Artemisic acid (C₁₅H₂₂O_2, purity ≥99.73%) was obtained from MedChemExpress (Shanghai, China). Macrophage colony-stimulating factor (M-CSF) and recombinant mouse RANKL were purchased from R&D Systems (Minneapolis, MN, United States). Phosphate-buffered saline (PBS), Eagle’s minimum essential medium, alpha modification medium (α-MEM), and penicillin/streptomycin were purchased from Gibco BRL (Gaithersburg, MD, United States). Fetal bovine serum (FBS) was obtained from Avantor (Radnor, PA, United States). The Cell Counting Kit-8 (CCK-8) was obtained from Beyotime Biotechnology Inc. (Shanghai, China). TRIzol reagent was purchased from Invitrogen (Carlsbad, CA, United States). The Prime Script RT reagent kit and TB Green™ Premix Ex Taq™ II were obtained from Takara Biomedical Technology (Beijing, China). DAPI, the Actin-Tracker Red-Rhodamine, and DCFH-DA probes were obtained from Beyotime Biotechnology Inc. (Shanghai, China). Primary antibodies targeting p-p65, p-ERK, p-JNK, p-p38, total ERK, total JNK, total p38, and total p65 were purchased from Cell Signaling Technology, Inc. (Beverly, MA, United States). NFATc1, c-fos, IκBα, nrf2, p-nrf2, NQO1, and β-actin antibodies were purchased from Affinity (San Jose, CA, United States). Ti particles (<20 μm diameter) were purchased from Alfa Aesar (Ward Hill, MA, United States). To remove possible adherent endotoxins, the particles were baked for 8 h at 180°C and then washed in 75% ethanol for 48 h. Subsequently, Ti particles were resuspended in sterile PBS at a concentration of 100 mg/mL using an LP vortex mixer (ThermoFisher Scientific, Scoresby, Australia) and stored at 4°C until further use.

We obtained primary BMMs from 6-week-old C57BL/6J mice by irrigating the marrow cavities of the femur and tibia with a sterile syringe (Toro et al., 2012; Li et al., 2013). Cells were cultured in α-MEM complete medium, which is α-MEM with 10% (v/v) FBS and 1% (w/v) penicillin/streptomycin) in a volume containing 30 ng/mL M-CSF in a 5% CO₂ humidified incubator at 37°C for the first 3–5 days. The medium was changed every 2–3 days until the cells reached approximately 90% confluency in the culture dish. After dissociation with trypsin, the cells were transferred to culture plates or dishes for subsequent studies. A schematic design regarding different in vitro incubation times is displayed in Supplementary Figure S1.

The cytotoxic effects of ArA on BMM cells were investigated using a CCK-8 assay. Briefly, BMMs (8 × 10³ cells/well) were transferred to 96-well plates in triplicate and incubated for 24 h in α-MEM complete medium containing 30 ng/mL M-CSF. Cells were then treated with different concentrations of ArA (0, 2.5, 5, 10, 20, 40, 60, 80, and 160 μM) for 72 h, and the culture medium changed after the first 48 h. The CCK-8 assay was performed according to the manufacturer’s instructions. Culture medium (100 μL) containing 10% CCK-8 regent (v/v) was added to each well and incubated at 37°C for 2 h. Absorbance at 450 nm was measured using an Infinite M200 Pro microplate reader (Tecan Life Sciences, Männedorf, Switzerland).

Intracellular ROS were detected using a DCFH-DA probe, according to the manufacturer’s instructions. Briefly, BMMs were seeded into confocal dishes (Cellvis, Mountain View, CA, United States) and cultured with or without 100 ng/mL RANKL, 30 ng/mL M-CSF, and different concentrations of ArA for 24 h. DCFH-DA was diluted 1:1,000 with serum-free medium to a final concentration of 10 μmol/L. The cells were incubated with this medium at 37°C for 20 min and then washed three times with serum-free cell culture medium to remove the uncombined probes. Fluorescence images were captured using a DM4000B epifluorescence microscope (Leica Microsystems, Wetzlar, Germany). Semi-quantitative analysis of ROS-positive cells was performed using ImageJ 1.53t (NIH, Bethesda, MD, United States).

