blotting detection system (ECL kits) purchased from Amersham Pharmacia Biotech. For Northern blot analysis, total RNAs were extracted from the cell with 4 M guanidium thiocyanate homogenization buffer (4 M guanidium thiocyanate, 0.5% sodium laurel sarcosinate, 1% β-mercaptoethanol in 100 mM Tris-HCl, pH 7.5). 20 μg of the total RNA was loaded and then electrophoresed on formaldehyde agarose gels. The separated RNAs were transferred to Duralon UV membrane and hybridized with COX-1– or COX-2–specific cDNA probes. These membranes were later probed with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) to normalize the RNA loading of each sample.

Primary embryonic fibroblast cells were seeded in duplicate in 12-well plates and grown for 24–48 h in DMEM containing 10% fetal bovine serum. Cells were treated with 100 ng/ml phorbol ester (12-o-tetradecanoylphorbol 13-acetate [TPA]) or vehicle for 10 h. PGE2 accumulations in the medium were measured using RIA kits purchased from Amersham Pharmacia Biotech. Cyclooxygenase activities were quantified by adding 30 μM free AA to the culture and measuring PGE2 production after 15 min had elapsed. The quantity of cells per well was counted using a hemacytometer. The experiments were conducted using primary cultured cells from several different cyclooxygenase-deficient embryos, and similar data were recorded.

Early passage embryonic fibroblasts were plated in 100- or 60-mm dishes, and 10–20 μg of plasmids containing pEJ-ras and SV40 oncogenes were transfected at a 1:1 ratio into the cells by calcium-phosphate precipitation methods. The cells were allowed to grow for 2 d after transfection, then split in a 1:3 ratio into new tissue culture plates and cultured for an additional 2 wk. Plates were fixed with ethanol and subsequently stained with Giemsa dyes; foci formed on the plate were counted. Growing foci were also picked and allowed to grow to confluence to measure cyclooxygenase expressions.

Cells expressing various levels of cyclooxygenase were plated at the same density in 0.33% soft agar with different NSAIDs: indomethacin (Sigma Chemical Co.), NS-398 (Cayman Chemicals), sulindac (Sigma Chemical Co.), ibuprofen (Sigma Chemical Co.), and piroxicam (Sigma Chemical Co.). The drugs were dissolved in DMSO, and identical volumes of top agar containing various concentrations of drugs were replenished at days 2 and 5. Cells were allowed to grow for 2 wk, and foci were counted. Duplicate plates were counted for each type of drug at each different concentration.

DNA synthesis was estimated by [³H]thymidine incorporation into cellular DNA. Cells in triplicate were plated in 24-well plates for 24 h and then treated with vehicle or a different NSAID for 24 h. [³H]Thymidine (1 mCi/ml) was added and allowed to label the cells for 24 h. The cells were washed twice with cold DMEM, twice with 10% TCA, and three times with ethanol. The cells were then lysed in NaOH/SDS, and radioactivity was counted.

Postconfluent cells were treated with each of the NSAIDs for 72 h. NSAID-treated and vehicle-treated cells were washed with PBS and then lysed in buffer containing 10 mM Tris-HCl (pH 8.0), 1% SDS, and proteinase K. Lysates were digested for 3 h at 55°C and then extracted with phenol/chloroform. The genomic DNA was precipitated in ethanol and resolved on a 1.5% agarose gel at 60 V for 4 h.

Cells from each embryo were collected and grown as described in Materials and Methods, the genotype of each cell line was identified by PCR, and total RNA was used for Northern blot analyses. As anticipated (Fig. 1, a and b), wild-type cells express both TPA-inducible COX-2 and constitutive COX-1 mRNA, whereas cells from COX-1^−/− or COX-2^−/− embryos each lack expression of their respective wild-type COX genes. Cells from COX-1^−/−/COX-2^−/− deficient embryos lack expression of both COX-1 and COX-2 transcripts. A small amount of abnormal COX-2 transcript is detected by Northern blot analysis in COX-2^−/− cells; however, Western blot analyses and measurements of PGE2 production confirm the disruption of the cyclooxygenase genes.

