Construction and Validation of a Universal Applicable Prognostic Signature for Gastric Cancer Based on Seven Immune-Related Gene Correlated With Tumor Associated Macrophages

Abstract

Background: Tumor-associated macrophages (TAMs) play a critical role in the progression of malignant tumors, but the detailed mechanism of TAMs in gastric cancer (GC) is still not fully explored.

Methods: We identified differentially expressed immune-related genes (DEIRGs) between GC samples with high and low macrophage infiltration in The Cancer Genome Atlas datasets. A risk score was constructed based on univariate Cox analysis and Lasso penalized Cox regression analysis in the TCGA cohort (n=341). The optimal cutoff determined by the 5-year time-dependent receiver operating characteristic (ROC) curve was considered to classify patients into groups with high and low risk. We conducted external validation of the prognostic signature in four independent cohorts (GSE84437, n=431; GSE62254, n=300; GSE15459, n=191; and GSE26901, n=109) from the Gene Expression Omnibus (GEO) database.

Results: The signature consisting of 7 genes (FGF1, GRP, AVPR1A, APOD, PDGFRL, CXCR4, and CSF1R) showed good performance in predicting overall survival (OS) in the 5 independent cohorts. The risk score presented an obviously positive correlation with macrophage abundance (cor=0.7, p<0.001). A significant difference was found between the high- and low-risk groups regarding the overall survival of GC patients. The high-risk group exhibited a higher infiltration level of M2 macrophages estimated by the CIBERSORT algorithm. In the five independent cohorts, the risk score was highly positively correlated with the stromal cell score, suggesting that we can also evaluate the infiltration of stromal cells in the tumor microenvironment according to the risk score.

Conclusion: Our study developed and validated a general applicable prognostic model for GC from the perspective of TAMs, which may help to improve the precise treatment strategy of GC.

Keywords: gastric cancer; immune; macrophages; prognostic; signature.

Figure 1

The workflow chart of this study.

Figure 2

Identification of differential expressed immune-related genes(DEIRGs) associated with macrophage infiltration. (A) The high infiltration by Macrophages is unfavorable for the OS of GC. (B). The vol plot DEIRGs. (C) The circle plot of GO terms up-regulated in the high macrophage infiltration group. (D) The circle plot of GO terms up-regulated in the low macrophage infiltration group.

Figure 3

The building process of the seven immune gene prognostic signature in the TCGA cohort. (A) The forrest plot of the univariate Cox analysis. (B) The corrplot of the prognostic related genes. (C) Lasso penalized COX regression analysis. (D) The optimal cutoff determined by 5-year time-depend ROC curve. (E, F) The Kaplan–Meier survival analysis and time‑dependent ROC analysis of the signature for predicting the OS of patients in the TCGA cohort. (G) The risk score distribution of patients in in the TCGA cohort.

Figure 4

The correlation analysis of the signature and the macrophage infiltration.

Figure 5

Independence validation of the risk score in the TCGA cohort. (A, B) The heatmap, and the survival status of patients in in the TCGA cohort. (C) The forrest plot of the univariate Cox analysis. (D) The forrest plot of the multivariate Cox analysis. (E, F) Subgroup validation based on the clinical stage.

Figure 6

External validation of the prognostic model. (A, B) The Kaplan–Meier survival analysis and the time‑dependent ROC analysis of the signature for predicting the OS of patients in the GSE84437 cohort. (C) The heatmap, distribution of risk score, and the survival status of patients in in the GSE84437 cohort. (D, E) The Kaplan–Meier survival analysis and the time‑dependent ROC analysis of the signature for predicting the OS of patients in the GSE62254 cohort. (F) The heatmap, distribution of risk score, and the survival status of patients in in the GSE62254 cohort. (G, H) The Kaplan–Meier survival analysis and the time‑dependent ROC analysis of the signature for predicting the OS of patients in the GSE15459 cohort. (C) The heatmap, distribution of risk score, and the survival status of patients in in the GSE15459 cohort. (J, K) The Kaplan–Meier survival analysis and the time‑dependent ROC analysis of the signature for predicting the OS of patients in the GSE26901 cohort. (L) The heatmap, distribution of risk score, and the survival status of patients in in the GSE26901 cohort.

Figure 7

External independence validation of the prognostic model in the (A) GSE84437 cohort (B) GSE62254 cohort (C) GSE15459 cohort (D) GSE26901 cohort. *green represents the univariate Cox analysis, red represents the multivariate Cox analysis.

Figure 8

The difference of immune cell infiltration between high- and low-risk groups. (A) The circos plot of risk score and the infiltration of six types of immune cells. (B) The vioplot showed the difference of the abundances of six immune infiltrates (TIMER algorithm) between high- and low risk groups in the TCGA cohort. (C) The heatmap of 22 kinds of immune cells infiltration. (D) The corHeatmap of 22 kinds of immune cells infiltration. (E) The vioplot showed the difference of the abundances of 22 types of immune cells infiltrates between high- and low risk groups[CIBERSORT algorithm, red represent high risk(n=527), blue represent low risk(n=509)].

Figure 9

The landscape of tumor microenvironment for difference of immune cell infiltration between high- and low-risk groups (A) TCGA cohort (B) GSE84437 cohort (C) GSE62254 cohort (D) GSE15459 cohort (E) GSE26901 cohort.

Figure 10

The correlation between genes and immune function.

Figure 11

The correlation between risk score and Stromal score (A) TCGA cohort (B) GSE84437 cohort (C) GSE62254 cohort (D) GSE15459 cohort (E) GSE26901 cohort.