Three-week-old male Sprague-Dawley rats were humanely euthanized, and the femurs and tibias were harvested under sterile conditions. Culture method was previously reported (Aghaloo et al., 2010; Wu et al., 2024), and operation flow is shown in Supplementary Figure S1. The cells from the third or fourth passage were used for the subsequent experiments.

The levels of cell surface markers, including CD44, CD34, CD45, and CD90, were analyzed following the guidelines provided by the manufacturer (Shi et al., 2014). Additionally, the capacity for multi-lineage differentiation was assessed by Alizarin red staining (ARS), Oil red O staining, and Alcian blue staining (Cyagen, China) after 7-day, 21-day, and 28-day stimulations. Colony-forming unit assays were initially performed with 1 × 10³ single-cell suspensions seeded in a 10 cm diameter culture dish (Corning, United States), and reflected by toluidine blue (Beyotime, China) after 14 days.

The PEMFs device (School of Manufacturing Science and Engineering, Sichuan University, China) (Huang et al., 2022; Wang et al., 2022) consists of a pulse generator, a stepper motor driver, and a Helmholtz coil, and produces spatially homogeneous, time-varying magnetic fields in the incubator (seen in Supplementary Figure S2; Supplementary Table S1). There are two trays inside the circular Helmholtz coil chamber, and the output waveform from the signal generator produced a pulsed burst with a duty ratio of 50% (burst width: 6.67 ms; pulse width: 6.67 ms), repeated at a frequency of 75 Hz. The magnetic flux density increased to a predetermined maximal level within approximately 50 μs (with a rise rate of ∼17 T/s) when driving field amplitudes ranged between 0.5 and 3.8 mT. This was measured using a hand-held Gaussmeter (HT201, Hengtong, China). According to previous studies, BMSCs at passage 4 hungered with 5% EVs-free FBS for 48 h, and then were placed in the incubator under PEMFs with 0 (negative control), 1 mT (Celik et al., 2021), 1.6 mT (Huang et al., 2021; Yang et al., 2018), or 3 mT (Parate et al., 2020) amplitudes for 60 min.

BMSC-EVs were isolated via ultra-centrifugation method as previously described (Guo et al., 2017; Xu et al., 2021) and flow is shown in Supplementary Figure S3. The culture media were harvested, centrifuged at 300 × g and 2,000 × g to remove debris, and then filtered through a 0.22 μm filter (Merck-Millipore). The supernatant was aliquoted into 15 mL Amicon Ultra-15 devices with a 100 kDa membrane and subjected to centrifugation at 4,000 × g. Then, it was ultra-centrifuged at 100,000 × g for 70 min (SW32Ti, Beckman Coulter), washed with PBS, and ultra-centrifuged again at the same speed for 70 min. The EVs were carefully resuspended in sterile PBS and stored at −80 °C for subsequent experiments. Transmission electron microscopy (TEM) (JEM-1400FLASH, Japan) was used for observing morphology operating at 80–120 kV. For particle size and number analysis, nanoparticle trafficking analysis (NTA) was completed using the ZetaView system (Particle Metrix, Germany) following the manufacturer’s instructions. Western blot was employed to confirm the presence of markers, such as positive expression of TSG101, CD81, and CD9, and negative expression of Calnexin.

Next process involved incubating EVs with DIO (Beyotime, China) for 15 min at room temperature. After washing with PBS and centrifuging at 100,000 × g for 70 min, the various BMSC-EVs were suspended in basal medium (10¹⁰/mL) and incubated with MC3T3-E1 cells for 48 h at 37 °C (Wu et al., 2019; Li et al., 2022). Then stained by 0.1 g/mL DAPI (Beyotime, China) for 5 min, the MC3T3-E1 cells were placed under a confocal system of high-content screening (PE/Opera Phenix Plus, PerkinElmer, United States) for image capture.

