Screening and characterization of circFIRRE in OS samples and cells
Initially, we studied the circRNAs expression profile using ribosomal RNA-depleted RNA-seq analysis from 4 paired OS and control samples. In the circRNAs profile database constructed after sequencing, most sequence-length of circRNAs were less than 2000 nucleotides (Figure S1A). Aberrantly expressed circRNAs are presented in Volcano Plot (Figure 1A). Upregulated circRNAs were further screened according to the criteria of a basemean value > 3, a fold change in expression > 2 and a detection Padj value < 0.05. The eligible 29 circRNAs are presented in Heat map (Figure S1B). Among the top three circRNAs with minimum Padj value, circFIRRE (Circbase[20] ID: hsa_circ_0001944) had the biggest fold change (Figure 1A), thus was chosen as the candidate and validated by RT-qPCR using back-spliced junction–specific primers. In 104 matched clinical OS samples and adjacent normal samples, circFIRRE was significantly upregulated as shown by RT-qPCR and FISH assays (Figure 1B, Figure 1D). Meanwhile, the expression level of circFIRRE was significantly elevated in 5 OS cell lines compared to osteoblast cell line (hFOB1.19), especially in MG63 and U2OS (Figure 1E). Afterwards, in orthotopic and tail vein metastasis mouse xenograft models (15 mice in each group) that we established, circFIRRE was significantly upregulated in both orthotopic and lung OS lesions compared to normal tissues (Figure 1C). Interestingly, the expression level of circFIRRE in lung metastasis OS was significantly higher than that in orthotopic lesions from xenograft models, indicating that circFIRRE probably played a role in both tumorigenesis and hematogenous lung metastasis in OS.
Subcellular localization of circFIRRE is essential for its cellular function, lncLocator 2.0[21] was firstly applied to predict circFIRRE subcellular localization, and the result showed that circFIRRE was preponderantly distributed in the cytoplasm (Figure S1C). In addition, robust expression of cytoplasmic circFIRRE in MG63 and U2OS was confirmed by both RT-qPCR following nuclear/cytosol separation and RNA FISH (Figure 1F-H).
circFIRRE is originated from firre intergenic repeating RNA element (FIRRE) locus on chromosome X (chrX: 130883333 - 130928494), and generated by the exon 5 to 10 of gene FIRRE region. The exact size of circFIRRE is 1096 bp and its specific trans-splicing was confirmed by Sanger sequencing (Figure 1I). The stability of circFIRRE in MG63 and U2OS cells was examined after treatment with actinomycin D, knowing that it can transcriptionally inhibit RNA synthesis. circFIRRE had a half-life of >24 hours, which was more stable than linear transcript FIRRE with a half-life of <5 hours (Figure 1J-K). circFIRRE was also resistant to digestion with exonuclease ribonuclease R (RNase R), while FIRRE was readily degraded after treatment (Figure 1L-O). These results suggested that circFIRRE could sustainedly and stably express in OS, which may be an eligible biomarker for diagnosis or prognosis prediction.
Knowing that back-spliced junction could come from either trans-splicing or genomic rearrangements, agarose gel electrophoresis (AGE) was applied to eliminate the possibility of genomic rearrangements. circFIRRE and liner FIRRE were separately amplified by divergent primers and convergent primers in cDNA and genomic DNA extractives from OS cells. circFIRRE was only found in cDNA but not in genomic DNA by AGE assay (Figure 1P-Q).
Upregulated circFIRRE expression portends higher metastatic risk and worse prognosis in OS patients
According to the circFIRRE expression in OS lesions, the 104 patients were equally divided into two groups. The baseline characteristics of the enrolled patients are displayed in Figure S2. To evaluate risk factors in overall survival (OS) and disease-free survival (DFS), univariate and multivariate analyses were applied. Univariate analysis showed that age, surgical modalities, tumor metastasis and high circFIRRE expression were associated with OS and DFS. In compromising these variables, multivariate analysis illustrated that metastasis and elevated circFIRRE expression were independent risk factors for prognosis of OS patients, together with age and surgical modalities. (Figure 2A-D). Additionally, Kaplan-Meier survival curves showed worse OS and DFS in patients in high circFIRRE expression than those with low circFIRRE expression group (p<0.001) (Figure 2E-F). Specifically, high expression level of circFIRRE was positively associated with shorter survival, poorer clinical outcomes and higher risk scores (Figure 2G). In addition, clinical baseline demonstrated that a high expression level of circFIRRE was also correlated with a high metastatic ratio (Figure S2).
