Classification of organoid transcriptomic profiles unravelling colorectal cancer molecular subtypes

Abstract

Colorectal cancer (CRC) is genetically and transcriptomically heterogeneous disease. Molecular subtyping of colorectal cancer using consensus molecular subtype (CMS) system demonstrated the potential predictive value for tumor progression and treatment response. However, the CMS system was developed from data of whole tissues containing both cancer and non-tumor transcripts components for classification which does not directly represent intrinsic heterogeneity of cancer cells. In this study genetic profiles of CRC organoids were investigated first, and the results indicate chromosomal instability (CIN) and microsatellite instability (MSI) as pathogenic pathways of CRC. Furthermore, the results also revealed diverse patterns of somatic mutations of these CRC organoids. Subsequently, we evaluated a strategy of subtyping CRCs based on transcriptomics data from patient-derived CRC organoids, which mainly contain cancer cells. We demonstrated that using non-negative matrix factorization (NMF) CRC cancer organoids could be classified into four groups (P1-P4). Cluster-specific genes and Gene Set Enrichment Analysis (GSEA) displayed different characteristics of each group. P1 exhibit enriched lipid and cholesterol metabolism pathways and P2 and P3 presented high TGF-β pathway. Lastly, P4 show stem cell-like properties and highly expressed genes in the DNA repair pathway associated with chemotherapy and radiation resistance. Moreover, P4 organoids present a hyperactivated ribosome biogenesis pathway which may be developed as a biomarker of P4 and a target of CRC treatment. Then, LASSO logistic regression was built to identify gene signatures and developed a classifier of each group of organoids. These results suggested that the signature gene of organoid groups has the potential to be developed into a useful tool for CRC subtyping and developing more specific therapeutic strategies.