Osteoclast formation was visualized with the tartrate-resistant acid phosphatase (TRAP) regent. BMMs were seeded in 96-well plates in triplicate (8 × 10³ cells/well) for 24 h before α-MEM complete medium was removed from each well. Then, based on the cell viability results, BMMs were incubated with varying non-cytotoxic concentrations of ArA (0, 20, 40, and 80 µM), M-CSF (30 ng/mL), and RANKL (100 ng/mL). The culture medium was changed every 2 days. After approximately 6–8 days, the mature osteoclasts were observed and were subsequently fixed with 4% paraformaldehyde and stained with TRAP regent at 37°C for 1 h. The quantity and surface of TRAP⁺ multinucleated cells were calculated using ImageJ 1.53t (NIH, Bethesda, MD, United States).

Bone resorption was observed through bone slice resorption pits. Bovine bone slices (JoyTech Bio Co., Ltd., Zhejiang, China) were sterilized in 70% ethanol for 2 h, then soaked in PBS overnight to remove the alcohol, and then soaked in α-MEM for 4 h before use. BMMs were transferred in groups of three to 96-well plates with pre-positioned, prepared bone slices. After incubation for 24 h, the osteoclasts were incubated with RANKL (100 ng/mL), M-CSF (30 ng/mL), and different concentrations of ArA (0, 20, 40, and 80 μM) for 8 days. The cells were then removed from the bone slices with sodium hypochlorite. After washing three times with PBS, bone slices were photographed using a scanning electron microscope (Zeiss Sigma 500/Oxford EDS). The area of the resorption pits was determined using ImageJ 1.53t (NIH, Bethesda, MD).

F-actin was visualized by immunofluorescence. BMMs (2 × 10⁴ cells/dish) were seeded in confocal dishes and incubated with M-CSF (30 ng/mL), RANKL (100 ng/mL), and different concentrations of ArA (0, 20, 40, and 80 μM). After the mature osteoclasts were observed, the cells were fixed with 4% paraformaldehyde for 10 min, followed by permeabilization with PBS containing 0.1% Triton X-100 (v/v) (Sigma-Aldrich) for 5 min. F-actin rings were stained with Actin-Tracker Red-Rhodamine, and cell nuclei were stained with DAPI. Actin rings were visualized with a DM4000B epifluorescence microscope (Leica Microsystems, Wetzlar, Germany), and ImageJ 1.53t (NIH, Bethesda, MD) was used to analyze the images.

The transcript levels of osteoclast-related genes were detected by RT-qPCR. BMMs (2 × 10⁵ cells/well) were seeded into 6-well plates supplemented with M-CSF (30 ng/mL), RANKL (100 ng/mL), and different concentrations of ArA (0, 10, 20, 40, and 80 μM) for 3 days. The TRIzol reagent was used to lyse the cells and extract total RNA. The entire experimental process of RT-qPCR strictly followed the MIQE guidelines (Taylor et al., 2010). RNA concentration and purity were determined using a Nanodrop 2000 instrument (Thermo Fisher Scientific, Waltham, MA, United States), ensuring that OD260/280 was between 1.8 and 2.0. Equal amounts of RNA (1,000 ng) were then reverse transcribed to synthesize complementary DNA (cDNA) to avoid inflating the differences between groups. According to the manufacturer instructions for TB Green™ Premix Ex Taq™ II, a 10 μL reaction system was established and used for the RT-qPCR assay on an ABI 7500 Sequencing Detection System (Thermo Fisher Scientific). Cycling conditions were set to 40 cycles (95°C for 5 s and 60°C for 30 s). Melting curves were examined to verify amplification specificity. Relative gene expression was calculated using the comparative 2^−ΔΔCT method (Livak and Schmittgen, 2001). The primer sequences were as shown in Table 1, among which Gapdh was used as the housekeeping gene.

