Circular RNA circFOXO3 promotes prostate cancer progression through sponging miR‐29a‐3p

Abstract Circular RNA FOXO3 (CircFOXO3, also termed as Hsa_circ_0006404) is derived from exon 2 of forkhead box O3 (FOXO3) gene, and abnormal expression is shown in different diseases. However, whether circFOXO3 plays important roles in tumorigenesis and progression of prostate cancer (PCa) remains unclear. In this study, we found that circFOXO3 was up‐regulated in both PCa tissues and serum samples. Moreover, circFOXO3 was positively correlated with the Gleason score in PCa samples. CircFOXO3 was observed to be up‐regulated in Gleason score > 6 PCa samples compared with Gleason score = 6 PCa samples. Knock‐down circFOXO3 could remarkably inhibit PCa cell cycle, proliferation and promote cell apoptosis in vitro. Furthermore, we demonstrated circFOXO3 could act as miR‐29a‐3p sponge to up‐regulate SLC25A15 expression by bioinformatics analysis, dual‐luciferase reporter assays and biotinylated RNA pull‐down assays. SLC25A15 could reverse the tumour suppressing roles of knock‐down circFOXO3 in PCa. Of note, we found that miR‐29a‐3p was down‐regulated; however, SLC25A15 was overexpressed in PCa samples compared with normal tissues. In conclusion, circFOXO3 acts as a miR‐29a‐3p sponge to exhibit oncogenic activity that affects the cell cycle and cell apoptosis in PCa through transcriptional up‐regulation of SLC25A15. Our analysis suggests circFOXO3 could act as promising prostate cancer biomarkers.

Recent reports have showed that 80%-90% RNA molecules are non-coding RNAs (ncRNAs). 5 Circular RNAs (circRNAs), a newly emerging endogenous ncRNA, are originated from its parental linear genes by RNA polymerase II and harbour covalently closed circular structure without poly(A) tail and 5′-3′ polarity. 6,7 Studies have exhibited that circRNAs are conserved, stable and stage-/tissue-specific expression. 8,9 Recently, scholars paid more attention on circRNAs and identified that lots of circRNAs were significantly differential expression in various cancers, suggested circRNAs might have crucial effect on cancer development. 10 For example, our group identified circSMARCA5 was dysregulated and promoted PCa cell growth. 11 Furthermore, accumulating studies have shown that circRNAs directly combine with miRNAs as 'miRNA sponges' and regulate their target genes expression and cancer progression. For example, circMTO1 suppresses hepatocellular carcinoma progression through regulating p21-mediated proliferation and invasion by sponging miR- 9. 12 circFOXO3 is derived from exon 2 of the FOXO3 gene and contains 1435 nucleotides. Furthermore, Burton Yang et al 13 14 However, the function and mechanism of circFOXO3 in PCa remain unclear.
In this study, we discovered the expression of circFOXO3 was highly expressed in PCa tissue samples and serum samples than controls. Therefore, we knock down circFOXO3 expression to identify its potential roles and explore possible mechanisms in carcinogenesis of PCa. Here, we demonstrated that circFOXO3 acted as a miR-29a-3p sponge to up-regulated solute carrier family 25 member 15 (SLC25A15) and played an oncogenic role in PCa.  Table 1 and Table S2. The study was approved by the Research Ethics Committee of Fudan University Shanghai Cancer Center. Informed consent was provided by all patients. All samples were collected and used for gene expression analysis by qRT-PCR.

| Cell culture
All cell lines (WPMY-1, LNCaP, 22Rv1, DU145 and PC-3) were kindly provided by Stem Cell Bank, Chinese Academy of Sciences. WPMY-1 (the human normal prostate epithelial cell) was expanded in DMEM medium (HyClone), and the PCa cell lines (LNCaP, 22Rv1, DU145 and PC-3) were maintained in RPMI 1640 medium (HyClone). All of the medium should be added 10% foetal bovine serum before use (FBS, Biological Industries). Then, all cell lines were maintained at 37°C, 5% CO 2 incubator.

| Cell proliferation assay
Cell proliferation was assessed using the CCK-8 (Dojindo) as our previous report. 11 In brief, transfected cells were maintained in 96well plates at a density of 5000 cells per well and at 0, 24, 48 and 72 hours post-treatment, 10 μL CCK-8 was added to each well and then incubated for 2 hours at 37°C. The optical density was measured at 450 nm by Microplate Reader ELx808 (BioTek).

| Annexin V-FITC apoptosis detection
PC-3, LNCaP-AI and DU145 cells were maintained in 6-well plates and transfected with sicircFOXO3 or NC by HilyMax transfection reagents.
At 48 hours after transfection, cells were treated using the FITC-Annexin V Apoptosis Detection Kit (Dojindo) for 15 minutes at room temperature.
The cell apoptosis was measured on FACSCalibur flow cytometer (BD).

