The aging-related risk signature in colorectal cancer

Abstract

Background: Colorectal cancer (CRC) is the third most common cancer worldwide. The opening of the TCGA and GEO databases has promoted the progress of CRC prognostic assessment, while the aging-related risk signature has never been mentioned.

Methods: R software packages, GSEA software, Venn diagram, Metascape, STRING, Cytoscape, cBioPortal, TIMER and GeneMANIA website were used in this study.

Results: Aging-related gene sets, GO_AGING, GO_CELL_AGING and GO_CELLULAR_SENESCENCE, were activated significantly in CRC tissues. We constructed an aging-related risk signature using LASSO COX regression in training group TCGA and validated in testing group GSE39582. The risk score was significantly associated with the overall survival of CRC patients, whose stability was clarified by stratified survival analysis and accuracy was demonstrated using the ROC curve. The risk score was significantly increased in the advanced stage, T3-4, N1-3 and M1 and positively correlated with the richness of immune cell infiltration in the tumor microenvironment. We further investigated the molecular characteristics of 15 hub genes at the DNA and protein levels and performed GSEA between high- and low-risk groups.

Conclusions: The aging-related signature is a reliable prognostic analysis model and can predict the severity and immune cell infiltration of CRC patients.

Keywords: GEO; TCGA; aging; colorectal cancer; risk signature.

Figure 1

Gene Set Enrichment Analysis (GSEA). Three aging-related gene sets were significantly activated in colorectal cancer (CRC) tissues compared with normal tissues. The significance criteria were nominal P-value < 5% and FDR q-value<25%.

Figure 2

Identification of differentially expressed and prognostic aging-related genes in CRC. (A) Aging-related genes shared by the training group TCGA and testing group GSE39582. (B) Differentially expressed aging-related genes in the TCGA was displayed in the heatmap and (C) the volcano map. (D) Gene ontology (GO) analysis of these genes. (E) Forest plot of prognostic aging-related genes in the training group.

Figure 3

Protein-protein interaction (PPI) of differentially expressed aging-related genes. (A) In the PPI network, the darker the color was, the greater the number of neighboring nodes was. (B) Top 30 genes with the most neighboring nodes. (C) The first MCODE component identified in this gene list and pathway and process enrichment analysis of this MCODE component was significantly related to the cell cycle.

Figure 4

Prognostic signature based on 15 hub genes. (A) Distribution of groups based on the aging-related risk score. (B) The scatter plot demonstrated the differences in the survival status of CRC patients between high- and low-risk groups. (C) Heatmap showed differential expression of included 15 hub genes in both groups. (D) The overall survival (OS) of the high-risk group was significantly shorter than that of the low-risk group. (E–H) The second verification was performed in the testing group GSE39582.

Figure 5

Validation of prognostic signature. (A, B) Univariate and multivariate COX regression analysis in the training group. (C, D) Univariate and multivariate COX regression analysis in the testing group. The receiver operating characteristic (ROC) curve and the areas under the curve verified the accuracy of prognostic signature in the (E) training and (F) testing groups.

Figure 6

Stratified survival analysis adjusted to age, gender, stage, T, N and M. All CRC patients in the training and testing groups were summarized in the stratified survival analysis. 68 years old was the median age of 1,015 CRC patients.

Figure 7

Relationships between risk score and clinicopathological traits. The aging risk score of (A) stage III & IV, (B) T3-4, (C) N1-3 and (D) M1 were significantly higher than that of stage I&II, T1-2, N0 and M0 in the training group (P < 0.05). (E) The same conclusion was obtained in the testing group.

Figure 8

Pearson correlation analysis between the risk score and infiltration abundances of 6 types of immune cells in the training group. (A) B cells; (B) CD4+T cells; (C) CD8+T cells; (D) neutrophils; (E) macrophages and (F) dendritic cells. P < 0.05 was considered statistically significant.

Figure 9

Mutation and copy number alteration (CNA) analysis of hub genes. (A) Frequency of mutation and CNA in hub genes in 3 types of CRC patients; (B) Mutation and CNA of each hub gene.

Figure 10

GeneMANIA website was used to identify functionally similar genes and establish a PPI network. The 20 functionally similar genes were located in the outer circle, while hub genes were located in the inner circle. The color of nodes was related to the protein function while line color represented the type of protein interaction.

Figure 11

In the 50 hallmark gene sets, we conducted GSEA between the high- and low-risk groups. (A) Significant enrichment of 27 hallmark gene sets in the high-risk group of training group TCGA; (B) Significant enrichment of 18 hallmark gene sets in the low-risk group of the training group; (C) Significant enrichment of 24 hallmark gene sets in the high-risk group of testing group GSE39582; (D) Significant enrichment of 9 hallmark gene sets in the low-risk group of the testing group. Dark black represented the enrichment results common to both datasets (Nominal P-value < 5% and FDR q-value<25%).