Introduction

The classification of endometrial carcinoma has evolved over time, with the goal of more precisely predicting patient prognosis and guiding management. The evolution of endometrial cancer classification has raised several questions such as (1) Can histotype of endometrial carcinoma be reproducibly defined? (2) How can stratification of patient risk of recurrence or death from disease be improved? (3) Which patients can be cured by surgery alone and who are candidates for fertility-preserving therapy? and (4) What is the appropriate surveillance for the patient after initial treatment? The Cancer Genome Atlas (TCGA)-based classification of endometrial carcinoma has shown promise in refining endometrial carcinoma classification and more accurately reflecting patient outcome. We review the evolution of endometrial carcinoma classification, from purely morphological to the recently proposed TCGA genomic-based classification, and provide an overview of immunohistochemical and molecular markers used in endometrial carcinoma classification. We also comment on future directions in endometrial carcinoma subclassification arising from these results.

The history of endometrial cancer classification

Uterine cancer was first subclassified based only on anatomical location, so that tumours from the cervix and uterine corpus were treated as separate entities; prior to this, cancers of the uterus were viewed by physicians as a single disease [1]. With regard to carcinoma of the uterine corpus, Bokhman first described two clinicopathological types of endometrial carcinoma based on epidemiological studies [2], namely, Type I endometrial carcinoma, associated with unopposed estrogen stimulation and corresponding to low-grade endometrioid carcinoma, and Type II which is unrelated to estrogen stimulation and corresponds to serous carcinoma [3]. This landmark study identified a high-grade endometrial carcinoma variant that was associated with increased frequency of myometrial invasion, metastasis and likelihood of death due to disease [2]; however, the Type I and Type II classification did not enter diagnostic practice as there was no clear boundary between the types and too many cases defied classification as either Type I or Type II. The corresponding histotypes, i.e. endometrioid or serous, are also not clearly separable in practice with some tumours having ambiguous morphology [4]; especially for high-grade tumours, interobserver variability in histotype diagnosis is considerable [5,6,7,8]. In one study, 56 cases of high-grade endometrial cancer were reviewed by three pathologists, and a consensus major histotype diagnosis was reached in only 62.5% of cases [5]. Another study showed that interobserver agreement in classification of high-grade endometrial carcinoma histotype based on morphology alone was only moderate [7]. Finally, Thomas et al. showed that upon review of 131 cases of grade 3 endometrioid carcinoma by two gynecologic pathologists, reclassification occurred 38% of the time [8]. Despite this, current endometrial carcinoma classification relies heavily on morphological features and the WHO 2014 classifies endometrial carcinoma based on histological subtype (Table 1), with endometrioid and serous carcinoma accounting for a large majority of cases.

Table 1 WHO 2014 histologic classification of endometrial carcinoma
Full size table

Endometrial carcinoma histotypes (Fig. 1)

Endometrioid carcinoma is identified by its complex, branching glandular or villoglandular architecture composed of back to back glands with no intervening stroma. The cells lining the glands are crowded, stratified, columnar cells with mild to moderate nuclear atypia, inconspicuous nucleoli and eosinophilic cytoplasm. One characteristic feature of this tumour is the smooth contour of the lumen of the glands. The precursor lesion (atypical hyperplasia) may be seen. Variants of endometrioid carcinoma include those with squamous or secretory differentiation. The squamous component can be in the form of squamous morules or at the stromal interface. It is often a helpful feature in identifying the tumour as endometrioid carcinoma, but is not considered when grading the tumour. Secretory differentiation resembles the endometrium in the secretory phase of the menstrual cycle and occurs in less than 2% of endometrioid carcinomas [9].

Fig. 1
figure 1

Histotypes of endometrial carcinoma. a Low-grade endometrioid adenocarcinoma. b High-grade endometrioid adenocarcinoma. c Serous carcinoma. d Clear cell carcinoma. e Dedifferentiated carcinoma, showing low-grade (endometrioid) and high-grade (undifferentiated) components. f High-grade (undifferentiated) component of dedifferentiated carcinoma. a–e H&E, 100× magnification; f H&E, 200× magnification

Full size image

Serous carcinoma is distinguished from endometrioid carcinoma by its marked nuclear pleomorphism, prominent nucleoli and scant cytoplasm. It characteristically has a papillary architecture, but can be solid and/or microcystic. In contrast to the round, smooth glandular lumens in endometrioid carcinoma, the luminal surfaces in serous carcinoma are irregular and slit-like. Mitoses are prominent. Serous carcinomas typically arise in a polyp or atrophic endometrium.