MC3T3-E1 osteoblastic cells were purchased from CCTCC (Subclone 14, GDC0188, China) and cultured in α-MEM supplemented with 10% FBS (Lonsera, Uruguay) and 1% penicillin-streptomycin in humidified incubators at 37 °C and 5% CO₂. Cells were treated with 10 μM DEX (TargetMol, China) in vitro to mimic the disease for 24 h (Tao et al., 2017). BMSC-EVs, BMSC-EVs^{PEMFs (1 mT)}, BMSC-EVs^{PEMFs (1.6 mT)} and BMSC-EVs^{PEMFs (3 mT)} at a dosage of 10¹⁰/mL were co-cultured with MC3T3-E1 osteoblastic cells for 48 h.

Annexin V-FITC/PI apoptosis detection kit (Cell Signaling Technology, Danvers, MA, United States) was used for distinguishing between live, early apoptotic, and late apoptotic/necrotic cells (Zhu et al., 2021). Then the results were obtained using a flow cytometer (FACSAria III, BD, United States).

Following osteogenic induction for periods of 7 and 21 days, the cells were fixed with a solution of 4% paraformaldehyde (Biosharp, China). Subsequently, alkaline phosphatase (ALP) activity was assessed with a BCIP/NBT staining kit (Beyotime, China). To evaluate the formation of mineralized nodules, Alizarin Red S (ARS) staining (Beyotime, China) was conducted for 20 min. The cells were then examined microscopically (Ti2, Nikon, United States) to assess osteogenic differentiation.

Total proteins were extracted and examined using a protein extraction kit (Beyotime, China) and BCA method with a commercial kit (Thermo Fisher Scientific, United States) following established protocols. Then the proteins were separated using 10% SDS-PAGE (EpiZyme, China) and were then transferred onto polyvinylidene fluoride membranes with a pore size of 0.22 μm. GAPDH was used for normalization. The experiments were performed in triplicate. The information concerning the antibodies used and their concentrations is presented in Supplementary Table S2.

Total RNA was extracted with TRIzol reagent (Takara, Japan) from femoral heads and cultured cells, followed by reverse transcription to generate the first-strand cDNA using the Stand cDNA Synthesis SuperMix for qPCR Kit (Hifair III, Yeasen, China). PCR was conducted with the SYBR Green PCR master mix (HieffUNICON, Yeasen, China) utilizing a Bio-Rad CFX Connect real-time system (Bio-Rad, United States). Primer sequences are shown in Supplementary Table S3 and Supplementary Table S4. The experiments were conducted in three replicates, and the data were analyzed by the 2^−ΔΔCT method (Bernáldez et al., 2017).

Briefly, after being fixed, permeabilization and blocked, cells were incubated with primary antibodies: RUNX2, CON, RIPK1 or RIPK3 (1:200) (Beyotime, China). After being washed twice, the 96-well (PerkinElmer, United States) were subsequently treated with secondary goat anti-rabbit antibody (Alexa Fluor 488, Beyotime, China) at 37 °C for 1 h. F-action and nuclei were co-stained for 20 min with phalloidin (Actin-Tracker Red-594, Beyotime, China) and for 5 min with DAPI, respectively.

MC3T3-E1 cells treated with DEX or with proper EVs were compared. The sequence and filtering of clean reads were completed as previously described (Thompson et al., 2020). A cDNA library was generated using pooled RNA from two groups and sequenced utilizing the Illumina Novaseq™ 6000 sequencing platform (LC-Biotechnology CO., Ltd., Hangzhou, China) (Kim et al., 2019; Kovaka et al., 2019). The raw sequence data have been submitted to the NCBI Short Read Archive (SRA) with the access number PRJNA1115973 (SAMN41518211-SAMN41518218). DESeq2 soft were utilized for analyzing the differentially expressed genes (DEGs). A volcano plot was performed by DEGs (Fold change ≥ 1.1 and false discovery rate (FDR) < 0.1). DEGs were included for further functional analysis based on GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases.

