A Proteogenomic View of Synchronous Endometrioid Endometrial and Ovarian Cancer

Abstract

Purpose: Increasing genomics-based evidence suggests that synchronous endometrial and ovarian cancer (SEOC) represents clonally related primary and metastatic tumors. A systematic analysis of the global protein landscape of SEOCs, heretofore lacking, could reveal functional and disease-specific consequences of known genetic alterations, the directionality of metastasis, and accurate histologic markers to distinguish SEOCs from single-site tumors.

Experimental design: We performed a systematic proteogenomic analysis of 29 patients diagnosed with SEOC at three international gynecologic oncology treatment centers (Chicago, Vancouver, and Tübingen). For direct comparison with single-site tumors, we included 9 patients with single-site endometrioid ovarian and 26 patients with single-site endometrioid endometrial cancer (EEC). For all 64 patients, we performed sequencing of a 275-gene cancer panel combined with compartment-resolved mass spectrometry-based proteomics of consecutive tissue sections to compare global (6,000+ proteins), tumor, and stromal proteomes.

Results: DNA-based panel sequencing confirmed that most SEOCs are clonally related. Global proteome profiling uncovered pronounced differences between SEOCs and single tumors and underscored the importance of the stromal proteome in defining and identifying SEOCs. We identified molecularly unique SEOC stromal proteomes, which were globally more related to single endometrial cancers. We finally derived a proteomic predictor distinguishing SEOCs from single-site ovarian and uterine tumors.

Conclusions: The integrated proteogenomic data show that SEOCs are distinguishable from endometrioid endometrial or endometrioid ovarian cancer. Based on their proteogenomic similarity to EECs, we conclude that most SEOCs represent primary EECs that have metastasized to the ovary.

Figure 1.. Proteogenomics elucidates the molecular landscape of synchronous cancers.

A, Study design. Tumor and stromal samples from dual, single endometrial endometrioid, and single endometrioid ovarian cancers were laser micro-dissected and analyzed using label-free shotgun proteomics. In addition, dual and single endometrioid cancer samples were subjected to panel sequencing using the QIASeq Comprehensive Cancer Panel. B, Number of patients and tissue sites used for proteomic analyses. Asterisks indicate the number of samples for which targeted sequencing was performed. C, Top ten mutated genes in SEOCs found in the uterus and ovary. Percentages indicate mutation frequencies per group. Mutation types are color-coded D, Shared, and unique mutations found between the uterine and ovarian anatomical sites of SEOCs. Asterisks indicate cases for which no shared mutations were identified between the uterine and ovarian anatomical sites of SEOCs. E, Probabilistic variant classification across two patients with SEOCs comparing uterine and ovarian cancer sites. Blue rectangles indicate variants present, while white rectangles indicate unknown mutation status. Brighter colors denote higher probability. Genes colored in red indicate likely cancer drivers with increasing support. Genes highlighted in bold indicate genes present in the Cancer Gene Census list. Duplicate genes indicate different mutations found within the same gene. SEOC: Synchronous endometrial and ovarian cancers. Figure 1A Created in BioRender. Coscia, F. (2025) https://BioRender.com/2e8ci03

Figure 2.. Synchronous cancers share mutational signatures with single uterine cancers of endometrioid histology.

A, Representative H&E-stained tissues from dual (patient-matched sites) and single endometroid cancers of the uterus and ovary. Black boxes indicate areas of higher magnification as shown to the right. Scale bar 100-μm. B, Top 30 altered genes shared or unique to dual and single endometroid uterine and single ovarian cancer by site. C, TCGA pan-cancer mutation comparison of genes shared across dual and single endometroid cancers of the uterus and ovary shown in panel B. Uterine endometrioid carcinoma shows the highest alteration frequency for identified shared genes (11 shared genes, 83,3 % of cases, n = 19). D, Heatmap. Unsupervised hierarchical clustering of the top 30 mutated genes based on mutation frequencies. E, Mutation frequencies of selected genes by dual and single endometroid cancer sites (related to panel D).