Osteoblastogenesis was visualized by alkaline phosphatase (ALP) and alizarin red staining of bone mesenchymal stem cells (BMSCs). BMSCs were obtained from marrow cavities of the femur and tibia in 4–6-week-old male C57BL/6 mice, which were cultured with a complete medium consisting of α-MEM supplemented with 10% (v/v) FBS and 1% (w/v) penicillin/streptomycin. BMSCs (10 × 10⁴ cells/well) were seeded in 24-well plates and cultured with osteogenic medium (50 μg/mL ascorbic acid, 5 mM glycerol phosphate) with or without different concentrations of ArA. ALP and alizarin red staining were respectively performed on the 7th and 21st days of osteogenic differentiation induction, and the culture medium was changed every 2 days.

Protein expression levels during osteoclast formation were detected using Western blotting. To detect the effect of ArA on the Nrf2, MAPK, and NF-κB signaling pathways, BMMs were seeded into 6-well plates at a density of 3 × 10⁵ cells/well with M-CSF (30 ng/mL) and different concentrations of ArA for 24 h. BMMs were then incubated with RANKL (100 ng/mL) for 25 min. To explore the time-dependent effects of ArA on NFATc1 and c-fos, the BMMs were incubated with ArA (40 µM), RANKL (100 ng/mL), and M-CSF (30 ng/mL) for different time periods. To determine the concentration-dependent effects of ArA on NFATc1 and c-Fos, cells were incubated with RANKL (100 ng/mL), M-CSF (30 ng/mL), and various doses of ArA for 4 days. The cells were washed with PBS and lysed with RIPA buffer with a protease and phosphatase inhibitor cocktail (Beyotime Biotechnology Inc.). After centrifugation, the collected proteins were dissolved in loading buffer, separated by 10% SDS-PAGE, and then transferred to 0.22 μm polyvinylidene fluoride membranes. The membranes were blocked with 1×TBST (Tris-buffered saline with Tween 20) containing 5% BSA on a shaker for 1 h at room temperature and were incubated with the relevant primary antibodies overnight at 4°C. After washing three times with TBST, the membranes were incubated with secondary antibodies for 70 min at room temperature. Blot detection was performed using an Odyssey V3.0 image scanning system (Li-COR Inc., Lincoln, NE, United States). Blotting results were analyzed and quantified using ImageJ 1.53t (NIH, Bethesda, MD, United States).

The animal protocol was approved by the Animal Care and Experiment Committee of the Ninth People’s Hospital Affiliated with the Shanghai Jiao Tong University School of Medicine (SH9H-2023-A862-1). All experiments were conducted in accordance with the American Psychological Association’s guidelines for Ethical Conduct in the Care and Use of Nonhuman Animals in Research. Based on previous reports (Liao et al., 2019; Liao et al., 2021; Wu et al., 2022), a Ti particle-induced mouse calvarial osteolysis model was used to evaluate the in vivo osteoprotective effect of ArA. Twenty 6-week-old male C57BL/6J mice (Jihui Laboratory Animal Breeding Co., Ltd., Shanghai, China), approximately weighing 22 ± 2 g, were randomly assigned to four different treatment groups with five mice each: 1) sham group (with PBS injection); 2) vehicle group (with 25 mg of Ti particles on the calvarial surface); 3) low-ArA group (with 25 mg of Ti particles on the calvarial surface and 5 mg/kg/d ArA in the abdominal cavity); 4) high-ArA group (with 25 mg of Ti particles on the calvarial surface and 10 mg/kg/d ArA intraperitoneally). To reduce the risk of ArA toxicity in mice, the doses used were determined by previous reports (Lee et al., 2017; Zhang, 2020) and our pilot study. Briefly, under 1.4% isoflurane general anesthesia, the mouse periosteum was separated from the calvaria, and a collagen sponge soaked in sterile PBS or titanium particle PBS solution was embedded under the periosteum at the midline of the calvaria. On the day after implantation, PBS with or without ArA (5 or 10 mg/kg/d) was injected intraperitoneally for 14 days. After this treatment, the mice were sacrificed, and the entire calvariae were dissected and washed with PBS. After removing the titanium particles, the calvariae were fixed by soaking in 4% PFA for 48 h for subsequent analysis.

