Circular RNA CircFgfr2 Promotes Osteogenic Differentiation of Rat Dental Follicle Cells via Sponging MicroRNA miR-133a-3p and Up-regulating Distal-less Homeobox DLX3

Abstract


Conclusion
We demonstrated that circFgfr2 promoted the osteogenic differentiation of rDFCs via circFgfr2 / miR-133a-3p / DLX3 / RUNX2 axis.CircFgfr2 has the potential to be the molecular target for the bone tissue engineering.

Introduction
The dental follicle is a loose connective tissue surrounding the tooth germ responsible for cementum, periodontal ligament, and alveolar bone formation in tooth development.Dental follicle cells (DFCs) are a group of heterogeneous cells from dental follicle and have the capacity for self-renewal and the potential for multi-directional differentiation [1,2] .In 2002, dental follicle cell was rstly isolated by Zhao et al. [3] from the mice molar region and differentiated toward osteoblast in vitro with exogenous bone morphogenetic protein 2 (BMP2) induction.In 2005, Morsczeck et al. [4] found precursor cells isolated from dental follicle of human third molar teeth expressed putative stem cell markers Notch Receptor 1 (Notch1) and Nestin.In 2010, Honda et al. [5] reported new bone formation after DFCs transplantation in calvarial defects created in immunode cient rats.Tsuchiya et al. [6] found that there were no statistically signi cant differences between DFCs and Mesenchymal stem cells (MSCs) with respect to capacity for osteogenic ability.Compared with other dental-derived stem cells like dental pulp stem cells (DPSCs), periodontal ligament stem cells (PDLSCs), stem cells from exfoliated deciduous teeth and stem cells from apical papilla, DFCs exhibited strong proliferative capacity, superior pluripotency, and high immunosuppressed effect [7] .Due to these properties, DFCs has the potential for clinical application in bone regeneration treatment.
Circular RNAs (circRNAs) are produced by a speci c splicing called "backsplicing" which is different from the canonical splicing of linear RNAs [8] .Unlike linear RNAs, circRNAs lack 5'caps and 3'tails and form closed loop structures [9] .CircRNAs almost located in the cytoplasm and are highly conserved and tissuespeci c [10] .In 1976, Sanger et al. [11] rstly found circRNAs in RNA viruses.CircRNAs were originally thought to be errors in RNA splicing.On account of high-throughput sequencing technologies and bioinformatics progression, large numbers of circRNAs are reported to be endogenous, abundant, conserved and stable in cells.Researchers found circRNAs function as competing endogenous RNAs or miRNAs sponges and modulate the expression of target genes [12] .CircRNAs can competitively bind with miRNAs to reverse the inhibition of miRNAs on target genes.The mechanism is called "competing endogenous RNAs (ceRNAs)".It is becoming evident that circRNAs may be involved in many diseases [13] such as diabetes [14] , cardiovascular diseases [15] , cancers [16] , etc.
There is increasing evidence that circRNAs function during osteogenic differentiation of periodontal tissue (or dental pulp) derived progenitor cells [17] .Chen et al. [18] found 43 upregulated circRNAs and 144 downregulated circRNAs during the differentiation process of human dental pulp stem cells (hDPSCs).Ji et al. [19] revealed that has_circ_0026827 was upregulated during osteogenic differentiation of hDPSCs in vitro and overexpression of has_circ_0026827 promoted bone formation in vivo.Gu et al. [20] found that circRNA BANP and circRNA ITCH may interact with miR-34a and miR-146a to promote PDLSCs osteogenic differentiation via the mitogen-activated protein kinase (MAPK) pathway.Hence, we proposed to explore the difference in circRNA expression during osteogenic differentiation of rDFCs.We have previously found that the level of circFgfr2 signi cantly increases after osteogenic differentiation induction for 28 days in rDFCs [21] .Therefore, our target is to further investigate the mechanism of circFgfr2 in rDFCs osteogenesis.

Ethics Statement
The study was conducted in accordance with the National Institutes of Health guidelines on the ethical use of animals.All experiments were performed with the approval of the Ethics Committee of Sun Yat-sen University.The experimental rats were purchased from the Laboratory Animal Center of Southern Medical University (License No: SCXK [Yue] 20160041).

