Adenocarcinomas of the Gynecologic Tract Involving the Urinary Bladder: A Series of 16 Cases Potentially Mimicking Urothelial Malignancy

Institutional review board protocols were obtained and followed. The surgical pathology databases from 2 institutions (University of Miami, Miami, Florida, and Johns Hopkins Medical Institutions, Baltimore, Maryland) were used to perform natural language searches including a combination of words: “bladder”; and “carcinoma” or “adenocarcinoma”; and “endocervical,” or “endometrial,” or “fallopian,” or “ovarian,” or “gynecologic.” The search period extended from 2011 to 2021 and included both in-house and consultation files. All available cases were centralized for review by 2 authors (D.H.R., A.P.) to confirm the diagnosis, describe the main morphologic features, and interpret the immunohistochemistry (IHC). Specimens without mucosal involvement (only serosa and/or muscularis propria) were excluded.

IHC for paired box gene 8 (PAX8) (Proteintech, Rosemont, Illinois), GATA-binding protein 3 (GATA3) (Cell Marque, Rocklin, California), estrogen receptor (ER; Ventana Roche Diagnostics, Rotkreuz, Switzerland), progesterone receptor (PR; Ventana Roche Diagnostics), p16 (Ventana Roche Diagnostics), tumor protein 63 (p63; Leica Biosystems, Buffalo Grove, Illinois), keratin 7 (Leica Biosystems), and keratin 20 (Leica Biosystems) was retrospectively performed on 4-μm-thick formalin-fixed, paraffin-embedded whole sections if material was available, and when no relevant immunostaining was performed at time of diagnosis. Tumors were stained with a standard protocol, using a Leica automated Bond System (Table 1). In cases without available unstained slides of paraffin blocks, information regarding IHC was retrieved from pathology reports. Clinical and radiographic information and follow-up, as well as time to bladder relapse in pertinent cases, were retrieved when obtainable.

We identified 16 cases with a median patient age of 61 years (range, 42–82 years). Cases were diagnosed on transurethral resection (n = 9), biopsy (n = 5), and cystectomy (n = 2). Most (14 of 16; 88%) had a documented prior history of gynecologic malignancy at the time of bladder sampling, with time to vesical relapse ranging from 3 weeks to 120 months (median, 22 months). No cases came provided with prior history of gynecologic malignancy or a suspicion of gynecologic malignancy stated in the pathology report. Final classification revealed endometrial endometrioid carcinoma (9 of 16; 56%) (2 low-grade and 7 high-grade), endometrial serous carcinoma (4 of 16; 25%), tubo-ovarian high-grade serous carcinoma (2 of 16; 13%), and cervical adenosquamous carcinoma (1 of 16; 6%). When available, imaging studies suggested nonbladder primary in 8 of 14 (57%), bladder origin in 3 of 14 (21%), but findings were unclear in the remaining 3 of 14 (21%). In those with obtainable cystoscopy reports, 10 of 12 (83%) tumors were favored to originate extravesically and 2 of 12 (17%) intravesically (3 tumors from 2 patients: 1 flat, 1 papillary, and 1 lesion obstructing the bladder) (Table 2).

In 12 cases with material available for review, morphologic findings consisted of uniform glandular histology without heterogeneity (9 of 12; 75%), tumor undermining and overrunning bladder mucosa with either minimal or no remaining normal urothelium (8 of 12; 67%) (2 of 12 indeterminate), reactive urothelium (4 of 12; 33.3%), and glandular and/or squamous differentiation (4 of 12; 33%) (Figures 1 through 3). One case appeared confined to the mucosa (1 of 12; 8.3%) with subsequent transurethral resection revealing overtly invasive disease (case 4; Figure 1, A through D). Eight cases had uniformly basophilic cytoplasm (8 of 12; 67%); 4 had eosinophilic cytoplasm (4 of 12; 33%), 2 of which had foci of clear cytoplasm. Lymphovascular invasion was not identified in any case (0 of 12) (Table 3).

