The sequence of the 3′UTR of IGF1R was retrieved from the University of California Santa Cruz database (http://genome.ucsc.edu/) in the human genome assembly GRCh37/hg19. The targets for miRNA binding in the IGF1R 3′UTR containing two polymorphisms, rs1815009 or rs2684788, were predicted using the TargetScan7.1 (http://www.targetscan.org) (19) and miRanda databases (http://www.microrna.org) (20). Mature human sequences were downloaded from the National Center for Biotechnology Information (https://www.ncbi.nlm.nih.gov/). A parameter-free thermodynamic model was used (21,22) to investigate the binding affinity of miR-133a and miR-133b for the C or T allele of rs1815009, and miR-455 for the C or T allele ofrs2684788. In this model, E_Target represented the energy of the dissociated mRNA target region with no miRNAs interacting with it, E_Intermediate indicated the energy required to make the target region accessible for microRNA binding, and E_Complex is the energy of the microRNA-target complex, the binding of which was consistent with the constraints imposed by the seed region (6,7,23). RNAfold software (Vienna RNA Package; http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi) was used to compute all ensemble free energies with the partition function algorithm for folding and the default settings. The RNAfold was modified to calculate the E_Intermediate and E_Complex in order to construct the connectivity matrix of the intermediate or complex product secondary structures, and the corresponding structure-associated energies were calculated. The target sequence size contained the relevant miRNA target seed region, the relevant IGF1R SNP site and the nucleotides surrounding the SNP site. The site length provided for the RNAfold was 42 nt for rs1815009 and 60 nt for rs2684788. The value of the selected target size for rs1815009 and rs2684788 was chosen to reduce the time complexity of the aforementioned computations, particularly as the results were not altered when the length of the target was extended (Fig. 1). The ΔE_a is the difference between E_Target and E_Intermediate, and ΔE_b is the difference between E_Target and E_Complex (6,7).

A thorough literature search was performed using PubMed on miRNA expression profiling studies for PCa with the following key words: Prostate and (cancer OR tumor OR tumor) and (mirna OR microrna OR mir-). Additionally, miRNA expression profiles for PCa were retrieved by searching the Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) (24). Profiles whose data published between June 14, 2007 and April 28, 2015 were selected. The inclusion criteria were as follows: i) Experimental samples consisting of prostate tumor and non-tumor tissues, for which the complete raw or normalized microarray data was available; ii) data included genome-wide miRNA expression; iii) Homo sapiens as the studied organism; iv) samples that had not been subjected to adjuvant therapy; and v) the number of PCa and normal controls was >15. For the present study, the non-tumor tissues included adjacent tissues of PCa or tissues from independent healthy donors, but excluded benign prostatic hyperplasia. The data extracted from each GEO dataset was as follows: Tumor sample type, country, platform, experimental materials and number of samples (Table I). The list of miRNAs with statistically significant expression in PCa was extracted from the publications. All statistical analyses were performed using GEO2R.

The normalized miRNA-HiSeq expression values and PCa clinical information was downloaded from the Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov/). The reads/million miRNA mapped (RPM) normalized values for miRNA expression were further log₂-transformed. The fold change in miRNA expression between tumors and normal tissues was calculated using the median-centered RPM values. The information for TCGA dataset is presented in Table I. The difference in expression between PCa and normal tissue was analyzed using an independent sample t-test.

