Proteomic analysis uncovers biological diversity in molecularly defined endometrial carcinomas

Abstract

While endometrial cancer has an overall favorable prognosis, some patients have poor outcomes and may benefit from further refinements of the current classification systems. Molecular classification stratifies endometrial cancer patients into four prognostic subtypes: POLEmut, MMRd (mismatch repair deficient), p53abn, and NSMP (no specific molecular profile), where patients with POLEmut have the best prognosis and p53abn has the worst prognosis. We used proteomic profiling to assess if additional prognostic or predictive information could be identified across or within molecular subtypes. Global proteome profiling of formalin fixed, paraffin embedded samples, that had clinicopathologic and outcome data, was performed on 184 endometrial cancers encompassing all four molecular subtypes, including replicate samples of the same tumor, and both biopsy and final hysterectomy specimens. To ensure representation of each subtype, we profiled an approximately equal distribution in the 148 unique tumors; 34 (23%) POLEmut, 40 (27%) MMRd, 35 (24%) p53abn and 39 (26%) NSMP, rather than the population-based distributions. There was high reproducibility in the proteomic profiles of intra-tumor replicate samples, and between matched biopsy and hysterectomy tumor samples. Consensus clustering identified four clusters with different prognosis, named 'Adhesion', 'Immune', 'Proliferation', and 'Metabolic' based on the functional characteristics of the enriched proteins. We associated protein expression features with common mutations, molecular subtype, and outcomes. These results demonstrate the biologic diversity within endometrial cancers, both between and within molecular subtypes, and provide candidate features for functional and clinical investigation.

Keywords: Endometrial carcinoma; Molecular subtype; Proteomics.

Fig. 1

Cohort Description. A. Schematic of proteomic workflow depicting the number of samples across the 4 molecular cohorts. Of the 182 samples in the study, 159 were hysterectomies and 23 were biopsies. The study included 11 replicate hysterectomy samples, defined as being from a different tumor block of the same case. There were 21 biopsy samples which were matched to hysterectomy samples and 2 biopsies that were not matched to hysterectomy. For subtype analysis, the proteomic data from 148 hysterectomy samples was used, consisting of 34 POLEmut, 40 MMRd, 35 p53abn, and 39 NSMP. The cohort features including histotype, grade, lymphovascular invasion (LVI), lymph node positivity and progression free survival (PFS) are displayed for individual samples (B) and as bar plots (C). PFS of 1 indicates a progression event within 5 years, while 0 indicates no progression event. MMMT indicates malignant mixed Mullerian tumor.

Fig. 2

Clustering of the proteomic data. A. Heatmap of consensus clustering of the top 25% most variable proteins (N = 927) resulting in four clusters. The histotype, molecular subtype, and stage are shown above. B. Heatmap showing the Normalized Enrichment Score of pathways associated with each of the four clusters. The clusters were renamed based on gene and pathway enrichment as Adhesion (Cluster 1), Immune (Cluster 2), Proliferation (Cluster 3), and Metabolic (Cluster 4). C. Kaplan-Meier survival curves are shown for the four clusters, disease specific survival (DSS, top) and overall survival (OS, bottom). The p-value for each pair of clusters is shown at the bottom left of the plots.

Fig. 3

Differential protein expression among molecular subtypes. A. Volcano plot comparing protein expression between p53abn (right) and the other molecular subtypes (left). B. Protein abundance of p53 measured from proteomics among the four molecular subtypes. C. Protein abundance of Grb7 among the molecular subtypes. D. Endometrial tumor TMA (16-005, 475 cases) was immunostained for Grb7 and intensity scored as 0 (negative), 1 (weak), 2 (moderate) and 3 (strong). Proportion of the different intensities of Grb7 staining among the molecular subtypes and representative image of staining (20X magnification). E. Kaplan-Meier survival curve for disease specific survival for Grb7 H-score (calculated by multiplying intensity by percent positive staining) in all molecular subtypes.

Fig. 4

Correlation between ARID1A expression and retinoic acid pathway and neutrophil infiltration. A. Volcano plot of proteins showing higher abundance in ARID1A positive (right side) or ARID1A negative (left side) tumors with proteins related to the retinoic acid pathway and bacterial response/neutrophils highlighted. B. Endometrial tumor TMA (16-005, 475 cases) was immunostained for ARID1A, RBP1, CRABP2 and MPO with representative images shown (20X magnification).

Fig. 5

Proteins associated with clinical outcomes. A. Volcano plot of disease specific survival hazard ratios within the NSMP subtype. B. Forest plot showing hazard ratios within MMRd, NSMP, and p53abn subtypes for selected proteins. C. Endometrial tumor TMA (16-005, 475 cases) was stained for NCL. Images of different percent positive NCL staining are shown (Magnification 20X). D. Kaplan-Meier curve for progression free survival for high and low NCL percent positive.