CircFOXP1/FOXP1 promotes osteogenic differentiation in adipose‐derived mesenchymal stem cells and bone regeneration in osteoporosis via miR‐33a‐5p

Abstract Osteoporosis (OP) is defined by bone mass loss and structural bone deterioration. Currently, there are no effective therapies for OP treatment. Circular RNAs (circRNAs) have been reported to have an important function in stem cell osteogenesis and to be associated with OP. Most circRNA roles in OP remain unclear. In the present study, we employed circRNA microarray to investigate circRNA expression patterns in OP and non‐OP patient bone tissues. The circRNA‐miRNA‐mRNA interaction was predicted using bioinformatic analysis and confirmed by RNA FISH, RIP and dual‐luciferase reporter assays. ARS and ALP staining was used to detect the degree of osteogenic differentiation in human adipose‐derived mesenchymal stem cells (hASCs) in vitro. In vivo osteogenesis in hASCs encapsulated in collagen‐based hydrogels was tested with heterotopic bone formation assay in nude mice. Our research found that circFOXP1 was significantly down‐regulated in OP patient bone tissues and functioned like a miRNA sponge targeting miR‐33a‐5p to increase FOXP1 expression. In vivo and in vitro analyses showed that circFOXP1 enhances hASC osteogenesis by sponging miR‐33a‐5p. Conversely, miR‐33a‐5p inhibits osteogenesis by targeting FOXP1 3′‐UTR and down‐regulating FOXP1 expression. These results determined that circFOXP1 binding to miR‐33a‐5p promotes hASC osteogenic differentiation by targeting FOXP1. Therefore, circFOXP7ay prevent OP and can be used as a candidate OP therapeutic target.


| INTRODUC TI ON
Osteoporosis (OP) is a common bone disease identified by low bone mass and defective bone structure, leading to bone fragility or high risk of fracture. 1 It develops as an age-related bone malady and is usually followed by menopause in women and occurs later in life among men. OP seriously affects patient life quality and leads to considerable financial burden. 2 Some progress has been made in preventing OP and developing therapeutic methods for reducing fracture risk. However, currently available treatments do little to cure OP completely. 3,4 With the rapid development of cell-based therapy, mesenchymal stem cells (MSCs) have become the focus of new treatments for osteoporosis. Recently, adipose-derived stem cells (ASCs) have become a popular cell source of MSCs because of their minimally invasive acquisition, greater abundance and high production. 5 Ye et al found that ASCs increased bone mineral density and new bone formation in an OVX-induced osteoporotic rabbit model. 6 However, lots of molecules and signalling pathways involved in osteogenic differentiation of ASCs remain unknown, and considerable research is needed to reveal the associated mechanisms and to induce ACS osteogenesis effectively and safely for manipulating ASC-based cell therapy in osteoporosis treatment. In recent years, more studies have demonstrated that circular RNAs (circRNAs) have a critical function in regulating various signalling pathways related to osteogenesis. 7 Therefore, circRNAs are regarded as a new focus point in bone research. The circRNAs belong to a particular class of noncoding RNA with a covalently closed loop structure. They are found in eukaryotic transcriptome and play various roles in multiple physiological and pathological processes. 8,9 The essential physiological circRNA functions include protein translation templates, miRNA sponges and parental expression regulation. 10,11 Recent studies have revealed that mRNAs, lncRNAs, circRNAs and miRNAs are abnormally expressed in peripheral blood lymphocytes in postmenopausal OP patients and are correlated with disease occurrence. 12 Several circRNAs have been identified to be essential in regulating stem cell osteogenic differentiation. 13,14 For instance, circPOMT1 and circMCM3AP inhibit human adipose-derived stem cell (hASC) osteogenic differentiation by targeting hsa-miR-6881-3p via the bone morphogenetic protein (BMP) signalling pathway. 15 Moreover, circRNA runt-related transcription factor 2 (circRUNX2) is down-regulated in OP patient bone tissues and enhances human bone mesenchymal stem cell (hBMSC) osteogenic differentiation by sponging miR-203 and regulating the expression of RUNX2 and osteocalcin (OCN, as a late marker for bone formation). 16 As a master regulator for osteogenesis, Runx2 plays a key role in co-ordinating multiple signalling pathways involved in osteoblast differentiation. 17,18 These studies suggest that circRNAs might take part in the osteogenesis process.
Studies have revealed that multiple microRNAs (miRNAs) play a pivotal role in bone homeostasis such as MSC differentiation, survivability and apoptosis, adipogenesis and osteogenesis. 19,20 Previous studies have shown that miR-21 promoted osteogenic differentiation of MSCs derived from human umbilical cord, 21 but enhanced adipogenic differentiation in ASCs, 22 suggesting that the same miRNA may play different roles in different MSCs. It is known that in bone tissue, miR-33a plays a role in osteosarcoma chemoresistance by down-regulating TWIST 23 and in osteoblast differentiation after mechanical stimulations. 24,25 Furthermore, miR-33a, which is the only one miR-33 isoform, is expressed in mice and conserved in humans. Human miR-33a has two subtypes, miR-33a-3p and miR-33a-5p, which correspond to miR-33-3p and miR-33-5p in mice, respectively. The miR-33a family (3p and 5p) has recently been found to be a possible modulator of YAP/TAZ during hMSCs osteoblast differentiation. 26 However, its role in osteogenic differentiation of ASCs still needs to be identified.  32 In this study, the role of circ-FOXP1 was investigated in hASC osteogenic differentiation and bone regeneration. It was discovered that circFOXP1 might act like an miR-33a-5p sponge to up-regulate FOXP1 expression and consequently promote osteogenesis. These findings enhance the current understanding of circRNA and miRNA function in osteogenesis. In addition, circRNA might be a key therapeutic target in OP patients.