Approval for the animal experiments in this study was obtained from the Animal Ethical Committee. According to the guidelines for sample size calculations of Boston University and the study of Li and colleagues (Li et al., 2023), a total of 40 male SD rats (8 weeks, male; purchased from HFKbio, China) were randomly divided into four groups (n = 10) by simple randomization method, using a computer-generated table: normal (saline); MPS (Methylprednisolone hemisuccinate (MPS) + saline); EVs^{No PEMFs} (MPS+EVs without preconditioning); and EVs^PEMFs (MPS+proper EVs^PEMFs). The model and treatment protocols were established as previously reported. MPS (20 mg/kg per day; TargetMol, China) or saline was injected intramuscularly into rats to induce ONFH during the first 3 days of each week, continuing until the third week (Zhao et al., 2023). Beginning with the first injection of MPS, the EVs (100 μL, 10¹⁰ particles) were injected into the rats through the tail vein, thrice a week for 3 weeks (Chen et al., 2020; Guo et al., 2016). Five rats were dissected after isoflurane overdose for examination at 2 weeks and the remaining rats were dissected at 6 weeks after the first injection (Chen et al., 2022).

The femoral heads were scanned using micro-CT (Quantum GX, PerkinElmer, United States), with a voltage of 80 kV and a current of 100 μA. The scanner software was configured to achieve high resolution, utilizing a voxel size of 20 μm and a field of view measuring 10 mm. Three-dimensional (3D) images of the femoral heads were reconstructed and the bone volume/total volume (BV/TV), bone surface/bone volume (BS/BV), bone mineral density (BMD), trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular separation (Tb.Sp) of the region of interest was calculated by Caliper Analyze (BIR, Mayo Clinic, United States).

The specimens for histological evaluation, including hematoxylin and eosin (H&E) and Masson’s trichrome staining, were processed according to earlier protocols (Li G. et al., 2020). Images were captured using a light optical microscope (Ni-E, Nikon, United States). The specimens were decalcified, fixed in formaldehyde, dehydrated, and embedded in paraffin. After dewaxing and antigen retrieval, 10% bovine serum albumin was used to block nonspecific binding for 30 min. The sections were incubated overnight at 4 °C with primary antibodies (seen in Supplementary Table S5) and then with an HRP-conjugated secondary antibody (PV9001, ZSGB Biotechnology, China), counterstained with hematoxylin. All tests were done on at least three sections per specimen and regions of interesting were selected randomly by two independent observers to ensure consistency and representativeness while minimizing sampling bias.

The data are expressed as the mean ± standard deviation (SD) based on a minimum of three separate experiments. Statistical analysis was conducted using GraphPad Prism 9 (La Jolla, CA, United States). One-way ANOVA was used to compare group means, with Bonferroni’s post hoc analysis for significance between pairs. A p-value <0.05 was deemed significant.

BMSCs expressed CD44 and CD90, while showing no expression of CD34 and CD45 (Figure 1A). They demonstrated the capacity to differentiate into adipocytes, osteoblasts, or chondrocytes upon appropriate induction (Figure 1B). The microscopy view indicated typical shape of BMSCs, and good viability from colonies (Supplementary Figure S4A). The EVs (BMSC-EVs and BMSC-EVs^PEMFs) were smaller than 200 nm (Figure 1C) with a mean diameter of 97.30 ± 11.34 nm, 95.00 ± 8.26 nm, 103.50 ± 10.20 nm and 99.60 ± 6.78 nm in BMSC-EVs^{No PEMFs}, BMSC-EVs^{PEMFs (1 mT)}, BMSC-EVs^{PEMFs (1.6 mT)}, and BMSC-EVs^{PEMFs (3 mT)} group, respectively. The EVs exhibited a round-shaped morphology (Figure 1D). No obvious difference of particles numbers was detected (p > 0.05, Supplementary Figure S4B). Moreover, particles were positive for EVs markers including CD9, CD81, and TSG101, and negative for Calnexin (endoplasmic marker, Supplementary Figure S4C). The result of immunofluorescence manifested that the DIO-labeled EVs could transfer to the perinuclear region of BMSCs after incubation with BMSCs (Figure 1E). The viability of osteoblasts was not reduced when incubated with EVs and most EVs transferred to osteoblasts after 24 h incubation (Supplementary Figure S4D). These results indicated that BMSC-EVs were isolated and incorporated into MC3T3-E1 cells.

