The rise of a novel classification system for endometrial carcinoma; integration of molecular subclasses

(1)

*Correspondence to: Tjalling Bosse, Leids Universitair Medisch Centrum (LUMC), Department of Pathology, Albinusdreef 2, 2300 RC Leiden, The Netherlands. E-mail: t.bosse@lumc.nl

Abstract

Endometrial cancer is a clinically heterogeneous disease and it is becoming increasingly clear that this heterogeneity may be a function of the diversity of the underlying molecular alterations. Recent large-scale genomic studies have revealed that endometrial cancer can be divided into at least four distinct molecular subtypes, with well-described underlying genomic aberrations. These subtypes can be reliably delineated and carry significant prognostic as well as predictive information; embracing and incorporating them into clinical practice is thus attractive. The road towards the integration of molecular features into current classification systems is not without obstacles.

Collaborative studies engaging research teams from across the world are working to define pragmatic assays, improve risk stratification systems by combining molecular features and traditional clinicopathological parameters, and determine how molecular classification can be optimally utilized to direct patient care. Pathologists and clinicians caring for women with endometrial cancer need to engage with and understand the possibilities and limitations of this new approach, because integration of molecular classification of endometrial cancers is anticipated to become an essential part of gynaecological pathology practice. This review will describe the challenges in current systems of endometrial carcinoma classification, the evolution of new molecular technologies that define prognostically distinct molecular subtypes, and potential applications of molecular classification as a step towards precision medicine and refining care for individuals with the most common gynaecological cancer in the developed world.

Keywords: endometrial cancer; molecular classification; TCGA; prognostic biomarkers; molecular diagnostic test

Received 29 November 2017; Revised 21 December 2017; Accepted 27 December 2017

No conflicts of interest were declared.

Introduction

It is an exciting era in cancer care; we are moving beyond treatment of disease according to anatomic site, beyond dependence on ‘histomorphology’ alone, and towards tailoring therapy according to molecular features within an individual’s tumour. This progression has been especially critical in endometrial carcinoma (EC), where current standard pathological classification by histological subtype, grade, disease extension, and lymphovascular space invasion (LVSI) is considered highly subjective and has challenges in reproducibility. In contrast to the high inter-observer agreement in histological subtype assignment documented for epithelial ovarian cancers [1,2], major disagreement or complete lack of consensus of histotype assignment has been observed in approximately one-third of high-grade EC cases, with substantial immunophenotype diversity noted [3–7]. Disagreement in grade assignment, both between pathologists and between diagnostic endometrial specimens and final hysterectomy specimens, is

also common [8–10]. Histology and grade, in addition to stage (determined after surgical staging proce- dures), form the basis for risk group stratification in EC [11–14], directing which women should receive adjuvant therapy [radiation (vaginal or external beam), chemotherapy or both] [11,14–17]. Additional features, such as patient age, depth of myometrial invasion, endocervical stromal involvement, lymph node status, and LVSI, are increasingly being incorporated [18–20]. However, assessment of the five most commonly used risk stratification systems in EC demonstrated that none are highly accurate at stratifying risk of recurrence or metastases [21]. This may be sec- ondary to the aforementioned irreproducibility of histotype and grade assignment such that the same woman might receive very different treatment and have a very different clinical outcome depending on how her tumour was interpreted. In addition, even within a defined grade and histotype category, e.g. grade 3 EC, tremendous diversity in clinical outcomes is observed. Biological diversity of tumours is not recognized in the current

(2)

tumours.

We need to do better if we are to improve outcomes for women with EC. A more reproducible, objective, biologically informative system to classify tumours and enable stratification of past and future clinical trials is urgently needed. Given what has been described as a global epidemic of increasing incidence and mortality of EC in the developed world, it is time to focus our atten- tion on a ‘novel classification system’ for this disease site. Herein, we describe an evolution of EC classification from conventional pathology assignment through comprehensive genomic/proteomic/transcriptomic characterization in The Cancer Genome Atlas (TCGA) project to new pragmatic methods of integrating molecular classification into conventional pathology assessment. We share new applications and research advances as well as challenges to address as we move forward.

TCGA classification of endometrial carcinomas

The TCGA network performed a multiplatform analysis of EC [22]. This was a landmark study in many ways, providing a new framework from which to approach EC going forward. The TCGA used a wealth of genomic information and exploited somatic copy number alterations (SCNAs) and tumour mutation burden (TMB) to propose four distinct EC classes with distinct genomic aberrations, breaking with the traditional dualistic view [23]. The first molecular subgroup consisted of copy number-stable, but ultra-mutated EC with recurrent mutations in the exonuclease domain of DNA poly- merase epsilon (POLE), a gene involved in nuclear DNA replication and repair [24–27]. These tumours had extraordinarily high somatic mutation frequencies, exceeding 100 mutations per megabase (Mb) in the majority of cases [28]. Subsequent studies showed that, histologically, POLE-mutant ECs are typically high-grade endometrioid EC (EEC) with a superficial broad front invasion pattern and the presence of tumour giant cells and prominent tumour-infiltrating lymphocytes (TILs) [29–32] (Figure 1). It has also been noted that ECs that are morphologically difficult to classify (histomorphologically ambiguous) are more likely to be POLE-mutant [33]. The second molecular EC sub- group reported by the TCGA was hypermutated, with microsatellite instability (MSI) due to dysfunctional mismatch repair (MMRd) proteins: MLH1, PMS2, MSH2, and MSH6 [22,34]. The predominant underly- ing mechanism was epigenetic silencing of MLH1 (by promoter hypermethylation). In a minority of EC the MMRd is caused by somatic or germline mutations in the MMR genes, the latter being diagnostic for Lynch syndrome (LS), an autosomal-dominant cancer suscep- tibility disorder associated with a markedly increased risk of colorectal carcinoma (CRC) and EC [35,36].

ECs with MSI or MMRd are mostly endometrioid; however, non-endometrioid subtypes have been described [38–40]. Like POLE-mutant ECs, these tumours typ- ically display TILs and peri-tumoural lymphocytes.

Furthermore, a so-called ‘microcystic elongated and fragmented (MELF)’ pattern of invasion and LVSI have also been associated with MMRd EC [32,41–43].

The third molecular subgroup was genomically rel- atively stable, MMR-proficient, and had a moderate number of mutations, mostly within the PI3K/Akt and Wnt signalling pathways. This subgroup was almost exclusively composed of EECs with oestrogen (ER) and progesterone receptor (PR) positivity. The fourth subgroup had high SCNA, very similar to high-grade serous ovarian carcinomas (HGSOCs), and frequent TP53 mutations (92%). Morphologically, these were high-grade (grade 3) ECs, including most serous ECs, but interestingly EECs (26%) were also classified in this group, justifying the now commonly used term

‘serous-like cancers’.