Overall, tumors expressed keratin 7 (12 of 13; 92%); PAX8 (11 of 12; 92%) (9 strong and diffuse, 2 focal); ER (11 of 15; 73%) (9 diffuse, 2 focal); p16 (8 of 11, strong block positivity; 73%); PR (8 of 14; 57%) (5 focal, 3 diffuse); GATA3 (5 of 12; 42%) (2 strong and diffuse, 3 focal); and p63 (3 of 11; 27%). All tumors were negative for keratin 20 (0 of 12) (Table 4).

Four cases (4 of 16; 25%) were initially misclassified as genitourinary primary, including 3 as high-grade urothelial carcinoma (2 invasive, 1 papillary noninvasive) with squamous differentiation (2 of 3; 67%) or without squamous differentiation (1 of 3; 33%), and 1 as metastatic clear cell renal cell carcinoma (Figures 1 through 3).

Urothelial carcinomas may exhibit overlapping morphologic features with gynecologic adenocarcinomas⁹  (Figure 4, A through D, and Figure 5, A and B), which may make recognition challenging, particularly without provided clinical history. Given the presence of glandular or squamous differentiation normally found in a subset of urothelial tumors, detecting such areas may not necessarily prompt hesitation on the final diagnosis, which may account for this pitfall.

A relatively common feature seen in urothelial carcinomas is the presence of intratumoral heterogeneity with marked cell pleomorphism and variability of cytoplasmic quality. In our series, 75% of cases demonstrated relatively homogeneous morphology, while 67% were composed of tumor undermining urothelium with either no residual or minimal normal mucosa; both findings may prompt consideration of an extravesical primary. In addition, only 17% of cases demonstrated “classic” morphology of urothelial carcinoma, including papillary architecture, transitional-like appearance, and morphologic heterogeneity. Such heterogeneity, however, should not be mistaken for the presence of squamous and/or glandular differentiation, as these are common features identified in some gynecologic adenocarcinomas and observed in some of the cases from our series, particularly those with endometrioid histology.

Other features that may prompt consideration for involvement by an extravesical primary include uniform basophilic or dark amphophilic cytoplasm, present in 67% of cases, whereas urothelial carcinomas generally, although not always, have lighter eosinophilic cytoplasm, as well as more nuclear pleomorphism than what we observed in this series. Finally, squamous (morular or nonmorular) metaplasia is relatively common in endometrioid tumors and arises abruptly in the background of an adenocarcinoma, in contrast to the solid foci observed in urothelial carcinomas. In addition, squamous metaplasia in gynecologic adenocarcinomas does not contain overtly malignant cytology, in contrast to squamous differentiation in urothelial carcinomas, which are cytologically malignant.

