CircPVT1 up‐regulation attenuates steroid‐induced osteonecrosis of the femoral head through regulating miR‐21‐5p‐mediated Smad7/TGFβ signalling pathway

Abstract Steroid‐induced osteonecrosis of the femoral head (SIONFH) has been a common disease following corticosteroid therapy. Presently, we aim to explore the functions of circular RNA (circ) PVT1 in SIONFH rats and the underlying mechanism. Glucocorticoid (GC) was used to treat SD rats and bone marrow‐derived mesenchymal stem cells (BMSCs) to construct SIONFH model in vitro and in vivo, respectively. The pathological injury of the femoral head in the SIONFH rats was detected via haematoxylin‐eosin (HE) staining and immunohistochemistry (IHC). The osteogenic differentiation, proliferation and apoptosis of BMSCs were detected. Western blot was used to detect Smad7, Bax, Bcl2 and Smad2/3. The potential targets of circPVT1 and miR‐21‐5p were validated through luciferase reporter gene assay and RNA pull‐down assay, respectively. We found that CircPVT1 was decreased in the femoral head of SIONFH rats and GC‐treated BMSCs, while miR‐21‐5p was markedly up‐regulated. Overexpressed circPVT1 attenuated the apoptosis and cell viability inhibition of BMSCs induced by GC, while miR‐21‐5p up‐regulation had the opposite effects. What's more, the in vivo experiments confirmed that up‐regulating circPVT1 repressed osteonecrosis in SIONFH rats through repressing apoptosis. Mechanistically, circPVT1 functioned as a ceRNA of miR‐21‐5p, which targeted at the 3'untranslated region of Smad7. CircPVT1 enhancing Smad7 and mitigating GC activated TGFβ/Smad2/3 pathway through inhibiting miR‐21‐5p. In conclusion, CircPVT1 exerts protective effects against SIONFH via modulating miR‐21‐5p‐mediated Smad7/TGFβ pathway.

of orthopaedic diseases. Taking CircRNA-9119 as an example, its overexpression up-regulates PTEN expression by down-regulating microRNA-26a to protect IL-1β-treated osteoarthritis-mediated chondrocytes from apoptosis. 9 Besides, CircGCN1L1 promotes synovium cell proliferation and chondrocyte apoptosis by targeting miR-330-3s and TNFα in TMJ osteoarthritis. 10 CircPVT1 is an essential member of the circRNA family and has shown its important role in several tumours. For instance, the down-regulation of CircPVT1 represses the growth of liver cancer cell via modulating miR-3666/SIRT7 axis. 11 Additionally, circPVT1 functions as a ceRNA of miR-526b, thus promoting the metastasis of osteosarcoma via up-regulation FOXC2. 12 Kun-Peng Z et al also found that circPVT1 enhances the chemoresistance of osteosarcoma cells to adriamycin and cisplatin through ABCB1. 13 However, the distinct expression of CircPVT1 within SIONFH and its specific regulatory mechanism are still unknown. target Runx2 and CXCL12 and prevent SIONFH by promoting osteoblast differentiation and angiogenesis. 15 MiR-21-5p, as an important member of the microRNA family, is located at 17q21.1 with 72 bp in length. Also, papers have shown that miR-21-5p has a significant effect on improving cognitive impairment, 16 regulating extracellular matrix degradation and angiogenesis, 17 but its role in SIONFH has not been reported.
Transforming growth factor β (TGFβ) is a multifunctional cytokine, including three receptors TGFβ R-I, TGFβ R-II and TGFβ R-III. Among them, only type I and type II are necessary for the Smad signalling pathway and have membrane-bound serine and threonine activity. 18 Smad protein is the only known intracellular kinase substrate of TGFβ receptor, and it is a signal transduction factor downstream of TGFβ receptor. 19 Smad7, one of Smad's proteins, acts as an intracellular inhibitory protein that antagonizes signal transduction between the TGFβ family. Researches prove that Huogu I formula (I) can modulate the expression of BMP2, TGFβ/Smads and OPG/RANKL, thus promoting the repair of necrotic bones. 20 Additionally, the study conducted by Bai et al suggested that miR-27a regulates steroid-induced ONFH via TGFβ/Smad7 signalling. 21 Collectively, the above studies suggest that TGFβ/Smad7 signalling pathway plays an important role in SIONFH.
Here, both in vivo and in vitro SIONFH models were established to explore the specific role and mechanism of circPVT1 regulating SIONFH. We found circPVT1 was down-regulated in the femoral head of rats with osteonecrosis as well as GC-treated BMSCs. Upregulating circPVT1 significantly reduced GC-mediated apoptosis of BMSCs. Additionally, the bioinformatic analysis showed that miR-21-5p is a potentially target at the downstream of circPVT1 and miR-21-5p was inhibited following circPVT1 up-regulation. Therefore, we supposed that the circPVT1/miR-21-5p axis might exert a vital role in SIONFH via modulating the Smad7/TGFβ pathway. We hope that this finding provides a new reference for the prognosis and molecular targeted therapy of SIONFH.

