Synovial membranes from the involved hip were obtained from four patients who underwent total hip revision in our institute from Jan 1st, 2019 to Dec 31st, 2019 due to aseptic loosening of their implant. Samples of synovial membranes were also obtained from four patients who underwent primary total hip replacement or hemiarthroplasties because of hip fracture. Demographic data for these eight patients are shown in Supplementary Table S1. This study was approved by the institutional review board of our hospital. Signed informed consent for participation and publication were obtained from all patients and their relatives.

Wear particles were made from the femoral head implant (CoCrMo alloy, Modular prosthesis head, Link) of patient No. 2 (details shown in Supplementary Table S1) who was diagnosed with aseptic loosening. All experiments were performed according to Chinese legal requirements. The removed femoral head was sterilized and subsequently transformed into nanoscale wear particles using a fabricated high-vacuum three-electrode direct current previously described by our group (Deng et al., 2021). The appearance of the wear particles was assessed by field emission scanning electron microscopy (Merlin, Carl Zeiss AG, Germany) and by transmission electron microscopy (JEM-2100, JEOL, Japan). Size distribution analysis was performed with a Nano Sizer and zeta-potential tester (Zetasizer Nano ZS, Malvern, United Kingdom). The chemical composition of wear particles was analyzed using tungsten filament scanning electron microscopy (Q25, FEI, OR, United States). All wear particles were washed with 75% ethanol and sterilized with ethylene oxide. The absence of endotoxin was confirmed using a commercial limulus assay kit (QCl-1000; Bio Whittaker, Walkersville, MD, United States). CoCrMo particles were then suspended in PBS and adjusted to a concentration of 8 × 10^–3 g/ml for in vitro experiments.

Bone marrow cells were collected from the femur and tibia of 12-week-old C57BL/6J female mice. BMDMs were obtained after co-culture of bone marrow cells in Dulbecco’s modified Eagle’s medium containing 10% fetal bovine serum, 10 ng/ml recombinant M-CSF, penicillin, and streptomycin for 7 days. All BMDMs were cultured at 37°C in 5% CO₂, followed by seeding for 24 h in high-glucose Dulbecco’s modified Eagle’s medium containing 10% fetal bovine serum, penicillin, and streptomycin. Cell counters (Countstar BioTech) were used to show at least 98% cell viability in all trials. The identification of BMDMs was performed by flow cytometry analysis with the marker being F4/80 + CD11b+.

ICA was administrated simultaneously with wear particles in both PCR and ELISA testing. At specific time points after stimulation, RNA was extracted using an RNA extraction kit according to the manufacturer’s instructions (9109, TaKaRa Biotechnology). PrimeScript RT Master Mix (RR036D, TaKaRa Biotechnology) was used to reverse-transcribe RNA to cDNA. Real-time PCR was performed using UNICONTM qPCR SYBR Green Master Mix (11198ES08, Yeasen) and a LightCycler 96 real-time PCR system (Roche Molecular Systems, Inc) to measure the expression of TNF-α and IL-6. GAPDH was used as the housekeeping gene. All reactions were run in triplicate. Primers for the target genes are listed in Supplementary Table S2.

For ELISA measurement, the cell supernatants were harvested and centrifuged to remove cellular debris. TNF-α and IL-6 levels were subsequently assessed using an instant enzyme-linked immunosorbent assay (ELISA) kit (Neobioscience Technology Co., Ltd) according to the manufacturer’s instructions.

Following stimulation by wear particles for 1 h and other interventions, BMDMs were lysed using a RIPA buffer (9806S, Cell Signaling Technology) containing PMSF (P0100-1, Solarbio) and a phosphatase inhibitor cocktail (CW2383, CW Biotech). Proteins were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and then transferred to polyvinylidene fluoride membranes. The membranes were cut into small pieces according to different protein molecular masses and then probed with the relevant antibodies overnight at 4°C. Details of the antibodies are listed in Supplementary Table S3. A chemiluminescence detection system (Syngene, Cambridge, United Kingdom) was used to visualize the protein bands and further analysis was performed using Gene Tools (Syngene, Cambridge, United Kingdom).

