Osteoporosis treatment via programmed delivery of IRF5 siRNA to macrophages in bone tissue combined with ROS scavenging

Abstract

Postmenopausal osteoporosis (PO) is not merely a consequence of increased osteoclast activity. Estrogen deficiency induces oxidative stress and pro-inflammatory M1 macrophage polarization in bone tissue, thereby sustaining osteoclastogenic stimuli and pathological bone resorption. Interferon regulatory factor 5 (IRF5) is a key regulator of M1 polarization, and silencing its gene by small interfering RNA (siIRF5) may attenuate the pathogenesis of PO. However, selective delivery of siIRF5 to M1 macrophages within the osteoporotic bone microenvironment is challenging. Herein, we develop a nanomedicine, termed siP@WM, for programmed sequential targeting of bone tissue and macrophages. The siP@WM integrates a platinum nanozyme core with mesoporous silica shell loading siIRF5. The shell with enlarged pores (10–15 nm) provides effective siRNA encapsulation and programmable surface grafting of two kinds of poly(ethylene glycol) (PEG) derivatives. The longer poly(aspartic acid)-terminated PEG chain promotes bone enrichment. ROS-triggered thioketal cleavage shortens this chain, reduces bone-matrix adhesion, and exposes the shorter folic acid-functionalized PEG chain to enhance M1 macrophage uptake. In vitro, siP@WM scavenges ROS, facilitates endosomal escape, achieves approximately 60% IRF5 knockdown, and promotes M2 polarization. In ovariectomized mice, systemic administration of siP@WM ameliorates bone loss, restores trabecular microarchitecture and mechanical strength, and shows no overt short-term toxicity in major organs. Transcriptomic and histological analyses support attenuated AP-1-associated inflammatory signaling, suppressed osteoclast differentiation, and modulated extracellular matrix remodeling. Together, programmed siIRF5 delivery to macrophages in bone tissue combined with ROS scavenging restores bone homeostasis and provides an adaptive biomaterial strategy for PO treatment.

Graphical abstract

Highlights

• •Mesoporous silica shells encapsulate platinum nanozyme and load siRNA.
• •Reactive oxygen species switch the carrier from bone retention to macrophage uptake.
• •Sequential bone–macrophage targeting improves siRNA delivery efficiency.
• •ROS scavenging and M2 polarization jointly suppress osteoclast overactivation.
• •Nanotherapeutic system ameliorates bone loss in ovariectomized mice.

1. Introduction

Postmenopausal osteoporosis (PO) is a prevalent metabolic bone disease characterized by decreased bone mass and deterioration of bone microarchitecture caused by estrogen deficiency [1]. PO-associated osteoporotic fractures significantly impair patients’ quality of life [2] and represent a growing public health burden [3]. PO is not merely a consequence of increased osteoclast activity. Increasing evidence indicates that oxidative stress and macrophage polarization imbalance play central roles in the progression of PO. Estrogen deficiency promotes the accumulation of reactive oxygen species (ROS) in bone tissue and shifts macrophages toward a pro-inflammatory M1 phenotype. Excessive ROS activates inflammatory signaling pathways, including MAPK, PI3K, and NF-κB, thereby promoting M1 polarization [4]. In turn, M1 macrophages produce additional ROS and pro-inflammatory cytokines, including TNF-α, IL-1β, and IL-6 [5], [6]. This cycle sustains osteoclastogenic stimuli and increases pathological bone resorption. Current antiresorptive drugs, including bisphosphonates and denosumab, effectively inhibit osteoclast-mediated bone loss. However, prolonged or excessive suppression of osteoclast activity may impair physiological bone remodeling and contribute to complications such as atypical femoral fractures and osteonecrosis of the jaw [7], [8]. Therefore, an intervention that simultaneously scavenges accumulated ROS and modulates pro-inflammatory M1 macrophages in bone tissue may provide an alternative strategy for PO treatment. By attenuating the upstream stimuli that drive osteoclast overactivation, this approach may reduce pathological bone resorption and improve the balance of bone remodeling.

Interferon regulatory factor 5 (IRF5) is a promising target for reprogramming pro-inflammatory M1 macrophages. IRF5 interacts with NF-κB signaling [9] and has been identified as a key determinant of M1 polarization [10]. It promotes the expression of pro-inflammatory cytokines [11] while suppressing anti-inflammatory pathways [12]. IRF5 silencing may therefore attenuate pro-inflammatory activation and facilitate macrophage phenotypic modulation [13]. These findings support the use of small interfering RNA targeting IRF5 (siIRF5) to modulate the osteoporotic bone microenvironment.

However, selective delivery of siIRF5 to M1 macrophages in bone tissue is challenging. Conventional nucleic acid delivery systems, including unmodified liposomes, lack bone-targeting capability and may distribute broadly after systemic administration. Bone accumulation can be improved by conjugating carriers with ligands such as alendronate or oligopeptides [14], [15], [16]. These ligands bind to hydroxyapatite, the major inorganic component of bone, and therefore provide tissue-level targeting [17], [18]. Hydroxyapatite affinity alone cannot selectively direct siRNA toward M1 macrophages within the bone microenvironment. Non-selective uptake by other cell populations may reduce the effective intracellular dose delivered to the intended cells and increase the risk of off-target effects. Therefore, a delivery system that combines bone enrichment with preferential macrophage uptake may improve the efficiency of siIRF5 delivery within the osteoporotic bone microenvironment.

To address this challenge, we design siP@WM, a nanomedicine that integrates a platinum nanozyme core with siIRF5 delivery and enables programmed sequential targeting of bone tissue and M1 macrophages. The mesoporous silica nanoparticle (MSN) shell stabilizes the Pt core, whereas its enlarged pores (10–15 nm) and polyethyleneimine (PEI) functionalization enable siIRF5 loading and support endosomal escape. A longer poly(aspartic acid) (PASP)-terminated poly(ethylene glycol) (PEG) chain (PEG₂₀₀₀–TK–PEG₁₀₀₀–PASP) promotes bone enrichment. In response to accumulated ROS in bone tissue, thioketal (TK) linker cleavage shortens the outer PEG chain, reduces bone-matrix adhesion, and exposes the shorter FA-functionalized PEG chain (PEG₂₀₀₀–FA), thereby enhancing M1 macrophage uptake. In vitro, siP@WM effectively scavenges ROS and robustly knocks down IRF5 (∼60%), leading to marked M2 polarization. In ovariectomized (OVX) mice, siP@WM significantly ameliorates bone loss and restores trabecular microarchitecture and mechanical strength. Transcriptomic and histological analyses support macrophage-mediated suppression of osteoclastogenesis associated with attenuation of AP-1 signaling (Scheme 1a). This study presents an adaptive biomaterial strategy for restoring bone homeostasis in PO and may inform the development of targeted treatments for other metabolic bone disorders.

Scheme 1.