All intervention groups displayed an enhancement of osteogenic activity, with the effect of BMSC-EVs^{PEMFs (3 mT)} being significantly stronger than that of different intensities and BMSC-EVs^{No PEMFs}. The activity of ALP was higher in BMSC-EVs^{PEMFs (3 mT)} group (p < 0.001, Figures 2A,B). The protein level of RUNX2 after 7-day induce were significantly higher than other groups with DEX (p < 0.001, Figures 2C,D). RT-qPCR determined that the expression of Runx2 and Bmp2 in BMSC-EVs^{PEMFs (3 mT)} group were significantly higher than that in the DEX, EVs, and other amplitudes groups (p < 0.001, Figure 2I). Subsequently, the result of ARS staining after 21-day differentiation demonstrated that the calcium nodules in EVs, BMSC-EVs^{PEMFs (1.6 mT)} and BMSC-EVs^{PEMFs (3 mT)} groups were significantly higher than that in the DEX groups (p < 0.001, Figures 2E,F). OCN was high-expressed in BMSC-EVs^{PEMFs (3 mT)} group than EVs and BMSC-EVs^{PEMFs (1.6 mT)} group (p < 0.001, Figures 2G,H). Moreover, the expression of Ocn and Col1a1 in the BMSC-EVs^{PEMFs (3 mT)} group was be increased comparing to DEX, EVs, and other amplitude groups (p < 0.001, Figure 2i).

Based on mentioned results, BMSC-EVs^{PEMFs (3 mT)} and the DEX group were conducted for RNA sequencing. The heatmap demonstrated the top 100 genes according to relative expression (Figure 3A). The volcano plot analysis indicated 58 DEGs were upregulated and 54 DEGs were downregulated in the DEX compared with the BMSC-EVs^{PEMFs (3 mT)} (Figure 3B). There results confirmed that the programmed cell death-related gene, Ripk3, was significantly reduced in BMSC-EVs^{PEMFs (3 mT)} group. Furthermore, GO database analysis indicated mechanism involved in bone development and cell death (Figure 3C). Reactome pathway database analysis determined several pathways were involved, such as ECM synthesis, metabolism, immune system, collagen synthesis (Figure 3D). The relevant enriched KEGG pathways analysis in Figure 3E indicated that the regulated genes clustered in cell adhesion and communication, immune-inflammatory regulation, and bone mentalism (such as TGF-beta, Wnt, and Hippo signaling pathways).

Annexin V/PI staining indicated that higher percentage of cell with programmed cell death in the DEX group than that other groups and the BMSC-EVs^{PEMFs (3 mT)} significantly ameliorated the ratio of dying cells (p < 0.001, Figures 4A,C). Western blot in Figure 4B (full-length gels are presented in Supplementary Figures S5–S9) determined a decrease in RIPK1, RIPK3, and mixed lineage kinase domain-like (MLKL). Immunofluorescence analysis in Figures 4E,F and RT-qPCR (Figure 4G) indicated, and RT-qPCR indicated BMSC-EVs^{PEMFs (3 mT)} abolished the increase of RIPK3 and MLKL (p < 0.001) than DEX group, suggesting a therapeutic effect on inhibiting necroptosis.

