Journal of Translational Medicine (Jun 2021)

Molecular subtype identification and prognosis stratification by a metabolism-related gene expression signature in colorectal cancer

  • Dagui Lin,
  • Wenhua Fan,
  • Rongxin Zhang,
  • Enen Zhao,
  • Pansong Li,
  • Wenhao Zhou,
  • Jianhong Peng,
  • Liren Li

DOI
https://doi.org/10.1186/s12967-021-02952-w
Journal volume & issue
Vol. 19, no. 1
pp. 1 – 15

Abstract

Read online

Abstract Background Metabolic reprograming have been associated with cancer occurrence and progression within the tumor immune microenvironment. However, the prognostic potential of metabolism-related genes in colorectal cancer (CRC) has not been comprehensively studied. Here, we investigated metabolic transcript-related CRC subtypes and relevant immune landscapes, and developed a metabolic risk score (MRS) for survival prediction. Methods Metabolism-related genes were collected from the Molecular Signatures Database and metabolic subtypes were identified using an unsupervised clustering algorithm based on the expression profiles of survival-related metabolic genes in GSE39582. The ssGSEA and ESTIMATE methods were applied to estimate the immune infiltration among subtypes. The MRS model was developed using LASSO Cox regression in the GSE39582 dataset and independently validated in the TCGA CRC and GSE17537 datasets. Results We identified two metabolism-related subtypes (cluster-A and cluster-B) of CRC based on the expression profiles of 539 survival-related metabolic genes with distinct immune profiles and notably different prognoses. The cluster-B subtype had a shorter OS and RFS than the cluster-A subtype. Eighteen metabolism-related genes that were mostly involved in lipid metabolism pathways were used to build the MRS in GSE39582. Patients with higher MRS had worse prognosis than those with lower MRS (HR 3.45, P < 0.001). The prognostic role of MRS was validated in the TCGA CRC (HR 2.12, P = 0.00017) and GSE17537 datasets (HR 2.67, P = 0.039). Time-dependent receiver operating characteristic curve and stratified analyses revealed the robust predictive ability of the MRS in each dataset. Multivariate Cox regression analysis indicted that the MRS could predict OS independent of TNM stage and age. Conclusions Our study provides novel insight into metabolic heterogeneity and its relationship with immune landscape in CRC. The MRS was identified as a robust prognostic marker and may facilitate individualized therapy for CRC patients.

Keywords