The roles of circRFWD2 and circINO80 during NELL‐1‐induced osteogenesis

Abstract Bone defects caused heavy social and economic burdens worldwide. Nel‐like molecule, type 1 (NELL‐1) could enhance the osteogenesis and the repairment of bone defects, while the specific mechanism remains to be elucidated. Circular RNAs (circRNAs) have been found to play critical roles in the tissue development and serve as biomarkers for various diseases. However, it remains unclear that the expression patterns of circRNAs and the roles of them played in recombinant NELL‐1‐induced osteogenesis of human adipose‐derived stem cells (hASCs). In this study, we performed RNA‐sequencing to investigate the expression profiles of circRNAs in recombinant NELL‐1‐induced osteogenic differentiation and identified two key circRNAs, namely circRFWD2 and circINO80. These two circRNAs were confirmed to be up‐regulated during recombinant NELL‐1‐induced osteogenesis, and knockdown of them affected the positive effect of NELL‐1 on osteogenesis. CircRFWD2 and circINO80 could interact with hsa‐miR‐6817‐5p, which could inhibit the osteogenesis. Silencing hsa‐miR‐6817‐5p could partially reverse the negative effect of si‐circRFWD2 and si‐circINO80 on the osteogenesis. Therefore, circRFWD2 and circINO80 could regulate the expression of hsa‐miR‐6817‐5p and influence the recombinant NELL‐1‐induced osteogenic differentiation of hASCs. It opens a new window to better understanding the effects of NELL‐1 on the osteogenic differentiation of hASCs and provides potential molecular targets and novel methods for bone regeneration efficiently and safely.

emerged as a promising method in bone regeneration. 6,7 Human adipose-derived stem cells (hASCs), acting as the seed cells, have received widespread attention as a result of their minimally invasive acquisition and the ability of multiple differentiation. 8,9 How to induce the osteogenesis of hASCs effectively and safely, however, remains a research focus.
Nel-like molecule, type 1 (NELL-1), as a secreted osteogenic growth factor, was detected initially in patients diagnosed with unilateral coronal synostosis. 10 Nell-1 deficiency mice revealed major skeletal anomalies, including decreased calvaria mineralization, depressed vertebral disc and even perinatal death. 11 In turn, the overexpression of NELL-1 could promote the bone formation in vitro and in vivo models. [12][13][14] It is reported that NELL-1 was not only specific to the osteochondral lineage, 15,16 but also served as a critical functional mediator in the downstream of Runx2 which was essential for osteoblast differentiation. 17 These factors may contribute to fewer complications of NELL-1induced osteogenesis, especially in the heterotopic bone formation, while the specific mechanism remains to be elucidated. Circular RNAs (circRNAs) are generated by splicing during pre-mRNAs maturity. Unlike linear RNAs, they hold a covalently closed loop, keeping them stable and resistant to RNase R digestion. 18,19 High throughput sequencing has demonstrated that circRNAs were widely and abundantly distributed in the eukaryotic transcriptome with tissue-specific expression. 20 Functionally, circRNAs could regulate the protein function via cap-independent mechanism, affect the splicing of parental genes and serve as the microRNAs (miRNAs) sponges to cross-talk with miRNAs. [21][22][23] Emerging evidence has proven that cir-cRNAs played important roles in the development of normal tissues and various diseases, such as cardiovascular diseases, age-related diseases, autoimmune diseases and tumours. [24][25][26][27][28][29][30] Their functions as potential biomarkers were gradually discovered in cancers and neurodegenerative disorders. 31,32 However, it remains unclear that the comprehensive expression patterns of circRNAs and the roles of them played in the NELL-1-induced osteogenesis of hASCs.
In present study, we performed RNA-sequencing and compared the expression profiles of circRNAs between recombinant NELL-1-induced osteogenic differentiation group (NG) and general osteogenic differentiation group (OG). The potential genomic function was investigated by Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. Two key circRNAs, namely circRFWD2 and circINO80, were identified to be up-regulated during recombinant NELL-1-induced osteogenesis. Our results indicated that circRFWD2 and circINO80 could regulate the expression of hsa-miR-6817-5p and influence the recombinant NELL-1-induced osteogenesis of hASCs.

