Self-Promotion? Intimate Connections Between APC and Prostaglandin H Synthase-2

have high level expression of COX-2 (Eberhart et al., Colon cancer is a paradigm for the multistep genetic 1994; Sano et al., 1995; Kutchera et al., 1996). The work basis of cancer. Several initiating mutations have been reported by Oshima et al. (1996) clearly demonstrates identified by studying families with inherited predisposithat COX-2 lies on the standard route from mutations tion to colon cancer and its long-lived intermediate, the in APC to colon tumors. adenomatous polyp. For example, in Familial AdenomaIn the APC mouse created by Oshima et al. (1995) tous Polyposis the affected individuals develop hunand in the Min mouse, early adenomas have lost both dreds of polyps, some of which progress to cancer. copies of APC and overexpress COX-2 (Oshima et al., The FAP gene was mapped to chromosome 5q21 and 1996; Dove et al., 1995). Are the two causally related? mutations in the APC gene were identified as the cause That is, does the rise in COX-2 require that there be (reviewed by Kinzler and Vogelstein, 1996). Insight into loss of function at both alleles of APC, or could COX-2 how these mutations result in cancer has been accelerexpression occur when the cells are still heterozygous? ated by the discovery of the Min mouse, which has a The upstream boundary for COX-2 expression (see Figmutated APC gene and a similar syndrome (reviewed ure 1) was set in experiments that examined adenomas by Dove et al., 1995). Studies of the biochemical mechaand normal tissue from APC1/2 mice. Two groups found nisms downstream of APC mutations have turned up that there was no COX-2 in the uninvolved colon, but important leads (Kinzler and Vogelstein, 1996); the overthat it was present in the tumor (Williams et al., 1996; expression of prostaglandin H synthase-2 (COX-2) is Oshima et al., 1996). Earlier, Oshima et al. (1995) showed one of the most significant since it has immediate implithat even small adenomas showed loss of heterozygoscations for prevention of colon cancer. ity at the normal APC allele. Taken together, these findA paper in this issue of Cell (Oshima et al., 1996) ings tentatively place the time of COX-2 induction after provides key evidence on the role of COX-2 in colon the loss of the second APC allele. In support of this, carcinogenesis. They bred mice carrying an APC mutaOshima et al. (1996) did not detect COX-2 protein in tion with a mouse with a disrupted COX-2 gene. All of polyps smaller than 2 mm, which would indicate that the animals were APC; if homozygous for wild type the tumor has been initiated before COX-2 is expressed. COX-2, they developed an average of 652 polyps at 10 However, Boolbol et al. (1996) obtained a directly conweeks, while heterozygotes had 224 polyps and homoflicting result—that the uninvolved epithelium in Min zygously deficient mice had only 93 polyps. This experimice did have increased COX-2, which will need to be ment provided definitive genetic evidence that induction assessed further in mice and humans. We conclude that of COX-2 is an early, rate-limiting step for adenoma the current evidence supports this sequence of events: formation. As supporting evidence, a drug that inhibits loss of both APC alleles, early polyp formation, expresCOX-2 but not COX-1 also markedly reduced the numsion of COX-2, polyp growth, and additional mutations ber of polyps. Thus, a mouse model again has proven to reach an invasive tumor (Figure 1). Thus, COX-2 exvaluable for measuring a quantitative genetic effect and pression would be past the boundary of tumor establishPtgs2 (the COX-2 gene in mouse terminology) can be ment (Dove et al., 1995) and would provide a mechanism added to the list of genes involved in colon neoplasia for self-promotion by the early adenoma. (Dove et al., 1995). The events between loss of the second APC allele and Overexpression of COX-2 Is an Early, Central Event COX-2 expression are unknown. One likely possibility is in Carcinogenesis that transcription of COX-2, which does not occur conA connection between prostaglandins and colon carcistitutively, is switched on. We found that a colon cancer nogenesis has been recognized for some time; epidecell line that expresses COX-2 also transcribes a COX-2 miological studies have shown that chronic intake of reporter gene, which is consistent with this hypothesis nonsteroidal anti-inflammatory drugs (NSAIDs), which (Kutchera et al., 1996). There is substantial information function as COX inhibitors, reduce colon cancer and about the transcriptional regulation of COX-2 (Herschpolyps by as much as 40%. NSAIDs prolong the lag man, 1994), but it is not clear how loss of APC function phase prior to the appearance of colon cancer induced would activate these pathways. Another mechanism is by chemical carcinogens in animals and decrease the suggested by recent studies of APC function: APC binds number of polyps in Min mice (Jacoby et al., 1996). to b-catenin and it, in turn, binds and regulates the Finally, the size and number of polyps in FAP patients transcription factor LEF-1 (Kinzler and Vogelstein, 1996). decrease markedly in response to the NSAID sulindac LEF-1 is not known to regulate COX-2 expression, but