Ovarian Endometrioid Adenocarcinomas With Infiltrative "Adenofibroma-like" Morphology and Aberrant ß-catenin Expression: Tumors That Coexpress CDX2 and LEF1 With Frequent Neuroendocrine Marker Expression, Diminished/Lost PAX8 and Possible Association With Endometrioid Type II Stem Cell Outgrowths in the Fallopian Tube.

It is with great interest that we read the report by McMullen et al describing unusual ovarian tumors with endometrioid proliferations co-expressing ER and CDX2 1 . We have also noticed several cases of endometrioid adenocarcinoma with very similar morphology to their first case 1 that showed a highly infiltrative “adenofibroma-like” morphology, characterized by small neoplastic glands set in a fibrous stromal background 2 , 3 ( Figures 1A – 1F ). Much like the glandular proliferations described by McMullen et al 1 , all of the carcinomas in our cohort showed strong, aberrant cytoplasmic and nuclear β-catenin expression and co-expression of CDX2. However, we also noted co-expression of LEF1, frequent co-expression of neuroendocrine (NE) markers and diminished to lost PAX8 expression ( Figure 2 and Table 1 ). Of particular interest, 2 of 5 cases with fallopian tubes available for testing had multifocal type II stem cell outgrowths (SCOUTs), a relatively recently described endometrioid lesion associated with aberrant β-catenin expression 4 (see Figures 1G – 1I ). Our cases arose in 6 patients, ranging in age from 64 to 76 years (mean: 71 years). In all cases with concurrent hysterectomy performed at the time of surgery (5/6 cases), the endometrium was benign; the remaining patient had undergone hysterectomy for benign indications decades prior to developing her ovarian malignancy. Grossly, the tumors ranged from 9 to 30 cm (mean: 14.3 cm). Adenofibroma-like foci of carcinoma were present as highly infiltrative small glands arranged singly or in small clusters or trabecular cords, separated by fibrous stroma ( Figures 1A – 1E ). Two cases also had prominent retraction spaces around tumor cells resulting a micropapillary-like appearance ( Figures 1B , 1C ). As noted, all cases showed aberrant cytoplasmic and nuclear β-catenin expression, which was often diffuse, as well as co-expression of CDX2 and LEF1; 5/6 cases had at least focal synaptophysin expression and 3/6 cases had chromogranin expression ( Figure 2 ). While ER was consistently positive (albeit often with weak expression), PAX8 was markedly diminished to negative in all but 1 case ( Figure 2 ). CK20 was negative in all cases. The immunohistochemical results are summarized in Table 1 . While the adenofibroma-like carcinomatous component of each case was FIGO grade 1, all cases also had at least focal areas with conventional FIGO grade 1–3 endometrioid morphology (see Table 1 ). Of the histological patterns typically associated with aberrant β-catenin expression in endometrioid adenocarcinomas (i.e. squamoid morules 5 , spindled 6 , corded and hyalinized 7 and pilomatrix-like high grade endometrioid adenocarcinoma 8 – 10 ) only a single case showed spindled elements. One case had focal intracytoplasmic mucin ( Figure 1F ). Precursor lesions were present in 2/6 cases in the form of endometrioid adenofibromas (accompanied by a borderline endometrioid tumor in 1 case); definitive endometriosis was not present in any of our cases. However, multifocal endometrioid type II SCOUTs were seen in 2/5 cases with fallopian tube tissue available for immunostaining; at least 5 in one tube, and 3 in the other. Tumor stage ranged from pT1a to pT3b. No lymph node metastases were present and none of the patients with clinical follow up information (4 of 6) experienced tumor recurrence with a mean follow up of 10 months (range 2–75 months). The clinicopathological features of these cases are summarized in Table 1 .

The features we describe herein, 1) an adenofibroma-like pattern in endometrioid adenocarcinoma 2 , 3 , 11 , 2) the association of aberrant β-catenin with positive CDX2 1 , 12 and LEF1 12 , and 3) the propensity for some endometrioid adenocarcinomas to express CDX2 and NE markers with diminished to negative PAX8 expression 2 , are by themselves not novel concepts. However, what is novel, and what we would like to emphasize, is the link among all three occurring together. The importance of this morphologic and immunophenotypic signature lies in the propensity of these tumors to mimic ovarian metastases from a gastrointestinal origin, especially appendiceal goblet cell adenocarcinomas. Metastatic appendiceal goblet cell adenocarcinoma often shows a similar immunophenotype 13 and one of our cases even had focal intracytoplasmic mucin ( Figure 1F ). Certainly, the aberrant cytoplasmic and nuclear β-catenin expression in these ovarian endometrioid adenocarcinomas can be helpful in making this distinction, since aberrant β-catenin expression 14 and CTNNB1 mutations 15 , 16 are not characteristic of appendiceal goblet cell adenocarcinomas. Additionally, none of our tumors were positive for CK20, while goblet cell adenocarcinomas are frequently positive 13 , and all our cases expressed ER at least focally (albeit often with weak intensity). Other differentials that may be even more problematic are metastases from pancreatic, biliary, or upper gastrointestinal (i.e. gastric or gastroesophageal junctional) sources, as adenocarcinomas from these regions may harbor CTNNB1 mutations 17 – 19 . Reaching the correct diagnosis of endometrioid adenocarcinoma in these cases depends on generous sampling of the tumor to identify areas with more conventional endometrioid morphology, which were present, at least focally, in all our cases. In addition, at least focal ER expression was present, including in areas with adenofibroma-like morphology, which may be used in this context to support a diagnosis of endometrioid adenocarcinoma. Finally, primary Brenner tumors of the ovary can show some morphological overlap with these tumors; however, consistent positivity for urothelial markers (i.e. GATA3, p63 or uroplakin) 20 makes this distinction less problematic than some of the aforementioned considerations. The two cases stained for GATA3 in our series were negative. The origin of these morphologically and immunohistochemically variant tumors is also of interest. The fact that 2 of our cases had multiple endometrioid type II SCOUTs 4 in the fallopian tubes provides a potential explanation, not only for the pathogenesis of these unusual tumors in the absence of endometriosis (absent in all our cases), but may also explain the diffusely aberrant β-catenin pattern observed in the adenofibroma-like areas, which is not characteristic of low-grade endometrioid adenocarcinomas 4 . We note that the true frequency of type II SCOUTs in this cohort may be underestimated as only 5/6 cases had fallopian tube tissue available for testing and none had their tubes entirely submitted for microscopic examination. In summary, we highlight the link between 1) adenofibroma-like morphology in ovarian endometrioid adenocarcinomas, 2) aberrant β-catenin expression associated immunophenotype that includes co-expression of CDX2 and LEF1, frequent neuroendocrine marker expression and diminution or loss of PAX8 and 3) possible association with endometrioid type II SCOUTs. These ovarian tumors are morphologically on the well-differentiated end of the spectrum of endometrioid adenocarcinomas with Wnt/β-catenin activation as compared to their uniformly high-grade PiMHEC counterpart 8 – 10 . Though we had limited cases with follow-up, we cautiously speculate that these tumors may also be less aggressive clinically, recognizing the need for additional study.