Circular RNA circ_001422 promotes the progression and metastasis of osteosarcoma via the miR-195-5p/FGF2/PI3K/Akt axis

This study was approved by the Ethics Committee of the Affiliated Zhujiang Hospital of Southern Medical University (approval no. 2018-GJGBWK-002) and conducted in accordance with the Declaration of Helsinki. A total of 55 patients with OS were enrolled in the study between May 2018 and April 2020. All participants underwent diagnostic core needle biopsy using a disposable sterile biopsy instrument (Trauson Medical Instrument Corporation, China). Histological diagnosis was performed independently by two experienced pathologists. Written informed consent was obtained from all participants and, for minors, from their legal guardian. Surgically removed OS tissues and adjacent noncancerous tissues were collected from these patients before the commencement of chemotherapy, radiotherapy or immunotherapy. Morphologically normal muscle tissues that were more than 5 cm from the cancerous tissues were used as adjacent noncancerous tissues. Upon resection, tissues were immediately frozen in liquid nitrogen and stored at − 80 °C. Table 1 shows the patients’ clinical characteristics.

A human osteoblast cell line (hFOB1.19) and OS cell lines (143B, U-2 OS, MG-63, MNNG and Saos-2) were purchased from Jennio (Guangzhou, China). Osteoblastic hFOB1.19 cells were cultured in DMEM/F-12 (Gibco, USA) supplemented with 10% fetal bovine serum (FBS) (BI, Israel), 2.5 mM L-glutamine (Invitrogen, USA) and 0.3 mg/ml geneticin (Gibco, USA). OS cells were maintained in DMEM supplemented with 1% penicillin/streptomycin (Invitrogen, USA) and 10% FBS in a humidified incubator at 37 °C in 5% CO₂.

High-throughput sequencing was performed to identify circRNAs through the following steps. Total cellular RNA was isolated from 3 matched OS tissues and adjacent noncancerous tissues with an RNAiso Plus Reagent Kit (TaKaRa, Japan). Ribosomal RNA was removed with a Ribo-Zero Magnetic Kit (Epicentre, USA), whereas linear RNA was digested with RNase R (Epicentre, USA). Ribosomal RNA depletion and total RNA quality were assessed using a TapeStation 2200 system (Agilent Technologies, USA) and a Qubit RNA high-sensitivity fluorimeter (Thermo Fisher Scientific, USA). Fragmentation was carried out using divalent cations in an Ambion proprietary fragmentation buffer at 94 °C for 5 min and was followed by ethanol precipitation. The RNA fragments were resuspended in nuclease-free water and purified using Agencourt RNA Clean XP Beads (Beckman Coulter, USA). First-strand complementary DNA (cDNA) was synthesized using random hexamer primers and SuperScript II reverse transcriptase (Thermo Fisher Scientific, USA) with the following thermal cycling conditions: 25 °C for 10 min, 42 °C for 15 min, and 70 °C for 15 min. Subsequently, second-strand cDNA synthesis was performed at 16 °C for 1 h after the addition of second-strand synthesis reaction buffer, dNTPs, RNase H, and DNA polymerase I. Double-stranded cDNA was purified using 1.8× Agencourt AMPure XP Beads (Beckman Coulter, USA) and was then subjected to end repair, 3′ adenylation and adaptor ligation by using an NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs, USA). To select cDNA fragments of the preferred length (200–300 bp), the ligation reaction product was treated with USERTM enzyme at 37 °C for 15 min and purified using Agencourt AMPure XP Beads. PCR was carried out by an initial denaturation step for 30 s at 98 °C followed by 12 cycles of denaturation for 10 s at 98 °C, annealing for 75 s at 65 °C, and extension for 5 min at 65 °C. PCR products were purified with 1× Agencourt AMPure XP beads and subjected to quality control using a High Sensitivity DNA Assay Kit (Agilent Technologies, USA). After quantification and pooling in a StepOnePlus Real-Time PCR System (Applied Biosystems, USA), the cDNA libraries were sequenced on the Illumina HiSeq 2500 platform by Gene Denovo Biotechnology (Guangzhou, China) according to the manufacturer’s instructions. Raw reads in FASTQ format were analyzed and preprocessed using fastp software [17] (version 0.19.1) to remove reads containing adaptors with more than 10% unknown nucleotides or more than 50% low-quality (Q-value ≤10) bases. The remaining high-quality clean reads were aligned to the Homo sapiens ribosomal RNA database and the human GRCh38 reference genome using Bowtie2 [18] (version 2.2.8) and TopHat2 [19] (version 2.1.1). The unmapped reads were extracted and processed with find_circ [20] (version 1) for circRNA identification. To quantify the expression levels of the circRNAs, the back-spliced junction reads were scaled to reads per million mapped reads (RPM). Differentially expressed circRNAs with | fold change | ≥ 2 and P value < 0.05 were identified using the limma package (version 3.42.0) (bioconductor.​org/​packages/​release/​bioc/​html/​limma.​html).

