PROSTAGLANDINS AND CANCER

Susan M. Fischer

The University of Texas M. D. Anderson Cancer Center, Science Park-Research Division, Smithville, TX 78957

Received 9/3/97 Accepted 9/23/97

4. HUMAN CANCER AND EXPERIMENTAL ANIMAL MODELS

The strongest evidence that PGs contribute to the development of cancer originates from the numerous studies in experimental animal models, as well as human epidemiologic studies, showing that the incidence of tumorigenesis is reduced in the presence of inhibitors of PG synthesis. Although a number of organ sites have been studied in rodents, the majority of this work has been carried out in skin and colon. The second line of evidence is the elevation in PG levels in many tumors. These models will be discussed below, followed by comments on breast cancer. Potential mechanisms of PG actions will be discussed in the final sections of this review.

4.1. Skin Carcinogenesis

Mouse skin has proven to be one of the best animal model systems for studying the multistage nature of carcinogenesis (60, 61). Skin tumors can be readily induced by the sequential application of a subthreshold dose of carcinogen (initiation stage) followed by repetitive treatment with a noncarcinogenic tumor promoter (promotion stage). In this model, promotion is most often accomplished by using the phorbol ester 12-O-tetradecanoylphorbol-13-acetate (TPA), although a variety of agents have been identified as skin tumor promoters, including nonphorbol esters such as benzoyl peroxide, anthralin, dihydroteleocidin B and ethyl phenylpropriolate (61-62). The promotion stage has been further subdivided into additional stages in which the sequential use of incomplete or partial promoters like the ionophore A23187 and mezerein can replace the use of a complete promoter such as TPA (60-61).

Investigations into the role of arachidonic acid metabolites in tumor promotion were spurred by the observation that TPA induces cytotoxicity, inflammation and increased vascular permeability (63). A crucial role for inflammation in phorbol ester promotion was suggested by several early studies showing that the anti-inflammatory steroids, dexamethasone and fluocinolone acetonide, were extremely effective in preventing tumor development (32, 64). Although other mediators of inflammation, including vasoactive amines, cytokines and growth factors have been and continue to be investigated, most of the emphasis has been on arachidonic acid and its oxidation products (65-67).

The early observations that steroidal anti-inflammatory agents are potent anti-promoters led to two approaches in assessing the importance of eicosanoids in tumor promotion: first, determination of whether tumor development could be modified through exogenous application of PGs and second, the effect on promotion of inhibitors of various parts of the arachidonic acid cascade, including use of a series of NSAIDs. In one study, skin tumor experiments were carried out in which PGE₂, PGF_2alpha, PGD₂ or arachidonic acid were applied either alone or with TPA during tumor promotion. The various PGs were found to be distinct in their modulation of the tumor response and that they depend both on the particular eicosanoid used as well as the dose and time of application (68). Other studies showed that PGE₂ and PGF_2alpha enhanced the development of papilloma and squamous cell carcinoma in mouse skin and promoted the growth of basal cell carcinomas in rats initiated with 3-methylcholanthrene (69, 70). Several studies also indicated that neither the PGs nor arachidonic acid were themselves tumor promoters (68, 71, 72).

While these exogenous application studies indicated that PGs could be used to modify tumor yield, understanding their action was confounded by the fact that TPA itself induces considerable PG synthesis. A series of studies have since been carried out using inhibitors of specific enzymes of the arachidonic acid cascade. The PLA₂ inhibitor dibromoacetophenone was shown by several laboratories to have strong inhibitory activity against phorbol ester promotion in several strains of mice (73). The effect of the COX inhibitor, indomethacin, however, is mouse-strain dependent: while inhibition occurs in CD-1 and NMRI mice, enhancement of promotion occurs in SENCAR mice (73). To show that increased PGE₂ synthesis is essential for DNA synthesis and tumor promotion, 'add-back' experiments were carried out in NMRI mice (74). While adding back PGE₂ had little effect on indomethacin reduction of tumor rate or yield, adding back PGF_2alpha produced a dose-dependent partial reversal (74, 75). Collectively, these inhibitor studies support the contention that arachidonic acid release and metabolism are essential components of the tumor promotion process.

Another observation supporting a critical role of PGs in the development of skin tumors is that carcinomas from animals no longer being treated with tumor promoters show constitutive overexpression of COX-2, but not COX-1, at the message and protein level (76). This correlates very well with the elevated PGE₂ levels measured in papillomas and especially in carcinomas (76). Although at least several classes of tumor promoters, induce COX-2 (but usually not COX-1) in keratinocytes in vivo and in vitro, the levels of expression or levels of PGE₂ synthesis are not as high as seen in tumor promoter-independent tumors (38, 76). The mechanisms involved in the constitutive upregulation of COX-2 expression in skin tumors are currently unknown. Although most skin tumors generated by the multistage protocol carry an activated H-ras gene, and activation of the Ras-MAP kinase signaling pathway can induce COX-2, there is not a strong evidence that this drives the overexpression in tumors. This is indicated by the lack of correlation between expression levels and ratios of normal to mutated H-ras alleles in several tumor cell lines (38, 76). Several other possibilities exist, including enhanced transforming growth factor alpha production and elevated epidermal growth factor receptor expression in tumors (77). It is expected that these and other studies on the regulation of COX-2 at the level of its promoter region soon will shed light on the mechanism(s) responsible for the elevated COX-2 expression and high PG production.

