Expression of Cytokeratins 7 and 20 in Ovarian Neoplasia

Transcription

1 Anatomic Pathology / EXPRESSION OF CYTOKERATINS 7 AND 20 IN OVARIAN NEOPLASIA Expression of Cytokeratins 7 and 20 in Ovarian Neoplasia Helen P. Cathro, MBChB, and Mark H. Stoler, MD Key Words: Ovary; Ovarian; Neoplasm; Cytokeratin immunohistochemistry Abstract To further delineate specific staining patterns and refine the differential usefulness of cytokeratin (CK) 7/20 staining, we studied multiple ovarian tumors and primary nongynecologic neoplasms likely to metastasize to the ovary. Immunohistochemical analysis with semiquantitative grading to give quartile scores (0-4) was performed on 127 cases. Subsequent analysis indicated that a more informative diagnostic segregation could be achieved with a biphasic grading system (>50% staining, positive; 50% or less, negative). Lower intestinal tumors were CK7 and usually CK20+, while upper gastrointestinal tumors, including those of pancreatobiliary origin, were mostly CK7+ and CK20. Serous papillary ovarian tumors were all CK7+ and CK20. Mucinous ovarian carcinomas were all CK7+ and slightly more often CK20, whereas the small number of ovarian borderline mucinous tumors studied were the most problematic, with no clear pattern. Multiple different tumor types from all nonovarian gynecologic sites were fairly consistently CK7+ and almost always CK20. Differential CK staining of mucinous tumors of the female genital tract using CK7 and CK20 is useful for predicting the site of origin, provided samples are adequate in size. The most specific usefulness is the identification of lower gastrointestinal vs other neoplasms. Ovarian carcinoma ranks high among causes of cancerrelated deaths among women in the United States. The overall long-term prognosis is usually poor, with limited chemotherapeutic options. Primary mucinous ovarian carcinomas constitute approximately 10% of all primary ovarian malignant neoplasms. 1 Mucinous carcinomas involving the ovary are, in fact, more likely to be secondary neoplasms, originating most commonly in the large intestine and appendix but also in the stomach, pancreas, gallbladder, kidneys, and breast, among other sites. 2 Metastatic clear cell carcinomas and tumors originating in other gynecologic sites can likewise pose diagnostic dilemmas with primary ovarian carcinomas. The issue of ovarian vs nonovarian primary site is reasonably common, as approximately 7% of malignant ovarian tumors found on exploration of an abdominal or pelvic mass are metastatic. 3 Despite the fact that metastatic ovarian tumors are described as bilateral in up to two thirds of cases and tend to have a multinodular pattern without peritoneal involvement, the gross features are sometimes misleading. For example, Daya et al 4 found that 75% of metastatic large intestinal adenocarcinomas to the ovary were unilateral. The microscopic features can be similarly misleading, with colorectal carcinoma classically mimicking endometrioid ovarian carcinoma but also primary ovarian mucinous carcinoma. Various authors have attempted to define microscopic criteria to distinguish these tumors, including a garland pattern and dirty necrosis in colorectal adenocarcinoma metastatic to the ovary. 3-6 Other authors have refuted the usefulness of the aforementioned microscopic features or at least cautioned against relying on their use. 7,8 Because metastatic large intestinal tumors portend a far worse prognosis than primary 944 Am J Clin Pathol 2002;117: American Society for Clinical Pathology