Paradigmatic low-grade endometrioid carcinomas and prototypic serous carcinomas are easily recognized based on their microscopic appearance, that correlates with different mutational and DNA expression profile [10, 11]. However, based on morphology alone, there is a subset of high-grade glandular tumours with ambiguous features that are difficult to classify (i.e. as endometrioid or serous) without ancillary tests [12].

An uncommon histotype is clear cell carcinoma. This is a high-grade neoplasm and is considered to be a Type II endometrial carcinoma using Bokhman’s classification, as it is not associated with increased estrogen. The tumours have tubulocystic, papillary or solid growth, and polygonal or hobnail cells with marked nuclear pleomorphism, conspicuous nucleoli and clear cytoplasm (although eosinophilic cytoplasm may be present). Eosinophilic extracellular globules or hyaline bodies are also a characteristic feature, present in approximately two thirds of these tumours. Similar to serous carcinoma, these tumours arise in polyps or atrophic endometrium. Clear cell carcinoma can be mistaken for serous carcinoma or endometrioid carcinoma with clear cell squamous or secretory differentiation [9]. Rigorous criteria are recommended before establishing the diagnosis of clear cell carcinoma, with special emphasis on the typical architectural patterns.

A mixed carcinoma category recognizes the occurrence of tumours with heterogeneity. In these tumours, there must be two histological components with the second component comprising at least 5% of the tumour. One component must be either serous or clear cell carcinoma [9]. Rigorous histomorphological criteria are also highly recommended in the diagnosis of mixed carcinoma. There is considerable variability in the frequency with which mixed carcinoma is diagnosed; as with ovarian mixed carcinomas [13], it is evident that mixed carcinomas of the endometrium are clonal and the immunophenotype and molecular features are uniform throughout, in most cases [14, 15]. As such, “true” mixed carcinomas, when strictly defined as having two components that are distinct based on both light microscopy and molecular biomarker expression, are relatively uncommon.

Undifferentiated endometrial carcinomas are those in which no differentiation is present. The tumour contains highly mitotic, small- to intermediate-sized cells with condensed chromatin, arranged in dyshesive sheets. Dedifferentiated carcinoma is a variant of undifferentiated carcinoma with both an undifferentiated component and a component of well- or moderately differentiated endometrioid carcinoma [9].

The WHO histological classification system correlates well with the natural history of the disease and patient prognosis [16,17,18,19]. However, interobserver variability in classifying endometrial carcinomas remains problematic, particularly in the subset of high-grade carcinomas, including the grey zone between high-grade endometrioid and serous carcinomas [5,6,7,8]. This has far-reaching implications as inaccurate classification can impact our understanding of the natural history of these entities and is reflected by the fact that some studies examining outcomes of patients with different histotypes of high-grade endometrial carcinomas have shown differences in patient outcome, while others have not [20,21,22,23,24,25]. In a study of 187 high-grade endometrial carcinomas including FIGO grade 3 endometrioid, serous and clear cell carcinoma, Soslow et al. showed that when several variables were controlled for, high-grade endometrial cancers of different histologic subtypes had similar clinical outcomes [24]. Voss et al. showed similar clinical presentation and poor survival when comparing high-grade endometrial carcinomas of different histotypes. Based on this, they suggested that FIGO grade 3 endometrioid cancer should be regarded as a Type II cancer and treated with similar adjuvant chemotherapy as serous and clear cell carcinoma [25]. In contrast, other studies showed a significant survival difference between grade 3 endometrioid carcinoma and serous carcinoma (75 vs. 41% at 5 years, respectively, in the study by Boruta et al.) [23, 26].

