The work was endorsed by the Ethics Committee of People’s Hospital of Pingxiang. 15 patients were diagnosed with glucocorticoid-related ONFH were recruited from 2016 to 2020. In addition, 13 individuals with femoral neck fractures due to trauma were recruited as controls. These patients received artificial femoral head replacement, and the femoral head tissues were obtained. The characteristics of participants were shown in Table 1. Informed consent was accordingly obtained from all patients prior to surgery. Bone marrow samples were collected intraoperatively and placed in heparin anticoagu-lant tubes.

To isolate the hBMSCs, serum-free minimum essential medium α (α-MEM) containing 100 UI/ml penicillin and 0.1 μg/ml streptomycin was mixed with the bone marrow samples. Bone marrow samples were centrifuged at room temperature, with the fat layer discarded. The cells were subsequently resuspended in α-MEM and Percoll was slowly loaded. The mixture was centrifuged for 30 min, and the intermediate monocyte layer was aspirated and discarded. After rinsing twice with phosphate buffer saline (PBS), the cells were resuspended in α-MEM with 10% fetal bovine serum, and routinely cultured. Subsequent to adjusting the cell concentration to 1×10⁶ /ml, the cells were inoculated in culture flasks until reached 80% con-fluence. Osteogenic differentiation was subsequently induced with α-MEM containing 10% fetal bovine serum, 0.1 μmol/l dexamethasone, 10 mmol/l β-glycerophosphate, and 50 mg/l vitamin C. In addition, we added 40 μM XAV939 (Selleckchem, Houston, Texas, USA) to the medium and examined the impacts of the Wnt/β-catenin pathway on cell growth and osteogenic differentiation.

MiR-29a-3p mimics, miR-29a-3p inhibitors and their control miR-NC, FOXO3 overexpression plasmid and empty plasmid were available from Genomeditech (Shanghai, China). The oligonucleotides and plasmids were transfected into hBMSCs by Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instruction. 48 h after the transfection, the cells were harvested for the subsequent experiments.

Total RNA extracted from hBMSCs by TRIzol kit (Invitrogen, Carlsbad, CA, USA) was reversely transcribed into cDNA by a Transcriptor First Strand cDNA Synthesis kit (Roche, Basel, Switzerland), and the SYBR Green PCR Master Mix (Qiagen, Valencia, CA, USA) was mixed with cDNA and primers to perform qRT-PCR. The relative RNA expression was calculated by 2^−ΔΔCt method. The primer sequences are: miR-29a-3p: 5’-CTGGTGTCGTGG AATTCAGTTGA-3’ (Forward), 5’-CCTGGCTCCTCACT TGGC-3’ (Reverse); miR-23a-3p: 5’-CGGATCACATTGC CAGGG-3’ (Forward), 5’-CAGTGCGTGTCGTGGAGT-3’ (Reverse); miR-96-5p: 5’-ATGCTTTCTCAACTTGTTGG-3’(Forward), 5’-TCACCGCTCTTGGCCGTCACA-3’ (Re-verse); miR-9-5p: 5’-GTGCAGGGTCCGAGGT-3’ (For-ward), 5’-GCGCTCTTTGGTTATCTAGC-3’ (Reverse); miR-182-5p: 5’-CCCAACTGTATGGTTT-3’(Forward), 5’-CGGA TGGCCCAACGG-3’(Reverse). FOXO3: 5’-CAGCCGAGG AAATGTTCGTC-3’ (Forward), 5’-AGAGTGAGCCGTTT GTCCG-3’ (Forward); U6:5’-GCTTCGGCAATACTAAAA T-3’ (Reverse), 5’-CGCTTCACGAATTTGCGTCATGT-3’ (Reverse); GAPDH: 5’-GGAGCGAGATCCCTCCAAAAT-3’ (Forward), 5’-GGCTGTCATACTTCTAATGG-3’ (Reverse).

The viability of hBMSCs was detected with Cell Coun-ting Kit-8 (CCK-8) (Yeasen, Shanghai, China). The transfected cells were seeded at 3×10³ cells per well into a 96-well plate and cultured, followed by addition of 10 μl CCK-8 solution at 0 h, 24th h, 48th h and 72nd h and incubation for 2 h at 37℃, and ultimately the absorbance value at 450 nm wavelength was measured by a microplate reader (Thermo-Fisher Scientific, Waltham, MA, USA).

Transfected hBMSCs were cultured in 6-well plates for 48 h. Subsequently, hBMSCs were rinsed three times with PBS, and next, the ALP activity of hBMSCs was detected by the ALP activity detection kit (Sigma-Aldrich, St. Louis, MO, USA).