All these results showed that circFIRRE could function as an independent risk factor of OS patients and may participate in OS progression and metastasis.
circFIRRE promotes OS progression in vitro
To investigate the biological role of circFIRRE, gene-set enrichment analysis (GSEA) (https://www.gsea-msigdb.org/gsea/index.jsp) was conducted and the result showed that circFIRRE was closely correlated with tumor progression and metastasis, especially in epithelial mesenchymal transition (EMT), cell cycle and angiogenesis (Figure S3A). To explore the impact of circFIRRE upregulation on tumor progression, loss-of-function assays were performed in OS cell lines. In the first step, we designed three siRNAs targeting the back-spliced junction of circFIRRE and conducted transfection in MG63 and U2OS cells. The results showed that si-circFIRRE-1 (si-1) and si-circFIRRE-2 (si-2) had relative a better knockdown effect (Figure S3B). Afterwards, si-1 and si-2 were chosen for the preliminary experiment in MG63 cells. CCK8 and wound healing assays were performed to test OS cell proliferation and migration capacities. CCK8 assays showed that both si-1 and si-2 significantly diminished the growth rate of MG63, which was consistent with the knockdown efficiency in transfection (Figure S3C). Wound healing assays showed that si-1 and si-2 retarded the rehabilitation of scratches 24 hours after scratching (Figure S3D). These results indicate that circFIRRE may contribute to OS progression in vitro.
In the second step, we infected lentivirus packaged shRNAs (sh-circFIRRE-1 and -2) (Figure S3E) and established circFIRRE stable knockdown cell lines in MG63 and U2OS (Figure 3A). Then, we performed CCK8 and EdU assays to evaluate the proliferation ability of OS cells. The result showed that sh-circFIRRE cell proliferation and EdU combination were reduced significantly as compared with the negative control cells (Figure 3B-F). Additionally, wound healing assay showed that silencing circFIRRE expression suppressed the migration capability of OS cells (Figure S3F); both migration and invasion abilities of OS cells were diminished by circFIRRE knockdown in Transwell migration and invasion assay (Figure 3G-I). We then conducted flow cytometry to evaluate the cell cycle and found that sh-circFIRRE cells were arrested in G0/G1 phase (Figure S3G-I), demonstrating that circFIRRE knockdown could promote cell cycle arrest. Collectively, these results demonstrate that circFIRRE knockdown inhibited OS tumor growth and motility in vitro.
In the third step, gain-of-function assays were performed to further explore the function of circFIRRE in vitro. Firstly, pGMLV-circRNA vector was designed to overexpress circFIRRE in OS cells (Figure S4A-B) and RT-qPCR verified the overexpression efficiency and confirmed that transfection did not influence linear FIRRE expression (Figure 3J). Next, CCK8 and EdU assays revealed that circFIRRE overexpression significantly facilitated MG63 and U2OS proliferation (Figure 3K-O). Additionally, wound healing assay (Figure S4C-D) and Transwell assays illustrated that circFIRRE overexpression promoted OS cell migration and invasion (Figure 3P-R). Flow cytometry demonstrated that circFIRRE promoted cell proliferation by promoting cell cycle (Figure S4E-G). In summary, circFIRRE contributed to the growth, migration and invasion of OS in vitro.
