Prostate cancer is the most diagnosed cancer in men worldwide and is the second leading cause of death in men (1,2). Prostate-specific antigen (PSA) is a traditional marker for the diagnosis of prostate cancer but has low specificity (3). This is partly attributed to the difficulties in PSA detection in premalignant lesions, including prostatic intraepithelial neoplasia that has a lower concentration of serum PSA (3). Periostin is also a candidate biomarker for prostate cancer in the early stages and prostate cancer stroma in the advanced stages, owing to its increased expression during such contexts (4). Androgens and androgen receptor signaling have been shown to promote prostate cancer progression. Subsequently, androgen deprivation therapy has become the main therapy for patients with prostate cancer at different stages (5). However, a considerable proportion of patients receiving these treatments ultimately progresses to more aggressive disease, leading to the development of castration-resistant prostate cancer (5). Patients with prostate cancer frequently exhibit resistance to androgen deprivation therapy, a condition known as castration-resistant prostate cancer (5). Therefore, the identification of biomarkers for the early detection of prostate cancer is essential. However, little is known about which candidate cell signaling pathway markers can be identified and which pathways may serve as therapeutic targets.

Here, we focused on the disulfidptosis pathway among numerous signaling pathways. This is because disulfidptosis is a mechanism that causes cell death and is associated with iron-dependent cell death or ferroptosis, the tricarboxylic acid cycle, energy metabolism, and glucose uptake; in other words, it is called the Warburg effect (6,7). Disulfidptosis is a novel type of cell death mediated by abnormal accumulation of intracellular disulfides and induced by glucose transporter inhibitors (8). It is triggered by the accumulation of reactive oxygen species (ROS) and relentless lipid peroxidation induced by disulfide-dependent mechanisms in tumor cells (9), and therefore, disulfidptosis suppresses tumor cell growth (10), suggesting a potential clinical application for use as advanced treatment strategy. Recent investigations have shed light on a distinctive form of programmed cell death known as disulfidptosis with high expression of SLC7A11 (SLC7A11^high) (8). During glucose starvation, overabundant intracellular disulfides accumulate in SLC7A11^high cells, leading to an uncharacterized form that is distinct from apoptosis and ferroptosis (8). Simultaneously, F-actin collapses during glucose starvation, and aberrant disulfide bonds are induced in an SLC7A11-dependent manner (8). Additionally, glucose transporter inhibitors suppress SLC7A11^high tumor growth by downregulating disulfidptosis (8). Besides, SLC7A11, SLC3A2, RPN1, and NCKAP1 have pivotal roles required for the progress of disulfidptosis in gastric cancer (11). A recent study showed that a disulfidptosis-related long non-coding RNA signature is a prognostic indicator for glioma immunotherapy (12) and that disulfidptosis-related genes could be potential prognostic biomarkers associated with tumor microenvironment and immunotherapy response (13). Long noncoding RNAs also contribute to establishing a prognostic risk prediction model in prostate cancer (14). However, promising biomarkers associated with disulfidptosis remain elusive in prostate cancer.

In this study, we analyzed transcriptomic and clinical data of prostate cancer patients to identify disulfidptosis-related signatures associated with prognosis. The genes were correlated with TNM stage, Gleason grade, and BCR-free survival. Here we would propose prognosis prediction models, and a hypothetical model in which disulfide stress may play roles in cell proliferation in prostate cancer. Findings from this study would suggest crucial roles in cell death or cell growth mechanisms induced by disulfide stress, offering insights into targeted therapies and personalized medicine.

Data collection. A dataset of gene expression and clinical information from patients with prostate cancer was used (15). Gene expression values of reads per kilobase of exon per million mapped reads (RPKM) were subjected to subsequent analyses. Analyses were performed on 324 patients in the Prostate Cancer, German Cancer Research Center (Deutsches Krebsforschungszentrum, DKFZ) dataset (https://www.dkfz.de/en/frueherkennung-prostatakarzinom/index.php) (16), and representative results were validated using additional dataset of the Prostate Adenocarcinoma, The Cancer Genome Atlas (TCGA), PanCancer Atlas (494 patients) (https://datacatalog.mskcc.org/dataset/10426) (16). Gene expression values of fragments per kilobase of exon per million mapped reads (FPKM) were subjected to subsequent analyses. Genes of interest (GOI) were annotated online using GOstat2.5 (http://gostat.wehi.edu.au/) (17) and Database for Annotation, Visualization, and Integrated Discovery (DAVID) 6.8 (https://david.ncifcrf.gov/) (18). Gene expression values ​​were used directly for calculations and then graphed in appropriate applications (19,20). However, the genes with RPKM=0 in all samples were excluded from all analyses. The workflow of this study is shown in Figure 1.