| Biotinylated RNA pull-down assays
Biotin-coupled circFOXO3 probe was designed according to the junction of circFOXO3 (Table S1). Cellular protein was extracted Then, 30 μL streptavidin-conjugated magnetic beads (11205D, Invitrogen) were added into the cell lysate and incubated at 4°C for 1 hour. The retrieved RNA was detected by qRT-PCR assay as described above.

| Statistical analysis
All data are reported as the mean ± standard deviation (SD) and represent average of at least three independent experiments. Statistical comparisons between two groups are carried out using Student's two-tailed unpaired t test, and P < .05 is considered statistically significant. Associations between circFOXO3 expression levels and the clinicopathologic features of PCa patients are summarized in Table 2.

| Expression pattern of circFOXO3 in PCa patients and cell lines
Higher expression levels of circFOXO3 were significantly correlated with the Gleason score ( Figure 1E and Table 2), but were not significantly associated with age and preoperative PSA level (Table 2).
CircFOXO3 was observed to be overexpressed in Gleason score > 6 PCa samples compared with Gleason score = 6 PCa samples. The expression of circFOXO3 was quantified by qRT-PCR with divergent primers (Table S1) after treatment with actinomycin D, a transcription inhibitor, by qRT-PCR. The result showed that circFOXO3 was much more stable than FOXO3 mRNA ( Figure 1H). Furthermore, we performed subcellular fractionation and detected the cellular localization of circFOXO3 by qRT-PCR in LNCaP, LNCaP-AI, PC-3 and DU145 cells. The result revealed that circFOXO3 was predominantly cytoplasmic ( Figure 1I).
Taken together, our results show that circFOXO3 is a stable, predominantly cytoplasmic circRNA and up-regulated in PCa.

| Knock-down of circFOXO3 inhibited proliferation and promoted apoptosis of PCa cells
To investigate the biological roles of circFOXO3 in PCa, we per-

| circFOXO3 knock-down affected cell cycle progression of PCa cells
Next, we detected the effect of circFOXO3 on the PCa cell cycle pro-  Figure 3). Taken together, these results indicate that the effect of silencing circFOXO3 on PCa cell proliferation could be attributed to its promotion of apoptosis and cell cycle arrest.

| Identification of circFOXO3 mediated ceRNA and confirmation of the sponging effect between circFOXO3 and miR-29a-3p
As mentioned in previous reports, natural circRNAs could function as efficient microRNA sponges to regulate protein-coding genes. 12 To explore potential mechanism of circFOXO3 regulating PCa progression, we constructed circFOXO3-mediated competing endogenous RNA (ceRNA) networks. By using RegRNA 2.0 database, we identified 14 miRNAs targeted to circFOXO3 ( Figure 4A-B).
Except as RNA sponge, mounting evidence suggested circRNAs could sequester proteins. 16 Here, we constructed a circFOXO3 interaction protein network. By using RBPDB database, eight RNA binding proteins (MBNL1, NONO, SFRS9, RBM4, SFRS1, FUS, EIF4B and RBMX) with relative score ≥ 0.99 were identified to interact with circFOXO3 ( Figure S1A). Next, we analysed the interaction proteins of these RNA binding proteins got from NCBI interaction protein database. As shown in Figure S1A,  splicing, mRNA processing, mRNA transport and mRNA catabolism) and cell proliferation (including cell cycle, anti-apoptosis, DNA replication, cell cycle arrest, DNA repair and mitosis) ( Figure S1B).
To test the hypothesis that circFOXO3 promoted PCa proliferation by regulating these miRNAs, we analysed these miRNA expression patterns in public data GSE21036, 17   were up-regulated in PCa ( Figure 4B-C). These results are consistent with our conclusion that circFOXO3 may act as an oncogenic circRNA in PCa.
In order to validate our ceRNA network analysis, we selected three of the most significantly down-regulated miRNAs (including

| The expression of miR-29a-3p was downregulated in PCa
Our previous studies had showed that miR-29a-3p played as a tumour suppressor in PCa. 18 However, the expression pattern of miR-29a-3p in PCa remained unclear. Therefore, we detected the expression of miR-29a-3p in PCa tissue. The result showed that miR-29a-3p was lowly expressed in PCa tissue sample compared with corresponding adjacent normal prostate tissues ( Figure 6A-B). However, we did not observe a significant correlation between the miR-29a-3p expression F I G U R E 4 CircFOXO3 acted as a miR-29a-3p sponge. A, Bioinformatics methods paragraph to describe the procedure of bioinformatics analysis. B, Intersections of circFOXO3 targeted miRNA. The red and green circles separately indicated up-and down-regulated miRNAs. C, Fourteen miRNAs were differently expressed in PCa samples using GSE21036. D, qRT-PCR analysis of circFOXO3 expression transfected with four tumour-suppressive miRNAs (including hsa-miR-143-3p, hsa-miR-23a, hsa-miR-29a-3p and hsa-miR-221-5p). E, Schematic illustration indicating the wild-type or mutant target site of circFOXO3 and base pairing of miR-29a-3p. F, Luciferase assay in DU145 cells.
Overexpression of miR-29a-3p decreased the luciferase activity of circFOXO3, while miR-29a-3p had no significant effect on luciferase activity of circFOXO3 with the mutated target site. G, Schematic illustration indicating the wild-type or mutant target site of circFOXO3 and base pairing of miR-221-5p. H, Luciferase assay in DU145 cells. Overexpression of miR-221-5p had no significant effect on luciferase activity of circFOXO3 with wild-type or mutated target site. I, miR-29a-3p was pulled down and enriched with circFOXO3 probe and then detected by qRT-PCR. Data are presented as the mean ± SD (n = 3). Significance is defined as P < .05 (*P < .05; **P < .01; ***P < .001) | 809 KONG et al.
and Gleason score, which may be due to the limited sample size ( Figure 6C). Very interestingly, we observed a significantly negative correlation between miR-29a-3p and circFOXO3 or SLC25A15 ( Figure 6D-E), suggested the existence of circFOXO3/miR-29a-3p/ SLC25A15 in PCa samples.