2 Anatomic Pathology / ORIGINAL ARTICLE ovarian mucinous tumors, reaching the correct diagnosis is not merely an academic exercise. 3,8 Various immunohistochemical stains have been used in an effort to resolve the aforementioned differential diagnosis. Mucin gene expression has been studied in ovarian and colonic carcinomas, but the results are somewhat contradictory. Mucinous colorectal carcinomas (defined as at least 50% mucinous) overexpress the intestinal mucin MUC2, while nonmucinous colorectal carcinomas and their metastases demonstrate decreased MUC2 expression. 9,10 One immunohistochemical study found MUC2 expressed in colonic carcinoma (not specified as to mucinous characteristics on routine histologic examination), while both MUC2 and MUC5AC were expressed in primary ovarian mucinous tumors. 11 Conversely, another study using Northern blot analysis found both mucin genes expressed in a wide variety of primary ovarian tumors but not in endometrioid ovarian carcinomas. 12 The expression of both mucin genes was substantiated in mucinous primary tumors of the ovary using in situ hybridization. 13 An abstract suggested combining MUC2 and MUC5AC with cytokeratin (CK) 7 and CK20 in an immunohistochemical panel to differentiate between pseudomyxoma peritonei and primary ovarian mucinous tumors. 14 The majority of mucinous ovarian tumors in that series did not express MUC2, unlike those in the previously cited studies. The usefulness of the MUC2/MUC5AC panel for differentiating primary ovarian from metastatic mucinous tumors of colorectal origin is thus uncertain. Previous literature suggests that CK7 and CK20 expression may be among the most useful discriminant tests for differentiating primary ovarian from metastatic epithelial tumors involving the ovary. 7,15-18 The use of CK staining patterns is feasible because epithelial tumors usually conserve their original CK subsets in primary and in metastatic settings. 19,20 An extensive but somewhat divergent literature exists on their use in various diagnostic situations. Clear evidence exists that the use of CK7 and CK20 as a panel can help differentiate nonmucinous primary ovarian carcinomas from lower gastrointestinal carcinomas metastatic to the ovary ,21 Differentiating primary mucinous ovarian tumors from metastatic mucinous lesions can be more problematic. In addition, variability between studies in defining cutoff points for positive vs negative staining have led to conflicting data. Most authors define a positive staining result using a cutoff of 5% to 10% cell staining. Furthermore, several studies have been limited by the use of only 1 of the 2 stains or by the scope of tumors analyzed. To clarify the concurrent use of these stains in the differential diagnosis of mucinous tumors of the ovary and peritoneum, we evaluated a comprehensive series of tumors, with an emphasis on mucinous neoplasms, for the expression of both CKs. These included mucinous neoplasms likely to metastasize to the ovary, as well as primary mucinous and nonmucinous ovarian tumors and tumors from other gynecologic sites. Materials and Methods We selected 147 cases, 93 of which were mucinous and 54 of which were nonmucinous neoplasms, from the surgical pathology files of the University of Virginia, Charlottesville Table 1. The mucinous tumor types (n = 12) included primary ovarian tumors and nonovarian tumors likely to metastasize to the ovary. The majority of nonmucinous tumor types (n = 8) were of gynecologic origin. Based on availability, an average of 7 cases each of the mucinous and 5 cases each of the nonmucinous tumors were examined. The majority of tissues were from the primary site. In addition, 7 mucinous and 13 nonmucinous tumors metastatic to the ovary were analyzed, some together with the primary site Table 2. A single tumor metastatic to a nonovarian site was analyzed. Three sets of cases included tumor tissue from 2 different tumor sites in the same patient; these cases are identified as concurrent. Three separate control tissue types (gastric, colonic, and cervical) were examined, as were internal controls consisting of uninvolved tissue at 9 other tumor sites. All tissues were fixed in 10% neutral buffered formalin (0.25% vol/vol zinc