1. Introduction

Prostate cancer is the leading cause of cancer-related deaths among men in Western countries, predominantly affecting men aged 45–60 years [1, 2]. Diagnosis typically involves methods such as prostate-specific antigen (PSA) testing and prostate biopsies, which can be invasive and often lead to overdiagnosis [3, 4]. The etiology of prostate cancer is complex, involving an interplay of genetic, environmental, and lifestyle factors that contribute to its heterogeneous nature and epidemiological disparities across different populations.

Genetic predisposition plays a significant role in prostate cancer, as evidenced by family inheritance and twin studies that highlight the genetic contribution to this disease. Androgen biosynthesis and metabolism, mediated through the androgen receptor pathway, are crucial in the development of prostate epithelium and cancer cells [5]. Understanding these genetic and molecular interactions is critical for developing targeted therapeutic and preventive strategies [6].

This study focused on the metabolic pathways involved in prostate cancer, with a particular emphasis on glutathione metabolism. Glutathione, the most abundant intracellular antioxidant, plays a pivotal role in cellular defense against oxidative stress. Glutathione is a tripeptide composed of L-glutamate, cysteine, and glycine [7]. It exists in both reduced (GSH) and oxidized (GSSG) states, maintaining a balance crucial for cell survival and function. In cancer cells, this balance is often disrupted, leading to an altered redox state that can promote tumor growth and resistance to chemotherapy. Abnormalities in glutathione levels and its related enzymes are linked to the etiology and progression of prostate cancer [8, 9].

Mendelian randomization (MR) is a powerful analytical tool that uses genetic variants as instrumental variables to infer causal relationships between modifiable exposures and clinical outcomes [10, 11]. This approach leverages the random assortment of genes at conception, which mitigates confounding factors typically encountered in observational studies [12]. MR provides a robust framework to explore how specific genetic and metabolic factors, such as those involved in glutathione metabolism, contribute to disease risk and progression.

By integrating MR with comprehensive metabolic profiling, this research aims to delineate the complex genetic and metabolic landscape of prostate cancer. This approach allows us to pinpoint metabolic alterations at the genomic level that could be leveraged for targeted intervention, ultimately paving the way for personalized medicine strategies in oncology.

2. Method

3. Results

3.1. MR Identifies Protective and Risk Factors for Prostate Cancer

Our analysis via MR revealed that elevated monocyte counts were identified as a potential risk factor, positively correlated with increased prostate cancer risk (b = 0.010, p = 0.046) (Figures 1(a), 1(b), 1(c), 1(d), 1(e)). Conversely, a statistically significant inverse association between cysteinylglycine disulfide levels and the risk of prostate cancer was found (IVW method: b = −0.0024, SE = 0.000635, p = 2.14 × 10⁴), indicating that higher levels of this metabolite may confer protection against the disease. The proportion of variance explained (PVE) was substantial at 46.24%, and the false discovery rate (FDR) was controlled at 4.98%, underlining the robustness of this finding. This finding aligns with the known roles of inflammation and immune surveillance in cancer progression (Figures 1(f), 1(g), 1(h), 1(i), 1(j)). Further mediation analysis indicated that monocytes may elevate prostate cancer risk by decreasing levels of cysteinylglycine disulfide (Figures 1(k), 1(l), 1(m), 1(n), 1(o)). These results collectively underscore the intricate interplay between metabolic dysregulation and immune responses in the etiology of prostate cancer, highlighting potential targets for therapeutic intervention and risk assessment.

