Special Considerations in Classification and Workup of Endometrial Carcinomas

Endometrial mesonephric-like carcinoma (MLCa) is a rare, aggressive carcinoma that often shows unimpressive cytologic features and may be misdiagnosed as International Federation of Gynecology and Obstetrics (FIGO) grade 1 to 2 endometrioid carcinoma. Endometrial mesonephric-like endometrial carcinoma frequently recurs and metastasizes, often to lung, but occasionally to unusual sites including the liver, brain, and eye.^23–27  The 5-year disease-specific survival for endometrial MLCa is 72%.²⁴  A single-institution study²⁶  of 23 endometrial MLCa cases reported median progression-free survival of 18.2 months (compared with 183 months for low-grade endometrioid carcinomas and 67.1 months for serous carcinomas) and median overall survival of 70.6 months (compared with 139.1 months for serous carcinomas). MLCa should be regarded as a high-grade tumor.^23,24 

MLCa typically shows variable architecture, including tubular, ductal, papillary, sievelike, corded, spindled, and glomeruloid patterns (Figure 1, A and B). As with its namesake, MLCa may show eosinophilic colloidlike secretions within tubules (Figure 1, C and D).^23–30  Cytologic atypia tends to be relatively mild, and monotonous oval vesicular nuclei reminiscent of papillary thyroid carcinoma may be seen (Figure 1, E).^23–30  Features reminiscent of corded and hyalinized endometrioid carcinoma (CHEC) or ovarian low-grade serous carcinoma have been reported in MLCa.^28,30  Because solid, spindled areas may be present within the spectrum of MLCa, McCluggage²⁴  recommends reserving a diagnosis of mesonephric-like carcinosarcoma for cases with heterologous sarcomatous elements. Some cases of MLCa are associated with endometrial atypical hyperplasia, supporting Müllerian origin.^24,31  Endometrial MLCa is not associated with mesonephric remnants.³² 

Immunohistochemical workup can facilitate accurate classification of these tumors. Expression of estrogen receptor (ER) and progesterone receptor (PR) should be absent or very limited; although some degree of ER staining is acceptable, PR staining is vanishingly rare.^26,29,32,33  The neoplastic cells may be positive for GATA3, TTF-1 (Figure 1, F), and/or luminal CD10.^24,25,28  TTF-1 and GATA3 tend to demonstrate an inverse staining pattern, and MLCa often shows diffuse staining for one of these markers.^23,24,28  TTF-1 and ER are good as first-line screens, but not all TTF-1–positive/ER-negative endometrial carcinomas are mesonephric-like carcinomas.²⁵  Immunohistochemical or molecular evidence of mismatch repair deficiency or microsatellite instability argues against a diagnosis of MLCa, as do areas of squamous or mucinous differentiation.^24,25,28  MLCa should show nonaberrant staining for p53 protein.^24,28  Napsin A and HNF1B staining have been negative in the cases reported in the literature.^26,32 

MLCa often harbors KRAS mutations without mismatch repair deficiency, TP53 mutation, or PTEN mutation; however, KRAS mutations are also often seen in endometrioid carcinoma with mucinous differentiation.^25,28,34 PIK3R1 mutation has been reported in MLCa but not in true mesonephric carcinoma.²⁸  MLCa is thought to represent approximately 1% of all endometrial carcinomas and 5% of KRAS-mutated microsatellite-stable endometrial carcinomas.²⁸  MLCa may show 1q amplification, but this is not a specific finding.²⁸  Most MLCa cases fall into the copy number low/no specific molecular profile category in TCGA molecular classification of endometrial carcinomas.²⁴  Two cases of KRAS-mutated endometrial MLCa with excellent and durable responses to lenvatinib and pembrolizumab have been reported,³⁵  although further studies are needed.