One hour after being stimulated with wear particles, BMDMs were incubated with p65 (1:400, rabbit, 8242, Cell Signaling Technology) or ERα (1:200, rabbit, ab32063, abcam) antibodies overnight at 4°C. Subsequently, BMDMs were washed three times with PBS followed by incubation with secondary IF antibody (goat anti-rabbit, A32740, Thermo Fisher Scientific) at 20°C for 1 h. Thereafter, BMDMs were treated with an anti-fade, fluorescence mounting medium with DAPI (HNFD-02, HelixGen Co., Ltd). A confocal microscope (LSM 710, Carl Zeiss) was used to visualize IF.

BMDMs were stimulated by wear particles for 24 h and then treated with TrypLE Express Enzyme (12604021, Thermo Fisher Scientific) to detach cells from the plates. They were then washed with Perm/Wash buffer, suspended in PBS and analyzed using flow cytometry (BD FACSVerse, BD Biosciences). After centrifugation at 1,000 rpm for 5 min, BMDMs were sequentially incubated with APC-conjugated CD206 antibody (0.25ug/test, 17-2061-82, Thermo Fisher Scientific) and PE-cyanine7-conjugated iNOS antibody (0.06ug/test, 25-5920-82, Thermo Fisher Scientific). The isotype controls used were APC-conjugated rat IgG2b, κ (17-4031-82, Thermo Fisher Scientific) and PE-cyanine7-conjugated rat IgG2a, κ (25-4321-82, Thermo Fisher Scientific).

To generate ERα knockdown macrophages, siRNA that targets Esr1 (GeneID: 13982) was constructed into the plasmid and recombinant lentiviruses aimed at Esr1 were obtained from Cyagen. The specific siRNA sequences are listed in Table 1. For transient transfections, 5×10⁵ BMDMs were seeded for 24 h and nurtured in Opti-MEMI Reduced Serum Medium (31985070, Thermo Fisher Scientific) for 12 h. Subsequently, 200 pmol siRNA together with 7.5 μL Lipofectamine RNAiMAX Transfection Reagent (13778075, Thermo Fisher Scientific) were added for a 24 h transfection period and the ERα knockdown efficiency then evaluated by qPCR assay. For the production of stable ERα knockdown macrophages, BMDMs were cultivated with 1 × 10⁷ transducing unit per mL lentiviruses for 24h and 3ug/mL puromycin was used to screen the transfected macrophages.

RNA sequencing was performed after the RAW264.7 macrophages had been stimulated by wear particles for 4 h. PPI network construction and identification of hub genes has been described previously (Qiu et al., 2020). Briefly, the STRING database and Cytoscape software were used to construct a PPI network of differentially expressed genes. The topology property of the network was analyzed using the MCODE application of Cytoscape software. The functional clustering of differentially expressed genes was performed using the Metascape online tool (https://metascape.org).

Twenty-five C57/BL6J female mice aged 10 weeks were purchased from the Laboratory Animal Research Center of the South China University of Technology. All experiments were approved by the Institutional Animal Care and Use Committee, Guangdong Province People’s Hospital. Guidelines for the care and use of laboratory animals were strictly followed. Each mouse weighed 20–25 g at the beginning of the study, and all had unlimited access to food and water.

