Pathologically and biologically distinct types of epithelium in intraductal papillary mucinous neoplasms: delineation of an "intestinal" pathway of carcinogenesis in the pancreas.

Although general characteristics of intraductal papillary mucinous neoplasms (IPMNs) and their delineation from other pancreatic tumors have been well established, several issues regarding their biology and management remain unresolved. It has been noted briefly by us and other authors that there are different types of papillae in IPMNs; however, their frequency, biologic significance, and clinical relevance are unknown. In this study, the association of different papillary patterns with clinical, pathologic, and biologic parameters was studied in 74 IPMNs, and the expression profile of CDX2 (a specific marker and one of the key determinants of intestinal "programming," and a tumor suppressor) was determined immunohistochemically in addition to MUC1 (a marker of an "aggressive" phenotype in pancreatic neoplasia) and MUC2 ("intestinal type mucin," a marker of the "indolent" phenotype, and a tumor suppressor). The patterns of papillae identified and their association with these parameters were as follows: 1) The intestinal-type (Yonezawa's dark-cell type), similar to villous adenomas, was seen in 26 of 74 (35%) cases. The majority harbored carcinoma in situ (85%) or borderline atypia (15%). They tended to be large (mean, 5.5 cm). Most expressed CDX2 (95%) and MUC2 (92%) but not MUC1 (8%). This type was more commonly associated with colloid-type invasion (14 of 16 invasive carcinomas were of colloid type). 2) The pancreatobiliary type, characterized by arborizing papillae lined by cuboidal cells resembling papillary neoplasms of the biliary tract, was present in 22% of the cases. These were mostly graded as carcinoma in situ (94%); they rarely expressed CDX2 (6%) or MUC2 (19%) but often showed MUC1 labeling (44%). This pattern was more commonly associated with the tubular type of invasive carcinoma and had a slight tendency for a more aggressive clinical course. 3) The null type was characterized by abundant apical mucin and basally located nuclei, similar to the gastric foveolar epithelium. Thirty-one percent of IPMNs had this type of papillae, but this pattern was also present in the background of other IPMNs and in the cystic components of most cases as well. Most pure null-type IPMNs were devoid of complexity and consequently classified as adenoma (48%). They tended to be small (mean, 2.6 cm), were often negative for CDX2, MUC1, and MUC2, and were rarely associated with invasive carcinoma. 4) Some IPMNs (12%) exhibited features that were difficult to classify, and 2 cases had a mixture of pancreatobiliary and intestinal types of papillae. In conclusion, IPMNs include pathologically and biologically distinct epithelial patterns. CDX2 and MUC2 expression is relatively specific for the intestinal type papillae, confirming that these IPMNs indeed exhibit intestinal differentiation. Their close association with colloid carcinoma, which also shows consistent MUC2 and CDX2 expression, supports the existence of an intestinal pathway of carcinogenesis. This "metaplastic" pathway may reflect different genetic events in the development of these IPMNs, and the presence of intestinal differentiation may potentially be used in prognostication and stratification of patients into appropriate treatment categories.