A total of 162 clinical samples of PCa tissues and adjacent normal tissues were acquired from the patients diagnosed with PCa and accepted operation in the First Affiliated Hospital of Chongqing Medical University from January 2019 to September 2021. The acquired samples of PCa patients were frozen in liquid nitrogen immediately and then stored at −80 °C until miRNA extraction.

Transcriptome RNA sequencing and miRNA data of PCa were downloaded from the TCGA data portal (https://portal.gdc.cancer.gov/), which contained data from 51 nontumor samples and 489 PCa samples. Clinical data of these PCa patients were downloaded and extracted. MiRNA and mRNA data were combined into matrix files by merging scripts of the Perl language (http://www.perl.org/). The limma package of R software (https://bioconductor.org/packages/release/bioc/html/limma.html) was used to screen for differentially expressed miRNAs and mRNAs in PCa and adjacent normal tissues. We performed differential analysis on all mRNA and miRNA data with the screening value of “FDR < 0.05, log| FC |  > 1, and p < 0.05”.

DEMIRs related to clinical prognosis values in PCa patients were selected as survival-related DMIRs (sDMIRs). Univariate Cox analysis was used to screen sDMIRs (p < 0.05). We determined hazard ratio (HR) to separate sDMIRs into protective and deleterious parts. sDMIRs were further analyzed by the multivariate Cox analysis (p < 0.05). To detect the clinical value of sDMIRs, we divided PCa patients into a high-risk group and a low-risk group based on the median risk score. We established RSM using the expression data multiplied by Cox regression coefficients. The formula is as follows: [Expression level of hsa-miR-10a-5p * (0.877309)] + [Expression level of hsa-miR-20a-5p * (0.792393)] + [Expression level of hsa-miR-508-3p * (−0.596601)].

Total RNAs were extracted from the PCa tumor and adjacent normal tissues by TRIzol (Takara, Beijing, China). RNAs were reverse-transcribed to cDNA by the PrimeScript RT reagent kit (Takara, Beijing, China). The reaction steps are as follows: 37°C for 15 min and 85°C for 5 s. The quantitative polymerase chain reaction (qPCR) was carried out on an ABI 7500 Real-Time PCR System (Applied Biosystems, Shanghai, China) by SYBR Green assay (Takara, Beijing, China). The reaction cycling conditions were 95°C for 30 s, 45 cycles of 95°C for 5 s, and 60°C for 30 s. The primer sequences are shown in Table 1. The relative quantification values of miRNAs were standardized to U6 using the −ΔΔ 2 Ct method.

Informed consent forms were signed by all patients before the study. The research protocol has been approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University and is based on the ethical principles of medical research involving human subjects in the Helsinki Declaration.

All statistical analyses were conducted by SPSS22.0 software (SPSS Inc., Chicago, IL) and GraphPad Prism8 (GraphPad Software Inc., La Jolla, CA). The clinical correlations were determined by ANOVA, a post hoc test (Bonferroni's method), and an independent t-test. The correlations between miRNA expression levels and clinicopathological characteristics of patients with PCa were evaluated using the chi-square test and Fisher's exact probability method. p < 0.05 was considered statistically significant.

A total of 132 differentially expressed PCa miRNAs were selected by the limma package of R software, of which 56 were downregulated and 76 were upregulated (Figure 1A). The 50 most upregulated and downregulated miRNAs were respectively demonstrated by the heatmap (Figure 1C). We further confirmed 4161 differentially expressed mRNAs, including 2032 downregulated and 2129 upregulated mRNAs (Figure 1B). The most 50 upregulated and downregulated mRNAs were shown in the heatmap (Figure 1D).