circFIRRE can induce angiogenesis
As mentioned before, GSEA analysis found that circFIRRE may contribute the tumor progression and metastasis, and angiogenesis may be the major biological procedure of tumor metastasis in which circFIRRE was involved (Figure S3A, Figure 4A). Hence, we hypothesized that circFIRRE may be involved in angiogenesis, which is vital in OS pulmonary metastasis. We chose HUVEC cell line for angiogenesis research to observe its proliferation, migration and tube formation capacities. First, we transferred three circFIRRE targeted siRNAs into HUVEC cells, and found that si-1 and si-2 diminished more than 50% circFIRRE expression level (Figure 4B), exhibiting a relatively good knockdown effect. Therefore, we selected them for functional experiments. Second, we verified the role of circFIRRE in HUVEC proliferation, motility and tube formation by CCK8 (Figure S5A), EdU assay (Figure S5B-C), wound healing assay (Figure S5D), Transwell migration assay (Figure S5E-F) and tube formation assay (Figure 4C-D). It was found that si-circFIRRE significantly compromised cell proliferation, migration and tube formation capabilities in HUVEC cells compared to the control. Third, we used ex vivo aortic ring endothelial cell sprouting assay and in vivo chick chorioallantoic membrane (CAM) assay to further verify our hypothesis. The aortic ring assay showed that si-circFIRRE transfected aortic rings were embedded in collagen for seven days, and less microvessel area was observed as compared with the normal group (Figure S5G, Figure 4E-F). CAM is a highly vascularized extraembryonic membrane of the chick embryo, in which newly formed vessels were obviously compromised after si-circFIRRE co-incubation (Figure S5H, Figure 4G-H). All these findings demonstrate that circFIRRE knockdown significantly restrained angiogenesis.
Then, we transfected circFIRRE plasmids into HUVEC cells and the overexpression efficiency assessed by RT-qPCR is shown in Figure 4I. Cell viability and migration capabilities of HUVEC cells were evaluated by CCK8, EdU, wound healing and Transwell migration assays, and the results showed that circFIRRE was vital in maintaining proliferation and migration of HUVEC cells (Figure S6A-F). Tube formation assay was performed to assess new sprout and the result showed that circFIRRE overexpression prominently increased the branch points and capillary length of HUVEC cells (Figure 4J-K). CAM and aortic ring assays illustrated that circFIRRE could promote angiogenesis as well (Figure 4L-O, Figure S6G-H). Finally, all these experiments revealed that circFIRRE can facilitate neovascularization in neoplasm metastasis.
The expression of circFIRRE in OS can be regulated by YY1
Given the significant upregulation of circFIRRE in OS, we assumed that circFIRRE was regulated by upstream transcription factor (TF) in human OS development. We retrieved the promoter sequence of circFIRRE from UCUS Genome Browser (http://genome.ucsc.edu/), and then applied three algorithms (UCSC, PROMO, JASPAR) to predict potential TFs that could combine the promoter sequence. It was found that Yin Yang 1 (YY1) was the only TF in the overlap of these algorithms (Figure 5A). Furthermore, we applied Gene Expression Profiling Interactive Analysis (GEPIA, https://gepia.cancer-pku.cn/) in The Cancer Genome Atlas (TCGA) database and found that the expression level of YY1 and gene FIRRE were elevated in sarcoma (SARC) (Figure S7A-B). Analysis of our RNA-seq results showed that YY1 and gene FIRRE expressions were elevated in OS samples compared with those in the adjacent no-tumorous samples, which is consistent with the upregulation of circFIRRE (Figure S7C-D). We then detected expression level of YY1 in 35 paired OS samples and found that circFIRRE was highly expressed in OS samples relative to the adjacent normal controls, and was positively correlated with the expression level of circFIRRE (Figure 5B-C). At cellular level, YY1 was relatively upregulated in MG63 and U2OS cells compared to hFOB1.19 cells (Figure 5D).
Subsequently, we compared promoter sequence of circFIRRE with YY1 binding motif in JASPAR and predicted two possible YY1 binding sites in circFIRRE promoter (Figure 5E), and dual-luciferase reporter assays were applied to verified two possible binding sites. In wild type (WT) group, the luciferase activity of circFIRRE promoter was strengthened under YY1 overexpression, and the mutation of either binding site 1 or 2 partially rescued the strengthened effect, whereas luciferase activity exhibited no significant change when we mutated both binding site 1 and 2 (Figure 5F).