Reports have shown that COX-1 and COX-2 isozymes exhibit distinctive selectivity over a different source of arachidonic acid (AA). COX-2 is more efficient at using mitogen-activated endogenous AA, whereas COX-1 is more efficient at using exogenously added AA, or AA released by stimuli such as calcium ionophore 17,18. Therefore, we investigated the production of PGE2 from these cells by measuring metabolism of either exogenous AA or endogenous AA. In the presence of 30 μM of added AA (Fig. 2 a), both COX-1– and COX-2–deficient cells produced less PGE2 than wild-type cells. TPA stimulation doubled the production of PGE2 in both wild-type and COX-2–containing cells. In the absence of added AA (Fig. 2 b), when 8-h PGE2 accumulation was measured, stimulation by TPA was found to dramatically increase PGE2 production in both wild-type and COX-2–containing (COX-1^−/−) cells, but the COX-2^−/− cells produced almost undetectable levels of PGE2 even with an active COX-1 enzyme. TPA-activated AA releases were also measured by labeling the cells with ³H-AA, and all of the cells, regardless of the presence of the enzymes, demonstrated similar induction of AA release in response to TPA or calcium ionophore (data not shown). Also, as expected, the COX-1^−/−/COX-2^−/− cells do not produce detectable PGE2 either in the presence or the absence of AA or TPA. The failure to detect any PG production in the COX-1^−/−/COX-2^−/− cells confirms the inactivation of both genes. Previous studies by Kirtikara et al. 19 showed that COX-1 or COX-2 is increased in the null cells to compensate for the deficiency of PG production. Our data support this finding by showing that in the COX-1^−/− cells, PGE2 production was increased both at the basal and the stimulated level (Fig. 2 b). However, we failed to detect any increased level of PGE2 production in the COX-2^−/− cells compared with the wild-type cells.

To examine the role of cyclooxygenases in neoplastic transformation, we conducted transformation assays in primary cultured embryonic cells. The oncogene ras was used in this study because its expression contributes to many malignancies, and several studies, including our unpublished data, have demonstrated that transformation by ras induces COX-2 gene expression 20. The expression of a mutant ras gene usually does not by itself lead to a transformed phenotype in primary cultured cells; additional mutations in other protooncogenes such as c-myc, or inactivation of tumor suppressor genes such as p53, are required to achieve a full malignant transformation 21. Thus, we evaluated the contribution of the cyclooxygenases to the transforming activity of activated EJ-ras by adding wild-type SV40 viral gene to complete the transformation. As shown in Table, transfection of EJ-ras alone did not induce any foci formation, and transfection of SV40 alone induced only a few small foci. These foci from SV40 transformation failed to grow in the 0.33% soft agar within 2 wk. In contrast, SV40 plus EJ-ras cooperatively led to dramatically increased foci formation, and these transformed cells developed a fully transformed phenotype, as shown by their loss of contact inhibition and their rapid growth in soft agar and nude mice. By comparing the wild-type cells and the cells deficient in either or both of the cyclooxygenases, our results demonstrate that cells lacking either or both of the cyclooxygenases can be readily transformed by SV40 plus ras at almost equal frequencies. Both cyclooxygenase-null cells and wild-type cyclooxygenase-containing cells showed similar morphology and growth behavior in soft agar. This result indicates that cyclooxygenases are not required for transformation per se, nor specifically for the operation of the ras-dependent transformation pathway.

Expression of the cyclooxygenase isozymes in transformed embryonic fibroblasts was analyzed using Western blots. As shown in Fig. 3 a, six of the seven transformed clones showed moderate or marked increases of COX-2 expression, and three of them had increases in the expression of both enzymes. However, one of the clones showed unchanged expression of both cyclooxygenases, again suggesting that fibroblasts can be transformed without altering the cyclooxygenase pathway. Interestingly, we found that the embryo fibroblasts immortalized only by SV40 expressed less cyclooxygenase, both COX-1 and COX-2, than the untransformed cells (Fig. 3b and Fig. c). Further studies confirmed that the corresponding cyclooxygenase product, PGE2, was dramatically reduced in the cells that were transformed by SV40 alone (data not shown). Isoenzyme expression was also measured in cells lacking either cyclooxygenase (Fig. 3c and Fig. d). In the cells lacking COX-1, transformation by EJ-ras plus SV40 led to an increased basal expression of COX-2 (Fig. 3 d, lanes 3–6), and in cells lacking COX-2, the expression of COX-1 was also markedly increased (Fig. 3 c, lanes 7–10). The results from the Western blot analyses are consistent with our Northern blot analyses, further confirming the genotype of each cell line we used in these studies. Taken together, these experiments illustrate that both cyclooxygenases can be selectively altered by oncogene ras or viral gene SV40; these actions are oncogene specific and might involve the alterations of signal transduction pathways that lead to the expression of both cyclooxygenases. However, despite the changes in cyclooxygenase gene expression, transformation does occur in the absence of both cyclooxygenases.

