Development of a prognostic risk model for predicting biochemical recurrence-free survival in patients with prostate cancer based on lysine acetylation

Abstract

Background

Lysine acetylation plays a critical role in prostate cancer (PCa) by modulating androgen receptor (AR) signaling. However, the exact mechanisms by which lysine acetylation impacts PCa prognosis remain unclear. The aim of this study was to investigate the mechanism by which lysine acetylation affects PCa prognosis by modulating the AR signaling pathway.

Methods

Data from The Cancer Genome Atlas-Prostate Adenocarcinoma (TCGA-PRAD), GSE54460, and lysine acetylation-related genes (LARGs) were obtained from public databases and literature. Differentially expressed genes (DEGs) were identified in TCGA-PRAD, and key module genes associated with LARGs were selected using weighted gene co-expression network analysis (WGCNA). Candidate genes were identified by overlapping DEGs and key module genes. A biochemical recurrence-free (BCR-free) prognostic model was constructed and validated using BCR-free survival data from patients with PCa. Prognostic genes were further confirmed through machine learning. PCa samples were stratified into high- and low-risk subgroups based on the median risk score. A nomogram model was developed integrating clinicopathological features and risk scores to identify independent prognostic factors. Enrichment analysis, tumor microenvironment profiling, and drug sensitivity assessments were performed for the two risk subgroups.

Results

A total of 2,658 DEGs and 723 key module genes were analyzed, yielding 105 overlapping candidate genes. Five genes—UBXN10, ACOX2, PLCL1, PLS3, and SLIT3—were identified as BCR-free-related prognostic markers in TCGA-PRAD. The prognostic risk model revealed significantly lower BCR-free survival rates in the high-risk subgroup compared to the low-risk subgroup. A nomogram incorporating Gleason score, tumor stage (T stage), and risk score effectively predicted BCR-free survival in patients with PCa. Notably, natural killer (NK) cells, myeloid dendritic cells, endothelial cells, and fibroblasts were significantly correlated with PLS3 (|Cor| >0.3, P<0.05). Drugs such as cisplatin, MK-1775, and ulixertinib were identified as potential therapeutic agents for PCa.

Conclusions

Five BCR-free-related prognostic genes were identified as potential therapeutic targets. Additionally, a BCR-free-related prognostic risk model was developed, offering a robust tool for predicting BCR-free survival in patients with PCa.

Highlight box.

Key findings

• UBXN10, ACOX2, PLCL1, PLS3, and SLIT3 were identified as biochemical recurrence-free (BCR-free)-related prognostic markers in prostate cancer (PCa).

What is known and what is new?

• BCR-free is a key clinical indicator for evaluating treatment efficacy and prognosis.

• Five BCR-free-related prognostic genes were identified as potential therapeutic targets for PCa.

What is the implication, and what should change now?

• A BCR-free-related prognostic risk model was developed, offering a robust tool for predicting BCR-free survival in patients with PCa.

Introduction

Background

Prostate cancer (PCa) is a prevalent malignancy originating in the prostate, presenting symptoms such as urinary frequency, urgency, dysuria, hematuria, hematospermia, prostate enlargement, and bone pain in metastatic cases (1). The World Health Organization reports millions of annual PCa diagnoses, with a higher incidence observed in men over 50 years. Risk factors, including age, family history, race, diet, obesity, inflammation, and elevated androgen levels, are associated with PCa development, though the exact etiology remains unclear (2,3).

Rationale and knowledge gap

The primary diagnostic tools for PCa include prostate-specific antigen (PSA) testing, digital rectal examination (DRE), and prostate biopsy (4). Biochemical recurrence-free (BCR-free) survival is a critical clinical endpoint for evaluating treatment effectiveness and prognosis (5). Recent advancements in targeted therapies, personalized treatments, and adjunct therapies such as immunotherapy have shown promise in improving BCR-free survival in patients with PCa (6). Moreover, artificial intelligence (AI) demonstrates significant potential across all stages of PCa. By analyzing pathological images using deep learning algorithms, it can assist doctors in more accurately identifying cancerous areas, thereby enhancing the accuracy and efficiency of biopsies (7,8). Simultaneously, AI prediction models constructed from multi-omics data enable personalized assessments of patient prognosis indicators, such as BCR-free survival, thereby expanding the potential for prognostic prediction (9). Ongoing research focuses on identifying novel biomarkers and therapeutic targets to further improve BCR-free survival while minimizing treatment-related side effects (10). Despite progress, further clinical validation and research are needed to optimize treatment strategies.

