Main

Among ovarian epithelial tumors, the mucinous tumors pose the greatest difficulty with regard to distinction of primary from metastatic tumors. The primary ovarian mucinous tumors, including the atypical proliferative (borderline) mucinous tumors of gastrointestinal type and mucinous carcinomas of usual type, are usually easily distinguished from metastatic mucinous carcinomas in the ovaries when they exhibit characteristic gross and microscopic features. However, when characteristic features are lacking, ancillary studies are often necessary to distinguish these tumors. Unfortunately, only a limited number of immunohistochemical markers have been shown to be useful in this distinction, including cytokeratins 7 and 20 (CK7 and CK20). In routine practice metastatic mucinous carcinomas are more common than primary ovarian mucinous carcinomas,1 with those of gastrointestinal and pancreaticobiliary tract origin being the most commonly encountered types.1, 2 Utility of CK7/CK20 coordinate expression profiles for distinction of these tumors is limited due to several factors: (1) both primary ovarian mucinous tumors and metastatic mucinous carcinomas of lower gastrointestinal tract origin (colon, rectum, and appendix) frequently express CK20, as do some metastatic mucinous carcinomas of upper gastrointestinal tract origin (stomach and pancreaticobiliary tract), (2) both primary ovarian mucinous tumors and metastatic mucinous carcinomas of upper intestinal tract origin have a high frequency of CK7 expression and overlapping CK7/CK20 profiles, and (3) subsets of all of these tumors exhibit the non-specific CK7+/CK20+ immunoprofile. Therefore, other markers that distinguish these tumors would be of value.

CDX2 is a nuclear transcription factor involved in intestinal-type differentiation, and a number of studies have evaluated this marker for its utility in distinguishing primary ovarian mucinous tumors from metastatic adenocarcinomas involving the ovary. CDX2 is expressed in a high proportion of colorectal carcinomas (61–100%)3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 whereas the reported range of expression in primary ovarian mucinous tumors (the majority reported being carcinomas) is broad (0-100%).3, 4, 7, 8, 9, 11, 12, 13, 14, 15, 16, 19, 22, 23 Most studies have evaluated CDX2 and CK20 in ovarian mucinous tumors as independent markers;7, 9, 12, 14, 16 very few studies have addressed coordinate expression with CK722, 23 and none has performed an analysis of coordinate expression profiles directly comparing the performance of CDX2 and CK20, when combined with CK7, in the same group of tumors.

Given the broadest possible range of results reported in the literature for CDX2 expression in primary ovarian mucinous tumors (0–100%),3, 4, 7, 8, 9, 11, 12, 13, 14, 15, 16, 19, 22, 23 it is not surprising that questions regarding the utility of CDX2 for distinction of primary ovarian mucinous tumors from metastases continue to be posed (we encounter this situation in our consultation practice and when lecturing on ovarian mucinous tumors). Thus, the aim of the current study was to compare CDX2 and CK20 expression in a large group of rigorously classified primary ovarian mucinous tumors and metastatic mucinous carcinomas of gastrointestinal tract origin and evaluate coordinate expression profiles of each of these markers combined with CK7 to determine whether CDX2 offers any diagnostic advantage over CK20.

Materials and methods

Case Selection

Ninety mucinous tumors involving the ovary were selected from the surgical pathology files of The Johns Hopkins Hospital, the Armed Forces Institute of Pathology (Department of GYN & Breast Pathology), and Washington Hospital Center from 1990 to 2006. Cases consisted of a mixture of routine in-house and consultation cases. Forty-two tumors were primary ovarian mucinous tumors (31 atypical proliferative (borderline) mucinous tumors of gastrointestinal type and 11 invasive mucinous carcinomas of usual type (referring to the ‘common’ type of ovarian mucinous carcinomas having gastrointestinal-type or non-specific mucinous differentiation rather than the rare mucinous carcinomas of seromucinous (‘endocervical-like’) type)).24, 25 Forty-eight tumors were metastatic mucinous adenocarcinomas involving the ovary, with primary sites including colon and rectum (n=25), appendix (n=4), pancreaticobiliary tract (n=14), and stomach (n=5). Those from the colon, rectum, and appendix were grouped as metastases of lower gastrointestinal tract origin. Those from the pancreaticobiliary tract and stomach were grouped as metastases of upper gastrointestinal tract origin. Of the 48 metastatic adenocarcinomas, eight were of signet ring cell type (four gastric, three appendiceal, one colorectal). Clinical data, data from imaging studies, and pathologic (gross and microscopic) criteria26 were used to rigorously classify the tumors as primary or metastatic.