4.2. Tumors of the Colon

The epithelium of the colon is similar to the epidermis of skin in that it is a self-renewing tissue, i.e., the epithelium in both cases continuously undergoes proliferation, differentiation and sloughing. Over the last decade, the multistage nature of human colon cancer has been defined and has similarities to the mouse skin model with regards to the progressive acquisition of mutations and chromosomal abnormalities. This raises the likelihood that there are similarities in physiological processes as well, including a role for PGs.

In rodent models of colon carcinogenesis COX inhibitors such as indomethacin, piroxicam and sulindac exhibit chemoprotective effects (78-83). Piroxicam is particularly effective, providing up to a 70% reduction in tumor incidence, although only at very high doses (84). Piroxicam is not an irreversible COX inhibitor like aspirin and it has been suggested that it may not be as effective as aspirin in certain cell types (85). Treatment with indomethacin of Sprague-Dawley rats with dimethylhydrazine-induced intestinal tumors reduced the tumor size and incidence by 40% (86). Indomethacin also inhibited the development of colon tumors induced by methyl nitrosourea in CD-Fischer rats (87).

The ability of human colonic mucosa to synthesize PGE₂, PGD₂, PGF₂alpha and thromboxane B₂ has been well demonstrated (87-89). Using different approaches, several laboratories have shown that human colon cancer tissue produces more PGE₂ than the surrounding normal tissue (87). One study measured PGE₂ levels in normal-appearing mucosa and in tumors from patients with adenomatous polyps or colon cancer. The level of PGE₂ in polyps was elevated by about 40% and in colon cancer by almost three fold (90). This increase in PGE₂ with progression from normal to malignant tumor is similar to that described for progression of tumors in the mouse skin model (76). However, when cell lines derived from various tumor types were compared, lines derived from colorectal adenocarcinomas were among the lowest producers of PGE₂ (91). Sine biopsy specimens and tumors contain several cell types, the origin of the PGs has been questioned (4). It has been recently shown that high tumor PGE₂ is due to the very high level of PGE₂ produced by the resident (not peripheral) mononuclear cells (89). Thus, a paracrine mechanism may exist in which the mononuclear cells produce the majority of the PGE₂ while the target or responding cells are the colonic epithelium.

The elevation of PGs in colon tumors suggested that the expression of COX-1 and/or COX-2 is altered in tumors. Elevated levels of COX-2 mRNA and protein have been found in rodent colon tumors induced by treatment with carcinogen (92). In humans, COX-1 can be detected in both normal and tumor colonic tissue, while COX-2, is rarely detectable in normal colonic tissue (93, 94). In a study by Eberhart et al (95), COX-1 and COX-2 mRNA levels were measured in normal human colonic mucosa from patients with tumors and in their polyps or carcinomas. COX-2 was increased over normal in 12 of the 14 carcinomas and in half of the adenomas. On the other hand, the levels of COX-1 mRNA were essentially the same in normal and tumor tissues. In another study by Kargman et al (94), similar results were reported although the percentage of colon cancers with increased COX-2 expression was lower (19 out of 25 tumors). These and other studies have contributed to the concept that COX-2 expression is responsible for the progressive increase in PGE₂ synthesis in colorectal neoplasia (96).

The importance of elevated COX-2 in the development of colonic tumors has been shown not only through studies using COX inhibitors, but also with a genetic model. Mutant mice carrying a germline mutation in the APC gene were generated that developed mulitple intestinal neoplasms (min) and exhibited a phenotype similar to human familial adenomatous polyposis (97). The adenomas that developed spontaneously in this mouse had a threefold higher level of mRNA and protein expression of COX-2 than normal mucosa. Immunohistochemical localization also indicated the presence of COX-2 in dysplastic foci, suggesting that COX-2 levels may be increased at an early stage in tumor development (98). Recently, the nonselective COX inhibitor, sulindac, was shown to reduce COX-2 expression and to inhibit tumor formation in the min mouse (99). In another min mouse study a COX-2 selective inhibitor was shown to be very effective in reducing the number of polyps (100). An even more convincing demonstration of the contribution of COX-2 to tumor development was the introduction of a knockout mutation of the COX-2 gene into APC knockout mice. These animals exhibited a dramatically reduced number of spontaneous tumors in a gene-dosage dependent manner, strongly suggesting that PGs play a major role in tumor development (100).