Substantial interobserver variability in histotype diagnosis makes it impossible to enrol patients in clinical trials based on histotype, or deliver histotype-specific treatment consistently. Histotype (or any variable that cannot be reproducibly diagnosed) is of limited use in guiding more individualized treatment, given that treatment recommendations will vary depending on the reviewing pathologist, rather than the underlying tumour biology. It is therefore important to improve reproducibility of subclassification of endometrial carcinoma and studies on molecular markers that could guide improved endometrial carcinoma subclassification have been undertaken with this aim in mind. The initial focus was on single markers, using immunohistochemistry or mutational analysis, to improve histotype-based assignment. It should be acknowledged, however, that a purely molecular-based classification may supplant histotype, as has happened with breast and lung carcinomas, where the molecular markers, whether assessed by immunohistochemistry, fluorescence in situ hybridization (FISH) or mutational analysis, have far greater importance in guiding treatment than histopathological variables [27, 28].

Immunohistochemical markers

p53

The characteristic molecular events in serous carcinogenesis include mutation of TP53, while endometrioid carcinomas are associated with mutations in PTEN. In particular, p53 immunostaining can serve as an aid in differential diagnosis of endometrioid and serous carcinoma, yet it undoubtedly remains true that approximately 30% of grade 3 endometrioid carcinomas (morphologically typical, with predominantly solid growth and squamous metaplasia) show abnormal, mutant-pattern, p53 staining [29]. It is not recommended that p53 alone be used when the differential diagnosis includes serous versus high-grade endometrioid carcinoma as it is not helpful in this context [30]. In low-grade endometrioid carcinomas (FIGO grade 1 or 2), p53 expression is rarely abnormal [30]; however, it is relatively uncommon for there to be a morphological problem in distinguishing low-grade (i.e. grade 1 or 2) endometrioid carcinoma from serous carcinoma [12].

p53 immunohistochemical staining should be reported as abnormal (mutated) or wildtype. Abnormal p53 staining occurs when there is strong nuclear staining in all (or almost all) of the tumour or complete loss of staining. Wildtype p53 staining is weak and patchy (Fig. 2). A third pattern of abnormal p53 immunostaining, associated with mutations that impact on the nuclear translocation domain of the protein, is moderate to intense cytoplasmic staining without strong nuclear staining, but this pattern is uncommon [31]. Occasionally, there can be heterogeneous staining for p53, with different clones showing different staining patterns (Fig. 2). Clear cell and undifferentiated or dedifferentiated carcinomas are usually p53 wildtype. In serous carcinoma, p53 is mutated and shows intense diffuse staining in 80–90% of tumours and complete absence of staining in 10% of tumours [29].

Fig. 2
figure 2

p53 immunohistochemical staining of endometrial carcinomas. a Mutant p53 staining, strong and diffuse (overexpression pattern). b Mutant p53 staining, null (loss of expression pattern—note positive staining of benign cells, which serves as an internal control). c Wildtype p53 staining, weak and patchy. d Heterogeneous p53 staining, with both overexpression and loss of expression. This pattern is encountered infrequently

Full size image

Estrogen receptor

Estrogen receptor (ER) is expressed in more than 95% of low-grade endometrioid carcinomas [32]; however, grade 3 endometrioid carcinomas can lack expression (15–50%) [33]. It was formerly thought that ER is usually negative in serous carcinomas; however, with more sensitive immunostaining protocols, more than half of serous carcinomas can show positivity [33]. Clear cell and undifferentiated or dedifferentiated carcinomas are typically negative for ER [3, 30, 34].

p16

Although HPV is not involved in the pathogenesis of endometrial carcinoma, p16 immunohistochemical staining of endometrial tumours may be useful. It can be diffusely and strongly positive in serous carcinoma (Fig. 3), with up to 82% specificity in some studies [30, 33, 35], but further validation, looking specifically at the differential diagnosis of grade 3 endometrioid and serous carcinoma, is needed. Endometrioid carcinoma usually exhibits focal positivity in less than 50% of tumour cells (Fig. 3) [36]. Occasional cases of mucinous adenocarcinoma of the endometrium or endometrioid adenocarcinoma with mucinous differentiation can show strong diffuse p16 immunoreactivity [37]. Clear cell carcinomas show a similar staining pattern to endometrioid carcinoma.