| Establishment of SIONFH model
The experimental animal centre of Honghui Hospital Xian Jiao Tong University Health Science Center provided 10-week-old male SD rats. The Animal Ethics Committee of Honghui Hospital Xian Jiao Tong University Health Science Center approved all experimental procedures in this study. In twelve hours light and dark circle, all rats were free to drink and eat. We randomly divided them into a SIONFH model group (N = 5) and a control group (N = 5). The SIONFH model was described as 22 and implemented after modification. Rats in the SIONFH model group were injected subcutaneously at day 1 with imiquimod (IMI 30 mg/kg), and intramuscularly at day 2 with methylprednisolone (MPSL 20 mg/kg). At the fourth week, all injections were made again. At the same time-point, the model group and the control group were injected with the same amount of normal saline. Six weeks after the first injection, the rats were killed, and the femurs were collected.

| BMSC isolation and culture
During total hip replacement, bone marrow tissue (5-10 mL) was removed from the proximal femur with a sterile syringe. The collected bone marrow tissue was transferred to a centrifuge tube containing phosphate-buffered saline (PBS) solution and carefully removed with a pipette to form a cell suspension. Then, suspension was transferred to a centrifuge tube containing the same amount of leucocyte separation liquid for centrifugation (30 minutes, 340 g). Next, the sample was resuspended in Dulbecco's Modified Eagle's Medium (DMEM; Gibco, Rockville, MD, USA) containing 10% foetal bovine serum (FBS; Gibco, Rockville, MD, USA). The cells were evenly inoculated in 6-well plates and grew in an incubator (37°C, 5% CO 2 , 95% humidity). The medium was changed every 3 days.
Twenty-four hours after transfection, we extracted total cellular RNA for real-time fluorescence quantitative PCR to monitor the expression changes of CircPVT1, miR-21-5p in cells after transfection.

| qRT-PCR
First, we used the TRIzol reagent to extract total RNA. Second, the RNA was reverse transcribed into cDNA using PrimeScript™ RT Reagent kit (Invitrogen, Shanghai, China) in line with the manufacturer's instructions. Bio-Rad CFX96 quantitative PCR system and SYBR were utilized to perform qRT-PCR in accordance with supplier's regulations. The conditions of qRT-PCR were as follows: pre-denatured at 95°C for 5 minutes, then denatured at 95°C for 15 seconds and annealed at 60°C for 30 seconds. GAPDH was the endogenous control for CircPVT1, ALP, RUNX2, CoL1a1, OCN, ACP5 and CTSK, and U6 was the endogenous control for miR-21-5p. The 2 (−ΔΔCt) method was applied for statistics. All qRT-PCRs were repeated three times.
Guangzhou Ruibo Company designed and synthesized the primers.
Primer sequences of each gene were shown in Table 1. were lysed in ice water by ultrasound. And the Bradford method was utilized to determine protein concentration. Equal amounts of protein from each group were subjected to 10% SDS-PAGE electrophoresis, and the protein on the gel was transferred to PVDF membranes (Millipore, Bedford, MA, USA). For one hour, the membranes were sealed at 4°C, then they were incubated with the addition of Anti-Smad7 antibody(ab216428), Anti-p-Smad2 antibody (ab18834), Anti-Smad2 antibody (ab40855), Anti-p-Smad3 antibody (ab52903), Anti-Smad3 antibody (ab40854), Anti-Bcl2 antibody (ab182858) and

