Downregulation of miRNA-205 Expression and Biological Mechanism in Prostate Cancer Tumorigenesis and Bone Metastasis

Background The expression and mechanism of microRNA-205 (miRNA-205) in prostate cancer (PCa) and its bone metastasis remain controversial. Materials and Methods The expression and discriminating capability of miRNA-205 were assessed by drawing a forest plot and a summarized receiver operating characteristic (SROC) curve, using data available from 27 miRNA-array and miRNA-sequencing datasets. The miRNA-205 target genes were acquired from online prediction tools, differentially upregulated genes in PCa, and differentially expressed genes (DEGs) after miRNA-205 transfection into PCa cell lines. Functional enrichment analysis was conducted to explore the biological mechanism of miRNA-205 targets. Immunohistochemistry (IHC) was applied to verify the protein level of the hub gene. Results The expression of miRNA-205 in the PCa group (1,461 samples) was significantly lower than that in the noncancer group (510 samples), and the downregulation of miRNA-205 showed excellent sensitivity and specificity in differentiating between the two groups. In bone metastatic PCa, the miRNA-205 level was further reduced than in nonbone metastatic PCa, and it showed a good capability in distinguishing between the two groups. In total, 153 miRNA-205 targets were screened through the three aforementioned methods. Based on the results of functional enrichment analysis, the targets of miRNA-205 were mainly enriched during chromosome segregation and phospholipid-translocating ATPase activity and in the spindle microtubule and the p53 signaling pathway. CDK1 had the highest connectivity in the PPI network analysis and was screened as one of the hub genes. A statistically significant negative correlation between miRNA-205 and CDK1 was observed. The expression of CDK1 in PCa samples was pronouncedly upregulated in terms of both the mRNA level and the protein level when compared with noncancer samples. Conclusion miRNA-205 may play a vital role in PCa tumorigenesis and bone metastasis by targeting CDK1.


Introduction
Prostate cancer (PCa) is one of the most prevalent malignancies, being the second most common cause of death among men [1]. The high mortality associated with PCa is closely related to distant metastasis, especially bone metastasis, which often develops in PCa [2][3][4]. Bone metastasis exists in almost 70% of advanced PCa patients [5,6]. There are currently a variety of treatments for localized PCa, including endocrine therapy, radiotherapy, and surgical treatment [7][8][9]. However, treatments are limited for patients with advanced stages of the disease and particularly bone metastatic PCa [10]. Therefore, there is an urgent need to elucidate the molecular mechanism of PCa tumorigenesis and bone metastasis and to identify new therapeutic targets.
MicroRNAs (miRNAs) are noncoding RNAs with a length of about 22 nucleotides, which affect targets' expression posttranscriptionally [11][12][13][14][15][16][17]. Previous studies have shown that dysregulated miRNA-205 is associated with the tumorigenesis and metastasis of a variety of malignancies. For example, the miRNA-205 level was found to be significantly higher in pancreatic cancer patients; it promotes the proliferation of pancreatic cancer cells by targeting APC [18]. Downregulated miRNA-205 is also involved in the epithelial-mesenchymal transition in breast cancer via the HMGB1-RAGE axis [19]. miRNA-205 was found to increase the proliferation and invasion capability of ovarian cancer cells by inhibiting SMAD4 and PTEN expression [20]. In PCa, some studies have observed that miRNA-205 is reduced and is a tumor suppressor [21][22][23][24]. However, Osipov et al. detected miRNA-205 expression to be predominantly higher in PCa patients than in control patients and found that the high expression of miRNA-205 distinguished PCa cases from control cases (sensitivity = 0:667; specificity = 0:773) [25]. In addition, in a study by Stephan et al., miRNA-205 expression was not statistically different among PCa and noncancer cohorts [26]. Therefore, the expression and biological mechanism of miRNA-205 in PCa need to be further explored.
In the present study, miRNA-array and miRNAsequencing datasets were used to comprehensively evaluate the expression and discriminating capability of miRNA-205 in PCa and bone metastatic PCa. miRNA-205 targets were collected, and functional enrichment analysis was applied to further understand the molecular mechanism of miRNA-205 targets in PCa tumorigenesis and bone metastasis. Protein-protein interaction (PPI) network analysis  BioMed Research International screened CDK1 as one of the hub genes, and its expression in PCa was verified at the messenger RNA (mRNA) level and protein level. The design flow of the study is shown in Figure 1(a). (1) studies based on animals were excluded, and (2) studies in which fewer than three samples were used were excluded.

