Molecular classification of endometrial cancers using an integrative DNA sequencing panel (271)

Abstract

<b>Objectives:</b> The Cancer Genome Atlas (TCGA) has identified four molecular subgroups of endometrial cancers (EC) with distinct differences in prognosis and response to adjuvant treatment. Stratification of patients according to molecular profiles is critical for individualized therapy at diagnosis, but adoption of molecular classification into clinical practice remains challenging due to the complexity and costs associated with sequencing. We aimed to develop a simple and affordable molecular technique to classify EC into four TCGA subgroups using our high-depth targeted sequencing panel. <b>Methods:</b> Women with newly diagnosed EC were prospectively recruited from three cancer centers in Ontario, Canada, between 2015 and 2018. Tumors were reflexively assessed for mismatch repair (MMR) protein expression by immunohistochemistry (IHC). Using our custom-designed panel, 181 formalin-fixed paraffin-embedded (FFPE) EC samples were sequenced. Our panel was able to identify somatic mutations, copy number variants, insertions and deletions, copy neutral loss of heterozygosity, structural rearrangements, and promoter methylation from a single aliquot of DNA. Variants were analyzed for pathogenicity, and clinical-pathologic information was collected through medical records. <b>Results:</b> Of 181, 86 (48%) were classified into microsatellite instability-high/hypermutated cohort (MSI-h or MMRd), of which 62 (72%) harbored <i>MLH1</i> promoter methylation, and 24 (27%) had pathogenic variants in MMR genes. Concordance with clinical IHC and methylation testing results was high at 99%. Of cases with a single classifier, three (1.6%; <i>n</i>=181) had pathogenic <i>POLE</i> exonuclease domain mutation (<i>POLE</i>mut), 15 (8%) had a pathogenic p53 variant (p53mut), and 61 (34%) had no specific molecular profile subtype (NSMP). Sixteen (9%; <i>n</i>=181) had more than one molecular classifying feature, with eight (4%) having MMRd-p53mut, six (3%) with MMRd-<i>POLE</i>mut, one (0.5%) with MMRd-<i>POLE</i>mut-p53mut, and one (0.5%) with <i>POLE</i>mut-p53mut. Overall, there were 11 (6%) with <i>POLE</i> mutations (three single and nine multiple classifiers). Survival outcomes of MMRd, <i>POLEmut,</i> p53mut, and NSMP groups stratified similarly to TCGA data, but when the MMRd group was further subclassified according to the mechanism of MMR loss, the MLH1 promoter methylated group did worse than those with somatic MMR pathogenic variants (5-year OS: 0.72 [0.57-0.92] vs 1.00 [1.00-1.00]; p=0.013 and RFS: 0.70 [0.58-0.85] vs 0.92 [0.82-1.00]; p=0.032). In contrast, those with somatic MMR pathogenic variants had similar outcomes to the <i>POLE</i>mut subtype (p=0.396) (Figure 1). <b>Conclusions:</b> Our NGS panel can classify EC into four TCGA subgroups through a simplified process and can be implemented reflexively in clinical practice. The higher rate of multiple classifiers (9%) in this study warrants further investigation. The difference in survival between MLH1 methylated group within MMRd cohort suggests that further subclassification may be required for accurate prognostication.