Sanger Sequencing and Agarose gel Electrophoresis
According to the circRNA circularization site, speci c divergent primers were designed.Sanger sequencing were carried out on the ampli ed products of divergent primers.Total RNA of rDFCs was extracted by Trizol reagent with or without RNaseR to obtain cDNA through inverse transcription of arbitrary primers.Then, cDNA was ampli ed by convergent primers and divergent primers.Agarose gel electrophoresis of the PCR product was carried out.

Fluorescence in situ Hybridization (FISH)
The probe labeled with CY3 were constructed for the circFgfr2 sequence and used for in situ hybridization.rDFCs were xed with 4% paraformaldehyde for 5 min at 4 °C, treated with 0.5% Triton X-100 15 min at room temperature, incubated with 100% alcohol for 1 min.Dehydration and air drying.The probe was prepared and the hybridization buffer with the probe was denatured at 88 °C for 5 min.After hybridization, the slide was washed with 2×SSC, 5 min, at 42 °C and then 2×SSC was washed at room temperature, 5 min×2.Inhale 50 µL DAPI-Antifade solution on the slide, cover the slide and incubate at dark room temperature for 20 min.Observation under uorescence microscope.

Cell transfection
The full-length cDNA of circFgfr2 was ampli ed from rDFCs cloned into the speci c vector pLC5-ciR (Geneseed, Guangzhou) between the BamHI and EcoRI sites for circFgfr2 over-expression (pLC5-ciR-circFgfr2).The mock plasmid pLC5-ciR (NC) without the circFgfr2 cDNA served as a control.Immuno uorescence was carried out for the veri cation of transfection according to the manufacturer's instructions.To over-express miR-133a-3p, miR-133a-3p mimics and NC-mimics were purchased from ThermoFisher.The rDFCs were transfected with the above plasmids and mimics according to the requirements of each experiment by using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions.After 48h, the rDFCs were harvested and used for further research.

RNA high-throughput sequencing
Total RNA was extracted from rDFCs of the circFgfr2 overexpression group and control groups.Then, the messenger RNA was puri ed by polyA selection, chemically fragmented and converted into singlestranded cDNA via random hexamer priming.After that, the second strand is generated to create doublestranded cDNA.TruSeq libraries were prepared using TruSeq DNA Library Preparation Kits (illumina, San Diego, CA) following with quality control using software Fastp (Shenzhen, China).During quality control, the sequence of the joint and the three bases at the beginning and end were removed.Then, 4bp width window was used to retrieve the sequencing read segments, and the base quality lower than Q30 was removed.Finally, the read segment with at least 50bp length is selected for the downstream analysis process.The data after quality control was compared to the Rat genome (HISAT2 index: Ensembl rnor_6.0Genome_TRAN) using software HISAT2 V.2.0.5 (https://ccb.jhu.edu/software/hisat2/index.shtml).The result (BAM format le) is then compared and sorted by chromosome position.Through StringTie V.1.3.3bsoftware (http://ccb.jhu.edu/software/stringtie/),we detected and quanti ed the known protein encoding gene transcripts.Differentially expressed genes were analyzed using edgeR software (https://bioconductor.org/packages/release/bioc/html/edgeR.html).Firstly, the relative expression level was generated according to the original count of the gene, and the unit was CPM (Counts Per Million).Then differentially expressed genes were were calculated and considered statistically signi cant when P-value<0.05 and |log(FC)|≧1.Metascape gene annotation and analysis resource (https://metascape.org/gp/#/main/step1), the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/)and Cytoscape V3.6.0 software (The Cytoscape Consortium, San Diego, CA) were used for gene annotation, enrichment and pathway analysis.

Real-time quantitative polymerase chain reaction (qRT-PCR)
Total RNA was extracted through Trizol reagent (Invitrogen).Total RNA was reverse transcribed to cDNA using Geneseed ® II First Strand cDNA Synthesis Kit (Geneseed, China).Relative mRNA and miRNA expression levels were detected by qRT-PCR using Geneseed ® qPCR SYBR ® Green Master Mix (Geneseed, China).GAPDH or U6 was used as an internal control.All reactions were run in triplicates and fold expression changes calculated via the comparative 2 -∆∆Ct method.Table 1 lists PCR primers used in this study.Primers were synthesized by Forevergen Co, Ltd (Guangzhou, China).