| Clinical samples
Samples of trabecular bone were acquired from the femoral trochanteric region located far from the periarticular bone in OP and non-OP patients undergoing hip arthroplasty for a fractured femoral neck. A total of 20 OP specimens were collected from ten females (age range: 60-87, average age: 73) and ten males (age range: 55-83, average age: 71). A total of 20 non-OP specimens from ten females (age range: 56-84, average age: 69) and ten males (age range: 52-85, average age: 70) who were suffering from external traumatic fracture were collected for the control group. Non-OP patients were known to have not suffered from any chronic condition or disease that may have affected their skeletal tissue.
Before enrolling in the study, none of the patients have experienced any medical therapy that affected their mineral metabolism or bone tissue. The Ethics Committee of Zhoushan Hospital of Traditional Chinese Medicine Affiliated to Zhejiang Chinese Medical University has approved the research protocol, and each patient signed the informed written consent form prior to their participation in this research.

| CircRNA microarray analysis
Sample preparation and microarray hybridization were performed following a standard protocol (Arraystar, Rockville, MD, USA).
Briefly, in order to eliminate linear RNAs and improve circRNAs, total RNA was digested with RNAse R (Epicentre, Madison, WI, USA). The improved circRNAs were then expanded and transcribed into fluorescent cRNAs using a random priming method with Arraystar Super RNA Labelling Kit (Arraystar). Labelled cRNAs were hybridized onto the Arraystar Human circRNA Array V1.0. Finally, the array was scanned using the Agilent Scanner G2505C and the resulting array images were analysed.

| QRT-PCR analysis
Total RNA and miRNA samples were collected from cultured cells  Table 1.

| Dual-luciferase reporter assay
PCR was used to amplify circFOXP1 or FOXP1 3′-UTR fragments containing wild-type (WT) or mutated (MUT) predicted potential miR-33a-5p binding sites. They were then cloned into the psiCHECK-2 vector (Promega Corp., Madison, WI, USA) to form a WT or MUT luciferase reporter plasmid. HEK293T cells were cultured in 24-well plates and cotransfected with 100-nmol/L miR-33a-5p mimics or negative control (miR-NC), 1 μg of MUT or WT luciferase reporter plasmid and Lipofectamine 3000 (Invitrogen). Two days later, dual-luciferase reporter assay system (Promega) was used to explore luciferase activity following a standard procedure. Luciferase activity was standardized using Renilla luciferase and expressed relative to basal activity.

| RNA-binding protein immunoprecipitation (RIP) assay
RIP assay was performed using the Magna RIP Kit (Millipore, USA) and Ago2 antibody (Cell Signaling Technology, USA) following a standard procedure. The transfected cells (2 × 10 7 ) were washed in ice-cold PBS twice and lysed in the same volume of RIP lysis buffer. Then, lysates TA B L E 1 Primers used in real-time RT-PCR analysis were incubated with 5 μg of anti-Argonaute-2 (AGO2) antibody or nonspecific anti-IgG antibody (Millipore) for 2 hours at 4°C. Subsequently, 50 μL of prepared magnetic beads were added to the cell lysates and incubated at 4°C overnight. The beads were then washed with RIP buffer five times and resuspended in 500 μL of TRIzol LS (Life Technology, USA) to obtain enriched RNA. Purified RNA was used for qRT-PCR.