| Western blot
Anti-Bax antibody (ab32503) (overnight, 4°C). Next, TBST was used to wash the membranes twice. They were incubated with fluoresceinlabelled goat anti-rabbit (ab205718, 1:2500) at room temperature for 1 hour. Abcam (Cambridge, UK) supplied the above antibodies. After being washed three times, they were exposed with ECL developer (Millipore, Bedford, MA, USA) and imaged with a membrane scanner.  and DAPI-positive cells in 3 fields were randomly selected and calculated. We took the average cell proliferation rate of 3 visual fields as the cell proliferation rate. Cell proliferation rate = the number of BrdU-positive cells/number of DAPI-positive cells.

| Flow cytometry for cell apoptosis
Annexin V-FITC double staining method was carried out to detect cell apoptosis. Twenty-four hours after transfection, the BMSCs underwent trypsinization and collection and were inoculated in a 6well plate (2 × 10 6 cells/well), followed with 24-hour culture. When
Then, the samples were dehydrated with gradient ethanol (1 minute/ time), permeated twice with xylene (1 minute/time). Finally, the sections were sealed with neutral resin. We observed the morphological differences of the FHs cells in the SIONFH group and the control group using an optical microscope. As for HE staining in animal experiments, the above steps were duplicated. SIONFH histopathological changes were monitored under an optical microscope, and bone analysis parameters were measured by image analysis. We select three view fields from each section and calculated the average value.

| TRAP staining
Tartrate-resistant acid phosphatase (TRAP) staining was conducted to assess osteoclast activity using Acid Phosphatase, Leukocyte (TRAP) Kit form Sigma-Aldrich. Methyl green was used to replace haematoxylin for cell nuclei staining, to give prominence to TRAPpositive cells labelled in purplish red.

| Data analysis
All statistical analyses were performed with SPSS22.0 statistical software (SPSS Inc, Chicago, IL, USA). The evaluation of the data was shown as mean ± standard deviation (x ± s). Univariate ANOVA was employed for the multi-factor comparison, and independent sample t test was used for the comparison between two groups. All differences in performance were statistically significant when P < .05.

| Bone histomorphology of SIONFH rats
Hormone mediation can have a severe effect on bone morphology, which may eventually lead to ONFH. We first examined the pathological injury of the bone by HE staining. It was found that the tra-  Figure 1A) and the TRAP-labelled osteoclasts were enhanced ( Figure 1B). The measurement results of bone cavity rate, cavity area and trabecular area showed that compared with the standard group, the cavity rate and cavity area of the SIONFH group increased, and the trabecular area decreased (P < .05, Figure 1C-E). At the same time, the bone volume fraction, trabecular bone number and thickness were notably reduced, while the trabecular bone separation and mode factor were significantly increased (P < .05, Figure 1F-J). The results of bone histomorphology revealed that the femoral head's hormone-induced necrosis severely damaged the typical morphology of bone tissue.

| Effects of GC on osteogenic differentiation ability of BMSCs and expression level of CircPVT1/ miR-21-5p
First, we tested osteogenic differentiation markers and genes related to osteoclast formation. The RT-PCR result illustrated that after 1600 mg GC treatment, the levels of osteogenic differentiationspecific markers ALP, RUNX2, CoL1a1 and OCN were downregulated (compared with the con group) (P < .05, Figure 2A-D). In comparison, the expression of osteoclast-related genes ACP5 and CTSK were up-regulated (P < .05, Figure 2E,F). Besides, the expression of CircPVT1 and miR-21-5p in BMSCs were examined via RT-PCR experiment. It was found that after GC treatment, CircPVT1 expression was decreased, and miR-21-5p expression was increased F I G U R E 1 Bone histomorphology examination of SIONFH rats. The SIONFH model was constructed in rats. A, HE staining was taken to check the pathological damage of femur. B, TRAP staining was adopted to detect osteoclast viability. C-E, Measurement of bone cavity rate, cavity area and trabecular area. F-J, Determination of bone volume fraction, trabecular bone quantity, trabecular bone thickness, trabecular bone separation and mode factor. *P < .05, **P < .01, ***P < .001 (vs.con group). N = 5 (P < .05, Figure 2G,H). The above results indicated that the osteogenic differentiation ability of BMSCs was considerably suppressed by GC, and the osteoclast ability was significantly enhanced.  Figure 3C).