Design, synthesis, and proposed therapeutic mechanism of the ROS-responsive nanomedicine (siP@WM) for programmed sequential bone–macrophage targeting. (a) In the ovariectomized (OVX) model, estrogen deficiency promotes pro-inflammatory M1 macrophage polarization, oxidative stress, and osteoclast-mediated bone resorption. siP@WM undergoes first-stage bone enrichment mediated by the poly(aspartic acid) (PASP)-terminated longer poly(ethylene glycol) (PEG) chain. ROS-responsive cleavage of the thioketal (TK) linker is designed to shorten the outer PEG chain, reduce bone-matrix adhesion, and expose of the shorter folic acid (FA)-functionalized PEG chain for preferential macrophage uptake. The Pt nanozyme core scavenges ROS, while delivery of small interfering RNA targeting interferon regulatory factor 5 (siIRF5) reprograms macrophages toward the reparative M2 phenotype. The combined effects of oxidative stress alleviation and M2 polarization are associated with reduced osteoclastogenesis and improved bone remodeling; (b) Synthetic route of siP@WM. PtNPs were first encapsulated within a mesoporous silica shell to form Pt@M, followed by the growth of a dendritic mesoporous silica layer to obtain Pt@WM. Surface functionalization with 3-glycidyloxypropyltrimethoxysilane (GPTMS) and polyethyleneimine (PEI) yielded Pt@WM-PEI, enabling siRNA loading. A ROS-responsive linker (HOOC–PEG₂₀₀₀–TK–MAL) was subsequently introduced via EDC/NHS coupling to produce Pt@WM-MAL. NHS–PEG₂₀₀₀–FA and rhodamine B isothiocyanate (RITC) were grafted to afford Pt@WM-FA/R for preferential macrophage uptake and fluorescence tracking. Finally, conjugation of PASP₁₆₀₀–PEG₁₀₀₀–SH via maleimide–thiol addition generated the PASP-terminated longer PEG chain (PEG₂₀₀₀–TK–PEG₁₀₀₀–PASP), and loading of siIRF5 produced the final nanotherapeutic system siP@WM.

2. Results

2.1. siP@WM enables nanozyme stabilization, siIRF5 loading, and programmed surface functionalization

To construct an integrated core–shell nanocarrier combining Pt nanozyme activity, siRNA loading, and programmed surface engineering, we designed a Pt-core mesoporous silica architecture with enlarged mesopores (10–15 nm). The Pt core provided ROS-scavenging activity, whereas the mesoporous silica shell was introduced to stabilize the core, create a confined space for siIRF5 loading, and provide a chemically addressable surface for stepwise ligand conjugation. Accordingly, Pt-based nanocarriers with enlarged pores, termed Pt@WM, were synthesized via a modified sol–gel strategy (Scheme 1b) [19], [20], [21]. PtNPs were first generated in situ by NaBH₄ reduction of K₂PtCl₄ (Fig. 1a), followed by the growth of a mesoporous silica shell using tetraethyl orthosilicate (TEOS) and the structure-directing agent cetyltrimethylammonium chloride (CTAC) to form Pt@M (Fig. 1a). Barrett–Joyner–Halenda (BJH) analysis revealed that Pt@M possessed pore sizes of approximately 2–3 nm (Fig. S2), which limit access of siRNA to the pore interior. To address this limitation, the solvent system and basic catalyst were adjusted to modulate CTAC micelle structure and enlarge pore size. Scanning electron microscopy (SEM) images and BJH measurements confirmed that Pt@WM exhibited mesopores of 10–15 nm (Figs. S1 and S2), providing sufficient space for siRNA encapsulation while facilitating subsequent surface functionalization.

Fig. 1.

Physicochemical characterization of the nanocarrier Pt@WM-PASP. (a) TEM images of PtNPs, Pt@M, Pt@WM, and Pt@WM-PASP, respectively; (b) SEM image showing the morphology of Pt@WM-PASP; (c) TGA curves of Pt@WM, Pt@WM-PEI, Pt@WM-MAL, Pt@WM-FA/R, and Pt@WM-PASP used to estimate the mass added during surface modification and normalize nanoparticle doses; (d) amino acid chromatograms of (i) a mixed amino acid standard and (ii) the acid hydrolysate of Pt@WM-PASP. Aspartic acid (Asp) was identified by matching its retention time with that of the standard; (e) Catalase (CAT)-like, peroxidase (POD)-like, and superoxide dismutase (SOD)-like activities of PtNPs, Pt@WM, Pt@WM-PEI, Pt@WM-FA/R, Pt@WM-PASP, and WM-PASP, respectively; (f) TEM-based energy dispersive spectroscopy (EDS) elemental mapping images of Pt@WM-PASP (i) and WM-PASP (ii), showing the distributions of silicon (Si) and platinum (Pt). BF: bright field; (g) Size distribution curves of (i) PtNPs, Pt@M, Pt@WM, and Pt@WM-PEI, and (ii) Pt@WM-MAL, Pt@WM-FA, Pt@WM-FA/R, and Pt@WM-PASP in DI water; (h) Agarose gel electrophoresis analysis of Pt@WM-PASP/siIRF5 complexes (i) and the corresponding quantification (ii). Nucleic acid binding (iii) and protection (iv) performance of Pt@WM-PASP/siIRF5 complexes. NPs: nanoparticles; HT: high temperature. Data are presented as mean ± SD (n = 3 independent experiments). Statistical significance was determined by one-way ANOVA followed by Tukey's multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, not significant.

To enable siRNA loading and programmed sequential targeting of bone tissue and macrophages, Pt@WM nanoparticles were modified in a stepwise manner. First, polyethyleneimine (PEI) was covalently grafted onto Pt@WM through 3-glycidyloxypropyltrimethoxysilane (GPTMS)-mediated coupling, which introduced abundant amine groups for siRNA loading and subsequent ligand conjugation. The PEI-modified nanoparticles were then reacted with HOOC–PEG₂₀₀₀–thioketal (TK)–maleimide (MAL) through EDC/NHS-mediated amide coupling to install a ROS-cleavable TK-containing maleimide linker, yielding Pt@WM-MAL. Next, NHS–PEG₂₀₀₀–folic acid (FA) was conjugated to the remaining surface amines to introduce the shorter chain for subsequent macrophage recognition. In selected formulations, rhodamine B isothiocyanate (RITC) was introduced for fluorescence tracking. Finally, poly(aspartic acid) (PASP)₁₆₀₀–PEG₁₀₀₀–SH was grafted via maleimide–thiol addition, generating Pt@WM-PASP. In this configuration, the longer PEG₂₀₀₀–TK–PEG₁₀₀₀–PASP chain mediates bone-matrix affinity. ROS-triggered TK cleavage is designed to remove the terminal PEG₁₀₀₀–PASP segment, shorten the outer chain, reduce bone-matrix adhesion, and expose the shorter PEG₂₀₀₀–FA chain for M1 macrophage uptake.

Surface functionalization was characterized by spectroscopy, chemical assays, thermogravimetric analysis (TGA), and ζ-potential measurements. Fourier transform infrared (FT-IR) spectra and derivative thermogravimetric (DTG) curves showed no readily detectable characteristic alkyl-chain signals or major decomposition peak of CTAC in Pt@WM (Figs. S3 and S4). Together with the negative ζ-potential of Pt@WM (Fig. S5b, Table S1), these results suggested minimal residual CTAC after extraction. Following PEI conjugation, additional C–H stretching and bending bands appeared at 2946, 2867, and 1461 cm⁻¹ (Fig. S3), the ζ-potential increased from −20.04 ± 0.69 for Pt@WM to +40.37 ± 1.30 mV for Pt@WM-PEI (Fig. S5b, Table S1), and the ninhydrin assay showed increased amine groups (Fig. S6a). Following EDC/NHS-mediated coupling, Pt@WM-MAL exhibited an additional amide II band at 1549 cm⁻¹ (Fig. S3), together with decreased amine groups and increased maleimide reactivity, supporting conjugation of the PEG₂₀₀₀–TK–MAL linker (Fig. S6). After NHS–PEG₂₀₀₀–FA conjugation, the ζ-potential decreased to +25.85 ± 0.06 mV for Pt@WM-FA and +22.11 ± 0.25 mV for Pt@WM-FA/R (Fig. S5b, Table S1). Ultraviolet–visible (UV–Vis) spectra of Pt@WM-FA/R showed absorption features attributable to FA relative to Pt@WM-MAL (Fig. S7). Subsequent PASP₁₆₀₀–PEG₁₀₀₀–SH conjugation reduced maleimide reactivity (Fig. S6b) and further decreased the ζ-potential to +13.09 ± 0.40 mV (Fig. S5b, Table S1). High performance liquid chromatography (HPLC)-based amino acid analysis of the Pt@WM-PASP hydrolysate revealed an aspartic acid (Asp) peak with the same retention time as the standard (Fig. 1d). Together, these results supported conjugation of PEI, the TK–MAL linker, FA, and PASP. TGA was used to estimate the mass added during surface modification and normalize the formulations to an equivalent mass of the Pt@silica core–shell (Fig. 1c). Based on this analysis, 100 μg of Pt@WM corresponded to 125.96 μg of Pt@WM-PEI, 127.62 μg of Pt@WM-MAL, 145.98 μg of Pt@WM-FA/R, and 148.81 μg of Pt@WM-PASP (Table S2). Dynamic light scattering (DLS) showed that formation of the enlarged mesoporous shell reduced the PDI from 0.448 for Pt@M to 0.156 for Pt@WM (Fig. S5a, Table S1). The Z-average diameters of the surface-functionalized intermediates ranged from 259 to 309 nm (Fig. 1g,Table S1), reflecting changes in surface hydration and colloidal interactions during modification. The final Pt@WM-PASP had a Z-average diameter of 295.57 ± 7.95 nm and a PDI of 0.059, indicating a narrow size distribution (Fig. 1g,Table S1).

