New classification of endometrial cancers: the development and potential applications of genomic-based classification in research and clinical care

Several research teams have defined immunohistochemical and/or mutation profiles to aid in distinguishing EC subtypes [50–58]. In one series, a set of seven immunohistochemical markers was able to improve the distinction between high-grade EC histotypes [28] and more recently, another team demonstrated a nine protein panel improved identification of both low and high-grade EC subtypes [57]. Sequencing has enabled further improvement, with a nine-gene panel, demonstrating distinct mutational profiles for the major EC histotypes [52]. Molecular data has also been used to further stratify risk categories; using gene expression profiling and copy number analysis to determine risk of recurrence [59, 60], even in apparent low stage disease [61]. Molecular characterization has also been pursued for potential therapeutic targets in EC, focusing on frequently mutated pathways such as PI3K/PTEN/AKT/mTOR. Further work is needed to define molecular biomarkers that more accurately reflect tumor susceptibility [62–66].

The most comprehensive molecular study of ECs to date has been The Cancer Genome Atlas (TCGA) project, which included a combination of whole genome sequencing, exome sequencing, microsatellite instability (MSI) assays, and copy number analysis [67]. Molecular information was used to classify 232 endometrioid and serous endometrial cancers into four groups – POLE ultramutated, MSI hypermutated, copy-number (CN) low, and CN high – that correlate with progression-free survival.

The ultramutated POLE subgroup was a novel finding from the TCGA, and generated interest due to its very favorable outcomes even within high-grade tumors. In TCGA, ultramutated cases were characterized by POLE exonuclease domain mutations (EDM), a high percent of C??A transversions, a low percent of C??G transversions, as well as more than 500 SNVs. POLE encodes the major catalytic and proofreading subunits of the Pol? (Polymerase Epsilon) DNA polymerase enzyme complex responsible for leading strand DNA replication. The exonuclease proofreading function and the high fidelity incorporation of bases by POLE ensures a low mutation rate in the daughter strand. In ECs, POLE EDMs are mostly found in hotspot regions with V411L and P286R being the most common mutations. Substitutions in DNA polymerases were shown to inactivate or suppress proofreading abilities, thus causing increased replicative error rates and resulting in the ultra-mutated phenotype. In the TCGA, whole genome or exome sequencing was used to assess POLE status. Other series have subsequently assessed POLE status using more focused methods including Sanger sequencing [68, 69], gene panels [69–71], digital PCR [72–74] or functional assays [75] and confirmed very favourable outcomes for women with POLE aberrant ECs.

TCGA also described a molecular subgroup that exhibited microsatellite instability (MSI). MSI arises from defects in post-replicative DNA mismatch repair system. In the TCGA, MSI was determined by a panel of four mononucleotide repeat loci (polyadenine tracts BAT25, BAT26, BAT40, and transforming growth factor receptor type II) and three dinucleotide repeat loci (CA repeats in D2S123, D5S346, D17S250) in addition to the recommended markers from the National Cancer Institute [76], tumor DNA was classified as microsatellite- stable (MSS) if zero markers were altered, low level MSI (MSI-L) if one to two markers (less than 40%) were altered and high level MSI (MSI-H) if three or more markers (greater than 40%) were altered. Mismatch repair deficiencies can result from i) an inherited cancer syndrome (e.g., Lynch), ii) acquired/somatic mutations or iii) epigenetic events e.g. methylation of one of the genes involved in mismatch DNA repair, most commonly MLH1.

Finally TCGA distinguished a distinct molecular subgroup by copy number analysis. Copy number was determined using Affymetrix SNP 6.0 microarrays using DNA originating from frozen tissue. Hierarchical clustering identified significantly reoccurring amplifications or deletions regions and a ‘copy number (CN) high’ subgroup. All remaining samples that did not belong to the POLE ultramutated group, the MSI group, or the CN high group, were termed CN low. The appeal of objective molecular categorization of new EC cases in to one of four prognostic subgroups was immediately apparent. However, methodologies used for the TCGA study were costly, complex and unsuitable for wider clinical application.

Two research teams, including our own, have subsequently developed more pragmatic methodologies to evaluate molecular features of ECs, working in standard formalin-fixed paraffin-embedded tissue. These methods do not identify molecular subgroups that are identical to TCGA but do recapitulate the four survival curves observed in TCGA [69, 71, 73, 77] (Fig. 1). Stelloo et al. [69, 71] used a combination of TP53 mutational testing and p53 IHC to determine p53 status obtained from sequencing as a surrogate for CN high TCGA subgroup. The promega MSI analysis system was used to determine MSI status. For tumors exhibiting low levels of instability or from which extracted DNA quality was poor, immunohistochemistry for mismatch repair (MMR) proteins (MLH1, MSH2, MSH6, and PMS2) was performed. POLE EDM hotspot mutations were identified by Sanger sequencing. This team also tested for hotspot mutations (159) across 13 genes (BRAF, CDKNA2, CTNNB1, FBXW7, FGFR2, FGFR3, FOXL2, HRAS, KRAS, NRAS, PIK3CA, PPP2R1A, and PTEN). Testing ultimately yielded four molecular subgroups: group 1 – p53 (mutation identified), group 2- MSI, group 3 –POLE (POLE EDM identified), and finally group 4 –NSMP, a group with ‘no specific molecular profile’ (Fig. 1a). Tumors with insufficient tissue to perform all molecular testing were not classified and tumors with more than one molecular feature, constituting 2–3% of the cohort, were also not classified. Due to this exclusion, the order of mutational testing was irrelevant. This research team initially assessed ECs from the PORTEC3 trial (n?=?116), with known high risk features. Recurrence-free survival and time to distance metastasis were assessed within the four molecular subgroups. They observed that patients belonging to the POLE and the MSI subgroups showed similar and much better survival outcomes in comparison to the p53 mutant group and the NSMP group which exhibited worse recurrence and distance metastasis outcomes even within the endometrioid histology cases. Differences in survival patterns relative to the TCGA results were attributed to a greater proportion of high-risk features in the PORTEC 3 cohort.

