[Frontiers in Bioscience 2, d482-500, October 1, 1997]

	[Frontiers in Bioscience 2, d482-500, October 1, 1997] Reprints PubMed CAVEAT LECTOR
	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 5. MECHANISMS OF PROSTAGLANDIN ACTION One of the primary approaches to the understanding of the function of PGs in any tissue has been through the use of inhibitors of COX activity. The NSAIDs that have most commonly been used in in vitro and in vivo studies include drugs such as aspirin, indomethacin, sulindac, ibuprofen and piroxicam, all of which inhibit PG synthesis albeit by different mechanisms. Aspirin inhibits the COX but not peroxidase activities of both COX-1 and COX-2 by acetylating a particular serine and thus blocking the channel that leads to the active site. This results in irreversible inhibition of PG synthesis although a concurrent increase in 15(R)-HETE occurs, particularly with COX-2 (127,128). Indomethacin forms a tight, slowly dissociable complex with COX that induces an inhibitory conformational change (129). Ibuprofen and piroxicam, on the other hand, compete with arachidonic acid for the active site (130, 131). In general, most of these classical NSAIDs are better inhibitors of COX-1 than COX-2, although some, like flurbiprofen and ibuprofen, have nearly equal IC₅₀ values (132). The recent demonstration that COX-2 is the inducible isoform and that its high expression occurs at sites of inflammation has spurred the development of isoform specific inhibitors. An additional impetus is the observation that inhibition of PG production, via inhibition of both COX-1 and COX-2, can lead to gastrointestinal lesions and nephrotoxicity. Thus, it was viewed as desirable to reduce the inflammatory effects of COX-2 while retaining the cytoprotective effects of COX-1 (133). As a result, a variety of COX-2 specific inhibitors have recently become available for testing One of the first inhibitors, NS-398, has an IC₅₀ of 3.8 x 10^-6 M in vitro for COX-2 and exerts no effect on COX-1 at 10^-4M (134). In vivo, NS-398 blocked COX-2 expression in inflammatory cells induced by exogenous stimuli and completely inhibited PG synthesis from these cells whereas it did not not affect PG production from COX-1 in the stomach (133). NS-398 and another COX-2 selective inhibitor, DuP697, cause conformational changes in COX-2 that lead to irreversible loss of activity (135). CGP28238 and L-745,337 also exhibit a dramatic COX-2 selectivity when tested on several cell types and in in vivo models of inflammation (136, 137). Together, these pharmaceutical studies strongly support the hypothesis that COX-1 is utilized for cytoprotection and that COX-2 primarily plays a pro-inflammatory role. However, the recent generation of knockout mice for COX-1 and COX-2 suggest that this view may be too simplistic. Contrary to the prediction that COX-1 knockouts would have spontaneous gastrointestinal ulcerations, this pathology was not found. Furthermore, these mice were less sensitive than normal mice to the ulcerating effect of indomethacin (138). The COX-2 knockout mice also produced unexpected results. In these animals topical TPA treatment caused edema to the same extent as in wild type mice (139). In keeping with high PG levels, COX-2 knockouts, however, did show a reduced endotoxin-induced hepatocellular cytotoxicity (140). It is clear from these animal models that, to a large extent, the tissue specific function of COX-1 and COX-2 remains undefined. 5.1. Cell Proliferation Prostaglandins have many biological effects on different tissues and thus act as endogenous biological modifiers. As mentioned above, specific receptors have been identified for each of the major PGs, as well as for prostacyclin and thromboxane (47), which are coupled to different signal transduction pathways (47, 53). The rapid metabolic breakdown in the circulation suggests that PGs are not classical hormones but act as autocrine or paracrine factors within a given tissue (141). In spite of extensive research on the effects of specific eicosanoids, little is known about their mechanism(s) of action at the cellular and molecular level. For most tissues, it is also not clear whether the effect on a given tissue of a particular PG is the same for normal and malignant cells. For some, but not all cells, a link has been established between synthesis of PGs and control of cell growth. In Balb/c 3T3 cells, epidermal growth factor-dependent proliferation is inhibited by the COX inhibitor, indomethacin (142). In these cells, PGF_2alpha is not a mitogen by itself but acts a permissive factor that allows the mitogenic action of a growth