Truncated UTR sequences containing the SNP region were amplified by polymerase chain reaction (PCR) from genomic DNA, then ScaI cloned into the 3′UTR of the pmirGLO-control vector (Promega Corporation, Madison, WI, USA) following the manufacturer's protocol. The 3′UTR reporter vector for IGF1R, containing the C allele of rs1815009 and seed region of miR-133a or miR-133b, was named pmirGLO-rs1815009-IGF1R-[C]. Plasmids carrying the 3′UTR with the T allele were designated as pmirGLO-rs1815009-IGF1R-[T]. Likewise, two 3′UTR constructs for rs2684788 containing the C or T alleles were designated as pmirGLO-rs2684788-IGF1R-[C] and pmirGLO-rs2684788-IGF1R-[T]. Each plasmid was co-transfected with the miRNA precursor (pre-133a, pre-133b, pre-455 or scrambled negative control; Shanghai GenePharma Co., Ltd., Shanghai, China) in 293 cells cultured in 24-well plates. The following sequences were used: Pre-133a, AATTCACAATGCTTTGCTAGAGCTGGTAAAATGGAACCAAATCGCCTCTTCAATGGATTTGGTCCCCTTCAACCAGCTGTAGCTATGCATTGAACCGG; pre-133b, AATTCCCTCAGAAGAAAGATGCCCCCTGCTCTGGCTGGTCAAACGGAACCAAGTCCGTCTTCCTGAGAGGTTTGGTCCCCTTCAACCAGCTACAGCAGGGCTGGCAATGCCCAGTCCTTGGAGAACCGG; pre-455, AATTCTCCCTGGCGTGAGGGTATGTGCCTTTGGACTACATCGTGGAAGCCAGCACCATGCAGTCCATGGGCATATACACTTGCCTCAAGGCCTATGTCATCACCGG; and scrambled negative control, AAATGTACTGCGCGTGGAGAC. Following transfection with Lipofectamine 2000 reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA), 1.5×10⁶ cells/ml cells were cultured for 24 h, then lysed in passive lysis buffer, according to the Dual-Luciferase^® Reporter Assay system kit (Promega Corporation), and luciferase activity was measured with a Tecan M1000 microplate reader (Tecan Group, Ltd., Mannedorf, Switzerland). A dual-luciferase system was utilized in this assay, containing two luciferase enzymes, one containing the luc₂ reporter gene that reports miRNA activity, and another containing the hRluc-neo reporter gene used as a control reporter for normalization of gene expression. Normalization of the empty pmirGLO vector was performed using the Renilla/firefly luciferase ratios. Each luciferase assay was performed in triplicate and the mean value was calculated.

The human PCa cell line PC3 was obtained from the American Type Culture Collection (Manassas, VA, USA), and the human PCa cell line LNCaP and the 293 cell line were purchased from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). The PC3 cells were maintained in Dulbecco's modified Eagle's medium (DMEM)/F12 (HyClone; GE Healthcare Life Sciences, Logan, UT, USA) supplemented with 10% (v/v) fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc.). LNCaP cells were cultured in RPMI-1640 (Gibco; Thermo Fisher Scientific, Inc.) with 10% (v/v) FBS, and 293 was cultured in DMEM supplemented with 10% (v/v) FBS, 1% (v/v) penicillin and streptomycin. All cultures were cultured in an incubator set to 37°C with 5% CO₂. Then, 0.5 µg knockdown synthetic oligonucleotides (pre-133a, pre-133b, pre-455, pre-455 inhibitor and pre-control) were transfected when cells reached 80–90% confluence, using Lipofectamine 3000 reagent (Invitrogen; Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol.

PC3 and LNCaP cells at the log growth phase were collected for DNA isolation. Genomic DNA was extracted using MiniBEST Universal Genomic DNA Extraction kit (Takara Bio, Inc., Otsu, Japan), following the manufacturer's protocol. The SNP-containing region was amplified with Premix Taq (Takara Bio, Inc.). Sequencing of the PCR product was performed using the BigDye Terminator Reaction Chemistry v3.1 sequencing kit (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol, and sequence analyses were performed with DNASTAR 7.0 (DNASTAR, Madison, WI, USA).