3.2. Single-Cell Analysis Reveals Differential Pathway Activation

We conducted a comprehensive single-cell analysis using samples obtained from the GEO database, focusing on monocytes to further explore their role in prostate cancer via immune and metabolic pathways. Initial quality control, normalization, and batch correction processes were applied rigorously, utilizing the Harmony algorithm to ensure consistent data quality across samples (Figures 2(a), 2(b)). The dimensionality reduction techniques, t-SNE and UMAP, effectively segregated the cell populations, allowing for detailed visualization and characterization of cellular subtypes within the prostate cancer and control groups (Figures 2(c), 2(d)). The analysis highlighted a notable increase in the proportion of monocytes within the tumor microenvironment compared to normal tissue, suggesting their enhanced role in tumor pathology (Figure 2(e)). We specifically analyzed glutathione and its metabolite, cysteinylglycine, within these cells. Pathway analysis, informed by gene sets from MsigDB, showed distinct metabolic activity profiles; monocytes in the tumor environment exhibited heightened activity in pathways related to glutathione transport, disulfide bond formation, metabolism, conjugation, and recycling. These findings were visualized using bubble plots, which clearly demonstrated the enhanced metabolic pathway activity in monocytes from prostate cancer samples compared to those from normal controls (Figures 2(f), 2(g), 2(h), 2(i), 2(j)). Furthermore, the overall pathway activity across different cell types was quantitatively assessed, with monocytes showing the highest activity particularly in metabolism and disulfide bond formation pathways (Figure 2(k)). The results suggested that monocytes may contribute to the oxidative stress response within the tumor microenvironment through altered glutathione metabolism, potentially affecting the overall progression and therapeutic response of prostate cancer. This enhanced metabolic activity in monocytes could serve as a novel biomarker for early diagnosis and as a potential therapeutic target in prostate cancer management. Further analysis divided cells into two groups, meta high and meta low, focusing on the distribution of other metabolites and visualizing the differences. It was observed that CCl4 was more prevalent in the meta low group (Figures 2(l), 2(m)), suggesting that CCl4 may act as an inhibitor in the meta high group, potentially impeding the monocyte-mediated metabolism of glutathione.

3.3. GSEA, Single-Cell Communication, and Transcription Factor Analysis

Continuing with the analysis, the GSEA identified key pathways differentially expressed between the “meta high” and “meta low” monocyte groups (Figure 3(a)). The “meta high” group exhibited upregulation in critical metabolic pathways, including those related to copper, zinc, and calcium ion metabolism, as well as elevated activity in glucose and amino sugar metabolism and lipoprotein clearance pathways. Conversely, the “meta low” group showed a downregulation of pathways involved in inositol lipid–mediated signaling and hydrogen peroxide metabolism, indicating a potential suppression of cellular defense mechanisms against oxidative stress. Transcription factor analysis using the Dorothea database revealed significant differences in the activation of transcription factors between the two groups. In the “meta low” group, PRDM14, a transcription factor known for its tumor-suppressing functions, was the most upregulated, suggesting a potential role in inhibiting tumor progression [15]. In contrast, the “meta high” group showed higher expression levels of MYC, a well-known oncogene, possibly reflecting a higher proliferative capacity and less differentiated state of the tumor cells [16] (Figure 3(b)). Single-cell communication analysis provided insights into the cellular interactions within the tumor microenvironment. Enhanced communication pathways were observed between monocytes in the “meta high” group and other cell types such as epithelial cells, vascular endothelial cells, and smooth muscle cells, suggesting a complex interplay that may promote tumor growth and metastasis (Figures 3(c), 3(d), 3(e), 3(f)). Plasma cells exhibit stronger cell communication with the meta high group, indicating that plasma cells may play a promoting role in the cysteinylglycine metabolism pathway of monocytes. Then, monocyte cells were reclustered (Figures 3(g), 3(h)). Pseudotime analysis sheds light on the developmental trajectory of monocytes within the tumor microenvironment. As monocytes matured, there was a marked increase in the activity of the cysteinylglycine metabolism pathway, indicating a progressive adaptation of these cells as the tumor evolves (Figures 3(i), 3(j), 3(k)). Through feature plot analysis, we observed that the activity of the cysteinylglycine metabolism pathway increases as cells mature. This reflects an enhancement in both monocyte maturity and cysteinylglycine metabolism pathway activity as tumors progress, aligning with our earlier analyses.

3.4. Bulk Data Analysis and Immune Infiltration Analysis

In the examination of bulk RNA sequencing data, we applied rigorous preprocessing and batch correction techniques to ensure data integrity across the different datasets (Figures 4(a), 4(b)). Subsequent Lasso regression analysis allowed us to identify significant genes that were used as final variables in organizing the training and validation sets (Figures 4(c), 4(d)). The deployment of machine learning techniques, specifically CatBoost, NGBoost, and XGBoost, further validated our findings. Each method was meticulously tested against the training and validation datasets, with the results displaying high AUC values, underscoring their predictive reliability. CatBoost emerged as the most effective, demonstrating the highest AUC value of 0.94, followed by XGBoost and NGBoost with AUC values of 0.88 and 0.86, respectively (Figures 4(e), 4(f), 4(g), 4(h)). The consistency of gene expression profiles across different datasets confirms the robustness of our analytical approach and the clinical relevance of these immune-related gene signatures.