Lysine acetylation, a widespread post-translational modification (PTM), regulates protein function by transferring acetyl groups to lysine residues through both enzymatic and non-enzymatic processes (11). In PCa, lysine acetylation modulates the expression and activity of oncogenes and tumor suppressor genes, influencing critical signaling pathways such as PI3K/Akt and MAPK, which are involved in cell proliferation, metastasis, and apoptosis (12). Moreover, lysine acetylation impacts protein stability and degradation, playing a pivotal role in cell cycle regulation through proteins like cyclins (13). Meanwhile, lysine acetylation can occur in the ligand-binding domain (LBD) or the DNA-binding domain (DBD) of the androgen receptor (AR), affecting its ability to bind androgens, such as testosterone and dihydrotestosterone, or altering the efficiency of AR binding to the promoter regions of target genes (14). In PCa, abnormal activation of the AR signaling pathway leads to the growth and spread of cancer cells (15), while the generation of AR variants, such as AR-V7, renders cancer cells androgen-independent, thereby conferring resistance to traditional endocrine therapy (16). Despite its significance, research on lysine acetylation in PCa remains limited (17), highlighting the need for further investigation. Understanding the mechanisms of lysine acetylation will provide valuable insights into PCa pathophysiology and guide the development of novel therapeutic strategies.

Objective

This study utilized transcriptomic data to identify lysine acetylation-associated genes in PCa. A comprehensive model incorporating machine learning techniques, such as least absolute shrinkage and selection operator (LASSO), univariate Cox analysis, and multivariate Cox regression, was developed to create a prognostic tool for assessing PCa recurrence risk, with a focus on BCR-free survival. Furthermore, identifying key genes involved in lysine acetylation in PCa offers a promising approach to advancing PCa management strategies and fostering personalized treatment options. We present this article in accordance with the TRIPOD reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-179/rc).

Methods

Data resource

Messenger RNA (mRNA) expression profiles, clinicopathological signatures, and BCR-free survival data for The Cancer Genome Atlas-Prostate Adenocarcinoma (TCGA-PRAD) were downloaded from the University of California, Santa Cruz (UCSC) Xena platform (https://xena.ucsc.edu/). TCGA-PRAD contained data from 499 patients with PCa and 52 normal controls (NC), with BCR-free survival information available for 464 patients with PCa. Additionally, the GSE54460 dataset (Illumina HiSeq 2000) was downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/), which included 100 patients with PCa with BCR-free survival data. A total of 33 lysine acetylation-related genes (LARGs) were sourced from published literature and listed in Table S1 (18,19). This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Identification of differentially expressed genes (DEGs)

DEGs between 499 PCa samples and 52 NC samples from TCGA-PRAD were identified using the ‘DESeq2’ R package (v.1.36.0) (20), with thresholds set at |log₂fold change (FC)| >1 and P<0.05. DEGs were visualized using volcano and heat maps generated with the ‘ggplot2’ (v.3.3.6) and ‘heatmap3’ (v.1.1.9) R packages, respectively (21,22).

Weighted gene co-expression network analysis (WGCNA)

The lysine acetylation-related score in TCGA-PRAD was calculated using the single-sample gene set enrichment analysis (ssGSEA) algorithm from the ‘GSVA’ R package (v.1.44.5) (23), with the 33 LARGs as the background gene set. The scores between the PCa and NC groups were compared using the Wilcoxon test (P<0.05). WGCNA was performed to identify the module genes related to lysine acetylation using the ‘WGCNA’ R package (v.1.71) (24). Outlier samples were excluded to ensure clustering validity. The soft threshold (β) was determined using the pickSoftThreshold function, and genes were classified into several modules based on dynamic tree cutting (minModuleSize =100). The module most strongly correlated with lysine acetylation was selected as the key module for further analysis (|Cor| ≥0.3, P<0.05). Gene significance (GS) and module membership (MM) were calculated for genes within the key module by intramodular analysis, and genes meeting the criteria |GS| >0.2 and |MM| >0.7 were selected as key module genes.