Materials and Methods
2.3. Identification of Potential miRNA-205 Target Genes. The miRWalk2.0 archive [27], which contains 12 prediction tools, was applied to predict the targets of miRNA-205. Genes predicted by at least four tools simultaneously were considered to be targets of miRNA-205. In our study, miRNA-205 was confirmed to be downregulated in PCa. In general, miRNAs play the biological role of negatively regulating target genes. As such, we next looked for genes that were differentially upregulated in PCa as miRNA-205 target genes. The screening terms and data processing steps are shown in Figure 1(b).

BioMed Research International
In order to more accurately obtain the targets of miRNA-205 in PCa, studies that interfered with miRNA-205 in PCa cell lines were included. After searching the GEO and ArrayExpress databases, only one array (GSE66498) was obtained. GSE66498 was performed in human PCa cell lines, including PC3 and DU145, and transfected with miRNA-205 and a negative control. In GSE66498, genes satisfying the standard log2FC > 1 or log2FC<−1 (p value < 0.05) were assumed to be the targets of miRNA-205. The overlapping of the above three units of genes was considered to represent targets of miRNA-205.  Meta-analysis estimates, given named study is omitted Begg's funnel plot with pseudo 95% confidence limits Discovery (DAVID). The PPI network analysis of miRNA-205 targets was built using the Search Tool for the Retrieval of Interacting Genes (STRING). In the PPI network, genes having the highest degree of connectivity were identified as hub genes that might play a critical role in PCa.

Correlation Analysis between miRNA-205 and Hub
Genes. The miRNA-205 and hub gene expression data were downloaded from the GSE21032 dataset and TCGA database. The raw data were converted by log2 ðx+1Þ . Spearman's correlation analysis was applied to assess the association between miRNA-205 and hub genes.
2.6. Validation of Hub Gene Expression in PCa 2.6.1. mRNA Expression of Hub Genes in PCa. As shown in Figure 1(b), 21 studies containing gene expression data in PCa were incorporated. Sixteen studies containing hub gene expression data were selected to assess the mRNA expression of hub genes in PCa.

mRNA Expression of Hub Genes in Bone Metastatic
PCa. Studies involving hub gene expression data in bone metastatic PCa were screened using the aforementioned databases and search formulas. The studies published before December 15, 2019, were incorporated. The data inclusion criteria were as follows: (1) each study had to include a bone metastatic PCa group and a nonbone metastatic PCa group, and (2) hub gene expression data had to be available. The exclusion criteria were as follows: (1) studies based on animals were excluded, and (2) studies in which fewer than three samples were used were excluded.   BioMed Research International and the raw data was processed by log2 ðx+1Þ . Student's t-test was employed to estimate the differential expression of miRNA-205 and hub genes between the two groups using SPSS 22.0 (IBM, Chicago, IL, USA). Based on the expression of miRNA-205 and hub genes, the cutoff value for each study was calculated. A receiver operating characteristic (ROC) curve was applied to test the discriminating capability of miRNA-205 and hub genes in each study. A SROC curve was employed to estimate the overall discriminating capability of miRNA-205 and hub genes. The standardized mean difference (SMD) and 95% confidence interval (CI) were determined using the Stata 12.0 software (Stata Corp., College Station, TX, USA) to evaluate the differential expression of miRNA-205 and hub genes between two groups. The I 2 statistic and chi-squared test were applied to determine the heterogeneity of the studies. When p > 0:05 and I 2 < 50%, the studies were considered homogeneous, and a fixedeffects model was used. When p < 0:05 and I 2 > 50%, the studies were considered heterogeneous, and a randomeffects model was used. A p value < 0.05 was considered statistically significant.

