3.1 Characteristics of PCa patients
Patient profiles are summarized as follows: patient set 1 (RP samples), n=53, including GP3: n=29, GP4: n=44, GP5: n=16, and adjacent normal tissue: n=53 (Table1), patient set 2 (biopsy samples), n=107, including GP3 from the real GS6 cancer tissues, n=32, as were all confirmed in the following RP, and GP3 from GS7 in the biopsy tissues, n=75 (Table1), patient set 3 (biopsy samples), n=119, incorporating 32 GP3 from the real GS6 PCa in the patient set 2, and all were GS6 in the biopsy, including 66 samples with the real GS6 and 53 samples with GS ≥ 7 in the subsequent RP (Table2).
3.2 Relative expression of miRNAs between the Gleason patterns 3, 4 and 5
To determine the expression of miRNAs associated with the GP of PCa, we investigated seven miRNAs, including miR-1-5p, miR-21-5p, miR-30d-5p, miR-100-5p, miR-145-5p, miR-224-5p, and miR-708-5p, all of which are implicated in PCa progression. Firstly, we explored the differential expression of seven miRNAs between GPs and adjacent normal tissues using RP samples (patient set 1). As shown in Figure 1, compared with normal samples, the relative expression of miR-224-5p in each GP was significantly decreased, the expression of miR-145-5p was downregulated in GP4 and GP5 but not in GP3, while the expression of miR-1-5p was downregulated in GP4 but not in GP3 or GP5. Then the miRNA expression among the different GPs was investigated. We found that the relative expression of miR-145-5p in GP4 was significantly lower than that in GP3, and no other differential expression was found in this study.
3.3 Altered miRNA expression among the same GP with different GS
To explore whether miRNA expression in PCa with the same GP differed according to GS, we analyzed miRNA expression in GP4 and GP3 separately. In GP4, samples from GS7 or GS ≥ 8 were evaluated, but no significant difference in miRNA expression associated with GS was detected (Supplementary Figure1). However, in GP3, expression of miR-145-5p, miR-224-5p and miR-708-5p in samples obtained from GS6 was significantly higher compared with that of GS7 (Figure2). Tissues from GP5 were excluded from the analysis, as all samples were obtained from GS9.
To verify the differential expression of miRNAs in the same GP3 between GS6 and GS7, a validation study using the biopsy samples was performed with patient set 2. We found that the expression of miR-145-5p was significantly higher in GP3 tissues from the real GS6 compared with that associated with GP4, which was similar to the result obtained from the RP group. However, statistically-insignificant difference in the expression of miR-224-5p and miR-708-5p was found between GP3 samples from the real GS6 and that associated with GP4 (Figure3).
3.4 Association between the expression level of miR-145-5p and Gleason upgrading of biopsy samples scored GS6
To further investigate the performance of miR-145-5p in predicting the Gleason upgrading (GU) from GS6 biopsy samples, we detected its expression level using the cohort 3. The 32 biopsy samples with the real GS6 from patient set 2 were also incorporated in this patient set. The total GU was 44.5% (n=53) and the comparison of the nonupgrading, and upgrading groups was summarized in Table 2. As shown in figure 4A, only the expression level of miR-145-5p decreased significantly in GU group. On ROC analysis, miR-145-5p yielded an AUC value of 0.705 (95%CI, 0.612-0.798, P < 0.001) with 50.8% sensitivity and 84.8% specificity (Figure 4B). The AUC for other clinical parameters in predicting GU was 0.592 for age, 0.554 for PSA, 0.567 for PSAD and 0.525 for percent of positive cores respectively (Supplement Figure 2). In multivariate analyses, miR-145-5p was an independent predictor of GU (Table 3).
3.5 Target gene prediction of miR-145-5p
To further determine the biologic functions affected by miR-145-5p in PCa, the potential target genes were predicted by miRwalk 3.0 and two other miRNA prediction tools. The intersection of these three predicted target gene sets was integrated and visualized using Venn diagram (Supplementary Figure 3). Finally, 125 overlapping genes were consistently predicted by the three databases, suggesting that these genes might be involved in the miR-145-5p-mediated functions and biological pathways in PCa.
3.6 Go annotation and KEGG enrichment analyses
To obtain insight into the biological roles of miR-145-5p in PCa, GO annotation and KEGG pathway enrichment were analyzed through Metascape (25) and KOBAS (26). As shown in Figure 5, these 125 target genes were mainly enriched in small GTPase mediated signal transduction and semaphoring-plexin signaling pathway with regard to biological processes (BPs) (Figure 5A). As for molecular functions (MFs), terms such as GTPase regulator activity and actin filament binding were enriched(Figure 5B). Regarding cellular components (CCs), the target genes were commonly enriched in the cell leading edge, postsynaptic specialization, contractile fiber and adherens junction (Figure 5C). Furthermore, the KEGG analysis revealed that adherens junction, endocytosis, dilated cardiomyopathy and HIF-1 signaling were significantly enriched pathways. Besides, the ecadherin stabilization pathway and smad2/3 nuclear pathway were enriched by Canonical Pathways enrichment, and the UV response DN and apical junction were enriched by HALLMARK enrichment (Figure 5D). In addition, databases such as KEGG DISEASE, and GWAS Catalog were also explored with KOBAS to further find out the relationship between miR-145-5p and diseases. The results showed that cardiovascular diseases and cancers were significantly enriched (Supplementary Figure4), suggesting a potential correlation between miR-145-5p and PCa.
In view of missing key information from merely taking the intersection of the predicted targets into consideration, an enrichment analysis was also performed on all predicted target genes with the only miRWalk(27). Consequently, 6,035 target genes with a miRWalk score >1 were used for the functional enrichment analysis. The results showed that cell junction assembly, neuron projection guidance and regulation of GTPase activity were enriched with regard to BPs(Supplementary Figure 5A). As for MFs, symporter activity, solute:cation symporter activity, GTPase regulator activity and small GTPase binding were significantly enriched (Supplementary Figure 5B), indicating a close association between miR-145-5p and the GTPase activity. Regarding CCs, the apical part of cell and synaptic membrane were enriched (Supplementary Figure 5C). In addition, terms such as pathway in focal adhesion were enriched In the KEGG pathway analysis, indicating that miR-145-5p might be involved in the structure remodeling in the carcinogenesis of PCa.