added) and paraffin embedded. The results were correlated with the clinicopathologically determined primary site. Clone OV-TL-12/30 (CK7) and clone IT-Ks 20.8 (CK20), both obtained from DAKO, Carpinteria, CA, were selected to improve comparability with previous studies. Immunohistochemical staining using the avidin-biotin complex and peroxidase methods was performed on 5-µmthick sections. Protease 1 (Ventana Medical Systems, Tucson, AZ) and quenching in hydrogen peroxide were used for both CK7 and CK20, with no antigen retrieval for either. Dilutions were 1:2,000 for CK7 and 1:400 for CK20, and positive and negative controls were performed on each batch stained. A semiquantitative grading of percentage-cell staining was used initially to yield quartile scores of 0 to 4 (0, 0%-4%; 1, 5%-24%; 2, 25%-49%; 3, 50%-74%; and 4, 75%-100%). Ultimately, however, greater than 50% cell staining was chosen as positive, because of the characteristically patchy CK20 staining discussed in the Results section. Moderate to strong staining of the tumor cells was required for positivity. These stringent criteria of percentage and strength of positivity were used throughout the study, ie, for both single and combination antibody analyses. American Society for Clinical Pathology Am J Clin Pathol 2002;117:

3 Cathro and Stoler / EXPRESSION OF CYTOKERATINS 7 AND 20 IN OVARIAN NEOPLASIA Table 1 Cytokeratin (CK) 7 AND CK20 in Mucinous and Nonmucinous Tumors * Organ Tumor Type Total No. of Cases CK7+ CK20+ Mucinous tumors Appendix Adenocarcinoma 3 (1 concurrent tumor) 0 (0) 3 (100) Breast Ductal adenocarcinoma 7 (1 metastatic to nonovarian site) 6 (86) 0 (0) Colon Adenocarcinoma 10 1 (10) 7 (70) Endocervix Adenocarcinoma 7 7 (100) 0 (0) Endometrium Adenocarcinoma 7 6 (86) 0 (0) Gallbladder Adenocarcinoma 6 5 (83) 1 (17) Pancreas Adenocarcinoma (100) 2 (20) Peritoneum Pseudomyxoma 8 (3 concurrent tumors) 2 (25) 7 (88) Rectum Adenocarcinoma 4 2 (50) 3 (75) Ovary Adenocarcinoma 9 9 (100) 4 (44) Ovary Borderline 5 (2 concurrent tumors) 3 (60) 2 (40) Stomach Adenocarcinoma 10 6 (60) 2 (20) Nonmucinous tumors Bartholin gland Adenocarcinoma 1 0 (0) 0 (0) Endometrium Clear cell carcinoma 6 4 (67) 0 (0) Endometrium Endometrioid 5 5 (100) 0 (0) Endometrium MMMT 6 3 (50) 0 (0) Endometrium Serous papillary 5 5 (100) 1 (20) Kidney Clear cell carcinoma 10 3 (30) 0 (0) Ovary Serous papillary 5 5 (100) 0 (0) Vagina Adenocarcinoma 3 3 (100) 0 (0) MMMT, malignant müllerian mixed tumor. * Data are given as number (percentage). Concurrent indicates tumor tissue from 2 different tumor sites in the same patient. Boldface denotes staining patterns of diagnostic utility. Results Single Antibody Patterns For mucinous tumors, nearly all frankly malignant noncolonic tumors were diffusely CK7+ (Table 1). Likewise, for nonmucinous tumors, only clear cell carcinoma of renal origin (30% positive) did not have a majority of cases positive for CK7 staining. Colon carcinomas were almost all (9/10 [90%]) negative. Colonic carcinomas of rectal origin had a somewhat higher frequency of CK7 positivity (2/4 [50%]). Appendiceal adenocarcinoma and pseudomyxoma peritonei cases stained mostly like colonic tumors. The large majority of colonic type adenocarcinomas and related tumors were positive for CK20. This includes tumors of the colon (7/10 [70%]), rectum (3/4 [75%]), appendix (3/3 [100%]), and pseudomyxoma peritonei (7/8 [88%]). In contrast, the vast majority of other tumor types (both mucinous and nonmucinous) were CK20 negative. Potentially problematic CK20 staining was seen in a minor but significant proportion of ovarian mucinous (4/9 [44%]), gastric (2/10 [20%]), and pancreatobiliary (3/16 [19%]) adenocarcinomas. Table 2 Cytokeratin (CK) 7 and CK20 in Tumors Metastatic to the Ovary Primary Site Tumor Type Total No. of Cases CK7+ * CK20+ * Mucinous tumors Stomach Signet ring 2 1 (50) 2 (100) Appendix Signet