MicroRNA‐539 functions as a tumour suppressor in prostate cancer via the TGF‐β/Smad4 signalling pathway by down‐regulating DLX1

Abstract Prostate cancer (PCa) is the second leading cause of cancer‐related death in males, primarily due to its metastatic potential. The present study aims to identify the expression of microRNA‐539 (miR‐539) in PCa and further investigate its functional relevance in PCa progression both in vitro and in vivo. Initially, microarray analysis was conducted to obtain the differentially expressed gene candidates and the regulatory miRNAs, after which the possible interaction between the two was determined. Next, ectopic expression and knock‐down of the levels of miR‐539 were performed in PCa cells to identify the functional role of miR‐539 in PCa pathogenesis, followed by the measurement of E‐cadherin, vimentin, Smad4, c‐Myc, Snail1 and SLUG expression, as well as proliferation, migration and invasion of PCa cells. Finally, tumour growth was evaluated in nude mice through in vivo experiments. The results found that miR‐539 was down‐regulated and DLX1 was up‐regulated in PCa tissues and cells. miR‐539 was also found to target and negatively regulate DLX1 expression, which resulted in the inhibition of the TGF‐β/Smad4 signalling pathway. Moreover, the up‐regulation of miR‐539 or DLX1 gene silencing led to the inhibition of PCa cell proliferation, migration, invasion, EMT and tumour growth, accompanied by increased E‐cadherin expression and decreased expression of vimentin, Smad4, c‐Myc, Snail1 and SLUG. In conclusion, the overexpression of miR‐539‐mediated DLX1 inhibition could potentially impede EMT, proliferation, migration and invasion of PCa cells through the blockade of the TGF‐β/Smad4 signalling pathway, highlighting a potential miR‐539/DLX1/TGF‐β/Smad4 regulatory axis in the treatment of PCa.

made in the treatment of localized malignancies, PCa remains to be incurable due to its high metastatic potential, with most PCa-related deaths being attributed to the development of metastasis and high resistance to the existing therapies. 4 Epithelial-mesenchymal transition (EMT) is closely related to the process of PCa metastasis; therefore, identifying agents that can effectively suppress EMT might be a candidate therapeutic option for metastatic PCa. 5 A number of microRNAs (miRNAs), including miR-29b, 6 miR-34a, 7 miR-21, 8 and miR-221 9 have been widely reported to regulate the development of PCa cells. Therefore, there has been a great deal of importance placed on identifying an underlying miRNA that can be a promising biomarker in order to provide an early detection and control of metastasis that occur from PCa. miRNAs play a major role in the pathogenesis and development of multiple cancers, which is due to their participation in cell differentiation and homoeostasis, as well as their involvement in complex regulatory networks commonly with transcription factors used as their direct targets. 10 There has been a close correlation detected between some miRNAs including miR-655, miR-130b and miR-590 with the EMT process during cancer progression. [11][12][13] Moreover, it has been reported that miR-539 can suppress EMT in oesophageal cancer. 14 The restoration of miR-539 has been proven to slow the progression of the proliferation and metastasis abilities of PCa cells by down-regulating sperm-associated antigen 5 (SPAG5). 15 Distal-less 1 (DLX) was initially discovered in Drosophila melanogaster, during which time it was found to regulate the development of nerves and embryo. 16 Previous studies have demonstrated that the depletion of DLX1 expression could cause PCa cell growth arrest, indicating that DLX1 could be a potential diagnostic target for PCa. 17 Furthermore, accumulating evidence shows that DLX1 could participate in the biological processes of PCa. 18 In addition, there is a correlation between the induction of EMT process and the transforming growth factor-β (TGF-β) and Smads. 19 The tumour suppressor Smad4, which has been identified as a pivotal transcription factor in the TGF-β signalling pathway, has been found to be mutated or deleted frequently in PCa. 20 Therefore, on the basis of the existing data mentioned above, we conducted the following study to confirm the potential mechanism of miR-539 in PCa by targeting DLX1 and to determine whether the TGF-β/ Smad4 signalling pathway is involved in the regulation of EMT and metastasis of PCa.