| Cell culture and osteogenic differentiation induction
The
They were then transfected with Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) following manufacturer instructions.

| Alkaline phosphatase staining and quantification
The hASCs were cultured in OM or PM for one week and evaluated using alkaline phosphatase (ALP) staining and quantification. ALP staining was performed with an NBT/BCIP staining kit (Beyotime Biotechnology, Shanghai, China). ALP assay kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) was used to detect the ALP concentration. Total protein content was measured using the Pierce BCA protein assay kit (Thermo Fisher Scientific, Rockford, IL). ALP activity was calculated relative to the control group by standardizing samples to the total protein content.

| Alizarin red S (ARS) staining and quantification
The hASCs were cultured in OM or PM for two weeks and then used for the mineralization assay. After fixation in 95% ethanol, the cells were stained with the 1% ARS staining solution (pH 4.2; Sigma-Aldrich) for 20 minutes at room temperature. To quantify mineralization levels, the stains were dissolved using 100-mmol/L cetylpyridinium chloride (Sigma-Aldrich) for 1 hour and measured at 562 nm with an EnSpire multimode plate reader (PerkinElmer, Waltham, MA). Relative ARS intensity was quantified after samples were standardized to the total protein content.

| Western blotting
Total protein samples were extracted using a radioimmunoprecipi-

| In vivo heterotopic bone formation assay
Female BALB/C homozygous nude (nu/nu) mice (aged six weeks, Abcam) to detect osteogenesis.

| Immunohistochemistry (IHC) assay
The slices (4-μm-thick) were dewaxed and rehydrated using xylene and ethanol, and high-pressure heat was applied for antigen retrieval. After blocking with goat serum, the sections were incu-

| Expression profile for circRNA in OP patients
To detect potential circRNA roles in OP, circRNA microarray analyses were conducted in OP and non-OP bone tissues (n = 20/group).
It was discovered that 2327 circRNAs were down-regulated in OP tissues with a fold difference of >2 (P < 0.05). A total of 2645 circR-NAs were up-regulated using the same cut-off. Hierarchical clustering illustrated top 20 down-regulated or up-regulated circRNAs in OP tissues ( Figure 1A). Subsequently, a subset of top differentially expressed circRNAs (the six most up-regulated and the six most downregulated) were chosen for verification by qRT-PCR ( Figure 1B,C).
The qRT-PCR analysis demonstrated that has_circ_0001320, a F I G U R E 2 CircFOXP1 serves as miR-33a-5p sponge. A and B, Putative binding site for miR-33a-5p in circFOXP1 and FOXP1 3′-UTR. Red colour indicates mutated miR-506-3p-binding site sequence. C and D, Relative luciferase activities analysed in 293T cells cotransfected with miR-33a-5p mimics or control miRNA (miR-NC) and WT or MUT luciferase reporter containing circMID1 (C) or FOXP1 3′-UTR (D, **P < 0.01). E, AGO2 RNA immunoprecipitation (RIP) assay for circFOXP1 levels in hASCs transfected with miR-33a-5p or miR-NC (**P < 0.01). F, Ago2 RIP assay for miR-33a-5p levels in hASCs overexpressing circFOXP1 or vector control (**P < 0.01). G, Fluorescence in situ hybridization assay conducted to determine co-localization between circFOXP1 and miR-33a-5p. Recent studies have discovered that circRNAs regulate corresponding linear transcript function and expression. 33,34 In this manner, regulatory correlation between circFOXP1 and its linear transcript (FOXP1) was explored in the present investigation. FOXP1 expression was found to be significantly down-regulated in OP bone tissues compared to non-OP bone tissues ( Figure 1E). Pearson's correlation analysis revealed a significant positive correlation between circFOXP1 and FOXP1 in OP bone tissues ( Figure 1F). Moreover, FOXP1 mRNA and protein levels were significantly up-regulated or down-regulated when circFOXP1 expression was artificially altered in hASCs with circFOXP1 overexpression or knockdown, respectively ( Figure 1G,H). These results suggest that FOXP1 may be a circFOXP1 target gene.