High-resolution μCT-100 micro-CT (SCANCO Medical AG, Brüttisellen, Switzerland) with an isometric resolution of 10 μm was used to analyze the fixed calvarias. Microstructural indicators of bone volume/tissue volume (BV/TV) and bone mineral density (BMD) were determined in the regions of interest using supporting software (version 6.5-3, SCANCO Medical). Both indices are key morphometric parameters of bone mass.

After micro-CT scanning, the calvaria samples were decalcified using 10% EDTA (pH = 7.4) for 14 days, embedded in paraffin, and prepared as histological sections before being stained with hematoxylin and eosin (H&E) and TRAP. Immunofluorescence staining was performed with antibodies against NFATc1 and Nrf2 (Affinity, San Jose, CA, United States). For the in vivo biosafety of ArA, the major organs of the mice, including the heart, liver, and kidneys, were collected at the same time when sacrificed 14 days after treatment and subsequently H&E stained. Images of stained sections were acquired with a Panoramic MIDI Scanner (3DHistech, Budapest, Hungary). The osteoclast/bone surface ratio (N.Oc/BS) and the osteoclast surface/bone surface ratio (OcS/BS) were calculated using ImageJ 1.53t.

GraphPad Prism v9.3.1 (GraphPad Software Inc., San Diego, CA, United States) was used for statistical analysis of the results. The data were all represented as the mean ± standard deviation (SD) of three or more independent experiments. The data were first tested for homogeneity of variance and then subjected to one-way analysis of variance (ANOVA), followed by Tukey’s post hoc analysis. Significant differences were determined at *p < 0.05, **p < 0.01, and ***p < 0.001.

To assess the effect of ArA on RANKL-induced osteoclastogenesis, osteoclast precursor BMMs were incubated with two cytokines necessary for osteoclast stimulation, namely, M-CSF and RANKL, and different concentrations of ArA. A CCK-8 assay was used to test the non-cytotoxic dose range of ArA on BMMs. The results showed that ArA at concentrations ≤80 μM did not impair BMM cell viability (Figures 1A, B), demonstrating that ArA had relatively low cytotoxicity and suggesting that the inhibitory effects of ArA on osteoclast differentiation and function were not caused by cytotoxicity. The calculated IC50 value was 1.147 mM (Figure 1C). Therefore, in subsequent in vitro experiments, different non-cytotoxic concentrations of ArA were used to observe their effect on multinucleate osteoclasts. TRAP staining revealed that ArA inhibited osteoclast abundance and size in a concentration-dependent manner (Figures 1D, E). Compared with that in the RANKL-only group, the formation of osteoclasts in the 20 μM ArA treatment group was reduced by approximately 30%, and osteoclast area was reduced by approximately 50%. In the 80 μM ArA treatment group, these two values increased to approximately 75%. The results showed that ArA mainly exerted an inhibitory effect in the first 4 days of osteoclast formation and had no obvious inhibitory effect during days 5–7 (Figures 1F, G). In summary, ArA inhibited RANKL-induced osteoclast formation in a dose- and time-dependent manner in vitro.

