A novel intronic circular RNA circFGFR1int2 up-regulates FGFR1 by recruiting transcriptional activators P65/FUS and suppressing miR-4687-5p to promote prostate cancer progression

Figures

• 

  Fig. 1. Discovery and characterization of circFGFR1int2. A Three of the FGFR1-derived circRNAs (hsa_circ_0002352, hsa_circ_0008016, hsa_circ_0005564) recorded in CircBase (n = 17) and two newly identified FGFR1-derived circRNAs by the present study (designated circFGFR1E2/E4/E5 and circFGFR1int2) were expressed in PCa cells. circFGFR1int2 was the most abundantly expressed and most significantly differentially expressed between PCa cells (PC-3, DU145, LNCaP, and 22Rv1) and normal prostate epithelium cell (RWPE-1). B Further characterization of the novel 875-nt circFGFR1int2 by RT-PCR and PCR-sequencing using divergent primers (circ-div-p) spanning splicing junction and convergent primers (circ-con-p). The circFGFR1int2 could be amplified from complimentary DNA (cDNA) obtained by reverse transcription, but not from genomic DNA (gDNA). GAPDH and globin were internal controls for PCR. C circFGFR1int2 was resistant to RNase digestion, as compared to the linear RNA internal control (GAPDH). D Cell fractionation showed that circFGFR1int2 was distributed in both the cytoplasm (Cyto) and the nucleus (Nuc) fractions, whereas FGFR1 mRNA was located in the cytoplasm. ISH (in situ hybridization)…

• 

  Fig. 2. CircFGFR1int2 promoted expression of FGFR1 mRNA and protein. A CircFGFR1int2 and FGFR1 mRNA levels were significantly higher in PC-3, DU145, LNCaP, and 22Rv1 PCa cells than in normal prostate RWPE-1 cells. B CircFGFR1int2 and FGFR1 mRNA were significantly up-regulated in PCa (n = 6) than in BPH (n = 5) tissue samples. C Expression of circFGFR1int2 and FGFR1 mRNA in PCa cells and tissues was significantly correlated. D–F Knockdown of circFGFR1int2 by ASO-circFGFR1int2 significantly reduced expression of FGFR1 mRNA (D, RT-PCR; E, qRT-PCR) and protein (F, Western blot), whereas artificial overexpression by OE-circFGFR1int2 rescued FGFR1 expression. Error bars for RT-PCR, qRT-PCR, and Western blot represented mean ± standard deviation (SD) of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001

• 

  Fig. 3. Analysis and identification of circFGFR1int2-interacting proteins. A Analysis by catRAPID, circAtlas, RBPDB and RBPmap databases revealed potential circFGFR1int2-binding proteins and enriched GO terms and KEGG pathways. B Analysis by Metascape database (http://metascape.org/gp/index.html#/main/step1) showed potential protein–protein interactions of the circFGFR1int2-interacting proteins and the biological processes they may be involved in. C The FUS protein was identified as a circFGFR1int2-interacting protein by all four databases. D RNA immunoprecipitation (RIP) showed retrieval of circFGFR1int2 from complexes obtained with both anti-P65 and anti-FUS. Knockdown of circFGFR1int2 by ASO-circFGFR1int2 decreased recovery of circFGFR1int2. Non-immune-IgG was used as negative control. E FUS and P65 were pulled down by biotin-labeled circFGFR1int2 probe, knockdown of which decreased the recovery. F Co-immunoprecipitation (Co-IP) showed retrieval of P65 by anti-FUS, and FUS by anti-P65, respectively. Non-immune-IgG was used as negative control. G Expression of P65 and FGFR1 in PCa tissue samples was significantly correlated (TCGA, n = 492, R = 0.29, P < 0.001)

• 

  Fig. 4. P65/FUS promoted FGFR1 transcription. A–C Knockdown of P65 or FUS by siRNAs significantly decreased FGFR1 mRNA and protein levels. D Predicted P65 binding sequence at -609 ~ -619 upstream of transcription start site by analysis with PROMO (https://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3) and JASPAR (https://jaspar.genereg.net/). E The predicted P65 binding sequence 5′-GACGTTCCCTA-3′ (in red box) was conserved across species (sequence alignment by UCSC). F Chromatin immunoprecipitation (ChIP) with anti-FUS or anti-P65 showed retrieval of FGFR1 promoter sequences containing the P65 binding site (right panels represented qRT-PCR analysis). Input and nonimmune IgG were used for control. G Dual-luciferase reporter assays showed that FUS or P65 knockdown by siRNAs significantly decreased the FGFR1 promoter activity. Error bars for qRT-PCR, Western blot, ChIP, and Dual-luciferase reporter assays represented mean ± standard deviation (SD) of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001

• 

  Fig. 5. CircFGFR1int2 facilitated FGFR1 transcription by recruiting FUS/P65. A Bioinformatics analyses by RNAhybrid showed multiple potential circFGFR1int2 binding sites (green boxes) enriched around P65 binding site (dotted line box) in the FGFR1 promoter. B Chromatin isolation by RNA purification (ChIRP) followed by Western blot analysis and PCR revealed that FUS, P65, circFGFR1int2, and FGFR1 promoter sequences containing the P65 binding site were enriched in the complexes obtained with biotin-labelled-circFGFR1int2 probes, but not by random-sequence control RNA probe. Knockdown of circFGFR1int2 by ASO-circFGFR1int2 decreased recovery of the above components. C ChIP with anti-FUS or anti-P65 showed that knockdown of circFGFR1int2 by ASO-circFGFR1int2 significantly reduced retrieval of FGFR1 promoter sequence containing the P65 binding site (lower represented qRT-PCR analysis). D Upper: dual-luciferase reporter constructs with wildtype FGFR1 promoter containing P65 binding site (pGL3-P691), or truncated sequence (pGL3-P500) or mutated binding sites (pGL3-P691-P65 MUT, pGL3-P691-circFGFR1int2 MUT). Lower: dual-luciferase reporter assays showed significant FGFR1 promoter activity…