After 48 h of transfection, the PC3 and LNCaP cells were harvested for RNA isolation. Total RNA was extracted using the mirVana^™ miRNA Isolation kit (Ambion; Thermo Fisher Scientific, Inc.), following the manufacturer's protocol. Total RNA from PC3 and LNCaP cells was used to synthesize cDNA using the PrimeScript^™ RT Master Mix (Takara Bio, Inc.). The PCR reaction conditions were as follows: 37°C for 15 min and 85°C for 5 sec. RT-qPCR was performed in triplicate using PowerUp™ SYBR^®-Green Master mix with a 7500 Real-Time PCR System (both form Applied Biosystems; Thermo Fisher Scientific, Inc.). The RT-qPCR reaction conditions were as follows: 50°C for 5 min; 95°C for 2 min; and 40 cycles of 95°C for 15 sec and 60°C for 1 min. The primers used are listed in Table II. The relative mRNA expression in the control and experimental groups of PC3 and LNCaP was determined using the 2^−ΔΔCq method (25), and normalized to GAPDH or β-actin mRNA expression levels.

Post-transfected cells (PC3 and LNCaP) were collected for protein isolation following 48 h of incubation. Total protein was extracted with radioimmunoprecipitation assay buffer (Beyotime Institute of Biotechnology, Beijing, China) with Halt Protease Inhibitor Cocktail (100X; Thermo fisher Scientific, Inc.) and quantified with the bicinchoninic acid assay, and 50 µg of protein/lane was separated on a 12% SDS-PAGE gel and transferred to a 0.2-µm polyvinylidene fluoride membrane (EMD Millipore, Billerica, MA, USA). Blots were blocked with 5% milk, TBS and 0.1% Tween-20 at room temperature for 2 h. Blots were incubated with the primary anti-mouse IGF-IR antibody (cat. no., 3G5C1; 1:2,000; Novus Biologicals, Ltd., Cambridge, UK) or tubulin (cat. no., 66240-1-lg; 1:2,000; ProteinTech Group, Inc., Chicago, IL, USA) at 4°C for 12 h. Subsequently, blots were incubated with the horseradish peroxidase-conjugated Affinipure Goat Anti-Mouse IgG (H+L) second antibody (cat. no., SA00001-1; 1:10,000; ProteinTech Group, Inc.) at room temperature for 2 h. BeyoECL Plus (Beyotime Institute of Biotechnology) was used as the visualization reagent. The intensity of the protein bands was quantified using Odyssey 2.1 software (LI-COR Biosciences, Lincoln, NE, USA).

Transfected LNCaP cells (pre-455, pre-455 inhibitor and pre-control) growing in the log phase were harvested for use in the migration and invasion assays. A total of 4×10⁵ cells in serum-free RPMI-1640 (200 µl) were seeded into the upper chamber of 8-µM pore Transwell plates (Costar; Corning Incorporated; Corning, NY, USA), coated with or without Matrigel (BD Biosciences, San Jose, CA, USA). RPMI-1640 (600 µl) filled with 10% (v/v) FBS was added to the lower chamber. The cells were allowed to migrate towards RPMI-1640 containing 10% (v/v) FBS for 48 or 72 h at 37°C. Cells on the top surface of the insert were removed by scraping, and the migrated and invasive cells were fixed for 10 min using 100% methanol and stained for 20 min with 1% crystal violet (both at room temperature). Subsequently, cells were imaged at ×20 magnification using an inverted phase contrast microscope (TS100-F; Nikon Corporation, Tokyo, Japan), removed the crystal violet dye retained on the filters using 33% acetic acid (600 µl) and then transferred the solution containing crystal violet to 96-well plates (150 µl/well). Values for migration and invasion were obtained by measuring absorbance of the solution at 570 nm using an ELISA reader (BioTek ELx-800; BioTek Instruments, Inc., Winooski, VT, USA) (26), and presented as the mean of at least three independent experiments.

The figures were generated with GraphPad Prism 5.0 (GraphPad Software, Inc., La Jolla, CA, USA). Data are presented as the mean ± standard error the mean and were compared using a Student's t-test (two-tailed). P<0.05 was considered to indicate a statistically significant difference.