ring 1 1 (100) 1 (100) Colon Adenocarcinoma 1 0 (0) 1 (100) Uterus MMMT 1 (1 primary site analyzed ) 1 (100) 0 (0) Unknown Adenocarcinoma 1 1 (100) 0 (0) Unknown Signet ring 1 1 (100) 0 (0) Nonmucinous tumors Uterus Endometrioid 6 (1 primary site analyzed ) 4 (67) 0 (0) Uterus Clear cell 2 (1 primary site analyzed ) 2 (100) 0 (0) Colon Adenocarcinoma 4 (2 primary sites analyzed ) 1 (25) 2 (50) Endocervix Adenocarcinoma 1 1 (100) 0 (0) MMMT, malignant müllerian mixed tumor. * Data are given as number (percentage). Staining identical to that of metastasis. Staining different from that of metastasis. 946 Am J Clin Pathol 2002;117: American Society for Clinical Pathology

4 Anatomic Pathology / ORIGINAL ARTICLE Combination Antibody Patterns The antibody staining patterns are summarized in Table 3. With the criterion of greater than 50% moderate to strong staining as positive, a CK7+/CK20 pattern was highly specific for tumors of other than the colonic type Image 1. No case of colonic or appendiceal carcinoma or pseudomyxoma peritonei had this pattern, although 1 of 4 cases of rectal adenocarcinoma exhibited this pattern. Nevertheless, an ovarian tumor with this pattern is almost certainly not of lower gastrointestinal origin. However, distinguishing upper gastrointestinal (eg, gastric or pancreatobiliary) from müllerian tumors based on this pattern is unreliable. For the pattern CK7 /CK20+, there was very high specificity for tumors of lower gastrointestinal origin (Image 1). The only significant exception was mucinous borderline tumor of the ovary, in which 2 (40%) of 5 tumors exhibited this phenotype. Rectal carcinomas again failed to consistently match the pattern of other colonic tumors, with only 2 (50%) of 4 having this pattern. Dual positive (CK7+/CK20+) tumors were relatively rare in this series. Among mucinous tumors, ovarian primary tumors seemed more likely than nonovarian tumors to display this pattern, but the reliability of this distinction was low (44% [4/9] dual positivity in ovarian primary tumors vs 11% [8/72] dual positivity in nonovarian mucinous tumors). In particular, tumors with this pattern are highly unlikely to be of endometrial origin. The least common phenotype was the dual negative (CK7 /CK20 ) pattern; renal cell carcinomas were the only exception (7/10 [70%] dual negative), and rare gastric, colonic, breast, endometrial, and Bartholin gland tumors had this phenotype. Metastatic lesions to the ovary (n = 20) stained with the CK pattern characteristic of the tissue of origin in the majority of cases (6 [86%] of 7 mucinous and 10 [77%] of 13 nonmucinous tumors; Table 2). The majority of those tumors for which tissue from the primary site was available (80%, n = 5) stained identically to the metastatic lesions at the primary site. Tumor tissue from 2 different sites in the same patient stained identically in those coincident cases (n = 3). Control Patterns and Grading The CK7 and CK20 staining patterns for the control tissues were as follows: gastric and colon tissue was invariably CK7 /CK20+, while cervix tissue was invariably CK7+/CK20. Normal gastrointestinal mucosa demonstrated only surface glandular positivity for CK20 and focal positivity for CK7 in both gastric and colonic mucosa, so only the superficial epithelium was assessed for control staining. Biopsy material was used for the control tissues, which underscored potential problems with tissue orientation Image 2. This factor, together with the heterogeneity of staining for CK20 in a variety of tumor types, supported the stringent requirements of a 50% cutoff for positivity to maintain specificity. Discussion Having examined a broad panel of mucinous ovarian tumors, mucinous tumors likely to metastasize to the ovary, Table 3 Cytokeratin (CK) 7/CK20 Positivity in Mucinous and Nonmucinous Neoplasms * Organ/Tumor Type CK7+/CK20 CK7 /CK20+ CK7+/CK20+ CK7 /CK20 