Identification and Validation of a Ferroptosis Prognostic Model for Prostate Cancer Patients through Screening the TCGA and scRNA-seq Datasets.

Abstract

Ferroptosis is currently a common mode of programmed cell death, and the induction of ferroptosis is a new strategy and idea for current tumor therapy. Therefore, the signaling pathways and genes regulated by ferroptosis are significant markers for current malignant tumor therapy. To construct a prognostic model for predicting the survival prognosis of prostate cancer (PCa) based on the molecules and markers related to ferroptosis, we combined with differentially expressed genes (DEGs) in PCa patients, and further analyze the correlation between this risk score and immune cell infiltration. Finally, to validate the expression of risk genes and analyze the expression and localization of risk genes in using the datasets of single cell RNA-sequencing (scRNA-seq). Firstly, we screened the DEGs in PCa patients by the expression profiles of TCGA database. Meanwhile, we collected the information of ferroptosis regulatory genes from FerrDb, and these two parts were intersected. Then the impact of genes on the survival and prognosis of PCa patients was confirmed and selected by LASSO regression, further screening of molecules and fitting the risk format. And the efficacy of the model was evaluated by ROC curves. The immune cell infiltration of PCa tissues was predicted using TIMER. Last, the scRNA-seq of PCa (GSM5155455, GSM3735993) were carried to reveal the expression of risk molecules in different cell types. Besides, the expression of risk molecules was validated using PCa cell lines. We found a total of 259 DEGs associated with ferroptosis in PCa tissues. After LASSO regression, we screened DRD5, LINC00336, ACSF2, RRM2, NOX1, GDF15, ALB, MIOX, and NOX4 as variables to establish a prognostic model, and the specific risk scores was calculated following this format: Risk score = (-1.9465)*DRD5+(-1.6806)*LINC00336+(0.3045)*ACSF2+(0.4747)*RRM2+(-0.2815)*NOX1+(-0.1871)* GDF15+(0.1846)*ALB+(0.2676)*MIOX+(0.1648)*NOX4 (lambda.min = 0.0032), with a 10-yr AUC value of 0.751 (95% CI, 0.549-0.953). Furthermore, we discovered the higher the scores, the fewer CD8+ T cells infiltrated as predicting, showing a negative relationship. By testing the gene sets of scRNA-seq forPCa, we discovered that RRM2, GDF15, ALB, and MIOX were mainly expressed in tumor cells, T cells, B cells and neutrophils of PCa tissues, and not in endothelial cells. Finally, we detected differences in protein expression of RRM2, GDF15, and MIOX in PCa cell lines compared to normal prostate cancer epithelium by WB. We constructed a novel prognostic model for PCa based on ferroptosis-related genes, which showed better predictive validity. And we analyzed the cellular expression of risk genes by scRNA-seq, which will be explored future in relation to prostate cancer radiotherapy.