To distinguish the contribution of the Pt core from that of the silica-based carrier, WM-PASP nanoparticles without PtNPs were prepared as a negative control. Compared with WM-PASP, Pt-containing formulations exhibited pronounced catalase (CAT)-, peroxidase (POD)-, and superoxide dismutase (SOD)-like activities, confirming that the nanozyme activity was attributable primarily to the Pt core and was preserved after surface modification (Fig. 1e). We next examined whether this multistep functionalization compromised the structural properties of the carrier. TEM-EDS mapping confirmed the core localization of Pt within Pt@WM-PASP (Fig. 1f). BJH analysis and SEM images showed that the enlarged mesopores remained accessible after surface modification (Fig. 1b,S2), indicating that ligand conjugation did not completely block the siRNA-loading space. TEM images combined with inductively coupled plasma optical emission spectrometry (ICP-OES) quantification of dissolved Si confirmed the time-dependent hydrolytic degradation of the mesoporous silica frameworks of both Pt@WM and Pt@WM-PASP in PBS (Fig. S8). Collectively, these results demonstrated that Pt@WM-PASP integrated preserved Pt nanozyme activity, accessible enlarged mesopores, and degradability, thereby providing a suitable carrier for subsequent siIRF5 loading.

To determine the optimal loading ratio for siIRF5, nucleic acid adsorption and gel electrophoresis assays were performed (Fig. 1h,S9). Although the ζ-potential decreased to +13.09 ± 0.40 mV after PASP conjugation, siRNA loading remained efficient. Free siIRF5 showed a clear migration band in the agarose gel, whereas siIRF5 migration was progressively retarded as the Pt@WM-PASP content increased, indicating concentration-dependent complexation between siIRF5 and the cationized nanocarrier. Quantification showed that the bound nucleic acid fraction increased with increasing nanoparticle mass and reached a near-plateau at siIRF5: Pt@WM-PASP mass ratios of 1:25–1:50 (Fig. 1h(i–iii)). Considering both loading efficiency and nanocarrier dosage, a mass ratio of 1:50 was selected for subsequent experiments. In the protection assay, free siIRF5 showed a markedly reduced nucleic acid signal after high-temperature treatment and was almost completely degraded after RNase A digestion. In contrast, Pt@WM-PASP-loaded siIRF5 retained a nearly unchanged nucleic acid signal after high-temperature treatment and preserved a high residual fraction after RNase A exposure (Fig. 1h(iv)). These results demonstrated that Pt@WM-PASP not only enabled efficient siIRF5 loading but also substantially protected siIRF5 from thermal and enzymatic degradation, supporting its use as a protective carrier for siRNA delivery. The final siIRF5-loaded Pt@WM-PASP formulation was termed siP@WM throughout this study.

2.2. Sequential Bone–Macrophage targeting strategy for bone enrichment, macrophage uptake, and siRNA delivery

A programmed sequential targeting strategy was designed to deliver siIRF5 and the Pt nanozyme to macrophages within bone tissue. The longer PEG₂₀₀₀–TK–PEG₁₀₀₀–PASP₁₆₀₀ chain mediates bone-matrix affinity and promotes first-stage bone enrichment after systemic administration. In response to accumulated ROS in the osteoporotic bone microenvironment, TK cleavage removes the terminal PEG₁₀₀₀–PASP segment, shortens the outer chain, reduces bone-matrix adhesion, and exposes the shorter PEG₂₀₀₀–FA chain for macrophage recognition. Bone affinity experiments using bovine bone slices showed that TK cleavage significantly reduced nanoparticle adhesion due to loss of the bone-targeting component PASP (Fig. 2a). Pretreatment of bone slices with sodium poly(aspartate) (PASP-Na) reduced Pt@WM-PASP retention at both concentrations tested, as shown by gross imaging and confocal laser scanning microscopy (CLSM) (Fig. S10). These results supported the contribution of PASP to nanoparticle binding to the bone matrix. A potential concern was that Pt-mediated ROS scavenging might inhibit TK cleavage. However, H₂O₂ could induce cleavage of the TK-containing chain under accelerated oxidative conditions (Fig. 2b). After adsorption to bone slices, Pt@WM-PASP showed markedly reduced surface retention following exposure to 100 μM H₂O₂ compared with PBS, whereas 1 μM H₂O₂ did not produce a significant reduction (Fig. S11). Together with the TK cleavage assay, these results suggested H₂O₂-induced nanoparticle release through cleavage of the TK linker. H₂O₂ pretreatment did not significantly affect siRNA protection or gene-silencing activity of the siP@WM. Gel electrophoresis showed no significant differences in the amount of intact siRNA remaining after pretreatment with 0, 1, or 100 μM H₂O₂, either with or without subsequent RNase A exposure (Fig. S12a and b). The three formulations also produced similar IRF5 knockdown in RAW264.7 cells, with no significant differences among groups (Fig. S12c). These findings indicated that H₂O₂-induced changes in the surface chains did not measurably impair siRNA protection or subsequent IRF5 silencing under the tested conditions.

Fig. 2.

Targeting and delivery capabilities of the nanocarrier Pt@WM-PASP. (a) Bone affinity of Pt@WM-FA/R, Pt@WM-PASP, and H₂O₂-treated Pt@WM-PASP evaluated by (i) digital imaging, confocal laser scanning microscopy (CLSM), and SEM, and (ii) quantification of fluorescence intensity. MFI: mean fluorescence intensity; (b) ROS-responsive thioketal (TK) cleavage in Pt@WM-PASP evaluated by the ninhydrin reaction; (c) Uptake of Pt@WM-MAL, Pt@WM-PASP, and H₂O₂-treated Pt@WM-PASP by M1-polarized RAW264.7 macrophages, assessed by CLSM. NP: nanoparticle; (d) Uptake of Pt@WM-MAL, Pt@WM-PASP, and H₂O₂-treated Pt@WM-PASP by M1-polarized RAW264.7 macrophages, assessed by (i) flow cytometry and (ii) corresponding quantification. *P < 0.05 versus Pt@WM-MAL and #P < 0.05 versus Pt@WM-PASP; (e) Intracellular ROS-scavenging ability of Pt@WM-PASP and WM-PASP evaluated using the 2′,7′-dichlorofluorescein (DCF) probe, shown as (i) representative flow cytometry histograms and (ii) corresponding quantification; (f) Endosomal escape ability of siP@WM evaluated by (i) CLSM and (ii) corresponding quantification using Manders' coefficient. BF: bright field. Data are presented as mean ± SD (n = 3 independent experiments). Statistical significance was determined by two-tailed unpaired Student's t-test or one-way ANOVA followed by Tukey's multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, not significant.

