Introduction Molecular classification of endometrial carcinoma has been proposed to predict survival. However, its role in patient management remains to be determined. We aimed to identify whether a molecular and immunohistochemical classification of endometrial carcinoma could improve decision-making for adjuvant therapy.
Methods All consecutive patients treated for endometrial carcinoma between 2010 and 2017 at Cochin University Hospital were included. Clinical risk of relapse was based on European Society for Medical Oncology-European Society of Gynaecological Oncology-European SocieTy for Radiotherapy & Oncology (ESMO-ESGO-ESTRO) consensus. The clinical event of interest was event-free survival. Formalin-fixed paraffin-embedded tissue samples were processed for histopathological analysis and DNA extraction. The nuclear expression of mismatch repair and TP53 proteins was analyzed by immunohistochemistry. Next-generation sequencing of a panel of 15 genes including TP53 and POLE was performed using Ampliseq panels on Ion Torrent PGM (ThermoFisher). Tumors were allocated into four molecular groups using a sequential method based on next-generation sequencing and immunohistochemistry data: (1) POLE/ultramutated-like; (2) MSI/hypermutated-like (mismatch repair-deficient); (3) TP53-mutated (without POLE mutations or mismatch repair deficiency); (4) not otherwise specified (the remaining tumors).
Results 159 patients were included; 125 tumors were available for molecular characterization and distributed as follows: (1) POLE/ultramutated-like: n=4 (3%); (2) MSI/hypermutated-like: n=35 (30%); (3) TP53-mutated: n=30 (25%); and (4) not otherwise specified: n=49 (42%). Assessing the TP53 status by immunohistochemistry only rather than next-generation sequencing would have misclassified 6 tumors (5%). TP53-mutated tumors were associated with poor prognosis, independently of International Federation of Gynecology and Obstetrics (FIGO) stage and histological grade (Cox-based adjusted hazard ratio (aHR) 5.54, 95% CI 2.30 to 13.4), and independently of clinical risk of relapse (aHR 3.92, 95% CI 1.59 to 9.64). Among patients with FIGO stage I–II tumors, 6 (38%) TP53-mutated tumors had low/intermediate clinical risk of relapse and did not receive adjuvant chemotherapy or radiotherapy.
Conclusion Endometrial carcinoma molecular classification identified potentially under-treated patients with poor molecular prognosis despite being at low/intermediate clinical risk of relapse. Consideration of molecular classification in adjuvant therapeutic decisions should be evaluated in prospective trials.
- endometrial neoplasms
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Endometrial carcinoma molecular characterization is feasible as a routine procedure.
Poor prognosis of TP53-mutated tumors is independent of tumor stage and histological grade, and of clinical risk of relapse.
Endometrial carcinoma molecular characterization identified potentially under-treated patients in the adjuvant setting.
Endometrial carcinoma is the most frequent gynecological cancer in developed countries.1 As surgery remains the first-line treatment in most of the cases, the main clinical challenge is to identify patients who will benefit from postoperative adjuvant treatment. Two main subtypes of endometrial carcinoma have been historically described, based on histological features and prognosis2: type 1 carcinoma, associated with endometrioid differentiation, low tumor grade, and good outcomes; and type 2 carcinoma, associated with non-endometrioid histology (serous adenocarcinoma, clear-cell carcinoma, and mixed Mullerian tumor) and poor outcomes. The European Society for Medical Oncology-European Society of Gynaecological Oncology-EuropeanSocieTy for Radiotherapy & Oncology (ESMO-ESGO-ESTRO) consensus recommends basing adjuvant treatment decisions on clinical risk of relapse, using International Federation of Gynecology and Obstetrics (FIGO) stage, tumor grade, lymphovascular space invasion, and histological subtype.3 4 Despite the improvement resulting from these recommendations, it remains mandatory to improve prognosis evaluation and prediction of adjuvant therapy efficacy, in order to avoid unnecessary morbidities related to adjuvant treatments, and to decrease the rate of relapse.
In 2013, the genomic characterization of endometrial carcinoma by The Cancer Genome Atlas (TCGA) was regarded as an important step towards handling this challenge. Four molecular groups were identified: (1) an ultra-mutated group, with DNA-polymerase ε (POLE) catalytic subunit A mutations, associated with an excellent prognosis; (2) a hypermutated group, characterized by a microsatellite instability (MSI), largely due to hypermethylation of MLH1 promoter; (3) a group characterized by low copy-number alterations (copy-number-low), and microsatellite stability; and (4) a group characterized by high copy-number alterations (copy-number-high (CNH)) and TP53 mutations, associated with poor prognosis.5 However, it has not been established whether this molecular characterization may provide additional information to indicate adjuvant therapies, independently of FIGO stage and clinical risk of relapse.
