一种用于预测子宫内膜癌预后和肿瘤免疫微环境的新型铁死亡相关lncRNA-miRNA-mRNA基因特征

Ferroptosis, an iron-dependent form of cell death, is known to be involved in cancer process and tumor immunity, and to be regulated by not only coding (mRNA) but also non-coding genes such as long non-coding RNA (lncRNA) and microRNA (miRNA). However, little is known about the involvement of ferroptosis in endometrial cancer (EC). The aim of this study is to identify comprehensive ferroptosis-related lncRNA-miRNA-mRNA interactions and to construct a ferroptosis-related lncRNA-miRNA-mRNA model for predicting overall survival (OS) and tumor immune microenvironment in EC. Tumor transcriptomes and corresponding clinical data of 544 EC patients were extracted from TCGA, and the ferroptosis database, FerrDb, was used to identify ferroptosis-related coding genes (FRGs) (mRNAs). Ferroptosis-related lncRNAs and miRNAs were selected based on their correlations with FRGs. Univariate, multivariate, and Lasso Cox regression analyses were conducted to construct a prognostic model based on ferroptosis-related lncRNA-miRNA-mRNA genes. EC patients were grouped into high- and low-risk categories based on the risk score which is constructed using the expression levels of ferroptosis-related transcripts. Kaplan-Meier (K-M) analysis and receiver operating characteristic (ROC) curves were used to evaluate the prognostic value. Gene Set Enrichment Analysis (GSEA) was conducted to explore biological pathways between the high- and low-risk groups. Besides, the proportion of infiltrating immune cells and the expression level of immune checkpoints between the risk groups were compared. All signature RNAs were validated using an independent CPTAC cohort (n = 213). Sixteen ferroptosis-related RNAs (10 lncRNAs, 2 miRNAs, and 4 mRNAs) were identified as prognostic markers. A ferroptosis-related lncRNA-miRNA-mRNA co-expression network was constructed. K-M analysis demonstrated that patients in the high-risk group had a worse OS (P < 0.001). ROC curves showed that the area under curve (AUC) values of the model were 0.731, 0.749, and 0.768 for 1, 3, and 5 years of survival, respectively, and the model had a better ability to predict the prognosis of EC patients than other clinical factors (age, grade, and stage). Moreover, the predictive nomogram suggested that our model could offer an independent prognostic evaluation with high accuracy. GSEA revealed that patients in the high-risk group had an enrichment of cancer-related pathways. Tumors in high-risk patients had lower levels of antitumor immunity, and there were several differences in the expression of immune checkpoints between the groups. In the CPTAC dataset, these RNAs were confirmed to be similarly associated with EC prognosis. This study provides new insight into ferroptosis-related molecular mechanisms and novel directions for prognostic assessments, immunotherapies, and targeted treatments of EC.