Subsequent CLSM and flow cytometry analyses demonstrated enhanced macrophage uptake of the nanoparticles after TK cleavage (Fig. 2c and d). In co-cultures with MC3T3-E1 or bEnd.3 cells, H₂O₂-treated Pt@WM-PASP showed greater preferential uptake by RAW264.7 macrophages than Pt@WM-MAL or intact Pt@WM-PASP (Figs. S13 and S14). These results supported enhanced macrophage uptake relative to osteoblast-lineage and endothelial cells following ROS-responsive exposure of FA. Pretreatment with excess free FA or an antibody against folate receptor β (FRβ) markedly reduced the uptake of H₂O₂-treated Pt@WM-PASP, produced a smaller reduction in the uptake of intact Pt@WM-PASP, and had little effect on Pt@WM-MAL uptake (Fig. S15). The response of intact Pt@WM-PASP was consistent with partial FA exposure during incubation. N-acetylcysteine (NAC) pretreatment markedly reduced the uptake of intact Pt@WM-PASP but had little effect on Pt@WM-MAL or H₂O₂-treated Pt@WM-PASP (Fig. S15), suggesting that endogenous ROS contributed to TK cleavage and FA exposure. Together, these findings supported ROS-responsive FA exposure and subsequent FA-mediated macrophage uptake, although changes in surface physicochemical properties may also have contributed.

Following internalization, Pt@WM-PASP effectively reduced intracellular ROS levels in M1-polarized RAW264.7 cells (Fig. 2e). CLSM showed progressively reduced colocalization of fluorescently labeled siIRF5 with LAMP2⁺ compartments at 8, 24, and 48 h, consistent with endosomal escape (Fig. 2f). These results supported the feasibility of the ROS-responsive programmed sequential delivery strategy.

2.3. siP@WM reverses M1 macrophage polarization and promotes bone enrichment in vivo

The regulatory effect of siP@WM on macrophage polarization was evaluated in RAW264.7 cells. First, IRF5 expression was confirmed to be elevated in M1 macrophages and suppressed in M2 macrophages (Fig. S16), supporting its role as a polarization-associated target. Among the concentrations tested, 37.5 μg/mL of siP@WM achieved effective IRF5 knockdown in RAW264.7 cells while limiting nanocarrier exposure and was selected for subsequent experiments (Fig. S17). Nanoparticle-mediated delivery of siRNA (siIRF5 (NP) and siCON (NP)) was compared with a commercial transfection reagent (siIRF5 and siCON) at the same siRNA concentration. RT-qPCR, Western blotting, immunofluorescence staining, and flow cytometry showed that siIRF5 (NP) produced IRF5 knockdown and changes in macrophage polarization similar to those produced by commercial reagent (Fig. 3a–d, S18). Compared with siCON, siCON (NP) significantly reduced iNOS expression at both the mRNA and protein levels and reduced IL-1β mRNA expression. Together with the comparison between Pt@WM-PASP and Pt-free WM-PASP in the intracellular ROS scavenging assay (Fig. 2e), these findings suggested a partial inhibitory effect of Pt-mediated ROS scavenging on M1-associated inflammatory activation. Compared with siCON (NP), siIRF5 (NP) further suppressed IRF5 and M1-associated markers and increased M2-associated markers, demonstrating an additional contribution of IRF5 silencing. These results suggested complementary effects of the Pt nanozyme and siIRF5 on macrophage polarization.

Fig. 3.

siP@WM reverses the M1 polarization of RAW264.7 cells and improves bone enrichment in vivo. (a) Relative mRNA expression levels of M1-associated markers (IRF5, TNF-α, iNOS, IL-1β, and IL-6) and M2-associated markers (Arg1, Egr2, IL-10, and TGF-β) in RAW264.7 cells after different treatments, as determined by RT-qPCR; (b) Immunofluorescence staining of IRF5, iNOS, and Arg1 in RAW264.7 cells after different treatments. NP: nanoparticle; (c) Western blot analysis of iNOS, IRF5, and Arg1 protein expression in RAW264.7 cells after different treatments, shown as (i) representative images and (ii) corresponding quantification. GAPDH was used as the internal control; (d) Flow cytometric analysis of macrophage polarization after different treatments, shown as (i) representative histograms of CD86 and CD206 expression and (ii) corresponding quantification of CD86⁺ and CD206⁺ cell populations; (e) Biodistribution of Pt@WM-FA/R and Pt@WM-PASP after intravenous injection into ICR mice, shown as (i) representative fluorescence images of major organs and bones collected at 1 h, 1 d, 7 d, and 14 d, and (ii) quantification of normalized bone radiant efficiency at the corresponding time points. Data are presented as mean ± SD (n = 3 independent experiments for a–d; n = 3 mice per group at each time point for e). Statistical significance was determined by two-tailed unpaired Student's t-test or one-way ANOVA followed by Tukey's multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, not significant.

Prior to PO treatment, we assessed the biocompatibility of Pt@WM, Pt@WM-PEI, Pt@WM-PASP, and H₂O₂-treated Pt@WM-PASP (Figs. S19 and S20). Pt@WM-PASP and H₂O₂-treated Pt@WM-PASP demonstrated low cytotoxicity over the range of 0–100 μg/mL and negligible hemolysis up to 200 μg/mL. A single dose of 1000 μg per mouse was used for the biodistribution and high-exposure safety study to provide sufficient fluorescence for ex vivo tracking over the 14-day observation period. No mortality or abnormal behavior was observed. Biodistribution analysis showed predominant nanoparticle-associated fluorescence in the liver and lungs initially (Fig. 3e), followed by a decrease over the observation period. At most time points, fluorescence signals in the heart, liver, spleen, lungs, and kidneys did not differ significantly between the two formulations (Fig. S21). Pt@WM-PASP showed higher bone fluorescence than the non-bone-targeting control (Fig. 3e). These results indicated favorable biocompatibility under the tested conditions and enhanced bone enrichment and retention of Pt@WM-PASP in vivo.

2.4. siP@WM ameliorates bone loss and normalizes bone turnover in OVX mice

The therapeutic performance of siP@WM was assessed in an OVX mouse model of osteoporosis. Briefly, 6–8-week-old female ICR mice were randomly assigned to undergo either sham surgery or bilateral ovariectomy. Eight weeks post-surgery, OVX mice were randomly divided into five groups (n = 6 per group) and received weekly tail-vein injections for 4 weeks (Fig. 4a). The treatments were as follows: Saline, 200 μg WM-PASP, 300 pmol siIRF5, 200 μg Pt@WM-PASP, and 200 μg siP@WM. One week after the final injection, all mice were euthanized. Daily monitoring during treatment revealed no mortality or abnormal behaviors, such as lethargy, hunched posture, or impaired locomotion. Histological examination (H&E staining) of major organs (heart, liver, spleen, lung, kidney) showed no lesions or degeneration in any group (Fig. S22a). Because prior biodistribution results indicated strong nanoparticle-associated fluorescence in the liver and lungs, silica-associated tissue burden was assessed by ICP-OES analysis of homogenized liver and lung tissues from the Saline and siP@WM groups. The results were comparable to those reported in similar studies (Fig. S22b) [21]. Serum biochemical analysis performed at the experimental endpoint showed no significant differences between any treatment group and the Sham group in ALT, AST, ALP, TP, ALB, TBIL, BUN, UA, or CREA (Fig. S23). These findings indicated no overt short-term toxicity under the tested dosing conditions.