Clinical and experimental evidence strongly suggests that NSAIDs are anticarcinogenic, antiproliferative, and antineoplastic. Although some evidence suggests that the effects of NSAIDs are achieved via their known inhibitory effects on COX-2, recent data from several independent groups demonstrate that some of the actions may be COX independent. Cells in which either or both cyclooxygenase genes have been disrupted offer distinct advantages for evaluation of the antineoplastic activities of these drugs. For this purpose, we examined the actions of several NSAIDs on transformed cells with different deficiencies in the expression of the cyclooxygenase isozymes. The antineoplastic activities of the drugs were evaluated by their influence on colony formation in soft agar, [³H]thymidine incorporation, and by DNA fragmentation analysis. NSAID dosages were varied from 1 to 200 μM, and these dosages are compatible with those used in several studies where antineoplastic effects were achieved 11,22,23,24,25. The therapeutic plasma concentration for NSAID therapy can be achieved with up to 10 μM indomethacin, 20 μM sulindac, 20 μM piroxicam, or 300 μM ibuprofen 26. However, it is reported that the local concentration of NSAIDs in certain organs such as intestines can be built up to an even higher level 27. Therefore, the dosages we used in these studies include both the physiological dosages and the higher dosages that might be achievable in certain organs and tumor tissues in vivo.

We evaluated NSAID actions by seeding the same number of cells in soft agar containing various concentrations of different NSAIDs. All of the NSAIDs used (NS-398, sulindac, indomethacin, piroxicam, and ibuprofen) reduced the size and the number of colonies derived from the cells containing either, neither, or both cyclooxygenase genes in a dosage-dependent manner, with the inhibition being statistically significant at the higher dosages (Fig. 4). Several cell clones of each cyclooxygenase-deficient cell line were used in this study. Quantitative differences in colony formation were observed in the presence of various NSAIDs, but these differences were not correlated with the presence or the absence of cyclooxygenase expression, as seen in Fig. 4. Thymidine incorporation was used to examine DNA synthesis of the cells in the presence of different NSAID doses. At the 20 and 100 μM doses of a COX-2–selective NSAID, NS-398, [³H]thymidine incorporation of all four cell lines examined was reduced by 15–68% regardless of the presence or absence of the two isozymes (Fig. 5). At these doses, where PGE2 production was reduced by 95% in the wild-type cells and the COX-1^−/− or COX-2^−/− cell lines, further addition of PGE2 up to 10 μM failed to restore DNA synthesis (data not shown). Finally, we examined the apoptotic responses of the various deficient cells upon treatment with the different NSAIDs. At higher doses (indomethacin 200 μM, sulindac 200 μM, piroxicam 200 μM, NS-398 100 μM, and ibuprofen 1 mM), these drugs potently induced apoptosis in postconfluent cells at 48–72 h (Fig. 6). The cell death was induced in cells with neither COX-1 nor COX-2 expression. Based on these results, we conclude that the antiproliferative and antineoplastic actions of NSAIDs in these cells at these concentrations are independent of cyclooxygenase inhibition.

The activating mutations of ras occur in ∼30% of all human tumors. In squamous carcinomas, these mutations appear at early or intermediate stages of neoplasia. In colon cancers, ras gene mutations were found in 50% of adenocarcinomas and larger adenomas 28. To initiate studies on the involvement of the cyclooxygenases in transformation and malignancy, we developed cells that lack COX-1, COX-2, or both, and studied ras-induced malignant transformation. We found that transformation by SV40 alone downregulated the expressions of both cyclooxygenases. However, addition of the activated ras oncogene induced malignant transformation accompanied by an increased expression of either or both cyclooxygenase enzymes. Although most of the clones exhibit increased expression of either or both cyclooxygenases, some do not express either of the genes at increased levels. This result by itself suggests that the expression of either cyclooxygenase does not have a direct implication on mechanisms of transformation. Furthermore, our transformation experiments conducted in the cyclooxygenase-null cells provide still stronger evidence that transformation can occur in the absence of either of the cyclooxygenases, and that cyclooxygenase overexpression is not necessary to induce neoplastic transformation.

Nevertheless, there is abundant evidence suggesting that high levels of cyclooxygenase activity do play a critical role in some malignancies. Recent reports show that forced overexpression of cyclooxygenase enzymes increases metastatic potential, invasiveness, and angiogenesis of tumor cells both in vitro and in vivo 29,30,31. Moreover, carcinogenesis is a chronic, sequential, and progressive process that usually involves accumulation of mutations of protooncogenes and tumor suppressor genes in later steps. Mechanisms involving cyclooxygenases in mutagenesis have been suggested 32,33. COX-2 overexpression also leads to resistance to apoptosis and increases in the cellular life-span 10,11; these phenotypic changes may result in the additional development of mutations leading to more malignant behavior. Elevated cyclooxygenase expression is also known to increase malondialdehyde generation, providing another mechanism to foster higher cellular mutation rates 34. So while activation of ras by itself is not usually sufficient for malignancy, the elevations in cyclooxygenases induced by the activation of ras could well predispose the cells to mutations in p53 or other genes that participate in the later steps towards high-grade malignancy. It is also noteworthy that in addition to COX-2, COX-1 expression is often induced by the activated ras oncogene, suggesting that both cyclooxygenases may serve some overlapping roles in tumorigenesis. Further studies are required to clarify the functional differences and relative importance of the two isozymes in cell malignancy and in the different stages of carcinogenesis.