Total RNA was extracted from cultured cells or clinical tissues using an RNAiso Plus Reagent Kit (TaKaRa, Japan) according to the manufacturer’s instructions. RNA integrity was confirmed through agarose gel electrophoresis, and RNA concentration and purity were determined with a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, USA). Final RNA quality was assessed in an Agilent 2100 Bioanalyzer (Agilent Technologies, USA), and a minimal RNA integrity number (RIN) of 8 was required. For quantitative analysis of circRNAs and mRNAs, cDNA was synthesized with a PrimeScript RT Reagent Kit (TaKaRa, Japan). Real-time amplification was conducted using SYBR Premix Ex Taq II (TaKaRa, Japan) in a LightCycler 96 System (Roche, Germany). For miRNA analysis, reverse transcription was performed using stem-loop RT primers specific for the miRNA of interest (GeneChem, China) based on an Evo M-MLV RT Kit (Accurate Biology, China), and qPCR was then performed with a SYBR Green Premix Pro Taq HS qPCR Kit (Accurate Biology, China). The following thermal cycling program was used for qPCR of all RNAs: denaturation at 95 °C for 30 s followed by 45 cycles of denaturation at 95 °C for 5 s, annealing at 55 °C for 30 s, and extension at 72 °C for 30 s. GAPDH (for circRNAs and mRNAs) and U6 (for miRNAs) were used as the endogenous controls. Relative quantification (RQ) values were calculated using the following equation: RQ = 2^−ΔΔCt, where ΔΔCt = [Ct (gene of interest, sample) - Ct (GAPDH or U6, sample)] - [Ct (gene of interest, calibrator) - Ct (GAPDH or U6, calibrator)]. All experiments were performed independently at least three times, and all samples were analyzed in triplicate. Primer sequences are provided in Additional file 1: Table S1.

Total RNA (2 μg) was extracted and digested with RNase R (3 U/μg) for 15 min at 37 °C. The control samples were processed in the same way as the experimental samples except that RNase R was not added. cDNA synthesis and real-time PCR were performed as described above. Additionally, circular and linear transcripts were amplified using specific divergent and convergent primers with or without RNase R. The PCR products amplified from cDNA or genomic DNA (gDNA) templates were added to 6× loading buffer with the nucleic acid dye GelRed (Biotium, USA). DNA fragments were separated by agarose gel (2%) electrophoresis at 100 V for 30 min and visualized with UV transillumination. For Sanger sequencing, PCR amplification products were excised from the agarose gel and purified using a GeneJET Gel Purification Kit (Thermo Fisher Scientific, USA). The nucleotide sequences of the purified fragments were determined by Sanger sequencing using standard approaches [21] by Geneseed (Guangzhou, China).

In brief, OS cells were cultured in six-well plates for 24 h, and fresh medium supplemented with 2 μg/ml actinomycin D (Sigma-Aldrich, USA) was then added. Total cellular RNA was extracted 0, 4, 8, 12 and 24 h after actinomycin D treatment for qRT-PCR analysis.

Extraction and purification of cytoplasmic and nuclear RNAs were performed with a PARIS kit (Life Technologies, USA) in accordance with the manufacturer’s protocols. Next, qRT-PCR was performed to quantify the expression of linear RNAs and circRNAs, with U6 and GAPDH as the internal references for nuclear and cytoplasmic RNAs, respectively.

RNA-FISH was performed to determine the subcellular localization of circ_001422. The Cy3-labeled circ_001422 probe was constructed by RiboBio (Guangzhou, China). Fluorescence signals were generated using a Fluorescence In Situ Hybridization Kit (RiboBio, China), and a Nikon A1 confocal laser scanning microscope (Nikon, Japan) was utilized to take pictures.