| SLC25A15 was up-regulated in PCa tissues
We compared SLC25A15 expression levels in 53 PCa tissue samples and corresponding adjacent normal prostate tissues by qRT-PCR.
Our analysis revealed that SLC25A15 mRNA expression was upregulated in PCa samples compared with corresponding adjacent normal prostate tissues ( Figure 7A-B). However, we did not observe a significant correlation between the SLC25A15 expression and Gleason score ( Figure 7C). To further compare SLC25A15 protein levels in PCa and normal tissues, we analysed SLC25A15 expression in Human Protein Atlas (https ://www.prote inatl as.org/). We also observed SLC25A15 was overexpressed in PCa samples compared with normal samples ( Figure 7D). More importantly, TCGA data set analysis revealed SLC25A15 was up-regulated in PCa samples ( Figure 7E) and increased SLC25A15 expression in PCa tissues was significantly correlated with shorter 5-year overall survival time of PCa patients, shown as Kaplan-Meier survival curve ( Figure 7F). Together, these data suggest that SLC25A15 is up-regulated in PCa tissues. Accumulating studies showed that circFOXO3 was involved in progression of diseases, such as cardiac senescence 21 and nonsmall-cell lung cancer. 22 Moreover, Burton Yang et al reported circFOXO3 interacted with p21 and CDK2, and retarded cell cycle progression. 15 They also detected that circFOXO3 induced tumour apoptosis though enhancing FOXO3 activity. 13 However, whether circFOXO3 plays a role in PCa remains unclear. Therefore, we knock down circFOXO3 and then explored the potential biological function in PCa development. Intriguingly, our data are not consistent with previous studies. In this study, we found circFOXO3 silencing significantly inhibited the growth of PCa cells by affecting cell cycle progression (LNCaP-AI and DU145 cells) and cell apoptosis.

| D ISCUSS I ON
These results suggested that circFOXO3 played an oncogenic role in the development of PCa. Importantly, this is the first study to investigate the molecular function and mechanism of circFOXO3 in PCa.
To explore the potential mechanisms of circFOXO3 regulating PCa progression, we constructed a circRNA-mediated ceRNA network in PCa by using bioinformatics analysis. Our results showed circFOXO3 may act as a miRNA sponge to bind miRNAs.
Interestingly, more than 71 per cent of these miRNAs (including  27,28 In our previous study, we also found hsa-miR-29a-3p suppressed cell proliferation and induced apoptosis in PCa. 18  Overexpressing SLC25A15 significantly suppressed cell apoptosis; however, this effect was significantly abrogated by co-transfection with sicircFOXO3. Data are presented as the mean ± SD (n = 3). Significance is defined as P < .05 (*P < .05; **P < .01; ***P < .001) and shared the similar seed region with miR-29b/c. Despite this study focused on the effect of circFOXO3 on miR-29a, we could reasonably hypothesize that circFOXO3 could also sponge miR-29b/c. miR-29b is a miRNA that regulates both osteoblastogenesis and osteoclastogenesis. 29 Furthermore, serum miR-29b is down-regulated in PCa patients and this increases the formation of osteolytic lesions, influencing osteoblast and osteoclast differentiation and activities. 30 These results suggested that overexpression in serum of circFOXO3 could influence osteolytic lesions formation through these processes in PCa. However, the further validation is still needed.
SLC25A15 transports ornithine across the inner membrane of mitochondria from the cytosol to the mitochondrial matrix and plays an important role in regulating the urea cycle. 31 However, the roles of SLC25A15 in cancers remain largely unclear. In this study, we found SLC25A15 was a downstream target of circFOXO3 and up-regulated in PCa samples compared with corresponding adjacent normal prostate tissues. Public data sets analysis also showed SLC25A15 mRNA, and protein levels were overexpressed in PCa samples. More important, we found highly expressed SLC25A15 in PCa tissues was significantly associated with poor prognosis in patients with PCa. We consider this study provides useful information for exploring potential therapeutic and prognostic targets for PCa intervention.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.