TABLE 1
The primers used in this study.
48h after the co-transfection, luciferase activity was measured via Dual-Glo@Luciferase Assay System E2940 (Promega, USA) according to the manufacturer's instructions.Fire y luciferase activities were normalized to Renilla luminescence in each well.All the experiments were independently repeated three times.
Chromatin Isolation by RNA Puri cation (ChIRP) circFgfr2 anti-sense DNA probes with biotin labeling at 3-prime end were designed and produced by Geneseed Biotech (Guangzhou, China): 1.GTGCTCCAACAACATCAAGG-Bio; 2.GTACGGTGCTCCAACAACAT-Bio. ChIRP analysis was performed according to protocols published by Chu et al [22] .LacZ probes were used as a non-speci c probe.Total RNA was collected, and qRT-PCR was performed to con rm circFgfr2 probes and interaction of circFgfr2 with miR-133a-3p and miR-133a-5p.
Alkaline phosphatase activity detection and staining ALP activity was assessed using AKP/ALP test kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China), according to manufacturer's protocol.Brie y, cells were seeded in 96-well plates, cell lysates were collected, samples were incubated at room temperature for 5-10 min, and dilution buffer was added to 100 µL of the sample in a 96-well plate.Finally, uorescence was measured at OD 405 nm.rDFCs were xed in 4% paraformaldehyde for 10 min, and then ALP staining was performed by ALP staining kit (Beyotime, Shanghai, China) according to the manufacturer's protocol.

Alizarin red staining
RDFCs were washed twice with PBS and xed with 4% paraformaldehyde for 10 min, and then stained with Alizarin red staining solution (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) for 30 min at room temperature.

Western blot analysis
Radioimmunoprecipitation assay lysis buffer (Fude, China) was used to extract proteins from cells.Protein concentrations were measured using a BCA protein assay kit (Abcam, USA), according to manufacturer's protocols.Each sample (20 μg) was mixed with 5x loading buffer (Cell Signaling Technology, USA), and boiled for 10 min.After separation using SDS-PAGE and 4%-20% gradient gels, the proteins were transferred to 0.22 μm polyvinylidene di uoride membranes (Millipore, USA).The nonspeci c binding sites were blocked with 5% (wt/vol) skim milk for 120 min.The membranes were incubated with the primary antibody of RUNX2 (1:1000, abcam, USA) at 4 ℃ overnight.Subsequently, the membranes were incubated with horseradish peroxidase A niPure goat anti-mouse IgG secondary antibody (Emarbio, Beijing, China) at room temperature for 1 h.An enhanced chemiluminescence kit (Millipore Corp, Bedford, MA) was used for imaging.

Statistical analysis
Results are presented as mean ± standard deviation.Statistical signi cance was determined using twotailed t-test.All data were analyzed using Prism 7.04 (GraphPad Software, USA).P<0.05 was considered statistically signi cant.