The wear particle-induced mouse calvaria osteolysis model was established when mice reached the age of 12 weeks. The mice were divided into 5 groups: control (particle-free), wear particle (Pa), wear particle + icariin (Pa + ICA), wear particle + icariin + ERα antagonist methyl-piperidino-pyrazole (MPP) (Pa + ICA + MPP), and wear particle + icariin + ERα-siRNA (Pa + ICA + siRNA) group. There were five C57BL/6J mice in each group. As we previously described (Fu et al., 2018), the center of the calvaria was exposed and the periosteum removed. In the Pa, Pa + ICA, and Pa + ICA + MPP groups, 50 ul of PBS containing 0.3 mg wear particles was spread onto the center of the calvaria. In the Pa + ICA + siRNA group, 50 ul PBS containing 0.3 mg wear particles, as well as 25 ul shRNA-ERα lentivirus (titer: 1×10⁸ TU/mL) were locally spread onto the center of the calvaria before suturing according to our established method (Zhang et al., 2018). In the Pa + ICA, Pa + ICA + MPP, and Pa + ICA + siRNA groups, ICA (2.8 mg/50 μL, B21576, YuanYeBio, China) and MPP (5.6 μg/50 μL, M7068, Sigma) were continuously and subcutaneously delivered by minipumps (micro-osmotic pump, model 1,004, 0.11 μL/h*28 days, Alzet, United States). The dosages of ICA and MPP used in this animal model were obtained from previous studies (Davis et al., 2008; Shao et al., 2015). Minipumps were subcutaneously implanted at a site slightly posterior to the scapulae. No mice died and no wound complications were observed during the experiments. All mice were sacrificed in a CO₂ chamber at day 14 after surgery.

After removing all the soft tissue, the calvarias were analyzed by micro-CT and associated analysis software (SkyScan 1176; SkyScan, Kontich, Belgium). The scanning protocol was set at an isometric resolution of 18 μm and radiographs were acquired at 45 kV and 500 uA through a 0.2 mm-thick aluminum filter with an exposure time of 240 ms. 3D image reconstructions were obtained using manufacturer’s software (NRecon, SkyScan, Kontich, Belgium). The region of interest (ROI) was set as previously described (Fu et al., 2018). BV/TV, BMD and the total porosity of each sample were obtained using CTAn software (SkyScan, Kontich, Belgium).

Cross-sections (4 µm) of the calvarias were cut and stained with hematoxylin and eosin (H&E) and with a commercial tartrate-resistant acid phosphatase (TRAP) staining kit (Jiancheng Bio Ins, Nanjing, China). Histological sections were analyzed using a standard, high-quality light microscope (DM 2000, Leica, Germany) at a magnification of 10X with the midline suture in its center. Histomorphometric analysis was performed on the most central section and on four adjacent sections using image analysis software (Image Pro-Plus 6.0, Media Cybernetics, Bethesda, MD). The definition of ROI, as well as measurements of the eroded bone area (mm²) and number of osteoclasts was performed as per earlier studies (Sawyer et al., 2003; von Knoch et al., 2004). Briefly, the eroded bone area (mm²) was determined by tracing the area of soft tissue between the parietal bones, including resorption pits on the superior surface of the calvaria. The presence of dark-purple-stained granules in the TRAP staining section, which is located on the bone perimeter within a resorption lacuna, was automatically recognized as osteoclasts by the image analysis software after manual setting of the color parameters.

The protocol for histological detection of TNF-α, IL-6, and ERα in human synovial membranes and mice calvarias was as follows. All sections first underwent endogenous peroxidase blocking and antigen retrieval. After being incubated with goat anti-mouse IgG, sections were then incubated overnight at 4°C with primary antibodies for TNF-α (ab183218, abcam), IL-6 (ab233706, abcam), and ERα (ab32063, abcam). Following incubation with the appropriate secondary antibodies, coloration of the sections was achieved with 3,3-diaminobenzidin. Rinsed sections were counterstained with hematoxylin. Three views of each section with the magnification of 100x were evaluated and the Bresalier’s semiquantitative scoring system was used to quantify the staining intensity (Bresalier et al., 1997). Briefly, the staining intensity (0 for no stain; 1 for weak stain; 2 for moderate stain; 3 for strong stain) and the percentage of cells with specific staining intensity (P0, P1, P2, P3) of each section were calculated independently by two pathologists. Subsequently, the immunoreactivity score for each section was calculated as follows: ∑(0×P0+1×P1+2×P2+3×P3).