| Cell culture and recombinant NELL-1-induced osteogenic differentiation of hASCs
The hASCs were purchased from Cyagen company. All the hASCs were separated into two groups: recombinant NELL-1-induced osteogenesis group (NG) and general osteogenesis group (OG).
The OriCell™ general medium, which was utilized to culture hASCs, contained 10% mesenchymal stem cell-qualified foetal bovine serum (FBS), 1% glutamine and 1% penicillin-streptomycin. The osteogenic induction was initiated once the confluence of hASCs reached 80%. The general osteogenic inductive medium included basic medium, 10 nM dexamethasone, 10 mM β-glycerophosphate and 50 μg/mL vitamin C (Sigma-Aldrich), which was applied to culture the hASCs in OG. The hASCs in NG was culture in the general osteogenic inductive medium supplemented with 300 ng/mL of recombinant human NELL-1 protein (R&D Systems). 33 The cells were maintained in an incubator with 5% CO 2 at 37℃.

| Alizarin red s (ARS) staining
Alizarin red s staining was performed to detect the mineral deposition. Based on the protocol, the samples were fixed in 4% paraformaldehyde solution at 4°C for 20 minutes. After washed three times with distilled water, the samples were stained with 0.1% ARS for 20 minutes at 25℃.

| Alkaline phosphatase (ALP) staining and ALP activity
Alkaline phosphatase staining was implemented with the Leukocyte The Alkaline Phosphatase Assay Kit (Beyotime) was used for the analysis of ALP activity. The samples were cracked with RIPA lysis buffer, and supernatant was transferred to a 96-well plate. The 96well plate was incubated at 37℃ for 30 minutes after p-nitrophenol, and the reaction substrates were added. Finally, the ALP activity was calculated using p-nitrophenyl substrate at 405 nm.

| CircRNAs sequencing
The total RNAs were isolated from NG and OG with the Trizol reagent (Invitrogen). The RNA purity and concentration of samples were determined by NanoDrop ND-1000 (NanoDrop Thermo). The RNA integrity of samples was detected by denaturing agarose gel electrophoresis. In general, 3 units/μg of RNase R (Epicentre, Inc) were applied to treat 5μg RNAs for 15 minutes at 37°C to remove linear RNA. The rRNAs of the Rnase R-treated RNAs were depleted using Ribo-Zero Magnetic Gold Kit (Epicentre, Inc). HiSeq 4000 sequencer was used to acquire paired-end reads.

| CircRNAs profiling analysis
Edger software was utilized to identify the differentially expressed circRNAs among NG and OG. 34 The hierarchical clustering analysis was performed based on the significant differences. Any circRNAs, presenting fold changes >2.0 with P values < .05, were considered as significant differential expression.
The GO and KEGG analyses were used to predict the functions of differentially expressed circRNA-associated genes. GO analysis measured biological processes, cellular components and molecular functions. KEGG pathway analysis was used to identify pathways related to the target mRNAs of circRNAs. To investigate the potential functions of the differentially expressed circRNAs, the prediction software was utilized to show the interactions of these circRNAs with the targeted miRNAs. The prediction of miRNAbinding sites of the targeted mRNAs was performed on the basis of TargetScanHuman 7.2 and miRanda.

| Cell transfection
The mimic and the inhibitor of hsa-miR-6817-5p, miRNA control, circRFWD2 siRNA, circINO80 siRNA and control vector were synthesized by GenePharma Co. and shown in Table S1. Cells were transfected by Lipofectamine 3000 Reagent (Invitrogen), when the cell density reached 80% confluency.

| Quantitative real-time PCR
For the selected circRNAs, total RNAs (3 μg) were employed for first strand cDNA synthesis with dNTP Mix (HyTestLtd), RNase inhibitor (Enzymatics) and SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). The qRT-PCR was performed on an Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems) using SYBR Green master mix (Cloudseq). The primers of circRNAs and genes were synthesized by Sangon and shown in Table S2.
The cDNA synthesis and quantitative detection of miRNAs were performed with the miRNA qRT-PCR Detection Kit (GeneCopoeia).
The primer of hsa-miR-6817-5p was designed by GeneCopoeia. U6 was used for normalization. The relative expression was calculated by the formula 2 −ΔΔCt .