The miR-195-5p mimic/inhibitor, miR-195-5p agomir/antagomir and their corresponding negative controls (NC) were purchased from RiboBio (Guangzhou, China). For construction of knockdown plasmids expressing short hairpin RNAs (shRNAs) against circ_001422 or FGF2, the pLshRNA-NC and pLKO.1 vectors were constructed by Geneseed (Guangzhou, China), and the annealed shRNA oligonucleotides targeting circ_001422 or FGF2, respectively, were ligated into these vectors. Additionally, the full-length cDNA sequences of circ_001422 and FGF2 were PCR amplified and cloned into the pLC5-ciR vector (Geneseed, China) or pcDNA3.1 vector (Geneseed, China), respectively, to construct the circ_001422 and FGF2 overexpression plasmids. The nontargeting pLC5-ciR and pcDNA empty vectors were used as the corresponding negative controls. 293 T packaging cells were transiently transfected with the abovementioned plasmids using EndoFectin Max reagent (GeneCopoeia, USA) for production of lentiviral particles. The viral supernatants were collected and concentrated using a Lenti-X Concentrator Kit (Clontech, USA). All lentiviral particles in this study had a titer of 1 × 10⁹ transducing units/ml and were stored at − 80 °C. Finally, the concentrated lentivirus was used to transduce 143B and Saos-2 cell lines in logarithmic growth phase. At 48 h post lentiviral transduction, OS cells were selected with puromycin (2 μg/ml, Invitrogen, USA) or geneticin (500 μg/ml, Gibco, USA) for 2 weeks to obtain stable cell lines. The knockdown or overexpression efficiency was verified using qRT-PCR. Notably, the OS cells transduced with sh-circ_001422#2 or sh-FGF2#1 showed the lowest transcription level of the corresponding target gene, and were therefore used in subsequent experiments.

The viability of 143B and Saos-2 cells after transfection was assessed using 5-ethynyl-2′-deoxyuridine (EdU) incorporation and colony formation assays. An EdU Apollo 488 Kit (RiboBio, China) was utilized to conduct the EdU incorporation assay. For evaluation with a Nikon inverted fluorescence microscope, EdU-positive cells were stained green, and nuclei were stained blue.

To evaluate the colony-forming ability of cells, transfected cells were counted and seeded into 6-well plates at 550 cells/well. Cells were cultured in complete DMEM supplemented with 10% FBS at 37 °C and 5% CO₂. The culture medium was replaced at 2-day intervals. After 10 days of incubation, the colonies were washed twice with phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde for 20 min. Then, the stationary liquid was removed, and 0.1% crystal violet (Solarbio, China) was added for 20 min of staining. The 6-well plates were gently rinsed with water, and colonies with > 50 cells were counted under an optical microscope (Olympus, Japan).

The cell cycle distribution of transfected cells was assessed by the following procedure. In brief, cultured cells were harvested, washed twice in PBS and fixed overnight at 4 °C with precooled 75% ethanol. After staining with propidium iodide, the cell cycle distribution was analyzed with a BD flow cytometer. 4′,6-Diamidino-2-phenylindole (DAPI) and Annexin-V-allophycocyanin (APC) double staining kits (BestBio, China) were used for apoptosis analyses.

The migration and invasion abilities of OS cell lines were evaluated using Transwell migration chambers (Costar, USA) and Transwell invasion chambers precoated with 50 μl of 2 mg/ml Matrigel (BD Biosciences, USA), respectively. In brief, transfected cells (4 × 10⁴ cells/well for the migration assay, 8 × 10⁴ cells/well for the invasion assay) suspended in 200 μl of serum-free DMEM were seeded into the upper chambers. A 600 μl volume of DMEM supplemented with 10% FBS was used as the attractant and was added into the lower chambers. After culture for 24 h, cells adhering to the lower surface of the membrane were fixed with paraformaldehyde (4%) and stained using crystal violet (0.1%), whereas cells on the upper surface of the membrane were removed by wiping with cotton swabs. At least three random fields of view containing cells that had migrated or invaded to the lower surface were imaged under an inverted light microscope.