| CircPVT1 overexpression promoted the differentiation and proliferation of BMSCs and inhibited cell apoptosis
Meanwhile, BrdU and flow cytometry were implemented to examine cell proliferation and apoptosis ability, respectively. It was found that overexpression of CircPVT1 promoted cell proliferation and inhibited apoptosis (compared with the GC group) (P < .05, Figure 3D,E). We carried out Western blot to determine the protein levels of Smad7, Smad2/3, Bcl2 and Bax. It was found that GC markedly enhanced the activation of Smad2/3 and Bax level, while reduced Smad7 and Bcl2 level. By contrast, overexpressing CircPVT1 reduced Smad2/3 and Bax, and promoted Smad7 and Bcl2 (compared with GC group) (P < .05, Figure 3F). The above outcome demonstrated that the overexpression of CircPVT1 could accelerate BMSCs differentiation and proliferation and suppress their apoptosis.  Figure 4K). These findings showed that transfection with miR-21-5p mimics inhibited the differentiation, proliferation and apoptosis of BMSCs.

| MiR-21-5p was the target of circPVT1 and Smad7
According to the lncRNA-miRNA-mRNA regulatory network diagram, we were urgent to understand the regulatory mechanism of CircPVT1 and Smad7. Through Venn diagram analysis, we found that 25 miRNAs were potential downstream targets of CircPVT1 ( Figure 5A). Meanwhile, RT-PCR results showed that the expression of miR-21-5p was suppressed most obviously after overexpressing CircPVT1 ( Figure 5B). By browsing the StarBase database (http:// starb ase.sysu.edu.cn/), the binding sites between miR-21-5p and CircPVT1 and Smad7 were found and shown in Figure 5C. To clarify the targeted relationship between the three, we conducted dualluciferase activity experiments. And the miR-21-5p mimics suppressed the luciferase activity of cells transfected with CircPVT1-WT or Smad7-WT vectors but had no significant effect on CircPVT1-MUT and Smad7-MUT (P < .05, Figure 5D). Besides, the RNA pulldown assay showed that CircPVT1 and Smad7 were obtained not from Bio-miR-21-5p-MUT but specifically from Bio-miR-21-5p-WT (P < .05, Figure 5E). This situation suggests that miR-21-5p contains binding sites for CircPVT1 and Smad7.

| The effect of overexpression of CircPVT1 on osteoblasts of femoral head necrosis in rats
To further verify the effect of CircPVT1 on osteoblastic differentiation of femoral head necrosis, we conducted experiments in vivo.
The results of HE staining showed that in the SONFH model, after the overexpression of CircPVT1, the trabecular structure in the femoral head sections of the rats was slightly degenerated and set in order, bone cells filled the lacunae, bone surface folded and collapse were not noticeable, and calcification zone was well connected to the subchondral trabecular structure ( Figure 7A), and TRAP activity was dampened ( Figure 7B). Meanwhile, the measurement results of bone cavity rate, cavity area and trabecular area showed that the CircPVT1 group had decreased cavity rate and area, while the trabecular area increased (compared with the vector group) (P < .05, Figure 7C-E). What's more, bone volume fraction, number and thickness of trabecular bone increased significantly, while trabecular separation degree and model factor decreased (P < .05, Figure 7F-J). and OCN (P < .05, Figure 7K-N) and decreased the expressions of osteoclast-related genes ACP5 and CTSK (P < .05, Figure 7O,P). The outcome demonstrated that CircPVT1 promotes BMSC differentiation in vivo and attenuates SIONFH.

