Section 5 -

Molecular Alterations in Non-Endometrioid Carcinomas of the Endometrium C. Blake Gilks Vancouver General Hospital and University of British Columbia Vancouver BC, Canada

The Simple Version There are two distinct pathways of endometrial carcinogenesis, leading to clinically and pathologically distinct subtypes of endometrial carcinoma. The first pathway (Type-1 cancers) is associated with hyperestrogenic states, leading to the development of atypical endometrial hyperplasia, which progresses to endometrioid carcinoma. The second pathway occurs independent of the growth promoting effects of estrogen, with the earliest recognisable pathological alteration being endometrial intra-epithelial carcinoma, which progresses to uterine serous carcinoma. The following table highlights the important differences between type-1 and type-2 cancers. TABLE 1. Endometrioid (Type-1) versus Non-Endometrioid (Type-2) Endometrial Carcinomas Endometrioid Non-Endometrioid Hyperestrinism/ER expression + - Complex atypical hyperplasia + - p53 mutation/loss - + PTEN mutation/loss + - Microsatellite instability +/- - The More Complex Version While the epidemiological data supports there being two distinct pathways of endometrial carcinogenesis, the situation in clinical practice is more complex, and much more interesting. Hendrickson et al. observed that the designations of type-1 and type-2 carcinoma describe "two loose clinicopathologic clusters or syndromes" [1] rather than reproducibly recognizable groups. The designations of type-1 and type-2, indicating tumours arising as a result of estrogenic stimulation, and independent of estrogen, respectively, have never come into routine use diagnostically; although they can be readily applied to many cases of endometrial carcinoma, a significant number of cases defy classification as either type-1 or type-2 carcinoma. Examples of such cases include uterine serous carcinoma associated with complex atypical hyperplasia, or mixed endometrioid and serous carcinoma. Therefore, these designations are useful in a general sense in understanding endometrial carcinogenesis but are an oversimplification and not clinically relevant for anatomical pathologists in routine day-to-day practice. In contrast, designation of the histological subtype of an endometrial carcinoma is routinely done in clinical practice and this cell-type assignment is used to guide patient therapy. Accordingly, in this presentation the focus will be on molecular abnormalities in the clinically important group of non-endometrioid carcinomas, rather than a discussion of general pathways of endometrial carcinogenesis. Before discussing these molecular abnormalities we will first review background information of relevance to surgical pathologists. The reproducibility of histopathological assessment of endometrioid vs non-endometrioid cell type. Meaningful studies on molecular abnormalities in subtypes of endometrial carcinoma can only be done if we, as diagnostic pathologists, can accurately diagnose those subtypes. Unless we can reproducibly recognise uterine serous and clear cell carcinoma, there cannot be progress in understanding the molecular events of importance during oncogenesis, or rational introduction of new treatments. We recently examined interobserver reproducibility in assignment of cell-type of endometrial carcinomas, based on three reviewers examining a single representative slide from fifty consecutive cases [2]. There was substantial but not perfect interobserver reproducibility in assignment of cell-type (kappa = 0.70). This would undoubtedly be lower if only high-grade tumours were considered, as in common, low-grade endometrial adenocarcinomas of endometrioid-type distinction from uterine serous carcinoma is not an issue. It is only within the smaller subset of higher grade tumours where this becomes problematic and although this group of tumours has not been examined specifically, it is likely that the reproducibility in assignment of cell-type in high grade endometrial cancer is suboptimal As patient management may vary depending on whether we diagnose uterine serous carcinoma or not, efforts to improve reproducibility are desperately needed. In the meantime, for research purposes it is possible to focus on very typical, pure examples of uterine serous or clear cell carcinoma for molecular studies, to minimise the effect of inter-observer variation in diagnosis, but it is important to remember that these cases may not be