As the linear transcript of gene FIRRE, linear FIRRE is also a well-known lncRNA. The RNA-seq results showed that the expression level of linear FIRRE in OS samples was more than treble that in adjacent normal samples, but its extent of upregulation was far less than that of circFIRRE (Figure S7E-F). We next explored the effect of YY1 on the expression change of circFIRRE and linear FIRRE. Three siRNAs of YY1 were designed and si-YY1-1 was selected for following experiments according to the relatively better knockdown effect (Figure 5G). The expression level of circFIRRE in MG63 and U2OS was decreased significantly after YY1 knockdown, while there was a minimal decreased in linear FIRRE expression, showing no significant difference between the two groups (Figure 5H-I), YY1 overexpression exhibited the similar results (Figure 5J-L). These results suggest the probability that the transcriptional regulatory effect of YY1 is mainly on circFIRRE, rather than on linear FIRRE. Altogether, YY1 may be an activator of circFIRRE transcription in OS.
circFIRRE may sponge miR-486-3p and miR-1225-5p in OS cells
Knowing that circFIRRE was predominantly expressed in cytoplasm (Figure 1F-H), we assumed that circFIRRE might act as miRNA sponge to neutralize miRNA-mediated gene silencing. Initially, RNA immunoprecipitation targeting AGO2 (argonaute RISC catalytic component 2) protein, a core component of RNA-induced silencing complex (RISC) was performed. AGO2 protein acted as an intermediary binding both miRNAs and target mRNAs. The results showed that circFIRRE was enriched specifically by AGO2 pull down instead of immunoglobulin G (IgG) (Figure 6A-C), suggesting that circFIRRE may bound to RISC and sponge corresponding miRNAs.
Furthermore, we applied five algorithms (miRanda, RNAhybrid, Interactome, circbank and RegRNA2.0) to predict the potential target miRNAs of circFIRRE, and identified miR-486-3p and miR-1225-5p as candidates from the overlap between the databases (Figure 6D). We sought to verify whether circFIRRE could regulate the expression levels of miR-486-3p and miR-1225-5p by RT-qPCR. The results illustrated that these two miRNAs were significantly declined in 35 paired OS samples and cell lines relative to the adjacent normal controls and osteoblasts (Figure 6E, Figure S8A-B), and they were negatively correlated with circFIRRE expression (Figure 6H-I). FISH assay further confirmed these patterns (Figure 6F-G). Then, the expression level of miR-486-3p and miR-1225-5p was significantly upregulated by circFIRRE knockdown (si-circFIRRE-1), and down-regulated by circFIRRE overexpression (Figure 6J-K). These results demonstrated that miR-486-3p and miR-1225-5p were comparatively low-expressed in OS, and negatively regulated by circFIRRE.
Additionally, we investigated whether circFIRRE could directly bind these two miRNAs by pull-down assay by using a specific biotin-labelled circFIRRE probe. It was found that the circFIRRE probe could specifically enrich circFIRRE, miR-486-3p and miR-1225-5p in cell lysate relative to the oligo probe (Figure S8C-D, Figure 6L-M). Specific biotin-labelled miR-486-3p and miR-1225-5p probes were applied in pull-down assay for further verification and successfully capture circFIRRE compared with the control probe (Figure 6N-O). Apart from RNA pull-down, binding sites for miR-486-3p or miR-1225-5p in circFIRRE were forecasted by circinteractome (https://circinteractome.nia.nih.gov/) algorithm (Figure S8E), and dual-luciferase reporter assay was performed via co-transfecting luciferase reporter plasmids with miR-486-3p or miR-1225-5p mimics into HEK-293 T cells. The results showed that miR-486-3p and miR-1225-5p synergistically diminished the luciferase activity by at least 50% compared with the negative control miRNA. We subsequently mutated the predicted miRNA binding sites from the luciferase reporter plasmid, finding that the mutant luciferase reporter activity remained unchanged after miRNAs transfection (Figure 6P). These results were supported by the colocalization of circFIRRE and corresponding miRNAs in MG63 and U2OS as sown by double FISH assay (Figure 6Q-R). All these findings suggest that circFIRRE may work as a sponge for miR-486-3p and miR-1225-5p.
circFIRRE promotes OS tumorigenesis and neovascularization via the miR-486-3p/miR-1225-5p-LUZP1 axis in vitro
It was reported in previous studies that both miR-486-3p and miR-1225-5p were downregulated in OS, and the latter was able to suppress OS progression[22-24]. Thus, we hypothesized that circFIRRE may promote OS progression and neovascularization by restraining the protective effect of two miRNAs and activating downstream gene. In the first place, we aimed to seek the target gene of miR-486-3p and miR-1225-5p in OS. By cross-analyzing three prediction algorithms (miRDB, TargetScan and RNAInter) and our RNA-seq data (fold-change≥1.5, p ≤ 0.05), we found that LUZP1, also known as Leucine Zipper Protein 1, was the sole predicted gene among overlapping gene (Figure 7A). It was reported that LUZP1 could regulate cancer features by modulating actin cytoskeleton stability and ciliogenesis[25, 26]. Our RNA-seq analysis revealed that LUZP1 was elevated in four paired OS samples relative to normal adjacent samples, and in sacoma from TCGA database (Figure S9A-B). The upregulated tendency was then testified in 35 paired clinical samples using RT-qPCR (Figure 7B), and they were found to be negatively correlated with miR-486-3p and miR-1225-5p expression level, and positively correlated with circFIRRE (Figure 7C-E). Moreover, we co-transfected luciferase reporter plasmids with miR-486-3p and miR-1225-5p mimics in HEK-293 T cells to investigate the interaction between these two miRNAs and LUZP1. The decreased luciferase activity by the LUZP1 3′-UTR was accompanied with miRNAs overexpression. In contrast, luciferase activity remained unchanged compared with the controls after LUZP1 3′-UTR mutation (Figure 7F, Figure S9C).
In the second place, we sought to demonstrate whether circFIRRE regulated LUZP1 via miR-486-3p and miR-1225-5p. Initially, we detected the regulatory relationship between circFIRRE and LUZP1, and found that LUZP1 was strikingly downregulated after circFIRRE knockdown (si-circFIRRE-1 and 2), and significantly upregulated when circFIRRE mRNA and protein levels were overexpressed (Figure S9D-G). Additionally, rescue assays were applied using RT-qPCR and Western blot. The results showed that endogenous LUZP1 down-regulation induced by si-circFIRRE-1 was partially rescued by miR-486-3p or miR-1225-5p inhibitors in mRNA and protein level (Figure 7G-H).
In the last place, we aimed to determine whether circFIRRE promoted tumor progression and angiogenesis in OS via sponging miR-486-3p and miR-1225-5p in vitro. Both circFIRRE stable knockdown OS cells (MG63, U2OS) and untreated HUVEC cells were applied in functional experiments. Functionally, CCK8 assay illustrated that miR-486-3p and miR-1225-5p inhibitors greatly abolished the inhibitory effects on sh-circFIRRE-1 cell proliferation (Figure S9H). Wound healing assay and Transwell migration and invasion assay illustrated that miR-486-3p and miR-1225-5p inhibitors markedly eliminated the inhibitory effects on migration and invasion in sh-circFIRRE-1 cells (Figure S9I, Figure 7I-J). Moreover, tube formation assay revealed that miR-486-3p and miR-1225-5p inhibitors markedly increased the reduced branch points and capillary length in si-circFIRRE-1 cells (Figure 7K-L).Collectively, these findings demonstrated that circFIRRE promoted OS progression and neovascularization in vitro, at least partially, through miR-486-3p/miR-1225-5p-LUZP1 pathway.
