COX-2 in Cancer--A Player That's Defining the Rules

Cyclooxygenase (COX) inhibitors are being developed as potential agents for the prevention and treatment of cancer after a century of widespread use for inflammation, fever, and pain. Beginning in the late 1970s, researchers noted elevated concentrations of prostaglandins in neoplastic lesions, which suggested a role for arachidonic acid metabolites in tumorigenesis. Multiple lines of evidence—in vitro, in vivo, observational, and clinical—now confirm that COX inhibitors reduce prostaglandin production and the risk of colorectal, skin, and other neoplasias (1). Owing to gastrointestinal safety concerns with traditional nonselective COX inhibitors, derivatives that selectively target COX-2 have been developed for applications in arthritis, analgesia, and the treatment of neoplasia. COX-2 selective inhibitors serve as a paradigm of molecularly targeted, cytostatic, antineoplastic agents (2). COX-2 is consistently overexpressed in a large percentage and variety of human and rodent tumors (3). Pathogenic relevance to in vivo carcinogenesis was demonstrated via genetic manipulations that either eliminated (4) or induced (5) COX-2 expression, resulting in substantial tumor reduction or stimulation, respectively. Food and Drug Administration (FDA)-approved selective COX-2 inhibitors, such as celecoxib (Celebrex™; Pharmacia, Peapack, NJ) or rofecoxib (Vioxx™; Merck, Whitehouse Station, NJ), effectively modulate inflammation and pain (6). Moreover, celecoxib has recently been shown to reduce the colorectal adenoma burden in high-risk patients (7). Cancer researchers, however, have reported COX-independent effects that are dose dependent, suggesting that mechanisms other than COX suppression alone may account for the observed efficacy of these agents. At the cellular level, COX inhibitors have been shown to inhibit proliferation, induce apoptosis, inhibit angiogenesis, reduce carcinogen activation, and stimulate the immune system (3,8). Although reductions in prostanoid concentrations may underlie these observed activities, non-COX targets and mechanisms may be involved as well. For example, nonselective COX inhibitors have been shown to modulate levels of cGMP (guanosine 3`,5`-cyclic monophosphate), NF-κB (nuclear factor-κB) activation, Bcl expression, and the binding of peroxisome proliferator-activated receptors (PPARs). Support for non-COX hypotheses is tempered by the fact that antineoplastic effectiveness in vitro has most often been described for concentrations of COX inhibitors typically exceeding those that might be clinically attainable (i.e., <5 μM). In this issue of the Journal, Song et al. (9) report a series of elegant studies with COX-2 antisense clones and structurally modified derivatives of celecoxib that support the existence of functionally distinct COX-2 inhibitory and proapoptotic effects. These results confirm previous reports demonstrating that COX-independent effects of celecoxib (50 μM) stimulate apoptosis in cells lacking either COX-1 or COX-2 (10). In the first set of experiments, they demonstrated that COX-2-deficient cells were viable yet underwent programmed cell death when exposed to various COX-2 inhibitors at concentrations of 50–100 μM. In the second set of studies, structural modifications to celecoxib yielded derivatives with distinct apoptotic properties when compared with the parent compound and each other. Subsequent structure–activity analyses found no association between the COX-2 inhibitory and proapoptotic activities of these celecoxib derivatives. The findings of Song et al. are highly provocative. First, for reasons yet unknown, the proapoptotic activity of celecoxib surpassed that of most of the other agents among the panel of COX-2 inhibitors tested. Currently, most publicly sponsored chemoprevention trials in this area are testing celecoxib, a circumstance that commends the foresight of its discoverers. If celecoxib proves to be the most active among a growing field of COX-2 inhibitors, the fact that it is already being tested in advanced clinical trials against a variety of epithelial malignancies (e.g., colon, esophagus, skin, and bladder cancers) may have enormous impact on the rate at which the true potential of this class of agents will be definitively assessed. Second, the authors' findings provide direction for further engineering derivatives of selective COX-2 inhibitors for anticancer applications. Celecoxib has already demonstrated efficacy in the reduction of colorectal adenomas, and newer generations of COX-2 inhibitors may achieve improved efficacy and safety. Needless to say, novel COX-2 derivatives will require intense evaluation to verify that apoptosis is preferentially induced in neoplastic rather than normal tissue, thereby ensuring a favorable therapeutic index. To resolve uncertainties regarding the relative contributions of COX-dependent and -independent mechanisms to antineoplastic activity, the National Cancer Institute convened a meeting of experts in 2001 (11). Although there was general consensus that COX-2 inhibition contributes to the anticancer activities of COX inhibitors, the role of COX-1 remained controversial. Two lines of evidence suggest that COX-1 may also play a role in neoplasia. First, knocking out COX-1 in a murine model of intestinal tumorigenesis led to an 80% reduction in intestinal tumor burden (12). Second, and perhaps more compelling, are in vivo and epidemiologic data demonstrating the efficacy of COX inhibitors that preferentially inhibit COX-1 (e.g., aspirin or piroxicam) in preventing cancer (2). In addition, several of the assembled experts argued that non-COX mechanisms of action likely contribute to the chemopreventive efficacy of celecoxib and other COX inhibitors. Because most of the supporting studies reported effects at drug concentrations only attainable in vitro (e.g., 50 μM or greater), additional research on the topic was recommended. Data published in this issue of the Journal are likely to fuel the COX-2 debate. Using relatively high concentrations of celecoxib (50 μM), the authors describe COX-2 independent activity. However, the central issue is whether comparable effects occur at celecoxib concentrations attainable in humans, i.e., plasma concentrations approaching 0.5–5 μM. This poses new challenges to the development of COX-2 inhibitors as anticancer agents, given that most cancer prevention and treatment trials have selected doses extrapolated from COX-2 inhibition achieved in inflammatory conditions, rather than through rigorous MTD (maximum tolerated dose) evaluations and efficacy against neoplastic biomarkers. Indeed, to date all clinical data on the value of COX-2 inhibitors for cancer prevention suggest that higher doses may be required for antineoplastic than for anti-inflammatory indications (7). This study underscores the biologic complexity of COX inhibition and the promiscuous activity of COX inhibitors in various contexts (e.g., inflammation, cancer treatment, and cancer prevention). Nevertheless, before COX-independent effects are definitively embraced, it is important to demonstrate the magnitude of antineoplastic effects achievable at clinically relevant doses, including back validation in relevant rodent models. For example, it would be worthwhile to determine whether selective COX-2 inhibitors reduce the intestinal tumor burden of Min mice lacking a functional COX-2 gene. Another issue concerns the potential limitations of using cell culture models to fully understand the proapoptotic effects of selective COX-2 inhibitors. For example, these agents have potent antiangiogenic effects (3,13). If selective COX-2 inhibitors such as celecoxib induce apoptosis in epithelial cells as a consequence of inhibiting angiogenesis, studies using cultured epithelial cells may yield results of unclear significance. Definitive data on the efficacy of celecoxib and rofecoxib in the contexts of cancer prevention and therapy will mature within the next few years. In addition, as the impact of COX-independent pathways is better defined, synergistic efficacy via serial or parallel inhibition becomes an increasingly achievable goal. Indeed, the chemopreventive potential of COX inhibitors administered with other agents was demonstrated 15 years ago with difluoromethylornithine and, subsequently, with other classes of agents (14–17). Two key trials are now testing this concept in persons at risk for colorectal neoplasia, the results of which will prove whether combinatorial strategies—such as those that have had enormous impact in other medical disciplines—succeed in cancer prevention as well.