Total proteins in the hBMSCs were extracted by radio-immunoprecipitation assay (RIPA) lysis buffer (Beyotime, Shanghai, China) on ice, and subsequently quantified by the bicinchoninic acid (BCA) protein assay. Then the loading buffer was added into the protein samples, and heated in boiling water for 10 min. Next, the same amount of protein samples in each group were respectively separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to the polyvinylidene fluoride membrane, which was firstly incubated overnight at 4℃ with primary antibodies (anti-FOXO3: ab109629, Abcam Cambridge, UK; anti-ALP: ab224335, Abcam secondary Cambridge, UK; anti-Runx2: ab192256, Abcam Cambridge, UK; anti-OCN: ab133612, Abcam Cambridge, UK; anti-GAPDH: ab9485, Abcam, Cambridge, UK), and secondly with horseradish peroxidase-labeled secondary antibody (ab6721, Abcam Inc., Cambridge, UK). Finally, the protein bands were shown by an ECL detection reagent (Beyotime, Shanghai, China).

For immunofluorescence analysis, cells were seeded in the 12-well plates, then fixed with 4% formaldehyde for 30 min and treated with 0.5% Triton X-100 for 20 min. Subsequently, the cells were incubated with anti-β-ca-tenin antibody (#8480, Cell Signaling Technology, Shanghai, China) at room temperature for 1 h. Then incubated with the corresponding secondary antibody (#4414, Cell Signaling Technology, Shanghai, China) in the darkness at room temperature for 1 h. The cell nuclei were then stained with DAPI. Images were captured by a fluorescent microscope.

For cell cycle assay, the cells were washed with PBS and fixed with cold ethanol for 2 h. Then the cells were stained with propidium iodide (PI) staining solution for 30 min at room temperature in the dark. Then the cells were detected by flow cytometry and the data were analyzed by the FCS 4 Express Flow Cytometry Software (De Novo Software, Glendale, CA, USA). For cell apoptosis assay, the Annexin V-fluorescein isothiocyanate (FITC)/PI Apoptosis Detection kit (Beyotime, shanghai, China) was used. Briefly, the cells were harvested and resupended in binding buffer, and stained with 5 μl of Annexin V-FITC staining solution for 15 min at room temperature in the dark, and then stained with 5 μl of PI staining solution for 5 min at room temperature in the dark. Then the cells were detected by flow cytometry and the data were analyzed by the FCS 4 Express Flow Cytometry Software (De Novo Software, Glendale, CA, USA).

FOXO3 3’UTR sequence containing a binding site with miR-29a-3p was subsequently amplified, and the PCR product was generally cloned into pGL3 Basic reporter vectors (Promega, Madison, WI, USA) to generate the wild type (WT) FOXO3-WT luciferase reporter vector. FOXO3 3’UTR complementary sequence was subsequently mutated and inserted into the reporter vector to generate mutant type (MUT) FOXO3-MUT luciferase reporter vector. Then, the above luciferase reporter plasmids were respectively transfected into 293T cells with miR-NC and miR-29a-3p mimics. 24 h after transfection, the luciferase activity was evaluated by the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA). The relative luciferase activity=(Firefly luciferase activity/Renilla luciferase activity)×100%.

Statistical analysis was performed with Statistical Product and Service Solutions (SPSS) 22.0 software (IBM, Armonk, NY, USA), and all experimental results were expressed as mean±standard deviation. Student t-test was subsequently executed to compare the differences between two groups, and one-way ANOVA test was adopted to compare the differences among three groups. Correlation analysis was accordingly performed by Pearson’s correlation analysis, and statistically, p<0.05 is meaningful.

To explore the differentially expressed genes (DEGs) in the serum of SAON patients, we downloaded microarray gene profiling dataset (GSE123568) which contains 30 SAON patients and 10 non-SAON individuals from the Gene Expression Omnibus (GEO) for bioinformatics analysis. The mRNAs with p-value<0.05 and |log₂FC|≥1.0 were identified as the DEGs. A total of 1402 DEGs were selected (Fig. 1A). Then, the down-regulated DEGs were analyzed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis, and GO analysis hinted that the DEGs were enriched in the oxidation-reduction process, protein phosphorylation, negative regulation of transcription from RNA polymerase II promoter, cell proliferation, protein ubiquitination and negative regulation of apoptotic process (Fig. 1B). KEGG pathway analysis indicated that the DEGs were enriched in FOXO signaling pathway (Fig. 1C). Previous studies have revealed that overexpression of a FOXO3 transgene in mature osteoblasts decreased oxidative stress and osteoblast apoptosis and increased the rate of bone formation (17), then FOXO3 was selected in our research, and notably it was also down-regulated in the SAON samples in GSE123568 (Fig. 1A). Next, we identified 5 common miRNAs which could target FOXO3 by searching miRDB, miRTarase, miRwalk and Targetscan databases. qRT-PCR revealed that FOXO3 mRNA expression was relatively lower in SAON group than that in the control group (Fig. 1E). miR-29a-3p and miR-23a-3p were expressed at higher levels in the hBMSCs isolated from 15 SAON patients than that from the 13 control subjects, however, the expression of miR-96-5p, miR-9-5p and miR-182-5p between two groups have no significant difference (Fig. 1E and Supplementary Fig. S1A∼D). Correlation analysis confirmed that miR-29a-3p expression and FOXO3 mRNA expression in the samples from SAON patients were negatively correlated, and there is no connection between miR-23a-3p and FOXO3 mRNA expression (Fig. 1F and Supplementary Fig. S1E). The above results suggest that miR-29a-3p and FOXO3 may be implicated in SAON progression.