factor. PGF_2alpha stimulates proliferation of osteoblast, MC3T3-E1, cells by increasing the number of high affinity binding sites for insulin-like growth factor-1 (143). Thromboxane A₂ has also been shown to stimulate mammary epithelial cell growth (144). Normal murine epidermal cell proliferation does not appear to depend on PGs. This concept is based on the inability of PGE₂ or PGF_2alpha or their more stable, 15(S)-15-methyl, derivatives to stimulate DNA synthesis and the lack of inhibition of normal proliferation by the PG synthetase inhibitor, indomethacin (145-147). In human keratinocyte cultures, however, PGs do appear to be required for normal cell proliferation. A correlation was observed between PGE₂ production and cell proliferation and this proliferation was inhibited by indomethacin in a manner that could be overcome by addition of PGE₂ but not PGF_2alpha (148). In rat skin, topical administration of PGE₁, PGE₂ and PGF_2alpha produced marked increases in DNA, RNA and protein synthesis, further suggesting that PGs are regulatory factors in epidermal cells (149). Unlike in normal mouse skin, in vivo topical application of TPA (and other irritants) results in a proliferation that is dependent on specific PGs. Indomethacin was shown to inhibit TPA-induced epidermal hyperproliferation and this inhibition could be overcome by topical application of PGE₂ but not PGF_2alpha (145, 146, 150). In addition, although not mitogenic by itself, PGE₂ is co-mitogenic when applied with TPA (151). With regards to colon, two transformed human colon cancer cell lines were recently evaluated. The in vivo growth and in vitro colony formation of one cell line, HCA-7, which constitutively expresses high levels of COX-2 protein was inhibited by the selective COX-2 inhibitor, SC-58125. The other cell line, HCT-116, however, does not express COX-2 protein and is not growth inhibited by SC-58125. This study suggests that SC-58125 is acting in a specific manner rather than via nonspecific cytostatic mechanisms and that PGs contribute to the growth of some, but not all, colon cancers (152). It was also noted that the responding HCA-7 cells are more differentiated than the HCT-116 cells, which may be an important determinant (152). The response of two human colon adenocarcinoma cell lines, SW111 and HT-29, to the proliferative or apoptotic effects of a series of eicosanoids was compared. While HT-29 cells showed increased proliferation in response to PGE₂ and leukotriene B₄, SW111 cells also responded to PGI₂ and PGF_2alpha, although neither distribution of cells in the cell cycle nor rate of spontaneous apoptosis was altered (153). On the other hand, the NSAID, sulindac, reduced the proliferation rate of HT-29 cells and increased the rate of apoptosis (154). Other NSAIDs such as aspirin, indomethacin and piroxicam also reduced cell cycling and all except aspirin induced apoptosis in these cell lines (155). In addition, another study suggests that COX-2 participates in the proliferation of colon cancer cells. Several gastrointestinal cancer cell lines that highly expressed COX-2 were inhibited by both COX-2 selective and nonselective inhibitors whereas those cell lines with low COX-2 expression showed minimal response (156). The growth stimulating effects of PGs in many cases appear to be linked to other biological modifiers, particularly the polyamines. Elevated polyamines are associated with an increase in DNA synthesis and result from induced ornithine decarboxylase (ODC) activity. Tumor promoters, both exogenous and endogenous (bile salts) induce the synthesis of ODC in a PGE₂-dependent process, based on inhibition by indomethacin and restoration by exogenous PGE₂ (157). In this respect, the colonic epithelium responds in a manner similar to murine epidermis, a tissue in which tumor promoter-induction of ODC is also PGE₂ dependent (71). In the skin multistage carcinogenesis model, ODC induction is required for tumor development; inhibitors of arachidonic acid metabolism inhibit both ODC induction and tumor development (158). Therefore, it seems that a clear link exists between PGE₂, enhanced proliferation and tumor development. In addition to effects on proliferation, PGs are associated with enhanced invasive and metastatic potentials of tumor cells. Recently, several studies have reported that increasing the expression of COX-2 was associated with alterations in cell behavior. When rat intestinal epithelial cells were transfected with a COX-2 expression vector, they showed enhanced cell adhesion due to upregulated E-cadherin expression; this was reversed by addition of sulindac sulfide (159). When human colon cancer cell line, Caco-2, was transfected with a COX-2 vector, they exhibited an enhanced invasive phenotype that was associated with an elevated expression of membrane metalloproteinase. Sulindac sulfoxide reversed both the increased invasiveness and PG synthesis (156). 