The Leiden/TransPORTEC group has since applied the same series of molecular tests to a larger, more diverse cohort [71]. However, survival analysis and assessment of prognostic ability was restricted to endometrioid subtype and stage 1 tumors of patients with intermediate clinical risk. Within this very specific group, the observed outcomes associated with each molecular subgroup more closely mirrored TCGA.

Our research team has also developed a molecular classification system that uses practical methodologies to assign ECs to one of four molecular subgroups with distinct survival outcomes. We have followed the Institute of Medicine (IOM) guidelines for the development of ‘omics based tests [78], initially exploring 16 models in a ‘discovery’ cohort (n?=?141) [73], next locking down sequence of testing and methods to a single model termed ProMisE (Proactive Molecular Risk Classifier for Endometrial Cancer) on a new ‘confirmation’ cohort (n?=?319) [77, 79] to prove feasibility and confirm the association with outcomes/prognosis, and finally testing in a large ‘validation’ cohort (n?=?~500) of ECs from collaborators at the University of Tübingen (Germany). Molecular decision tree analysis for ProMisE is outlined in Fig. 1b. Specific methodologies include immunohistochemistry (IHC) for the detection of the presence/absence of two mismatch repair (MMR) proteins: MSH6 and PMS2. This identifies ‘MMR-D’ (deficient) subgroup. Cases are then sequenced using digital PCR to identify POLE exonuclease domain mutations (‘POLE EDM’). Finally, cases are assessed using IHC for p53 (wild type vs. null or missense mutations; ‘p53wt’ and ‘p53abn’, respectively). We have demonstrated that women within each molecular subgroup have clinicopathological characteristic that have consistently been shown to be typical of that group. For example, the p53 abn subgroup usually encompasses the highest proportion of high grade, advanced stage, non-endometrioid histotypes and arises in older, thinner women. Similarly, the emerging phenotype of women whose EC harbor POLE EDMs is of particular interest since it generally includes younger, thinner and with surprisingly aggressive pathologic features (large proportion of grade 3 tumors, many with deep myometrial invasion and LVSI) yet consistently exhibit favorable outcomes. The MMR-D subgroup have very similar ‘uterine factors’ (clinicopathologic features in the uterus itself) [48] to the POLE subgroup, i.e. a comparable proportion of high grade tumors and deep myometrial invasion and LVSI, yet they have worst observed outcomes of any group next to p53abn [77, 79]. On multivariable analysis, ProMisE molecular subgroup assignment maintained its association with overall survival (OS), progression free survival (PFS) and recurrence free survival (RFS) even after correction of other clinicopathologic parameters of known prognostic significance available at time of diagnosis/collection of diagnostic specimen for molecular analysis (e.g., age, BMI, grade, histotype but not stage).

Both ProMisE (across all tumors tested), and the Leiden classifier (within the intermediate-risk group examined) demonstrate comparable risk discriminatory ability to the ESMO risk stratification system. Furthermore when clinical and pathological features were integrated with molecular features they resulted in improved risk stratification. Through evaluation of the collective cohort (discovery?+?confirmation?+?validation cohorts?=?~ 1000 ECs) we plan to evaluate which key clinicopathological parameters can add value to molecular classification giving high priority to those features available at time of diagnosis (e.g., age, BMI).

Our goal has consistently been to develop a molecular classification tool that could be applied to diagnostic specimens (endometrial biopsy or curettage) and therefore inform treatment at the earliest time point. Biologically relevant information about an individual’s tumor could guide surgical urgency and aggressiveness, fertility or hormonal function sparing management options, adjuvant therapy, and/or surveillance schedules. We have demonstrated high concordance between ProMisE molecular classification in diagnostic vs. final hysterectomy samples, far superseding concordance of grade, or histotype as assigned on original pathology reports or within or between reviews by expert gynecologic cancer pathologists [80]. The Leiden team has also shown high concordance of molecular tumor alterations between pre-operative curettage specimens and final hysterectomy specimens (13 gene panel and MSI assay) [81] and a multicenter, prospective trial in Holland is in process to see if surgical management can be improved [82]. As diagnostic specimens are fixed immediately (in contrast to a hysterectomy specimen that may sit for hours in an operating room before processing in pathology), the quality of DNA extracted and fixation for IHC is high. We believe one of the most exciting aspects of molecular classification and what will be most impactful in directing care for women with EC will be this capability of determining earlier prognostic (and possibly predictive) information.

Ultimately, integration of molecular classification by either method into current practice, as performed on diagnostic specimens or final hysterectomy, will need to be studied in the context of a prospective clinical trial; comparing survival outcomes, quality of life and health economic implications to conventional/historical standard of care.