Mucinous tumors Appendix/adenocarcinoma (n = 3) 0 (0) 3 (100) 0 (0) 0 (0) Breast/adenocarcinoma (n = 7) 6 (86) 0 (0) 0 (0) 1 (14) Colon/adenocarcinoma (n = 10) 0 (0) 7 (70) 1 (10) 2 (20) Endocervix/adenocarcinoma (n = 7) 7 (100) 0 (0) 0 (0) 0 (0) Endometrium/adenocarcinoma (n = 7) 6 (86) 0 (0) 0 (0) 1 (14) Gallbladder/adenocarcinoma (n = 6) 4 (67) 0 (0) 1 (17) 1 (17) Pancreas/adenocarcinoma (n = 10) 8 (80) 0 (0) 2 (20) 0 (0) Peritoneum/pseudomyxoma (n = 8) 0 (0) 5 (62) 2 (25) 1 (12) Rectum/adenocarcinoma (n = 4) 1 (25) 2 (50) 1 (25) 0 (0) Ovary/adenocarcinoma (n = 9) 5 (56) 0 (0) 4 (44) 0 (0) Ovary/borderline tumor (n = 5) 3 (60) 2 (40) 0 (0) 0 (0) Stomach/adenocarcinoma (n = 10) 5 (50) 1 (10) 1 (10) 3 (30) Nonmucinous tumors Bartholin gland/adenocarcinoma (n = 1) 0 (0) 0 (0) 0 (0) 1 (100) Endometrium/clear cell carcinoma (n = 6) 5 (83) 0 (0) 0 (0) 1 (17) Endometrium/endometrioid carcinoma (n = 5) 5 (100) 0 (0) 0 (0) 0 (0) Endometrium/MMMT (n = 6) 3 (50) 0 (0) 0 (0) 3 (50) Endometrium/serous papillary (n = 5) 4 (80) 0 (0) 1 (20) 0 (0) Kidney/clear cell carcinoma (n = 10) 3 (30) 0 (0) 0 (0) 7 (70) Ovary/serous papillary adenocarcinoma (n = 5) 5 (100) 0 (0) 0 (0) 0 (0) Vagina/adenocarcinoma (n = 3) 3 (100) 0 (0) 0 (0) 0 (0) MMMT, malignant müllerian mixed tumor. * Data are given as number (percentage). Boldface denotes staining patterns of diagnostic utility. American Society for Clinical Pathology Am J Clin Pathol 2002;117:

5 Cathro and Stoler / EXPRESSION OF CYTOKERATINS 7 AND 20 IN OVARIAN NEOPLASIA A B C D Image 1 The 2 most common patterns of cytokeratin (CK) 7 and CK20 staining. CK7+ (A) and CK20 (B) primary pancreatic carcinoma, an example of the predominant pattern seen in mucinous carcinomas other than those of lower intestinal origin (original magnification 200). CK7 (C) and CK20+ (D) primary colon carcinoma, typical of the pattern of tumors of lower intestinal origin (original magnification 200). and nonmucinous tumors from both gynecologic and nongynecologic sites, the usefulness of differential CK7 and CK20 staining is better defined. The principal conclusion of this study is that this immunohistochemical panel can reliably distinguish tumors metastatic to the ovary from the lower intestinal tract from other mucinous tumors metastatic to the ovary. Our findings agree with those of previous studies of the usefulness of CK7 and CK20 on many points. First, the majority of noncolonic mucinous carcinomas likely to metastasize to the ovary are CK7+/CK20, while the large majority of colorectal carcinomas are CK7 /CK20+. 7,15-18,21-23 Most nonmucinous gynecologic tumors are also CK7+/CK20. Clear cell carcinomas of the kidney are the one exception to the CK7+ in nongastrointestinal carcinoma rule, with negativity for both CK7 and CK20 in two thirds of cases. 21,22,24,25 Second, mucinous ovarian cystadenocarcinomas are invariably CK7+, but variably CK20+. Of those in our series, 44% were CK20+, while CK20 positivity in other series varied from 40% to 73%, with positivity being definitiondependent. 15,16,18 Third, tumors associated with pseudomyxoma peritonei are usually CK7 and nearly always CK20+, reflecting their appendiceal origin The mucinous nature of these tumors, together with the dispersed pattern of carcinomatous 948 Am J Clin Pathol 2002;117: American Society for Clinical Pathology