Function enrichment analysis and construction of protein-protein interaction (PPI) network

Candidate genes were identified by overlapping DEGs and key module genes using Venn diagrams from the Bioinformatics & Evolutionary Genomics website (http://bioinformatics.psb.ugent.be/webtools/Venn/). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed to investigate the biological functions and pathways of these candidate genes using the ‘clusterProfiler’ R package (v.4.7.1) (adjusted P<0.05) (25). The PPI network for the candidate genes was constructed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://cn.string-db.org/) with a medium confidence threshold (>0.9), and the PPI network was visualized using Cytoscape software (v.3.9.1) (26).

Establishment and verification of the BCR-free-related prognostic risk model

In TCGA-PRAD, to identify BCR-free-related prognostic genes in patients with PCa, LASSO regression analysis was performed. Initially, univariate Cox analysis was performed using the ’survival’ R package (v.3.4-0) (https://cran.r-project.org/web/packages/survival/index.html) to identify characteristic genes from the candidate gene set (P<0.05). After confirming the proportional hazards (PH) assumption, characteristic genes with P<0.05 were retained for further analysis. Subsequently, LASSO regression analysis was performed using the ‘glmnet’ R package (v.4.1.4) (27) to select genes most strongly associated with BCR-free survival. Multivariate Cox analysis was then performed to identify BCR-free-related prognostic genes and construct a BCR-free-related prognostic risk model as follows:

$$\text{Risk\hspace{0.17em}score}={\displaystyle {\sum }_{\text{i=1}}^{\text{n}}\left(\text{expression\hspace{0.17em}level\hspace{0.17em}of\hspace{0.17em}BCR-free-related\hspace{0.17em}prognostic\hspace{0.17em}genes}\times \text{coefficient}\right)}$$ [1]

Based on the median risk score in TCGA-PRAD from the BCR-free-related prognostic model, the 464 PCa samples were divided into high- and low-risk subgroups. Kaplan-Meier (KM) curves were generated using the ‘survminer’ R package (v.0.4.9) (https://CRAN.R-project.org/package=survminer). Receiver operating characteristic (ROC) curves were plotted to evaluate the risk model’s performance using the ‘survivalROC’ R package (v.1.0.3.1) (28). To validate the BCR-free-related prognostic risk model, an external dataset, GSE54460, was employed. The 100 patients with PCa were also divided into high- and low-risk subgroups, and the same analytical methods were applied as described above.

Independent prognostic value of the BCR-free-related prognostic risk model in TCGA-PRAD

The expression levels of BCR-free-related prognostic genes were analyzed between different Gleason scores (pattern 3, pattern 4, pattern 5, pattern 2) by the Wilcoxon test (P<0.05). Univariate Cox analysis was performed to assess the association between risk scores, clinical indicators [age, node stage (N stage), tumor stage (T stage), Gleason score], and BCR-free survival (P<0.05). Factors significantly correlated with BCR-free survival were selected based on the PH assumption (P>0.05). Independent prognostic factors related to BCR-free survival were then identified (P<0.05). A nomogram was constructed using the ‘rms’ (v.6.3-0) (https://CRAN.R-project.org/package=rms) and ‘survival’ R packages, incorporating these independent prognostic factors. This nomogram allows clinicians to calculate precise, individualized BCR-free survival probabilities for patients, thereby aiding clinical decision-making. Furthermore, calibration curves for the 1-, 3-, and 5-year BCR-free survival rates, along with decision curve analysis (DCA), were used to assess the nomogram’s accuracy and clinical utility.

Enrichment analysis between high- and low-risk subgroups

The c5.go.bp.v2023.2.Hs.symbols (GO) and C2: KEGG gene sets (KEGG) were sourced from the ‘msigdbr’ R package (v.7.5.1) as background gene sets. In TCGA-PRAD, DEGs between high- and low-risk subgroups were identified and ranked based on log₂FC using the ‘clusterProfiler’ R package (v.4.7.1) (adjusted P<0.05) (25). These DEGs were then subjected to GSEA using the aforementioned background gene sets. Additionally, gene set variance analysis (GSVA) scores for all GO terms and KEGG pathways were calculated using the ‘GSVA’ R package (v.1.44.5) (23). Significant differences in GSVA scores between the high- and low-risk subgroups were identified (P<0.05).