| Western blot
The protein levels of RUNX2 and bone sialoprotein (BSP) were de-

| Statistics
Quantitative data were expressed as means ± standard deviation (SD), and all the experiments were performed three times at least.
The statistical analysis was performed with SPSS 17.0 software. The differences between two groups were analysed by unpaired t test, while one-way analysis of variance (ANOVA) were utilized to identify the differences between more than two groups. P-value < .05 was considered as statistical significance.

| Recombinant NELL-1-induced osteogenic differentiation of hASCs
After induced for 7 days, the intensity of ALP staining was much more positive and the calcified nodules of ARS staining were apparently spotted in NG ( Figure 1A). The quantification of ALP activity were larger in NG than that in OG (P < .01) ( Figure 1B).
The expression of NELL-1 was significantly increased in NG, compared with OG (P < .01). Meanwhile, the expression levels of osteogenic markers, namely COL I and RUNX2, were up-regulated significantly in NG (P < .01, P < .01, respectively) ( Figure 1C). These results above suggested that NELL-1 could enhance osteogenesis of hASCs.  Figure 2B).

| Expression patterns of circRNAs in recombinant NELL-1-induced hASCs
The overview of differentially expressed circRNAs between NG and OG was displayed by heatmap after fold changes filtering ( Figure 2C). In total, 77 circRNAs were identified to express differentially, among which 19 circRNAs were up-regulated and 58 cir-cRNAs were down-regulated significantly. Four circRNAs, namely circRFWD2, circINO80, circHAGH and circDCBLD2, were selected based on their raw intensities, fold changes and P values. CircRFWD2 and circINO80 were up-regulated, while circHAGH and circDCBLD2 were down-regulated in NG. The results of qRT-PCR were consistent with RNA-sequencing ( Figure 2D).

| GO and KEGG pathways analyses of the host genes of circRNAs
Gene ontology analysis was performed to analyse the host genes of differentially expressed circRNAs. It contained three aspects, that is biological processes, cellular components and molecular function.
The top 60 enrichment GO analysis was shown in Figure S1. The most enriched biological processes terms were associated with the regulation of cell cycle process (GO:0010564), the organelle organization (GO:0006996) and the mitotic nuclear division (GO:0007067).
The most enriched cellular components terms were the intracellular part (GO:0044424), the nucleoplasm (GO:0005654) and the intracellular (GO:0005622). For molecular function aspect, the most enriched terms were related to the protein binding (GO:0005515), the guanyl-nucleotide exchange factor activity (GO:0005085) and the small GTPase binding (GO:0031267).
As for the KEGG pathway analysis, the top 39 pathways were listed according to the enrichment scores ( Figure S2). The most enriched pathways included Glutamatergic synapse (hsa04724), Thyroid hormones (THs) signalling pathway (hsa04919) and p53 signalling pathway (hsa04115).

| The effects of circRFWD2 and circINO80 on osteogenesis
After induced for 7 days, the expressions of circRFWD2 and cir-cINO80 were up-regulated in NG, when were compared with those in OG ( Figure 3A). Sanger sequencing confirmed the specific splicing sites of circRFWD2 and circINO80 ( Figure 3B).

F I G U R E 3
The effects of circRFWD2 and circINO80 on osteogenesis. A, Real-time qPCR was used to determine the expression levels of circRFWD2 and circINO80. B, Sanger Sequencing results of circRFWD and circINO80. C, ALP and ARS staining after 7 d of induction in si-circRFWD2 and si-circINO80 groups. D, Real-time qPCR was used to determine the expression levels of RUNX2 and OPN in si-circRFWD2 and si-circINO80 groups. E, Real-time qPCR was used to determine the expression levels of RUNX2 and OPN in si-circRFWD2 and si-circINO80 groups with NELL-1 induction. **P < .01 (scale bar = 200 μm) Transfection was conducted to knock down the expression of circRFWD2 and circINO80. After osteogenic induction for 7 days, the intensity of ALP and ARS staining decreased in both of cir-cRFWD2 and circINO80 knockdown groups ( Figure 3C). The outcomes of qRT-PCR indicated that the expression levels of RUNX2 and OPN were decreased in both of circRFWD2 and circINO80 knockdown groups, when compared to group NC ( Figure 3D).
To investigate the function of circRFWD2 and circINO80 on recombinant NELL-1-induced osteogenesis, the expressions of osteogenic markers were detected in circRFWD2 and circINO80 knockdown groups, respectively, after recombinant NELL-1-induced osteogenic differentiation. It showed that the expression levels of RUNX2 and OPN were reduced significantly in both of circRFWD2 and circINO80 knockdown groups, when compared with group NC ( Figure 3E).