All the results of in vitro studies were obtained from at least three independent experiments, with data expressed as mean ± SD. The values were first assessed using the Kolmogorov-Smirnov test to verify data normality. The two-sided Student’s t-test was used to compare two groups. One-way analysis of variance was used for three or more groups, and Tukey test was used for post hoc comparisons. Statistical analyses were performed using SPSS 20.0 software, with p < 0.05 considered to be statistically significant.

Scanning electron microscopy and transmission electron microscopy images of the nanoscale wear particles are shown in Figures 1A,B. The mean size of the particles was 150.2 ± 37.4 nm and their size distribution is shown in Figure 1C. The chemical composition of the wear particles was shown in Figure 1D and the normalized mass of each element was available in Table 2. Consistent with the known composition of the CoCrMo alloy femoral head (Link), the leading metallic elements of the nanoscale wear particles were Cr and Co.

Marked wear particle infiltration was found in H&E-stained sections of the synovial membranes obtained from the involved hip of patients diagnosed with aseptic loosening (Figure 2A). Immunohistochemical staining of the synovial membranes revealed significantly higher staining intensities for TNF-α (+2.10 ± 0.37, p < 0.01) and IL-6 (+1.57 ± 0.29, p < 0.01) in the aseptic loosening group, as calculated by Bresalier’s semiquantitative scoring system (Figures 2A,B).

Results of flow cytometry for the identification of BMDMs were provided in Supplementary Figure S1. TNF-α and IL-6 mRNA levels increased in a time-dependent manner following stimulation of BMDMs by CoCrMo wear particles (Figure 2C). Similarly, ELISA revealed that TNF-α and IL-6 secretion by BMDMs increased significantly and gradually after 4, 8, 12 and 24 h stimulation by wear particles. To further investigate the molecular mechanism underlying wear particle-induced activation of inflammation, WB was used to evaluate protein expression for NF-κB signaling pathway components. Marked increases in the expression of p-IKKβ, p-p65 and p-IκBα were observed, but no obvious changes in the total protein expression of IKKβ, p65 and IκBα (Figure 2D). Confocal microscopy also revealed translocation of p65 protein (red) from the cytoplasm into the nucleus (blue) following CoCrMo particle stimulation of BMDMs (Figure 2E).

Figure 3A shows the chemical structure of ICA. The inhibitory effect of ICA on CoCrMo wear particle-induced expression of TNF-α and IL-6 mRNA in BMDMs occurred in a dose-dependent manner (10⁻⁸ to 10⁻⁶ M), as shown in Figure 3B. The maximum inhibitory effect of ICA was found at a concentration of 10⁻⁶ M. CCK-8 assay was also performed to determine the influence of ICA concentration on the proliferation and survival of BMDMs. No significant changes in cell viability were observed at ICA concentrations ranging from 10⁻⁸ to 10⁻⁶ M (Figure 3B).

ICA was therefore used at a concentration of 10⁻⁶ M in the subsequent in vitro experiments. Compared to the CoCrMo group, the CoCrMo + ICA group showed significantly lower TNF-α and IL-6 mRNA levels and protein secretion at 4, 8, 12 and 24 h after CoCrMo particle stimulation (Figure 3C). These results suggest that the inhibitory effect of ICA on the CoCrMo wear particle-induced macrophage inflammatory response was related to the downregulation of NF-κB signaling pathways. Marked decreases in p-IKKβ, p-p65 and p-IκBα protein expression were observed after ICA treatment (Figure 3D). IF staining showed that CoCrMo-induced translocation of p65 (red) from the cytoplasm into the nucleus (blue) was consistently inhibited in the CoCrMo + ICA group (Figure 3E). Activation of the NF-κB pathway is thought to be responsible for M1 polarization of macrophages, which is pivotal for the wear particle-induced inflammatory response. Therefore, we also evaluated macrophage polarization in the CoCrMo and CoCrMo + ICA groups using flow cytometry. Following ICA treatment, a significantly lower mean fluorescent expression (-58.7 ± 12.2%, p < 0.01) of iNOS and a significantly higher mean fluorescent expression (70.8 ± 50.6%, p < 0.05) of CD206 were observed (Figure 3F).