Fig. 3
figure 3

p16 immunohistochemical staining of endometrial carcinomas. a diffuse, strong, nuclear p16 staining in serous carcinoma. b Focal patchy p16 staining in endometrioid carcinoma

Full size image

Napsin A/HNF-1β

Napsin A and HNF-1β have been used to identify clear cell carcinoma (> 90% specificity) and usually give a diffuse moderate to strong staining pattern [38,39,40]. These markers appear to be less specific for clear cell carcinoma of the endometrium, than for clear cell carcinoma of the ovary, where they are well validated, as they can be positive in a subset of other endometrial carcinoma histotypes including endometrioid and serous [38, 39]. Alpha-methylacyl-CoA racemase (AMACR) has also been suggested to be useful for diagnosing clear cell carcinomas, in some studies.

Abnormal SWI/SNF subunit expression: INI1 (SMARCB1), BRG1 (SMARCA4) and BAF250a (ARID1A)/ARID1B immunostaining—a feature of dedifferentiated endometrial carcinoma

Recent studies have examined the immunohistochemical and molecular profiles of undifferentiated and dedifferentiated carcinomas of the endometrium [41,42,43]. While these tumours often lose PAX8 and ER expression (seen in over 60% of these tumours), it has also been found that a large portion lose INI1 and BRG1 expression [41,42,43,44,45,46]. Aberrant p53 staining can also occur in a small subset of these BRG1/INI1 deficient tumours; however, wildtype p53 expression predominates [41,42,43,44].

PTEN

Loss of expression of the tumour suppressor, PTEN (phosphatase and tensin homolog), occurs most frequently in endometrioid carcinoma, but can also occur in undifferentiated and mixed carcinomas [30, 47]. PTEN immunohistochemistry detects tumours with genetic loss of PTEN, but also with functional PTEN loss as a consequence of epigenetic mechanisms, and may be superior to gene sequencing for identifying tumours with PTEN abnormalities [47]. PTEN immunohistochemistry has been reported to show variable results in different laboratories. However, recent studies have proposed new protocols designed to improve inter-laboratory performance [48].

Other markers and panels of antibodies

Immunohistochemistry may be used as an aid in histotype diagnosis and subclassification of endometrial carcinoma; however, no single marker is completely sensitive or specific for a given histotype [33]. Lack of expression of mismatch repair genes has been suggested as evidence against the diagnosis of serous carcinoma. Additional proteins that have been proposed in the differential diagnosis between endometrioid and serous carcinomas are HER2, claudin 3 and 4, FOLR-1, HMGA-2, cyclin E, IMP2 and IMP3, but none of them has shown good sensitivity and specificity. p53 is the most extensively validated immunomarker for serous versus endometrioid, but, as noted previously, is limited in that abnormal p53 expression is seen in a significant minority of high-grade endometrioid carcinomas.

Panels of immunostains have been recommended and have been shown to improve interobserver agreement regarding histotype diagnosis, but there is no agreement on the best composition of such a panel, and how to handle cases where the results of individual stains from the panel are not consistent [5, 49, 50]. The most frequent proteins included in these panels are p53, p16 and PTEN [51].

Genetic markers

CTNNB1 (β-catenin)

Several studies have identified the presence of β-catenin mutations in endometrioid carcinoma. Mutation can occur in up to 66% of low-grade endometrioid tumours [30, 52].

ARID1A

ARID1A mutations are acquired in both low- and high-grade endometrioid carcinomas, occurring in 40–46.7% of low-grade tumours and up to 60% of high-grade tumours [49, 52, 53]. It has also been found to be mutated in up to one quarter of clear cell carcinomas [39, 54]. These mutations are rarely seen in serous carcinoma.

PTEN

PTEN mutations are the most common recurrent genetic event in endometrioid carcinomas. Clear cell carcinoma also have PTEN mutations at a high frequency, whereas these are rarely found in serous carcinoma [30, 39, 52].