3.5. Screening of Key Genes and Clinical Prognostic Features

Our models highlighted the importance of specific genes, including ATP6V1G1 and CAPG, pinpointed in the NGBoost model as crucial for their predictive capabilities (Figure 5(a)). In the TCGA database, we further analyzed the diagnostic value of the genes, with TXN having the highest AUC (AUC = 0.928), followed by ATP6V1G1 (AUC = 0.925) (Figure 5(b)). The expression levels of different genes showed statistically significant differences between cancerous and adjacent noncancerous tissues (Figure 5(c)). Considering that ATP6V1G1 consistently demonstrated high diagnostic value across various models, we selected ATP6V1G1 as the key gene for further analysis. The Kaplan–Meier (KM) curve indicated that patients with high ATP6V1G1 expression have a poorer prognosis (Figure 5(d)). Additionally, ATP6V1G1 was significantly overexpressed in advanced T and N stages (Figures 5(e), 5(f)). In terms of progression-free interval (PFI), ATP6V1G1 expression was higher in patients who have died (Figure 5(g)). Figures 5(h), 5(i) demonstrate the differential expression of ATP6V1G1 in prostate cancer tissues and adjacent normal tissues via immunohistochemistry (IHC). The staining conferred a significant overexpression of ATP6V1G1 in prostate cancer tissues compared to normal tissues. Western blot analysis of prostate cancer cell lines (PC-3 and DU-145) and the normal prostate epithelial cell line (RWPE-1) further corroborates these findings, showing elevated ATP6V1G1 protein levels in the cancer cell lines. GAPDH was used as the loading control (Figure 5(j)). In summary, ATP6V1G1 was a key gene with significant diagnostic and prognostic value in cancer.

3.6. Immunological Characteristics of ATP6V1G1

To further analyze the immunological characteristics of ATP6V1G1, we divided the samples into high- and low-expression groups based on the median expression level. Using the ssGSEA algorithm, we found that most immune cells were significantly overexpressed in the low-expression group, and ATP6V1G1 showed a significant negative correlation with most immune cells (Figures 6(a), 6(b)). Additionally, most immune checkpoints were significantly overexpressed in the low-expression group, with ATP6V1G1 showing a significant negative correlation with these checkpoints (Figures 6(c), 6(d)). Tumor immune phenotype tracking, which involves monitoring and analyzing the immune characteristics of tumors over time, is crucial in tumor treatment and disease progression. We observed that the low-expression group exhibited higher expression levels across most steps compared to the high-expression group (Figure 6(e)). Furthermore, in the immune function scores, the low-expression group showed higher expression (Figure 6(f)). In the tumor microenvironment scores, the high-expression group exhibited significantly higher expression in tumor purity but lower expression in StromalScore, ImmuneScore, and ESTIMATEScore (Figure 6(g)). Given the critical role of immune checkpoints in determining immunotherapy response, we used the TCIA database and found that CTLA4 and PD1 immunotherapy were more effective in the low-expression group (Figures 6(h), 6(i), 6(j)). Immunotyping assessment showed that ATP6V1G1 had the highest expression in C4 (lymphocyte depleted) (Figure 6(k)). C4 indicates low levels of lymphocyte infiltration and high cell cycle activity, potentially reflecting an immunosuppressive microenvironment where tumor cells can grow more easily and evade immune surveillance. Therefore, ATP6V1G1 emerges as a promising target for immunotherapy, offering new prospects and pathways for advancing immunotherapeutic strategies.

3.7. Exploring the Role of ATP6V1G1 in Prostate Cancer Cell Lines

This study investigated the expression patterns and functions of ATP6V1G1 in prostate cancer cell lines through a series of experiments. Figures 7(a), 7(b) indicate that ATP6V1G1 was upregulated in prostate cancer tissues and cell lines, suggesting a potential role in promoting the progression of prostate cancer. ATP6V1G1 mRNA expression was silenced using two different siRNAs (siATP6V1G1#1 and siATP6V1G1#2) in DU145 and PC3 cells. The results, as shown in Figures 7(c), 7(d), demonstrated a significant reduction in ATP6V1G1 mRNA levels compared to the control group, confirming the effectiveness of the siRNA approach. The CCK8 and colony formation assay results, depicted in Figures 7(e), 7(f), 7(g), 7(h), revealed that silencing ATP6V1G1 significantly reduced the proliferative capacity of both DU145 and PC3 cells. The wound healing assay results evaluated the migration capacity of DU145 and PC3 cells following ATP6V1G1 knockdown. The migration distance was notably decreased in cells treated with siATP6V1G1#1 and siATP6V1G1#2 compared to the control, indicating that ATP6V1G1 supports the invasive properties of prostate cancer cells (Figures 7(i), 7(j)). These results collectively highlighted the oncogenic potential of ATP6V1G1 in prostate cancer, impacting both cell proliferation and migration.