Culture and Characterization of rDFCs and Location of circFgfr2 in rDFCs
The rDFCs had a polygon or long spindle morphology (Figure 1A).Flow cytometry analysis showed rDFCs were positive for MSC markers CD29, CD90, CD44, but were negative for CD34 and CD45 (Figure 1B).CircFgfr2 derived from FGFR2 gene exons5 and exons6 is located on chr1: 200648164-200658087 and 9630 bp in length (Figure 1C).Sanger sequencing and agarose gel electrophoresis con rmed the closed loop structure of circFgfr2 and resistance to RNaseR digestion (Figure 1D).The FISH examination showed that circFgfr2 primarily had a cytoplasmic location in rDFCs (Figure 1E).
Differentially expressed mRNAs and function enrichment analysis suggest circRNAs participant in osteogenic differentiation of rDFCs CircFgfr2 overexpression vector (pLC5-ciR-circFgfr2) was constructed and rDFCs were transfected with the vector which con rmed by immuno uorescence (Figure 2A).The mRNA sequencing results illustrated that 361 mRNAs were upregulated, and 391 mRNAs were downregulated (Figure 2B, C).Furthermore, GO analysis revealed that differentially expressed mRNAs are associated with several biological process including cellular process, biological regulation, metabolic process, multicellular organisms process, development process and so on (Figure 3A).GO pathway enrichment analysis revealed 181 pathways and process including IL-17 signaling, response to mechanical stimulus, positive regulation of MAPK and biomineralization.Heatmap was plotted by http://www.bioinformatics.com.cn, a free online platform for data analysis and visualization (Figure 3B, C, D).Moreover, KEGG pathway analysis indicated that 54 pathways were related to the differentially expressed mRNAs, including MAPK signaling, cAMP signaling, Calcium signaling and TNF signaling which are important for osteogenic differentiation (Figure 3E).Thus, function enrichment analysis indicated that circFgfr2 may play a important role in osteogenesis.
CircFgfr2 Promotes the Osteogenic Differentiation of rDFCs and Functions as a Sponge of miR-133a-3p in rDFCs Increasing evidence indicated that circRNAs can suppress the levels of functional miRNA by serving as miRNA sponges.The expression levels of miR-133a-3p and miR-133a-5p in rDFCs after transfection with pLC5-ciR-circFgfr2 were examined by qRT-PCR.Upregulated circFgfr2 signi cantly inhibited miR-133a-3p and miR-133a-5p mRNA levels (P < 0.001; Figure 4A).In order to identify the overexpression of circFgfr2, qRT-PCR was used to detect the expression levels of circFgfr2 in rDFCs.Compared with the group transfected with pLC5-ciR-circFgfr2, the overexpression (OE) group expressed a higher level of circFgfr2 (P < 0.001; Figure 4B).We previously found [21] circFgfr2 was remarkably increased during the osteogenic differentiation of rDFCs and speculated that circFgfr2 might stimulate the osteogenic differentiation.So, we evaluated the levels of osteogenesis markers including Runt-related transcription factor 2 (RUNX2), osteocalcin (OCN), Osterix (OSX), bone morphogenetic protein2 (BMP2) and BMP6.The mRNA expression levels of RUNX2, OCN and Osterix were upregulated signi cantly in circFgfr2 over-expression groups, when compared with NC group (Figure 4C).Moreover, the data showed that overexpression of circFgfr2 signi cantly upregulated BMP2 and BMP6 mRNA levels (P < 0.01; Figure 4D).These data imply that circFgfr2 promotes the osteogenic differentiation of rDFCs and positively regulates the expression of osteogenesis-related genes.Therefore, we investigated the potential mechanism by which circFgfr2 promoted the osteogenesis of rDFCs.MiR-133a has been proven to inhibit the osteogenic differentiation of vascular smooth muscle cells [23] and we found the levels of miR-133a-3p and miR-133a-5p decreased in circFgfr2 overexpressed rDFCs.Bioinformatics analysis revealed that miR-133a-3p has two binding sites and miR-133a-5p has one binding site for circFgfr2 (Figure 4E).The potential interaction models between miR-133a-3p, miR-133a-5p and circFgfr2 were predicted via RegRNA (Figure 4F).DLR assay showed that both miR-133a-3p and miR-133a-5p can effectively inhibit circFgfr2 luciferase activity (P < 0.001) and the inhibitions disappeared when the binding sites are mutated (Figure 4G, H).However, ChIRP assay illustrated that compared with LacZ, only miR-133a-3p was obviously enriched in the pellet pulled down by circFgfr2 probes (P < 0.05, Figure 4I).Hence, DLR and ChIRP assay showed that circFgfr2 functioned as sponge to miR-133a-3p.