Survival analysis. Correlations between gene expression and survival times were evaluated by Cox hazard regression analysis using R (16,19,21). Kaplan-Meier analysis was performed to estimate the survival distribution of the subgroups using R. Subgroups were divided by the median expression of GOI or median risk scores. The prognostic model genes were confirmed, and the risk scores were imputed as follows:

Risk score = ∑ i = 1 n k i,coef x m i,RPKM

where k_i,coef indicates the Cox regression coefficient, i indicates prognostic gene candidates, and m_i,RPKM indicates the gene expression level as RPKM (16). HR and CI were calculated using a Cox regression model according to patient survival times, which were assessed to compare the subgroups (22). BCR-free survival was defined as the time from the date of surgery for prostate cancer to the date of recurrence or the last follow-up.

Graphical lasso network analysis. Genetic interactions with hub networks among variables from gene expression were analyzed by graphical lasso estimation of Gaussian graphical models, such as a sparse inverse covariance matrix using a lasso (L1) penalty and the glasso package in R (16,23).

Survey for therapeutic targets and drug responses. Drug responses in pan-cancer were surveyed with Gene Set Cancer Analysis (GSCA) online at https://guolab.wchscu.cn/GSCA/#/drug, based on the Genomics of Drug Sensitivity in Cancer (GDSC) dataset 1 and 2 (https://www.cancerrxgene.org/) and the Cancer Therapeutics Response Portal (CTRP) v2 dataset (https://portals.broadinstitute.org/ctrp/). Positive and negative correlations with false discovery rate (FDR) <0.05 indicated drug resistances and sensitivities in the subgroups harboring higher expression of GOI, respectively.

Statistics. Statistical analyses were performed using R4.3.3. p-Values <0.05 were considered statistically significant.

Overview of analysis for disulfidptosis in prostate cancer. In this study, we developed a transcriptome-based prognosis prediction model for disulfidptosis, a type of cell death triggered by disulfide stress resulting in a disulfide bond-induced collapse of the cytoskeleton in prostate cancer, using the transcriptome data and clinical information of 324 patients (8,16). In particular, the detailed expression profile of 106 disulfidptosis-related genes (24) (Figure 2) was investigated for the classification using TNM stage, Gleason score, BCR-free survival, and PSA levels. The 106 genes examined were mainly involved in actin filament binding (GO: 0051015, p=4.69×10^-28), muscle filament sliding (GO: 0030049, p=1.53×10^-15), cytoskeletal motor activity (GO: 0003774, p=3.69×10^-17), and mitochondrial respiratory chain complex I assembly (GO: 0032981, p=2.22×10^-11) and were associated with other functions and pathways including tight junction (KEGG: hsa04530, p=6.70×10^-6), oxidative phosphorylation (KEGG: hsa00190, p=9.85×10^-6), ferroptosis (KEGG: hsa04216, p=1.65×10^-5), and chemical carcinogenesis - reactive oxygen species (KEGG:hsa05208, p=6.11×10^-5).

Differential expression of the disulfidptosis-related genes during progression and metastasis. First, the 106 disulfidptosis-related genes were investigated during the progression and metastatic stages and compared with those of relatively early stage pT2a or Gleason grade 3+3 cases. Of these, PRC1 expression was increased in pT3b (2.27-fold, p=0.008) cases and cases with Gleason grade 4+3 (2.11-fold, p<0.001) and ≥8 (2.85-fold, p<0.001) (Figure 3A). FANCI expression was also increased in cases with Gleason grade 4+3 (1.52-fold, p=0.001) and ≥8 (1.76-fold, p=0.027) (Figure 3A). However, the five genes were decreased in progression and metastatic stages as follows: NSUN2 in pT2c (0.89-fold, p=0.035), SPG20 in Gleason grade 4+3 (0.68-fold, p=0.040), GLUD1 in pT3b (0.46-fold, p=0.007), pT4 (0.42-fold, p=0.001), and Gleason grade ≥8 (0.47-fold, p=0.001), MYH11 in Gleason grade 4+3 (0.28-fold, p=0.030), and DSTN in pT3b (0.71-fold, p=0.046) and Gleason grade 4+3 (0.81-fold, p=0.007) cases (Figure 3B). These seven genes were considered progression marker candidates associated with disulfidptosis in prostate cancer progression. In addition, these genes were associated with the cytoplasm (GO: 0005737, p=0.002), cell cycle (UniProt: 0131, p=0.004), spindle (GO: 0005819, p=0.036), and actin filament binding (GO: 0051015, p=0.055) and were enriched in cytoskeletal regulators.