• 

  Fig. 6. miR-4687-5p was downregulated in PCa and interacted with circFGFR1int2. A Bioinformatics analysis (RegRNA 2.0 and RNA22 v2) of hsa-miR-4687-5p showing the potential circFGFR1int2-interacting sequence TGTGGGGTGAGGGCT (in red) at 762 to 779. B, C miR-4687-5p was significantly down-regulated in PCa cells (LNCap, PC-3, and DU145) and tissue samples (n = 6) than in normal prostate cell RWPE-1 and BPH tissues. D Expression of FGFR1 mRNA was negatively correlated with that of miR-4687-5p. E Dot blot hybridization showed biotin-labelled miR-4687-5p probe bind to wild-type circFGFR1int2 (WT) in dose-dependent manner, whereas mutation of the binding site (MUT) resulted in no or very weak signals. Blank (no RNA was added) was used as negative control. Error bars for qRT-PCR represented mean ± standard deviation (SD) of three independent experiments. **P < 0.01, ***P < 0.001, ns, not significant

• 

  Fig. 7. miR-4687-5p was a novel suppressor of FGFR1 by targeting FGFR1 3′UTR and CDS. A Bioinformatics analysis by TargetScan, miRwalk, Microt4, miRanda and RNAhybrid databases revealed two potential miR-4687-5p-interacting sites located in FGFR1 3′UTR and CDS, respectively. B The miR-4687-5p binding sequences in 3′UTR (5′-AGGGCU-3′, red) and CDS (5′-GAGGGCTG-3′, blue) were highly conserved across species. C Artificial overexpression of miR-4687-5p by mimics significantly suppressed expression of FGFR1 mRNA and protein. D Dot blot hybridization showing biotin-labelled miR-4687-5p probe bound to wild-type FGFR1 3′UTR and CDS fragments in a dose-dependent manner, while mutation of miR-4687-5p binding sites in FGFR1 mRNA (FGFR1 3′UTR-MUT and FGFR1 CDS-MUT) resulted in no or signals. Error bars for qRT-PCR represented mean ± standard deviation (SD) of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001

• 

  Fig. 8. CircFGFR1int2 suppressed the inhibitory effects of miR-4687-5p on FGFR1. A Artificial overexpression of miR-4687-5p by miR-4687-5p mimic significantly decreased expressions of FGFR1 mRNA and protein in PCa cells, which could be reversed by circFGFR1int2 artificial overexpression (OE-circFGFR1int2). B CircFGFR1int2 and miR-4687-5p were simultaneously precipitated by anti-AGO2 in PCa cells. Artificial overexpression of miR-4687-5p or knockdown of circFGFR1int2 decreased the recovery of circFGFR1int2. C Artificial overexpression of miR-4687-5p increased AGO2 pull down by biotin-labeled miR-4687-5p probe, whereas knockdown of circFGFR1int2 decreased AGO2 retrieval. D Artificial overexpression of miR-4687-5p mimic significantly suppressed the promoter activity of the WT constructs containing the miR-4687-5p binding sited (pGL3-FGFR1 3′UTR-WT, pMIR-FGFR1 CDS-WT, and pGL3-circFGFR1int2-WT), which could be rescued by the respective mutant plasmids (pGL3-FGFR1 3′UTR-MUT, pMIR-FGFR1 CDS-MUT and pGL3-circFGFR1int2-MUT). Error bars for qRT-PCR and Dual-luciferase reporter assays represented mean ± standard deviation (SD) of three independent experiments. **P < 0.01, ***P < 0.001

• 

  Fig. 9. Upregulated circFGFR1int2 and downregulated miR-4687-5p promoted PCa cell proliferation, migration, and invasion. A–C Knockdown of circFGFR1int2 by ASO-circFGFR1int2 or artificial overexpression of miR-4687-5p by mimic significantly decreased PCa cell proliferation and DNA replication as shown by CCK-8 assays and EdU incorporation assays, which could be reversed by artificial overexpression of circFGFR1int2. D Knockdown of circFGFR1int2 by ASO-circFGFR1int2 or artificial overexpression of miR-4687-5p by mimic significantly decreased PCa cell migration and invasion as shown by wound healing, which could be reversed by artificial overexpression of circFGFR1int2. E Similar effects as shown by Transwell assays. Three microscopic fields were evaluated for each sample. Error bars for CCK-8, EdU incorporation, wound healing and Transwell assays represented mean ± standard deviation (SD) of three independent experiments. **P < 0.01, ***P < 0.001

• 

  Fig. 10. Prognostic significance of circFGFR1int2 overexpression in PCa and schematic summary of the major findings of the present study. A Survival analysis by Kaplan–Meier method with log-rank test revealed that disease-specific survival (DSS) and castration-resistance free survival (CFS) of PCa patients were significantly shorter when CircFGFR1int2 ≥ 4, Gleason score ≥ 8, and PSA ≥ 50 ng/ml. B Schematic summary of the major findings of the present study