RNA sequencing and PPI network construction was performed to further investigate the underlying mechanism of wear particle-induced osteolysis and to identify potential hub genes. The volcano plot and heat map for this analysis are shown in Figures 4A,B, respectively. A total of 302 upregulated mRNAs and 184 downregulated mRNAs were profiled (log 2 FC > 1, FDR <0.05). The PPI network from the STRING database showed that ERα has a close relationship with TNF-α and IL-6 (Figure 4C).

As shown in Figure 4D, ICA induced a marked increase in the protein expression of both phospho-ERα Ser118 and phospho-ERα Ser167 while inhibiting the wear particle-related macrophage inflammation response. Confocal microscopy also showed translocation of ERα from the cytoplasm into the nucleus following ICA treatment, whereas P65 translocation towards the nucleus was suppressed (Figure 4E).

Furthermore, both ERα antagonist (MPP) and ERα-siRNA interference reversed the suppressive effect of ICA on the CoCrMo particle-induced mRNA transcription and protein secretion of TNF-α and IL-6 in BMDMs (Figure 4F). Similar changes were also noted for activation of the NF-κB signaling pathway. No significant differences in p-IKKβ, p-p65 or p-IκBα protein expression were found between the CoCrMo, CoCrMo + ICA + MPP and CoCrMo + ICA + ERα-siRNA groups (Figure 4D). Consistent with this, the inhibitory effect of ICA on CoCrMo particle-induced translocation of P65 from the cytoplasm into the nucleus was reversed by both ERα antagonist and ERα-siRNA interference (Figure 4E). Similar trends were also observed regarding the influence of these two interventions on macrophage polarization (Figure 4G).

3D reconstruction of micro-CT images showed that CoCrMo particles induced profound changes in calvarial bone microarchitecture, as well as significantly decreased BMD (0.044 ± 0.034 g/cm³, p < 0.05) and BV/TV (−5.4 ± 2.5%, p < 0.01) (Figure 5). Consistent with the results from in vitro experiments, 3D images (Figure 5A) showed that ICA exerted a marked protective effect on CoCrMo particle-induced osteolysis. No significant differences in bone microstructure parameters were found between the ICA and control groups (Figures 5B–D). It was also noted that the protective effect of ICA on BMD, BV/TV and total porosity was reversed by ERα antagonist and by ERα-siRNA interference.

HE and TRAP staining showed that ICA treatment suppressed CoCrMo wear particle-induced mouse calvarial osteolysis and osteoclastogenesis (Figure 6A). This was confirmed by quantitative measures of eroded bone area (−0.43 ± 0.19 mm², p < 0.001) and of TRAP-positive cell number (−30.0 ± 19.4, p < 0.05) (Figures 6B,C). Significant increases in the eroded bone area and TRAP-positive cell numbers were found in the CoCrMo + ICA + MPP and CoCrMo + ICA + ERα-siRNA groups compared to the CoCrMo + ICA group. Representative images at a higher magnification of TRAP staining sections were provided in Supplementary Figure S2. Similar findings were observed for local inflammatory infiltration, as evaluated by TNF-α staining (Figure 6A) and subsequent semiquantitative analysis (Figure 6D). The efficiency of ERα knockout in the mice calvaria model by the injection of shRNA-ERα lentivirus was supported by the results of immunohistochemical staining for ERα in the calvarial sections (Figure 6A). Semiquantitative analysis showed that there was a significant decrease of the ERα immunoreactivity score in the ERα-siRNA targeting group when compared to the control group (Figure 6E).