Total protein from homogenized tissues or cell lysates was extracted using ice-cold RIPA solution (Fudebio, China) and protease inhibitors (Fudebio, China). The protein samples were diluted to equal concentrations, denatured in a boiling water bath, separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes. Membranes were blocked with 5% nonfat milk in Tris-buffered saline-Tween (TBST) buffer and were then incubated overnight at 4 °C with primary antibodies against cleaved CASP3 (1:1000) (Affinity Biosciences, USA), CCND1 (1:1200) (Proteintech, China), CDK4 (1:2000) (Abcam, USA), BAX (1:1000) (Abcam, USA), BCL2 (1:800) (Abcam, USA), E-cadherin (1:2000) (Proteintech, China), N-cadherin (1:2000) (Proteintech, China), Vimentin (1:3000) (Proteintech, China), FGF2 (1:200) (Santa Cruz Biotechnology, USA), PI3K (1:1000) (Cell Signaling Technology, USA), phosphorylated PI3K (p-PI3K, 1:1000) (Cell Signaling Technology, USA), Akt (1:1000) (Cell Signaling Technology, USA), phosphorylated Akt (p-Akt, 1:1000) (Cell Signaling Technology, USA) and GAPDH (1:10000) (Proteintech, China). After four washes with TBST buffer, membranes were incubated for 60 min at 25 °C with horseradish peroxidase (HRP)-conjugated secondary antibodies (1:10000) (Bioss, China). Protein bands were visualized with a chemiluminescence imaging system (Bio-Rad, USA).

The protocols for animal experiments were approved by the Medical Ethics Committee of Southern Medical University. Four-week-old female BALB/c athymic nude mice were purchased from Guangdong Medical Laboratory Animal Center (Guangdong, China) and reared in a pathogen-free facility with access to adequate standard food and water. The subcutaneous xenograft model was established by subcutaneous injection of 5 × 10⁶ stable 143B cells suspended in 100 μl of PBS into the right limbs of nude mice (n = 5 mice per group). Tumors were measured every 3 days, and tumor volumes were calculated using the following equation: volume = length × width² × 0.5. To establish the lung metastasis model, the abovementioned cells were injected into nude mice via the tail vein (2 × 10⁶ 143B cells in 200 μl of PBS per mouse, n = 5 mice per group) to mimic tumor metastasis. To evaluate the effect of miR-195-5p in vivo, the miR-195-5p antagomir/agomir or corresponding negative control was injected intratumorally (for the subcutaneous xenograft tumor model) or intravenously via the tail vein (for the lung metastasis model) twice weekly for 2 weeks in accordance with the manufacturer’s recommendations (RiboBio, China). Tumor tissues were extracted from sacrificed mice 4 weeks after inoculation. All mice were euthanized by CO₂ asphyxiation in accordance with the American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals (2013 Report of the AVMA Panel of Euthanasia).

Paraffin-embedded thin (4-μm) lung sections containing metastatic nodules were dewaxed using xylene and were then rehydrated through an alcohol gradient. Then, sections were stained using H&E for general histological examination by standard procedures.

The expression of target proteins in tissue samples from OS patients or animal xenograft models was determined using immunohistochemistry (IHC) as previously described [22]. Tissues were incubated with primary antibodies against FGF2 (1:100), Ki-67 (1:200), PCNA (1:300), N-cadherin (1:200), E-cadherin (1:200), and Vimentin (1:300). Except for the anti-FGF2 antibody (Santa Cruz Biotechnology, USA), all antibodies were purchased from Proteintech (China). After washing, the tissue sections were incubated with HRP-conjugated secondary antibodies (1:200) (Servicebio, China) and were then stained with diaminobenzidine (Zhongshan Golden Bridge, China). The tissues were observed and imaged using an optical microscope (Olympus, Japan).

To assess apoptotic DNA fragmentation, xenograft tumor tissues were first fixed for 24 h with 4% paraformaldehyde and were then embedded in paraffin. Apoptosis in situ was evaluated with a TUNEL Apoptosis Assay Kit (Alexa Fluor 488) (Yeasen, China). Corresponding images of the apoptotic cells were acquired using a fluorescence microscope. Analyses were performed with at least three random fields of view per sample.

To predict the potential miRNAs binding with circ_001422, five online bioinformatics tools were used: starBase (http://​starbase.​sysu.​edu.​cn/​), miRanda (http://​www.​microrna.​org/​), TargetScan (http://​www.​targetscan.​org/​vert_​72/​), RNAhybrid (http://​bibiserv.​techfak.​uni-bielefeld.​de/​rnahybrid/​), and RNA22 (https://​cm.​jefferson.​edu/​rna22/​). The potential downstream target mRNAs of miR-195-5p were predicted after analysis with the miRanda, RNAhybrid, TargetScan and miRTarBase (http://​mirtarbase.​mbc.​nctu.​edu.​tw/​) databases.