The alleles of rs1815009 and rs2684788 in the IGF1R 3′UTR in the present study were primarily based on GWAS data from ChinaPCa. In our previous study, rs1815009 and rs2684788 were identified to be significantly associated with PCa using the same dataset (18). Among the PCa cases, rs1815009 was identified to be a common variant, with a frequency of ~52.12% for C the allele and 47.88% for the T allele, and healthy individuals exhibited a common variant with a frequency of ~48.57% for the C allele and 51.43% for the T allele. In addition, among the PCa cases, rs2684788 was demonstrated to have a frequency of ~52.23% for the C allele and 47.77% for the T the allele, and the healthy population exhibited a common variant with a frequency of ~47.43% for the C allele and 52.57% for the T allele. The 3′UTR was demonstrated to be an important region for miRNA binding to genes. To clarify the function of these two polymorphisms on miRNA binding, the target miRNA sequence was predicted using TargetScan and miRanda. By combining the results obtained from TargetScan (19) and miRanda (20), miR-133a, miR-133b and miR-455 were predicted to bind to the IGF1R 3′UTR (Fig. 2A). Furthermore, rs1815009 was located near the binding seed regions of miR-133a and miR-133b, and rs2684788 was located near the binding seed region of miR-455 (Fig. 2B).

In order to analyze the binding affinity of miR-133a, miR-133b and miR-455 to the IGF1R 3′UTR of different variants (rs1815009 and rs2684788), a parameter-free thermodynamic model was generated (21). The free energy of secondary structure-based molecules was calculated for each stage of the binding process: E_Target represents the energy of the dissociated mRNA target region; E_Intermediate is the energy required to make the target region accessible for miRNA binding; and E_Complex is the energy of the miRNA-target complex (6,7,23). The activation energy (ΔE_a) was calculated as the difference between E_Intermediate and E_Target, whereas the binding energy (ΔE_b) was equivalent to the interaction score, which is the difference between E_Complex and E_Target (6,7,23). Compared with the C allele, the T allele of rs1815009 exhibited a higher E_target and lower ΔE_a, indicating that the binding affinity of miR-133a for the T allele was stronger compared with that of the C allele. The same phenomenon was observed for miR-133b. A higher E_target and lower ΔE_a was identified for the rs2684788-IGF1R C allele, suggesting that this allele was more accessible for miR-455 binding. Furthermore, the ΔE_b of miR-133a or miR-133b to the rs1815009-IGF1R C allele was increased compared with that of the T allele, suggesting that miR-133a and miR-133b exhibited a higher binding affinity for the C allele. Regarding the other SNP, the ΔE_b of the rs2684788-IGF1R C allele was reduced compared with that for the T allele, suggesting that miR-455 exhibited a higher binding affinity for the T allele (Fig. 3).

According to the inclusion criteria, four datasets [GSE36803 (27), GSE8126 (28), GSE23022 (29) and TCGA datasets] were selected to evaluate the miRNA expression in PCa. However, the GSE23022 dataset is primarily derived from moderately differentiated PCa samples, unlike the other three datasets which are from primary PCa samples, therefore, this dataset was removed. In total, 579 carcinoma samples and 89 normal samples were included in the present study. With the exception of miR-133b in the GSE8126 dataset, the expression levels of all three miRNAs (miR-133a, miR-133b and miR-455) in PCa tissues were significantly lower compared with that of normal tissues in the three datasets (Fig. 4). This result indicated that miR-133a, miR-133b and miR-455 are likely to be dysregulated in PCa.