PIK3CA

PIK3CA mutations can be found in both endometrioid and serous carcinomas, occurring in 30–60% of endometrioid carcinomas and 24–40% of serous carcinomas [29, 55,56,57,58]. High-grade endometrioid carcinoma shows a significantly higher PIK3CA mutation frequency when compared with serous carcinoma. Furthermore, there is an association between ARID1A and PTEN/PIK3CA mutation, further strengthening the relationship between these mutations and the endometrioid histotype [52].

TP53

While high- and low-grade endometrioid carcinomas have a somewhat similar mutation profile, several studies have shown a significantly increased TP53 mutation frequency with increased FIGO grade of endometrioid carcinoma [29, 49, 52, 55]. Furthermore, TP53 mutations occur much more frequently in serous carcinomas when compared to endometrioid carcinoma and is the hallmark of this histotype, being present in 80–90% or more of these tumours [29, 55, 58]. Interestingly, there are some differences in the spectrum of TP53 mutations between endometrioid and serous carcinomas, with hot-spot mutations more frequent in serous tumours [59].

PPP2R1A

Up to 41% of serous carcinomas have a missense mutation in PPP2R1A, while this occurs in a much smaller fraction (5%) of endometrial endometrioid carcinomas [55, 60, 61].

Molecular markers have aided in the classification of endometrial carcinoma and there are strong correlations between histotype and mutations in single genes; however, just as with immunohistochemistry, no single marker is completely sensitive or specific for classification (Table 2).

Table 2 Summary of single immunohistochemical and molecular markers in endometrial carcinoma histotypes
Full size table

Gene panel testing

The introduction of next-generation sequencing into clinical practice has enabled routine sequencing of multiple genes for the same cost as a single gene. This makes possible the use of a panel of genetic markers as an aid in classification, with the goal of decreased interobserver variation and greater diagnostic accuracy [52, 62]. For example, McConechy et al. studied 393 endometrial carcinomas with targeted exon sequencing of nine genes [52]. PTEN and ARID1A are associated with endometrioid type, while TP53 and PPP2RIA are associated with serous type, so that the presence of either or both of the former mutations and absence of the latter (i.e. TP53 and PPP2R1A) support a diagnosis of endometrioid carcinoma. Such an approach, of sequencing a small number of genes, is not able to reliably classify endometrial carcinoma, however, as there are large numbers of indeterminate cases where the mutation profile does not permit unequivocal assignment of histotype and occasional cases with outright discordance between histologic diagnosis and mutation profile. Various molecular profiles of endometrial carcinoma histotypes are shown in Fig. 4. Intra-tumoural heterogeneity has also been suggested as a potential problem, since some important mutational driver events may be restricted to specific regions of the tumour [63].

Fig. 4
figure 4

Mutation profiles of endometrial subtypes. a Low-grade endometrioid carcinoma, including grade 1 and 2 tumours. b High-grade endometrioid carcinoma, grade 3. c Serous carcinoma. d Carcinosarcoma. e Undifferentiated and mixed histology subtypes: undifferentiated carcinomas (a), mixed low-grade endometrioid carcinoma with serous carcinoma (b), mixed endometrioid and clear cell carcinoma (c) and mixed serous and clear cell carcinoma (d). Rows indicate genes and columns represent tumour cases. Coloured bars indicate mutations including missense, truncating, indels and splice site mutations. Grey bars indicate no mutations were detected. +carcinosarcomas with heterologous differentiation elements. *serous carcinoma outliers with ARID1A mutations. #low-grade endometrioid carcinoma and high-grade endometrioid carcinoma mutation outliers with serous-type mutations (TP53 or PPP2R1A). (Reproduced with permission. McConechy MK, Ding J, Cheang MCU, et al. (2012) Use of mutation profiles to refine the classification of endometrial carcinomas. J Pathol 228:20–30. doi: https://doi.org/10.1002/path.4056)

Full size image

The Cancer Genome Atlas (TCGA) genomic-based classification of endometrial carcinoma