Transfected cells were washed twice in ice-cold PBS and were then lysed in RIP lysis solution containing RNase and protease inhibitors. Cell lysates (200 μl) were incubated overnight at 4 °C with immunoprecipitation buffer containing anti-Argonaute2 (anti-Ago2)-conjugated magnetic beads (Millipore, USA) or negative control anti-IgG (Millipore, USA). Subsequently, the immunoprecipitated RNAs were extracted and purified to determine the abundances of the target RNAs by qRT-PCR.

A control probe and a biotinylated circ_001422 probe were constructed by RiboBio (Guangzhou, China). The probes were coated with C-1 magnetic beads (Life Technologies, USA) after incubation for 2 h at room temperature with the abovementioned beads. Transfected cells were harvested and treated with ice-cold lysis solution and were then incubated overnight at 4 °C with the circ_001422 or oligo probes. Finally, the precipitates were extracted and purified using an RNeasy Mini Kit (Qiagen, USA). The abundances of circ_001422 and miRNAs in the RNA complexes were evaluated by qRT-PCR.

The circ_001422 or FGF2 fragments with the mutant (MUT) or wild-type (WT) miR-195-5p binding sites were subcloned downstream of the Renilla gene in the psiCHECK-2 dual-luciferase reporter vector (Geneseed, China). 143B and Saos-2 cells in logarithmic growth phase were cotransfected with the reporter vectors and the miR-195-5p mimic or NC mimic. After 48 h of incubation, a Dual-Luciferase Reporter Assay System (Promega, USA) was utilized to measure luciferase activity.

To investigate the potential involvement of circRNAs in OS, three matched OS samples and adjacent noncancerous tissues were subjected to high-throughput sequencing analysis. Based on criteria of a P-value < 0.05 and | fold change | ≥ 2, 374 differentially expressed candidate circRNAs were identified: 192 upregulated and 182 downregulated circRNAs in OS tissues relative to the matched noncancerous tissues. The top ten most upregulated and downregulated circRNAs are presented in the cluster heatmap (Fig. 1a). By combining these results with the circRNA annotation in the circBase database (http://​www.​circbase.​org/​), we found that circ_001422 is located on chromosome 4 (chr4: 1900626–1,935,262), has a length of 1703 bp and is generated by circularization of exons 2–7 of the host gene NSD2. The existence and structure of circ_001422 were validated through primer-specific amplification and Sanger sequencing of the PCR products (Fig. 1b). In addition, qRT-PCR further demonstrated upregulated expression of circ_001422 in OS cells relative to normal osteoblasts (Fig. 1c). Thus, circ_001422 was selected for further analyses.

CircRNAs are noncoding RNAs (ncRNAs) and are characterized by their stable structure due to the absence of a 5′ cap and a 3′ polyadenylated tail. Herein, circ_001422 and NSD2 mRNA were amplified using divergent and convergent primers, respectively. After RNase R treatment, the level of linear NSD2 mRNA decreased sharply, whereas that of circ_001422 remained stable (Fig. 1d). When both cDNA and gDNA were used as templates, NSD2 mRNA was amplified by convergent primers from both cDNA and gDNA, whereas circ_001422 was amplified by divergent primers only from cDNA (Fig. 1e). In addition, the actinomycin D-based transcriptional inhibition assay revealed that circ_001422 had a longer half-life than NSD2 mRNA (Fig. 1f). Nucleocytoplasmic fractionation and FISH further revealed that circ_001422 was localized predominantly in the cytoplasm (Fig. 1g and h).

The upregulated expression of circ_001422 was validated after analysis of OS tissues and matched noncancerous tissues from the patient cohort (n = 55) (Fig. 2a). In particular, overexpression of circ_001422 was observed in 40 (72.7%) of the 55 OS tissues relative to their matched noncancerous tissues (Fig. 2b). Further analyses revealed that overexpression of circ_001422 in OS tissues correlated positively with advanced clinical stage (I vs. III-IV, P = 0.026; II vs. III-IV, P = 0.048, Fig. 2c), large tumor size (P = 0.003, Fig. 2d) and distant metastasis (P = 0.005, Fig. 2e). No significant relationship was found for other parameters, including age (P = 0.520), sex (P = 0.150) and primary tumor location (P = 0.818, Fig. 2f) (Table 1). Kaplan-Meier survival analysis with the log-rank test was performed on OS patients in the high and low circ_001422 expression groups (based on the median circ_001422 expression level) and revealed that patients with higher circ_001422 expression levels exhibited worse overall survival rates (Fig. 2g). Moreover, receiver operating characteristic (ROC) analysis was used to determine the diagnostic potential of circ_001422 based on the 55 paired tissue samples. As shown in Fig. 2h, the area under the ROC curve (AUC) was 0.752, implying that circ_001422 is a relatively accurate diagnostic marker for OS.