Tumor microenvironment

To investigate immune cell infiltration differences between the two risk subgroups in TCGA-PRAD, infiltration scores of 10 immune cell types in patients with PCa were calculated using the MCPcounter algorithm from the ‘IOBR’ R package (v.3.6.3) (29). Spearman’s correlations between the expression of prognostic genes and the infiltration scores of these 10 immune cells were assessed (|Cor| >0.3, P<0.05).

Drug sensitivity analysis

Half-maximal inhibitory concentration (IC₅₀) values for 198 drugs were obtained from the Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/) and analyzed using the ‘oncoPredict’ R package (v.0.2) (30). Spearman’s correlation between IC₅₀ values and risk scores in TCGA-PRAD was evaluated (|Cor| ≥0.3, P<0.05). Additionally, IC₅₀ values were compared between the high- and low-risk subgroups using the Wilcoxon test (P<0.05).

Statistical analysis

All statistical analyses were performed using R software (v.4.2.3). A P value of <0.05 (two-tailed) was considered statistically significant.

Results

There were 317 candidate genes related to lysine acetylation identified in patients with PCa

In the TCGA-PRAD dataset, a total of 2,658 DEGs were identified between the PCa and NC groups, with 1,117 up-regulated and 1,541 down-regulated genes (Figure 1A,1B). The lysine acetylation-related score in the PCa group was significantly higher than in the NC group (P<0.001) (Figure 1C). After excluding sample TCGA.EJ.7123.11A, all remaining samples were included in the clustering analysis (Figure S1). A total of 7 modules were generated when β=9 (R²=0.8) (Figure 1D,1E), with the MEblue module showing a negative correlation with LARGs (Cor =−0.3, P<0.05) (Figure 1F). From this module, 723 key module genes were selected for further analysis (|GS| >0.2, |MM| >0.7) (Figure 1G). By overlapping the 2,658 DEGs with these 723 key module genes, 317 candidate LARGs were identified (Figure 1H). GO enrichment analysis revealed that these genes were involved in pathways related to contractile fibers, myofibrils, and muscle contraction (Figure 1I, table available at https://cdn.amegroups.cn/static/public/tau-2025-179-1.xlsx). KEGG pathway analysis linked these genes to pathways such as vascular smooth muscle contraction, focal adhesion, and the cGMP-PKG signaling pathway (Figure 1J, Table S2). Additionally, the PPI network constructed for these 317 candidate genes revealed strong interactions, particularly between ABCC9-KCNJ8, CAV1-CAV2, and CAV1-CAVIN1 (combined score >0.999) (Figure 1K, Table S3).

Figure 1.

DEGs between the PCa group and the NC group. (A) Volcano plot of differential gene analysis. Red dots represent upregulated genes, green dots represent downregulated genes, and grey dots represent genes with no significant difference or small FCs. (B) Differential gene expression heat map. Color indicates normalized gene expression, with red representing high expression and blue representing low expression. (C) Lysine acetylation score box diagram. (D,E) Selection of the soft threshold power value. (F) Module and score correlation heat map. (G) GS-MM scatter plot. (H) Venn diagram of DEGs and WGCNA. (I) The candidate genes GO enrichment result. (J) The candidate genes KEGG enrichment results. (K) The candidate genes PPI network diagram. ****, P<0.0001. BP, biological process; CC, cellular component; DEG, differentially expressed gene; FC, fold change; GO, Gene Ontology; GS, gene significance; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function; MM, module membership; NC, normal controls; PCa, prostate cancer; PPI, protein-protein interaction; WGCNA, weighted gene co-expression network analysis.

Five BCR-free-related prognostic genes were selected by the LASSO-Cox regression analyses