representative of all cases of uterine serous or clear cell carcinoma encountered in practice, especially mixed tumours. Association of non-endometrioid carcinomas with inherited cancer susceptibility syndromes (HNPCC, BRCA). Our insight into the molecular events underlying a number of different carcinomas has been greatly advanced by understanding the genetic basis of inherited susceptibility to these cancers. For non-endometrioid carcinomas of endometrium there are two relevant cancer susceptibility syndromes. Familial breast and ovarian cancer, associated with mutations in BRCA1 and BRCA2, is associated with little or no increase in the likelihood of developing endometrial carcinoma of either endometrioid or non-endometrioid type. Case reports have documented patients with germ-line mutations in BRCA genes who have developed uterine serous carcinoma, but, unlike serous carcinoma or ovary, tube or peritoneum, uterine or cervical serous carcinomas are not a major contributor to cancer risk in these patients. Although this is perhaps counterintuitive, given the morphological similarities of serous carcinoma of ovary/peritoneum/fallopian tube vs serous carcinoma of endometrium, there are other molecular differences (e.g. frequent expression of WT-1 in the former but not the latter, more frequent Her-2 amplification in the latter than the former) suggesting that there are fundamental differences between these serous tumours arising at different anatomic sites. Endometrial carcinoma is a very significant part of the spectrum of disease associated with Hereditary Non-Polyposis Colon Cancer (HNPCC) syndrome, being the second most common malignancy after colonic adenocarcinoma. Endometrial carcinomas associated with HNPCC are more frequently higher grade, and present with higher advanced stage disease, compared to sporadic cases of endometrial cancer. In a recent study by Carcangiu et al., endometrial carcinomas arising in HNPCC patients were often of non-endometrioid type (46.3 % of vs. 53.6 % of endometrioid type) [3]. Many of these non-endometrioid carcinomas were mixed (e.g. 6 of 11 clear cell carcinomas were mixed clear cell and endometrioid tumours). This indicates that that mismatch repair defects can contribute to development of some non-endometrioid carcinomas (more about this later). Specific genetic abnormalities in non-endometrioid carcinomas P53: Mutations in P53 are found in most non-endometrioid carcinomas, and are the most characteristic genetic abnormality in this set of tumours; the precise percentage of cases with loss of P53 is impossible to state due to differences in case selection and methods of testing for P53 mutations. For surgical pathologists, p53 immunostaining is an inexpensive surrogate for P53 mutational testing; p53 immunoreactivity (defined as strong diffuse nuclear immunoreactivity) is seen in most cases of non-endometrioid carcinoma [4]. It has been suggested that p53 immunoreactivity can be used diagnostically to support a diagnosis of uterine serous carcinoma but this requires correlation with morphology as endometrioid carcinomas also can have mutations of p53, and not all non-endometrioid carcinomas exhibit p53 immunoreactivity. Loss of p53 function leads to genetic instability, with the result that there is widespread loss of heterozygosity/allelic imbalance in non-endometrioid carcinomas with mutant p53, and frequent aneuploidy. PTEN, microsatellite instability (MSI), k-RAS: Mutations in PTEN and k-RAS, while common in endometrioid carcinoma, are rarely encountered in non-endometrioid carcinomas. MSI is more common in endometrioid carcinomas but is also present in occasional cases of non-endometrioid carcinomas (11% in the study of Catasus et al. [5]) and, as noted previously, non-endometrioid carcinomas are encountered with increased frequency in patients with HNPCC. Her-2 amplification: Her-2 amplification was identified in 47% of cases of uterine serous carcinoma by Santin et al. [6], and Her-2 amplification was associated with a significantly worse outcome, compared to those patients whose tumours did not show amplification. This observation is of great potential impact. It may be that gene amplification events are critical in the genesis of uterine serous carcinoma, an important discovery given how little we know about genetic events important in the development of these tumours. Interestingly, Slomovitz et al. did not find this high frequency of Her-2 amplification in their series of cases [7], however