Development and validation of a prognosis prediction model based on 18 endoplasmic reticulum stress-related genes for patients with lung adenocarcinoma

Long Shu; Shuang Liu; Yongguang Tao

doi:10.3389/fonc.2022.902353

Development and validation of a prognosis prediction model based on 18 endoplasmic reticulum stress-related genes for patients with lung adenocarcinoma

Front Oncol. 2022 Aug 30:12:902353. doi: 10.3389/fonc.2022.902353. eCollection 2022.

Authors

Long Shu^{1

2}, Shuang Liu³, Yongguang Tao^{1

2

4

5}

Affiliations

¹ NHC Key Laboratory of Carcinogenesis (Central South University), Cancer Research Institute School of Basic Medicine, Central South University, Changsha, China.
² Hunan Key Laboratory of Cancer Metabolism, Hunan Cancer Hospital and The Affiliated Cancer Hospital of Xiangya School of Medicine, Central South University, Changsha, China.
³ Department of Oncology, Institute of Medical Sciences, National Clinical Research Center for Geriatric Disorders, Xiangya Hospital, Central South University, Changsha, China.
⁴ Key Laboratory of Carcinogenesis and Cancer Invasion, Ministry of Education, Department of Pathology, Xiangya Hospital, School of Basic Medicine, Central South University, Changsha, China.
⁵ Hunan Key Laboratory of Early Diagnosis and Precision Therapy in Lung Cancer, Department of Thoracic Surgery, Second Xiangya Hospital, Central South University, Changsha, China.

Abstract

Background: Endoplasmic reticulum (ER) stress had a crucial impact on cell survival, proliferation, and metastasis in various cancers. However, the role of ER stress in lung adenocarcinoma remains unclear.

Method: Gene expression and clinical data of lung adenocarcinoma (LUAD) samples were extracted from The Cancer Genome Atlas (TCGA) and three Gene Expression Omnibus (GEO) datasets. ER stress score (ERSS) was constructed based on hub genes selected from 799 ER stress-related genes by least absolute shrinkage and selection operator (LASSO) regression. A Cox regression model, integrating ERSS and the TNM stage, was developed to predict overall survival (OS) in TCGA cohort and was validated in GEO cohorts. Gene set enrichment analysis (GSEA), single-sample GSEA (ssGSEA), and gene mutation analyses were performed to further understand the molecular features of ERSS. The tumor immune infiltration was evaluated by ESTIMATE, CIBERSORT, and xCell algorithms. The receiver operating characteristic (ROC) curves were used to evaluate the predictive value of the risk model. p< 0.05 was considered statistically significant.

Results: One hundred fifty-seven differentially expressed genes (DEGs) were identified between tumor and para-carcinoma tissues, and 45 of them significantly correlated with OS. Next, we identified 18 hub genes and constructed ERSS by LASSO regression. Multivariate analysis demonstrated that higher ERSS (p< 0.0001, hazard ratio (HR) = 3.8, 95%CI: 2.8-5.2) and TNM stage (p< 0.0001, HR = 1.55, 95%CI: 1.34-1.8) were independent predictors for worse OS. The prediction model integrating ERSS and TNM stage performed well in TCGA cohort (area under the curve (AUC) at five years = 0.748) and three GEO cohorts (AUC at 5 years = 0.658, 0.717, and 0.739). Pathway enrichment analysis showed that ERSS significantly correlated with unfolded protein response. Meanwhile, pathways associated with the cell cycle, growth, and metabolism were significantly enriched in the high ERSS group. Patients with SMARCA4, TP53, and EGFR mutations showed significantly higher ERSS (p = 4e-04, 0.0027, and 0.035, respectively). Tissues with high ERSS exhibited significantly higher infiltration of M1 macrophages, activated dendritic cells, and lower infiltration of CD8+ T cells and B cells, which indicate an activated tumor antigen-presenting but suppressive immune response status.

Conclusion: We developed and validated an ER stress-related risk model that exhibited great predictive value for OS in patients with LUAD. Our work also expanded the understanding of the role of ER stress in LUAD.

Keywords: GEO; TCGA; endoplasmic reticulum stress; lung adenocarcinoma; prediction model.