Similarly, metastatic stage-specific differential expression of these genes was examined. Genes with increased expression levels were as follows: ANP32E in pT4 (3.11-fold, p=0.040), PCBP3 in pT4 (4.34-fold, p=0.001), UBASH3B in pT4 (2.23-fold, p=0.031) and Gleason grade ≥8 (1.65-fold, p=0.031), ACTB in pT4 (1.57-fold, p=0.005), and DBN1 in Gleason grade ≥8 (1.62-fold, p=0.041) cases (Figure 4A). Conversely, genes with decreased expression were NDUFS2 in Gleason grade ≥8 (0.92-fold, p=0.036), GYS1 in pT4 (0.55-fold, p=0.030), TNSK1BP1 in Gleason grade ≥8 (0.62-fold, p=0.043), C12orf51 in pT4 (0.38-fold, p=0.001), FLNC in pT4 (0.45-fold, p=0.040) and Gleason grade ≥8 (0.27-fold, p=0.032), RPN1 in pT4 (0.73-fold, p=0.004) and Gleason grade ≥8 (0.80-fold, p=0.014), TARDBP in pT4 (0.91-fold, p=0.005), MYH13 in pT4 (0.012-fold, p=0.008), and CAPZB in Gleason grade ≥8 (0.82-fold, p<0.001) cases (Figure 4B). These genes could be considered metastatic marker candidates associated with disulfidptosis in prostate cancer. These genes were also involved in cortical cytoskeleton (GO: 0030863, p<0.001), actin-binding (UniProt: 0009, p=0.002), cytoskeleton (GO: 0005856, p=0.002), methylation (UniProt: 0488, p=0.004), actin filament binding (GO: 0051015, p=0.005), and regulation of apoptotic process (GO: 0042981, p=0.006). This observation was interesting in apoptosis and methylation, compared to advanced stages.

Candidates of prognostic signatures with a single disulfidptosis-related gene. Next, the univariate Cox hazard regression analysis was used to clarify several candidates of prognostic signatures for BCR-free survival. These candidates included ANP32E [hazard ratio (HR)=0.28, 95% confidence interval (CI)=0.09-0.83, p=0.022], ATG5 (HR=0.20, 95% CI=0.06-0.69, p=0.011), DCTN (HR=0.27, 95% CI=0.09-0.83, p=0.022), IPO7 (HR=0.32, 95% CI=0.10-0.98, p=0.047), and RPA1 (HR=0.28, 95% CI=0.09-0.87, p=0.028) as better prognosis factors (Figure 5A-C, E, and G), and FANCI (HR=3.03, 95% CI=1.07-8.62, p=0.037) and PRC1 (HR=4.08, 95%CI=1.33-12.51, p=0.014) as poor prognosis factors (Figure 5D and F). Besides, the receiver operating characteristic (ROC) curve analysis showed the following areas under the curves (AUCs): ANP32E (AUC=0.67, 95% CI=0.49-0.85), ATG5 (AUC=0.65, 95% CI=0.50-0.81), DSTN (AUC=0.65, 95% CI=0.48-0.82), IPO7 (AUC=0.58, 95% CI=0.41-0.74), and RPA1 (AUC=0.70, 95% CI=0.53-0.83) (Figure 5A-C, E, and G). For the poor prognosis factors, the results were as follows: FANCI (AUC=0.67, 95% CI=0.50-0.83) and PRC1 (AUC=0.74, 95% CI=0.57-0.91) (Figure 5D and F). These gene expression patterns could be used as appropriate prognostic indicators of disulfidptosis in prostate cancer.

