Stromal-Immune Score-Based Gene Signature: A Prognosis Stratification Tool in Gastric Cancer

Abstract

Background: A growing amount of evidence has suggested the clinical importance of stromal and immune cells in the gastric cancer microenvironment. However, reliable prognostic signatures based on assessments of stromal and immune components have not been well-established. This study aimed to develop a stromal-immune score-based gene signature in gastric cancer. Methods: Stromal and immune scores were estimated from transcriptomic profiles of a gastric cancer cohort from TCGA using the ESTIMATE algorithm. A robust partial likelihood-based Cox proportional hazard regression model was applied to select prognostic genes and to construct a stromal-immune score-based gene signature. Two independent datasets from GEO were used for external validation. Results: Favorable overall survivals were found in patients with high stromal score (p = 0.014) and immune score (p = 0.045). Forty-five stromal-immune score-related differentially expressed genes were identified. Using a robust partial likelihood-based Cox proportional hazard regression model, a gene signature containing SOX9, LRRC32, CECR1, and MS4A4A was identified to develop a risk stratification model. Multivariate analysis revealed that the stromal-immune risk score was an independent prognostic factor (p = 0.018). Based on the risk stratification model, the cohort was classified into three groups yielding incremental survival outcomes (log-rank test p = 0.0004). A nomogram integrating the risk stratification model and clinicopathologic factors was developed. Calibration and decision curves showed a better performance and net benefits for the nomogram. Similar findings were validated in two independent cohorts. Conclusion: The stromal-immune score-based gene signature represents a prognosis stratification tool in gastric cancer.

Keywords: gastric cancer; immune; microenvironment; prediction; prognosis; stromal.

Figure 1

Association of stromal and immune scores with gastric cancer pathology and prognosis. (A) Distributions and comparisons of stromal and immune scores among different tumor stages, differentiation grades, and histology classifications. (B) An illustration of optimal cutoff identification for stromal score. The upper scatter plot shows the standardized log-rank statistic value for each corresponding expression cutoff. The optimal cutoff (stromal score = −186.209) with the maximum standard log-rank statistic is marked with a vertical dashed line. The lower histogram shows the density distribution for low- and high-stromal score groups divided by the optimal cutoff. (C) Kaplan-Meier plot of overall survival for patients with low vs. high stromal scores. (D) Kaplan-Meier plot of overall survival for patients with low vs. high immune scores. (E) Kaplan-Meier plot of overall survival for patients with simultaneously low stromal and immune scores vs. patients with high stromal and immune scores.

Figure 2

Expression profiles and biological functions of stromal and immune score-related DEGs. (A,B) Heatmaps showing expression profiles for stromal score- and immune score-related DEGs with unsupervised hierarchical clustering analyses, using the complete linkage method to measure distances between clusters. (C) Overlap of stromal score- and immune score-related overexpressed DEGs. (D) Overlap of stromal score- and immune score-related underexpressed DEGs. (E) Distribution of prognostic and non-prognostic DEGs among all DEGs. (F) Top six Gene Ontology terms and KEGG pathways enriched by the overexpressed and underexpressed DEGs. P-values were adjusted by false discovery rate.

Figure 3

Forest plot of hazard ratios for 45 stromal-immune score-related prognostic DEGs. Hazard ratios and corresponding 95% confidence intervals were estimated by using the Cox proportional hazard regression model.

Figure 4

Stromal-immune score-based gene signature and risk stratification model. (A) Kaplan-Meier plots of overall survival for patients grouped by expression of four signature genes, SOX9, LRRC32, CECR1, and MS4A4A. (B) Algorithm for risk stratification model based on the expression of four stromal-immune score-related signature genes. (C) Correlation between risk score and gastric cancer patient living months. (D) Kaplan-Meier plot of overall survival for all patients according to risk group. (E) Kaplan-Meier plot of overall survival for patients with stage II gastric cancer according to risk group. (F) Kaplan-Meier plot of overall survival for patients with stage III gastric cancer according to risk group.

Figure 5

Nomogram for predicting overall survival of gastric cancer patients and decision curve analysis. (A) Nomogram to predict 1-, 3-, and 5-years overall survival probability by integrating the stromal-immune score-based risk group and clinicopathologic risk factors. (B) Plot depicting the calibration of the nomogram in terms of the agreement between predicted and observed outcomes. Nomogram performance is shown by the plot relative to the dotted line, which represents perfect prediction. (C) Decision curve analysis of the nomogram. None: assume an event will occur in no patients (horizontal solid line); All: assume an event will occur in all patients (dash line). The graph shows the expected net benefits based on the nomogram prediction at different threshold probabilities.

Figure 6

Validation of stromal-immune score-based risk stratification model in two independent cohorts. (A) Kaplan-Meier plots of overall survival for patients from GEO Series GSE84437, grouped by expression of four signature genes, SOX9, LRRC32, CECR1, and MS4A4A. (B) Correlation between risk scores and living months for patients from GEO Series GSE84437. (C) Kaplan-Meier plot of overall survival according to risk group for patients from GEO Series GSE84437. (D) Kaplan-Meier plots of overall survival for patients from GEO Series GSE62254 grouped by expression of four signature genes, SOX9, LRRC32, CECR1, and MS4A4A. (E) Correlation between risk score and living months for patients from GEO Series GSE62254. (F) Kaplan-Meier plot of overall survival according to risk group for patients from GEO Series GSE62254.