The difficulty in correctly interpreting these cases, particularly without adequately provided clinical history, is demonstrated in cases 4, 7, 9, and 13, which were originally misinterpreted as primary urologic malignancies. Case 4, originally diagnosed as noninvasive high-grade papillary urothelial carcinoma, appeared entirely confined to the mucosa and had uniform glandular papillary morphology with focal squamous differentiation. We received an in-house follow-up transurethral resection and reclassified the case as grade 3 endometrial endometrioid carcinoma (Figure 1, A through D). This patient had a known history of endometrial endometrioid adenocarcinoma diagnosed 18 months prior (FIGO [Fédération Internationale de Gynécologie et d'Obstétrique] stage IB) and documented in the electronic medical record. Case 7 was originally diagnosed as invasive high-grade urothelial carcinoma. This patient also had a known prior history of FIGO stage IVB grade 3 endometrial endometrioid carcinoma documented in the medical chart. Retrospectively the tumor was reevaluated after receiving subsequent clinical information and inquiry, and was diffusely positive for PAX8 and reclassified as grade 3 endometrial endometrioid carcinoma. Case 9 consisted of a poorly differentiated invasive carcinoma that was focally positive for GATA3, PAX8, p63, and p16, but negative for keratin 7, keratin 20, ER, and PR (Figure 2, A through D). Such an ambiguous phenotype led the contributor to render a diagnosis of invasive high-grade urothelial carcinoma involving muscularis propria, a diagnosis that was re-reviewed after receiving subsequent clinical information and inquiry and reclassified as grade 3 endometrioid adenocarcinoma arising in endometriosis. In this instance, it was the patient's initial diagnosis of gynecologic malignancy. This patient also had suspicious imaging at the site of prior hysterectomy performed 9 years prior for endometriosis, as well as concurrent cystoscopy that demonstrated findings of a “chocolate cyst.” Case 13 was originally diagnosed as metastatic clear cell renal cell carcinoma (Figure 3, A through D) but was diffusely positive for PAX8 and ER and negative for GATA3. This patient had undergone neoadjuvant chemotherapy for a mass centered in the lower uterine segment and extending into the bladder, strongly considered to be of gynecologic origin by imaging, which was information not provided to the pathologist at the time of sign-out. We subsequently reclassified this case as endometrial serous carcinoma. It is worth briefly mentioning here that bona fide cases of occult metastatic renal cell carcinomas in the setting of negative renal imaging are exceptional, outside of a handful of case reports.^7,10 

Infrequently, urothelial carcinomas exhibit diffuse glandular morphology to such an extent as to trigger consideration of a primary vesical or urachal adenocarcinoma, or involvement from an extravesical site. Primary bladder adenocarcinomas are difficult to definitively diagnose, particularly when there is a substantial invasive component present that precludes evaluation of uninvolved and/or neoplastic urothelium. They are thought to arise in some instances from chronic cystitis cystica et glandularis with intestinal metaplasia and recapitulate intestinal neoplasia, both morphologically and immunophenotypically, requiring clinical, radiographic, and endoscopic correlation for the diagnosis.^4,8,11–14  It is important to note, however, that an intestinal phenotype is uncommon in gynecologic adenocarcinomas, with the exception being ovarian mucinous carcinomas,^8,15,16  and exceptionally uncommon in endometrial adenocarcinomas. When a tumor is centered in the wall of the dome and consideration is given to urachal adenocarcinoma, the same features used for distinguishing a primary vesical adenocarcinoma can be applied to urachal adenocarcinomas, as they share a common immunophenotype. Frequently, urachal adenocarcinomas demonstrate mucinous pools with floating signet ring cells, a morphology not typical of gynecologic adenocarcinomas.

Some examples of nephrogenic adenoma with prominent hobnail cytology should also be considered in the differential diagnosis. While these lesions may have some morphologic and immunophenotypic overlap with gynecologic clear cell carcinomas (positive for keratin 7, PAX8, and AMACR, and negative for GATA3 and keratin 20), nephrogenic adenomas lack nuclear atypia, hyperchromasia, and mitotic activity, and a solid component definitively excludes the diagnosis.

On occasion, high-grade serous carcinomas with mutations in breast cancer gene (BRCA) or other homologous recombination deficiency genes present with urothelial-like (termed transitional) morphology.^17–19  However, most patients have a prior diagnosis of tubo-ovarian cancer, and as such, caution is advised when rendering a diagnosis of urothelial carcinoma for a patient with a prior history of adnexal malignancy. Subsequent to acceptance for publication we received a human papillomavirus (HPV)–associated endocervical adenocarcinoma involving the bladder, originally considered to be invasive high-grade urothelial carcinoma with extensive glandular differentiation, arising from carcinoma in situ with glandular differentiation (Figure 5, A and B). Differentiation from endocervical adenocarcinoma can be made on morphologic grounds or with IHC, as most endocervical adenocarcinomas are HPV-driven malignancies. These tumors contain distinctive morphology consisting of columnar nuclei with cytoplasmic mucin depletion, brisk apical mitoses, and easily identifiable apoptotic bodies. When in doubt, positive in situ hybridization for high-risk HPV supports a lower gynecologic tract primary.^20,21 