Immunohistochemistry

Immunohistochemical stains were performed at both PhenoPath Laboratories and The Johns Hopkins Hospital (JHH) Immunopathology Laboratory using formalin-fixed, paraffin-embedded tissue sections. For the vast majority of cases, the immunohistochemical stains (CK7 and CK20) were performed at PhenoPath Laboratories. For a minority of cases, immunohistochemical stains (CK7 and CK20) were performed at JHH as part of the routine diagnostic evaluation of the cases; due to the reliable performance of these stains and appropriate controls in the JHH laboratory, they were not repeated at PhenoPath Laboratories. However, all CDX2 stains were performed at PhenoPath Laboratories. The manufacturer, clone, dilution, and pretreatment details for each primary antibody are summarized in Table 1. Unstained sections were deparaffinized and rehydrated prior to subjecting them to antigen retrieval using protocols optimized for each antibody.

Table 1 Details of immunohistochemical analysis
Full size table

The immunohistochemical method used at PhenoPath Laboratories was performed as follows. Following pretreatment, primary antibodies (CDX2, CK7, and CK20) were applied to respective sections and incubated for 30 min at room temperature. Antibody localization was achieved by incubating slides for 30 min at room temperature in Envision+ labeled polymer (DakoCytomation, Carpinteria, CA, USA) using a Dako autostainer. The slides were then incubated for 10 min at 37°C in a solution containing 3% hydrogen peroxide and 3,3′-diaminobenzidine.

The immunohistochemical method used at The Johns Hopkins Hospital laboratory was performed as follows. Following pretreatment, primary antibody (CK7 and CK20) was applied to respective sections and incubated for 20 min, respectively, at room temperature. Antibody localization was achieved by incubating slides for 16 min at room temperature in iVIEW labeled conjugate (Ventana; Tucson, AZ, USA) using a Ventana autostainer (Benchmark XT). The slides were then incubated for 8 min at room temperature in a solution containing hydrogen peroxide (prediluted; Ventana) and 3,3′-diaminobenzidine.

Interpretation and Scoring of Immunohistochemical Preparations

Nuclear expression of CDX2 and cytoplasmic expression of CK7 and CK20 were considered positive staining reactions. As the extent of staining for CK7 and CK20 has been shown to be diagnostically relevant for distinction of primary and secondary mucinous tumors of the ovary,27 immunohistochemical results were initially scored in a semiquantitative fashion by one of the authors (RV) based on the percentage of tumor cells positive (0: ≤5; 1+: 6–25; 2+: 26–50; 3+: 51–75; and 4+: 76–100%). However, due to the complexity of reporting both staining distribution and coordinate expression profiles for multiple tumor types, this 5-tiered scoring system was compressed into a 3-tiered system (0: ≤5% (negative); focal +: ≤50%; and diffuse +: >50%) for reporting results for single marker expression and the markers in the coordinate expression profiles were simply reported as positive (>5% of tumor cells) or negative (≤5% of tumor cells). In general, intensity of staining was not used for scoring; however, only cases with greater than weak intensity were considered positive. Intensity was arbitrarily designated as weak when staining was present but only barely detectable. As staining intensity was not recorded, we did not correlate it with the distribution of staining.

For purposes of analyzing coordinate expression, atypical proliferative (borderline) tumors and primary ovarian carcinomas were combined into a single category of ‘primary ovarian mucinous tumors’, metastatic colorectal carcinomas and appendiceal carcinomas were combined into a single category of ‘metastatic lower gastrointestinal tract adenocarcinomas’, and metastatic pancreaticobiliary and gastric carcinomas were combined into a single category of ‘metastatic upper gastrointestinal tract adenocarcinomas’. This was done because the individual tumor categories grouped in each of these combined categories have been shown to display similar patterns of CK7/CK20 coordinate expression.27

Statistical Analysis

χ2 and Fisher exact tests were used to compare the differences in percentages of positive results between groups. All data was analyzed in SAS v9.1 (SAS Institute Inc, Cary, NC, USA). P-values <0.05 were considered statistically significant.