Some cases of low-grade endometrial endometrioid carcinoma show significant morphologic overlap with endocervical microglandular hyperplasia (MGH).^36–40  Because of the bland appearance of these endometrial carcinomas, misdiagnosis as MGH is not uncommon.^36–38  Additionally, rare cases of MGH have atypical features, simulating carcinoma.^39,40  Microglandular proliferations in samples from postmenopausal women are concerning, as even MGH is an abnormal finding in the absence of hormone therapy. MGH may be indicative of hormone excess,^39,41  which warrants further sampling to exclude the possibility of concurrent hyperplasia or malignancy.

Classic features of MGH include proliferation of small, tightly packed glands lined by bland cuboidal or low columnar mucinous epithelium, often with subnuclear vacuoles, rare to absent mitotic figures, and abundant acute and chronic inflammation.^39,42  Unusual features reported in so-called atypical MGH include signet ring cells, hobnail cells, nuclear atypia, mitotic figures, and solid, reticular, cribriform, or pseudoinfiltrative architecture.^40,43 

Although MGH is often associated with reserve cell hyperplasia, luminal squamous metaplasia is seen more often in low-grade endometrioid carcinoma.⁴²  Cytologic atypia, mitotic activity, foamy histiocytes (stromal foam cells), and medium to large complex glands are more common in endometrioid carcinoma (Figure 2, A and B).^38,42,44,45  Rare mitotic figures can be seen in either entity.^42,44,46  If a subcolumnar reserve cell layer can be identified in the microglandular proliferation, a diagnosis of MGH can be made.⁴⁴  Small, tightly packed glands and subnuclear vacuoles are more common in MGH cases.^38,44,45 

The immunohistochemical stains typically used to assess for endometrial versus endocervical origin do not assist with this differential diagnosis. MGH is not associated with high-risk human papillomavirus; therefore, p16 assessment is not particularly helpful. MGH and low-grade endometrioid carcinomas may show overlapping results for ER, PR, vimentin, and Ki-67.^42,43  Consistent p63-positive subcolumnar reserve cells⁴⁶  are seen only in MGH (comparison shown in Figure 2, C through F), whereas loss of PTEN⁴⁴  and/or PAX2⁴³  expression is much more commonly seen in endometrioid carcinoma.

KRAS mutation has been detected in a large subset of microglandular endometrioid carcinoma cases but has been absent in mitotically active MGH cases.⁴⁷ 

Polymerase epsilon (POLE) plays a role in proofreading and error correction during DNA replication.⁴⁸  Mutations in the exonuclease domain of POLE lead to an ultramutated phenotype. POLE-ultramutated (POLEmut) endometrial carcinoma tends to be diagnosed at a younger age (median, 55 years), in patients with a lower body mass index, and at low stage.^49,50  Although POLEmut endometrial carcinomas frequently show concerning histopathologic features, the clinical outcomes are excellent in retrospective studies (with close to 100% recurrence-free survival).^{15–17,51–57}  Post hoc analysis of samples from the PORTEC-3 clinical trial did not show benefit of adding chemotherapy to radiation in patients with POLEmut endometrial carcinoma, implying that the favorable outcome was not attributable to chemotherapy.¹⁵ 

Nevertheless, there are rare reports of adverse outcomes in POLEmut endometrial carcinomas. In one study of 23 patients with POLEmut endometrial carcinoma, 4 patients had recurrent disease and 1 had progressive disease after a median follow-up of 30 months.⁵⁷  One patient initially diagnosed with stage I disease underwent adjuvant intravaginal radiation therapy and subsequently developed isolated brain lesions that responded to treatment (surgical resection and stereotactic radiation, then stereotactic radiation and pembrolizumab). The other patient with brain metastasis died 33 months after being diagnosed with stage IV disease and being treated with chemotherapy. A meta-analysis of data from 294 patients with POLEmut endometrial carcinoma has indicated high salvage rates in the rare recurrences of POLEmut ECs.^57,58  In the recurrent or metastatic setting, POLEmut tumors may respond to immunotherapy.^55,59–63  PORTEC-4a (Molecular Profile–Based Versus Standard Adjuvant Radiotherapy in Endometrial Cancer; https://classic.clinicaltrials.gov/ct2/show/NCT03469674) is randomizing patients between the standard treatment and molecular-integrated risk-based individualized treatment and will show if omitting treatment in cases of favorable molecular profiles is safe and cost-effective.