We aimed to assess whether a sequential molecular and immunohistochemical classification of endometrial carcinoma might be associated with survival, independently of the previously established prognostic factors, and to highlight its potential clinical usefulness through the identification of potentially under-treated patients.
Patients and Endometrial Carcinoma Tissue Selection and DNA Extraction
All consecutive patients who underwent endometrial biopsy and/or hysterectomy for endometrial carcinoma in the Department of Gynecologic Surgery of Cochin University Hospital between 2010 and 2017, and for whom formalin-fixed paraffin-embedded (FFPE) tissue samples were suitable for molecular analysis, were included in the cohort. Genomic DNA extraction was performed on the sample area with the highest cellularity, using the Maxwell 16 FFPE Plus LEV DNA Purification Kit (Promega, Charbonnières-les-Bains, France), according to the manufacturer’s instructions. This study was performed in accordance with ethical guidelines and approved by the Paris Ethics Committee (CPPIDF1-2015-DAP22). All patients signed an informed consent.
Follow-up and Data Collection
Patients were prospectively followed as per routine practice, from the date of surgery (or endometrial biopsy for patients who were not operated on), until the date of last news or date of database lock on October 31, 2017.
Clinical and Pathological Data
Surgical procedures and non-surgical treatments were retrospectively collected from medical records. Pathological features (histological type, grade, lymphovascular space invasion, myometrial invasion, immunohistochemistry of hormone receptors, TP53, and mismatch repair status) were collected from pathology reports. All pathological diagnostics were retrospectively reviewed by a pathologist with a specific expertise in gynecological oncology (P-AJ). Missing immunohistochemistry data were completed retrospectively from archived FFPE tissue samples using the same routinely applied method (see below). Tumor stage was determined according to the 2010 FIGO staging system.
Clinical Risk of Relapse
Classification of the clinical risk of relapse was applied according to the ESMO-ESGO-ESTRO consensus on endometrial carcinoma.3 4 This classification considers lymphovascular space invasion, histological grade and type, myometrial invasion, and FIGO stage. Patients with FIGO stage I–III tumors were classified as low, intermediate, high-intermediate, and high-risk of relapse. Patients with advanced diseases (stage IVA/IVB) were classified as advanced/metastatic.
Immunohistochemistry assays were performed on a Leica Bond-III Autostainer using the Bond Polymer Refine Detection Kit (Leica Biosystem), according to the manufacturer’s instruction. The following primary antibodies were applied on 5 µm FFPE tissue slides after appropriate heat-induced epitope retrieval (ER1 or ER2, Leica Biosystems): TP53 (DO-7; DAKO; 1:800 dilution), PMS2 (A16-4; Pharmingen; 1:300), MSH6 (44; LSBio; 1:50), estrogen receptor (1D5; DAKO; 1:50), and progesterone receptor (PgR636; DAKO; 1:300). Staining was categorized using the staining intensity (+ to +++) and the percentage of stained tumor cells (0–100%). Hormone receptor staining was considered positive if >+ and >10%, and TP53 staining was considered aberrant if >+++ and >90%, or if there was complete staining loss (suggesting truncating mutations).6 Mismatch repair deficiency (MMRd) was defined by the loss of staining of PMS2 or MSH6 with positive endogenous control7 8 (MLH1 and MSH2 assays were available for 109 patients, with only one discordance). Technical failures for immunohistochemical analyses are reported as “technically uninformative immunostaining”.
DNA sequencing was performed using a standard system for targeted sequencing.9 An AmpliSeq gene panel was designed using AmpliSeq Designer (v4.47) on Human genome Hg19 to detect somatic point mutations (single nucleotide variants and short indels) in genes significantly mutated in endometrial carcinoma5: TP53, KRAS, PTEN, PIK3CA, PIK3R1, POLE, ERBB2, CTNNB1, RPL22, PPP2R1A, ARID1A, ARID5B, CTCF, FBXW7, FGFR2. Overall, 440 amplicons were designed in two pools (online supplementary method 1). Libraries were prepared with Ion AmpliSeq protocols and clonally amplified on sequencing beads in emulsion PCR with Ion OneTouch. Enriched beads were loaded on Ion 316 chips. Sequencing was then performed using Ion Torrent technology with Ion Torrent Personal Genome Machine (PGM) System (Life Technology, ThermoFisher Scientific, Courtaboeuf, France).