An in vitro luciferase reporter assay was performed to verify whether the two SNPs, rs1815009 and rs2684788, are able to affect miRNA binding. For each SNP, pmirGLO constructs were generated carrying the C or T allele for rs1815009 and rs2684788. These plasmids were designated as pmirGLO-rs1815009-IGF1R-[C] and pmirGLO-rs1815009-IGF1R-[T] for rs1815009, and pmirGLO-rs2684788-IGF1R-[C] and pmirGLO-rs2684788-IGF1R-[C] for rs2684788 (Fig. 2). Each of the pmirGLO constructs was analyzed for luciferase activity levels in 293 cells that overexpressed the predicted interacting miRNAs or the scrambled negative control. When miR-133a or the scrambled negative control was co-transfected with pmirGLO-rs1815009-IGF1R constructs, the luciferase activity for the C allele was significantly increased compared with the T allele (Fig. 5A), suggesting a stabilizing rather than inhibitory role, resulting in stronger binding of miR-133a to the rs1815009-IGF1R C allele. Conversely, according to the parameter-free thermodynamic model, it was predicted that the rs1815009-IGF1R C allele may exhibit a higher binding affinity and decreased miR-133b binding energy compared with the T allele, which was in agreement with the results of the luciferase assay. These results suggest a significantly higher suppressive effect of miR-133b when interacting with the construct containing the C allele (Fig. 5B). Furthermore, the luciferase assay demonstrated a statistically significant suppressive effect of miR-455 on the construct carrying the T allele (Fig. 5C), indicating that miR-455 had a primary binding efficacy for the rs2684788-IGF1R C allele. Overall, the results of the in vitro luciferase reporter assays were consistent with the MFE prediction.

To determine the effect of rs1815009 and rs2684788 on the mRNA and protein expression levels of endogenous IGF1R in vitro, PCa cell lines (PC3 and LNCaP) were transfected with miRNA precursors of the interacting miRNAs. RT-qPCR analysis demonstrated that, with the exception of overexpression of miR-133a in PC3 cells, forced expression of the interacting miRNAs significantly suppressed IGF1R expression at the transcriptional level in PCa cells (PC3 and LNCaP) compared with the miR-control (Fig. 6). In addition, western blot analysis demonstrated that miR-133b markedly downregulated the expression of the IGF1R protein in LNCaP cells containing the rs2684788-IGF1R C allele (Fig. 6D, right). A similar suppressive effect was observed in PC3 cells that overexpressed miR-455 (Fig. 6E, right). Taken together, these results suggest that miR-133b directly regulates IGF1R gene expression at the mRNA and protein levels in LNCaP cells, whereas miR-133a regulates IGF1R gene expression at the mRNA level in LNCaP cells, but not in PC3 cells. Therefore, the rs2684788 SNP exhibited a functional effect on the miR-455 binding efficacy in PC3 cells.

In order to analyze the effect of miR-133a, miR-133b and miR-455 on the expression of genes associated with the biological activity of tumors, including cell invasion, migration and apoptosis, miRNA precursors (pre-133a, pre-133b, pre-455 or scrambled negative control) were transfected into PC3 and LNCaP cell lines. Expression levels were determined 48 h after transfection. Matrix metalloproteinase 2 (MMP-2) in transfected LNCaP cells was demonstrated to be significantly suppressed by miR-133a, miR-133b and miR-455, and was suppressed by miR-455 in PC3 cells compared with the miR-control. E-cadherin (CDH1) was significantly upregulated in LNCaP cells that overexpressed miR-133a, and in PC3 cells that overexpressed miR-455, but exhibited mild downregulation in LNCaP cells that overexpressed miR-455 compared with the miR-control. Forced expression of miR-133b and miR-455 in LNCaP cells resulted in mild suppression of vascular endothelial growth factor A (VEGFA), but the overexpression of miR-133a in LNCaP cells led to a more significant downregulation at the transcriptional level. B-cell lymphoma-2 (BCL-2) was significant upregulated in LNCaP cells with forced expression of miR-133a, but a suppressive effect was observed in post-transfected LNCaP cells that overexpressed miR-133b compared with the miR-control. These results are presented in Fig. 7.

To analyze the effect of miR-455 on prostate cells migration and invasion, post-transfected LNCaP cells were harvested to a Transwell assay was performed. After 48 h incubation, the migration rate was significantly increased in LNCaP cells with miR-455 overexpression, and significantly suppressed in cells that were treated with miR-455 inhibitor compared with the miR-control (Fig. 8A). Similarly, after 72 h incubation, the invasion rate in LNCaP cells with miR-455 overexpression was significantly increased, and significantly decreased in LNCaP cells treated with miR-455 inhibitor compared with the miR-control (Fig. 8B).