they included mixed tumours, while the study of Santin et al. included only pure examples of uterine serous carcinoma. It remains to be seen whether Herceptin therapy will be effective against uterine serous carcinomas with Her-2 amplification; the amplification ratios were relatively low in most cases, and heterogeneity of Her-2 amplification was seen in 4/14 cases. Use of Panels of Immunomarkers in Classification of Endometrial Carcinoma TABLE 2. Immunomarkers in Endometrial Carcinoma Marker % expression p-value Total Grade 1&2 endometrioid Grade 3 endometrioid USC & Clear cell p53 16% (25/155) 2.8% (3/107) 38% (13/34) 64% (9/14) <0.0001 ER 89% (139/156) 97% (106/109) 79% (27/34) 46% (6/13) <0.0001 Her2Neu 4.5% (7/155) 0% (0/107) 3% (1/34) 43% (6/14) <0.0001 E2F-1 24% (37/153) 18% (19/106) 32% (11/34) 54% (7/13) 0.008 Table 2 shows the immunophenotype of different subsets of endometrial adenocarcinoma stained at our centre. It is clear that no single immunomarker can serve as a discriminator between endometrioid and non-endometrioid carcinomas. An obvious next step is to use panels of immunomarkers that each, individually, show differences in staining between endometrioid and non-endometrioid carcinomas. When such data is generated, either supervised or unsupervised analytical approaches can be applied to the data. With supervised approaches the cases are assigned to categories (e.g. endometrioid vs non-endometrioid) based on morphological examination. This approach assumes that our morphological assessment is accurate. In unsupervised analysis, no assumptions are made based on the clinical or pathological features of the cases, and the immunostaining results are considered completely independently. Darvishan et al. used a supervised approach to data analysis to demonstrate that the combination of p53, PR and PTEN were the most useful in distinguishing between uterine serous carcinoma and endometrioid carcinoma [8]. Using unsupervised hierarchical clustering analysis, and a panel of 9 immunomarkers, we were able to demonstrate that three subsets of endometrial cancer are identifiable [9]. There is a large group composed of predominantly low grade endometrioid adenocarcinomas that are characterized by ER +ve, p53 –ve immunophenotype. These are classic type-1 tumors. Another smaller group of tumours, consisting predominantly of uterine serous carcinoma, shows an ER -ve, p53 +ve immunophenotype. These are classic type-2 tumors. The third group is characterised by positivity for ER but also positivity for p53 and/or other oncogenes not typically seen in low-grade endometrioid carcinoma. It is likely that these tumours start off as lower grade, hormonally sensitive tumours, and through accumulation of new mutations become higher grade. I suspect that these cases are under-represented in many of the studies of molecular events during endometrial carcinogenesis, as more "typical" examples of pure endometrioid or non-endometrioid carcinoma are selected for study. This third group would account for the finding of atypical hyperplasia in association with non-endometrioid carcinoma, the mixed endometrioid/non-endometrioid carcinomas in general, and more specifically those arising in patients with HNPCC syndrome, and the differences in frequency of Her-2 amplification in studies with different case characteristics [6, 7]. It also suggests that the definition of mixed adenocarcinoma of the endometrium as "a tumour composed of an admixture of a type-1……and a type-2 carcinoma" [10] should be revised, as mixed tumors are more likely to be type-1 tumors with progression to higher grade malignancy. Selected References: Hendrickson MR, Longacre TA, Kempson RL. The uterine corpus. Sternberg's Diagnostic Surgical Pathology 2004:2460-84. Alkushi A, Abdul-Rahman ZH, Lim P, et al. Am J Surg Pathol, 2005;29:295-304 . Carcangiu ML, Dorji T, Radice P, et al. Mod Pathol 2006;19S1:173A. Alkushi A, Lim P, Coldman A, et al. Int J Gynecol Pathol 2004;23: 129-137. Catasus L, Machin P, Matias-Guiu X, Prat J. Human Pathol 1998;29:1160-64. Santin AD, Bellone S, Van Stedum S, et al. Cancer 2005;104:1391-7. Slomovitz RM, Broaddus RR, Burke TW, et al. J Clin Oncol 2004;22:3126-32. Darvishan F, Hummer AJ, Thaler HT, et al. Am J Surg Pathol 2004;28:1568-78. Alkushi A, Lim P, Coldman A, et al. manuscript submitted. Silverberg SG, Kurman RJ, Nogales F, et al. Epithelial tumours and related lesions. World Health Organization Classification of Tumours: IARC Press, 2003:221-32.