A total of 25 BCR-free-related characteristic genes were identified through univariate Cox analysis and PH assumption (Figure 2A, Table S4). Subsequently, 10 characteristic genes strongly associated with BCR-free survival were selected via LASSO analysis, with lambda.min =0.01 (Figure 2B,2C). Multivariate Cox analysis identified five prognostic genes: UBXN10, ACOX2, PLCL1, PLS3, and SLIT3 (Figure 2D). The prognostic risk model was formulated as: risk score = UBXN10 × (−0.8874) + ACOX2 × (−0.4662) + PLCL1 × (−0.7002) + PLS3 × (0.7728) + SLIT3 × (0.5854). Based on the median risk score, the 464 patients with PCa in TCGA-PRAD were divided into high-risk (n=232) and low-risk (n=232) subgroups (table available at https://cdn.amegroups.cn/static/public/tau-2025-179-2.xlsx), and the 100 patients with PCa in the GSE54460 dataset were divided into high-risk (n=49) and low-risk (n=51) subgroups (table available at https://cdn.amegroups.cn/static/public/tau-2025-179-3.xlsx). Risk curves showed that the BCR-free survival rates were significantly lower in the high-risk subgroups in both the TCGA-PRAD and GSE54460 cohorts (Figure 2E,2F). KM curves further confirmed that the BCR-free survival probabilities were significantly lower in the high-risk subgroups compared to the low-risk subgroups in both cohorts (P<0.05) (Figure 2G,2H). The area under the ROC curves (AUCs) for 1-, 3-, and 5-year BCR-free survival in TCGA-PRAD and GSE54460 exceeded 0.6, indicating that the BCR-free-related prognostic risk model effectively predicted BCR-free survival probabilities in patients with PCa (Figure 2I,2J).

Figure 2.

The identification of prognostic genes. (A) The univariate Cox analysis forest map. (B,C) The LASSO regression results. (D) The multifactor Cox forest map. (E,F) Risk curve. red representing the high-risk group, and a circle representing a sample. The bottom half of the picture shows the status and time of non-recurrence corresponding to the samples. The vertical coordinate is the time of non-recurrence. Blue represents non-recurrence samples, and red represents recurrence samples (BCR). (E) Training set. (F) Validation set. (G,H) The no-recurrence curve of the high- and low-risk group. (G) Training set. (H) Validation set. (I,J) The ROC curve. (I) Training set. (J) Validation set. AUC, area under the ROC curve; BCR, biochemical recurrence; CI, confidence interval; HR, hazard ratio; LASSO, least absolute shrinkage and selection operator; ROC, receiver operating characteristic.

The nomogram effectively predicted the BCR-free ratio of patients with PCa

Prognostic genes showed significant differences between different Gleason score groups. The expression levels of ACOX2 were significantly different between pattern 3 and pattern 4, pattern 3 and pattern 5, as well as pattern 4 and pattern 5 (Figure S2A). The expression levels of PLCL1 and UBXN10 were significantly different between pattern 3 and pattern 4, and between pattern 3 and pattern 5 (Figure S2B,S2C). The expression level of PLS3 was significantly different between pattern 4 and pattern 5 (Figure S2D). The expression level of SLIT3 was significantly different between pattern 3 and pattern 4 (Figure S2E). In TCGA-PRAD, three independent prognostic factors were identified through univariate Cox analysis, the PH assumption, and multivariate Cox analysis: Gleason score, T stage, and risk score (Figure 3A,3B, Table S5). A nomogram was then developed based on these three factors to predict the BCR-free survival rate of patients with PCa (Figure 3C). The nomogram’s predicted BCR-free survival rates at 1-, 3-, and 5-years closely matched the actual outcomes in patients with PRAD (Figure 3D). Additionally, DCA suggested that the nomogram demonstrated stable and reliable prediction performance (Figure 3E).

Figure 3.

Construction of nomogram. (A,B) The independent prognostic analysis of forest maps. (A) Unifactor Cox forest map. (B) Multifactor Cox forest map. (C) The nomogram. (D) Correction curve of the nomogram. (E) DCA curve. BCR, biochemical recurrence; CI, confidence interval; DCA, decision curve analysis; HR, hazard ratio; N, node; T, tumor.

There were significant differences in GO items and KEGG pathways between the high- and low-risk subgroups