5.2. Role in Apoptosis The biochemical basis for the antineoplastic characteristics of NSAIDs has generally been attributed to their ability to reduce PG levels by inhibiting COX. However, this category of drugs may have other effects that are responsible for, or at least contribute to, their anti-cancer properties. Studies using cultured tumor cell lines have shown that a variety of NSAIDs can induce apoptosis (160-164). Apoptosis, frequently referred to as programmed cell death, is a specific morphological and biochemical form of cell suicide. The apoptotic pathway can be triggered by extracellular agents, pathological processes and also occurs during normal development and tissue remodeling. Apoptosis differs from necrosis in that the apoptotic cell kills itself in manner that does not harm neighboring cells nor elicits an inflammatory response. Activation of apoptosis in tumors has thus been a target for some chemotherapies (165). Although NSAIDs can both inhibit PG synthesis and induce apoptosis, there is recent evidence that the apoptotic effects is likely not due to COX inhibition. The most striking example comes from studies with sulindac and its metabolites. Sulindac sulfoxide is consumed as a prodrug that has no COX inhibitory activity. It is readily reduced to the sulfide form in the liver and in the colon via bacterial microflora; it is this sulfide form that is responsible for the anti-inflammatory characteristics of sulindac (166, 167). However, sulindac sulfoxide is also oxidized in an irreversible manner in the liver to a sulfone that is excreted in the bile and intestine. The sulfone metabolite is devoid of COX inhibitory or other anti-inflammatory activity (168) although it still retains the ability to inhibit tumor cell growth and induce apoptosis (161). Thompson et al (168) were the first to report that the sulfone metabolite has cancer chemoprotective activity. In a chemically induced rat mammary cancer model, sulindac sulfone was found to be comparable to sulindac sulfoxide in reducing tumor incidence and numbers and in inducing apoptosis (168). Sulindac sulfone also inhibits formation of aberrant crypts in a rodent colon carcinogenesis model and recently was shown to inhibit azoxymethane-induced colon carcinogenesis in rats (160, 169). In a recent study, it was also shown that in addition to having no COX-inhibitory activity, the sulfone metabolite does not inhibit lipoxygenases or phospholipase A₂ (170). The relationship between NSAIDs, PGs and apoptosis was further explored in the HT-29 colon adenocarcinoma cells which expresses COX and synthesizes PGs and in the HT-15 carcinoma cell line which does not exhibit these properties. Addition of PGE₂ to HT-15 cells enhanced proliferation only slightly (25%); addition of sulindac or piroxicam severely reduced proliferation and this could not be reversed by addition of PGs. The PG producing HT-29 cells were also growth inhibited by sulindac and piroxicam but unlike similar experiments described above for indomethacin treatment of keratinocytes, growth could not be restored by addition of PGs. The conclusion drawn is that these NSAIDs reduce proliferation in human colonic cells through a PG-independent mechanism (171). These and related studies have raised the question of exactly how the NSAIDs exert their anti-proliferative effects. As reviewed by Abramson and Weissmann (172), NSAIDs affect a variety of membrane processes not related to PG synthesis. These include inhibition of superoxide anion generation by NADPH oxidase, phospholipase C and 12-HPETE peroxidase and coupling of mitochondrial oxidative phosphorylation (172). The ability of NSAIDs to have biological activity independent of PGs was clearly demonstrated with the drug, flurbiprofen. As a chiral compound, it can exist in the R and S enantiomeric forms. The S enantiomer was shown to be 500 fold more potent than the R form in inhibiting COX-1 and COX-2. However, when both forms were tested for anti-proliferative activity in the colon, they were equally potent. Although the R form has not been tested in a carcinogenesis model, it has been suggested that it could a be useful chemopreventive agent that is nonulcerogenic and does not interfere with normal PG production (4). 