Hypoxia-Related Gene Signature Predicts Biochemical Recurrence Of Localized Prostate Cancer After Radical Prostatectomy

Background and ObjectiveProstate cancer(PCa)is the second most common malignant tumor in men worldwide,and most of the newly diagnosed patients are localized PCa.At present,the main basis for predicting biochemical recurrence(BCR)after radical prostatectomy(RP)for localized PCa is based on clinical parameters,and its prediction accuracy rate is low.In view of the important role of hypoxia in the occurrence and development of PCa,we aim to establish a new prognostic model based on hypoxia-related gene characteristics and clinicopathological parameters to improve the risk stratification of BCR and assist clinical decision-making.MethodsWe downloaded the RNA sequences and clinicopathological data of 323 and 87 PCa patients from The Cancer Genome Atlas(TCGA)database and Gene Expression Omnibus(GEO)database,respectively.Hypoxia-related gene sets were mainly obtained from the Gene Set Enrichment Analysis(GSEA)website.Protein-protein interaction network(PPI)and Cytoscape software were used to screen the key genes in hypoxia-related genes.Univariate Cox analysis and Lasso regression analysis were used to screen the prognostic genes associated with BCR and established a risk model,and according to the median value of the risk score,patients were divided into high and low hypoxia risk groups.We used the risk curve,Kaplan-Meier(KM)survival analysis,univariate Cox analysis and multivariate Cox analysis to evaluate the prognostic value of the model.Gene expression heat maps were used to analyze the expression of each hypoxia gene in the high and low risk groups.Box plots were used to analyze the relationship between hypoxia genes and clinicopathology.GSEA was used to analyze the main enrichment pathways of genes in the high hypoxia risk group.The time-dependent receiver operating curve(ROC)was used to analyze the predictive ability of the risk model in various time periods and compared with various clinical indicators.In order to better assist clinical diagnosis and treatment,we synthesized the hypoxia risk score and various clinical indicators to establish a nomogram,and used the consistency index(C index)and the calibration curve for evaluation.Finally,we analyzed the relationship between hypoxia and immunity.The CIBERSORT tool was mainly used to estimate the proportion of 22 immune cell types in the high-risk and low-risk groups,and analyzed the expression of immune negative regulatory genes(including immune checkpoints and immunosuppressive factors)in the high-and low-risk groups.ResultsWe obtained 200 hypoxia-related genes from the GSEA website,and then selected 20 key genes from the 200 hypoxia-related genes,and finally selected 7prognostic-related genes to build models through univariate Cox analysis and Lasso regression analysis.Univariate and multivariate Cox analysis results showed that the prognostic model of hypoxia-related genes we established exhibited a high prognostic value and was able to act as an independent risk factor for BCR [Training set:P<0.001,hazard ratio(HR)=2.529;Validation set: P=0.002,HR=1.908].The risk curve showed that the proportion of BCR in PCa patients in the high hypoxia risk group was significantly higher.The results of survival analysis showed that the recurrence-free survival rate(RFS)of the high hypoxia risk group was worse than that of the low hypoxia risk group(p<0.05).The gene expression heat map showed that except for MT1 E,the expression levels of the other 6 genes in the high-risk group were significantly higher than those in the low-risk group.The box plot showed that the expression of MT1 E in low clinicopathological grades was higher than that of high clinicopathological grades,and the expression of the other 6 genes in high clinicopathological grades was higher than that of low clinicopathological grades.GSEA showed that the genes in the high hypoxia risk group were mainly enriched in Notch signaling pathway,cancer-related pathway and TGF-β signaling pathway.The area under the ROC curve(AUC)for 1,3,and 5 years in the TCGA cohort were 0.795,0.821,and 0.828,respectively,and the AUC values for 1,3,and 5 years in the GEO cohort were 0.771,0.769,and 0.764,respectively.The AUC values of the two cohorts were higher than clinical parameters in the same time period.Nomogram based on the hypoxia risk score and clinical parameters was capable of distinguishing high-risk BCR patients and its concordance index(C index)were 0.79,and the calibration curve had no obvious deviation.In addition,the ratio of immunosuppressive cells[regulatory T cells(Tregs),tumor-associated macrophages(TAMs)and neutrophils],resting T cells and resting natural killer(NK)cells in the high hypoxia risk group was significantly higher,but significantly lower in the proportion of activated NK cells(p<0.05).At the same time,immune negative regulatory genes(including immune checkpoints and immunosuppressive factors)were highly expressed in the high-risk group(p<0.05).Conclusionwe developed and validated a hypoxia risk model,which served as an independent prognostic indicator and could effectively predict BCR after RP in patients with localized PCa.In addition,our research had clarified that tumor hypoxia would form immunosuppression to further worsen the prognosis of PCa,which might provide a new direction for future hypoxic targeted therapy for PCa.