We next explored whether the DMIRs were correlated with the prognosis of PCa patients. We detected four sDMIRs by univariate Cox regression analysis. As shown in Figure 2, miR-508-3p demonstrated a positive correlation with poor prognosis, while miR-20a-5p, miR-10a-5p, and miR-4784 illustrated reversed results. To establish the risk score model, three sDMIRs (miR-10a-5p, miR-20a-5p, and miR-508-3p) were screened by the multivariate Cox regression analysis (p < 0.05). Moreover, the survival curves of these sDMIRs demonstrated that the higher expression levels of sDMIRs were correlated with poor prognosis except for miR-508a-3p. (Supplementary Figure 1A–C). The selected three sDMIRs were selected to establish the RSM, of which the PCa patients were divided into a high-risk group and a low-risk group based on the median risk score (Figure 3A). The mortality rate enhanced significantly in the high-risk group (Figure 3B). The survival probabilities of patients in the low-risk group were remarkably higher than that of the high-risk group (Figure 4A). The ROC curve was used to investigate the sensitivity of the risk score model, and the AUC of the ROC curve was 0.742, highlighting the satisfied sensitivity of RSM based on the three sDMIRs in predicting the prognosis (Figure 4B).

The relationships between the RSM and clinical features of PCa were detected by an independent t-test. We found that patients with advanced T-stages (Figure 5A) were correlated with remarkably higher expression of miR-10a-5p. High expression of miR-508-3p was observed in early T-stages (Figure 5B). The univariate Cox regression and multivariate Cox regression analyses were displayed to verify further whether the risk score could serve as an independent prognostic factor. As shown in Table 2, the risk score was related to prognosis by the univariate Cox regression analysis. The results of the multivariate Cox regression analysis also showed that the risk score has the potential for predicting the prognosis of PCa patients. These results highlighted that the risk score could be a sensitivity and accuracy independent prognostic factor for Pca patients.

To further find the regulatory relationships between sDMIRs and their target genes, we selected the target genes of sDMIRs from TargetScan, miRTarBase, and miRDB databases, and the results were demonstrated in the Venn diagrams (Figures 6A–C). Meanwhile, the regulatory networks of target genes and sDMIRs are shown in Figure 7. Some of the target genes had a significant correlation with the prognosis of Pca patients, and we further detected the survival curve of these target genes. The results showed that the higher expression levels of JAZF1, PRDM6, RBMS3, and TSHR were correlated with the poor prognosis of Pca patients (Supplementary Figure 2). We used Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses to investigate the potential molecular mechanisms of these target genes. The functional enrichment analysis results showed that “fat cell differentiation,” “adherens junction,” and “DNA-binding transcription activator activity, RNA polymerase II-specific” were the most enriched in biological processes (BP), cellular components (CC), and molecular functions (MF), respectively (Figure 8A). “Axon guidance” was the most enriched pathway in the KEGG pathways (Figure 8B).

To investigate the expression levels of sDMIRs in different prostate cell lines and clinical samples, we detected the expression of miR-508-3p, miR-20a-5p, and miR-10a-5p in nontumorigenic human prostate epithelial cell line (RWPE1), PCa cell lines (PC-3 and DU145), and the tumor tissues and adjacent normal tissues of PCa. Compared to RWPE1, the expression levels of miR-10a-5p (Figure 9A) and miR-20a-5p (Figure 9B) were obviously higher in PCa cell lines (PC-3 and DU145) but miR-508-3p (Figure 9C) showed the opposite result. Besides, we found that the expression of miR-508-3p was prominently lower in tumor tissues than that in the paired adjacent normal tissues, but we detected higher expression levels of miR-20a-5p and miR-10a-5p in tumor tissues (Figure 9D).

To further determine the clinical significance of these sDMIRs, we next examined the associations of miR-508-3p, miR-20a-5p, and miR-10a-5p with T-stage. As illustrated in Figure 9E, miR-20a-5p and miR-10a-5p expressed increasingly as the more advanced T-stages, but the expression of miR-508-3p altered inversely. Furthermore, we investigated the clinical significance of these sDMIRs by analyzing their associations with different clinicopathological characteristics. As shown in Table 3, the higher expression of miR-20a-5p and miR-10a-5p remarkably correlated to the high level of prostate-specific antigen (PSA) and Gleason score, as well as the more advanced T-stage and distant metastasis status.