Loss-of-function experiments were first performed by transducing 143B and Saos-2 cells with three shRNAs designed to specifically target the junction site of circ_001422. Among the shRNAs, sh-circ_001422#2 (hereafter denoted sh-circ_001422) showed the highest knockdown efficiency and was used in further analyses (Additional file 2: Fig. S1a). After knockdown of circ_001422, both the DNA synthesis and colony-forming abilities of 143B and Saos-2 cells were dramatically reduced (Fig. 3a and b). Flow cytometric analyses further revealed that circ_001422 knockdown induced G0/G1 phase arrest (Fig. 3c) and enhanced OS cell apoptosis (Fig. 3d). Moreover, the results of Transwell assays showed that circ_001422-knockdown cells had greatly decreased migration and invasion capacities (Fig. 3e). By western blot analysis (Fig. 3f), we found that circ_001422 silencing decreased the protein levels of CCND1, CDK4, BCL2, N-cadherin and Vimentin but upregulated the protein levels of cleaved CASP3, BAX and E-cadherin. Given the physiological role of these proteins, these findings suggest that circ_001422 regulates cell cycle progression, apoptosis and epithelial-mesenchymal transition (EMT), all of which impact OS development. For in vivo experiments, 143B cells transduced with stable sh-circ_001422 lentivirus or control lentivirus were injected subcutaneously (for the subcutaneous xenograft tumor model) or intravenously via the tail vein (for the lung metastasis model) into nude mice. Compared with mice in the control group, mice injected with circ_001422-knockdown cells exhibited significantly decreased tumor volumes and tumor weights (Fig. 3g-i) and a decreased number of metastatic pulmonary colonies (Fig. 3j and k). Immunohistochemical analysis further revealed that circ_001422 knockdown reduced the protein expression levels of Ki-67, PCNA, Vimentin and N-cadherin, but increased that of E-cadherin (Fig. 3l). Moreover, in the TUNEL assay more apoptotic cells were detected in the circ_001422-knockdown tumors than in the control tumors (Additional file 2: Fig. S1b).

In addition, OS cell lines were transduced with the circ_001422 overexpression vector or control vector, and the overexpression efficiency was validated using qRT-PCR (Additional file 3: Fig. S2a). The results of gain-of-function experiments consistently showed that overexpression of circ_001422 markedly suppressed the apoptosis but promoted the proliferation and metastasis of OS cells (Fig. 4, Additional file 3: Fig. S2b). Collectively, these findings demonstrate that circ_001422 regulates the oncogenic and metastatic properties of OS cells.

Accumulating studies have revealed that circRNAs contain abundant binding sites for miRNAs and thus may act as miRNA sponges. Given that circ_001422 was localized primarily and stably in the cytoplasm, we speculated that circ_001422 might be a competitive endogenous RNA. An anti-Ago2 RIP assay was performed by transfecting 143B cells with the Ago2 overexpression plasmid or the control vector. The qRT-PCR results demonstrated that endogenous circ_001422 was immunoprecipitated more effectively from the 143B cell clones with Ago2 overexpression than from the corresponding control cells (Fig. 5a). This finding suggested that circ_001422 might bind to miRNAs via the Ago2 protein. Next, 4 candidate miRNAs were identified by determining the overlap of the prediction results from 5 bioinformatics databases (TargetScan, miRanda, starBase, RNAhybrid and RNA22) (Fig. 5b). To validate our findings, we performed a pulldown assay using a biotinylated probe specific for circ_001422. As shown in Fig. 5c and d, overexpression of circ_001422 markedly enhanced the pulldown efficiency, and only miR-195-5p was substantially pulled down by the biotinylated circ_001422 probe in both Saos-2 and 143B cells. After validating the transfection efficiencies of the miR-195-5p mimic and inhibitor (Additional file 4: Fig. S3), we performed a dual-luciferase reporter assay, which demonstrated that the miR-195-5p mimic suppressed the luciferase activity of the WT circ_001422 reporter vector (circ_001422-WT vector) but not the MUT reporter vector (circ_001422-MUT vector) (Fig. 5e and f). In addition, the results of an anti-Ago2 RIP assay revealed that Ago2, circ_001422 and miR-195-5p were efficiently immunoprecipitated in the presence of anti-Ago2 but not anti-IgG antibodies. Moreover, circ_001422 and miR-195-5p were significantly enriched in 143B and Saos-2 cells transfected with the miR-195-5p mimic compared to those transfected with the NC mimic (Fig. 5g-i). qRT-PCR further revealed exceptionally low levels of miR-195-5p in the 55 OS samples relative to the corresponding paired noncancerous tissues, and miR-195-5p expression was inversely related to circ_001422 expression (Fig. 5j and k). Furthermore, miR-195-5p expression was downregulated in OS cells compared to normal osteoblasts (Fig. 5l).