miRNA-205 Expression in PCa.
A total of 27 eligible studies were included (Table 1) (Figure 2(a)). Due to the significant heterogeneity that was observed, we first performed a subgroup analysis based on the sample type, and the results showed that the sample type was not the source of the heterogeneity (Figure 2(b)). To further probe the origin of the heterogeneity, we implemented a sensitivity analysis; however, the result showed that no study caused the heterogeneity (Figure 2(c)). The results of the funnel plot indicated no publication bias (Figure 2(d)). The SROC curve showed that miRNA-205 had an excellent capacity to discriminate PCa samples from noncancer samples (area under the curve ðAUCÞ = 0:92; sensitivity = 0:86; and specificity = 0:84) (Figure 3(a)). No publication bias was observed (Figure 3(b)). The diagnostic capacity of miRNA-205 in each study is shown in Figs. S3 and S4.

Identification of miRNA-205 Target
Genes. The miRNA-205 targets predicted by at least 4 tools were screened from the miRWalk2.0 database, resulting in a total of 10,959 genes. According to the screening terms and data processing   Meta-analysis estimates, given named study is omitted  procedure in Figure 1(b), a total of 21 studies were incorporated, and 723 genes were identified as targets of miRNA-205. In the GSE44698 dataset, 9,063 genes satisfying the standard log2FC<−1 (p value < 0.05) were screened as miRNA-205 targets. After intersecting the above 3 parts of genes, 153 genes were finally obtained as potential targets of miRNA-205 (Figure 4(a)).

GO, KEGG, and PPI Analyses.
To further understand the biological function of miR-205 in PCa, GO and KEGG analyses were performed in DAVID. Among the GO analysis results, chromosome segregation; transcription, DNAtemplated; and mitotic metaphase plate congression were the three most significant biological processes (BP). The results of the cellular component (CC) analysis indicated that miRNA-205 potential targets were mainly concentrated in the spindle microtubule; spindle midzone; and chromosome, centromeric region. For molecular function (MF), the three significantly involved entries were phospholipidtranslocating ATPase activity, methylenetetrahydrofolate dehydrogenase activity, and methenyltetrahydrofolate cyclohydrolase activity (Table S2 and Figure 5). The KEGG results indicated that miRNA-205 targets were mainly concentrated in the p53 signaling pathway (Table S2 and Figure 5), while the PPI network analysis showed that CDK1 had the highest connectivity among the miRNA-205 targets and was screened as one of the hub genes (Figure 4(b)).

Correlation Analysis between miRNA-205 and CDK1.
Spearman's correlation analysis was applied to detect the association between miRNA-205 and CDK1 in PCa. The result showed that there was a significantly reverse connection between miRNA-205 and CDK1 in the GSE21032 dataset and TCGA database (Figures 4(c) and 4(d)). In addition, we acquired the expression of miRNA-205 and CDK1 in different types of cancer from starBase v3.0. A negative correlation between miRNA-205 and CDK1 was observed in thy-moma, brain lower grade glioma, colon adenocarcinoma, uterine corpus endometrial carcinoma, and breast invasive carcinoma ( Figure 6).

Validation of CDK1 Expression in PCa
3.6.1. CDK1 mRNA Expression in PCa. Sixteen studies from the mRNA-array and mRNA-sequencing datasets were used to verify the mRNA expression level of CDK1 (Table 2). Of the 16 studies, 6 showed that CDK1 expression was predominantly increased in PCa samples compared with noncancer samples, while in the other 10 studies, no statistical difference in CDK1 expression was detected between PCa samples and noncancer samples (Fig. S6). The result (SMD = 0:50; 95% CI = 0:28 to 0.73; and I 2 = 57:4%) indicated that CDK1 expression was significantly higher in the PCa group (1,159 samples) than in the noncancer group (491 samples) (Figure 7(a)). Due to the existing heterogeneity, a sensitivity analysis and subgroup analysis were applied to investigate the origin of the heterogeneity, but neither analysis found the origin of the heterogeneity (Figures 7(b) and 7(c)). The results of the funnel diagram showed no publication bias (Figure 7(d)). The SROC curve indicated that CDK1 had a good capability in discriminating PCa samples from noncancer samples (AUC = 0:76; sensitivity = 0:70; and specificity = 0:71) (Figure 7(e)). No publication bias was observed (Figure 7(f)). The diagnostic capacity of CDK1 in various studies is shown in Fig. S7.

