Construction and clinical application of a risk model based on N6-methyladenosine regulators for colorectal cancer

Colorectal cancer (CRC) shows a high incidence in developed countries. This study established a prognosis signature based on N6-methyladenosine (m6A) regulators involved in CRC progression.

The bulk RNA-seq data from the Atlas and Compass of Immune-Colon cancer-Microbiome interactions (AC-ICAM) and GSE33113 CRC datasets were obtained from the cBioportal and GEO databases, and a total of 21 m6A regulators genes were collected from a previous study. The scRNA-seq analysis of the GSE146771 cohort was conducted applying the Seurat and harmony R packages. Consensus clustering based on the expressions of m6A regulators was performed with the ConsensusClusterPlus package. The ggGSEA package was used for the Gene Set Enrichment Analysis (GSEA). The un/multivariate and LASSO Cox analysis were performed applying the "survival" and "glmnet" packages for developing a risk model. The pRRophetic and GSVA packages were utilized to analyze potential drugs for CRC and immune infiltration in different risk groups, followed by the Kaplan-Meier (KM) survival and ROC analysis with the "survival" and "timeROC" packages. In vitro assays included the quantitative polymerase chain reaction (qPCR), wound healing and transwell were performed.

CRC patients in the AC-ICAM cohort were assigned into three molecular subtypes (S1, 2 and 3) based on nine m6A regulator genes. Specifically, the prognostic outcome of the S3 was the most favorable, while that of the S1 was the worst and this subtype was associated with the activation of NF-kB, TNF-α and hypoxia pathways. Three key genes, namely, methyltransferase-like 3 (METTL3), insulinolike Growth Factor2 mRNA-Binding Protein 3 (IGF2BP3) and YTH domain-containing protein 2 (YTHDC2), selected from the 9 m6A regulator genes were combined into a RiskScore, which showed a high classification effectiveness in dividing the patients into high- and low-risk groups. Inhibition of the expression of METTL3A or that of IGF2BP3 suppressed the invasion and migration of CRC cells. Notably, the high-risk patients had higher immune cell infiltration to support the activation of multiple immune responses and exhibited significantly poor prognosis. Meanwhile, a nomogram with practical clinical value was developed based on the RiskScore and other clinical features. Finally, eight potential drugs associated with the RiskScore were identified, and CD4+ cells and Tregs were found to be closely associated with CRC progression.

The RiskScore model developed based on m6A regulators played a critical role in CRC development and can be considered as a prognosis predictor for patients with the cancer. The present discoveries will facilitate the diagnosis and clinical management of CRC patients.