| Ethics statement
The experiments conducted in the present study have been ap-

| Microarray-based gene expression profiling
Gene expression datasets related to PCa including GSE55945, GSE45016 and GSE38241 were retrieved from Gene Expression Omnibus (GEO) database (https ://www.ncbi.nlm.nih.gov/geo/). Differential analysis of these datasets was conducted using the 'limma' package in the R Language Programming with |logFC|> 2 and P value < 0.05 as the screening threshold. Next, the expression heatmap of the differentially expressed genes was constructed using the 'pheatmap' in the R Language Programming. The Venn diagram of the genes with differential expression retrieved from the chips was constructed using the Venn diagram website (http://bioin forma tics.psb. ugent.be/webto ols/Venn/) and the results were then intersected.
As a convenient and interactive database for analysis of transcriptome in cancers, UALCAN (http://ualcan.path.uab.edu/index. html), which could be used to retrieve the expression of specific genes in Cancer Genome Atlas (TCGA) database, was used to retrieve the expression of DLX1 in PCa in the present study. miRNAs regulating DLX1 were predicted by means of microRNA. org (http://34.236.212.39/micro rna/home.do), TargetScan (http:// www.targe tscan.org/vert_71/), mirDIP (http://ophid.utoro nto.ca/ mirDI P/index.jsp#r) and starBase (http://starb ase.sysu.edu.cn/) databases. Next, the intersection was determined for the first 50% miRNAs predicted from the microRNA.org and starBase databases as well as the first 200 miRNAs predicted from the TargetScan and mirDIP databases. Afterwards, the Venn diagram was constructed using the Venn diagram website and the intersection of the miRNAs predicted from these four databases was obtained. The mirSVR score of the miRNA regulating DLX1 predicted from the microRNA. org database was analysed to further determine the candidate miRNA.

| Immunohistochemistry
Paraffin-embedded specimens were cut into 4-μm thick sections after being fixed in 10% formaldehyde and placed in a 60°C oven for l hour. The sections were dewaxed routinely with xylene, dehydrated with graded alcohol and incubated in 3% H 2

| Dual luciferase reporter gene assay
The target gene of miR-539 was conducted using the microRNA.
org. Dual luciferase reporter gene assay was conducted to confirm whether DLX1 was the target gene of miR-539. The 3′-untranslated region (3′UTR) of the DLX1 gene was cloned into pmirGLO vector (E1330, Promega Corp., Madison, WI, USA) and was given the name pDLX1 wild-type (Wt). Site-directed mutagenesis was performed on the potential binding sites of miR-539 and the pDLX1-mutant (Mut) vector was also constructed. pRL-TK vector expressing Renilla luciferase (E2241, Promega Corp., Madison, WI, USA) was constructed to serve as an internal control. The miR-539 mimic or the NC plasmids were transfected into cells together with pDLX1-Wt or pDLX1-Mut.
After 48 hours of transfection, the cells were collected and luciferase activity was detected according to the instructions of the luciferase assay kit (GeneCopoeia, Inc, Rockville, Maryland, USA).

| Reverse transcription quantitative polymerase chain reaction (RT-qPCR)
Levels of DLX1 and downstream target genes of TGF signalling pathway were determined using RT-qPCR. The primers for RT-qPCR are listed in (1 cycle), 45 cycles of 95°C for 20 seconds, 60°C for 1 minute and 72°C for 30 second. β-Actin and U6 served as internal references.
The fold changes were calculated by means of relative quantification (2 −ΔΔCT method), using the following formula: ΔCt = Ct target gene -Ct internal control gene and ΔΔCT = ΔCt experimental group -ΔCt control group .

| Western blot analysis
Once the total protein had been extracted, the protein concentration was measured using the bicinchoninic acid (BCA) Protein Assay Relative protein levels of the target genes were expressed as the ratio of grey values of the target protein bands to those of the β-actin band.

| Scratch test
After 48 hours of treatment, the cells were seeded into 6-well plates. Upon reaching full adherence, the cells were cultured in serum-free DMEM culture medium. When the cell confluence reached 90%-100%, the cells were scratched slowly with a sterile pipette tip (10 μL) perpendicular to the plate bottom. Each well was expected to have 4-5 scratches with the same width. Next, the plates received three washes with PBS, followed by cell collection. An inverted microscope was used to measure the metastasis distance at 0 hour and 24 hours following scratching with various randomly selected visual fields. The photographs were obtained and the experiment was conducted in triplicate.

| Tumour formation in nude mice
The single cell suspension was prepared using PCa cell lines PC3 and

| Statistical analysis
Statistical analysis was conducted by SPSS 19.0 (IBM Corp. Armonk, NY, USA). Measurement data were presented as mean ± standard deviation. Data between two groups were compared using independentsample t test and data among multiple groups were compared using one-way analysis of variance (ANOVA). Fisher's least significant difference (LSD) method was adopted for pairwise comparison. Enumeration data were presented as percentage and analysed using chi-square test.
A value of P < 0.05 indicated statistical significance.