POLEmut endometrial carcinoma typically shows features of endometrioid carcinoma (often high-grade/FIGO grade 3; Figure 3, A and B), though POLEmut endometrial carcinomas with serous and clear cell histotypes have also been documented.⁵⁸  Mixed differentiation, morphologic heterogeneity, and ambiguous histologic features are also commonly described.^49,58  The nuclear grade and mitotic index are often high, and lymphovascular invasion is common.^49,58  Peritumoral lymphocytes and tumor-infiltrating lymphocytes are typically present (Figure 3, C and D),^49,58  and scattered bizarre cells and serouslike atypia are also characteristic of these tumors (Figure 3, E and F). Other unique morphologic patterns that have been reported include lymphoepithelioma-like.⁴⁹  Although these features are enriched in POLEmut tumors, they can be seen across all 4 molecular groups, emphasizing the importance of molecular testing.²² 

The Proactive Molecular Risk Classifier for Endometrial Cancer separates ECs into the following 4 categories using a limited panel of tests (immunohistochemistry for mismatch repair proteins and p53 and testing for POLE exonuclease domain pathogenic variants): POLE mutated, mismatch repair deficient, p53 abnormal, and p53 wild type.^16,19  For a tumor to be classified as POLEmut EC, the detected mutation should be within the exonuclease domain and should be recognized as a pathogenic mutation. The most common pathogenic POLE exonuclease domain mutations are P286R, V411L, S297F, A456P, and S459F.⁶⁴  ECs with nonpathogenic POLE exonuclease domain mutations do not show the same excellent prognosis as those ECs with pathogenic POLE exonuclease domain mutations. Microsatellite instability–high (MSI-H) ECs with POLE mutation outside the exonuclease domain have similar genomic characteristics to MSI-H ECs without POLE mutation.⁶⁴  No specific POLE mutations have been associated with disease recurrence.

POLEmut endometrial carcinoma may acquire additional mismatch repair–associated mutational signatures during clonal evolution. A subset of POLEmut tumors shows only focal abnormal p53 staining with an abrupt transition to nonaberrant/wild-type staining patterns.⁵⁸  Of note, TP53 mutations have been reported in approximately one-third of POLEmut tumors.⁴⁹  Some authors believe that the serouslike or ambiguous features in these tumors may result from somatic mutations occurring secondary to the POLE mutation.^49,58  It has been postulated that this may also be the reason for heterogeneous patterns of staining for p53 and mismatch repair proteins.⁵⁸  Double-classifier POLEmut-p53abn tumors and POLEmut MSI-H tumors behave like POLEmut tumors.^64,65  Tumor molecular classification of early-stage high-risk EC using next-generation sequencing for POLE exonuclease domain mutation in combination with immunohistochemical assessment of p53 and mismatch repair proteins has been shown to be cost-effective based on a Markov decision model.⁶⁶ 

Among the less common appearances of endometrial endometrioid carcinoma is a biphasic pattern with spindling or cording of the epithelial cells, extensive hyalinization, and sometimes even osteoid formation, which may lead to misdiagnosis as carcinosarcoma.⁶⁷  In 2005, Murray et al⁶⁸  described a series of 31 cases of endometrioid adenocarcinoma with areas forming “cords, spindle cells, fusiform cells, or a striking hyalinized stroma,” invariably associated with a more typical villoglandular endometrioid carcinoma (often low-grade) comprising 10% to 90% of the tumor, which they termed CHEC. A total of 62 cases have been reported in the literature,^69–73  although the pattern is assuredly more common than this number suggests.

