Integrated multi-omic analyses reveals clinical relevance of endometrial cancer cell line models

Abstract

<h3>Objectives:</h3> Endometrial cancer (EC) cell lines models established from diverse ancestral backgrounds and reflective of <i>in situ</i> disease characteristics are necessary to support preclinical investigations with broad generalizability. We performed multi-omic analyses of EC cell line models established from women representing diverse ancestries and compared these with clinical and molecular profiling data from the TCGA UCEC patient cohort to assess the relevance of models to <i>in situ</i> clinical and molecular disease characteristics. <h3>Methods:</h3> Whole genome sequencing (WGS), transcriptome (RNAseq) and proteome (LC-MS/MS and RPPA) analyses were performed for primary cell line models established from endometrioid EC patients (NCI-EC1, ACI-80, ACI-181, ACI-52, ACI-61 and ACI-68) as well as commercial EC models (MFE-296, SNG-M, HEC1A, RL95-2, Ishikawa, AN3-CA, KLE). Microsatellite instability (MSI) was determined from WGS data using MSISensor2 and standard ancestry estimates by comparison with reference populations from the 1000 Genomes Project. Hierarchical cluster analyses was performed using Pvclust and differential analyses using LIMMA. TCGA UCEC clinical and transcriptome data (n=371 patients) were downloaded from the TCGA UCEC manuscript supplement or cbioportal.org and odds ratio significance was assessed using Fisher's exact test (p<0.05). <h3>Results:</h3> EC cell line models are MSI-high except ACI-61, ACI-68 and KLE cells which are microsatellite stable. Standard ancestry estimates showed most cell lines are from women of European ancestry except Ishikawa, HEC1A and SNG-M which are of East Asian ancestry and NCI-EC1, ACI-181 and ACI-80 which are of African ancestry. Unsupervised hierarchical analysis of proteome data revealed distinct clusters comprising MFE-296, AN3-CA, ACI-68, ACI-61, NCI-EC1 cells (cluster 1) and ACI-80, ACI-52, HEC1A, SNG-M, RL95-2, KLE, ACI-181 and Ishikawa cells (cluster 2). Differential analyses showed elevated levels of mesenchymal markers (eg, vimentin) and lower levels of epithelial markers (eg, E-cadherin) in cluster 1 versus 2 cells. Correlation of transcript-level data with TCGA UCEC patients showed all cell lines are significantly correlated with endometrioid versus mixed/ uterine serous carcinoma tumors except NCI-EC1, ACI-52, HEC1A and KLE cells. Most cell lines significantly correlated with MSI-H patient tumors and largely trended as being correlated with tumors harboring high mutational loads, except ACI-68 and KLE cells. Cluster 2 cell lines are largely correlated with tumors exhibiting low somatic copy-number alterations except KLE, Ishikawa and HEC1A cells and are significantly correlated with tumors classified as hormonal versus mitotic or immunoreactive molecular subtypes except KLE cells. <h3>Conclusions:</h3> We describe EC cell line models established from women of diverse patient ancestries including those from women of African descent (i.e. NCI-EC1, ACI-181 and ACI-80 cells) that represent novel preclinical resources to investigate ancestry-linked molecular alterations underlying racial disparities between European and African-American women diagnosed with EC. We further define EC cell line models correlating with endometrioid tumors that exhibit more epithelial-like (elevation of E-Cadherin) as well as hormonal characteristics that will benefit preclinical investigations of endometrioid EC disease biology.