4. Discussion

Prostate cancer is a widespread health issue affecting millions of men globally [17, 18]. It is the second most common cancer in men after lung cancer, comprising 7% of newly diagnosed cancers in men globally and rising to 15% in developed regions. Each year, over 1.2 million new cases of prostate cancer are diagnosed globally, with more than 350,000 deaths attributed to the disease annually. This makes prostate cancer one of the leading causes of cancer-related death among men [19, 20].

MR studies have been widely used to explore the causes of various diseases, but their combination with tissue cell single-cell analysis is still relatively uncommon [21, 22]. Our study integrated MR results from both single-cell and bulk data levels, ultimately validating the reliability of associated genes through machine learning, providing a more comprehensive understanding of the causes of prostate cancer.

Our findings reveal that elevated levels of cysteinylglycine disulfide, a storage form of cysteinylglycine and a metabolite of glutathione, potentially reduce the risk of prostate cancer, highlighting its protective role. This association underscores the importance of glutathione metabolism in prostate cancer, as fluctuations in the levels of cysteinylglycine disulfide correlate with changes in glutathione metabolism. This is supported by our MR analysis, which robustly associates higher levels of this metabolite with a decreased incidence of the disease. Conversely, the increase in monocyte counts correlates with heightened cancer risk, suggesting that immune system dysregulation contributes to tumor progression. These results emphasize the importance of targeted therapies that could modulate the glutathione metabolism pathway, potentially curbing prostate cancer development.

The integration of single-cell RNA sequencing and bulk transcriptome analysis has provided a comprehensive view of the cellular and molecular landscapes within prostate tumors [23, 24]. Particularly, the differential pathway activation in monocytes underlines the metabolic plasticity of these cells in the tumor microenvironment. Our machine learning models corroborate these findings, identifying gene signatures that predict disease presence and progression with high accuracy. Such predictive biomarkers are invaluable for developing personalized medicine approaches, enhancing both diagnostic precision and treatment efficacy.

Notably, our study highlights the role of ATP6V1G1, which emerged as a key gene with significant diagnostic and prognostic value. In vitro experiments demonstrated that silencing ATP6V1G1 significantly reduced the proliferation and migration of prostate cancer cells, suggesting its oncogenic potential. ATP6V1G1 is known to play a critical role in the acidification of intracellular organelles and the regulation of autophagy, processes that are often dysregulated in cancer. Studies have shown that overexpression of ATP6V1G1 is associated with poor prognosis in various cancers, including breast cancer and glioma [25, 26]. This gene offers new prospects for therapeutic intervention in prostate cancer, underscoring the importance of further research on its role in disease progression.

Despite these advances, several limitations must be acknowledged. The causal relationships derived from MR rely on the assumption that the genetic variants used as instrumental variables affect the disease outcome only through the exposure being studied. Unmeasured confounding or pleiotropy could skew these interpretations. Additionally, while our models demonstrate significant predictive power, the external validity of these findings needs to be tested in broader, more diverse populations to confirm their generalizability. Future research should focus on validating these metabolic and immune signatures in longitudinal studies, potentially integrating more comprehensive genetic, environmental, and lifestyle data. This would not only confirm the clinical relevance of our findings but also might unveil new therapeutic targets within the intricate network of pathways influencing prostate cancer.

5. Conclusion

This study revealed the dual roles of cysteinylglycine disulfide and monocytes in prostate cancer, showing that elevated cysteinylglycine disulfide levels protect against the disease, while increased monocyte counts heighten risk. Additionally, ATP6V1G1 was identified as a key gene with significant diagnostic and prognostic value, suggesting its potential as a therapeutic target. The integration of MR, transcriptome analysis, and machine learning provided a comprehensive understanding of the metabolic and immune pathways involved in prostate cancer. Future research should validate these findings in diverse populations to confirm their clinical relevance and potential for personalized treatment strategies.

Conflicts of Interest

The authors declare no conflicts of interest.

Author Contributions

Rijian Guan: conceptualization, writing – original draft, funding acquisition, and software. Dongdong Wang: investigation, data curation, writing – original draft, and formal analysis. Mingwei Zhan: investigation, data curation, writing – original draft, and resources. Pengfei Liu: writing – review and editing, investigation, validation, and visualization. Yi Yu: writing – review and editing, supervision, project administration, and funding acquisition. Dongdong Wang and Mingwei Zhan contributed equally to this work.

Funding

This work was supported by the Zhejiang Medical Association Clinical Research Fund project (2022ZYC-A119).