Principal component analysis for the expression of disulfidptosis-related genes. The disulfidptosis-related genes were also classified into several subgroups using principal component analysis (PCA). The PCA biplots indicated that the top three dimensions (Dim) included Dim1, Dim2, and Dim3 (Figure 6A-C). Furthermore, by adding Dim4 and Dim5, the matrix of the five principal components was summarized (Figure 6D). The following dimension classification of the genes were obtained: Dim1 harbored ACTN4 (relative percent contribution to the dimension: 11.23%), ATG5 (10.29%), IPO7 (9.99%), SMPD4 (8.18%), DSTN (8.10%), GLUD1 (7.07%), DBN1 (6.73%), and ANP32E (4.84%), associated with acetylation (UniPlot: 0007, p=0.0012), actin cytoskeleton (GO: 0015629, p=0.003), actin-binding (UniPlot: 0009, p=0.006; GO:0003779, p=0.006), and cytoplasm (GO: 0005737, p=0.019); Dim2 harbored PRC1 (18.21%), HNRNPH1 (13.75%), DHX9 (12.24%), and UBASH3B (10.94%), associated with identical protein binding (GO: 0042802, p=0.023) and nucleus (GO: 0005634, p=0.025; UniPlot:0539, p=0.034); and Dim3 harbored ZHX2 (16.64%), SCO2 (16.22%), CHD4 (14.60%), and FANCI (9.87%) associated with chromatin (GO: 0000785, p=0.008), DNA binding (GO: 0003677, p=0.014), isopeptide bond (UniPlot: 1017, p=0.047), and metal ion binding (GO: 0046872, p=0.057). In addition, Dim4 included SAMM50 (27.22%), RPA1 (11.43%), RUFY1 (9.31%), and SPG20 (7.80%), whereas Dim5 included MYH13 (34.31%), SQSTM1 (12.20%), and PCBP3 (13.77%). The Dim5-relevant genes were associated with extracellular exosomes (GO: 0070062, p=0.011); however, Dim4-relevant genes were not detected in statistically significant ontology terms. Additionally, the analysis of the percentage contribution of the dimensions showed five representative dimensions with >5% contribution (Figure 6E). The percentage contributions of Dim1, Dim2, and Dim3 were highly consistent with the biplot results (Figure 6F-H). These results suggested that the examined disulfidptosis-related genes were successfully distinguished into five main subgroups using PCA.

Approximate combined classification using expression correlation and graphical lasso network. Twenty-three representative genes were extracted from the results of differential expression patterns, Kaplan-Meier survival curves, ROC curves, and percent contributions to principal components. The matrix of correlation patterns of the genes was divided into two subgroups (Figure 7A). This indicates that the 23 representative genes could be divided into two subgroups. The first group included DHX9, HNRNPH1, ANP32E, SPG20, RPA1, IPO7, GLUD1, ATG5, and DSTN. The second group included FANCI, PRC1, UBASH3B, MYH13, PCBP3, CHD4, ZHX2, SCO2, ACTN4, SMPD4, DBN1, SQSTM1, RUFY, and SAMM50. In contrast, the graphical lasso model constituted a dense hub network including ATG5, DHX9, ACTN4, DBN1, DSTN, GLUD1, IPO7, and ZHX2 (edge number >4, edge weight >0.15) (Figure 7B). Interestingly, the positive correlation cluster that included DHX9, IPO7, GLUD1, ATG5, and DSTN was negatively associated with another positive correlation cluster that included ZHX2, ACTN4, and DBN1. These results suggested that there are approximately two cellular or molecular functions that reciprocally support disulfidptosis in prostate cancer.

A construction of the lasso-cox prognostic prediction model with disulfidptosis-related genes. These 23 genes were used to reduce the sparse groups in the lasso analysis (Figure 8A and B). Eleven genes with estimated HRs (Figure 7C) were used for the multivariate Cox regression analysis (Figure 8D). Finally, by using coefficient (Coef) values with p<0.05 in the cox analysis, prognosis prediction formula was obtained as: Risk score=-2.76 ZHX2 + 1.50 SMPD4 + 1.15 CHD4. The subgroup with scores higher than the median risk score exhibited shorter BCR-free survival times than the subgroup with lower scores (HR=5.43, 95% CI=1.56-18.93, p=0.008) (Figure 8E).

These results suggested that the three representative genes could be effective in predicting BCR-free survival, that is, progression of prostate cancer. The similar results were nearly replicated in disease-free survival (DFS) (HR=2.51, 95% Cl=0.89-7.08), p=0.082) of the patients with tumor stages T3 and T4 in the Prostate Adenocarcinoma, TCGA, PanCancer Atlas. On the other hand, statistically significant hazard ratios were validated in overall survival (OS) of all patients (HR=1.74, 95% CI=1.21-2.51, p=0.003) and the patients with tumor stages T3 and T4 (HR=1.43, 95% CI=1.01-2.03, p=0.043), disease-specific survival (DSS) of all patients (HR=1.82, 95% CI=1.18-2.79, p=0.007), and the patients with tumor stages T3 and T4 (HR=1.89, 95% CI=1.23-2.91, p=0.004), progression-free survival (PFS) of all patients (HR=1.63, 95% CI=1.13-2.34, p=0.009) and the patients with tumor stages T3 and T4 (HR=1.63, 95% CI=1.14-2.34, p=0.008) in the TCGA, PanCancer Atlas dataset. Thus, it is possible to predict several survival times with some subgroups in prostate cancer using a mixed expression model including ZHX2, CHD4, and SMPD4. However, statistically significant HRs were estimated even if SMPD4 was excluded from formulas with the three gene candidates, so it may be possible for a better combination of genes to exist. Since these results show statistical prognostic prediction models on the gene expression levels, the importance as molecular markers and biological significance of genes should be clarified in vitro and vivo experiments in future. Alternatively, it would be explored in prospective cohort studies.