Post-TCGA pragmatic molecular classification of endometrial carcinomas: clinically applicable surrogate markers

The TCGA applied methods that are too costly and cumbersome for widespread implementation into routine clinical practice. Post-TCGA, two research teams have developed more pragmatic approaches that can enable molecular subtyping in pathology laboratories.

These molecular classification systems are feasible on formalin-fixed, paraffin-embedded (FFPE) material and designate four molecular subclasses of ECs with distinct prognostic outcomes, analogous but not identical to the TCGA subgroups [41,42]. A summary of the character- istics of each molecular subtype according to TCGA criteria and subsequent post-TCGA publications is given in Table 1. Importantly, there is little intratumoural heterogeneity in molecular subtype assignment when testing these surrogate markers and they can be assessed reliably on diagnostic endometrial specimens as obtained by office biopsy (e.g. pipelle) or dilatation and curettage [44,45].

Diagnostic tests to identify ultra-mutated POLE -mutated ECs

Clinically relevant POLE exonuclease domain muta- tions have been identified adjacent to highly conserved exo motifs, required for proper proofreading activity.

The most frequent somatic mutations in the proof- reading domain in human cancers are P286R, V411 L, and S459F. Sequencing methods that limit analysis to these three hotspots would identify 67–92% of all POLE-mutant cancers [42,46]. However, for clinical implementation, in order not to miss detecting this

(3)

Figure 1. POLE-mutant ECs have a diverse morphological spectrum. Different histological appearances of three independent POLE-mutant ECs are shown. (A) Grade 1 EEC. (B) Grade 3 EEC with prominent tumour-infiltrating lymphocytes (TILs). (C) Morphologically ambiguous EC with serous-like features and tumour giant cells. (All images ×20 objective magnification.)

prognostic subgroup, it may be preferred to cover the complete exonuclease domain. Sanger sequencing is an acceptable option, as well as panel-based next-generation sequencing (NGS), incorporating POLE exons 9–14 or 9, 13, and 14.

Regardless of the extent to which POLE is assessed, it will be critical to restrict the POLE designation to tumours associated with pathogenic POLE variants.

This is particularly challenging in cases with rare variants, of which the effect on proofreading function is unclear. For this reason, it has been proposed to only classify POLE variants as pathogenic if the variant has been demonstrated to result in a high mutational frequency (greater than 100 mutations per Mb), an increased proportion of C→ A transver- sions (exceeding 20% of all substitutions), and POLE mutation at a residue that is recurrently mutated in cancer [28].

Diagnostic tests to identify mismatch repair-deficient (MMRd), MSI-high ECs

Two common laboratory assessment tools for the identification of tumours with deficient mismatch DNA repair include (i) MSI assays and (ii) immunohistochemistry (IHC) for the presence/absence of MMR protein expression. Historically, both of these tests have been optimized and designed to triage CRC patients, followed by an assay that tests the methylation status of the MLH1 promoter, to identify LS patients [35,36]. MMR proteins heterodimerize into MutSa (MSH2–MSH6) and MutLa (MLH1–PMS2) [47,48], where MSH2 and MLH1 sta- bilize MSH6 and PMS2, respectively [49]. The prognostic and predictive significance of MMR/MSI status, as well the value of identifying patients at higher risk of LS (increased screening, risk-reducing interventions, referral of family members), has prompted universal testing for all CRC patients [50,51]. Given the observation that EC is the most common presenting tumour in women with LS, routine testing by MMR-IHC in all ECs is increasingly being advocated [52–57]. As the MSI assay is more expensive and requires DNA purifi- cation, MMR-IHC has become the preferred test in most pathology laboratories. Furthermore, the IHC approach is reliable, with reported concordance rates with the MSI

assay ranging from 86% to 99% [58,59]. Recently, subclonal/regional loss of MMR expression has been appre- ciated [58,60] (Figure 2). This indicates that MMRd can occur during the progression of disease [58]. The concordance between MMR-IHC and MSI assays is greater when subclonal/regional loss of expression is accounted for. MMR-IHC is also the preferred approach for LS triage cascade testing, because it allows identification of the affected gene. Moreover, MMR-IHC is a more reliable method for the identification of ECs with MSH6 mutations than MSI assay [58]. The triage cas- cade includes an assay in which the methylation status of the promoter of MLH1 is assessed. Given the afore- mentioned binding properties of mismatch repair het- erodimer complexes, performing IHC for only MSH6 and PMS2 saves cost and has been demonstrated to be both sensitive and specific [58,59,61].

Diagnostic tests to identify copy number-high analogous subgroup

Given the high frequency of TP53 mutations (92%) [22]

in the copy number-high molecular subgroup, analysing the mutational status of TP53 appears to be sufficient to identify a poor prognosis analogous molecular subclass.

TP53 mutations elicit two functional consequences: gain and loss of function. Gain of function is due to non- synonymous mutations, with accumulation of nuclear p53 in the nucleus due to a failure in the degradation of the protein by MDM2-induced ubiquitination [62].

This results in strong and diffuse nuclear overexpres- sion, easily identified by IHC. Loss of p53 function is present in nonsense mutations, indel, and splicing mutations, which interfere with correct protein translation.

These mutations are associated with loss of p53 expression (the so called ‘null-pattern’) or very rarely cyto- plasmic accumulation [63,64]. Recently, the two func- tional effects of TP53 mutations have been described to correlate well with IHC staining in HGSOC, demon- strating a sensitivity of 0.96 and a specificity of 1.00 [64]. A subclonal/regional pattern can be identified in some cases, showing aberrant p53 expression in a dis- crete geographical area, associated with TP53 mutations [44] (Figure 3). Currently, inter-observer and concordance studies in p53 assessment are on-going, of which

(4)

