An Immune Cell Signature Is Associated With Disease-Free Survival and Adjuvant Chemosensitivity of Patients With Resectable Gastric Cancer

Abstract

Increasing evidence has indicated that current tumor-node-metastasis (TNM) stage alone cannot predict prognosis and adjuvant chemotherapy benefits accurately for stages II and III gastric cancer (GC) patients after surgery. In order to improve the predictive ability of survival and adjuvant chemotherapy benefits of GC patients after surgery, this study aimed to establish an immune signature based on the composition of infiltrating immune cells. Twenty-eight types of immune cell fractions were evaluated based on the expression profiles of GC patients from the Gene Expression Omnibus (GEO) database using single-sample gene set enrichment analysis (ssGSEA). The immunoscore (IS) was constructed using a least absolute shrinkage and selection operator (LASSO) Cox regression model. Through the LASSO model, an IS classifier consisting of eight immune cells was constructed. Significant difference was found between high-IS and low-IS groups in the training cohort in disease-free survival (DFS, P < 0.0001) and overall survival (OS, P < 0.0001). Multivariate analysis showed that the IS classifier was an independent prognostic indicator. Moreover, a combination of IS and TNM stage exhibited better prognostic value than TNM stage alone. Further analysis demonstrated that low-IS patients who had more tumor-infiltrating lymphocytes had better response to adjuvant chemotherapy. More importantly, we found that patients with high-IS were more likely to benefit from a Xeloda plus cisplatin regimen after surgery. Finally, we established two nomograms to screen the stage II and III GC patients who benefitted from adjuvant chemotherapy after surgery. The combination of IS classifier and TNM stage could predict DFS and OS of GC patients. The IS model has been proven as a promising tool that can be used to identify the patients with stages II and III GC who may benefit from adjuvant chemotherapy.

Keywords: DFS; XP; adjuvant chemotherapy; gastric cancer; immunoscore.

Figure 1

The flowchart through the identification procedure and analyses. LASSO least absolute shrinkage and selection operator.

Figure 2

Construction of the IS model. (A) Forest plots of HRs for tumor infiltrating cells by univariate Cox analysis; (B) LASSO coefficient profiles of the 28 immune cell fractions; Immune cell types: 1.CD4⁺Ta; 2.CD8⁺Ta; 3.aDC; 4.CD56⁺ NK; 5.CD4⁺ Tcm; 6.CD8⁺ Tem; 7. CD4⁺ Tem; 8.CD8⁺Tcm; 9.NK; 10.NKT; 11.Th1; 12.Th17; 13.CD56⁻NK; 14.DCi; 15. Macrophages; 16.MDSC; 17.Neutrophils; 18.pDC; 19.Tregs; 20.Th2; 21.Ba; 22.Eosinophils; 23.γδT; 24.Bi; 25.Mast cells; 26.Bm; 27.Monocytes; 28.Tfh; (C) Ten-fold cross-validation for tuning parameter selection in the LASSO model. Error bars represent confidence intervals for partial likelihood deviance as Λ was changed. The dotted line indicates the optimal values; (D) The IS measured by time-dependent ROC curves in the training cohort. The area under the ROC curve is defined as AUC. HR, hazard ratio; LASSO, least absolute shrinkage and selection operator; ROC, receiver-operating characteristic.

Figure 3

The predictive value of IS for GC patients with surgery. (A) Heatmaps summarizing the distribution of IS and clinical pathological characteristics in the training cohort; (B, C) Kaplan-Meier analysis for the DFS and OS of GC patients in the GSE62254 dataset; (D, E) Kaplan-Meier analysis for the DFS and OS of GC patients in the GSE26253 dataset; (F) The IS measured by time-dependent ROC curves in the testing cohort. The area under the ROC curve was defined as AUC; (G) Distribution of 28 immune cells transformed using ssGSEA in high-IS and low-IS groups. ADJC, adjuvant chemotherapy; Intraabd_LN intraabdominal lymph nodes; DFS, disease free survival; OS, overall survival; GC, gastric cancer; ssGSEA single-sample gene set enrichment analysis; IS, immunoscore; ROC, receiver-operating characteristic. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 4

Kaplan-Meier survival analysis of DFS (left pane) and OS (right pane) according to IS in subgroups of patients with GC in the training cohort. (A) Stage I (n = 30); (B) Stage II (n = 97); (C) Stage III (n = 146); (D) The survival impact of IS in TNM stage subgroup. HR, hazard ratio; IS, immunoscore; DFS, disease free survival; OS, overall survival.

Figure 5

Comparing adjuvant chemotherapy benefit by DFS in total and TNM stage subgroups. Kaplan-Meier curves of (A) total cohort, (B) stage I and (C) stages II and III patients with GC stratified by the receipt of adjuvant chemotherapy; (D) Bubble diagrams describing the association between IS value and immune checkpoint regulators (left pane) and inflammatory cytokines (right pane). Bubble size represents the degree of correlation, bubble color denotes P-value. DFS, disease free survival; GC, gastric cancer.

Figure 6

Kaplan-Meier survival analysis of DFS stratified by types of adjuvant chemotherapy among stages II and III patients. (A) Kaplan-Meier curves of patients in XP group (left panel) and 5-Fu group (right panel) stratified by IS; (B) Kaplan-Meier curves of patients in Low-IS group (left panel) and high-IS group (right panel) stratified by chemotherapy regimen; (C) Forest plot describing the benefit of chemotherapy regimen in different IS groups. IS, immunoscore; XP, Xeloda plus cisplatin; 5-Fu 5-fluorouracil.

Figure 7

Treated and untreated nomograms to predict the probability of 1, 2, and 3-year recurrence with or without adjuvant chemotherapy in GC. (A) Treated nomogram predicting 1-, 2-, and 3-year DFS for GC patients after surgery with adjuvant chemotherapy; (B) Calibration curves to validate treated nomogram for 1-, 2-, and 3-year DFS; (C) Untreated nomogram predicting 1-, 2-, and 3-year DFS for GC patients after surgery without adjuvant chemotherapy; (D) Calibration curves to validate untreated nomogram for 1-, 2-, and 3-year DFS. DFS, disease free survival; OS, overall survival; GC, gastric cancer.