5.3. Carcinogen Metabolism Besides metabolizing arachidonic acid, the peroxidase component of PG synthetase can, by a co-oxidation reaction, oxidize a wide range of xenobiotics, including several classes of carcinogens (5, 173, 174). Among these, heterocyclic and aromatic amines and dihydrodiol derivatives of polycyclic hydrocarbons are activated to mutagenic derivatives by PG synthetase (173, 175). These oxidations are inhibited by aspirin by preventing cyclooxygenase-catalyzed generation of the hydroperoxide substrate for the peroxidase. Notable examples of this extrahepatic metabolism include the urinary tract carcinogens benzidine and the nitrofuran derivative (N-[4,5-nitro-2-furyl]-2-thiazole) formamide (FANFT), both of which are activated by renal COX (176). FANFT-induced bladder tumors are prevented by administration of aspirin (177). With regards to the colon, COX can activate the heterocyclic aromatic amine IQ, found in food, to mutagens and the NSAID indomethacin blocks both activation and tumors induced by IQ (178). Thus, inhibiting COX may have a direct preventive effect on colon cancer which is independent of the level of PGE₂. Prostaglandins may participate in carcinogenesis via the generation of peroxyl radicals. Peroxyl radicals are amongst the more stable oxy radicals, which also allows for greater diffusion from the site of formation. Peroxyl radicals can, for example, cause epoxidation of the procarcinogen benzo[a]pyrene 7,8-diol to its ultimate carcinogenic form, a diol epoxide. Peroxyl radicals can also epoxidize other carcinogens, such as 3,4-dihydroxy-3,4dihydrobenzo[a]anthracene and aflatoxin B1 to their ultimate carcinogenic forms. Because differences in the relative amounts of DNA adducts generated from peroxyl radicals or cytochrome P450 can be measured, the relative contribution of COX activity can be determined. In mouse skin, peroxyl radicals play a significant role in the metabolism of benzo[a]pyrene 7,8-diol (179). This is likely to be true for a number of carcinogens and in other tissues as well. Prostaglandins may also contribute more directly to increased DNA damage and mutations via the non-enyzmatic breakdown of PGH to malondialdehyde. Thromboxane sythase also carries out this conversion enzymatically (5). MDA binds to DNA and produces a diversity of base-pair substitutions and frameshift mutations which has led to the suggestion that it may be an important contributor to endogenous mutagenesis (180). 5.4. Immune Modulation High levels of PGE₂ have been shown to suppress immune surveillance (181-183) and to impair killing of malignant cells (184, 185). These effects appear to be specific to PGE₂ since other eicosanoids have not been shown to have a clear role in the regulation of cellular and humoral immune responses (186). In support of a suppressive effect of PGE₂ is the observation that drugs that inhibit PG synthesis enhance immune responses (184-186). PGE₂ can regulate immune function by acting as a negative feedback inhibitor for such processes as T cell proliferation, lymphokine production and cytotoxicity and for macrophage and natural killer cell cytotoxicity (183, 187). Since the growth of various tumors is associated with immune suppression in animals and humans (181, 188), inhibition of COX may reduce tumorigenesis. This tumor-associated immune suppression may occur as a result of factors such as colony stimulating factor which is released by tumor cells and causes monocytes and macrophages to synthesize PGE₂; this elevated PGE₂ inhibits blastogenesis of T cells and the cytotoxic activity of natural killer cells (188, 189). Examples include the use of indomethacin to reduce the size of tibial bone tumors in maloney sarcoma virus infected mice (190). Indomethacin was also shown to augment the impact of BCG treatment in mice with 3-methylcholanthrene induced fibrosarcomas, which produce large amounts of PGs (191). Conversely, administration of PGE₂ and PGF_{2 alpha}to syngeneic mice bearing transplanted squamous cell carcinomas enhanced the transplantability of these tumors (192). Although the growth of some transplanted tumors is reduced by COX inhibitors, when given soon after transplantation, the inhibitory effect is lost over time. At this point, macrophages have been shown to lose their ability to synthesize PGE₂ and their immunosuppressive activity (193). A second phase of immunosuppression ensues that results not from PG production but from enhanced production of bone-marrow derived monocyte-like suppressor cells (193, 194). Thus, while COX inhibitors may be good agents to inhibit initial tumor development, do not seem to be effective chemotherapeutic agents against existing large tumor burdens.