A total of 8,727 DEGs were identified between the two risk subgroups in TCGA-PRAD (adjusted P<0.05) (table available at https://cdn.amegroups.cn/static/public/tau-2025-179-4.xlsx). GO enrichment analysis revealed that these DEGs were enriched in processes such as striated muscle cell development, sarcomere organization, and muscle contraction (Figure 4A, table available at https://cdn.amegroups.cn/static/public/tau-2025-179-5.xlsx). Notably, the top 10 GO terms were down-regulated in the high-risk subgroup. KEGG enrichment analysis highlighted that these DEGs were linked to pathways involved in cardiac muscle contraction, hypertrophic cardiomyopathy (HCM), and dilated cardiomyopathy (Figure 4B, Table S6). The GO and KEGG pathways with significant differences in GSVA scores between the high- and low-risk subgroups are listed in tables available at https://cdn.amegroups.cn/static/public/tau-2025-179-6.xlsx and https://cdn.amegroups.cn/static/public/tau-2025-179-7.xlsx (P<0.05). GO results showed differential activity in processes such as the positive regulation of wound healing, epidermal cell spreading, and blood pressure regulation by epinephrine and norepinephrine (Figure 4C, table available at https://cdn.amegroups.cn/static/public/tau-2025-179-6.xlsx). KEGG analysis revealed differences in metabolic pathways like limonene and pinene degradation, beta-alanine metabolism, and propanoate metabolism (Figure 4D, table available at https://cdn.amegroups.cn/static/public/tau-2025-179-7.xlsx).

Figure 4.

The GSEA enrichment analysis between high and low risk groups. (A,B) The GSEA enrichment trend diagram. (C,D) The high-low rating group top 10 GO and KEGG enrichment pathway. BP, biological process; CC, cellular component; GO, Gene Ontology; GSEA, gene set enrichment analysis; GSVA, gene set variance analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function; PRAD, Prostate Adenocarcinoma.

Immune mechanism was explored in patients with PCa

Natural killer (NK) cells, myeloid dendritic cells, endothelial cells, and fibroblasts showed significant correlations with the five BCR-free-related prognostic genes (|Cor| >0.3, P<0.05) (Figure 5A). Among these, PLS3 and fibroblasts exhibited the highest Spearman’s correlation coefficient (Cor =0.77, P<0.001). Notably, PLS3 was more strongly associated with immune cell infiltration. The Spearman’s correlation between PLS3 and the top 4 immune cells with the highest correlation coefficients is shown in Figure 5B-5E.

Figure 5.

The tumor microenvironment analysis in PCa patients. (A) Correlation heat maps of biomarkers and immune cells. (B-E) Scatter plot of correlation between PLS3 and four kinds of immune cells. *, P<0.05; **, P<0.01; ***, P<0.001; ns, not significant (P>0.05). NK, natural killer; PCa, prostate cancer.

Seven drugs showed strong correlations with the risk score

The IC₅₀ values of 43 drugs differed significantly between the high- and low-risk subgroups. Seven drugs were strongly correlated with the risk score (|Cor| ≥0.3, P<0.05) (Figure 6). Cisplatin, MK-1775, ulixertinib, ABT737, and cediranib were negatively correlated with the risk score, while SB505124 and selumetinib showed positive correlations with the risk score.

Figure 6.

The screen of chemotherapy drugs of PCa. (A-G) Association of drug IC₅₀ values with risk scores and differences between groups. IC₅₀, half-maximal inhibitory concentration; PCa, prostate cancer.

Discussion

PCa is a prevalent malignancy in men with multifactorial etiology. Despite extensive research, its precise pathogenesis remains incompletely understood, complicating both diagnosis and treatment. Early detection is hindered by the lack of evident symptoms in the early stages, while treatment, particularly in advanced or recurrent cases, is limited by a shortage of effective therapeutic options. Lysine acetylation, a vital cellular process that influences gene expression, protein activity, and cell signaling, has emerged as a key factor in PCa progression. Thus, further investigation into the relationship between lysine acetylation and PCa, including its underlying mechanisms, is crucial.

This study identified five prognostic genes—UBXN10, ACOX2, PLCL1, PLS3, and SLIT3—associated with lysine acetylation. The MCPcounter algorithm was employed to analyze the tumor microenvironment in PCa. Seven potential drugs closely associated with risk scores were predicted for the treatment of PCa patients: cisplatin, MK-1775, ulixertinib, ABT737, cediranib, SB505124, and selumetinib.