To determine whether circ_001422 promotes the malignant phenotype of OS cells by targeting miR-195-5p, functional rescue experiments were performed. The miR-195-5p inhibitor was first transfected into circ_001422-knockdown cells. In the EdU incorporation and colony formation assays, circ_001422 knockdown dramatically decreased the number of EdU-positive cells and inhibited colony formation in both 143B and Saos-2 cells, whereas miR-195-5p silencing reversed these effects (Fig. 6a and b). Flow cytometric analyses revealed that reintroduction of the miR-195-5p inhibitor abolished the circ_001422 knockdown-induced increases in the proportions of G0/G1 phase and apoptotic cells (Fig. 6c and d). The suppressive effects of circ_001422 knockdown on the migration and invasion of OS cells were effectively weakened by exogenous inhibition of miR-195-5p expression (Fig. 6e and f). Western blot analysis further validated the downregulated expression of cell cycle-associated proteins (CCND1 and CDK4), an antiapoptotic protein (BCL2) and mesenchymal proteins (N-cadherin and Vimentin) after circ_001422 knockdown, but miR-195-5p inhibition restored the normal expression of these proteins. In contrast, miR-195-5p knockdown markedly reversed the sh-circ_001422-mediated upregulation of proapoptotic proteins (cleaved CASP3 and BAX) and an epithelial cell marker (E-cadherin) (Fig. 6g). Furthermore, the results of functional rescue experiments revealed that the miR-195-5p mimic markedly attenuated the promotive effects of circ_001422 on cell proliferation and metastasis (Additional file 5: Fig. S4). Collectively, these findings demonstrate that circ_001422 enhances the malignant phenotypes of OS cells at least partially by sponging miR-195-5p.

Given the above results, we further explored probable downstream molecules regulated by the circ_001422/miR-195-5p axis. Prior evidence demonstrates that circRNAs function as miRNA sponges to terminate the regulatory effects of miRNAs on target genes. In addition, the interaction between a miRNA and its target mRNA usually leads to degradation of the mRNA or posttranslational inhibition of gene expression. Thus, we hypothesized that the expression of circ_001422 is positively related to that of its target genes. To test this hypothesis, we first analyzed mRNA expression profiles by sequencing the differentially expressed genes (DEGs) in 143B cells with or without circ_001422 knockdown. A total of 2758 downregulated DEGs were identified based on a P-value < 0.05 and fold change ≤ − 2 (Fig. 7a). Further bioinformatic analysis of four databases (TargetScan, miRanda, RNAhybrid and miRTarBase) revealed 466 potential miR-195-5p target genes (Fig. 7b). After determining the overlap of the downregulated genes identified by mRNA-seq (n = 2758) and the potential miR-195-5p target genes predicted by the bioinformatics databases (n = 466), we identified 71 overlapping genes and further subjected these genes to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. As shown in Fig. 7c, “PI3K/Akt signaling pathway” was the most enriched pathway after “MicroRNAs in cancer”. All genes included in “PI3K/Akt signaling pathway” (CCND2, ITGA2, PRKAA1, FGF2, LAMC1 and GNB1) were considered candidate mRNAs potentially regulated by the circ_001422/miR-195-5p axis.

Subsequently, the levels of these 6 candidate mRNAs in 143B and Saos-2 cells were determined using qRT-PCR. Circ_001422 knockdown markedly altered the mRNA levels of FGF2 and LAMC1 in both 143B and Saos-2 cells compared to the corresponding control cells (Fig. 7d and e). After introduction of exogenous circ_001422, both 143B and Saos-2 OS cells exhibited a significant increase in FGF2 mRNA, whereas no significant change in LAMC1 mRNA was detected in 143B cells (Fig. 7f and g). Interestingly, FGF2 has been widely identified as an oncogene that participates in tumorigenesis and metastasis in multiple cancer types [23‐25]. Thus, we presumed that circ_001422 might contribute to the malignant progression of OS by protecting FGF2 against degradation induced by miR-195-5p.