[Frontiers in Bioscience 2, d482-500, October 1, 1997]
Reprints
PubMed
CAVEAT LECTOR

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

5. MECHANISMS OF PROSTAGLANDIN ACTION

One of the primary approaches to the understanding of the function of PGs in any tissue has been through the use of inhibitors of COX activity. The NSAIDs that have most commonly been used in in vitro and in vivo studies include drugs such as aspirin, indomethacin, sulindac, ibuprofen and piroxicam, all of which inhibit PG synthesis albeit by different mechanisms. Aspirin inhibits the COX but not peroxidase activities of both COX-1 and COX-2 by acetylating a particular serine and thus blocking the channel that leads to the active site. This results in irreversible inhibition of PG synthesis although a concurrent increase in 15(R)-HETE occurs, particularly with COX-2 (127,128). Indomethacin forms a tight, slowly dissociable complex with COX that induces an inhibitory conformational change (129). Ibuprofen and piroxicam, on the other hand, compete with arachidonic acid for the active site (130, 131). In general, most of these classical NSAIDs are better inhibitors of COX-1 than COX-2, although some, like flurbiprofen and ibuprofen, have nearly equal IC₅₀ values (132).

The recent demonstration that COX-2 is the inducible isoform and that its high expression occurs at sites of inflammation has spurred the development of isoform specific inhibitors. An additional impetus is the observation that inhibition of PG production, via inhibition of both COX-1 and COX-2, can lead to gastrointestinal lesions and nephrotoxicity. Thus, it was viewed as desirable to reduce the inflammatory effects of COX-2 while retaining the cytoprotective effects of COX-1 (133). As a result, a variety of COX-2 specific inhibitors have recently become available for testing One of the first inhibitors, NS-398, has an IC₅₀ of 3.8 x 10^-6 M in vitro for COX-2 and exerts no effect on COX-1 at 10^-4M (134). In vivo, NS-398 blocked COX-2 expression in inflammatory cells induced by exogenous stimuli and completely inhibited PG synthesis from these cells whereas it did not not affect PG production from COX-1 in the stomach (133). NS-398 and another COX-2 selective inhibitor, DuP697, cause conformational changes in COX-2 that lead to irreversible loss of activity (135). CGP28238 and L-745,337 also exhibit a dramatic COX-2 selectivity when tested on several cell types and in in vivo models of inflammation (136, 137). Together, these pharmaceutical studies strongly support the hypothesis that COX-1 is utilized for cytoprotection and that COX-2 primarily plays a pro-inflammatory role. However, the recent generation of knockout mice for COX-1 and COX-2 suggest that this view may be too simplistic. Contrary to the prediction that COX-1 knockouts would have spontaneous gastrointestinal ulcerations, this pathology was not found. Furthermore, these mice were less sensitive than normal mice to the ulcerating effect of indomethacin (138). The COX-2 knockout mice also produced unexpected results. In these animals topical TPA treatment caused edema to the same extent as in wild type mice (139). In keeping with high PG levels, COX-2 knockouts, however, did show a reduced endotoxin-induced hepatocellular cytotoxicity (140). It is clear from these animal models that, to a large extent, the tissue specific function of COX-1 and COX-2 remains undefined.

5.1. Cell Proliferation

Prostaglandins have many biological effects on different tissues and thus act as endogenous biological modifiers. As mentioned above, specific receptors have been identified for each of the major PGs, as well as for prostacyclin and thromboxane (47), which are coupled to different signal transduction pathways (47, 53). The rapid metabolic breakdown in the circulation suggests that PGs are not classical hormones but act as autocrine or paracrine factors within a given tissue (141). In spite of extensive research on the effects of specific eicosanoids, little is known about their mechanism(s) of action at the cellular and molecular level. For most tissues, it is also not clear whether the effect on a given tissue of a particular PG is the same for normal and malignant cells.

For some, but not all cells, a link has been established between synthesis of PGs and control of cell growth. In Balb/c 3T3 cells, epidermal growth factor-dependent proliferation is inhibited by the COX inhibitor, indomethacin (142). In these cells, PGF_2alpha is not a mitogen by itself but acts a permissive factor that allows the mitogenic action of a growth factor. PGF_2alpha stimulates proliferation of osteoblast, MC3T3-E1, cells by increasing the number of high affinity binding sites for insulin-like growth factor-1 (143). Thromboxane A₂ has also been shown to stimulate mammary epithelial cell growth (144).