A previous study emphasized several forms of programmed cell death, such as apoptosis, ferroptosis, pyroptosis, and cytoproptosis, which exhibit unique morphological, biochemical, and functional traits (25). Programmed cell death can be divided into suicide and sabotage programs (25). However, the boundaries between the two are vague, and these mechanisms may intersect or overlap under certain circumstances (26). The most recently discovered type of programmed cell death is disulfide stress-induced cell death, which is marked by an abnormal buildup of disulfide bonds in intracellular molecules and proteins (8). Apoptosis is typically repressed in tumor cells, which is one of the causes of their infinite cell proliferation; however, in some cases, it causes various types of cell death, including disulfidptosis. In this study, we showed the first construction of a prognosis prediction model based on disulfidptosis in prostate cancer.

Here, we present a disulfidptosis-related prognostic signature for prostate cancer. From the expression profiles of cancer progression and metastasis stages, PRC1 and FANCI levels increased with cancer grade and were associated with shorter BCR-free survival times. Increased PRC1 protein levels have been observed during the S and G₂/M phases of the cell cycle, followed by a dramatic decrease at the mitotic exit and entrance into the G1 phase (27,28). The FANCI protein participates in the meiotic recombination of germ cells (29), and deletion of the Fanci gene causes a strong meiotic phenotype and severe hypogonadism (30). Therefore, these genes are likely reliable prognostic factors. MYH11 and MYH13 are hexameric proteins that consist of two heavy-chain subunits and two pairs of non-identical light-chain subunits (31-33). MYH family members are major contractile proteins that convert chemical energy into mechanical energy via hydrolysis of ATP (26,34). Cancer metastasis requires cytoskeletal flexibility, and relevant gene alterations are likely to influence patient survival. Considering their molecular functions and possible contributions to lifespan, decreased MYH11 and MYH13 expression levels in the prostate cancer transcriptome during advanced and metastatic stages may reciprocally contribute to cell migration and/or invasion through cell plasticity and energy metabolism for morphological change and cell movement, respectively. ANP32E is an ANP32 family member that shares N-terminal leucine-rich repeats and a C-terminal variable anionic region and forms a complex with SET domain proteins that stabilize short-lived mRNAs containing AU-rich elements (35). ANP32E also has an acetyltransferase inhibitory activity, which plays a role in chromatin remodeling and transcription (35). In the present study, ANP32E mRNA was increased in the metastatic stage, and the subgroup with higher expression of ANP32E showed better prognosis, which is consistent with its function as a transcriptional repressor. However, why ANP32E is increased in the metastatic stage, which seems contradictory, is hard to specify and will need further investigations in vitro and vivo. ZHX2 or CHD4 repress transcriptional activity by interacting with the A subunit of nuclear factor-Y (NF-YA) (36) or constituting the nucleosome remodeling and deacetylase (NuRD) complex for epigenetic alteration (37-39), respectively. Sphingomyelin phosphodiesterase 4 (SMPD4) induces epithelial-mesenchymal transition, which is also a late-stage marker of clear cell renal cell carcinoma (ccRCC) (40). Thus, a selection of extremely unique genes may construct a hypothetical model. Briefly, disulfide stress induces epithelial-mesenchymal transition (EMT) with SMPD4, and represses tumor suppressor genes [e.g., BRCA2, CDKN1A (p21^CIP1/WAF1), CDKN1B (p27^KIP1), FOXO1, LATS2, NKX3.1, PTEN, RB1, TP53, and ZBTB16 (PLZF) in prostate cancer] with the transcriptional repression complex ZHX2-NF-YA targeting the CCAAT DNA motif, and the chromatin remodeling complex CHD4-NuRD targeting the methylated CpG island, thereby guaranteeing tumor cell proliferation (Figure 9A and B). Therefore, the three genes are related to cancer proliferation with EMT and chromatin remodeling, which may be associated with disulfidptosis. Our findings provide insights into the biological significance of disulfidptosis in cancer resistance.

