A SERPINE1-Based Immune Gene Signature Predicts Prognosis and Immunotherapy Response in Gastric Cancer

Abstract

Immune checkpoint inhibitors (ICIs) therapy has been successfully utilized in the treatment of multiple tumors, but only a fraction of patients with gastric cancer (GC) could greatly benefit from it. A recent study has shown that the tumor microenvironment (TME) can greatly affect the effect of immunotherapy in GC. In this study, we established a novel immune risk signature (IRS) for prognosis and predicting response to ICIs in GC based on the TCGA-STAD dataset. Characterization of the TME was explored and further validated to reveal the underlying survival mechanisms and the potential therapeutic targets of GC. The GC patients were stratified into high- and low-risk groups based on the IRS. Patients in the high-risk group, associated with poorer outcomes, were characterized by significantly higher immune function. Further analysis showed higher T cell immune dysfunction and probability of potential immune escape. In vivo, we detected the expressions of SERPINE1 by the quantitative real-time polymerase chain reaction (qPCR)in tumor tissues and adjacent normal tissues. In vitro, knockdown of SERPINE1 significantly attenuated malignant biological behaviors of tumor cells in GC. Our signature can effectively predict the prognosis and response to immunotherapy in patients with GC.

Keywords: gastric cancer; immune checkpoint inhibitor; immune risk signature; tumor microenvironment; tumor mutation load.

Figure 1

Flow chart of the study design and construction of the IRS. (A) Flow chart of our study. (B) DEGs in GC vs. adjacent normal tissues. (C) These DEGs were intersected with a combined IRGs set. (D) LASSO coefficient profiles of genes in TCGA-STAD. (E) Heatmap of the risk score consisting of four IRGs. (F) Selecting the best parameters by ten-fold cross-validation in the LASSO model. (G) Distribution of risk score, survival time, and status of patients in TCGA-STAD cohort.

Figure 2

Relationship between risk score and patients’ survival in different cohorts of GC patients. K–M curves of OS according to risk groups in the TCGA-STAD discovery cohort (A), GSE62254 validation cohort (B), and GSE26253 validation cohort (C). Forest plot of multivariate regression analyses showing the association between risk score and patients’ survival in the three cohorts (D–F).

Figure 3

Transcriptome traits and clinical characteristics of TME phenotypes in the TCGA-STAD cohort. (A) ssGSEA identified the relative infiltration of 28 tumor-infiltrating immune cell types with different risk groups in the TCGA-STAD cohort. (B) Correlation matrix of risk score, four IRGs and seven immune-checkpoint-relevant genes (Spearman test, p < 0.05). (C,D) The relationship between risk score and 28 tumor-infiltrating immune cell types or pathways using ssGSEA. (E–G) Comparison of Immune, TIDE scores, and IPS between high- and low-risk groups in TCGA-STAD cohort. Wilcoxon test, ** p < 0.01; *** p < 0.001; **** p < 0.0001; ns, not statistically significant.

Figure 4

Comparison of tumor mutations between high- and low-risk groups. (A,B) The oncoPrint of high- and low-risk groups in the TCGA-STAD cohort. (C) TML difference in the high- and low-risk groups. Wilcoxon test, *** p < 0.001. (D) Kaplan–Meier curves for high- and low-TML groups of the TCGA-STAD cohort. Log-rank test, p = 0.011. (E) The relationship between risk score and TML in TCGA-STAD cohort (Spearman test, p < 0.0001). The dotted color indicates the ACRG molecular subtypes of GC. (F) TML difference in different ACRG molecular subtypes of the TCGA-STAD cohort. Steel–Dwass test, * p < 0.05; *** p < 0.001; ns, not statistically significant. (G,H) Distribution of the risk score and percentage of the high-risk group in different ACRG molecular subtypes of the TCGA-STAD cohort. Steel–Dwass test, **** p < 0.0001; ns, not statistically significant.

Figure 5

Response to immunotherapy in patients of urothelial cancer and melanoma with different risk groups divided by risk score. (A) K–M curve for patients of urothelial cancer in the high- and low-risk groups in the IMvigor210 cohort. Log-rank test, p = 0.0013. (B) Risk score in groups with a different clinical response to immunotherapy in the IMvigor210 cohort. Wilcoxon test, * p < 0.05. (C) The proportion of patients with a response to immunotherapy in high- and low-risk groups in the IMvigor210 cohort. Pearson’s Chi-squared test, p = 0.0002. (D) K–M curve for patients of melanoma in the high- and low-risk groups in the GSE91061 cohort (Log-rank test, p = 0.024). (E) Risk score in groups with a different clinical response to immunotherapy in the GSE91061 cohort. Wilcoxon test, * p < 0.05. (F) The proportion of patients with response to ICIs therapy in high- and low-risk groups in the GSE91061 cohort. Fisher’s exact test, p = 0.0315.

Figure 6

Construction and evaluation of a nomogram based on the IRS to predict the 3-year and 5-year OS for patients with gastric cancer. (A) Nomogram was constructed with the TCGA-STAD cohort for predicting the probability of 3-year and 5-year OS for GC patients. (B) Calibration plot for 3-year and 5-year OS in the TCGA-STAD cohort. (C) Calibration plot for 3-year and 5-year OS in the GSE62254 cohort.

Figure 7

SERPINE1 promotes the malignant biological behaviors of GC cells. SERPINE1 (A), APOD (B), GANI1 (C) and BMP1 (D) mRNA and SERPINE1 (E) protein levels were measured in GC tissues, and adjacent normal tissues were paired by qPCR and Western blotting, respectively. Western blotting analyses of SERPINE1 protein levels in GC cell lines (F). Western blot and qPCR analyses of SERPINE1 levels in GC cell lines transfected with the siSERPINE1 (G–J). CCK8 assay of cell growth with SERPINE1 silencing and control group (K,L). Representative images and quantification of migration and invasion in SERPINE1 silencing and control GC cell lines (M,N,P,Q). Apoptosis assay and the quantitative analysis of SERPINE1 in silencing and control group (O,R,S). Data were presented as the mean ± SD. Wilcoxon test, * p < 0.05; ** p < 0.01; **** p < 0.0001.