Results

Comparison of CDX2 vs CK20 Expression as Individual Markers

Results are detailed in Table 2. Expression of CDX2 was most common in metastases from the stomach and appendix (100%) and was frequently seen in metastases from the colon, rectum, and pancreaticobiliary tract (64–88%). Expression was also encountered in primary ovarian mucinous tumors but less frequently so (36–42%). Primary ovarian mucinous tumors as a group expressed CDX2 (40%) less frequently than CK20 (83% (P<0.0001)) and this frequency of CDX2 expression was lower than CDX2 expression in metastatic mucinous carcinomas of both upper (74% (P=0.016)) and lower gastrointestinal tract origin (90% (P<0.0001)). Expression of CK20 was frequent in tumors from all anatomic sites (73–100%), with all metastases from the stomach and appendix showing expression. In particular, CK20 expression was similar in primary ovarian mucinous tumors (83%) and metastases of upper (89% (P=0.071)) and lower gastrointestinal tract origin (93% (P=0.29)). Thus, as a single marker CDX2 offered some advantage over CK20 because it was less frequently positive in primary ovarian mucinous tumors.

Table 2 CDX2 and CK20 expression in primary ovarian mucinous tumors and metastatic mucinous carcinomas of gastrointestinal tract origin (n=90)
Full size table

In CDX2+ tumors, expression was usually diffuse (>50% of tumor cells positive) in colorectal and appendiceal tumors, usually focal (≤50% of tumor cells positive) in ovarian atypical proliferative (borderline) tumors and metastases from the pancreaticobiliary tract and stomach, and variable in primary ovarian mucinous carcinomas. In CK20+ tumors, expression was usually diffuse (>50% of tumor cells positive) in metastases from the colon, rectum, appendix, and stomach and in primary ovarian carcinomas; expression was usually focal (<50% of tumor cells positive) in ovarian atypical proliferative (borderline) tumors and variable in metastases from the pancreaticobiliary tract.

Comparison of CDX2 vs CK20 in Coordinate Expression Profiles with CK7 (Figures 1, 2 and 3)

Figure 3
figure 3

(a) Metastatic pancreatic carcinoma. The tumor simulates an ovarian atypical proliferative (borderline) mucinous tumor of gastrointestinal type in this field. (b) Focal expression of CDX2 (1+ positive overall). (c) Focal expression of CK20 (1+ positive overall). Tumor exhibited diffuse CK7 expression (4+ overall; not shown).

Full size image
Figure 2
figure 2

(a) Metastatic colorectal carcinoma. The tumor simulates an ovarian atypical proliferative (borderline) mucinous tumor of gastrointestinal type in this field. (b) Diffuse expression of CDX2 (4+ overall). (c) Nuclear staining for CDX2. (d) Focal expression of CK20 (1+ overall). Tumor lacked CK7 expression (not shown).

Full size image
Figure 1
figure 1

(a) Primary ovarian mucinous tumor. (b) Focal expression of CDX2 (1+ overall). (c) Diffuse expression of CK20 (4+ overall). Tumor exhibited diffuse CK7 expression in this case (4+ overall; not shown).

