Abstract
INTRODUCTION: Autophagy can be triggered by oxidative stress and is a double-edged sword involved in the progression of multiple malignancies. However, the precise roles of autophagy on immune response in gastric cancer (GC) remain clarified. METHODS: We endeavor to explore the novel autophagy-related clusters and develop a multi-gene signature for predicting the prognosis and the response to immunotherapy in GC. A total of 1505 patients from eight GC cohorts were categorized into two subtypes using consensus clustering. We compare the differences between clusters by the multi-omics approach. Cox and LASSO regression models were used to construct the prognostic signature. RESULTS: Two distinct clusters were identified. Compared with cluster 2, the patients in cluster 1 have favorable survival outcomes and lower scores for epithelial-mesenchymal transition (EMT). The two subtypes are further characterized by high heterogeneity concerning immune cell infiltration, somatic mutation pattern, and pathway activity by gene set enrichment analysis (GSEA). We obtained 21 autophagy-related differential expression genes (DEGs), in which PTK6 amplification and BCL2/CDKN2A deletion were highly prevalent. The four-gene (PEA15, HSPB8, BNIP3, and GABARAPL1) risk signature was further constructed with good predictive performance and validated in 3 independent datasets including our local Tianjin cohort. The risk score was proved to be independent prognostic factor. A prognostic nomogram showed robust validity of GC survival. The risk score was significantly associated with immune cell infiltration status, tumor mutation burden (TMB), microsatellite instability (MSI), and immune checkpoint molecules. Furthermore, the model was efficient for predicting the response to tumor-targeted agent and immunotherapy and verified by the IMvigor210 cohort. This model is also capable of discriminating between low and high-risk patients receiving chemotherapy. CONCLUSION: Altogether, our exploratory research on the landscape of autophagy-related patterns may shed light on individualized therapies and prognosis in GC.