The adhesion of osteoclasts to the bone surface is crucial for their bone resorption function. The formation of dense, ribbon-like F-actin structures is one of the changes in mature osteoclast polarization and morphology (Teitelbaum, 2000; Vaananen et al., 2000) and plays an essential part in the resorption function of osteoclasts (Garbe et al., 2012). Therefore, we examined whether ArA affected F-actin ring formation using immunofluorescence analysis. ArA dose-dependently inhibited osteoclasts from forming intact F-actin rings, which was not observed in the control group treated with RANKL only (Figure 2A). The quantity of osteoclasts and their nuclei was reduced after ArA treatment (Figure 2B). Consistent results were observed for the erosion of bovine bone slices; scanning electron microscopy images (Figure 2C) showed that the area of resorption pits decreased as the concentration of ArA increased (Figure 2D). Therefore, ArA inhibited osteoclast F-actin formation and bone resorption function in a concentration-dependent manner.

RT-qPCR was used to further explore the effect of ArA on formation- and function-related gene transcription in RANKL-induced osteoclasts at the mRNA level. As shown in Figure 3A, stimulation of RANKL caused BMMs to differentiate into osteoclasts. The expression of osteoclast-related genes in each group induced by RANKL was significantly increased compared with the control group (p < 0.001). The mRNA transcript levels of osteoclast formation- and function-related genes in the ArA-treated group were lower than those with only RANKL stimulation. TRAP and CTR are closely related to differentiation, whereas DC-STAMP and CTSK are related to osteoclast fusion and bone resorption, respectively (Garbe et al., 2012). c-Fos and NFATc1 are the key downstream transcription factors that initiate osteoclast differentiation (Gohda et al., 2005). In the control RANKL-stimulated osteoclasts, protein levels of NFATc1 and c-Fos were enhanced, reaching the highest level on day 3 and then decreasing slightly by day 5. Protein expression at these two time points was inhibited by ArA (Figures 3B, C). To further explore the effects of ArA, BMMs were treated with different concentrations of ArA. The protein levels of NFATc1 and c-Fos were reduced upon ArA treatment, and as the concentration increased, the inhibition became more significant (Figures 3D, E). To investigate the effects of ArA on osteoclast formation at different stages, we have included experiments using RAW264.7 murine macrophages alongside BMMs. We verified by the CCK-8 kit that ArA has no cytotoxicity to RAW264.7 murine macrophages in the concentration range of 0–320 μM. Then we stimulated RAW264.7 murine macrophages with RANKL and co-cultured them with different concentrations of ArA for 6 days, and we detected the levels of osteoclast-related transcription factors by RT-qPCR. The results are shown in Supplementary Figure S2, which, together with the BMMs experimental results, indicate that ArA not only has a significant effect on inhibiting the differentiation of BMMs into osteoclasts but also inhibits the differentiation of RAW264.7 murine macrophages into osteoclasts. Therefore, ArA suppressed the expression of osteoclast formation- and function-related genes and proteins in a time- and concentration-dependent manner.

Bone homeostasis is a complex process that requires the coordinated actions of osteoblasts and osteoclasts (Simon et al., 2017). Therefore, in addition to the effect on osteoclast differentiation and bone resorption function, it is also important to study the effect of ArA on osteoblast differentiation and mineralization. First, a CCK-8 assay was performed to assess the cytotoxicity of ArA on BMSCs (Supplementary Figure S3A). The results showed that ArA at concentrations ≤160 μM did not impair BMSC cell viability. To explore the effect of ArA on osteoblastogenesis and mineralization, BMSCs were cultured in osteogenic induction medium with different concentrations of ArA (0, 40, and 80 μM). Cells were stained with ALP and alizarin red after 7 and 21 days of culture, respectively (Supplementary Figures S3B, C). The results indicated that ArA had no significant effect on osteoblast differentiation and mineralization.

Western blotting was used to study the role of the NF-κB and MAPK pathways, two important osteoclastogenic signaling pathways upstream of NFATc1 and c-Fos, respectively, in ArA’s inhibitory effect on osteoclast differentiation and function. The results showed that the phosphorylation levels of MAPK pathway elements (i.e., ERK, JNK, and p38) were significantly upregulated after RANKL stimulation for 25 min (Figures 4A, B). This upregulation was inhibited by ArA treatment in a concentration-dependent manner. Specifically, the phosphorylated protein ratio in the 80 μM ArA treatment group was approximately 50% that in the control group. Similar results were observed for the phosphorylation of p65 in the NF-κB pathway. Furthermore, ArA treatment visibly inhibited the RANKL-induced degradation of IκBα (Figures 4C, D), which is an inhibitor of NF-κB. Therefore, ArA inhibited RANKL-induced MAPK and NF-κB pathway activation in a dose-dependent manner.