Fig. 4.

siP@WM ameliorates bone loss in ovariectomy-induced osteoporotic mice. (a) Schematic illustration of the animal experimental design. After a 7-day acclimatization period, mice underwent ovariectomy. Eight weeks after surgery, OVX mice received treatment once weekly for 4 weeks, followed by sacrifice and sample collection; (b) Micro-CT analysis of femora after 4 weeks of treatment, including (i) representative horizontal reconstruction images, (ii) representative coronal reconstruction images, and (iii) quantitative analysis of trabecular bone parameters, including bone mineral density (BMD), bone volume fraction (BV/TV), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), and trabecular number (Tb.N); (c) Histological evaluation of femora after treatment, including (i) Von Kossa staining and (ii) Masson trichrome staining of femoral sections; (d) Mechanical properties of femora evaluated by three-point bending tests, including (i) representative load–deflection curves and (ii) mechanical parameters including maximum loading, Young's modulus, and fracture energy. Data are presented as mean ± SD (n = 3 mice per group). Statistical significance was determined by one-way ANOVA followed by Tukey's multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, not significant.

Systemic assessments were performed to evaluate the therapeutic performance of siP@WM. Micro-CT scanning and 3D reconstruction of femora demonstrated that siP@WM substantially ameliorated OVX-induced bone loss (Fig. 4b). Quantitative parameters, including BMD, BV/TV, Tb.Th, Tb.Sp, and Tb.N, showed that siP@WM-treated mice reached levels comparable to those of the Sham group (Fig. 4b). Representative images of Von Kossa, H&E, and Masson trichrome staining further indicated the improved trabecular number and density in the siP@WM group (Fig. 4c,S24). Mechanical properties assessed by three-point bending tests showed that maximum loading, fracture energy, and Young's modulus of the siP@WM group did not differ significantly from the Sham group and were significantly better than those in the Saline and other treatment groups (Fig. 4d). At the therapeutic dose of 200 μg per injection, nanoparticle-associated fluorescence was detected along trabecular surfaces and in adjacent marrow spaces in the distal femoral metaphysis of OVX mice treated with WM-PASP, Pt@WM-PASP, or siP@WM (Fig. 5a). Additional immunofluorescence staining showed colocalization of nanoparticle-associated fluorescence with F4/80⁺ macrophages (Fig. S25). Tartrate-resistant acid phosphatase (TRAP) staining and quantitative analysis further showed that osteoclast density in the siP@WM group was comparable to that in the Sham group and significantly lower than in the Saline and other treatment groups (Fig. 5b). ELISA measurement of serum bone turnover markers indicated no significant differences between the siP@WM and Sham groups (Fig. 5c). Double fluorochrome labeling with calcein and tetracycline demonstrated that bone formation rate (BFR) in the siP@WM group was similar to that in the Sham group and significantly higher than in the Saline and other treatment groups (Fig. 5d). Collectively, these findings indicated that siP@WM ameliorated OVX-induced bone loss, restored femoral mechanical properties, and normalized bone turnover.

Fig. 5.

siP@WM normalizes the high bone turnover state in ovariectomy-induced osteoporotic mice. (a) Confocal fluorescence images showing nanoparticle-associated fluorescence (red dots) in undecalcified bone sections after 4 weeks of treatment; (b) Osteoclast activity in femora after treatment, shown as (i) representative tartrate-resistant acid phosphatase (TRAP) staining images of decalcified femur sections and (ii) quantification of TRAP ⁺ cell number per mm²; (c) Serum levels of bone turnover markers, including C-terminal telopeptide of type I collagen (CTX-Ⅰ) and procollagen type I N-terminal propeptide (P1NP), measured by ELISA; (d) Dynamic histomorphometric analysis of bone formation by sequential fluorescent labeling, shown as (i) quantification of mineral apposition and (ii) representative fluorescence images of femur sections. Green: calcein; yellow: tetracycline. Data are presented as mean ± SD (n = 3 mice per group). Statistical significance was determined by one-way ANOVA followed by Tukey's multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, not significant.

2.5. siP@WM reprograms the osteoporotic bone microenvironment by attenuating M1 macrophage polarization, oxidative stress, and osteoclastogenic signaling

To elucidate the underlying mechanisms of siP@WM in treating PO, transcriptomic analysis was conducted on femora from the Saline and siP@WM groups. Analysis of differentially expressed genes (DEGs) (Fig. 6a, S26), together with Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Reactome pathway enrichment analyses (Fig. 6b,S27), identified treatment-associated changes in genes related to osteoclast differentiation, extracellular matrix (ECM) remodeling, macrophage polarization, and oxidative stress-induced inflammatory responses. To validate the transcriptomic findings, immunofluorescence staining was conducted on femoral sections. Within the F4/80⁺ macrophage population, the percentages of IRF5⁺ and CD86⁺ cells were significantly lower and the percentage of CD206⁺ cells was significantly higher in the siP@WM group than in the Saline and other treatment groups, with no significant differences from the Sham group (Fig. 6d). Consistently, siP@WM treatment reduced the number of M1 macrophages (F4/80⁺IRF5⁺CD86⁺CD206^-) and increased the number of M2 macrophages (F4/80⁺CD206⁺IRF5⁻CD86⁻) (Fig. 6d). The results indicated successful promotion of M2 macrophage polarization by siP@WM. Fluorescence intensities of IL-1β and Ptgs2 were significantly lower in the siP@WM group than in the Saline and other treatment groups, with no difference from the Sham group (Fig. S28), supporting reduced inflammation in the bone microenvironment. Immunofluorescence staining showed significantly higher levels of 8-hydroxy-2′-deoxyguanosine (8-OHdG) and 4-hydroxynonenal (4-HNE) in the Saline group than in the Sham group, indicating increased oxidative DNA damage and lipid peroxidation. Both signals were reduced following siP@WM treatment (Fig. S29). Similarly, fluorescence intensities of HMOX-1 and c-Jun were significantly lower in the siP@WM group than in the Saline and other treatment groups, with levels comparable to the Sham group (Fig. S28), suggesting effective alleviation of oxidative stress and consequent reduced activation of endogenous antioxidant system components. Fluorescence intensities of RANKL and c-Fos were significantly reduced in the siP@WM group relative to the Saline and other treatment groups, whereas those of OPG and OCN were significantly elevated (Fig. S28). Compared with the Saline group, the siP@WM group showed significantly lower MMP13 and significantly higher P4HA3 fluorescence intensities, with neither marker differing significantly from the Sham group (Fig. S29). These changes were consistent with reduced collagen degradation and altered collagen post-translational processing following siP@WM treatment. These tissue-level findings suggested that siP@WM might suppress osteoclast differentiation and modulate ECM remodeling by regulating macrophage polarization and attenuating oxidative stress-induced inflammatory responses, thereby contributing to bone regeneration (Fig. 6c). However, because the transcriptomic and immunofluorescence analyses were performed at the tissue level, they did not establish whether macrophage reprogramming functionally altered AP-1-associated signaling and osteoclast differentiation in osteoclast precursors.

Fig. 6.

Transcriptomic analysis reveals the potential mechanism of siP@WM in the treatment of osteoporotic mice. (a) Heatmap showing differentially expressed genes (DEGs) in the femora of siP@WM-treated mice compared with saline-treated mice. Genes related to osteoclast differentiation, inflammatory response, oxidative stress, macrophage polarization, osteoblast differentiation, and extracellular matrix (ECM) organization are highlighted; (b) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of DEGs in the siP@WM versus Saline groups; (c) Schematic illustration of the proposed mechanism by which siP@WM alleviates osteoporosis; (d) Characterization of macrophage polarization in femoral sections after 4 weeks of treatment, shown as (i) representative immunofluorescence staining images of F4/80, IRF5, CD86, and CD206 and (ii) percentages of IRF5⁺, CD86⁺, and CD206⁺ cells among F4/80⁺ macrophages and densities of M1 macrophages (F4/80⁺IRF5⁺CD86⁺CD206^-) and M2 macrophages (F4/80⁺CD206⁺IRF5⁻CD86⁻), expressed as cells per mm² (n = 3 mice per group). Data are presented as mean ± SD. Statistical significance was determined by one-way ANOVA followed by Tukey's multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, not significant.