Bio-informatic Processing of Sequencing Data
Torrent Suite Software (v5.6) was used for sequencing data processing (online supplementary method 2). TMAP (Torrent Mapping Alignment Program) software was used to perform reads processing and mapping on loaded genome (Hg19), using default parameters. Samples with mean read length <90 bp, mean depth <150X, and uniformity of coverage <70%, were removed in subsequent analysis (see online supplementary Figure S1 for flow-chart). Variant Caller Plugin was used for variant calling. Variant Caller’s parameters were chosen to allow variant detection with minimal stringency, to get a high sensitivity (online supplementary method 3). Variant call used COSMIC database reference transcripts. Variants were filtered to include in the final review all significant (p<0.0001, based on Phred quality score logarithmic transformation), somatic non-synonymous variants (Global minor allele frequency <0.1% in 1000 Genomes Project database), with a coverage ≥50X and a vaf ≥5%. All selected variants were manually reviewed one by one by two authors (GB, KL) using the Integrative Genomics Viewer (IGV) tool10 11 and automated annotations (dbSNP, 1000 Genome Project, UCSC common single nucleotide polymorphisms (SNPs)).
Molecular and Immunohistochemical Classification of Endometrial Carcinoma
We sequentially classified endometrial carcinoma on the basis of the following steps, consistently with the TCGA analysis5: (1) detection of POLE exonuclease domain mutations; (2) MMRd phenotype; (3) detection of TP53 mutations.5 12 Tumors mutated within the POLE exonuclease domain were classified as POLE/ultramutated-like. POLE exonuclease domain wild-type tumors with an MMRd phenotype were considered as MSI/hypermutated-like. Among the remaining tumors, the detection of somatic point mutations within TP53 coding sequence identified a third molecular group, TP53-mutated classified tumors. In other words, among TP53-mutated tumors, only those without POLE mutations and MMRd phenotype belonged to the TP53-mutated molecular group. Tumors not yet classified were allocated to a fourth group: not otherwise classified tumors (online supplementary Figure S1).
Quantitative data were described using mean or median. Qualitative variables were described using absolute values and ratio. Distribution of categorical variables such as mutation rates in molecular groups were compared using Fisher’s exact test. Median follow-up was estimated using the reverse Kaplan–Meier method. Survival analyses were performed using the Cox logistic regression model. Primary event of interest was event-free survival, defined as any event (progression, relapse, or death), censored by date of last news. Variables reaching trends toward significance in univariate analysis (p<0.10) were considered for inclusion in multivariate models, and subsequently selected based on clinical relevance and collinearity. Proportional hazard assumptions were checked. Survival curves were analyzed using the Kaplan-Meier method, and compared using the two-tailed log-rank test. Patients with missing data for one variable were not considered in statistical analyses with this variable (missing data summarized in Table 1). Statistical significance was defined by p<0.05. Statistical analyses were performed using R (version 3.6).
We identified 159 patients with a diagnosis of endometrial carcinoma and included 125 of these for further analyses (flow chart: online supplementary Figure S1). The population’s characteristics were consistent with known epidemiology of endometrial carcinoma (Table 1). Median age was 66.2 years. Most patients had type 1 carcinoma (82%) and were at an early stage of disease at the time of diagnosis (stage I: 69%). Median follow-up was 33.9 months (interquartile range 15.1–49.5) and 34 patients (27%) experienced an event of interest (progression, relapse, or death). All patients with stage I–III disease underwent at least hysterectomy and bilateral oophorectomy. Lymph node dissection and adjuvant treatments (radiotherapy, brachytherapy, and chemotherapy) were performed according to the latest published international guidelines3 4 (online supplementary Table S1).