6 Anatomic Pathology / ORIGINAL ARTICLE A B Image 2 Surface vs deep glandular staining with cytokeratin (CK) 20 in lower intestinal epithelium. A, Surface epithelial staining with CK20 in normal colonic mucosa adjacent to a tumor (original magnification 100). B, Surface glandular staining in fragmented biopsy of normal colon, which in some situations could be misleading (original magnification 200). cell clusters, often causes a heterogeneous staining pattern with CK20, reducing the number of cases called positive using the 50% cutoff point. These findings underscore the importance of using CK7 and CK20 in parallel to maximize the information obtained. However, our findings differ somewhat from findings in other studies using CK7 and CK20 in several respects. First, 40% of borderline mucinous tumors of the ovary in our series were CK7, unlike those in another study examining these tumors using the same immunohistochemical panel. 26 Three other studies examining only CK7 found all mucinous borderline tumors of the ovary to be CK7+. 27,28,30 All 4 studies used low cutoff points for CK7 positivity (1 used >0%; 1, >1%; and 2, >5%). Second, rare cases of colorectal carcinoma can have CK7 positivity even using a stringent cutoff of 50%. One each of our cases of rectal (1/4 [25%]) and colonic carcinoma (1/10 [10%]) were both CK7+ and CK20+. The former finding may reflect the small number of cases or may indicate that some rectal tumors are phenotypically distinct. Others 5,16,31 have also observed CK7 positivity in this group of tumors. One study 16 found focal CK7 positivity in a single primary colonic adenocarcinoma; in a second study, 5 a metastatic clear cell adenocarcinoma of the intestine had patchy strong CK7 staining, while a third study 31 found 53% CK7 positivity (>5% cutoff) in a group of rectal adenocarcinomas. Other differences from the reported literature include an absence of CK7+ appendiceal adenocarcinomas in our small group, as opposed to 50% in another study that used a 10% cutoff for positivity. 29 A minority of our pancreatic carcinomas (2/10 [20%]) were both CK7+ and CK20+, while pancreatic tumors in 2 other studies had more than 60% dual positivity, but the studies used 1% and 5% cutoffs. 21,22 These differences emphasize the importance of the threshold values used to define positivity and negativity. Therefore, the most specific distinction that can be made using this immunohistochemical panel is between tumors metastatic to the ovary from the lower gastrointestinal tract and other mucinous tumors metastatic to the ovary. Caution should be used when rectal carcinoma, tumors associated with pseudomyxoma peritonei, or primary mucinous tumors of the ovary are in the differential diagnosis. Although the use of the 2 stains in parallel will reduce misclassification, some ambiguity still exists using this limited panel. Using a stringent cutoff of greater than 50% of carcinomatous cells staining for positivity can strengthen the differential utility of this panel. Although we specified moderate to strong intensity of staining for purposes of stringency, the percentage of tissue staining is known to be more important than the intensity of staining for true positivity. Goldstein and Bassi 23 made a strong case for the use of majority staining as a cutoff for positivity in a study using CK7 and CK20 in a panel with CK17. They also alluded to the risks of placing too much reliance on computerized immunohistochemical literature searches, as many of the databases do not take into account the varying definitions of positive vs negative. Yaziji et al 32 asserted that because of the steady increase of antibody sensitivity owing to improved pretreatment methods, focal expression of CK7 American Society for Clinical Pathology Am J Clin Pathol 2002;117:

7 Cathro and Stoler / EXPRESSION OF CYTOKERATINS 7 AND 20 IN OVARIAN NEOPLASIA and CK20 should be given the same significance as negative immunostaining. Another practical point based on our experience relates to the size of the sample. Small samples can be problematic when using an immunohistochemical stain that is positive in a mosaic or histologically limited pattern. CK20 is known to stain in a patchy manner, and the intensity of staining has been variable in proportion to the heterogeneity of staining. 23,25 Distinctions also have to be made between benign and neoplastic elements in each section, as both CK7 and CK20 stain mostly surface gastrointestinal epithelium in normal tissue. 