2.6. Macrophage IRF5 silencing attenuates AP-1-associated signaling and osteoclastogenesis through paracrine regulation

Conditioned media (CM) from differently treated RAW264.7 macrophages were used to assess the effects of macrophage IRF5 silencing on RANKL-induced osteoclast differentiation in bone marrow-derived macrophages (BMMs). RANKL significantly increased the p-JNK/JNK and p-c-Jun/c-Jun ratios relative to the Control group. Neither ratio differed significantly between the M1 CM and RANKL groups, although a numerical increase was observed following M1 CM treatment. In contrast, siIRF5 (NP) CM and M2 CM significantly reduced both ratios relative to RANKL alone (Fig. 7a). siIRF5 (NP) CM and M2 CM treatments also reduced nuclear c-Fos accumulation (Fig. 7b) and significantly reduced the RANKL-induced expression of Nfatc1, Acp5, Mmp9, Oscar, Dcstamp, and Ocstamp (Fig. 7c). These changes were accompanied by fewer TRAP⁺ multinucleated cells, a lower percentage of the culture area occupied by these cells, and markedly impaired F-actin ring formation (Fig. 7d and e). The AP-1 inhibitor T-5224 also reduced osteoclastogenic gene expression, TRAP⁺ multinucleated cell formation, and F-actin ring formation in the presence of M1 CM. Together, these results supported a paracrine mechanism by which IRF5 silencing in macrophages attenuated AP-1-associated signaling and osteoclast differentiation.

Fig. 7.

Macrophage IRF5 silencing attenuates AP-1-associated signaling and osteoclast differentiation through paracrine regulation. Bone marrow-derived macrophages (BMMs) were induced with M-CSF (20 ng/mL) and, where indicated, RANKL (50 ng/mL) in the presence or absence of conditioned media (CM) derived from differently treated RAW264.7 macrophages. CM was mixed with fresh complete medium at a volume ratio of 1:3. (a) Western blot analysis of phosphorylated and total JNK and c-Jun at 30 min after RANKL stimulation: (i) representative protein bands and (ii) corresponding quantification. GAPDH was used as the loading control. (b) Representative immunofluorescence images showing c-Fos localization at 24 h. Red: c-Fos; green: phalloidin; blue: DAPI. (c) Relative mRNA expression levels of the osteoclastogenic markers at 48 h, as determined by RT-qPCR. (d) Representative fluorescence images of F-actin ring formation on day 6. Green: phalloidin; blue: DAPI. (e) Tartrate-resistant acid phosphatase (TRAP) staining on day 6: (i) representative images and (ii) quantification of the number and occupied area of TRAP⁺ multinucleated cells (≥3 nuclei). Data are presented as mean ± SD (n = 3 independent experiments). Statistical significance was determined by one-way ANOVA followed by Tukey's multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, not significant.

3. Discussion

Our results showed that siP@WM achieved bone enrichment, enhanced macrophage uptake in response to ROS, and ameliorated bone loss in OVX mice. Structurally, the MSN shell with enlarged pores stabilized the Pt nanozyme core, enabled efficient siIRF5 encapsulation, and provided a modular interface for sequential surface functionalization. Functionally, the programmed bone–macrophage targeting strategy mediated first-stage bone enrichment, subsequent preferential uptake by pro-inflammatory M1 macrophages in bone tissue, intracellular ROS scavenging, and effective IRF5 gene silencing, promoting the shift from pro-inflammatory M1 macrophages toward a reparative M2 phenotype. Through combined ROS scavenging and siIRF5 delivery to macrophages in bone tissue, siP@WM markedly ameliorated OVX-induced bone loss. Mechanistically, transcriptomic analyses, histological staining, and CM transfer experiments further suggested that siP@WM disrupted the cycle between oxidative stress and pro-inflammatory macrophage activation, thereby attenuating AP-1-associated inflammatory signaling, suppressing osteoclast differentiation, and modulating ECM remodeling.

3.1. Mesoporous silica shells enable stable nanozyme integration and protected siIRF5 delivery

The design of the MSN shell was intended to address the intrinsic instability of bare PtNPs. Although PtNPs possess strong ROS-scavenging capability, unprotected PtNPs tend to aggregate and precipitate in body fluids, posing a major challenge for delivering sufficient quantities of PtNPs to target tissues and cells [22]. PtNPs typically rely on stabilizing agents such as citrate, polyvinylpyrrolidone (PVP), polyvinyl alcohol (PVA), and cetyltrimethylammonium bromide (CTAB) to maintain optimal dispersion [23]. These stabilizers may lose stability upon dilution or introduce potential toxicity. In contrast, the MSN shell provides structural stability, hydrolytic degradability, and tolerable biological toxicity [24].

The MSN shell also offers a versatile platform for surface modification and siRNA loading. Pore size plays a critical role in determining siRNA loading behavior. Given that siRNA molecules are approximately 7–8 nm in length and 2–3 nm in diameter [25], the small pores (∼2 nm) of Pt@M are expected to restrict access to the pore interior and favor adsorption on the external surface, making it vulnerable to degradation and detachment [26]. By enlarging the pore size to 10–15 nm in Pt@WM, siRNA could be encapsulated within the mesopores, thereby enhancing both loading efficiency and protection [27], [28]. This structural optimization established Pt@WM as a suitable nanocarrier for nucleic acid delivery.

Surface modification is also essential for efficient siRNA loading and in vivo delivery. Unmodified Pt@WM carried a high negative charge, which is unfavorable for adsorption of the similarly negatively charged siIRF5. A common strategy involves surface modification with aminosilanes such as 3-aminopropyltriethoxysilane (APTES) via covalent attachment [21] or with cationic polymers such as PEI via electrostatic adsorption [29] to impart a positive surface charge and enable subsequent grafting. Following cellular uptake, siRNA must escape from endosomes into the cytosol to exert its function—a process known as endosomal escape. This requirement constitutes a major bottleneck limiting siRNA therapeutic efficiency [30], [31]. PEI-based strategies outperform aminosilane-based approaches by promoting endosomal escape through the proton sponge effect, umbrella effect, and disruptive interactions with the endosomal membrane [32]. However, electrostatically adsorbed PEI may dissociate in vivo due to competitive interactions with biomolecules [33]. We employed a one-pot method in which epoxy silane was premixed with PEI to form covalent linkages, followed by grafting onto the Pt@WM surface, thereby achieving stable covalent immobilization of PEI. Based on previous reports [34], 10 kDa PEI was selected to balance transfection efficiency and cytotoxicity. This design supported siRNA loading, protection, and intracellular delivery.

Mesoporous inorganic carriers have been investigated for siRNA delivery in osteoporosis. Kim et al. [35] used amine-functionalized mesoporous bioactive glass nanoparticles to deliver siRNA against RANK and suppress osteoclastogenic gene expression in vitro. Mora-Raimundo et al. [15] developed alendronate (ALN)-functionalized, PEI-coated MSNs for co-delivery of osteostatin and siRNA targeting SOST, enhancing osteogenic gene expression and improving bone microarchitecture in OVX mice. These studies support the use of mesoporous carriers to protect siRNA and regulate bone remodeling. Pt@WM combines enlarged mesopores for siRNA loading, a catalytically active Pt core for ROS scavenging, and covalently immobilized PEI. The silica shell therefore supports siRNA delivery, stabilizes the nanozyme core, and provides a surface for sequential functionalization.