Discussion
In recent years, researchers have reported the participation of circRNAs in the progression of many diseases [13] .However, the role and mechanism of circRNAs in osteogenesis stilled remain to be clari ed.In this study, we found that circFgfr2 overexpression promoted the osteogenesis and circFgfr2 can act as a sponge for miR-133a-3p and reverse the inhibition of miR-133a-3p on DLX3 and RUNX2 in rDFCs.Thus, circFgfr2 has the potential to be the molecular target for the bone tissue engineering.
We previously found circFgfr2 upregulated 7.7 times in rDFCs during osteogenic differentiation via RNA high-throughput sequencing [21] .In order to further investigate the function of circFgfr2 in osteogenesis, plasmids containing circFgfr2-overexpresssion fragments were transfected into rDFCs and mRNA highthroughput sequencing analysis was carried out.The mRNA sequencing results illustrated that 361 mRNAs were upregulated, and 391 mRNAs were downregulated.GO analysis revealed that differentially expressed mRNAs in circFgfr2 overexpressed rDFCs are associated with several biological process.Moreover, GO and KEGG pathway analyses revealed that mitogen-activated protein kinase (MAPK) signaling was enriched which consistent with our prior studies.The MAPK signaling pathway involves a variety of cellular biological processes, including proliferation, differentiation, and migration.ERK, JNK, and p38 MAPK are some of the main sub-families [25] .Jin et al. [26] reported that ERK/p38 MAPK signaling pathway is an important factor for regulating osteogenic differentiation in hPDLSCs.The activation of the MAPK signaling pathway by phosphorylation of p38 and ERK1/2 with BMP9 increases ALP activity and calcium deposition in rDFCs [27] .Meng et al. [28] revealed the inhibition of JNK and p38 MAPK pathways decreased the expression of osteoblast-related genes (ALP, OPN, RUNX2) in dental follicle cells.Along with former studies, our results demonstrated that circFgfr2 participates in MAPK pathway during osteogenic differentiation, however, the exact mechanism needs further research.
Studies have demonstrated that miRNAs act as a regulator of osteogenic differentiation.Zhao et al. [29] transfected miR-129-5p mimics into BMSCs and found the expression of osteogenic genes (Runx2, Bmp2, and OCN) improved which indicated miR-129-5p promoted osteogenic differentiation of BMSCs and bone regeneration.On the contrary, Hu et al. [30] transfected BMSCs with miRNA-132-3p mimics or inhibitors.The upregulation of miRNA-132-3p expression inhibited osteogenic differentiation, whereas the downregulation of miRNA-132-3p expression enhanced osteogenic differentiation.Moreover, miR-100-5p overexpression inhibited the osteogenesis of hBMSCs and silencing miR-100-5p expression rescued the reduction in osteogenesis [31] .MiRNAs regulate gene expression via inhibiting mRNA translation or promoting mRNA degradation following binding to the 3'untranslated of target genes [32] .We speculated whether miRNAs played a positive or negative role in osteogenesis is determined by downstream targets.In this study, we present evidence supporting DLX3 as a target of miR-133a-3p using bioinformatics analysis.DLX3 is a member of the DLX family transcription factors and plays an important role in osteogenic differentiation.DLX3 mRNA is expressed at relatively high levels in osteoblasts and can stimulate osteoblastic differentiation [33] .Northern blot showed transduction with DLX3 containing virus improved the expression of bone sialoprotein (BSP) and osteocalcin (OC) in MSCs.Sun et al. [34] revealed that DLX3 overexpression promoted the expression of ALP, RUNX2, OSX and OCN and induced the formation of calci ed matrix in BMSCs.However, Duverger et al. [35] provided the opposite evidence in vivo.Neural crest deletion of DLX3 increased bone formation and mineralization in craniofacial bones which suggested DLX3 inhibited osteoblastic differentiation.Consistent with our study, Qadir et al. [24] found two binding sites for miR-133a in DLX3 3'-UTR by performing bioinformatics analysis and luciferase reporter assays.Then C2C12 premyoblast cells were transfected with miR-133a mimics in the myogenic medium and western blotting indicated that DLX3 protein expression decreased.We found miR-133a-3p overexpression suppressed DLX3 expression and osteogenesis in rDFCs, and miR-133a-3p inhibition by circFgfr2 overexpression reversed the suppression.Therefore, miR-133a-3p plays a negative role in the regulation of osteogenic differentiation via targeting DLX3.
RUNX2 is a BMP response gene essential for early bone formation [36] .DLX3 regulates the transcription of RUNX2 by binding to its promoters during osteogenic differentiation [24] .Issac et al. [37] carried out RNA sequencing and chromatin immunoprecipitation-Seq analyses and found DLX3 regulated RUNX2 which is crucial for bone formation.In this study, transfection of miR-133a-3p mimics suppressed the osteogenic differentiation and RUNX2 protein expression in rDFCs.Zhou et al. [38] inhibited the expression of miR-133a with ibandronate and found the mRNA level of ALP, type I collagen (COL-1), osteoprotegerin (OPG), OCN, and Runx2 enhanced in PDLSCs.Liao et al. [23] revealed overexpression of miR-133a suppressed the osteogenic differentiation of vascular smooth muscle cells and caused a decrease in alkaline phosphatase activity, osteocalcin secretion and Runx2 expression.Zhang et al. [39] carried out 3'UTR luciferase reporter, immunoblotting, and mRNA stability assays and found Runx2 protein accumulation reduced when transfected with miR-133a in osteoblasts and chondrocytes.Hence, miR-133a-3p inhibits RUNX2 expression via regulating DLX3 which regulates RUNX2 transcription.
Recent studies revealed that circRNAs have been involved in mandiseases such as diabetes mellitus [40] , neurological disorders [41] , cardiovascular diseases [42] and cancer [43] .Abundant circRNAs exert important biological functions by acting as microRNA or protein inhibitors ('sponges'), by regulating protein function or by being translated themselves.However, most of circRNAs which have investigated were proposed to act as miRNA sponges [44] .Liu et al. [45] found circ_AFF4 stimulated Irisin expression through complementary binding to its downstream target molecule miR-135a-5p in BMSCs.In our study, FISH analysis showed that circFgfr2 primarily had a cytoplasmic location.Furthermore, using dual-luciferase reporter assay and ChIRP assay, we found circFgfr2 bond to miR-133a-3p and regulated its function.
Chen et al. [46] revealed that circFgfr2 overexpression accelerated the myogenic differentiation and the formation of myotubes via eliminating the inhibition effect of miR-133a-5p and miR-29b-1-5p in chicken myoblasts.We found circFgfr2 overexpression promoted the osteogenic differentiation by sponging miR-133a-3p in rDFCs which indicated circFgfr2 was a positive regulator during osteogenic differentiation.However, whether circFgfr2 expedites osteogenesis and bone formation in vivo need to be con rmed in future studies.
This study indicates circFgfr2 acts as a sponge for miR-133a-3p and positively regulates the osteogenic differentiation in rDFCs.The circFgfr2 / miR-133a-3p / DLX3 / RUNX2 axis may be a new therapeutic strategy for the treatment of bone regeneration.