Table1.Clinicopathologicalandmolecularcharacteristicsofthemolecularsubgroups(TCGAanalogous) POLE-mutantMMRdNSMPp53-aberrant (i.e.POLEEDM)(i.e.MSI)(i.e.p53wt)(i.e.p53abn,p53-mutant) Mutationalfrequency>100mutations/Mb100–10mutations/Mb<10mutations/Mb<10mutations/Mb Somaticcopy-number alterationsVerylowLowLowHigh TopfiverecurrentgenePOLE(100%)PTEN(88%)PTEN(77%)TP53(92%) mutations(%)DMD(100%)PIK3CA(54%)PIK3CA(53%)PIK3CA(47%) CSMD1(100%)PIK3R1(42%)CTNNB1(52%)FBXW7(22%) FAT4(100%)RPL22(37%)ARID1A(42%)PPP2R1A(22%) PTEN(94%)ARID1A(37%)PIK3R1(33%)PTEN(10%) Associatedhistological featuresEndometrioidEndometrioidEndometrioidSerous Grade3Grade3Grade1–2Grade3 AmbiguousmorphologyLVSIsubstantialSquamousdifferentiationLVSI BroadfrontinvasionMELF-typeinvasionER/PRexpressionDestructiveinvasion TILs,peri-tumouralLymphocytesTILs,Crohn’s-likeperi-tumouralreactionHighcytonuclearatypia Gianttumouralcells GianttumouralcellsLowuterinesegmentinvolvementHobnailing Slit-likespaces Associatedclinical featuresLowerBMIHigherBMIHigherBMILowerBMI Earlystage(IA/IB)LynchsyndromeAdvancedstage EarlyonsetLateonset Prognosisinearly stage(I–II)ExcellentIntermediateExcellent/intermediate/poorPoor DiagnostictestSanger/NGS(exons9,13,14or9–14)MMR-IHC(MLH1,MSH2,MSH6,PMS2)p53-IHC TumourmutationburdenMSIassayNGS TumourmutationburdenSCNA Suggestedtreatment optionsinrecur- rent/metastatic disease*

CheckpointinhibitorsCheckpointinhibitorsHormonaltherapymTORinhibitorsSmallmoleculeactivatorsofp53 PARPi LVSI:lymphovascularspaceinvasion;MELF:microcysticelongatedandfragmentedtypeofinvasion;TILs:tumour-infiltratinglymphocytes;BMI:bodymassindex;IHC:immunohistochemistry;MSI:microsatelliteinstability. *Theseareproposedtreatmentoptionsinrecurrentand/ormetastaticdiseasethatcanbeexplored.

(5)

Figure 2. Example of subclonal/regional loss of MMR proteins. (A, B) EC with a discrete area with loss of PMS2 expression. (A) ×2 objective magnification; (B) ×10 objective magnification. MLH1 expression (not shown) showed an identical pattern. (C) Subsequent MSI assay of both tumour areas. The area with intact PMS2 expression (#) was microsatellite-stable (MSS), whereas the area with loss of PMS2 expression (*) was microsatellite-unstable (MSI).

the results are eagerly awaited, as p53 IHC is an attrac- tive low-cost surrogate to determine TP53 mutational status.

Diagnostic test to identify the copy number-low subgroup

ECs that lack pathogenic mutations in POLE and TP53, and are microsatellite-stable (MSS), by default fall within the category of tumours analogous to the ones the TCGA had annotated as ‘copy number-low’. We feel that this annotation is not accurate and somewhat misleading, since there are focal SCNAs. In fact, this subgroup of ECs has more SCNAs than MMRd and POLE-mutant ECs [65]. As a result, subsequent research teams have come up with alternative molecular defi- nitions and associated names, such as ‘p53wt group’, referring to the complete lack of TP53 mutations in this subgroup. This name, however, is also imperfect, as TP53 mutations may also be encountered in the MSI and POLE-mutant ECs, albeit at low frequency. There- fore, to this end, and in the absence of a better alternative, this group has been referred to as ‘no specific molecular profile/no surrogate marker profile, NSMP’, as this is the group of ECs that remain after exclud- ing those ECs with a specific profile or with a surrogate marker.

Novel testing strategies for POLE and MMRd sub- types such as assessment of the TMB [37,66,67] may have the advantage of identifying the overriding phenotype, overcoming the genotyping issues such as rare variants with uncertain pathogenicity. This may become

an attractive option to simplify laboratory workflow in the future.

Prognostic refinement

The molecular classification of the TCGA provided an attractive new context to interpret ECs. Four prognostic subgroups were identified, not all of which would have been predicted based on traditional clinicopathological factors. As the TCGA classification was the result of unsupervised clustering of genomic aberrations of a small and clinically heterogeneous cohort, it was insuf- ficiently powered to address clinical utility. Subsequent studies by the Vancouver and PORTEC groups have corroborated the prognostic value in large cohorts with mature follow-up data [41,42,68,69].

The Vancouver cohorts were unselected, and con- tained EECs (83%), serous, mixed, and undifferentiated ECs, with a mixture of all grades and stages, in a dis- tribution weighed slightly towards higher-risk tumours reflecting their status as a tertiary referral cancer cen- tre. A discovery phase, testing 16 different models of surrogate molecular marker testing, yielded a pragmatic

‘best’ molecular algorithm that was then confirmed and validated according to the Institute of Medicine Guidelines [42,68,70]. This molecular classifier (coined ProMisE, for Proactive Molecular Risk Classifier for Endometrial cancer) stratified EC into four prognostic groups mirroring TCGA outcomes. Patients with POLE-mutant ECs had an excellent prognosis, with a

(6)

Figure 3. Example of subclonal/regional aberrant expression of p53. (A, B) EC with areas of p53 wild-type IHC expression (#) and areas with aberrant expression of p53 (*). (A) ×2 objective magnification; (B) ×10 objective magnification. (C) Sanger sequencing results of both microdissected areas, showing the presence of a TP53 (R248Q, c.743G> A) (arrow) in the area with aberrant expression, whereas the area with wild-type expression did not carry this mutation.

hazard ratio (HR) of 0.15 for recurrence-free survival (RFS), while p53abn EC had a HR of 1.64 (p = 0.024) compared with the p53 wt/NSMP group [42].

The PORTEC cohort was different, in that it assessed ECs from women included in the PORTEC-1 and -2 trials, therefore benefiting from a focused analysis of intermediate and intermediate–high risk patients [71,72]. An integrated molecular model resulted in an improved risk stratification compared with a model using established clinicopathological factors. In this large trial cohort, patients with POLE-mutant EC had no locoregional recurrences, and p53-mutant ECs had a HR of 6.79 (p< 0.001) for locoregional recurrence compared with NSMP ECs [41].

The results obtained from both research teams show that the pragmatic molecular classifiers carry independent prognostic significance and, importantly, the prognostic ability of these novel variables further improved by integrating them with existing prognostic clinicopathological variables (e.g. LVSI and stage) [41,42,68,69]. The data appear to be particularly strong for women with ECs that would currently be catego- rized as ‘high–intermediate-risk’ and may improve the prognostic ability in ‘high-risk’ EC patients [41,69]. In the context of low-risk disease (e.g. grade 1, stage IA EEC), the available data do not show that molecular classification adds prognostic value. This argues in favour of cost-conscious selected testing for EC patients with higher risks based on conventional pathology.

