Abstract 3193: Development of a colon cancer model system reveals epithelial contribution to poor-prognosis gene signatures

Abstract

Abstract Background: Recent consensus on molecular classification categorizes colorectal cancer (CRC) into 4 robust subtypes: CMS1 (epithelial-MSI), CMS2 (epithelial-canonical), CMS3 (epithelial-metabolic) and CMS4 (mesenchymal)1. CMS4 is linked to poor cancer prognosis and characterized by mesenchymal and epithelial-to-mesenchymal transition (EMT) gene expression2,3. Recent attempts to deconvolute the transcriptome from CRC tumors have suggested that the mesenchymal gene expression results from a large stromal compartment and is not due to epithelial cells with EMT-like features4,5. This challenges the classic notion that tumor cells activate the EMT program to undergo trans-differentiation into mesenchymal cells as observed in a number of in vitro and in vivo studies6. To resolve these conflicting views a detailed investigation of the tumor or stromal cell contributions to the mesenchymal signature is important. Material & methods: We developed patient-derived xenografts (PDX) from multiple subtypes of CRCs. Transcriptome analysis was performed to assign each tumor and PDX to a CMS subtype, and further dissect the contribution of both epithelial and stromal gene expression to the overall expression phenotype. Additionally, we established multiple cultures from primary human CRCs to further assess subtypes-differences. Results: We developed a panel of human CRC PDXs from all subtypes (CMS1-4). The expression profiles of the patient's tumor and PDXs remain stable for multiple serial passages in vivo. Evaluating the gene expression in the PDXs, which are free of human stroma, our analyses identified tumor specific expression markers that were differentially expressed in the mesenchymal subtype (CMS4). This finding suggests the presence of an epithelial tumor derived signal in mesenchymal tumors which contributes to CMS4 specific gene expression. This set of epithelial specific CMS4 tumor markers has been validated in other CRC expression datasets and associated with disease relapse. Our panel of primary cell lines representing various subtypes showed differential sensitivity to certain therapeutic agents thus indicating a subtype-specific drug response. Conclusions: The identification of diverse subtypes in our panel of CRCs and corresponding PDXs, shows the presence of distinct tumor specific gene expression in mesenchymal CMS4 tumors. Furthermore, the primary CRC cell line panel comprising all subtypes, provides great insight into CRC heterogeneity and serves as a valuable model system to accelerate the development of effective treatment.