Although it is often difficult to extrapolate results from animal experiments to humans, a number of clinical trials and epidemiological studies strongly implicate PGs in the development of human colon cancer. Sulindac substantially reduces the number and size of rectal polyps in individuals with familial polyposis. Nevertheless, when sulindac administration was discontinued, tumors recurred (101-103). These polyp regression studies indicate that sulindac is inhibiting a process associated with preneoplastic colonic tumor development. Sulindac was shown to reduce PG synthetic capacity in colonic mucosal samples from patients on long-term sulindac therapy (102). However, in another study, with a small number of patients, the COX inhibitor, indomethacin, did not induce regression of polyps (104). This discrepancy has been explained by differences in the pharmacokinetics of sulindac and indomethacin (3). Relatively high concentrations of sulindac metabolites are found in the colon while indomethacin is primarily excreted in the urine, with little amounts reaching the colon (105).

Numerous epidemiologic studies looked for an association between chronic aspirin ingestion and reduced incidence of colon cancer. In one of the earliest reports Kane et al (106) found that in a case-control study (n=715) regular aspirin intake was associated with a nearly 50% reduction in incidence of colon cancer. Rosenberg et al (107) reported similar findings in an even larger study (n=1326). One of the more recent studies, by Thun et al (108), examined mortality from colon cancer. They reported that regular aspirin ingestion was associated with a 50% reduction in colon cancer deaths over a six year follow-up period. There are some caveats in drawing strong conclusions from this study, however. There was little association between the use of acetaminophen and fatal colon cancer even though acetaminophen is a moderate inhibitor of PG synthesis. It was suggested that possibly aspirin induced bleeding, leading to earlier diagnosis and improved survival (108). Greenberg et al (109) showed that the protective effect of aspirin was not influenced by the number of prior adenomas and was the same among men and women. Suh et al (110) further showed a frequency-of-use effect, i.e., consumption of aspirin several times a day was more protective than less frequent use. Giovannucio et al (111) reported that use of aspirin > 2 times per week resulted in a reduced risk of total and advanced (metastatic and fatal) colorectal cancers in male health professionals. The issue of dose was also addressed by Gann et al (112) who observed that regular use of aspirin at low doses, that recommended for prevention of myocardial infarction, did not reduce the incidence of colon cancer during a five year randomized treatment trial. In disagreement with the above reports, a small study (n=181) by Paganini-Hill et al (113) found that there was a 50% increased risk of colon cancer among daily aspirin users. Overall, however, the large majority of these epidemiologic studies and clinical trials show that regular use of aspirin is associated with significant reduction in risk of development of colorectal cancer. This protective effect was originally thought to be limited to the gastrointestinal tract in that little association of aspirin and reduced risk of development of cancers of lung, breast, endometrium, ovary, testis, bladder, lymph nodes, hematopoietic cells or melanocytes were reported by Rosenberg et al (107). However, a more recent study found a significant inverse association with cancers of all sites (114).

4.3. Tumors of the Breast

The role of PGs has also been evaluated, in both rodents and humans, in tissues and tumors other than those from the skin and colon, notably the breast. Many of these studies are contradictory, however. For example, although, in carcinogen-induced rat mammary tumors, the PGE₂ content was elevated (115) and could be reduced by indomethacin, indomethacin was reported to have either no effect (116), an inhibitory effect only with diets high in linoleate (117), or to promote tumor proliferation, resulting in increased tumor size (118). The level of PGs in human breast tissue and tumors has also been determined by several laboratories (119-121). As with the rat mammary carcinogenesis model, some contradictions have been noted which have been attributed, at least in part, to the lack of standardization of methodologies (121). The majority of the PGE₂ appears to be synthesized by non-epithelial cells, particularly fibroblasts (122). It has also been suggested that the elevated PG production could be used as a marker of high metastatic potential for breast cancer (119, 123). COX expression levels have been reported for two human breast cancer cell lines, estrogen-dependent MCF-7 and estrogen-independent and aggressive, MDA-MB-231, cells. MCF-7 cells had high COX-1 but barely detectable COX-2 expression; the reverse was seen in the MDA-MB-231 cells. The high PG synthesis by the MDA-MB-231 cells was suggested to be causatively associated with their highly invasive and metastatic phenotype (124). However, another study found no significant correlation between estrogen receptor status, tumor size, lymph node involvement and length of the post-treatment disease-free state, suggesting that PG levels are not of prognostic value (125).

Strong support for a role of PGs in human breast cancer comes from a recent epidemiology study. This case-control study compared 511 cases of breast cancer with 1534 control subjects and found that the relative risk of breast cancer was reduced by the use of NSAIDs (126). A greater risk reduction was seen with more frequent use (daily for > five years). It is likely that this study will provide impetus for further investigation into the regulation of COX expression in breast tissue and the function of PGs in breast tumor development.