The molecular characterization may become especially relevant for patients with grade 3 EEC, a diagnosis

suffering from considerable inter-observer variability [5,7]. Recent data show that grade 3 EEC is remarkably heterogeneous, with representation of all molecular subgroups and a corresponding wide range in clinical outcomes [73]. This illustrates the capacity of the molecular classification to add clinically relevant information as well as enhance diagnostic accuracy in grade 3 EEC (Figure 4).

How the integrated molecular classification may impact treatment

The integration of biologically relevant molecular information into the EC diagnoses provides opportunities for molecular subclass-specific treatment stratification.

Depending on the clinical context, this integrated information may inform treatment in the preoperative setting (lymphadenectomy yes or no), post-operative setting (adjuvant treatment yes or no), and additionally may carry information that can be exploited for surveillance advice or targeted treatment options in the recurrent or metastatic setting (Figure 5).

Impact in the preoperative setting

As more and more guidelines are enforcing LS screening on all ECs, reflex testing on preoperative material, such as an endometrial curetting or pipelle biopsy, seems opportune. Adding p53 and POLE status may become desirable as well, as this information may

(7)

Figure 4. Example of molecular heterogeneity among grade 3 EECs. Top. Case I: H&E of a grade 3 EEC with wild-type p53 expression and presence of a POLE mutation by Sanger sequencing (P286R mutation, arrow). Bottom. Case II: H&E of a grade 3 EEC with aberrant p53 expression and absence of a POLE mutation. (All images ×20 objective magnification.)

Figure 5. Integrated molecular endometrial cancer classification. A proposed model of the integration of molecular subtype and clinico- pathological features. The relative weight of each molecular or pathological parameter and the optimal treatment algorithms within each molecular subtype need to be further refined through clinical trials. However, this integrated model yields opportunities in research and clinical care to interrogate within consistently classified, molecularly similar EC subsets.

become relevant for deciding upon the optimal surgical approach. The surrogate markers have already been validated to be reliable using preoperative material [44,45]. Lymphadenectomy (total, sentinel, none) is recommended for high-grade EC in most countries without factoring in molecular features. Disease spread beyond the uterus is extremely rare in high-grade EC harbouring POLE mutations; thus, simple hysterectomy and bilateral salpingoophorectomy may be worthy for prospective evaluation. In contrast, for women with p53-aberrant ECs, more aggressive surgery with lymph node assessment and omentectomy seems appropriate. Another clinical scenario in which molecular classification may aid in decision making is young EC

patients who wish to preserve fertility by hormonal man- agement and delayed hysterectomy. Further prospective studies to define the safety of these measures are required.

Impact on adjuvant treatment

The adjuvant treatment choices in the post-operative setting are likely to be the first to be impacted by the integrated molecular classification, as most studies have focused on this scenario. As mentioned before, there is no evidence that POLE or TP53 status has significant impact on the risk allocation of low-risk EC patients based on conventional pathology.

(8)

impacted by the integrated molecular classification, as risk stratification showed significant improvement.

This observation has been the basis for the currently recruiting randomized PORTEC-4a trial [74]. This is the first EC clinical trial that uses molecular features as an integral component to study the effects of differential adjuvant treatments. PORTEC-4a will compare standard adjuvant treatment for HIR-EC patients (vaginal brachytherapy) with differential adjuvant treatments based on an integrated molecular profile. It is prospective trials like PORTEC-4a that will be crucial in the years to come, as these will provide the required evidence for subclass-specific treatments.

Currently, the optimal adjuvant treatment of high-risk EC is largely unknown. Platinum-based adjuvant chemotherapy regimens are being used for high-risk EC patients, including those with non-endometrioid subtypes, but clinical benefit to the addition of chemotherapy has been disappointing [75,76]. Two randomized controlled trials (PORTEC-3 and GOG 249 [77,78]), which are in the final stages of evaluation, have analysed the benefit of adjuvant treatment with chemoradiotherapy versus radiotherapy alone in high-risk EC patients. The results of these trials will be critical to inform future adjuvant treatment guidelines.

Impact on surveillance

Tailored surveillance, adjusted to the patient’s risk of recurrence instead of a uniform protocol, would enhance patient care. Based on the different outcomes of the molecular EC subgroups, it would seem appropriate to follow up women with p53-aberrant ECs more closely than those with POLE-mutant tumours. Health eco- nomic implications, patient reported outcomes, and clinical outcomes relevant to these practice changes need to be assessed.

Impact in recurrent disease

An integrated molecular classification would provide value for women with recurrence of their EC, informing targeted treatment and focusing resources specifically on women most likely to derive benefit. Novel clinical trials can be designed, with molecular features considered in the treatment design.

Perhaps the best example of an opportunity for subtype-directed care is the application of immune therapy [32,79,80]. Several studies have shown that the high TMB present in POLE-mutant and MMRd EC results in an abundance of neoepitopes/neoantigens that bind to MHC and induce an enhanced anti-tumour CD8 T-cell response [81,82]. These cancers appear to escape the immune response by up-regulating immune checkpoint molecules, such as PD-1 and PDL1, which act as nega- tive regulators of effector T cells [83]. Agents blocking these molecules, termed checkpoint inhibitors, restore anti-tumour immunity, leading to tumour regression

cancer patients with progressive metastatic disease with nivolumab or pembrolizumab. Patients showed a rapid response to treatment, with partial regression of metastatic mass, and response was maintained for 7 and 14 months for nivolumab and pembrolizumab, respectively [86,87]. This has led to the accelerated approval by the Food and Drug Administration (FDA) of pembrolizumab for unresectable or metastatic MMRd tumours that have progressed following prior treatment and that have no satisfactory alternative treatment [88].

These exciting developments are particularly relevant as MMRd ECs are not infrequently advanced stage and may recur. Although POLE-mutated tumours repre- sent a therapeutic opportunity for immune checkpoint inhibitors, the prototypical early stage and excellent outcomes observed in these women make it harder to justify costly intervention.

Challenges that still need to be overcome

Multiple classifiers

One unresolved issue with the proposed molecular EC stratification is how to classify and manage ECs with more than one molecular feature, commonly referred to as ‘multiple classifiers’ (Figure 6). The frequency of multiple classifiers is low, ranging from 3% to 5% of all EC cases in the published series that used surrogate markers. In contrast, the frequency of multiple classifiers reported by the TCGA was higher, which likely is the result of the different ways the subgroups were defined [89]. Currently, there are insufficient data on the clinical outcome to allocate these multiple classifiers to one of the molecular subclasses. Therefore, as we move forward towards implementing the molecular EC model, it will be imperative to know the clinical behaviour of these multiple classifiers in order to help guide treatment decisions. International collaboration is underway to address this challenge.