The mechanisms through which genes such as UBXN10, ACOX2, PLCL1, PLS3, and SLIT3 influence PCa are complex. UBXN10 may regulate cell proliferation, apoptosis, and metastasis, with aberrant expression potentially contributing to PCa development (31). UBXN10 belongs to the ubiquitin-associated X domain family and is involved in the ubiquitin-proteasome pathway, potentially regulating protein degradation through interactions with ubiquitin ligases (E3) or deubiquitinating enzymes (DUBs) (32). Lysine acetylation often competes with ubiquitination for the same modification sites, such as lysine residues (33). For instance, acetylation may stabilize proteins by inhibiting ubiquitination markers, such as K48-linked ubiquitin chains, while UBXN10, as a component of the ubiquitin pathway, may indirectly influence the stability of acetylated proteins, including AR and p53 (34). In hepatocellular carcinoma, UBXN10 inhibits tumor progression by promoting the ubiquitination and degradation of β-catenin (35). In PCa, the aberrant stabilization of AR is one of the key oncogenic mechanisms (36). UBXN10 may influence the protein stability of AR by regulating its ubiquitination levels, thereby modulating the AR signaling pathway; however, its specific role in the pathogenesis of PCa requires further investigation for elucidation.

ACOX2 is a key enzyme in peroxisomes involved in the β-oxidation of fatty acids. Lysine acetylation can directly modify metabolic enzymes to regulate their activity. In the liver, acetylation modification of ACOX1 can inhibit its enzymatic activity, while SIRT3 (mitochondrial deacetylase) can activate ACOX1 through deacetylation (37). ACOX2 may influence the occurrence and progression of PCa by modulating lipid metabolism pathways (38).

PLCL1 belongs to the phospholipase C family and specifically hydrolyzes lysophosphatidic acid (LPA) to generate diacylglycerol (DAG). It is involved in cell signal transduction, proliferation, and survival (39). Additionally, lysine acetylation of PLCL1 may inhibit its binding to substrates or promote its interaction with regulatory proteins (40). In the context of PCa, PLCL1 participates in phosphatidylinositol signal transduction and may influence the proliferation and invasion of PCa cells (41). PLCL1 may promote the invasion and metastasis of cancer cells by regulating cytoskeletal remodeling (such as through calcium signaling regulating actin polymerization) or the expression of extracellular matrix-degrading enzymes, and acetylation modifications may enhance this effect (42).

PLS3 is an actin-binding protein that belongs to the plastin family, involved in cytoskeletal remodeling, migration, and invasion, closely related to tumor metastasis (43). The acetylation modification of actin and its binding proteins can regulate the dynamics of the cytoskeleton (44). For example, the lysine acetylation of α-tubulin is associated with microtubule stability, while PLS3, as an actin crosslinker, its acetylation may affect the assembly or disassembly of actin filaments (45). In PCa, PLS3 may influence immune cell function and the tumor microenvironment, with elevated expression potentially promoting the migration and invasion of tumor cells, thereby affecting tumor progression and metastasis (46).

SLIT3 is a secreted glycoprotein of the Slit family that acts as an axon guidance molecule, regulating neuronal migration by binding to ROBO receptors (47). In tumors, SLIT3 can exert anti-cancer effects by inhibiting angiogenesis or regulating cell polarity (48). In PCa, the expression of SLIT3 may be regulated by the AR signaling pathway, and the acetylation modification of AR can affect the transcription of SLIT3 (e.g., AR enhances binding to the SLIT3 promoter through acetylation) (49). SLIT3, a signaling molecule involved in nervous system development, is frequently dysregulated in various tumors and may modulate tumor cell migration and invasion (50). Further investigation is needed to elucidate the mechanisms and clinical significance of SLIT3 in PCa. These findings suggest that the genes mentioned may serve as potential diagnostic biomarkers for PCa.