Initially, CHECs were thought to have a good prognosis, as they are often associated with a low-grade endometrioid carcinoma, but a small subset of reported cases has displayed more aggressive behavior. There are 29 reported cases with follow-up data. While the majority of patients (78%) have shown no evidence of disease at last follow-up, approximately 8% of patients were alive with disease and approximately 8% of patients had died of disease at last follow-up.⁷³  All patients who died of disease had a FIGO grade 2 endometrioid component and presented at stage II or higher. Poor outcomes in some cases are less surprising in light of reports that EC with CTNNB1 alterations, common in CHEC, portends a worse prognosis.^74,75 

These tumors most often show anastomosing cords of tumor cells within a variably dense hyalinized background (Figure 4, A). In some cases, the hyalinization may be minimal, resulting in a sex cord–like appearance (Figure 4, B). A spindle cell component may be present, either along with the corded component or in lieu of it (Figure 4, A and C). Myxoid, osteoid, and chondroid stroma may be present.^73,76  The CHEC pattern is often a minor component of the tumor, with the remainder composed of gland-forming endometrioid carcinoma, usually FIGO grade 1 or 2 (45 of 47 cases; 96%). In rare cases, a higher-grade endometrioid carcinoma may be present (Figure 4, D). That said, the corded and hyalinized component is often localized to the superficial aspect of the tumor and therefore may be disproportionately represented on endometrial biopsies and curettings.⁷³  A squamous component (frequently abrupt and keratinizing) is present in the majority of cases, and lymphovascular space invasion is not uncommon, seen in approximately one-quarter of cases.^70,73  The current recommendation is to grade CHEC based on the architecture of the conventional glandular component.⁷⁶ 

Immunohistochemically, the corded and hyalinized areas often show less extensive cytokeratin expression and more vimentin expression than the typical endometrioid component. The corded and hyalinized areas typically show nuclear positivity for β-catenin (Figure 4, E), and the glandular areas can also show at least focal nuclear staining. Expression of hormone receptors (ER, PR) is variable, but the corded component often shows lower expression than the glandular component and may be negative or only focally positive (Figure 4, F).

The majority of CHECs demonstrate alterations in the CTNNB1 gene, as suggested by their nuclear β-catenin staining. Unsurprisingly, most CHECs fall into the no specific molecular profile TCGA category. Cases with high-grade features (cytologic atypia, brisk mitotic activity) in the corded and hyalinized component (high-grade CHEC) may show TP53 and RB1 alterations in addition to CTNNB1 and PI3K pathway mutations,⁷⁰  and Safdar et al⁷¹  described 1 case with low-grade endometrioid carcinoma and low-grade CHEC components showing strong and diffuse p53 staining in both components. A study of 6 cases reported as CHEC with high-grade morphologic features in the corded and hyalinized component demonstrated mismatch repair deficiency in 3 (2 of which had high-grade endometrioid components) and aberrant overexpression of p53 in 2 (both with high-grade endometrioid components).⁷²  To our knowledge, there have been no reported ultramutated CHECs. Further studies are required to clarify the appropriate classification of tumors with high-grade corded and hyalinized and/or glandular components.

Unlike endometrial endometrioid carcinoma, which is often associated with endometrial atypical hyperplasia/endometrioid intraepithelial neoplasia, uterine serous carcinoma (USC) does not have a well-defined precursor lesion. As in the fallopian tube, p53 signatures—stretches of endometrial glandular epithelium with aberrant p53 staining—have been identified in association with USC and even show identical TP53 alterations.^77,78  Endometrial glandular dysplasia, a morphologically atypical lesion within the endometrium that falls short of serous carcinoma but harbors similar molecular findings,^79,80  is considered by some to represent a premalignant serous lesion.^81,82  However, unlike endometrial atypical hyperplasia/endometrioid intraepithelial neoplasia, these diagnoses may not be reproducible in the absence of coexisting serous carcinoma.