Relevant biomarkers, beyond the TCGA classifiers The novel molecular classification system described in this review by no means captures the molecular heterogeneity in EC. Intra-class heterogeneity may further refine prognosis, which is particularly relevant for the large and heterogeneous group of ECs that fall into molecular class 3, p53wt/NSMP. CTNNB1 muta- tional status, L1CAM expression, ER/PR expression, and amplification of 1q32.1 have been put forward to refine the prognostic model. A significant proportion of ECs in NSMP (26–52%) carry activating muta- tions in exon 3 of CTNNB1. In low-risk and HIR-EC, CTNNB1-mutated ECs are associated with signifi- cantly lower RFS [41,90,91]. As CTNNB1 mutations are likely early drivers in endometrial carcinogenesis [92,93], CTNNB1-mutated EC may be regarded as

(9)

Figure 6. Example of a ‘multiple classifier’ endometrioid endometrial cancer. (A–C) H&E of a grade 1 EEC with loss of expression of MSH6 and aberrant expression of p53, thus an MMRd p53-aberrant EEC. The tumour also had loss of expression of MSH2 (not shown). (All images

×20 objective magnification.)

the fifth molecular subclass, dividing the NSMP into biologically and clinically divergent tumours. L1CAM can be considered a pan-cancer prognostic biomarker, which when overexpressed is strongly associated with advanced stage, lymph node invasion, and aggressive histological subtypes [94,95]. There are sufficient data that L1CAM is not merely a proxy of TP53 mutations, and carries independent prognostic value in EC [96].

In stage I EEC, expression of L1CAM (using a 10%

cut-off) is associated with lower overall survival and RFS [96–98], and in the PORTEC series of HIR-ECs, multivariable analysis showed independent prognostic value [41]. It is for that reason that L1CAM expression and CTNNB1 mutational status are incorporated into the molecular profiles of the PORTEC-4a trial [74].

ER/PR status is among the oldest prognostic markers in EC, where high PR expression has been correlated with improved survival and lymph node metastases [99–101]. Loss of ER has been correlated to lymph node metastases, and some studies correlate loss of expression to adverse prognosis [100,102]. However, in more recent studies, multivariate analysis showed no added prognostic significance for ER/PR expression in models that include molecular EC classification [41,101].

Finally, focal SCNA, such as 1q32.1 amplification, may improve the recurrence risk prediction model [65].

Non-endometrioid endometrial cancers

The molecular classification presented by the TCGA was built upon EC with endometrioid and serous morphology, leaving the question of whether it can be translated to other rare EC subtypes. Several recent studies have applied the same classification to non-endometrioid EC, resulting in separation of clinical outcome [38,42,68,103–106]. These studies are limited in size, but appear to demonstrate ‘proof of concept’ that molecular classification may be clinically meaningful across all histotypes.

Conclusion

The molecular classification of ECs, through practical clinically applicable assays, points towards a paradigm shift in EC pathology practice in the near future. Most

of the required tools used to classify these tumours are already in place in pathology laboratories, facilitating easy implementation and interpretation. We foresee a system where every new EC may be assessed conven- tionally (including histotype, grade, LVSI status, and stage) but will also be assigned to one of four molecular subclasses. The relative weight and importance of these parameters will continue to be studied. Regardless, this integrated approach will provide EC patients and their clinician team with a reliable and objective classification that carries information to guide their surgery, adjuvant treatment(s), and subsequent cancer surveillance sched- ule.

Although more work is needed to define the optimal implementation of molecular EC classification, we anticipate significant changes from standard practice in the near future. The ultimate goal is to improve outcomes for EC patients: facilitating LS identification, reducing unnecessary toxic therapies, and implementing biological rationale for conventional and targeted therapies. We encourage pathologists and clinicians to embrace molecular classification and look forward to a new era of precision approach to both research and clinical care.

Acknowledgements

This work was supported by the Dutch Cancer Society, Young Investigator Grant (TB, KWF/YIG, 10418).

Author contributions statement

JMcA and TB conceived the idea and were in charge of overall direction and planning. AL took the lead in writ- ing the manuscript. All authors provided critical feed- back and helped shape the manuscript. TB supervised the project and served as the corresponding author.

References

1. Gilks CB, Ionescu DN, Kalloger SE, et al. Tumor cell type can be reproducibly diagnosed and is of independent prognostic signifi- cance in patients with maximally debulked ovarian carcinoma. Hum Pathol 2008; 39: 1239–1251.

(10)

Am J Surg Pathol 2010; 34: 984–993.

3. Reid-Nicholson M, Iyengar P, Hummer AJ, et al. Immunopheno- typic diversity of endometrial adenocarcinomas: implications for differential diagnosis. Mod Pathol 2006; 19: 1091–1100.

4. Alkushi A, Kobel M, Kalloger SE, et al. High-grade endometrial carcinoma: serous and grade 3 endometrioid carcinomas have differ- ent immunophenotypes and outcomes. Int J Gynecol Pathol 2010;

29:343–350.

5. Han G, Sidhu D, Duggan MA, et al. Reproducibility of histological cell type in high-grade endometrial carcinoma. Mod Pathol 2013;

26:1594–1604.

6. Bartosch C, Manuel Lopes J, Oliva E. Endometrial carcinomas:

a review emphasizing overlapping and distinctive morphological and immunohistochemical features. Adv Anat Pathol 2011; 18:

415–437.

7. Gilks CB, Oliva E, Soslow RA. Poor interobserver reproducibility in the diagnosis of high-grade endometrial carcinoma. Am J Surg Pathol 2013; 37: 874–881.

8. Wang XY, Pan ZM, Chen XD, et al. Accuracy of tumor grade by preoperative curettage and associated clinicopathologic factors in clinical stage I endometriod adenocarcinoma. Chin Med J (Engl) 2009; 122: 1843–1846.

9. Karateke A, Tug N, Cam C, et al. Discrepancy of pre- and postoper- ative grades of patients with endometrial carcinoma. Eur J Gynaecol Oncol 2011; 32: 283–285.

10. Batista TP, Cavalcanti CL, Tejo AA, et al. Accuracy of preoperative endometrial sampling diagnosis for predicting the final pathology grading in uterine endometrioid carcinoma. Eur J Surg Oncol 2016;

42:1367–1371.

11. Colombo N, Preti E, Landoni F, et al. Endometrial cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up.

Ann Oncol 2013; 24(suppl 6): vi33–38.