| The effects of hsa-miR-6817-5p on osteogenesis
After transfected with hsa-miR-6817-5p mimic, the protein levels of BSP and RUNX2 were down-regulated markedly ( Figure 5A), and the intensity of ALP staining and ARS staining were obviously weaker in the mimic group ( Figure 5B). When inhibiting hsa-miR-6817-5p, we found the expression of BSP and RUNX2 were up-regulated ( Figure 5A). ALP staining and ARS staining were more positive in the inhibitor group ( Figure 5B).
After induced for 7 days, the expression level of hsa-miR-6817-5p was lower significantly in NG than that in OG (P < .01) ( Figure 5C). Inhibition of the expression of hsa-miR-6817-5p could partially reverse the suppressive effect of si-circRFWD2 ( Figure 5D) and si-circINO80 ( Figure 5E), respectively, on the expression of RUNX2.
F I G U R E 4 The co-targeted miRNAs of circRFWD2 and circINO80. A, Co-targeted miRNAs of circRFWD2 and circINO80. B, The binding sites of hsa-miR-6817-5p in circRFWD2 and circINO80. C, Real-time qPCR detected the expression level of hsa-miR-6817-5p in mimic, inhibitor, and NC group. D, Real-time qPCR detected the expressions of circRFWD2 and circINO80 in mimic, inhibitor and NC groups. *P < .05, **P < .01