NSAIDs such as sulindac and indomethacin have been shown to be beneficial for the prevention and in some cases for the treatment of certain cancers, although the underlying mechanisms are still unknown. Inhibition of cellular proliferation and/or induction of apoptosis are reported to be the mechanisms that cause the regression of tumors in vivo 35. The antiproliferative and apoptotic effects of NSAIDs have been documented in a variety of cells and cell lines, including colon cancer cells, breast cancer cells, and fibroblasts 22,23. Considerable recent data suggest that some of the NSAIDs actions may not be entirely explained by inhibition of cyclooxygenases. First, some NSAID-related compounds, which are not cyclooxygenase inhibitors, can also induce antineoplastic changes in both the cell cycle and cellular apoptotic responses 36,37. Second, the antineoplastic or anticarcinogenic activity of NSAIDs may not coincide with the inhibition of PG production in some tumor cells and animals 38. Finally, Hanif et al. 39 found that sulindac and piroxicam inhibit the growth of and induce apoptosis in HCT-15 cells, which were shown to produce neither COX-1, COX-2, nor PGs. Elder et al. 40 also showed that NS-398 induced apoptosis in S/KS cells, which lack COX-2 protein. In contrast, Murphy et al. 41 showed that the growth of the cells expressing COX-2 was not suppressed by COX-2–selective inhibitors. In this study, we further illustrate that a selective knockout of cyclooxygenase activity does not change the ability of NSAIDs to inhibit transformation or to induce apoptosis. These results confirm that inhibition of cyclooxygenase is not the only mechanism responsible for the anticarcinogenic and antineoplastic activity of NSAIDs.

NSAID actions that are independent of cyclooxygenase have been reviewed by Abramson and Weissmann 42. New evidence suggests that many of the traditional NSAIDs have targets other than cyclooxygenases. A recent report by Herrmann et al. 24 demonstrated that sulindac sulfide inhibits the ras signaling apparatus, potentially via an eicosanoid-independent pathway. Lehmann et al. 43 demonstrate that many NSAIDs are ligands of peroxisome proliferation–activated receptors (PPARs) α and γ. Chan et al. 25 demonstrated that the elevation of the PG precursor AA could be involved in the induction of apoptotic response by the NSAIDs (i.e., sulindac and indomethacin) through production of ceramide. However, due to the higher concentration of NSAIDs used in their study, it is not clear whether the induction of AA is a result of cyclooxygenase enzyme inhibition or direct production of AA through activation of phospholipases. NSAIDs are a large group of structurally diverse compounds; therefore, differences in the actions of individual NSAIDs are also expected. For example, aspirin inhibits the activation of the transcription factor NF-κB, whereas indomethacin does not 44. Recently COX-2–specific inhibitors were demonstrated to have potent antiproliferative effects and to induce apoptosis in cultured cells 20,45,46. Whether or not actions exerted by these drugs are mediated through inhibition of COX-2 needs additional investigation.

We have also observed in our studies that there are differences between cell lines in their sensitivity to NSAIDs that are independent of cyclooxygenase expression (Fig. 5). Thus, it seems likely that the differences in the sensitivity of tumor cell lines to NSAIDs as reported in the literature may also be independent of cyclooxygenase expression, particularly when NSAIDs are used at high concentrations.

Finally, the effects of NSAIDs on apoptosis, cell growth, and DNA synthesis studied here are those often cited to account for the antitumor activity of this class of compounds, yet in accord with the results of others, our data show that NSAID inhibition of the cyclooxygenases does not account for the antiproliferative and apoptotic actions seen in transformed embryonic fibroblasts. Our results further demonstrate that although cellular transformation by ras and SV40 usually leads to a dramatic increase in the expression of COX-1 and COX-2 when present, neither these inductions nor the presence of either of the enzymes is required for neoplastic transformation per se. These results suggest that the involvement of cyclooxygenase expression in neoplastic growth may occur at steps beyond those involved in transformation.

This study is supported by U.S. Public Health Service grant RO1 AM-16177 and institutional funds.