Considering that increased intracellular ROS levels are indispensable for osteoclastogenesis (Lee et al., 2005) and that ArA reportedly has antioxidant effects (Ferreira and Luthria, 2010), we further explored the effect of ArA on intracellular ROS levels during osteoclast formation using a DCFH-DA probe. Intracellular ROS levels were significantly increased by RANKL stimulation. The relative fluorescence intensity showed that ArA treatment significantly suppressed this effect, with the ROS level in the 40 μM treatment group suppressed to less than 5% of its normal level (Figures 5A, B). To further investigate the molecular mechanism by which ArA inhibits intracellular ROS levels, Western blotting was used to study the Nrf2 signaling pathway, which is closely related to the cellular restriction of ROS production. The results demonstrated that RANKL stimulation-induced Nrf2 phosphorylation was attenuated by ArA treatment. In addition, ArA increased the levels of Nrf2 proteins and its downstream target, NQO1 (Figure 5C). Our analysis indicated that RANKL stimulation did not significantly impact the protein expression of Nrf2 or NQO1, whereas treatment with 20 μM ArA increased Nrf2 and NQO1 by approximately 40% (Figure 5D). Therefore, ArA suppressed intracellular ROS levels in cells undergoing osteoclastogenesis by activating Nrf2 and downstream NQO1 dose-dependently.

Based on the above in vitro experiments showing that ArA inhibits osteoclast formation, we constructed a mouse Ti particle-induced calvarial osteolysis model to investigate the effect of ArA on osteoclast-associated osteolysis in vivo and explore its potential applications in treating PPO. Micro-CT results indicated that the Ti particles induced extensive bone erosion on the calvarial surface, with large, deep pits in the vehicle-treated group. However, ArA treatment rescued osteolysis in a concentration-dependent manner, as the extent of bone erosion in the 5 mg/kg ArA treatment group was visibly lower than that in the vehicle group, and this effect was even more significant in the 10 mg/kg ArA treatment group (Figure 6A). Quantitative analysis of bone parameters further assured the effect of ArA. Specifically, the BMD and BV/TV of the ArA-treated group were significantly increased compared with those of the vehicle group (Figure 6B).

The results of the histomorphological evaluation were consistent with those of the micro-CT analysis, further confirming that ArA inhibited Ti particle-induced osteolysis. Specifically, H&E staining showed that localized Ti particles caused extensive osteolysis, whereas ArA alleviated this effect (Figure 6C). TRAP staining showed that TRAP⁺ osteoclasts in the vehicle group were abundant and widely distributed, and clear bone resorption was observed. In the ArA-treated group, the number of multinucleated TRAP⁺ cells and the OcS/BS ratio were both significantly reduced (Figure 6D), as was the degree of osteolysis. Immunofluorescence analysis was used to assess the molecular mechanisms of the ArA effect in vivo. Figure 7 shows that compared with the sham operation group, the NFATc1 level in the Vehicle group was significantly increased (p < 0.001), while there was no significant difference in Nrf2 levels between the two groups. Compared with the vehicle group, the NFATc1 level in the ArA treatment group decreased significantly (p < 0.05), and the 10 mg/kg ArA treatment group decreased more significantly (p < 0.01). Moreover, we tested the biosafety of ArA in vivo. H&E staining revealed no obvious adverse effects on the histological structures of major organs, including the heart, liver, and kidneys, as shown in Supplementary Figure S4. Overall, immunohistochemical results further demonstrated the concentration-dependent protective effects of ArA on Ti particle-induced calvarial osteolysis.