The western blot analysis results indicated that circ_001422 silencing effectively reduced the FGF2 protein level and inhibited the phosphorylation of PI3K and Akt (Fig. 7h). In contrast, overexpression of circ_001422 significantly enhanced FGF2 expression and PI3K/Akt signaling pathway activity (Fig. 7i). Additionally, introduction of the exogenous miR-195-5p mimic into 143B and Saos-2 cells decreased the protein levels of FGF2, p-PI3K and p-Akt. Conversely, miR-195-5p silencing exerted the opposite effects in both cell lines (Fig. 7j and k). As expected, the circ_001422-mediated promotive effect on FGF2/PI3K/Akt axis activity was attenuated after introduction of miR-195-5p (Fig. 7l and m). Next, the putative miR-195-5p binding sites in FGF2 were identified (Fig. 7n). Transfection of the FGF2-WT reporter vector with the miR-195-5p mimic but not the scrambled oligonucleotide sequence dramatically decreased luciferase activity. Mutation of the binding sequence, however, abolished the disruption of luciferase activity (Fig. 7o). The expression levels of FGF2 mRNA and the corresponding protein were increased in OS tissues compared with the corresponding noncancerous tissues (Fig. 7p and q, Additional file 6: Fig. S5a). In addition, FGF2 expression showed significant positive and negative correlations with circ_001422 expression and miR-195-5p expression, respectively, in OS tissues (Additional file 6: Fig. S5b and S5c).

We performed rescue experiments to explore whether circ_001422 promotes the malignant phenotypes of OS cells by competitively interacting with miR-195-5p and then upregulating the expression of FGF2. The FGF2 overexpression and knockdown efficiencies were verified by qRT-PCR (Additional file 7: Fig. S6a and S6b). First, cells with stable circ_001422 silencing were transfected with the miR-195-5p inhibitor alone or in combination with sh-FGF2. The results of the EdU assay were consistent with previous data showing that the reduced proliferation ability of stable sh-circ_001422 cells was significantly attenuated by the miR-195-5p inhibitor but this rescue effect of the miR-195-5p inhibitor on stable sh-circ_001422 cells was effectively reversed by cotransfection with sh-FGF2 (Fig. 8a). Additionally, miR-195-5p silencing abolished circ_001422 knockdown-induced cell cycle arrest and apoptosis, while reintroduction of sh-FGF2 successfully attenuated these effects of miR-195-5p silencing (Additional file 7: Fig. S6c and S6d). The Transwell assay results revealed that the inhibitory effects of circ_001422 depletion on the cell migration and invasion capacities were dramatically weakened through exogenous inhibition of miR-195-5p. However, in cells with stable circ_001422 silencing, the suppressive effects of sh-circ_001422 on cell motility were restored after cotransfection with the miR-195-5p inhibitor and sh-FGF2 (Fig. 8b, Additional file 7: Fig. S6e). Similar trends were observed for the regulatory effects of the circ_001422/miR-195-5p/FGF2 axis on the expression of cell cycle-, apoptosis-, EMT- and PI3K/Akt signaling pathway-associated proteins (Fig. 8c). In the in vivo experiments, circ_001422 deficiency decreased subcutaneous tumor formation and lung metastases relative to those in mice injected with control cells. In addition, administration of the miR-195-5p antagomir abolished the effects of circ_001422 silencing and accelerated the growth and metastasis of OS cells, while knockdown of FGF2 reversed the pro-oncogenic role of the miR-195-5p antagomir (Fig. 8d-h). The results of immunohistochemical analysis (Fig. 8i) and TUNEL assays (Additional file 7: Fig. S6f) further validated the role of the circ_001422/miR-195-5p/FGF2 axis in OS. Moreover, functional rescue experiments were performed in cell lines with stable circ_001422 overexpression. The results, as shown in Fig. 9 and Additional file 8: Fig. S7, revealed that circ_001422 competitively interacted with miR-195-5p to upregulate FGF2 expression and activate the PI3K/Akt signaling pathway, thus facilitating the metastatic and proliferative abilities of OS cells. A schematic diagram was generated to visualize the role of the circ_001422/miR-195-5p/FGF2 axis in OS (Fig. 10).