Endometrial Carcinoma Classification and Prognosis
Four patients (3%) had tumors with POLE exonuclease domain mutations (online supplementary Table S2). Thirty-five patients (30%) had tumors with MMRd phenotype with the following distribution: n=29 (83%) loss of PMS2 expression and n=6 (17%) loss of MSH6. Thirty patients (25%) were assigned to the TP53-mutated molecular group—that is, TP53 mutated tumors without POLE mutation or MMRd phenotype (online supplementary Table S3). The 49 remaining patients (42%) were considered with not otherwise classified tumors (Table 1). Analysis of mutation rates in significantly mutated genes in endometrial carcinoma throughout the molecular groups was consistent with TCGA findings (online supplementary Figure S2). Worthy of note, we observed discrepancies between TP53 immunostaining and TP53 mutations. Seven out of 37 (19%) TP53-mutated tumors had normal TP53 immunostaining, and five out of 35 (14%) tumors with TP53 abnormal staining had no detected TP53 mutations (online supplementary Table S3 and Figure S3). Two out of 30 (7%) TP53-mutated classified tumors had normal TP53 staining, and seven non-TP53-mutated classified tumors had abnormal TP53 staining. Overall, assessing the TP53 status by immunohistochemistry only rather than next-generation sequencing would have misclassified six tumors (5%).
Tumor FIGO stage III–IV, histological high-risk tumors (type 2 or grade 3 endometrioid carcinomas), and abnormal TP53 immunostaining were associated with poorer event-free survival (online supplementary Table S4). Clinical risk of relapse classification was not associated with outcome in our cohort (p=0.069). Regarding molecular alterations, only TP53 mutation was significantly associated with worse prognosis (online supplementary Table S5). None of patients with POLE tumors had relapsed (the number of patients was too small to perform the relevant statistical test).
This molecular classification was significantly associated with event-free survival both in the whole cohort (online supplementary Figure S4) and in patients with stage I–III tumors (Figure 1A), while clinical risk of relapse classification was not (Figure 1B). Patients affected by TP53-mutated classified tumors had the poorest event-free survival, while those with MSI/hypermutated-like and not otherwise classified tumors had intermediate outcomes. None of the patients with POLE mutations have relapsed to date. Because of the correlation between FIGO stage and clinical risk of relapse (Pearson correlation coefficient=0.56 between stage III–IV and clinical high risk of relapse), two multivariate models were performed to assess the independent association between the TP53-mutated molecular group and poor event-free survival. Either adjusted on FIGO stage and histological risk or on clinical risk of relapse, TP53-mutated classified tumors were independently associated with poorer event-free survival (Table 2).
Clinical Risk of Relapse Versus Molecular Classification in Therapeutic Management
While TP53-mutated classified tumors were more frequent in high-intermediate/high clinical risk of relapse groups (n=16, 31%), they still represented a clinically meaningful proportion of tumors belonging to intermediate/low clinical risk groups (n=6, 10%) (p=0.01) (Table 3). Thus, six out of 16 (38%) stage I–II TP53-mutated classified tumors were clinically classified as having intermediate/low risk of relapse. TP53-mutated classified tumors were associated with poorer event-free survival both in patients with or without high clinical risk of relapse (online supplementary Figure S5).
We observed an extensive use of adjuvant treatments in patients belonging to the high clinical risk of relapse group (online supplementary Table S1). However, when considering patients with TP53-mutated classified tumors, regardless of clinical risk of relapse, only 17 (57%) received adjuvant chemotherapy, 10 (33%) received pelvic radiotherapy, six (20%) received para-aortic radiotherapy, and 10 (33%) received brachytherapy. Thirty-three (83%) patients with high risk of relapse underwent pelvic and para-aortic lymph node dissection. Only 16 (53%) patients with TP53-mutated classified tumors underwent lymph node dissection.
In patients with localized disease, when stratified on clinical risk of relapse, exploratory analysis showed that chemotherapy administration was associated with better event-free survival in patients with TP53-mutated classified tumors (HR 0.17, 95% CI 0.03 to 1.00; p=0.05), but not in patients with non-TP53 tumors (HR 1.70, 95% CI 0.38 to 7.86; p=0.4) (online supplementary Figure S6). The TP53-mutated group was also independently associated with worse event-free survival in chemotherapy-untreated patients (online supplementary Table S6).
This study showed that a classification of endometrial carcinoma combining targeted sequencing and immunohistochemistry could refine the prognostic assessment based on current standard criteria (myometrial invasion, lymphovascular space invasion, histological type and grade), independently of tumor stage, and could be relevant for decision-making in the adjuvant setting for stage I–II tumors. TP53-mutated classified tumors, better identified by TP53 sequencing than immunohistochemistry, were strongly associated with poor outcomes, independently of clinical risk of relapse. Almost one-third of patients with low/intermediate risk tumors and TP53-mutated classified tumors could have been under-treated. In contrast, high-intermediate and high-risk tumors included 31% of TP53-mutated classified tumors, but also 35% of MSI-like tumors and one POLE tumor, with potentially expected good outcomes.