Figure 3.. Proteomic analysis of the stroma and tumor of single uterine and ovarian as well as synchronous uterine and ovarian cancer.

A, Principal component analysis of all tumor (n=88) and stromal (n=56) proteomes. PC1 (x-axis) and PC2 (Y-axis) account for 31.2% of the total data variability. B, Heatmap showing unsupervised hierarchical clustering of all tumor and stromal samples based on 5,918 unique protein groups. Following data filtering, compartment-specific protein signatures in the tumor and stroma are revealed. Relative protein levels were z-scored to depict up- (red) or down (blue) regulated proteins. C, Boxplots of relative protein levels (z-score) for canonical stromal and tumor markers, related to panel B. D, Number of significantly changed proteins based on all groups (EEC, EOC, SEOC_Ut, EOC_Ov), organ site (uterus versus ovary), cancer type (single versus dual) and collection site (Vancouver, Tübingen, Chicago). Proteins with a false-discovery rate (FDR) of less than 5% were considered significant. E, Global proteome correlations (Pearson r) of synchronous cancers (patient-matched comparison) or between dual (across all patients) and single uterine and ovarian cancers (across all patients). Of note, patient-matched synchronous cancers show significantly higher proteome correlations compared to the other groups. F, Venn diagram showing the overlap between laser micro-dissected tumor and stroma-derived significant proteins (<10%) from panel D.

Figure 4.. Proteomic analysis of the stroma and tumor of single uterine and ovarian as well as synchronous uterine and ovarian cancer.

A, PCA of 353 cancer type (single or dual) specific tumor proteins as shown in Fig. 3D. B, PCA of 418 cancer type (single or dual) specific stroma proteins as shown in Fig. 3D. C, Protein loadings of the stromal PCA shown in panel A. Top proteins along principal component 1 are shown. D, Upper panel: Unsupervised hierarchical clustering of tumor samples after stringent data filtering (Methods) and normalization by anatomical site. The four groups EEC, EOC, SEOC_Ut, SEOC_Ov are represented here by “Group” (Single vs. SEOC) and “Tissue” (Endometrium vs. Ovary). Note, SEOCs show no apparent clustering trend. Lower panel: Corresponding analysis for stromal samples. Note, two clusters were identified with 85% of SEOCs grouping together in the same cluster (dark color). Color bars depict sample groups and molecular subtype according to ProMisE (38,39) determined by the presence of pathogenic POLE mutations (40) (Suplementary table S2), low protein levels of mismatch repair complex (MMR) proteins, namely MLH1, MSH2, MSH6, but not PMS2 since it was not measured, or TP53 mutations. Stromal protein levels are superior to tumor protein levels in differentiating between dual and single tumors. E, Heatmap using unsupervised hierarchical clustering of group averaged proteomes based on 47 histology markers from healthy endometrium and ovary based on enrichment profiles from the Human Protein Atlas (HPA) (upper panels) or the 1,000 most variably expressed proteins in the dataset (lower panels). Of note, synchronous cancers from the ovary are closer to uterine derived single and synchronous cancers than to single ovarian cancers. Stromal samples cluster by anatomic site.

Figure 5.. Machine learning can predict the origin of synchronous cancers using proteomics data.

A, Overview of a machine learning-based stratification strategy to predict dual and single cancers based on compartment-resolved MS-based proteomics using a training and test dataset, CV: cross-validation. B, Receiver operating curve (ROC) curve of SEOC predictor based on the 20 top featured proteins of the stromal compartment. C, Relative protein quantification values of the top 20 stroma-specific protein candidates that discriminate dual from single cancers. D, ROC curve of SEOC predictor based on the 20 top featured proteins of the tumor compartment E, Relative protein quantification values of the top 20 tumor-specific protein candidates that discriminate dual from single cancers.