| The effect of overexpressing CircPVT1 on SMAD7 and SMAD2/3 in vivo
First, we adopted IHC to detect the expression of Smad7 and the pro-apoptotic protein Bax. We found that overexpressing CircPVT1 in the SONFH model significantly increased Smad7-positive labelled cells ( Figure 8A) and reduced Bax positive labelling cells ( Figure 8B).
Western blot results illustrated that there was no significant difference in Smad7 expression and Smad2/3 phosphorylation after CircPVT1 overexpression in the normal group. Nevertheless, overexpressing CircPVT1 in SONFH model significantly elevated Smad7 and Bcl2 expression and abated Smad2/3 phosphorylation and Bax level (P < .05, Figure 8C), indicating that CircPVT1 up-regulated Smad7 and inactivated Smad 2/3 in vivo.

| D ISCUSS I ON
As a subtype of non-traumatic ONFH, steroid-induced ONFH (SIONFH) is associated with increased intra-osseous pressure caused by bone marrow adipocyte proliferation and increased adipogenesis, which can slow the femoral head blood flow and eventually lead to The effect of overexpressed CircPVT1 on osteogenic differentiation of rat femoral head necrosis was verified in vivo. The CircPVT1 overexpression plasmid was injected into the ventricle of SIONFH rats. A, HE staining was taken to check the pathological damage of the femur. B, TRAP staining detected osteoclast viability. C-E, Measurement of bone cavity rate, cavity area and trabecular area. F-J, Determination of bone volume fraction, trabecular bone quantity, trabecular bone thickness, trabecular bone separation and mode factor. K-P, RT-PCR was used to detect osteogenic differentiation-related genes ALP, RUNX2, CoL1a1, OCN and osteoclastogenesis-related genes ACP5 and CTSK. NS P > .05, **P < .01, ***P < .001 (vs.vector group). N = 5 F I G U R E 8 Effect of overexpressing CircPVT1 on SMAD7 and SMAD2/3 in vivo. The treatment method was the same as that in Figure 7. A,B, IHC was implemented to verify the positive expression of Smad7 and pro-apoptotic protein Bax. C, Western blot was applied to monitor the expression of Smad7, Smad 2/3, Bcl2 and Bax. NS P > .05, ***P < .001 (vs.vector group). N = 5

F I G U R E 9
The mechanistic diagram. CircPVT1 is down-regulated in SIONFH.
Here, we discovered miR-21-5p expression was raised in SIONFH, and the overexpression of miR-21-5p significantly suppressed the differentiation and proliferation of BMSCS and promoted apoptosis.  39 Similar to the above studies, this study found a targeted relationship between CircPVT1 and miR-21-5p through database analysis, and overexpressed CircPVT1 can inhibit miR-21-5p-induced osteonecrosis of BMSCs and reduce apoptosis. These findings confirmed that miR-21-5p, as a downstream molecule of CircPVT1, induces osteoclast formation and inhibits osteogenic differentiation in SIONFH.
TGFβ is an important cytokine involved in osteocyte function and metabolism. Previous papers have proven that TGF not only contributes to osteocyte mitosis, but also can reduce bone protein loss, increase bone deposition rate and promote osteoblast differentiation. 40 Meanwhile, Smad7, as an inhibitory Smad, is a key regulator of TGFβ/bone morphogenetic protein (BMP) signalling through the negative feedback loop. 41 Studies have shown that MOTS-c promotes the synthesis of type I collagen in osteoblasts by regulating the TGFβ/Smad signalling pathway, thus improving osteoporosis. 42 He G et al reported that miR-877-3p promoted osteoblast differentiation of TGF-β1-induced MC3T3-E1 cells by targeting Smad7. 43 Besides, Cui H et al found that exosome-derived miR-21-5p secreted by bone marrow macrophages directly targets Smad7 and causes the activation of fibrosis in tendon cells. 44 In this study, it was found that over- To sum up, our study shows that CircPVT1 is down-regulated in SIONFH and down-regulates TGFβ/Smad2/3 expression and promotes Smad7 activation by targeting miR-21-5p, thereby reducing SIONFN ( Figure 9). This study provides a new reference for the diagnosis and treatment of SIONFH, but more relevant mechanisms of action and treatment strategies need to be further explored.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no competing interests.

E TH I C A L S TATEM ENT
Our study was approved by the Ethics Review Board of Honghui Hospital, Xi'an Jiaotong University Health Science Center.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data sets used and analysed during the current study are available from the corresponding author on reasonable request.