| CircFOXP1 functions as miR-33a-5p sponge
Given that circRNAs act as an miRNA sponge to regulate gene expression, potential miRNA associated with circFOXP1 and FOXP1 targeting was predicted using bioinformatics analysis. Among these target miRNAs, circFOXP1 and FOXP1 had a binding site for miR-33a-5p, and the predicted miR-33a-5p binding site was mutated ( Figure 2A,B). To ensure miR-33a-5p and circFOXP1/FOXP1 target relationships, luciferase reporter plasmids with MUT or WT circFOXP1 or FOXP1 sequence were cotransfected with miR-33a-5p mimics or negative control. The luciferase activity was significantly lower in the miR-33a-5p mimic group cotransfected with WT circFOXP1 or WT FOXP1 3′UTR ( Figure 2C,D). RIP assay was performed to pull down RNA transcripts bound to Ago2 in hASCs. The circFOXP1 and miR-33a-5p were pulled down efficiently by anti-Ago2, but not by the non-specific anti-IgG antibody ( Figure 2E,F). Moreover, RNA FISH assay results demonstrated that circFOXP1 and miR-33a-5p were co-located in the cytoplasm ( Figure 2G). Overexpression or silencing of circFOXP1 reduced or enhanced miR-33a-5p expression, respectively ( Figure 2H). In addition, miR-33a-5p expression levels in the OP samples were much lower than those in the non-OP bone tissues ( Figure 2I).
To evaluate circFOXP1, miR-33a-5p and FOXP1 roles in hASC osteogenesis, these genes were overexpressed or inhibited in hASCs. ALP and ARS staining was used to predict the mineraliza-

| CircFOXP1 and miR-33a-5p affect hASC osteogenic differentiation in vivo by regulating FOXP1
To further confirm that circFOXP1 osteogenesis regulation in vivo is similar to that in vitro, hASCs with circFOXP1 overexpression and/ or miR-33a-5p mimics were encapsulated in collagen-based hydrogels and implanted into the nude mouse dorsal subcutaneous space.
Implantation samples were extracted after eight weeks for analysis.
HE staining revealed more newly constructed bone in the circ-FOXP1 or miR-33a-5p inhibitor group and less osteoid tissue in the miR-33a-5p or si-circFOXP1 group compared to the other three groups ( Figure 5A). Masson's trichrome staining demonstrated more collagen fibre bundles arranged compactly in the circFOXP1 or miR-33a-5p inhibitor group and less collagen organization (in blue) in the miR-33a-5p or si-circFOXP1 group compared to the other three groups ( Figure 5B). Furthermore, IHC staining for OCN revealed a  Figure 5D). There were no significant differences between the circFOXP1 + miR-33a-5p or si-circFOXP1 + miR-33a-5p inhibitor group and the NC group, implying that miR-33a-5p can eliminate the circFOXP1 effect on osteogenic differentiation in hASCs in vivo ( Figure 5A-D). The qRT-PCR and Western blot analysis results showed that FOXP1 expression level was markedly increased by the circFOXP1 or miR-33a-5p inhibitor and decreased by the miR-33a-5p or si-circFOXP1. It remained unchanged in the circFOXP1 + miR-33a-5p or si-circFOXP1 + miR-33a-5p inhibitor group compared to the NC group ( Figure 5D,E). In summary, circ-FOXP1/miR-33a-5p was able to regulate hASC osteogenesis in vivo by regulating FOXP1.

| D ISCUSS I ON
OP is a progressive metabolic skeletal condition that increases the fragility fracture risk caused by low bone mass architectural deterioration of bone microstructure because of an imbalance between bone resorption and bone formation. [35][36][37]  miR-33a-5p expression is up-regulated with TNF-α treatment in BMP-2-induced hBMSC osteogenic differentiation. These conflicting findings may be explained by the differences in cell sample sources and different treatments.
MSCs are a good cell source for cell-based therapy for various conditions, including OP, because of their self-renewal properties, multilineage differentiation potential and low immunogenicity. 48,49 Extensive investigations have shown that biomaterial scaffolds loaded with bone marrow-derived MSCs (BMSCs) enhance bone regeneration and cartilage repair. [50][51][52] When comparing to BMSCs, osteogenic proliferation and differentiation of adipose-derived MSCs (ASCs) are less affected by multiple passage and age, making ASCs a candidate source for cell-based therapy, particularly in elderly OP patients. 53

| CON CLUS ION
In this study, circFOXP1 was down-regulated in OP. Study results demonstrated that circFOXP1 might promote hASC osteogenesis in vitro and in vivo by targeting the miR-33a-5p/FOXP1 pathway.
Taken together, these data suggest that circFOXP1 may regulate osteogenesis and bone regeneration and may be a candidate target for hASC-based therapy for OP treatment.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated for this study are available on request to the corresponding author.