Normal murine epidermal cell proliferation does not appear to depend on PGs. This concept is based on the inability of PGE₂ or PGF_2alpha or their more stable, 15(S)-15-methyl, derivatives to stimulate DNA synthesis and the lack of inhibition of normal proliferation by the PG synthetase inhibitor, indomethacin (145-147). In human keratinocyte cultures, however, PGs do appear to be required for normal cell proliferation. A correlation was observed between PGE₂ production and cell proliferation and this proliferation was inhibited by indomethacin in a manner that could be overcome by addition of PGE₂ but not PGF_2alpha (148). In rat skin, topical administration of PGE₁, PGE₂ and PGF_2alpha produced marked increases in DNA, RNA and protein synthesis, further suggesting that PGs are regulatory factors in epidermal cells (149). Unlike in normal mouse skin, in vivo topical application of TPA (and other irritants) results in a proliferation that is dependent on specific PGs. Indomethacin was shown to inhibit TPA-induced epidermal hyperproliferation and this inhibition could be overcome by topical application of PGE₂ but not PGF_2alpha (145, 146, 150). In addition, although not mitogenic by itself, PGE₂ is co-mitogenic when applied with TPA (151).

With regards to colon, two transformed human colon cancer cell lines were recently evaluated. The in vivo growth and in vitro colony formation of one cell line, HCA-7, which constitutively expresses high levels of COX-2 protein was inhibited by the selective COX-2 inhibitor, SC-58125. The other cell line, HCT-116, however, does not express COX-2 protein and is not growth inhibited by SC-58125. This study suggests that SC-58125 is acting in a specific manner rather than via nonspecific cytostatic mechanisms and that PGs contribute to the growth of some, but not all, colon cancers (152). It was also noted that the responding HCA-7 cells are more differentiated than the HCT-116 cells, which may be an important determinant (152). The response of two human colon adenocarcinoma cell lines, SW111 and HT-29, to the proliferative or apoptotic effects of a series of eicosanoids was compared. While HT-29 cells showed increased proliferation in response to PGE₂ and leukotriene B₄, SW111 cells also responded to PGI₂ and PGF_2alpha, although neither distribution of cells in the cell cycle nor rate of spontaneous apoptosis was altered (153). On the other hand, the NSAID, sulindac, reduced the proliferation rate of HT-29 cells and increased the rate of apoptosis (154). Other NSAIDs such as aspirin, indomethacin and piroxicam also reduced cell cycling and all except aspirin induced apoptosis in these cell lines (155). In addition, another study suggests that COX-2 participates in the proliferation of colon cancer cells. Several gastrointestinal cancer cell lines that highly expressed COX-2 were inhibited by both COX-2 selective and nonselective inhibitors whereas those cell lines with low COX-2 expression showed minimal response (156).

The growth stimulating effects of PGs in many cases appear to be linked to other biological modifiers, particularly the polyamines. Elevated polyamines are associated with an increase in DNA synthesis and result from induced ornithine decarboxylase (ODC) activity. Tumor promoters, both exogenous and endogenous (bile salts) induce the synthesis of ODC in a PGE₂-dependent process, based on inhibition by indomethacin and restoration by exogenous PGE₂ (157). In this respect, the colonic epithelium responds in a manner similar to murine epidermis, a tissue in which tumor promoter-induction of ODC is also PGE₂ dependent (71). In the skin multistage carcinogenesis model, ODC induction is required for tumor development; inhibitors of arachidonic acid metabolism inhibit both ODC induction and tumor development (158). Therefore, it seems that a clear link exists between PGE₂, enhanced proliferation and tumor development.

In addition to effects on proliferation, PGs are associated with enhanced invasive and metastatic potentials of tumor cells. Recently, several studies have reported that increasing the expression of COX-2 was associated with alterations in cell behavior. When rat intestinal epithelial cells were transfected with a COX-2 expression vector, they showed enhanced cell adhesion due to upregulated E-cadherin expression; this was reversed by addition of sulindac sulfide (159). When human colon cancer cell line, Caco-2, was transfected with a COX-2 vector, they exhibited an enhanced invasive phenotype that was associated with an elevated expression of membrane metalloproteinase. Sulindac sulfoxide reversed both the increased invasiveness and PG synthesis (156).