12. Keys HM, Roberts JA, Brunetto VL, et al. A phase III trial of surgery with or without adjunctive external pelvic radiation therapy in intermediate risk endometrial adenocarcinoma: a Gynecologic Oncology Group study. Gynecol Oncol 2004; 92: 744–751.

13. Mariani A, Dowdy SC, Cliby WA, et al. Prospective assessment of lymphatic dissemination in endometrial cancer: a paradigm shift in surgical staging. Gynecol Oncol 2008; 109: 11–18.

14. Kwon JS, Qiu F, Saskin R, et al. Are uterine risk factors more important than nodal status in predicting survival in endometrial cancer? Obstet Gynecol 2009; 114: 736–743.

15. Creutzberg CL. GOG-99: ending the controversy regarding pelvic radiotherapy for endometrial carcinoma? Gynecol Oncol 2004; 92:

740–743.

16. Creutzberg CL, van Putten WL, Koper PC, et al. Surgery and post- operative radiotherapy versus surgery alone for patients with stage-1 endometrial carcinoma: multicentre randomised trial. PORTEC Study Group. Post operative radiation therapy in endometrial car- cinoma. Lancet 2000; 355: 1404–1411.

17. Todo Y, Kato H, Kaneuchi M, et al. Survival effect of para-aortic lymphadenectomy in endometrial cancer (SEPAL study): a retro- spective cohort analysis. Lancet 2010; 375: 1165–1172.

18. Colombo N, Creutzberg C, Querleu D, et al. Appendix 5: Endome- trial cancer: eUpdate published online 8 June 2017 (www.esmo .org/Guidelines/Gynaecological-Cancers). Ann Oncol 2017; 28:

iv153–iv156.

19. AlHilli MM, Mariani A, Bakkum-Gamez JN, et al. Risk-scoring models for individualized prediction of overall survival in low-grade and high-grade endometrial cancer. Gynecol Oncol 2014; 133:

485–493.

20. Kong TW, Chang SJ, Paek J, et al. Risk group criteria for tailoring adjuvant treatment in patients with endometrial cancer: a validation

21. Bendifallah S, Canlorbe G, Collinet P, et al. Just how accurate are the major risk stratification systems for early-stage endometrial cancer? Br J Cancer 2015; 112: 793–801.

22. Cancer Genome Atlas Research Network, Kandoth C, Schultz N, et al. Integrated genomic characterization of endometrial carci- noma. Nature 2013; 497: 67–73.

23. Bokhman JV. Two pathogenetic types of endometrial carcinoma.

Gynecol Oncol 1983; 15: 10–17.

24. Church DN, Briggs SE, Palles C, et al. DNA polymerase epsilon and delta exonuclease domain mutations in endometrial cancer. Hum Mol Genet 2013; 22: 2820–2828.

25. Henninger EE, Pursell ZF. DNA polymerase epsilon and its roles in genome stability. IUBMB Life 2014; 66: 339–351.

26. Rayner E, van Gool IC, Palles C, et al. A panoply of errors:

polymerase proofreading domain mutations in cancer. Nat Rev Cancer 2016; 16: 71–81.

27. Shevelev IV, Hübscher U. The 3^′5^′exonucleases. Nat Rev Mol Cell Biol 2002; 3: 364–376.

28. Shinbrot E, Henninger EE, Weinhold N, et al. Exonuclease muta- tions in DNA polymerase epsilon reveal replication strand specific mutation patterns and human origins of replication. Genome Res 2014; 24: 1740–1750.

29. Howitt BE, Shukla SA, Sholl LM, et al. Association of polymerase e-mutated and microsatellite-instable endometrial cancers with neoantigen load, number of tumor-infiltrating lymphocytes, and expression of PD-1 and PD-L1. JAMA Oncol 2015; 1: 1319–1323.

30. van Gool IC, Eggink FA, Freeman-Mills L, et al. POLE proofread- ing mutations elicit an antitumor immune response in endometrial cancer. Clin Cancer Res 2015; 21: 3347–3355.

31. Bakhsh S, Kinloch M, Hoang LN, et al. Histopathological features of endometrial carcinomas associated with POLE mutations: impli- cations for decisions about adjuvant therapy. Histopathology 2016;

68:916–924.

32. Eggink FA, Van Gool IC, Leary A, et al. Immunological pro- filing of molecularly classified high-risk endometrial cancers identifies POLE-mutant and microsatellite unstable carcinomas as candidates for checkpoint inhibition. Oncoimmunology 2017;

6:e1264565.

33. Hussein YR, Weigelt B, Levine DA, et al. Clinicopathological anal- ysis of endometrial carcinomas harboring somatic POLE exonucle- ase domain mutations. Mod Pathol 2015; 28: 505–514.

34. Simpkins SB, Bocker T, Swisher EM, et al. MLH1 promoter methy- lation and gene silencing is the primary cause of microsatellite insta- bility in sporadic endometrial cancers. Hum Mol Genet 1999; 8:

661–666.

35. Ligtenberg MJ, Kuiper RP, Chan TL, et al. Heritable somatic methylation and inactivation of MSH2 in families with Lynch syn- drome due to deletion of the 3’ exons of TACSTD1. Nat Genet 2009;

41:112–117.

36. Liccardo R, De Rosa M, Izzo P, et al. Novel implications in molecular diagnosis of Lynch syndrome. Gastroenterol Res Pract 2017; 2017: 2595098.

37. Campbell BB, Light N, Fabrizio D, et al. Comprehensive analysis of hypermutation in human cancer. Cell 2017; 171: 1042–1056.e10.

38. DeLair DF, Burke KA, Selenica P, et al. The genetic land- scape of endometrial clear cell carcinomas. J Pathol 2017;

243:230–241.

39. Rosa-Rosa JM, Leskelä S, Cristóbal-Lana E, et al. Molecular genetic heterogeneity in undifferentiated endometrial carcinomas.

Mod Pathol 2016; 29: 1594.

40. Kobel M, Tessier-Cloutier B, Leo J, et al. Frequent mismatch repair protein deficiency in mixed endometrioid and clear cell carcinoma of the endometrium. Int J Gynecol Pathol 2017; 36: 555–561.

(11)

44. van Esterik M, Van Gool IC, de Kroon CD, et al. Limited impact of intratumour heterogeneity on molecular risk assignment in endome- trial cancer. Oncotarget 2017; 8: 25542–25551.

45. Talhouk A, Hoang LN, McConechy MK, et al. Molecular classifi- cation of endometrial carcinoma on diagnostic specimens is highly concordant with final hysterectomy: earlier prognostic information to guide treatment. Gynecol Oncol 2016; 143: 46–53.