Serous endometrial intraepithelial carcinoma (SEIC) was originally described as a finding seen adjacent to carcinoma in up to 90% of cases of USC and was thought to represent a precursor lesion and/or intraepithelial spread.^83–86  Soon after, Zheng et al⁸⁶  suggested the term uterine surface carcinoma instead of endometrial intraepithelial carcinoma to emphasize the “unique aggressive nature” of SEIC and its potential for metastasis. Nonmyoinvasive USC was already established to have the potential for spread, with lymph node metastasis in 36% of cases and intraperitoneal disease in 43% in a broad study of USC by Goff et al,⁸⁷  and 3 bona fide cases of SEIC with extrauterine disease in the absence of USC were later reported as well.⁸⁸ Minimal USC (MUSC) was proposed by Wheeler and colleagues⁸⁹  as an umbrella term for USC that was nonmyoinvasive and measured 1 cm or less in greatest extent.

MUSC is, as the name suggests, a serous carcinoma—morphologically, immunophenotypically, and molecularly indistinguishable from the more invasive USC. The tumor is small but otherwise meets cytologic criteria for serous carcinoma, with enlarged epithelial cells having eosinophilic cytoplasm, variably sized nuclei, prominent nucleoli, and frequent mitotic activity. They almost invariably harbor TP53 alterations with aberrant expression of p53 by immunohistochemistry.

There is a close association between MUSC and endometrial polyps, with 80% to 90% of cases involving a polyp and around half (48%–53%) confined to one.^89–91  Compared with patients whose MUSC extends from a polyp to the background endometrium, patients with polyp-confined disease are much less likely to have extrauterine spread at the time of staging resection (17% versus 41.7% in non–polyp-confined cases) or to die of disease (9.4% versus 29.2%).⁹² 

The subtle presentation of MUSC belies its potential for high-stage disease. Between 33% and 45% of patients with MUSC have extrauterine disease at presentation, most commonly peritoneal spread but occasionally lymph node metastasis,^89,90  and therefore patients with any serous carcinoma on endometrial sampling or polypectomy, even SEIC, should undergo a full staging resection. For this reason, the 5th edition of the World Health Organization classification of female genital tumors discusses SEIC in the “serous carcinoma of the uterine corpus” section rather than listing it as a distinct diagnostic entity.⁹³  Endometrial polyps should be entirely submitted for histologic evaluation, as should the endometrium in a hysterectomy for endometrial serous carcinoma without a grossly identifiable lesion. Fortunately, noninvasive USC, including MUSC, that is limited to the uterus has a good prognosis, with a 5-year survival rate of around 90%.⁹⁴  Spread beyond the uterus dampens the prognosis considerably.

Evaluation for HER2 overexpression may be a reasonable consideration even in early-stage disease, as HER2 positivity has been shown to predict for lymphovascular invasion and recurrence.⁹⁵  Trastuzumab added to conventional chemotherapy improves progression-free survival in patients with advanced-stage or recurrent HER2-positive USC.⁹⁶  Neratinib, alone or in combination with olaparib, has shown promise in HER2-positive USC cell lines as well.⁹⁷ 

As described above, there are multiple situations in which accurate classification of endometrial carcinomas can be challenging, and awareness of pitfalls and immunohistochemical or molecular tests are of critical importance. Variants of endometrioid carcinoma may show morphologic similarity to benign endocervical MGH or carcinosarcoma. Aggressive carcinoma subtypes such as mesonephric-like endometrial carcinoma can have a low-grade histologic appearance, whereas POLEmut carcinomas (with expected favorable clinical course) often have a high-grade appearance. Even carcinomas that are limited in extent, such as MUSC, can have a poor prognosis. Recognition of these cancer subtypes allows for more accurate prognostication and can guide clinicians’ approach to treatment.