5.2. Role in Apoptosis

The biochemical basis for the antineoplastic characteristics of NSAIDs has generally been attributed to their ability to reduce PG levels by inhibiting COX. However, this category of drugs may have other effects that are responsible for, or at least contribute to, their anti-cancer properties. Studies using cultured tumor cell lines have shown that a variety of NSAIDs can induce apoptosis (160-164). Apoptosis, frequently referred to as programmed cell death, is a specific morphological and biochemical form of cell suicide. The apoptotic pathway can be triggered by extracellular agents, pathological processes and also occurs during normal development and tissue remodeling. Apoptosis differs from necrosis in that the apoptotic cell kills itself in manner that does not harm neighboring cells nor elicits an inflammatory response. Activation of apoptosis in tumors has thus been a target for some chemotherapies (165).

Although NSAIDs can both inhibit PG synthesis and induce apoptosis, there is recent evidence that the apoptotic effects is likely not due to COX inhibition. The most striking example comes from studies with sulindac and its metabolites. Sulindac sulfoxide is consumed as a prodrug that has no COX inhibitory activity. It is readily reduced to the sulfide form in the liver and in the colon via bacterial microflora; it is this sulfide form that is responsible for the anti-inflammatory characteristics of sulindac (166, 167). However, sulindac sulfoxide is also oxidized in an irreversible manner in the liver to a sulfone that is excreted in the bile and intestine. The sulfone metabolite is devoid of COX inhibitory or other anti-inflammatory activity (168) although it still retains the ability to inhibit tumor cell growth and induce apoptosis (161). Thompson et al (168) were the first to report that the sulfone metabolite has cancer chemoprotective activity. In a chemically induced rat mammary cancer model, sulindac sulfone was found to be comparable to sulindac sulfoxide in reducing tumor incidence and numbers and in inducing apoptosis (168). Sulindac sulfone also inhibits formation of aberrant crypts in a rodent colon carcinogenesis model and recently was shown to inhibit azoxymethane-induced colon carcinogenesis in rats (160, 169). In a recent study, it was also shown that in addition to having no COX-inhibitory activity, the sulfone metabolite does not inhibit lipoxygenases or phospholipase A₂ (170).

The relationship between NSAIDs, PGs and apoptosis was further explored in the HT-29 colon adenocarcinoma cells which expresses COX and synthesizes PGs and in the HT-15 carcinoma cell line which does not exhibit these properties. Addition of PGE₂ to HT-15 cells enhanced proliferation only slightly (25%); addition of sulindac or piroxicam severely reduced proliferation and this could not be reversed by addition of PGs. The PG producing HT-29 cells were also growth inhibited by sulindac and piroxicam but unlike similar experiments described above for indomethacin treatment of keratinocytes, growth could not be restored by addition of PGs. The conclusion drawn is that these NSAIDs reduce proliferation in human colonic cells through a PG-independent mechanism (171).

These and related studies have raised the question of exactly how the NSAIDs exert their anti-proliferative effects. As reviewed by Abramson and Weissmann (172), NSAIDs affect a variety of membrane processes not related to PG synthesis. These include inhibition of superoxide anion generation by NADPH oxidase, phospholipase C and 12-HPETE peroxidase and coupling of mitochondrial oxidative phosphorylation (172). The ability of NSAIDs to have biological activity independent of PGs was clearly demonstrated with the drug, flurbiprofen. As a chiral compound, it can exist in the R and S enantiomeric forms. The S enantiomer was shown to be 500 fold more potent than the R form in inhibiting COX-1 and COX-2. However, when both forms were tested for anti-proliferative activity in the colon, they were equally potent. Although the R form has not been tested in a carcinogenesis model, it has been suggested that it could a be useful chemopreventive agent that is nonulcerogenic and does not interfere with normal PG production (4).