Validation of CDK1 mRNA Expression in Bone
Metastatic PCa. According to the inclusion and exclusion criteria, a total of three studies were incorporated to verify the expression level of CDK1 in bone metastatic PCa (Table S3). All three studies showed that CDK1 was increased in bone metastatic PCa compared with nonbone metastatic PCa; however, only GSE32269 had significant statistical significance (Figs. S8A, S8C, and S8E). Next, a comprehensive evaluation of CDK1 expression was 11 BioMed Research International performed by calculating the SMD, which was 1.43 (95% CI = 0:32 to 2.54; I 2 = 77:0%). The SMD value showed that the expression of CDK1 in bone metastatic PCa was predominantly increased compared with that in nonbone metastatic PCa (Figure 8). Three studies, including GSE32269, PMID: 26000489, and TCGA, showed that CDK1 had a good capability for discriminating between bone metastatic patients and nonbone metastatic patients, with AUC of 0.926, 0.903, and 0.785, respectively (Figs. S8B, S8D, and S8F).  samples (Figures 9 and 10). In addition, we observed that upregulated CDK1 in PCa was significantly associated with pathological T, N, and M stages (Table 3). However, there was no significant correlation between CDK1 and other clinical characteristics.

Clinical Significance of miRNA-205 and CDK1 in PCa.
The association between miRNA-205 and clinical characteris-tics was analyzed in PCa patients using TCGA database. While miRNA-205 was significantly related to the pathological T stage and Gleason score, no statistical correlation was observed between miRNA-205 and the overall survival (OS) rate or other clinical characteristics (Table S4 and Figure 11(a)). The correlation between CDK1 and clinical characteristics in PCa patients based on TCGA database was also explored. CDK1 was significantly associated with age, pathological T and N stages, Gleason score, and recurrence (Table S5). A survival analysis showed that the PCa group with upregulated CDK1 had a shorter OS time than the PCa group with downregulated CDK1 (Figure 11(b)).