One additional caveat worth mentioning here is that endometriosis may infrequently involve the bladder, as in case 9 in our series. Endometrioid adenocarcinoma, clear cell carcinoma, and mesonephric-like carcinomas are recognized as endometriosis-associated neoplasms,²²  and as such, lack of a prior gynecologic primary does not necessarily preclude the diagnosis of a gynecologic adenocarcinoma. In these instances, finding a combination of endometrial glands, stroma, and hemosiderin-laden macrophages in association with a tumor centered in the wall of the bladder, with supporting PAX8, CD10, ER, and PR IHC, are clues to the correct diagnosis.²³ 

In challenging cases in which the diagnosis remains uncertain after correlation with clinical history, cystoscopy, and imaging, a small panel of immunostains including PAX8, GATA3, and ER will likely confirm the correct diagnosis, as it is uncommon for urothelial carcinomas of the bladder to express either PAX8 or ER, although faint PAX8 expression in upper tract urothelial carcinoma is well known.²⁴  In cases of urothelial carcinoma with extensive glandular differentiation, GATA3 may be positive (focal or diffuse) or negative, while most cases of bladder adenocarcinomas are negative.²⁵  In our series, GATA3 was strongly and diffusely positive in 2 of 12 cases (cases 3 and 11).

GATA3 may not be quite as impactful in the differential diagnosis as PAX8, however, as immunoreactivity in gynecologic adenocarcinomas has been reported²⁶  and is in fact a diagnostic aid to mesonephric and mesonephric-like carcinomas, a rare histologic type.^27,28  Positive PAX8 with negative estrogen and PR panel should raise that diagnostic consideration if the tumor demonstrates the appropriate histologic features, which include the presence of small tubules with intraluminal secretions, along with areas of papillary, solid, and ductal growth patterns with round and spindled cells.^28–30  In this setting, TTF1 should also be included in the immunopanel as it is frequently positive in mesonephric-like tumors but negative in urothelial carcinomas.^28–32 

As it relates to other commonly used immunohistochemical stains, p63 may have some discriminatory power, as it is positive in 80% of urothelial carcinomas but is rare in adenocarcinomas of Müllerian origin,^12,18,33,34  even in cases with squamous differentiation. In contrast, p16 should be cautiously interpreted in this setting, as urothelial carcinomas may show diffuse blocklike nuclear and cytoplasmic expression,^20,35–37  overlapping with endometrial serous carcinomas, tubo-ovarian high-grade serous carcinomas, a subset of high-grade endometrioid carcinomas, and HPV-associated tumors of the uterine cervix.^37,38  Finally, while some have found keratin 20 to be beneficial to support urothelial origin,¹³  others have found it to be unhelpful in a substantial proportion of challenging cases, although strong keratin 20 expression would argue against gynecologic origin.^39–43  Similarly, keratin 7 may best be used in this setting to exclude a gynecologic primary if negative,^{8,12,39–41,44}  as a positive result is not discriminatory.

In conclusion, when approaching a bladder tumor with extensive glandular and/or squamous differentiation in a female patient, in addition to the “traditional” differential diagnoses (urothelial carcinoma, primary bladder adenocarcinoma, extension from a colorectal tract), malignancy involvement by a gynecologic adenocarcinoma should also be considered when tumors demonstrate morphologic homogeneity and lack of a surface mucosal urothelial precursor component. Prior history of gynecologic malignancy and cystoscopy suggestive of a nonbladder primary should be specifically sought out in these cases and, in equivocal cases, judicious use of a small immunohistochemical panel composed of GATA3, PAX8, and ER, with or without PR and p63, should resolve the issue, with caution given to the limitations and potential pitfalls caused by overlapping results.