3.2. IRF5 gene silencing for macrophage reprogramming

IRF5 acts as a key regulatory node of macrophage polarization and connects multiple downstream inflammatory pathways [10]. In human macrophages, IRF5 is required for pattern recognition receptor (PRR)-mediated activation of NF-κB signaling [36]. It can also cooperate with RelA/p65 to form an IRF5–RelA cistrome, thereby driving transcription of pro-inflammatory genes [9]. However, comparative analyses of IRF5- and RelA-deficient macrophages revealed overlapping but distinct transcriptional programs, indicating that their contributions to inflammatory regulation are not interchangeable [9]. In addition, IRF5 promotes Akt2 activation during PRR signaling, enhancing expression of glycolytic pathway genes and HIF1α, as well as pro-inflammatory cytokine production and M1 polarization [37]. IRF5 deficiency impairs TLR8-induced inflammatory responses while largely preserving TNF-α-induced NF-κB reporter activity, which is strongly suppressed by p65 knockdown [38]. These findings suggest that targeting IRF5 can modulate particular inflammatory responses without uniformly suppressing NF-κB signaling. However, this distinction depends on the stimulus and experimental context and does not guarantee a general advantage in immune selectivity. Collectively, these features support IRF5 silencing as a strategy for macrophage reprogramming. Courties et al. [39] showed that nanoparticle-mediated siIRF5 delivery reduced inflammatory macrophages and improved healing after myocardial infarction. Ma et al. [40] used PEI-conjugated, diselenide-bridged MSNs for ROS-responsive siIRF5 delivery, promoting M1-to-M2 polarization and reducing inflammation after spinal cord injury. These studies demonstrate the feasibility of nanoparticle-mediated IRF5 silencing and its integration with redox-responsive delivery. Combining catalytic ROS scavenging with sequential bone–macrophage delivery of siIRF5, the present study applies this approach to chronic bone loss caused by estrogen deficiency, and links macrophage reprogramming to the suppression of osteoclastogenesis.

3.3. Sequential Bone–Macrophage delivery strategy enhances macrophage uptake in the osteoporotic bone microenvironment

The surface PEG chains were designed to achieve programmed sequential recognition of bone tissue and M1 macrophages within the pathological bone microenvironment. The PASP-terminated longer PEG chain mediated first-stage bone enrichment. The TK linkers are presented in the external layer and are therefore readily accessible to oxidants in the surrounding microenvironment, whereas the Pt nanozyme is located within the internal mesoporous framework. This arrangement is expected to provide an opportunity for initial ROS exposure to induce sufficient TK cleavage to shorten the outer chain, weaken bone anchoring and expose the shorter FA-functionalized chain for subsequent macrophage uptake. Several molecules with strong affinity for hydroxyapatite—the primary inorganic component of bone—such as bisphosphonates, tetracycline, PASP, and aptamers, have been widely employed for bone targeting. Previous studies indicate that PASP preferentially interacts with highly crystalline hydroxyapatite characteristic of bone-resorption surfaces, whereas tetracycline binds more readily to forming bone surfaces [17], [41]. Accordingly, PASP serves as the initial bone-enrichment component in our design. TK is the sulfur analog of a ketal and cleaves under ROS exposure to yield thiol-containing products and acetone [42]. The ROS-responsive TK-PASP composite chain enabled conditional release of nanoparticles from the bone matrix. FA was introduced to target folate receptor beta (FRβ)-expressing macrophages. In normal tissues, resting macrophages involved in homeostasis exhibit low or negligible FRβ expression, whereas M1 macrophages show markedly upregulated FRβ levels. FRβ⁺ macrophages also produce high levels of TNF-α and ROS [43]. Meanwhile, PEG segments improve colloidal stability, reduce reticuloendothelial clearance, and prolong circulation time [44]. Together, these components form a coordinated system that combines bone enrichment with ROS-responsive macrophage uptake.

Previous bone-targeted RNA delivery systems have used affinity for different mineral surfaces to favor delivery to nearby bone-remodeling cells. D-Asp₈-functionalized liposomes bind preferentially to highly crystalline hydroxyapatite at bone-resorption surfaces, facilitating antagomir-148a delivery to osteoclasts [17]. In contrast, (DSS)₆-functionalized liposomes preferentially localize at bone-formation surfaces and enhance delivery of Ckip-1 siRNA to osteogenic cells [16], [45]. ALN-functionalized MSNs and polyelectrolyte nanoparticles also use bisphosphonate affinity for bone mineral to promote delivery of siRNA targeting SOST or Dcstamp to bone-resident cells [14], [15]. In these systems, preferential cellular delivery depends largely on the spatial association between the targeted remodeling interface and the resident cell population. In siP@WM, PASP first promotes bone accumulation, after which ROS-responsive TK cleavage reduces bone-matrix binding and increases FA exposure. FA–FRβ binding then provides an additional mechanism for receptor-mediated macrophage uptake. Therefore, our strategy is not merely a consequence of spatial proximity but additionally involves ligand–receptor recognition at the cellular level. Other ROS-responsive platforms for osteoporosis primarily use oxidant-sensitive linkers to release antioxidants or inhibitors of osteoclast activity after bone accumulation [46], [47]. Here, ROS responsiveness changes ligand accessibility to connect initial bone retention with subsequent macrophage uptake. This sequential targeting strategy may improve the efficiency of siIRF5 delivery within the osteoporotic bone microenvironment.

3.4. Coordinated ROS scavenging and macrophage reprogramming suppress osteoclast overactivation and modulate ECM remodeling

The improvement in bone mass observed in OVX mice highlighted the therapeutic benefit of coordinated ROS scavenging and siIRF5-mediated macrophage reprogramming. Transcriptomic analysis further identified the molecular pathways associated with this response. The AP-1 transcription factor forms homo- or heterodimers through interactions among members of the Fos, Jun, activating transcription factor (ATF), and musculoaponeurotic fibrosarcoma (MAF) families. AP-1 can be activated by ROS [48] and interacts with classical inflammatory mediators such as IL-1β and Ptgs2 within TNF, IL-17, and NF-κB pathways, thereby playing a critical regulatory and amplifying role in related inflammatory signaling [49]. In the siP@WM group, key AP-1 components (c-Fos, c-Jun, FosB, JunB) and downstream inflammatory mediators (IL-1β, Ptgs2) were suppressed. Together with immunostaining for IRF5, CD86, CD206, HMOX-1, 8-OHdG, and 4-HNE, these results were consistent with reduced inflammation and oxidative stress in bone tissue. These effects may reflect both ROS scavenging and changes in macrophage polarization.

The AP-1 transcription factor also links inflammatory processes to the activation of osteoclast signaling pathways, thereby serving as a key pathogenic factor in inflammatory bone diseases such as PO [50]. c-Fos, a critical component of AP-1, is essential for osteoclast differentiation by initiating the necessary signaling, and c-Fos-deficient mice exhibit severe osteopetrosis due to a lack of osteoclasts [51]. Sirpb and Pira encode critical co-stimulatory receptors in osteoclasts. The signals transduced by these receptors promote the autoamplification of the transcription factor NFATc1, which in turn initiates and sustains the transcription of osteoclast-specific genes [52]. Reduced expression of c-Fos, Sirpb, and Pira in the siP@WM group was consistent with suppressed osteoclast differentiation. However, whole-femur analyses could not identify the cells responsible for the reduction in AP-1 signaling or directly establish its effect on osteoclastogenesis.