Conclusion
We concluded that circFgfr2 promoted the osteogenic differentiation of rDFCs via sponging miR-133a-3p.Our study indicated that a circFgfr2 / miR-133a-3p / DLX3 / RUNX2 axis may serve as a potential therapeutic target for improving bone regeneration.Therefore, circFgfr2 has the potential to be the molecular target for the bone tissue engineering.JL participated in the high-throughput sequencing.HJ, YD, YN conceived of the study, and participated in its design and coordination and helped to draft the manuscript.All authors read and approved the nal manuscript.Figure 4 circFgfr2 promotes the osteogenic differentiation of rDFCs and functions as a miRNA sponge and negatively regulates miR-133a-3p in rDFCs.(A) The expression levels of miR-133a-3p and miR-133a-5p of rDFCs after transfection with pLC5-ciR-circFgfr2.(B) qRT-PCR was used to determine the expression levels of (C) qRT-PCR was used to determine the expression levels of RUNX2, OPN and osterix.(D) qRT-PCR was used to determine the expression of BMP2 and BMP6.(E) The potential binding site sequence of miR-133a-3p, miR-133a-5p on and mutant sequence for DLR.(F) The potential interaction model between miR-133a-3p, miR-133a-5p and circFgfr2 from RNAhybrid.(G) Relative luciferase activity in rDFCs was detected with luciferase assays after with miR-133a-3p mimics.(H) Relative luciferase activity in rDFCs was detected with luciferase assays after transfected with miR-133a-5p mimics.(I) ChIRP assay was conducted to prove the interaction between miR-133a-3p or miR-133a-5p and circFgfr2 in rDFCs and followed by qRT-PCR to detect circFgfr2 expression level.*P < 0.05, **P < 0.01, ***P < 0.001.

Figure 1 Characteristics
Figure 1