| miR-539 participates in the development of PCa by regulating DLX1
Through differential analysis, 33, 667 and 1008 differentially expressed genes were obtained from the GSE55945, GSE45016 and GSE38241 datasets which were retrieved from the GEO database respectively. Next, the expression heatmap of 30 differentially expressed genes was constructed ( Figure 1A-C). In order to screen out the genes associated with PCa, Venn analysis was conducted to analyse all the differentially expressed genes in GSE55945 and GSE45016 datasets as well as the first 500 significantly differentially expressed genes in the GSE38241 dataset ( Figure 1D), the results of which showed that DLX1 was the only one of them that was simultaneously present and up-regulated in all the three datasets. Then the expression of DLX1 in PCa samples and normal control samples from the TCGA database was analysed and the results revealed that there was a high expression in DLX1 in PCa samples ( Figure 1E). It has been reported that DLX1 could influence disease development via the TGF-β/Smad4 signalling pathway 22,23 and the TGF-β/Smad4 signalling pathway was previously found to be associated with F I G U R E 1 Microarray-based gene expression profiling identifies the involvement of hsa-miR-539 in the development of PCa by mediating DLX1. A-C, expression heatmap of PCa-related gene expression profile, in which the abscissa refers to sample number, the ordinate refers to gene name and the upper crossband refers to sample type. Blue represents normal control sample and red represents PCa sample. The right histogram represents colour grade, in which red indicates high expression and green indicates poor expression. The upper dendrogram refers to sample cluster and the left dendrogram refers to gene expression cluster. Each circle represents the expression of a gene in one sample; D, Venn analysis of differentially expressed genes from three datasets, in which blue refers to the number of differentially expressed genes in GSE55945 datasets, red refers to the number of differentially expressed genes in GSE45016 datasets and green refers to the number of differentially expressed genes in GSE38241 datasets. The middle part is the intersection of three datasets and the number in the image is the number of differentially expressed genes in each region; E, expression of DLX1 in PCa samples and normal control samples from the TCGA database, in which the abscissa represents sample cluster, the ordinate represents the expression of DLX1 gene, the left blue box plot refers to the expression of DLX1 in normal control samples and the right red box plot refers to the expression of DLX1 gene in PCa samples; F, the predicted miRNAs regulating DLX1 gene, in which blue represents the miRNAs predicted from the mirDIP database, red represents the miRNAs predicted from the starBase database, green refers to the miRNAs predicted from the microRNA. org database and yellow refers to the miRNAs predicted from the TargetScan database. The region with number 7 is the intersection of the prediction results from four databases, suggesting that there are seven miRNAs existing simultaneously in these four databases. *P < 0.05, compared with primary tumour group or normal group; PCa, prostate cancer; miR-539, microRNA-539; TGF-β, transforming growth factor-β; DLX1, distal-less homeobox 1; TCGA, The Cancer Genome Atlas PCa. 24,25 Hence, there is a high possibility that DLX1 could affect the development of PCa through the TGF-β/Smad4 signalling pathway.
In order to figure out the potential miRNAs regulating DLX1, the top 50% miRNAs were predicted from the microRNA.org and star-

Base databases and top 200 miRNAs predicted from the mirDIP and
TargetScan databases were subjected to Venn analysis. The results displayed that there were seven miRNAs in the intersection of the prediction results from four databases ( Figure 1F). The binding abilities of these seven miRNAs to DLX1 were scored <−0.6 ( Table 2).
Therefore, miR-539 was selected for subsequent studies and we investigated whether miR-539 could influence the development of PCa through the TGF-β/Smad4 signalling pathway by mediating DLX1.

| PCa tissues presented with a high expression of DLX1
The results from the immunohistochemistry, which was performed in order to detect the positive expression of DLX1 in PCa, suggested that DLX1 was positively stained brown and located at the cytoplasm (Figure 2A). PCa tissues exhibited a positive rate of DLX1 of 76.92% and the adjacent normal tissues displayed 18.46% ( Figure 2B). Compared with the adjacent normal tissues, the protein expression of DLX1 in PCa tissues was considerably increased (P < 0.05). These findings indicate that there was a higher positive rate of DLX1 protein expression in the PCa tissues.

| PCa tissues exhibit reduced miR-539 but elevated DLX1
RT-qPCR and Western blot analysis were carried out in order to fur- However, the expression levels of miR-539 and E-cadherin were greatly reduced in the PCa tissues (P < 0.05). These findings provided evidence that miR-539 level is reduced in PCa tissues.

