Transcriptomic and Cellular Content Analysis of Colorectal Cancer by Combining Multiple Independent Cohorts

By linking cellular content and molecular subtypes of colorectal cancer (CRC), we aim to uncover novel features useful for targeted therapy. Our first goal was to evaluate gene expression alterations linked to CRC pathogenesis, and then, we aimed to evaluate the cellular composition differences between normal colon mucosa and tumor and between different colon cancer molecular subtypes.

We collected microarray and RNA sequencing data of patients with CRC from the Genome Expression Omnibus and The Cancer Genome Atlas. We combined all cases and performed quantile normalization. Genes with a fold change of >2 were further investigated. We used xCell for cellular decomposition and CMScaller for molecular subtyping. For statistical analyses, the Kruskal-Wallis H test and Mann-Whitney U tests were performed with Bonferroni correction.

We established an integrated database of normal colon and CRC using transcriptomic data of 1,082 samples. By using this data set, we identified genes showing the highest differential expression in colon tumors. The top genes were linked to calcium signaling, matrix metalloproteinases, and transcription factors. When compared with normal samples, CD4+ memory T cells, CD8+ naive T cells, CD8+ T cells, Th1 cells, Th2 cells, and regulatory T cells were enriched in tumor tissues. The ImmuneScore was decreased in tumor samples compared with normal samples. The CMS1 and CMS4 molecular subtypes were the most immunogenic, with the highest ImmuneScore but also high infiltration by CD8+ T cells, Th1 cells, and Th2 cells in CMS1 and B-cell subtypes and CD8+ T cells in CMS4.

Our analysis uncovers features enabling advanced treatment selection and the development of novel therapies in CRC.