25,33 These problems can be circumvented partially by restricting the use of this panel to larger operative specimens and using a 50% cutoff for positivity. Although 1 or 2 studies have examined the use of these antibodies on cytology specimens, at present, data are somewhat lacking to endorse their use in this setting. 24 Finally, it should be kept in mind that new molecular pathology techniques are demonstrating expression of messenger RNAs in tumors for which immunohistochemical stains for the corresponding proteins are negative. This is possibly an example of gene expression without translation leading to biologically ambiguous results to date. For example, Zemur et al 34 demonstrated 17 of 18 endometrial carcinomas expressing CK20 by reverse transcriptase polymerase chain reaction, despite negative immunohistochemical staining in all cases. Others demonstrated expression of CK20 in human ovarian carcinoma cell lines, although CK20 was absent in the original tumors. 35 While protein and complementary DNA microarrays may revolutionize our understanding of the expression profiles of an ever-widening set of tumors, these recent studies underscore the complexity of interpretation and validation that will be required to implement this technology. From the Robert E. Fechner Laboratory of Surgical Pathology, University of Virginia Health System, Charlottesville. Presented in part at the 90th Annual Meeting of the United States and Canadian Academy of Pathology, Atlanta, GA, March Address reprint requests to Dr Stoler: Dept of Surgical Pathology, PO Box , University of Virginia, Charlottesville, VA References 1. Russel P. Surface epithelial-stromal tumors of the ovary. In: Kurman RJ, ed. Blaustein s Pathology of the Female Genital Tract. 4th ed. New York, NY: Springer-Verlag; Hoerl HD, Hart WR. Primary ovarian mucinous cystadenocarcinomas: a clinicopathologic study of 49 cases with longterm follow up. Am J Surg Pathol. 1998;22: Ulbright TM, Roth LM, Stehman FB. Secondary ovarian neoplasia: a clinicopathologic study of 35 cases. Cancer. 1984;53: Daya D, Nazerali L, Frank GL. Metastatic ovarian carcinoma of large intestinal origin simulating primary ovarian carcinoma: a clinicopathologic study of 25 cases. Am J Clin Pathol. 1992;97: Young RH, Hart WR. Metastatic intestinal carcinomas simulating primary ovarian clear cell carcinomas and secretory endometrioid carcinoma: a clinicopathologic and immunohistochemical study of five cases. Am J Surg Pathol. 1998;22: Scully RE, Young RH, Clement PB. Tumors of the Ovary, Maldeveloped Gonads, Fallopian Tube and Broad Ligament. Washington, DC: Armed Forces Institute of Pathology; Atlas of Tumor Pathology, Third series, Fascicle DeConstanzo DC, Elias JM, Chumas JC. Necrosis in 84 ovarian carcinomas: a morphologic study of primary versus metastatic colonic carcinoma with a selective immunohistochemical analysis of cytokeratin subtypes and carcinoembryonic antigen. Int J Gynecol Pathol. 1997;16: Gersell DJ. Metastatic ovarian tumors: still a challenge [editorial]. Am J Clin Pathol. 1992;97; Hanski C, Hofmeier M, Schmidtt-Graff A, et al. Overexpression or ectopic expression of MUC2 is the common property of mucinous carcinomas of the colon, pancreas, breast and ovary. J Pathol. 1997;182: Hanski C, Riede E, Gratchev A, et al. MUC2 gene suppression in human colorectal carcinomas and their metastases: in vitro evidence of the modulatory role of DNA methylation. Lab Invest. 1997;77: Albarracin CT, Jafri J, Montag AG, et al. Differential expression of MUC2 and MUC5AC mucin genes in primary ovarian and metastatic colonic carcinoma. Hum Pathol. 2000;31: Giuntoli RL, Rodriquez GC, Whitaker RS, et al. Mucin gene expression in ovarian cancers. Cancer Res. 1998;58: Boman F, Buisine M-P, Wacrenier A, et al. Mucin gene transcripts in benign and borderline mucinous tumors of the ovary; an in situ hybridization study. J Pathol. 