46. McConechy MK, Talhouk A, Leung S, et al. Endometrial carcino- mas with POLE exonuclease domain mutations have a favorable prognosis. Clin Cancer Res 2016; 22: 2865–3873.

47. Sameer AS, Nissar S, Fatima K. Mismatch repair pathway:

molecules, functions, and role in colorectal carcinogenesis. Eur J Cancer Prev 2014; 23: 246–257.

48. Chang DK, Ricciardiello L, Goel A, et al. Steady-state regulation of the human DNA mismatch repair system. J Biol Chem 2000; 275:

18424–18431.

49. Jiricny J. The multifaceted mismatch-repair system. Nat Rev Mol Cell Biol 2006; 7: 335–346.

50. Sepulveda AR, Hamilton SR, Allegra CJ, et al. Molecular biomark- ers for the evaluation of colorectal cancer: guideline from the Amer- ican Society for Clinical Pathology, College of American Patholo- gists, Association for Molecular Pathology, and the American Soci- ety of Clinical Oncology. J Clin Oncol 2017; 35: 1453–1486.

51. Molecular Testing Strategies for Lynch Syndrome in People with Colorectal Cancer. National Institute for Health and Care Excel- lence (NICE) Diagnostics Guidance (DG27). [Accessed 20 Decem- ber 2017]. Available from: https://www.nice.org.uk/

52. oncology NCCN Uterine Neoplasms. Version 3.2017. [Accessed 20 December 2017]. Available from: https://www.nccn.org

53. Committee on Practice Bulletins-Gynecology; SoGO. ACOG Prac- tice Bulletin No. 147: Lynch syndrome. Obstet Gynecol 2014; 124:

1042–1054.

54. Weissman SM, Burt R, Church J, et al. Identification of individuals at risk for Lynch syndrome using targeted evaluations and genetic testing: National Society of Genetic Counselors and the Collabo- rative Group of the Americas on Inherited Colorectal Cancer joint practice guideline. J Genet Couns 2012; 21: 484–493.

55. Leenen CH, van Lier MG, van Doorn HC, et al. Prospective eval- uation of molecular screening for Lynch syndrome in patients with endometrial cancer≤ 70 years. Gynecol Oncol 2012; 125: 414–420.

56. Lu KH, Dinh M, Kohlmann W, et al. Gynecologic cancer as a "sen- tinel cancer" for women with hereditary nonpolyposis colorectal cancer syndrome. Obstet Gynecol 2005; 105: 569–574.

57. Anagnostopoulos A, McKay VH, Cooper I, et al. Identifying Lynch syndrome in women presenting with endometrial carcinoma under the age of 50 years. Int J Gynecol Cancer 2017; 27: 931–937.

58. Stelloo E, Jansen AML, Osse EM, et al. Practical guidance for mis- match repair-deficiency testing in endometrial cancer. Ann Oncol 2017; 28: 96–102.

59. McConechy MK, Talhouk A, Li-Chang HH, et al. Detection of DNA mismatch repair (MMR) deficiencies by immunohistochemistry can effectively diagnose the microsatellite instability (MSI) phenotype in endometrial carcinomas. Gynecol Oncol 2015; 137:

306–310.

TP53 mutations in ovarian carcinoma: an immunohistochemical and nucleotide sequencing analysis. Mod Pathol 2011; 24: 1248–1253.

64. Kobel M, Piskorz AM, Lee S, et al. Optimized p53 immunohis- tochemistry is an accurate predictor of TP53 mutation in ovarian carcinoma. J Pathol Clin Res 2016; 2: 247–258.

65. Depreeuw J, Stelloo E, Osse EM, et al. Amplification of 1q32.1 refines the molecular classification of endometrial carcinoma. Clin Cancer Res 2017; 23: 7232–7241.

66. Nowak JA, Yurgelun MB, Bruce JL, et al. Detection of mismatch repair deficiency and microsatellite instability in colorectal adeno- carcinoma by targeted next-generation sequencing. J Mol Diagn 2017; 19: 84–91.

67. Stadler ZK, Battaglin F, Middha S, et al. Reliable detection of mismatch repair deficiency in colorectal cancers using mutational load in next-generation sequencing panels. J Clin Oncol 2016; 34:

2141–2147.

68. Talhouk A, McConechy MK, Leung S, et al. Confirmation of ProMisE: a simple, genomics-based clinical classifier for endome- trial cancer. Cancer 2017; 123: 802–813.

69. Stelloo E, Bosse T, Nout RA, et al. Refining prognosis and identify- ing targetable pathways for high-risk endometrial cancer; a Trans- PORTEC initiative. Mod Pathol 2015; 28: 836–844.

70. Committee on the Review of Omics-Based Tests for Predicting Patient Outcomes in Clinical Trials; Board on Health Care Services;

Board on Health Sciences Policy; Institute of Medicine. Micheel CM, Nass SJ, Omenn GS (eds). Evolution of Translational Omics:

Lessons Learned and the Path Forward. National Academies Press:

Washington, DC; 2012. Available from: https://www.ncbi.nlm.nih .gov/books/NBK202168/

71. Nout RA, Smit VT, Putter H, et al. Vaginal brachytherapy ver- sus pelvic external beam radiotherapy for patients with endometrial cancer of high-intermediate risk (PORTEC-2): an open-label, non-inferiority, randomised trial. Lancet 2010; 375: 816–823.

72. Creutzberg CL, van Putten WLJ, Koper PCM, et al. Surgery and postoperative radiotherapy versus surgery alone for patients with stage-1 endometrial carcinoma: multicentre randomised trial.

Lancet 2000; 355: 1404–1411.

73. Bosse T, Nout RA, McAlpine JN, et al. Molecular classifica- tion of grade 3 endometrioid endometrial cancers identifies distinct prognostic subgroups. Am J Surg Pathol 2018; doi:

10.1097/PAS.0000000000001020 [Epub ahead of print].

74. PORTEC-4a: Randomised Trial of Standard or Molecular Profile-Based Recommendation for Radiotherapy after Surgery for Women with Early Stage Endometrial Cancer; 2016 Jun 01.

[Accessed 20 December 2017]. Available from: https://www.isrctn .com/ISRCTN11659025

75. Susumu N, Sagae S, Udagawa Y, et al. Randomized phase III trial of pelvic radiotherapy versus cisplatin-based combined chemotherapy in patients with intermediate- and high-risk endometrial cancer:

a Japanese Gynecologic Oncology Group study. Gynecol Oncol 2008; 108: 226–233.

76. Maggi R, Lissoni A, Spina F, et al. Adjuvant chemotherapy vs radiotherapy in high-risk endometrial carcinoma: results of a ran- domised trial. Br J Cancer 2006; 95: 266–271.