Full size image

Coordinate CK7/CDX2 and CK7/CK20 profiles for primary ovarian mucinous tumors and metastatic mucinous carcinomas, divided into those of upper and lower gastrointestinal tract origins, are detailed in Table 3. CK7/CDX2 coordinate expression profiles offered limited advantage over CK7/CK20 profiles for distinction of primary ovarian tumors from metastases. The CK7+/CDX2− profile was more common than the CK7+/CK20− profile in primary ovarian tumors (60 vs 17%, respectively (P<0.0001)) and thus better for distinguishing the primary ovarian tumors from lower gastrointestinal tract metastases, which lacked these profiles, but only when CDX2 was negative; however, CK7 coupled with either marker offered generally similar distinction of primary ovarian tumors from lower gastrointestinal tract metastases due to the discriminatory value of CK7. CK7/CDX2 profiles provided some advantage for distinction of primary ovarian tumors from upper gastrointestinal tract metastases (CK7+/CDX2− in 60 vs 26%, respectively (P=0.016) and CK7+/CDX2+ in 36 vs 63%, respectively (P=0.046)) because CK7/CK20 profiles were essentially identical in both tumor types (CK7+/CK20− in 17 and 11%, respectively (P=0.42) and CK7+/CK20+ in 79% for both types). However, the presence of shared CK7/CDX2 coordinate expression profiles in both the primary ovarian tumor and upper gastrointestinal tract metastases categories, with more than rare numbers of tumors in individual expression profile subgroups, limits the utility. For tumors lacking CK7 expression, CK7/CDX2 coordinate expression offered no advantage over CK7/CK20 because the CK7−/CDX2+ and CK7−/CK20+ profiles characterized most metastases of lower gastrointestinal tract origin (83 and 86%, respectively (P=1.00)) and were thus identical in their ability to distinguish most metastases of lower gastrointestinal tract origin from both metastases of upper gastrointestinal tract origin and primary ovarian mucinous tumors which infrequently had either profile (11 and 5%, respectively, for both tumor types).

Table 3 Comparison of coordinate CK7/CDX2 and CK7/CK20 profiles in primary ovarian mucinous tumors and metastatic mucinous carcinomas of gastrointestinal tract origin
Full size table

Discussion

The current study is the first immunohistochemical analysis of primary and secondary mucinous ovarian tumors to directly compare CDX2 with CK20 in the context of coordinate expression with CK7 (ie, CK7/CDX2 vs CK7/CK20). The results demonstrate that CDX2 offers some but limited advantage over CK20 in conjunction with coordinate expression of CK7 for distinction of primary ovarian mucinous tumors from metastatic mucinous carcinomas of upper gastrointestinal tract origin but not those of lower gastrointestinal tract origin. A number of studies have reported a variety of results for both frequency and extent of CDX2 and CK20 in primary ovarian mucinous tumors and colorectal carcinomas, without comparison of each of these markers with one another in the setting of coordinate expression with CK7.3, 4, 7, 8, 9, 11, 12, 14, 15, 16, 19, 22, 23 Our results for primary ovarian mucinous tumors are in keeping with some studies3, 4, 7, 9, 11, 14, 15, 16, 19 but not others;4, 7, 8, 9, 12, 14, 22, 23 possible reasons for the differences observed include criteria for classifying tumors as mucinous, inclusion of endometrioid tumor types, and differing immunohistochemical techniques and scoring methods.

Although there has been much investigation of CDX2 as a diagnostic marker for the distinction of primary ovarian mucinous tumors and colorectal carcinomas,3, 4, 7, 8, 9, 11, 12, 14, 15, 16, 19 expression of this marker is not limited to these tumor types. In the current study, CDX2 was also expressed in metastatic appendiceal, pancreaticobiliary, and gastric carcinomas. In the literature, CDX2 expression has been reported in adenocarcinomas of various anatomic sites.3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42 CDX2 expression has even been studied in ovarian atypical proliferative seromucinous tumors (referred to as endocervical-like mucinous borderline tumors or mixed epithelial borderline tumors of mullerian type by some authors). These types of ovarian tumors are typically diffusely positive for CK7, ER, and CA-125, variably positive for PR and mesothelin, and negative for CK20 and CDX2.43 However, atypical proliferative seromucinous tumors are usually not in the histologic differential diagnosis of metastatic mucinous carcinomas involving the ovary; thus, immunohistochemical evaluation is generally not needed for that type of primary ovarian tumor in routine practice.

Given the ubiquity of expression in a variety of adenocarcinomas, CDX2 cannot be used as an individual marker for the distinction of primary ovarian mucinous tumors from metastases in the ovary and appears to offer limited advantage over CK20 in conjunction with coordinate expression of CK7 in this differential diagnosis. If CDX2 is to be used for diagnostic purposes, it should be assessed in the context of coordinate expression with CK7, similar to the manner in which CK20 is typically used (ie, CK7/CK20 coordinate expression).