Exploratory analysis showed that patients with high-intermediate/high-risk tumors could have better outcomes with adjuvant chemotherapy, especially in the subgroup gathering TP53-mutated classified tumors. Despite this observation needing further confirmation, it could be related to the lower rate of events in the non-TP53-mutated classified molecular group and/or to a specific sensitivity to platinum salts of the TP53-mutated classified tumors.
We propose a method to identify tumor’s molecular group using routine systems (immunohistochemistry, targeted next-generation sequencing). Although further feasibility studies are warranted, the method presented here could provide a tumor’s molecular classification to the clinician within delays compatible with adjuvant therapeutic constraints. Since the first publication of molecular classification in endometrial carcinoma,5 a handful of reports aimed to develop a simplified and less costly molecular characterization.12–16 However, few of these reports described its relevance in therapeutic decision-making. A re-classification of high-intermediate risk patients has been proposed by integrating molecular features in patient stratification,15 as the authors suggested the existence of patients potentially over- or under-treated. However, no predictive marker for postoperative chemotherapy or radiotherapy benefit has been proposed. Indeed, the evidence of a benefit for adjuvant chemotherapy is still weak.17 This could be due to a poor decision-making process for chemotherapy indications, affected by over-treatments and under-treatments that could bias outcomes.
We determined TP53 status, a surrogate marker of the CNH molecular group, using sequencing on the TP53 coding sequence. Most of the previously reported molecular classifications, including the ProMisE classification, were based on TP53 immunostaining analysis.14 15 This assay is of complex interpretation, particularly for truncating variants with TP53 loss, for which distinguishing TP53 loss of expression and physiological degradation requires caution.18 On the other hand, some variants may not affect TP53 function/regulation and/or may not lead to an abnormal protein stabilization.19 The potentially misclassified cases theoretically support the use of TP53 sequencing, but further studies are needed to clarify the discrepancies and evaluate the clinical impact of using one or the other technology to assess TP53 status.
Our work showed weaknesses relative to the retrospective data collection and the limited statistical power. More mature results should be published after reaching a median follow-up consistent with PORTEC3 time-point for primary analysis (5 years). The retrospective design of molecular classification from archived FFPE samples could have biased our population by precluding the analysis of biopsies with low tissue material (constraints limiting our capabilities to assess a four protein panel for mismatch repair), or impairing molecular analyses from weakly fixed samples. For these patients, surgical specimens were analyzed, with potential bias regarding fixation artifacts.
Beside these limitations, our study showed strengths. All consecutive patients treated in one center were screened for inclusion, with standardized surgical and non-surgical treatments. Pathology analysis was centralized and retrospectively checked for all cases. The therapeutic strategy was consistent overall throughout the inclusion period and followed international guidelines.
In conclusion, our study provides evidence that endometrial carcinoma molecular characterization could be useful not only to evaluate patient prognosis, but also to improve adjuvant therapeutic strategies. Further prospective studies are required to assess the predictive impact of TP53 genotype on chemotherapy benefit, such as the PORTEC4a trial.20
This work has been performed thanks to the scientific collaborative CARPEM (CAncer Research for PErsonalized Medicine) framework.
BR and P-AJ contributed equally.
KL and JA contributed equally.
Contributors Planning and design : Karen Leroy, Jérôme Alexandre, Bruno Borghese. Cohort constitution: Guillaume Beinse, Pierre-Alexandre Just, Catherine Durdux, Jérôme Alexandre, Bruno Borghese. Data acquisition, analysis, and interpretation: Guillaume Beinse, Bastien Rance, Pierre-Alexandre Just, Brigitte Izac, Franck Letourneur, Nathaniel Edward Bennett Saidu, Sandrine Chouzenoux, Carole Nicco, Eric Passmant, Karen Leroy, Jérôme Alexandre, Bruno Borghese. Manuscript drafting : Guillaume Beinse, Bruno Borghese. All authors have significantly contributed to the work, and approved the final version of the mansucript.
Funding This study was funded by the Ligue Contre le Cancer and the Cancer Research and Personalized Medicine (CARPEM) program. Guillaume Beinse and Nathaniel EB Saidu received a grant to perform this work by the Association pour la Recherche contre le Cancer (ARC) and the CARPEM program, respectively. None of these had any role in the design and conduction of the study, the collection, management, analysis, and interpretation of the data, the preparation, review and approval of the manuscript or the decision to submit the manuscript for publication.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information. Data are available upon reasonable request to firstname.lastname@example.org.