Discussion
Previous research has detected that miRNA plays a vital role in the tumorigenesis and metastasis of PCa. However, the expression and biological function of miRNA-205 in PCa remain controversial. The present study used 27 miRNAarray and miRNA-sequencing datasets containing 2,031 samples to verify the expression and diagnostic capability of miRNA-205 in PCa. In addition, the expression and distinguishing capability of miRNA-205 in bone metastatic PCa were also explored. We screened 153 miRNA-205 target genes using 3 different methods, which increased the reliability of the targets. Functional enrichment analysis was applied to clarify the biological mechanism of miRNA-205 in PCa. CDK1 had the highest connectivity in the PPI network analysis and was screened as one of the hub genes. In PCa, the mRNA expression level of CDK1 was verified using 1,650 samples from 16 datasets. Meanwhile, IHC was used to verify the CDK1 protein expression level. The clinicopathological significance of miRNA-205 and CDK1 in PCa were also explored. This study is the first extensive analysis of the significance of miRNA-205 and CDK1 in PCa.
Our study found that PCa samples had a lower expression of miRNA-205 compared with noncancer samples, which was inconsistent with the results of studies by Osipov et al. and Stephan et al. Osipov et al. tested the miRNA-205 expression in 48 PCa patient blood samples and 47 healthy volunteer blood samples and found that miRNA-205 expression was significantly higher in PCa patients than in controls. The discrepancies between their results and those of the present study may be due to small sample sizes (n < 50) and the different detection methods. In the study by Stephan et al., urine sediment from 38 PCa patients and 38 noncancer patients was collected to assess the expression of miRNA-205 in PCa, and no difference in miRNA-205 expression was detected between the two groups. Stephan et al. explained that the possible reason for this was that they only measured the miRNA-205 in urine sediment and excluded that from exosomes. The AUC value of the SROC curve of the miRNA-205 was 0.92, which indicated that miRNA-205 had an excellent capability in differentiating between PCa samples and noncancer samples. In addition, miRNA-205 was significantly associated with the pathological T stage and Gleason score. The above results showed that miRNA-205 might be a tumor suppressor gene involved in PCa.   [29], consistent with our result. In addition, miRNA-205 had a good capability in distinguishing between bone metastatic PCa samples and nonbone metastatic PCa. However, large sample cohorts are needed to further verify our results.
At present, the mechanism of PCa tumorigenesis and bone metastasis is not clear. GO and KEGG analyses were performed to elucidate the role of miRNA-205 targets in PCa. Some important GO and KEGG entries have been reported that play important roles in PCa. For example, the expression of FZD8 was found to be significantly higher in bone metastatic PCa samples and promoted the occurrence of bone metastatic PCa samples by activating the Wnt signaling pathway [30]. Both miRNA-182 and miRNA-200 negatively regulated the expression of GNA13, thereby reducing the invasion and migration capability of PCa [31]. TMSG1 regulated ATPase activity and inhibited the invasion and metastasis of PCa cells [32]. Long noncoding RNA LINP1 promoted the progression of PCa by regulating the p53 signaling pathway [33]. Therefore, it is meaningful to further study the role of miRNA-205 targets in PCa.
Previous studies have reported that miRNA-205 may affect the progression and metastasis of PCa by targeting HMGB3, c-SRC, and CENPF [21][22][23]. However, no other targets of miRNA-205 have been reported. In this study, CDK1 was selected as the hub gene and the most likely to be targeted by miRNA-205. CDK1 is a member of the serine/threonine protein kinase family, which is necessary for G1/S and G2/M phase transitions in the eukaryotic cell cycle [34]. CDK1 plays an important role in a variety of malignancies, including PCa. In human melanoma, the interaction between CDK1 and SOX2 could promote tumorigenesis [35]. CDK1 was significantly increased in ovarian cancer, and knocking down CDK1 reduced the growth of ovarian cancer cells [36]. The upregulation of CDK1 was associated with poor survival in patients with renal clear cell carcinoma [37]. MicroRNA-1271 was increased in endometrial cancer and reduced proliferation by inhibiting CDK1 expression [38]. In PCa, CDK1 phosphorylated the androgen receptor (AR) and promoted PCa cell proliferation, while CDK1 inhibitors reduced AR phosphorylation and protein expression in PCa cells [39,40]. Decreased levels of CDK1 reduced tumor growth activity, and increased levels of CDK1 promoted PCa cells to develop in the direction of G2/M, which may be the cause of everolimus resistance [41]. In Sarwar et al.'s research, CDK1 complexes with AR-V7 and PIP5K1α to accelerate PCa growth and metastasis [42]. CDK1 was overexpressed in metastatic castration-resistant PCa versus hormone-sensitive PCa and could be used as a prognostic indicator of PCa [43].
In our study, the mRNA level and protein expression level of CDK1 were significantly higher in PCa samples compared with noncancer samples, according to the SMD and IHC. The high expression of CDK1 can distinguish PCa samples from noncancer samples with 70% sensitivity and 71% specificity. CDK1 expression was further increased in bone metastatic PCa samples compared with nonbone metastatic PCa samples, and it had a good capability in distinguishing between the two groups. The results of our study showed CDK1 was significantly associated with pathological T and N stages, recurrence, and overall survival. The above results indicate that CDK1 plays a vital role in PCa occurrence and bone metastasis and might be a therapeutic target for PCa.
Although we obtained valuable findings, this study does have some limitations. First, there was significant heterogeneity, which would reduce the reliability of our results. Even though we tried to solve this problem, the results of subgroup analysis and sensitivity analysis failed to identify the source of heterogeneity. The following factors may be sources of heterogeneity. (1) Datasets were collected from 10 different countries, including the USA, Turkey, Spain, Mexico, Italy, Iran, Germany, Finland, China, and Brazil. (2) The control group had different sources, including normal prostate tissues, paracancerous tissues, benign prostatic hyperplasia

Conclusions
Overall, the present study demonstrated that miRNA-205 is downregulated in both PCa and bone metastatic PCa. In addition, miRNA-205 may be involved in the carcinogenesis and bone metastasis of PCa by negatively regulating CDK1 expression.

Data Availability
The raw data is available at the GEO, ArrayExpress, Oncomine, SRA, and TCGA databases.