Disulfidptosis-related subtype and prognostic signature in prostate cancer

Abstract

Background: Disulfidptosis refers to cell death caused by the accumulation and bonding of disulfide in the cytoskeleton protein of SLC7A11-high level cells under glucose deprivation. However, the role of disulfidptosis-related genes (DRGs) in prostate cancer (PCa) classification and regulation of the tumor microenvironment remains unclear.

Methods: Firstly, we analyzed the expression and mutation landscape of DRGs in PCa. We observed the expression levels of SLC7A11 in PCa cells through in vitro experiments and assessed the inhibitory effect of the glucose transporter inhibitor BAY-876 on SLC7A11-high cells using CCK-8 assay. Subsequently, we performed unsupervised clustering of the PCa population and analyzed the differentially expressed genes (DEGs) between clusters. Using machine learning techniques to select a minimal gene set and developed disulfidoptosis-related risk signatures for PCa. We analyzed the tumor immune microenvironment and the sensitivity to immunotherapy in different risk groups. Finally, we validated the accuracy of the prognostic signatures genes using single-cell sequencing, qPCR, and western blot.

Results: Although SLC7A11 can increase the migration ability of tumor cells, BAY-876 effectively suppressed the viability of prostate cancer cells, particularly those with high SLC7A11 expression. Based on the DRGs, PCa patients were categorized into two clusters (A and B). The risk label, consisting of a minimal gene set derived from DEGs, included four genes. The expression levels of these genes in PCa were initially validated through in vitro experiments, and the accuracy of the risk label was confirmed in an external dataset. Cluster-B exhibited higher expression levels of DRG, representing lower risk, better prognosis, higher immune cell infiltration, and greater sensitivity to immune checkpoint blockade, whereas Cluster A showed the opposite results. These findings suggest that DRGs may serve as targets for PCa classification and treatment. Additionally, we constructed a nomogram that incorporates DRGs and clinical pathological features, providing clinicians with a quantitative method to assess the prognosis of PCa patients.

Conclusion: This study analyzed the potential connection between disulfidptosis and PCa, and established a prognostic model related to disulfidptosis, which holds promise as a valuable tool for the management and treatment of PCa patients.

Fig. 1

Connection between disulfidptosis-related genes and prostate cancer: a Mutation status of DRGs in the TCGA-PRAD cohort. b Chromosomal positions of DRGs. c Copy number variations of DRGs, with green balls indicating that "loss" is more frequent than "gain" in DRGs, and vice versa. d Differential DRGs between normal tissue and tumour tissue in the TCGA-PRAD cohort. e Impact of SLC7A11 on PFS. f, g qPCR and WB results of SLC7A11 in PC-3, DU-145, and RWPE-1 cells. h Results of single-cell sequencing of prostate cancer cells in the GSE141445 dataset and UMAP dimensionality reduction analysis of the distributions of various immune cells and malignant cells. i SLC7A11 expression in different prostate cancer cell subsets. j, k qPCR and WB were used to verify the plasmid transfection efficiency of SLC7A11 in DU145 and PC3 cells. l, n A scratch assay confirmed that SLC7A11 promoted the migration of DU145 and PC3 cells. m, o Transwell assays confirmed the role of SLC7A11 in promoting the invasion of DU145 and PC3 cells (triplicates). p, q IC50 values of BAY-876 in PC-3 and DU145 cells. r Inhibitory ability of BAY-876 on PC-3 cells in the SLC7A11-OE, SLC7A11-EV, and control groups. s Inhibitory ability of BAY-876 on DU145 cells in the SLC7A11-OE, SLC7A11-EV, and control groups

Fig. 2

Consensus clustering and machine learning based on DRGs in Co-PCa: a Co-PCa patients were divided into two clusters via a consensus clustering algorithm. b Principal component analysis (PCA) showing the distribution of the two clusters. c Differences in PFS between the two clusters. d GSVA of biological pathways between the two clusters. e Differences in clinical information and DRG expression according to the heatmap. f Immune infiltration differences between the two clusters. g, h Random forest trees select the minimum set of genes. i Cox univariate analysis was used to screen 64 prognosis-related DEGs. j KEGG analysis of the biological pathways associated with the DEGs

Fig. 3

Gene cluster and prognostic risk label in Co-PCa: a Co-PCa patients were divided into three gene clusters on the basis of 64 DEGs via a consensus clustering algorithm. b Differences in PFS among the three gene clusters. c Heatmap of clinicopathological features and expression of DEGs. d DRG expression levels in the 3 gene clusters. e, f Selection of the LASSO model; the simulation parameters were set to 1000, tenfold cross-validation was selected, and 4 risk genes were screened. g–i Differences in PFS between the low-risk group and the high-risk group in the training group, testing group, and external testing group. j–l ROC curve for predicting the 1-, 3-, and 5-year PFS of patients in the training group, testing group, and external testing group. m Risk score distribution in two DRG clusters. n Risk score distribution in three gene clusters. o Connections among the DRG cluster, gene cluster, and risk group

Fig. 4

Prognosis and immune microenvironment analysis among the clusters: a Nomogram for predicting 1-, 3-, and 5-year PFS in the Co-PCa population on the basis of the risk signature and clinical features. b ROC curves of the prognostic nomogram for 1-, 3-, and 5-year survival. c Calibration chart of the nomogram. d Landscape of DRG mutations in the high- and low-risk groups. e Pearson correlation between the risk score and RNAs. f The TME score in the high- and low-risk groups. g Expression levels of 21 types of immune cells in low-risk and high-risk tumour samples, with the Wilcoxon rank-sum test used for significance. h PD-L1 expression levels in the high- and low-risk groups. i Heatmap of the correlations among 21 types of immune cells. j Heatmap of the differences in the expression of risk signature genes between the high-risk group and the low-risk group. k Heatmap of the correlations between 21 types of immune cells and the risk signature. l Boxplot of TMB differences between the high-risk group and the low-risk group. (M) PFS of high-TMB and low-TMB patients in the TCGA-PRAD cohort. n Comparison of PFS between TCGA-PRAD patients in different risk groups and different TMB levels

Fig. 5

Expression levels of risk labels in prostate cancer cells and tissues. a–f Annotation of all cell types and the percentage of each type in GSE141445 and the expression of PGM5, COL4A1, ANTXR1, and CTSB in each cell type. g Connection between risk signature genes and DRGs, with purple circles representing risk factors, green circles representing favourable factors, pink lines indicating positive correlations between two factors, and blue lines indicating negative correlations. h–k RNA expression levels of the 4 risk signature genes in multiple prostate cancer cell lines. l Protein expression levels of the 4 risk signature genes in multiple prostate cancer cell lines, with GAPDH as the reference protein and the thickness of the band representing the protein level. m–p Immunohistochemistry results of risk signature genes in the HPA