A multiplatform analysis of 373 endometrial carcinomas through the TCGA resulted in a proposed molecular classification of endometrial carcinoma. Four groups were described based on integrated genomic architecture rather than single genetic mutations, and this subclassification has direct clinical and prognostic implications [55, 64]. The four groups are (1) ultramutated/polymerase ε mutated (POLE), (2) hypermutated/microsatellite instability, (3) low-copy number abnormalities and (4) high-copy number abnormalities (Table 3). The ultramutated/POLE group comprised approximately 10% of the studied tumours and was frequently of high-grade endometrioid histologic subtype, with few somatic copy number alterations but a very high mutation burden (ten times the number of mutations, on average, as hypermutated tumours and 100-fold more mutations than the tumours in the low-copy number group). Furthermore, these tumours had a significantly better prognosis than the other three TCGA groups [65,66,67]. The hypermutated/microsatellite instability (MSI) group is characterized by the histologic features of Lynch syndrome-associated carcinomas, including tumour heterogeneity, ambiguous histology and tumour-infiltrating lymphocytes [68, 69]. High-somatic-copy number abnormalities were seen in serous-like tumours and correspond broadly to the Type II tumours described by Bokhman, while the low-somatic-copy number tumours correspond to Type I endometrial carcinomas. Notably, however, up to 35% of high-grade endometrioid carcinomas also had high-copy number alterations and were therefore more accurately classified, genomically, as being within the high-copy number group (despite having unequivocal endometrioid features, including squamous differentiation in some instances) [55].

Table 3 Genomic-based classification of endometrial carcinoma by The Cancer Genome Atlas (TCGA)
Full size table

This genomic approach to classification is potentially more robust than single-gene or single-protein analysis, has shown significant correlation with patient outcome and is clinically actionable. Unfortunately, however, due to logistic and resource constraints, such as the need for fresh frozen tissue, cost, long turn-around time and lack of applicability to biopsies or curettings, so that classification is not available before definitive surgical treatment, this approach is not currently applicable in routine practice. Identification of surrogate markers that accurately reflect molecular subtype is the only feasible way to overcome this barrier.

Development of a clinically applicable surrogate for TCGA classification

More cost-effective and convenient methods, allowing testing on formalin-fixed paraffin-embedded tissue, are required for genomic-based classification of endometrial carcinomas into molecular subtype to enter routine practice. Two groups, working independently, have identified the same approach to molecular classification [64, 70, 71]. Abnormal p53 staining by immunohistochemistry is a surrogate for identifying tumours with high-copy number alterations, immunohistochemistry for mismatch repair proteins (MLH1, MSH2, MSH6, PMS2) can be used to detect hypermutated/MSI tumours [72], sequencing for mutations in the POLE exonuclease domains is a surrogate for identifying the ultramutated/POLE group [61], and the copy number low group consists of those tumours lacking any of the above molecular features.

These methods are more widely accessible and inexpensive and can be performed expeditiously and on small samples, with results based on biopsy/curettings being highly concordant with results based on the hysterectomy specimen [73]. Although further validation is required, the evidence to date indicates the potential for this classification to be highly reproducible. For example, p53 immunohistochemistry can show a very high correlation with TP53 mutation status [31]; however, this is not being achieved in all clinical laboratories, and further improvements in quality of staining and interpretation should be sought [74]. Another caveat is that this classification does not apply to all endometrial carcinomas, specifically dedifferentiated/undifferentiated endometrial carcinoma. A unique molecular profile with frequent aberrant mismatch repair protein expression and loss of SWItch-sucrose non-fermentable (SWI/SNF) protein expression characterizes these tumours, and they do not fit into the TCGA-based classifier [42,43,44].

This TCGA surrogate seems to be particularly informative in the group of high-grade endometrial carcinomas, in the spectrum of tumours that includes high-grade endometrioid carcinoma and serous carcinoma. It is also a good tool to identify patients with tumours (ultramutated and hypermutated) that may benefit from immunotherapy [75, 76].

Summary and conclusions

We have seen the evolution of endometrial cancer classification from being purely based on anatomical location to histologic cell type-based classification, to classification incorporating ancillary molecular testing, such as gene panel testing and genomic analysis. Molecular classification holds promise for more accurately subtyping endometrial carcinoma to better reflect patient prognosis and outcome and may be the mainstay of endometrial carcinoma classification in the future.