To identify the role of miR-29a-3p in the proliferation, cell cycle and osteogenic differentiation of hBMSCs, we transfected miR-29a-3p mimics, miR-29a-3p inhibitors and control miR-NC into hBMSCs and successfully constructed miR-29a-3p overexpression and low expression models (Fig. 2A). We examined Runx2, ALP and OCN expression after the induction of osteogenic differentiation of hBMSCs, and observed that miR-29a-3p mimics remarkably inhibited the expression of Runx2, ALP and OCN, as well as ALP activity, while the transfection of miR-29a-3p inhibitors functioned oppositely (Fig. 2B∼F). Flow cytometry revealed that overexpression of miR-29a-3p induced apoptosis of hBMSCs, and miR-29a-3p inhibitor showed the opposite effect (Fig. 2G and 2H). CCK-8 assay revealed that the transfection of miR-29a-3p mimics dramatically restrained the viability of hBMSCs, while miR-29a-3p inhibitors promoted the viability of hBMSCs (Fig. 2I). Cell cycle analysis revealed that overexpression of miR-29a-3p induced cell-cycle arrest at G1-G0 phase, and miR-29a-3p inhibitor showed the opposite effect (Fig. 2J and 2K). These findings confirmed that miR-29a-3p restrains hBMSCs growth and osteogenic differentiation, as well as induces cell apoptosis and cell-cycle arrest.

FOXO3 3’UTR contained the potential binding site complementary to miR-29a-3p (Fig. 3A). To further confirm their specific binding relationship, a dual-luciferase reporter assay was performed, and it was found that the transfection of miR-29a-3p mimics greatly repressed the luciferase activity of FOXO3-WT-containing luciferase reporter plasmids but had no impacts on that of FOXO3-MUT-containing luciferase reporter plasmids (Fig. 3B). qRT-PCR and western blot showed that the transfection of miR-29a-3p mimics significantly decreased FOXO3 expression compared with miR-NC, while miR-29a-3p inhibitors functioned oppositely (Fig. 3C and 3D). The above results confirmed that FOXO3 was a downstream target of miR-29a-3p in hBMSCs.

Next, we co-transfected the FOXO3 overexpression plasmid and miR-29a-3p mimics into hBMSCs and found that FOXO3 overexpression plasmid partially counteracted the inhibitory impact of miR-29a-3p mimic on the expression of osteogenic differentiation markers, including Runx2, ALP and OCN (Fig. 4A∼D), and the viability of hBMSCs (Fig. 4E), as well as the ALP activity (Fig. 4F). Additionally, FOXO3 overexpression significantly reversed the apoptosis (Fig. 4G) and cell-cycle arrest (Fig. 4H) of hBMSCs induced by miR-29a-3p overexpression. The above results implied that miR-29a-3p repressed hBMSCs’ viability and osteogenic differentiation through FOXO3.

Some previous studies report that miR-29a-3p can regulate the activity of Wnt/β-catenin signaling (18, 19), and considering Wnt/β-catenin is a crucial pathway to modulate the osteogenic differentiation of hBMSCs (20), it is also interesting to explore whether miR-29a-5p inhibited hBMSCs cell growth and osteogenic differentiation by regulating the Wnt/β-catenin signaling pathway. We detected β-catenin protein expression in hBMSCs transfected with miR-29a-3p inhibitors and miR-29a-3p mimics, and found that as against the miR-NC group, the transection of miR-29a-3p inhibitors significantly promoted β-catenin expression, while the transfection of miR-29a-3p mimics worked oppositely, and the immunofluorescence assay showed that transfection of miR-29a-3p inhibitor could promoted β-catenin nuclear translocation (Fig. 5A∼C). Subsequently, XAV939, a specific inhibitor of Wnt/β-catenin signaling pathway was used to treat the cells, and the results showed that XAV939 could partially reverse the impact of miR-29a-3p inhibitors on the osteogenic differentiation (Fig. 5E∼G), viability (Fig. 5H), ALP activity (Fig. 5I), apoptosis (Fig. 5J) and cell-cycle arrest (Fig. 5K) of hBMSCs. These data suggested that miR-29a-3p could also regulate the phenotypes of hBMSCs via repressing Wnt/β-catenin signaling.