2001;193: Kuan SF, Hart J, Montag AG. Can CK7/CK20 cytokeratin and MUC2/MUC5AC mucin immunohistochemistry predict the primary site of pseudomyxoma peritonei in women [abstract]? Mod Pathol. 2001;14:139A. 15. Loy TS, Calaluce RD, Keeney GL. Cytokeratin immunostaining in differentiating primary ovarian carcinoma from metastatic colonic adenocarcinoma. Mod Pathol. 1996;9: Berezowski K, Statsky JF, Kornstein MJ. Cytokeratins 7 and 20 and carcinoembryonic antigen in ovarian and colonic carcinoma. Mod Pathol. 1996;9: Lagendijk JH, Mullink H, Van Diest PJ, et al. Tracing the origin of adenocarcinomas with unknown primary using immunohistochemistry: differential diagnoses between colonic and ovarian carcinomas as primary sites. Hum Pathol. 1998;29: Wauters CCAP, Smedts F, Gerrits LGM, et al. Cytokeratins 7 and 20 as diagnostic markers of carcinomas metastatic to the ovary. Hum Pathol. 1995;26: Moll R, Franke WW, Schiller DL. The catalog of human cytokeratins: patterns of expression in normal epithelia, tumors and cultured cells. Cell. 1982;31: Am J Clin Pathol 2002;117: American Society for Clinical Pathology

8 Anatomic Pathology / ORIGINAL ARTICLE 20. Moll R, Lowe A, Laufer J, et al. Cytokeratin 20 in human carcinomas. Am J Pathol. 1992;140: Wang NP, Zee S, Zarbo RJ, et al. Coordinate expression of cytokeratins 7 and 20 defines unique subsets of carcinomas. Appl Immunohistochem. 1995;3: Chu P, Wu E, Weiss LW. Cytokeratin 7 and cytokeratin 20 expression in epithelial neoplasms: a survey of 435 cases. Mod Pathol. 2000;13: Goldstein NS, Bassi D. Cytokeratins 7, 17 and 20 reactivity in pancreatic and ampulla of Vater adenocarcinomas. Am J Clin Pathol. 2001;115: Baars JH, De Ruijter JLM, Smedts F, et al. The applicability of keratin 7 monoclonal antibody in routinely Papanicolaoustained cytologic specimens for the differential diagnosis of carcinomas. Am J Clin Pathol. 1994;101: Miettinen M. Keratin 20: immunohistochemical marker for gastrointestinal, urothelial, and Merkel cell carcinomas. Mod Pathol. 1995;8: Ronnett BM, Shmookler BM, Diener-West M, et al. Immunohistochemical evidence supporting the appendiceal origin of pseudomyxoma peritonei in women. Int J Gynecol Pathol. 1997;16: Guerrieri C, Franlund B, Fristedt S, et al. Mucinous tumors of the vermiform appendix and ovary, and pseudomyxoma peritonei: histogenetic implications of cytokeratin expression. Hum Pathol. 1997;28: Ueda G, Sawada M, Ogawa H, et al. Immunohistochemical study of cytokeratin 7 for the differential diagnosis of adenocarcinomas in the ovary. Gynecol Oncol. 1993;51: Ronnett BM, Kurman RJ, Shmookler BM, et al. The morphologic spectrum of ovarian metastases of appendiceal carcinomas: a clinicopathologic and immunohistochemical analysis of tumors often misinterpreted as primary ovarian tumors or metastatic tumors from other sites. Am J Surg Pathol. 1997;21: Guerrieri C, Franlund B, Boeryd B. Expression of cytokeratin 7 in simultaneous mucinous tumors of ovary and appendix. Mod Pathol. 1995;8: Ashraf M, Zhang PJ. Cytokeratin 7 expression in rectal adenocarcinoma [abstract]. Mod Pathol. 1999;12:187A. 32. Yaziji H, Davie PL, Gown AM. On the use of cytokeratins 7 and 20 in the differentiation between transitional cell carcinoma and prostatic adenocarcinoma [letter]. Am J Clin Pathol. 2001;115: Ramaekers F, Van Niekerk C, Poels LG. On the use of monoclonal antibodies to keratin 7 in the differential diagnosis of adenocarcinomas. Am J Pathol. 1990;136: Zemur R, Fishman A, Bernheim J, et al. Expression of cytokeratin-20 in endometrial carcinoma. Gynecol Oncol. 1998;70: Yanagibashi T, Gorai I, Nakazawa T, et al. Complexity of expression of the intermediate filaments of six new human ovarian carcinoma cell lines: new expression of cytokeratin 20. Br J Cancer. 1997;76: American Society for Clinical Pathology Am J Clin Pathol 2002;117: