CircCHD7 expression varied among osteosarcoma cell lines
First, we analyzed data from the GSE96964 dataset and found that 20 circRNAs were significantly expressed based on the standards of fold change (Supplemental Table 1). To elucidate the functions of these circRNAs, we performed apoptosis assays after circRNA knockdown and that hsa_circ_0084582 (circCHD7) had the greatest effect on promoting apoptosis (Supplemental Figure S1A, Fig. 1a). Moreover, analysis of circCHD7 expression in three pairs of chondroma and osteosarcoma tissues by reverse transcription quantitative polymerase chain reaction (RT-qPCR; Fig. 1b) showed that circCHD7 expression was low in osteosarcoma. Similar results were observed in different cell lines (Fig. 1c). To further confirm differences in the expression of circCHD7 in chondroma and osteosarcoma, we performed RNA fluorescence in situ hybridization (FISH) and found that circCHD7 showed low expression in osteosarcoma tissues compared with chondroma tissues (Fig. 1d). Indeed, circCHD7 expression was lower in osteosarcoma cell lines than in h-FOB1-19 cells. Although these findings differed from the data in the GSE96964 dataset, we verified our results several times.
Among cell lines showing low expression, we chose 143B and HOS cells, which exhibited relatively higher expression, for subsequent experiments. Additionally, analysis of the effects of RNase R on total RNA showed that circCHD7 was more stable than linear CHD7 mRNA (Fig. 1e). Further analyses of total RNA and genomic DNA from 143B and HOS cell lines (Fig. 1f) showed that circCHD7 could be amplified from cDNA but not genomic DNA, demonstrating that circCHD7 existed in 143B and HOS cells. Sanger sequencing of total cDNA (Fig. 1g) revealed that the sequence of circCHD7 contained a back-splicing site, again supporting the existence of circCHD7. In addition, FISH analysis of circCHD7 in 143B and HOS cells (Fig. 1h) showed that circCHD7 was present in the cytoplasm of 143B and HOS cells. Overall, these results demonstrated that circCHD7 showed differential expression in chondroma and osteosarcoma and that circCHD7 was expressed in the cytoplasm of 143B and HOS cells.
eIF4A3 regulated circCHD7 expression
In order to explore the mechanisms regulating circCHD7 expression, we searched the Circinteractome database (https://circinteractome). Eight RNA binding sites for eIF4A3 were found to match the flanking regions of circCHD7 (Fig. 2a). A previous study suggested that eIF4A3 could act as an RNA editing protein to promote the formation of circRNAs (27, 28). Therefore, we evaluated whether eIF4A3 regulated circCHD7 expression using a probe containing the upstream 500 bp and downstream 500 bp of circCHD7. RNA pulldown assays (Fig. 2b) and western blotting (Fig. 2c) showed that eIF4A3 could be pulled down by this probe. Furthermore, we transfected cells with small interfering RNA (siRNA) targeting eIF4A3 and found that circCHD7 expression was downregulated, whereas CHD7 mRNA levels did not change significantly (Fig. 2d). Collectively, these findings suggested that eIF4A3 bound with the mRNA of CHD7 and promoted circCHD7 expression.
CircCHD7 suppressed the proliferation and migration of osteosarcoma cells
To assess the functions of circCHD7 in osteosarcoma, cells were transfected with siRNA targeting circCHD7. Importantly, circCHD7 was knocked down by this siRNA, whereas CHD7 mRNA levels were not altered (Fig. 3a), demonstrating that the siRNA affected circCHD7 only. Next, after knockdown of circCHD7, we assessed the migration of osteosarcoma cells using transwell assays (Fig. 3b). The results demonstrated that migration ability decreased when circCHD7 was knocked down in 143B and HOS cells. Analysis of cell proliferation by clone formation assays (Fig. 3c) and Cell Counting Kit-8 (CCK8) assays (Supplemental Fig. 3A) also showed that proliferation was reduced in 143B and HOS cells after circCHD7 knockdown. And the apoptosis rates of osteosarcoma cells were increased after the expression of circCHD7 was knocked down (Fig. 3d). These results suggested that circCHD7 acted as a protective factor in the progression of osteosarcoma.
CircCHD7 regulated proliferation and apoptosis by sponging miR-608
CircRNA plays roles in cancer by sponging downstream miRNAs (29). In our study, analysis of the Circinteractome database suggested that circCHD7 may bind with argonaute RISC catalytic component 2 (AGO2; Supplemental Figure S4A), which has been reported to bind with circRNAs to sponge miRNAs and further regulate gene expression(30). Therefore, we speculated that circCHD7 may regulate apoptosis and proliferation in osteosarcoma cells by sponging miRNA. Thus, to explore whether circCHD7 bound to miRNA to regulate downstream target genes, we performed RIP. The results showed that circCHD7 could indeed bind with AGO2 (Fig. 4a). Analysis of RNAhybrid, TargetScan, MiRanda, and Circbank databases showed that 37 miRNAs could bind with circCHD7 (Fig. 4b). In addition, in the GSE28423 database, 257 miRNAs were found to exhibit differential expression between osteosarcoma and adjacent tissues (Fig. 4c Supplementary Table 2). we take the intersection of the bioinformation and the GSE28423 and found the miR-608 (Fig. 4d). MiR-608, which is known to promote cell proliferation, migration, and invasion in ovarian cancer(31). Therefore, we speculated that miR-608 may have similar functions in osteosarcoma cells. To confirm the interaction of circCHD7 and miR-608, we conducted RNA pulldown assays and found that miR-608 did bind to circCHD7 in 143B and HOS cells (Fig. 4e). Six binding sites for circCHD7 and miR-608 were predicted by the TargetScan database (Fig. 4f). Luciferase reporter assays (Fig. 4g) further confirmed that circCHD7 could bind to miR-608. Finally, FISH experiments in 143B and HOS cells (Fig. 4h) showed that circCHD7 and miR-608 were localized in the same regions in cells. Taken together, these findings demonstrated that circCHD7 could bind to miR-608 and that miR-608 may act downstream of circCHD7 to regulate molecular functions in osteosarcoma.
miR-608 regulated apoptosis and proliferation in osteosarcoma cells
miR-608 regulates the proliferation of ovarian cancer (32). However, the roles of miR-608 in osteosarcoma remain unknown. Therefore, we transfected miR-608 mimics into 143B and HOS cell lines and the qRT-PCR verified the efficiency of transfection (Fig. 5a). Migration assays (Fig. 5b), clone formation assays (Fig. 5c), and CCK-8 assays (Supplemental Figure S5A) demonstrated that cell proliferation was decreased after transfection with miR-608 mimics in 143B and HOS cells. Moreover, flow cytometry analysis revealed that the rate of apoptosis was increased after transfection with miR-608 mimic into 143B and HOS cells (Fig. 5d). These results demonstrated miR-608 indeed regulated the proliferation and apoptosis of osteosarcoma cells.
Next, to determine whether circCHD7 modulated osteosarcoma by regulating miR-608, we transfected cells with Si-circCHD7 and miR-608 inhibitor. Importantly, our results showed that the effects of circCHD7 knockdown on blocking migration and proliferation were partly reversed by inhibiting the expression of miR-608 (Fig. 5e, 5f, Supplemental Figure S5B). Similar results were observed by flow cytometry (Fig. 5g). Collectively, these results demonstrated that circCHD7 regulated the molecular functions of osteosarcoma through miR-608.
MiR-608 regulated tumor migration and proliferation via MMP2 and FZD4
We then assessed the mechanisms through which miR-608 mediated molecular functions in osteosarcoma using the miRDIP website (http://ophid.utoronto.ca/mirDIP/index.jsp) to predict the downstream target genes of miR-608. Based on the criteria of integrated score greater than or equal to 0.5 and score class of very high, we identified 166 genes (Supplement Table 3). From these genes, we performed Kyoto Encyclopedia of Genes and Genomes analysis (Fig. 6a) and found that the proteoglycans in cancer pathway was associated with miR-608. We then further evaluated the 30 genes predicted to bind with miR-608 in this pathway.
To determine which genes were regulated by miR-608 and circCHD7, 143B cells were transfected with si-circCHD7 and miR-608 mimic, and the expression levels of the target genes were assessed by RT-qPCR (Supplementary Figure S6A). The results showed that MMP2 and FZD4 were stably downregulated by knockdown of circCHD7 or overexpression of miR-608 (Fig. 6b). In addition, the expression levels of MMP2 and FZD4 differed in the GSE126209 dataset (Fig. 6c). TargetScan database analysis identified binding sites for miR-608 in MMP2 and FZD4 mRNA sequences (Fig. 6d), and direct binding of miR-608 with MMP2 and FZD4 was confirmed by luciferase reporter assays (Fig. 6e). Taken together, these findings showed that miR-608 could bind to MMP2 and FZD4 and that MMP2 and FZD4 may act downstream of miR-608.
CircCHD7 regulated miR-608/FZD4 through the Wnt/β-catenin pathway
MMP2 regulates tumor metastasis by affecting the balance of the extracellular matrix and has been shown to promote the progression of colorectal cancer, neuroblastoma, and thyroid cancer(33–36). Therefore, MMP2 is a known biomarker of cancer and a potential target of antitumor activity. Additionally, FZD4 activates the Wnt/β-catenin pathway and directly or indirectly regulates downstream target genes (37), such as LRP receptor, glycogen synthase kinase-3β, and β-catenin (38). β-Catenin forms complex with lymphoid enhancer factor and T-cell factor proteins (39), which then activates the transcription of tumorigenesis-related target genes, including c-Myc, cyclin D1, vimentin, E-cadherin, and N-cadherin (40).
Next, we explored whether circCHD7 regulated the proliferation of osteosarcoma via the Wnt/β-catenin pathway and MMP2 by assessing the protein and mRNA expression of FZD4, MMP2, vimentin, cyclin D1, c-Myc, N-cadherin, and β-catenin after knockdown of circCHD7 in 143B and HOS cells (Supplemental Figure S7A and 7B). Our results demonstrated that circCHD7 regulated MMP2, FZD4, and Wnt/β-catenin pathway-related genes. Further exploration of the roles of miR-608 in modulating gene expression showed that transfection with miR-608 mimic had the same effect (Fig. 7a and 7b). To assess whether circCHD7 regulated the Wnt/β-catenin pathway by sponging miR-608, we co-transfected cells with si-circCHD7 and miR-608 inhibitor. Interestingly, the downregulation of vimentin, MMP2, cyclin D1, c-Myc, N-cadherin, and β-catenin was partly reversed by transfection with si-circCHD7 and miR-608 inhibitor compared with si-circCHD7 alone (Fig. 7c and 7d). Thus, these results suggested that circCHD7 regulated MMP2, FZD4, and Wnt/β-catenin signaling molecules by sponging miR-608.
CircCHD7 regulated tumor growth by sponging miR-608 in vivo
Finally, we verified the role of circCHD7 in vivo using a xenograft tumor model using cells transfected with sh-circCHD7 or with sh-circCHD7 and miR-608. The efficiency of transfection was verified by FISH (Fig. 8a). The size, volume, and weight of tumors transfected with sh-circCHD7 were much smaller than those in the control group, and co-transfection with miR-608 sponge partly reversed these effects (Fig. 8b–d). Western blot and RT-qPCR analyses of protein and total RNA from tumors (Fig. 8e and Supplemental Figure S8A) revealed similar effects as observed in vitro. Furthermore, immunohistochemistry (Fig. 8f) showed that decreased staining of sh-circCHD7-transfected cells was partly rescued by treatment with the circCHD7 and miR-608 sponge. Overall, these results showed that circCHD7 regulated osteosarcoma by sponging miR-608 through the Wnt/β-catenin pathway in vivo (Fig. 8g).