ONFH rat models were used to investigate if BMSC-EVs^{PEMFs (3 mT)} could effectively prevent the development of ONFH in vivo (Figure 5A). The normal, MPS, EVs^{No PEMFs}, and EVs^{PEMFs (3 mT)} group were established for comparison, and the weights of each group were recorded (Figure 5B). Images of the coronal, sagittal, and transverse planes of micro-CT analyzing short-term effect srevealed that the MPS group suffered significant bone mineral loss, decreased bone density, and the presence of osteonecrosis-like structures below the epiphyseal line compared to the normal group. However, these adverse effects were somewhat improved in the EVs^{No PEMFs} and EVs^{PEMFs (3 mT)} groups, although the results still fell short of expectations (Figure 5C). The BS/BV of the region of interest was significantly increased in EVs^{PEMFs (3 mT)} group compared to EVs^{No PEMFs} group (p < 0.01, Figure 5D). In the long-term observations, images of the planes revealed that the EVs^{PEMFs (3 mT)} group displayed a compact and evenly distributed trabecular bone structure. This suggests that the EVs^{PEMFs (3 mT)} eliminated the negative effect to the femoral head caused by MPS. However, the EVs^{No PEMFs} still did not meet expectations (Figure 5E). The BV/TV, BS/BV, Tb.Th, Tb.N and BMD were significantly increased in EVs^{PEMFs (3 mT)} group compared to EVs^{No PEMFs} group, while Tb.Sp was decreased (p < 0.05, Figure 5F).

H&E staining Figure 6A revealed significant formation of empty bone lacunae, localized disruption of bone trabeculae characterized by a sparse and disorganized structure, as well as the invasion of adipose tissue into the marrow cavity in the short-term samples of the femoral head from the MPS and EVs^{No PEMFs} groups. Most areas of the femoral head in EVs^{PEMFs (3 mT)} group also presented the above changes. Masson staining indicated some new bone trabeculae transitioning from blue to red in EVs^{PEMFs (3 mT)} group. For dissection at long-term shown in Figure 6B, HE staining revealed that the trabecular bone was still occupied by an abundance of cells resembling adipocytes, along with deteriorating cells exhibiting condensed nuclei and encircled by a lucid cytoplasmic area in MPS group, and the trabecular bone became sparse and thin in EVs^{No PEMFs} group. In contrast, rats after the treatment of EVs^{PEMFs (3 mT)} showed only slight osteonecrosis of the trabecular bone, as well as fewer empty lacunae and adipose cells. Short-term immunohistochemistry analysis revealed lower expression of MLKL, and higher expression of RUNX2 in EVs^{PEMFs (3 mT)} group (p < 0.05, Figures 6C,E). For long-term, lower expression of RIPK1 and RIPK3, and higher expression of RUNX2 and OCN in EVs^{PEMFs (3 mT)} group than others (p < 0.05, Figures 6D,F). Furthermore, more mRNA of Ripk1, Ripk3, and Mlkl were expressed in MPS group (p < 0.05), and more mRNA of Ocn and Runx2 were expressed in normal group than others (p < 0.05) (Figure 6G).

Immunohistochemical staining for CD31 and TRAP was conducted to investigate the distribution of blood vessels and osteoclasts within the femoral heads across various groups. The normal, EVs^{No PEMFs}, and EVs^{PEMFs (3 mT)} groups had higher expression of CD31, and the MPS group displayed the lowest levels (Figure 7A). TRAP staining indicated that the osteoclasts in normal and EVs^PEMFs group were inactive than others (Figure 7B). The results confirmed that the MPS injections in rats led to a significant reduction in blood vessel formation and increased activity of osteoclasts in the femoral head. In contrast, BMSC-EVs^{PEMFs (3 mT)} effectively mitigated blood vessel deficiency and osteoclast activity, avoiding the deterioration of ONFH. Additionally, HE staining of vital organs showed no significant biotoxicity from BMSC-EVs and BMSC-EVs^{PEMFs (3 mT)} to, although nephrocalcinosis was observed in the MPS group (Figure 7C).