Abstract
Background: Lung adenocarcinoma (LUAD) has a high mortality rate. The signaling of endoplasmic reticulum (ER) stress sensor can regulate cancer progression. We aimed to investigate the relationship between ER stress-related genes and LUAD prognosis. Methods: Gene expression and clinical data of LUAD patients were downloaded from The Cancer Genome Atlas (TCGA) database. GSE68465 from Gene Expression Omnibus (GEO) was used for validation analysis. Two ER stress-related datasets were downloaded from MSigDB. ER stress-related genes which significantly correlated with the prognosis of LUAD were selected to construct ER score model. Patients were divided into two groups based on the median value of ER sore. The prognosis of patients in different groups was compared. Unsupervised hierarchical clustering analysis was used to identify different LUAD subtypes. Features with significantly correlation of prognosis were used to construct a nomogram. The expression of key ER stress-related genes in LUAD cells was experimentally validated. Results: Seven genes (EIF2AK3, IGFBP1, SHC1, GSK3A, EIF4G1, MBTPS2, and PSMC6) significantly associated with prognosis were used to construct the ER score model. The patients in the high ER score group had worse prognosis in both TCGA and GEO datasets. Patients with higher ER scores tended to have higher pathologic_T, pathologic_N, pathologic_M, and stage. There was a significant difference in prognosis between the two subtypes identified according to the model gene (p < 0.05). The nomogram constructed with stage, ER score and cluster showed a great performance for predicting prognosis. Experimental validation confirmed the downregulation of EIF2AK3 and the upregulation of IGFBP1, SHC1, GSK3A, EIF4G1, MBTPS2, and PSMC6 in LUAD cells. Conclusion: The nomogram constructed by ER score and clinical features showed good predictive performance on LUAD. This study provided new insights into understanding the role of ER stress in LUAD.