Furthermore, expression analyses for drug responses in pan-cancer return the results as sensitive to prostatic adenocarcinoma cells in vitro as follows: Histone deacetylase (HDAC) inhibitors including Vorinostat (FDR=1.04×10^-20), AR-42 (FDR=1.16×10^-20, CAY10603 (FDR=1.02×10^-18), and Belinostat (FDR=4.27×10^-15) in CHD4^high; Vorinostat (FDR=7.74×10^-4) in SMPD4^high; and CUDC-101 (FDR=8.40×10^-8) in ZHX2^high. Similarly, pan-cancer cell lines also indicate sensitivities for protein kinase inhibitors including NPK76-II-72-1 (FDR=1.81×10^-18) and BX-912 (FDR=5.79×10^-13) in CHD4^high; NPK76-II-72-1 (FDR=2.51×10^-4), TG101348 (FDR=1.03×10^-3), TPCA-1 (FDR=1.76×10^-3), BX-912 (FDR=1.88×10^-3), GSK1070916 (FDR=2.50×10^-3), and KIN001-260 (FDR=4.76×10^-3) in SMPD4^high; ZSTK474 (FDR=2.63×10^-9), GSK690693 (FDR=3.82×10^-9), AZD6482 (FDR=7.87×10^-9), EKB-569 (FDR=1.44×10^-8), THZ-2-49 (FDR=1.49×10^-7), GSK2126458 (FDR=2.49×10^-7), PHA-793887 (FDR=5.52×10^-7),and PIK-93 (FDR=1.02×10^-6) in ZHX2^high. However, various mitogen-activated protein kinase (MAPK) kinase (MEK) inhibitors show resistance in CHD4^high, SMPD4^high, and ZHX2^high. These suggest a possibility that HDAC inhibitors may be effective to prostate cancer, but MEK inhibitors may not so, which is a hint for selective drugs in prostate cancer treatment and compatibilities/affinities with disulfidptosis-related signaling and chromatin remodeling but not MAPK-dependent cancer cell proliferation.

Recent studies have highlighted key molecular factors that may improve the diagnosis, prognosis, and treatment of prostate cancer. Huang et al. identify a significant association between the DNMT3A rs77993651 variant and survival outcomes in patients undergoing androgen deprivation therapy (ADT), suggesting DNMT3A as a potential prognostic biomarker and therapeutic target (41). Nakamura et al. explore early diagnostic biomarkers and find that BMP7 expression is significantly reduced in prostate cancer tissues, indicating the potential utility in early detection (42). Meanwhile, RHAMM expression enhances prostate cancer cell migration and is associated with poor prognosis, proposing RHAMM as a novel prognostic marker in metastatic hormone-sensitive prostate cancer (43). In this way, recent findings show promising indications for improving clinical management through molecular profiling in prostate cancer. This study would add prognostic factors linked to disulfidptosis in prostate cancer.

Based on the promising results from previous and current studies, disulfidptosis may have great prospects in the treatment of tumors, including prostate adenocarcinomas. However, due to limited studies, the underlying mechanism of disulfidptosis and the relevant phenomena remain largely unknown. Additionally, the clinical significance and value of disulfidptosis-related genes remain uncertain. Therefore, it is necessary to assess the roles of disulfidptosis-linked genes in tumor cells. Because disulfidptosis represents a novel form of programmed cell death, the detailed mechanisms by which disulfide stress triggers cell death in tumors remain unclear. Therefore, future molecular and animal studies are required. The novel hypothetical model for prostate cancer cell proliferation proposed in this study should also be verified through in vitro and animal experiments. The clinical relevance of disulfidptosis in human prostate cancer also requires further investigation.

The datasets used in the study are available from the Prostate Cancer at DKFZ and the Prostate Adenocarcinoma at TCGA.

The Authors declare no conflicts of interest.

Y.T. designed the experiments. Y.T., K.Y. and M.T. performed the experiments. Y.T., K.Y. and M.T. analyzed data. Y.T., K.Y., M.T. and K.T. wrote the manuscript.

This study was supported in part by MEXT/JSPS KAKENHI (23K08528) to Y.T. The funder had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

No artificial intelligence (AI) tools, including large language models or machine learning software, were used in the preparation, analysis, or presentation of this manuscript.