(12)

cuff brachytherapy followed by paclitaxel/carboplatin chemotherapy (VCB/C) in patients with high risk (HR), early stage endome- trial cancer (EC): a Gynecologic Oncology Group trial. Gynecol Oncol 2014; 134: 438–440.

78. de Boer SM, Powell ME, Mileshkin LR, et al. Final results of the international randomized PORTEC-3 trial of adjuvant chemotherapy and radiation therapy (RT) versus RT alone for women with high-risk endometrial cancer. J Clin Oncol 2017; 35(no. 15_suppl):

5502–5502.

79. van Gool IC, Eggink FA, Freeman-Mills L, et al. POLE proofread- ing mutations elicit an antitumor immune response in endometrial cancer. Clin Cancer Res 2015; 21: 3347–3355.

80. van Gool IC, Bosse T, Church DN. Neoepitopes and CD3-positive and CD8-positive cells in polymerase e-mutated and microsatellite-instable endometrial cancers. JAMA Oncol 2016; 2: 141.

81. Brown SD, Warren RL, Gibb EA, et al. Neo-antigens predicted by tumor genome meta-analysis correlate with increased patient survival. Genome Res 2014; 24: 743–750.

82. Schumacher TN, Schreiber RD. Neoantigens in cancer immunother- apy. Science 2015; 348: 69–74.

83. Sharpe AH, Pauken KE. The diverse functions of the PD1 inhibitory pathway. Nat Rev Immunol 2018; 18: 153–167.

84. Topalian SL, Drake CG, Pardoll DM. Immune checkpoint blockade:

a common denominator approach to cancer therapy. Cancer Cell 2015; 27: 450–461.

85. Ott PA, Bang YJ, Berton-Rigaud D, et al. Safety and antitumor activity of pembrolizumab in advanced programmed death ligand 1-positive endometrial cancer: results from the KEYNOTE-028 Study. J Clin Oncol 2017; 35: 2535–2541.

86. Mehnert JM, Panda A, Zhong H, et al. Immune activation and response to pembrolizumab in POLE-mutant endometrial cancer.

J Clin Invest 2016; 126: 2334–2340.

87. Santin AD, Bellone S, Buza N, et al. Regression of chemotherapy-resistant polymerase ε (POLE) ultra-mutated and MSH6 hyper-mutated endometrial tumors with nivolumab.

Clin Cancer Res 2016; 22: 5682–5687.

88. FDA Grants Accelerated Approval to Pembrolizumab for First Tissue/Site Agnostic Indication. Silver Spring, MD: US Food and Drug Administration. [Accessed 23 May 2017]. Available from:

https://www.fda.gov

89. Schultheis AM, Martelotto LG, De Filippo MR, et al. TP53 muta- tional spectrum in endometrioid and serous endometrial cancers. Int J Gynecol Pathol 2016; 35: 289–300.

90. Liu Y, Patel L, Mills GB, et al. Clinical significance of CTNNB1 mutation and Wnt pathway activation in endometrioid endometrial carcinoma. J Natl Cancer Inst 2014; 106: pii: dju245.

91. Kurnit KC, Kim GN, Fellman BM, et al. CTNNB1 (beta-catenin) mutation identifies low grade, early stage endometrial cancer patients at increased risk of recurrence. Mod Pathol 2017; 30:

1032–1041.

2010; 1: 674–684.

93. Russo M, Broach J, Sheldon K, et al. Clonal evolution in paired endometrial intraepithelial neoplasia/atypical hyperplasia and endometrioid adenocarcinoma. Hum Pathol 2017; 67: 69–77.

94. van der Putten LJ, Visser NC, van de Vijver K, et al. L1CAM expression in endometrial carcinomas: an ENITEC collaboration study. Br J Cancer 2016; 115: 716–724.

95. Tangen IL, Kopperud RK, Visser NC, et al. Expression of L1CAM in curettage or high L1CAM level in preoperative blood samples predicts lymph node metastases and poor outcome in endometrial cancer patients. Br J Cancer 2017; 117: 840–847.

96. Van Gool IC, Stelloo E, Nout RA, et al. Prognostic significance of L1CAM expression and its association with mutant p53 expression in high-risk endometrial cancer. Mod Pathol 2016; 29:

174–181.

97. Bosse T, Nout RA, Stelloo E, et al. L1 cell adhesion molecule is a strong predictor for distant recurrence and overall survival in early stage endometrial cancer: pooled PORTEC trial results. Eur J Cancer 2014; 50: 2602–2610.

98. Zeimet AG, Reimer D, Huszar M, et al. L1CAM in early-stage type I endometrial cancer: results of a large multicenter evaluation. J Natl Cancer Inst 2013; 105: 1142–1150.

99. Kobel M, Atenafu EG, Rambau PF, et al. Progesterone recep- tor expression is associated with longer overall survival within high-grade histotypes of endometrial carcinoma: a Canadian high risk endometrial cancer consortium (CHREC) study. Gynecol Oncol 2016; 141: 559–563.

100. Trovik J, Wik E, Werner HM, et al. Hormone receptor loss in endometrial carcinoma curettage predicts lymph node metastasis and poor outcome in prospective multicentre trial. Eur J Cancer 2013; 49: 3431–3441.

101. Karnezis AN, Leung S, Magrill J, et al. Evaluation of endometrial carcinoma prognostic immunohistochemistry markers in the context of molecular classification. J Pathol Clin Res 2017; 3: 279–293.

102. Wik E, Raeder MB, Krakstad C, et al. Lack of estrogen receptor-alpha is associated with epithelial– mesenchymal transition and PI3K alterations in endometrial carcinoma. Clin Cancer Res 2013; 19: 1094–1105.

103. Espinosa I, Lee CH, D’Angelo E, et al. Undifferentiated and dedif- ferentiated endometrial carcinomas with POLE exonuclease domain mutations have a favorable prognosis. Am J Surg Pathol 2017; 41:

1121–1128.

104. Jones S, Stransky N, McCord CL, et al. Genomic analyses of gynae- cologic carcinosarcomas reveal frequent mutations in chromatin remodelling genes. Nat Commun 2014; 5: 5006.

105. Cherniack AD, Shen H, Walter V, et al. Integrated molecular char- acterization of uterine carcinosarcoma. Cancer Cell 2017; 31:

411–423.

106. Le Gallo M, Rudd ML, Urick ME, et al. Somatic mutation profiles of clear cell endometrial tumors revealed by whole exome and targeted gene sequencing. Cancer 2017; 123: 3261–3268.