5.3. Carcinogen Metabolism

Besides metabolizing arachidonic acid, the peroxidase component of PG synthetase can, by a co-oxidation reaction, oxidize a wide range of xenobiotics, including several classes of carcinogens (5, 173, 174). Among these, heterocyclic and aromatic amines and dihydrodiol derivatives of polycyclic hydrocarbons are activated to mutagenic derivatives by PG synthetase (173, 175). These oxidations are inhibited by aspirin by preventing cyclooxygenase-catalyzed generation of the hydroperoxide substrate for the peroxidase. Notable examples of this extrahepatic metabolism include the urinary tract carcinogens benzidine and the nitrofuran derivative (N-[4,5-nitro-2-furyl]-2-thiazole) formamide (FANFT), both of which are activated by renal COX (176). FANFT-induced bladder tumors are prevented by administration of aspirin (177). With regards to the colon, COX can activate the heterocyclic aromatic amine IQ, found in food, to mutagens and the NSAID indomethacin blocks both activation and tumors induced by IQ (178). Thus, inhibiting COX may have a direct preventive effect on colon cancer which is independent of the level of PGE₂.

Prostaglandins may participate in carcinogenesis via the generation of peroxyl radicals. Peroxyl radicals are amongst the more stable oxy radicals, which also allows for greater diffusion from the site of formation. Peroxyl radicals can, for example, cause epoxidation of the procarcinogen benzo[a]pyrene 7,8-diol to its ultimate carcinogenic form, a diol epoxide. Peroxyl radicals can also epoxidize other carcinogens, such as 3,4-dihydroxy-3,4dihydrobenzo[a]anthracene and aflatoxin B1 to their ultimate carcinogenic forms. Because differences in the relative amounts of DNA adducts generated from peroxyl radicals or cytochrome P450 can be measured, the relative contribution of COX activity can be determined. In mouse skin, peroxyl radicals play a significant role in the metabolism of benzo[a]pyrene 7,8-diol (179). This is likely to be true for a number of carcinogens and in other tissues as well.

Prostaglandins may also contribute more directly to increased DNA damage and mutations via the non-enyzmatic breakdown of PGH to malondialdehyde. Thromboxane sythase also carries out this conversion enzymatically (5). MDA binds to DNA and produces a diversity of base-pair substitutions and frameshift mutations which has led to the suggestion that it may be an important contributor to endogenous mutagenesis (180).

5.4. Immune Modulation

High levels of PGE₂ have been shown to suppress immune surveillance (181-183) and to impair killing of malignant cells (184, 185). These effects appear to be specific to PGE₂ since other eicosanoids have not been shown to have a clear role in the regulation of cellular and humoral immune responses (186). In support of a suppressive effect of PGE₂ is the observation that drugs that inhibit PG synthesis enhance immune responses (184-186). PGE₂ can regulate immune function by acting as a negative feedback inhibitor for such processes as T cell proliferation, lymphokine production and cytotoxicity and for macrophage and natural killer cell cytotoxicity (183, 187). Since the growth of various tumors is associated with immune suppression in animals and humans (181, 188), inhibition of COX may reduce tumorigenesis. This tumor-associated immune suppression may occur as a result of factors such as colony stimulating factor which is released by tumor cells and causes monocytes and macrophages to synthesize PGE₂; this elevated PGE₂ inhibits blastogenesis of T cells and the cytotoxic activity of natural killer cells (188, 189). Examples include the use of indomethacin to reduce the size of tibial bone tumors in maloney sarcoma virus infected mice (190). Indomethacin was also shown to augment the impact of BCG treatment in mice with 3-methylcholanthrene induced fibrosarcomas, which produce large amounts of PGs (191). Conversely, administration of PGE₂ and PGF_{2 alpha}to syngeneic mice bearing transplanted squamous cell carcinomas enhanced the transplantability of these tumors (192).

Although the growth of some transplanted tumors is reduced by COX inhibitors, when given soon after transplantation, the inhibitory effect is lost over time. At this point, macrophages have been shown to lose their ability to synthesize PGE₂ and their immunosuppressive activity (193). A second phase of immunosuppression ensues that results not from PG production but from enhanced production of bone-marrow derived monocyte-like suppressor cells (193, 194). Thus, while COX inhibitors may be good agents to inhibit initial tumor development, do not seem to be effective chemotherapeutic agents against existing large tumor burdens.