| The expression of DLX1 and miR-539 is correlated with lymph node metastasis, pathological stage, tumour size and Gleason scores of PCa
The correlation between the clinical characteristics of PCa patients and the expression of DLX1 and miR-539 was determined. As shown in Table 3, the results found that the expression levels of DLX1 and miR-539 were not statistically related to age and family history of patients with PCa (P > 0.05), but were associated with lymph node metastasis, pathological stages, tumour size and Gleason scores of PCa (P < 0.05).

| DLX1 is a target gene of miR-539
Since DLX1 was predicted to be targeted by miR-539 ( Figure 4A), dual luciferase reporter gene assay was performed to verify the target relationship between miR-539 and DLX1. Compared with the NC group, there was a decrease in the luciferase activity of Wt-miR-539/ DLX1 in the miR-539 mimic transfection group (P < 0.05) ( Figure 4B).
However, the luciferase activity did not differ greatly in the Mut-miR-539/DLX1 transfection group (P > 0.05). Therefore, miR-539 could specifically bind to DLX1 gene.  Note: miRNA refers to the predicted miRNAs regulating DLX1; mirSVR score is thermodynamic stability score (the criterion is ≤−0.1) and the lower the score, the stronger of the binding ability of miRNA-mRNA, suggesting a higher possibility that miRNA down-regulates genes; PhastCons score represents the evolutionary conservatism of gene untranslated region in species and stronger evolutionary conservatism reflects better prediction; miR, microRNA; DLX1, distal-less homeobox 1.

| Up-regulation of miR-539 or silencing of DLX1 inhibits PCa cell proliferation
MTT assay was employed to determine cell viability. The results (shown in Figure 6) elucidated that both PC3 and DU145 cell lines presented with the consistent trend. There was no significant difference observed in the cell proliferation in the blank and NC groups (P > 0.05). Compared with the blank and NC groups, the miR-539 inhibitor group showed distinctly increased cell proliferation, but the miR-539 mimic and si-DLX1 groups had evidently inhibited cell proliferation (P < 0.05). The miR-539 inhibitor + si-DLX1 group had no significant difference (P > 0.05). These results suggested that overexpression of miR-539 or silencing DLX1 could result in the suppression of the proliferation of PCa cells.

| Up-regulation of miR-539 or silencing of DLX1 suppresses PCa cell migration
Scratch test was adopted to detect cell migration. As shown in Figure 7, migration of PC3 and DU145 cell lines displayed the consistent trend.
The blank and the NC groups had no significant difference concerning

| Up-regulation of miR-539 or silencing of DLX1 inhibits PCa cell invasion
The Transwell assay was conducted in order to measure PCa cell in-

| Up-regulation of miR-539 impedes PCa cell proliferation, migration and invasion through the inhibition of the TGF-β signalling pathway
MTT assay, scratch test and transwell assay were conducted to elucidate the role of miR-539 in PCa cell proliferation, migration and invasion through TGF-β signalling pathway. As shown in  The initial findings of our study indicated that the there was a poor expression in miR-539 while DLX1 was highly expressed in PCa tissues. Consistently, a previous study has also indicated that there In addition, the reduction in miR-539 expression has also been detected in triple-negative breast cancer and the up-regulated miR-539 could suppress tumour progression by negatively regulating and targeting the expression of LAMA4. 30 The expression of miR-539 was reported to be negatively linked to the lymph node metastasis and clinical phase of human colorectal cancer, suggesting that miR-539 could slow the progression of colorectal cancer. 29 Moreover, there is an overexpression in DLX1 in PCa and DLX1, which could be considered as an underlying marker for the treatment of PCa. 18 There has also been a study indicating that DLX1 can serve as a promising indicator for high-grade PCa detection. 31 In addition to PCa, highgrade serous ovarian cancer was also found to have presented with increased DLX1, indicating that it can also play a role as a therapeu-     In addition, our findings also suggested that the study con-

| CON CLUS ION
In conclusion, the aforementioned findings provided evidence that miR-539 could result in the suppression of cell migration, invasion, migration and EMT in PCa by inactivating the TGF-β/ Smad4 signalling pathway through the direct down-regulation of DLX1 ( Figure 11), which provides new insights for future study of PCa treatment. However, this study is still at the pre-clinical stage and the investigation on the mechanism of action is not yet well elucidated. Therefore, further large-scale studies are required to broaden the therapeutic perspective of PCa from the cellular level by understanding cell capacity and cell-cell interaction.

ACK N OWLED G EM ENTS
We would also like to thank all participants enrolled in the present study.

CO N FLI C T O F I NTE R E S T
None declared.