An independent prognostic model was developed using factors such as risk score, Gleason score, and T stage in PCa (51). These factors are critical in determining disease prognosis. The risk score evaluates a patient’s likelihood of developing PCa based on various clinical and biological features. High-risk patients are more susceptible to tumor invasion and metastasis, resulting in a worse prognosis (52). Close monitoring and appropriate intervention can improve the prognosis for these high-risk patients. The Gleason score assesses the histopathological grade of PCa, reflecting tumor malignancy. A higher Gleason score indicates greater cellular abnormality and often necessitates more aggressive treatment. T stage measures the extent of tumor invasion, including size and depth; more advanced stages are associated with deeper tissue involvement and a poorer prognosis. The Gleason score, as a core indicator of pathological grading in PCa, indicates that a higher score (such as pattern 4/5) represents poorer tumor differentiation and greater aggressiveness (53). The expression differences of the acetylation-related genes ACOX2, PLCL1, UBXN10, PLS3, and SLIT3 between high Gleason grade tumors (pattern 4/5) and low-grade tumors (pattern 3) may reveal their key roles in tumor progression. In high-grade tumors, the metabolic reprogramming effect of ACOX2 may work in conjunction with other acetylation genes to meet the metabolic demands of highly aggressive tumors (54). Clinically, patients with high Gleason scores often exhibit AR staining loss, which may be related to the decreased stability of AR protein mediated by UBXN10, thereby forming a malignant cycle of “AR signal inactivation-tumor malignant progression” (55). T stage measures the extent of tumor invasion, including size and depth, with advanced stages indicated deeper tissue involvement and a poorer prognosis. Patients with high-risk scores, high Gleason scores, and advanced T stages typically have a worse prognosis (56). Thus, this prognostic model assists physicians in evaluating disease severity, predicting patient risk, and formulating personalized treatment plans.

In our study, a strong correlation was observed between NK cells, bone marrow dendritic cells, endothelial cells, and fibroblasts with five BCR-free-related prognostic genes. Notably, the Spearman correlation coefficient between PLS3 and fibroblasts was the most significant. These findings suggest a close association between PLS3 and immune cell infiltration.

NK cells play a vital role in anti-tumor immunity through mechanisms such as direct cytotoxicity, the release of cytotoxic molecules, and modulation of immune responses. NK cells significantly influence the outcomes of immunotherapy in PCa (57). Bone marrow dendritic cells, as specialized antigen-presenting cells, are crucial in activating T cell responses. In PCa, these cells are key to regulating the tumor microenvironment and enhancing anti-tumor immunity, which holds important implications for PCa immunotherapy (58). Endothelial cells, which line blood vessels, are essential for tumor angiogenesis and metastasis in PCa. Angiogenesis is a critical process in tumor progression and dissemination, and dysfunctional endothelial cells can promote tumor vascularization and subsequent metastasis in PCa (59). Fibroblasts, as essential components of connective tissue, regulate the tumor microenvironment, extracellular matrix synthesis, and tissue repair. In PCa, fibroblasts influence tumor cell proliferation, invasion, and metastasis and may contribute to mechanisms of immune evasion and drug resistance (60). Overall, NK cells, bone marrow dendritic cells, endothelial cells, and fibroblasts may play important roles in PCa, but their specific roles need further validation.

Cisplatin, MK-1775, ulixertinib, ABT737, cediranib, SB505124, and selumetinib are known for their ability to inhibit DNA replication in cancer cells, enhance the efficacy of chemotherapy, disrupt tumor blood supply, block the MEK signaling pathway, and inhibit TGF-beta signaling. These actions collectively suppress tumor cell growth and metastasis and promote cancer cell apoptosis (61).

However, there are some limitations and flaws in this study. In terms of mechanism exploration, the biological functions of five LARGs have not been thoroughly elucidated through in vivo and in vitro experiments, resulting in a lack of direct experimental evidence supporting the intrinsic connection between gene function and the occurrence and development of diseases. In terms of prognostic value validation, the inability to incorporate more datasets for sufficient validation of the prognostic model has affected the broad applicability and reliability of the research conclusions. In the future, systematic in vivo and in vitro experiments will be conducted to explore the specific mechanisms of action of the five LARGs in the disease process through cell experiments and animal models, clarifying their upstream and downstream regulatory pathways and filling the gap in mechanism research. In terms of clinical validation and data expansion, efforts will be made to actively collect more datasets containing detailed survival information, incorporating case data with larger sample sizes and more clinical characteristics, to conduct comprehensive and rigorous validation of the prognostic model from multiple dimensions, and to carry out prospective clinical studies to further confirm the prognostic value of this feature, enhancing the clinical translation potential of the research findings and providing a more solid theoretical basis and practical guidance for precise diagnosis and treatment of diseases. Despite these limitations, this study successfully developed a nomogram to predict PCa recurrence. This tool can aid in making informed diagnostic and treatment decisions and provide a foundation for future research into the underlying pathogenic mechanisms.

Conclusions

This study identified five prognostic genes associated with BCR-free survival in patients with PCa, highlighting their potential as therapeutic targets. Additionally, a prognostic risk model was developed to predict BCR-free status in these patients.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.