| D ISCUSS I ON
CircRNAs are highly conserved among species due to the closed loop structures, 35 and they could be divided into five patterns, including exonic circRNAs, intronic circRNAs, intergenic circRNAs, sense overlapping circRNAs and antisense circRNAs, according to the different formation mechanisms. 19 It is reported that circRNAs could interact with RNA-binding protein and affect the protein function. 18 A few circRNAs could be directly translated into proteins. 36 Exonic-intronic circRNAs were associated with U1 small nuclear ribonucleoproteins, while intronic circRNAs were related to the polymerase II machinery, and both of them could regulate the host gene expression. 37 In addition, multiple circRNAs were known to harbour the binding sites of miRNAs, making them competently bind to miRNAs and buffer the inhibition of miRNAs on target genes, which played a critical regulatory role in the osteogenesis of mesenchymal stem cells. 38 Hsa_circRNA_33287 regulated the expression of Runx3 via targeting miR-214-3p to enhance the osteogenic differentiation of maxillary sinus membrane stem cells. 39 Hsa_circ_0002052 targeted miR-1205 to regulate the expression of APC2, which inhibited Wnt/beta-catenin signalling pathway. 40,41 Mm9_circ_009056 could act as a sponge of miR-22-3p, regulating the expression of BMP7 to promote bone formation. 42 CircRNA CDR1as could down-regulate miR-7 to activate p38 MAPK pathway, which enhanced the osteogenesis of periodontal ligament stem cells. 43 However, the roles of circRNAs in recombinant NELL-1-induced osteogenesis of hASCs remains unclear.
In this study, we performed GO analysis to annotate biological functions of parental genes of differentially expressed circRNAs.
The biological processes, in which most circRNAs were enriched, were the regulation of cell cycle process, the organelle organization and the mitotic nuclear division. The most enriched cellular components terms were the intracellular part, the nucleoplasm and the intracellular. For molecular function aspect, the most enriched terms were related to the protein binding, the guanyl-nucleotide exchange factor activity and the small GTPase binding. The KEGG pathway analysis suggested most of the differentially expressed circRNAs were enriched in the Glutamatergic synapse, THs signalling pathway and p53 signalling pathway. p53 signalling pathway was reported F I G U R E 5 The effects of hsa-miR-6817-5p on the osteogenesis of hASCs. A, Western blot detected the protein levels of RUNX2 and BSP in hsa-miR-6817-5p mimic, inhibitor and NC groups. B, The ALP and ARS staining of hASCs in hsa-miR-6817-5p mimic, inhibitor and NC groups. C, Real-time qPCR detected the expression of hsa-miR-6817-5p in OG and NG. D, The expression levels of RUNX2 of hASCs after co-transfection with si-NC or si-circRFWD2 plus hsa-miR-6817-5p mimic, inhibitor or NC groups. E, The expression levels of RUNX2 of hASCs after co-transfection with si-NC or si-circINO80 plus hsa-miR-6817-5p mimic, inhibitor or NC groups. **P < .01. (scale bar = 200 μm) to regulate bone formation negatively in vivo and in vitro, under the control of BMPs and IGFs pathways. 44 It has shown that p53 could suppress the expression of osterix (Osx) to influence osteoblastic differentiation. 45 Meanwhile, p53 could affect the cell fate of mouse bone marrow-derived mesenchymal stem cells and activate miRNA-34 family to repress the expression of Runx2. 46 It was known that THs played roles in the endochondral ossification by affecting chondrocytes and osteoblasts directly, and regulated the initiation and progression of the secondary ossification centre through promoting the chondrocytes to differentiate into bone matrix-producing osteoblasts via activating Indian hedgehog and Osx expression. 47 NR2F1, one of the steroid/THs nuclear receptor superfamily members, could impair the osteoblast differentiation, which could be rescued by BMP-2. 48 Therefore, the circRNAs identified might influence the recombinant NELL-1-induced osteogenesis of hASCs via p53 signalling pathway and THs signalling pathway, cross-talking with BMPs signalling pathway.
Additionally, circRFWD2 and circINO80 were identified to be up-regulated during recombinant NELL-1-induced osteogenesis by RNA-seq, which was validated by qRT-PCR. When inhibiting cir-cRFWD2 and circINO80, respectively, we found the osteogenic markers, RUNX2 and OPN, were down-regulated obviously. The ALP activity and calcium nodules staining were further confirmed the results above. During recombinant NELL-1-induced osteogenesis, the expression of RUNX2 and OPN were also decreased significantly in both of circRFWD2 and circINO80 knockdown groups than group NC. Therefore, it is suggested that these two circRNAs might influence recombinant NELL-1-induced osteogenesis positively.
Our study showed that hsa-miR-6817-5p was down-regulated during the recombinant NELL-1-induced osteogenesis, and it could inhibit the recombinant NELL-1-induced osteogenic differentiation.
It is suggested that hsa-miR-6817-5p was closely associated with hsa_circ_0061012, which was up-regulated in psoriasis significantly.
Psoriasis was related well to the regulation of NF-kappaB, 49 whose activity could be weakened by NELL-1. 50 Moreover, we found that the binding sites of hsa-miR-6817-5p could match the sequences of circRFWD2 and circINO80, respectively, and the expression levels of these two circRNAs were negatively correlated with hsa-miR-6817-5p. Inhibition of hsa-miR-6817-5p expression could partially reverse the suppressive effect of si-circRFWD2 and si-circINO80 on the osteogenesis. It is suggested that circRFWD2 and circINO80 could regulate hsa-miR-6817-5p and promote recombinant NELL-1-induced osteogenesis of hASCs.
Although we tried the best to design our experiments, there are some limitations in this study. It was difficult to obtain the clinical samples of unilateral coronal synostosis, where NELL-1 was detected to be over-expressed, and thus, we did not verify the expression profiles of CircRFWD2 and circINO80 axis in the clinical specimen data.
Furthermore, some interesting and valuable research issues, for example, whether and how NELL-1 could regulate miR-6817-5p, are need to be studied. Therefore, we would focus on miming and enriching the mechanism of circRFWD2 and circINO80 axis in recombinant NELL-1induced osteogenesis, and this study opened the window.

| CON CLUS ION
circRFWD2 and circINO80 were identified to up-regulated during the recombinant NELL-1-induced osteogenesis. Mechanistically, circRFWD2 and circINO80 could regulate hsa-miR-6817-5p and influence recombinant NELL-1-induced osteogenic differentiation of hASCs. Therefore, circRFWD2 and circINO80 promise to be the potential molecular targets for the regulation of osteogenesis and bone regeneration.

ACK N OWLED G M ENTS
This study was supported by grants from the National Natural Science

CO N FLI C T O F I NTE R E S T
The authors declare that there are no conflicts of interest.

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.