circFIRRE acts as miRNA sponge to promote both OS tumorigenesis in situ and metastasis in vivo
To verify the functions of circFIRRE-miR-486-3p/miR-1225-5p axis in vivo, both orthotopic xenograft tumor model and tail vein metastasis models were set up. Stable MG63 cells transfected with empty virus were used as negative control (Group-A), stable MG63 cells transfected with sh-circFIRRE-1 (Group-B) or co-transfected with sh-circFIRRE-1 and these two miRNA sponges (Group-C) were established. Considering the tibia is the common primary site of OS, we injected MG63 cells labelled with luminescent dye into the marrow cavity of the right tibia and established an orthotopic xenograft tumor model (ten mice in each group). Four weeks after injection, in vivo bioluminescence imaging (IVIS) assay was applied to monitor the process of tumorigenesis. The luciferase Intensities illustrated that circFIRRE knockdown reduced the tumor size in situ, videlicet, restrained OS cell proliferation, while miR-486-3p and miR-1225-5p suppression alleviated the impairment (Figure 8A). In addition, micro-CT scans and 3-D reconstruction were performed to evaluate the bone destruction caused by tumorigenesis 5 weeks after injection. The results exhibited grievous tibiofibula and joint destructions in the control group, and the bone destruction was alleviated in sh-circFIRRE group, while the repairment of sh-circFIRRE was rescued by miR-486-3p and miR-1225-5p sponges (Figure 8B), which is consistent with the consequence of IVIS. Five weeks after injection, mice were sacrificed and OS lesions were obtained for immunohistochemical (IHC) and protein analyses. The IHC result showed that the tumor proliferative activity was diminished as shown by the cell proliferation marker ki-67, and the EMT capacity was inhibited as shown by the mesenchymal markers N-cadherin and Vimentin, and the epithelial marker E-cadherin in sh-circFIRRE-1 group, while these tumor characteristics were rescued by miRNA sponges (Figure 8C, Figure S10). The down-regulated protein level of LUZP1 was observed in sh-circFIRRE-1 group relative to the control, while inhibition of miR-486-3p and miR-1225-5p partially abrogated the downregulation in IHC and Western bolt analyses (Figure 8D-E).
To further investigate how circFIRRE regulated tumor pulmonary metastasis and angiogenesis, we constructed a lung metastasis model by injecting stable MG63 cells labelled with firefly luciferase into the lateral tail vein of nude mice (ten mice in each group). MG63 cells were treated with (Group-B) or without sh-circFIRRE-1 transfection (Group-A) or with both circFIRRE and miR-486-3p & miR-1225-5p silenced (Group-C). According to the luciferase intensity in IVIS assay 4 weeks after injection, we found diffuse metastasis in the lung field of the mice in control group; the lung zone in the sh-circFIRRE group was relatively clear without pervasive metastasis; and silencing miR-486-3p and miR-1225-5p dramatically increased the extent of lung metastasis and the metastatic lesion size compared with sh-circFIRRE group (Figure 8F). Analogously, micro-CT scans and 3-D reconstruction exhibited a prominent shrinkage in the tumor volume and number in the sh-circFIRRE group relative to the control group, while the volume and number of metastatic lesions was increased in sh-circFIRRE and miRNAs sponge group as compared with sh-circFIRRE group (Figure 8G-I). The photograph under while light and haematoxylin and eosin (H&E) staining of the excised lungs further confirmed this conclusion (Figure 8J-K). The expression of circFIRRE and miRNAs was detected by FISH test, and fluorescence in the metastatic lesions displayed the knockdown effect of sh-circFIRRE-1, and the knockdown of circFIRRE was rescued by miR-486-3p and miR-1225-5p sponges (Figure S11). Vascular endothelial growth factor (VEGF) is an indispensable mediator of angiogenesis, and upregulated expression of VEGF means the course of aberrantly activated angiogenesis[27]. CD31 is the marker of vascular endothelial cells, which represents the formed blood vessel in lesions. VEGF and CD31 staining in metastatic OS showed that angiogenesis was significantly activated in both control and miRNAs sponge groups, and restrained in sh-circFIRRE-1 group (Figure 8L-M). In addition, IHC demonstrated significant changes in the expression of ki-67, N-cadherin, E-cadherin and Vimentin (Figure S12). Also, IHC and Western blot validated corresponding changes in LUZP1 expression in the metastatic lesions (Figure 8N-O). All these findings demonstrated that circFIRRE promoted the in-situ growth, lung metastasis and angiogenesis of OS by sponging miR-486-3p/miR-1225-5p in vivo.