A Robust Prognostic Signature of Tumor Microenvironment in Colorectal Cancer

Colorectal cancer (CRC) has been a major public health problem. Tumor microenvironment (TME) greatly contributes to the heterogeneity of CRC and is crucial for the regulation of CRC progression. The authors' study aimed to develop a robust prognostic signature for CRC patients based on TME-related genes. Gene expression data and clinicopathologic information of CRC patients were collected from Gene Expression Omnibus and The Cancer Genome Atlas databases. TME-related genes with prognostic value were identified by Cox regression and bootstrap method. The authors used the prognostic genes to construct a robust prognostic model using the least absolute shrinkage and selection operator (LASSO) regression method. The immune and stromal cell abundance of CRC samples were estimated by a microenvironment cell populations-counter method. Based on a training set that comprised 893 CRC samples and 4775 TME-related genes, they established a prognostic model consisting of 25 TME-related genes. With specific risk score formulae, the prognostic model divided CRC patients into high-risk and low-risk subgroups with significantly different survival, which were further confirmed in validation cohorts consisting of other 473 CRC cases or subpopulation of specific stages. The result of time-dependent receiver operating characteristic analysis demonstrated strong predictive accuracy of the prognostic model both in training and validation cohorts. Multivariate Cox regression analysis showed that the 25-gene signature was an independent prognostic factor for overall survival, which was validated through clinical subgroups analysis. Further analysis revealed that CRC samples of high-risk group was abundant of stromal-relevant processes and had a significantly higher proportion of fibroblasts and endothelial cells infiltration. The authors established a robust prognostic signature of 25 TME-related genes which may be an effective tool for prognostic prediction and CRC patient stratification to assist in making treatment decisions.