Nanotherapies can also modulate the osteoimmune microenvironment by sequestering extracellular mediators of osteoclastogenesis. Zhou et al. [53] developed preosteoclast membrane-coated poly(lactic-co-glycolic acid) (PLGA) nanodecoys that sequestered RANKL and TNF-α and restored the osteoclast–osteoblast balance in OVX mice. siP@WM instead modulates upstream M1 macrophages that promote osteoclastogenesis through combined ROS scavenging and IRF5 silencing. The CM experiments supported a paracrine effect on osteoclast precursors. M1 CM did not significantly increase JNK or c-Jun phosphorylation beyond that induced by RANKL alone, possibly because the strong RANKL response limited further detectable activation. In contrast, siIRF5 (NP) CM reduced JNK and c-Jun phosphorylation, nuclear c-Fos accumulation, osteoclastogenic gene expression, and the formation of TRAP⁺ multinucleated cells and F-actin rings. Similar effects were observed with M2 CM. Inhibition of AP-1 with T-5224 [54] also reduced osteoclastogenic gene expression and differentiation in the presence of M1 CM, supporting AP-1 involvement in this macrophage-mediated response. Together with the in vivo transcriptomic and histological findings, these results suggested that siP@WM suppressed osteoclastogenesis partly by altering macrophage paracrine signaling. Nevertheless, contributions from effects on other bone-resident cells cannot be excluded.

Recent studies highlight the important role of LepR-expressing cells in new bone formation [55], [56], [57], and LepR expression was upregulated in the siP@WM group. This change, together with increased mineral apposition and serum P1NP, was consistent with improved bone formation. Silica and its silicate hydrolysis products have also demonstrated potential to promote osteoblast proliferation, differentiation, and activity [58], [59]. Reduced inflammation and oxidative stress may contribute to a microenvironment that supports bone formation, alongside possible effects of the silica-based carrier.

Reduced inflammation following nanoparticle treatment may also affect ECM remodeling. MMP13, a major collagenase in bone tissue under inflammatory conditions, degrades collagen matrix following osteoclast-mediated dissolution of hydroxyapatite and is also regulated by AP-1 [60]. Both transcriptomic analysis and tissue immunofluorescence showed reduced MMP13 expression in the siP@WM group, supporting attenuation of collagen degradation. Although expression of several collagen synthesis-related genes such as Col5a2 was reduced in the siP@WM group, transcript and protein levels of P4ha3 increased. P4ha3 is involved in post-translational collagen modification, encoding one of the catalytic subunits of collagen prolyl 4-hydroxylases (P4H), which catalyze prolyl 4-hydroxylation essential for the thermal stability of the collagen triple helix [61]. These changes are consistent with reduced matrix degradation and altered collagen processing during normalization of high bone turnover. As bone mass recovers and collagen degradation slows, collagen synthesis demand decreases, while existing collagen might become more stable, ultimately contributing to the suppression of high bone turnover in PO.

3.5. Limitations

Several limitations should be acknowledged. This study evaluated treatment over four weeks, and the duration of the skeletal benefits after treatment withdrawal remains unknown. Because siP@WM does not correct estrogen deficiency, the need for maintenance treatment requires further investigation. The in vivo dose–response relationship and the effects of repeated dosing in healthy animals were not assessed. Safety evaluation was limited to short-term observations, and the long-term retention, clearance, and potential toxicity of Pt remain uncertain. In addition, the absence of a clinically used comparator, such as a bisphosphonate, precluded direct comparison with established osteoporosis treatments. Future studies should optimize dosing and assess the persistence of therapeutic benefits, effects on healthy bone, and efficacy and long-term safety relative to established therapies.

4. Conclusion

In summary, we develop a ROS-responsive nanomedicine (siP@WM) for programmed sequential bone–macrophage targeting that integrates a Pt nanozyme core with siIRF5 delivery to restore bone homeostasis in postmenopausal osteoporosis. The mesoporous silica shell provides structural stability, efficient siRNA encapsulation, and stepwise functionalization, while the surface PEG chains enable sequential bone enrichment and uptake by pro-inflammatory M1 macrophages in a ROS-rich microenvironment. Mechanistically, siP@WM alleviates oxidative stress and reprograms macrophage polarization through ROS scavenging and IRF5 gene silencing, thereby suppressing AP-1–associated inflammatory signaling, inhibiting osteoclast differentiation, and modulating ECM remodeling. This coordinated intervention effectively ameliorates bone loss and restores bone microstructural and mechanical integrity in OVX mice.

5. Materials and methods

5.1. Synthesis of Pt@WM nanoparticles

Potassium tetrachloroplatinate (K₂PtCl₄, 12 mL, 10 mM, Aladdin) aqueous solution was mixed with 12 mL of cetyltrimethylammonium chloride (CTAC, 25 wt%, Macklin) as a stabilizing agent, followed by the addition of 90 mL of deionized (DI) water. The mixture was stirred at room temperature for 10 min and then at 50 °C for an additional 10 min until the solution became transparent, indicating micelle formation. Subsequently, 7.2 mL of pre-cooled sodium borohydride (NaBH₄, 0.5 M, Aladdin) aqueous solution was carefully added dropwise. The reaction vessel was transferred to an ice bath and maintained under an open atmosphere for 20 min, followed by stirring at 50 °C overnight. After completion, the reaction mixture was centrifuged at 36000 × g for 1.5 h to collect Pt nanoparticles (PtNPs).

The PtNPs were redispersed in 96 mL of DI water, followed by the addition of 2.4 mL of sodium hydroxide (NaOH, 0.05 M, Aladdin) and stirring at room temperature for 10 min. Under vigorous stirring, 0.9 mL of tetraethyl orthosilicate (TEOS, Macklin) premixed with 16.2 mL of methanol (Macklin) was added dropwise as the silica source. The reaction proceeded at 35 °C for 3 h. The product was collected by centrifugation (36000 × g, 45 min) and washed three times with DI water to obtain Pt@M nanoparticles.

The Pt@M was redispersed in 12 mL of DI water, followed by the addition of 7.2 mL of CTAC (25 wt%) as the structure-directing agent and 5.4 mL of triethanolamine (TEA, 12 mg/mL, Aladdin) as the catalyst. After stirring at 60 °C for 1 h to form micelles, 1.2 mL of TEOS premixed with 15.8 mL of chlorobenzene (Macklin) was added dropwise under vigorous stirring. The reaction was maintained at 60 °C for 3 h. After centrifugation (36000 × g, 30 min), the precipitate was extracted with acidic methanol (37% HCl/methanol = 1:10, v/v) at 60 °C for 6 h. The process was repeated three times to remove the CTAC template. The Pt@WM nanoparticles were stabilized in ethanol (Macklin) and stored at room temperature.

5.2. Surface modification

3-glycidyloxypropyltrimethoxysilane (GPTMS, 1 mL, Macklin) and branched polyethyleneimine (PEI, 14 mg, M_w = 10,000, Macklin) were dissolved in 20 mL of N,N-dimethylformamide (DMF, Sangon) and stirred at 50 °C for 3 h. Subsequently, 80 mg of Pt@WM nanoparticles dispersed in 10 mL of DMF were added dropwise, followed by stirring at 50 °C for 24 h. The resulting Pt@WM-PEI was collected by centrifugation (36000 × g, 30 min) and washed three times with DMF.

The Pt@WM-PEI was redispersed in 10 mL of DMF. Separately, carboxyl–PEG₂₀₀₀–thioketal–maleimide (HOOC–PEG₂₀₀₀–TK–MAL, 6 mg, Xarxbio) was activated in 9 mL of DMF with N-hydroxysuccinimide (NHS, 3 mg, Sigma) and N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC·HCl, 6 mg, Sigma) at 37 °C for 1 h. The activated solution was added dropwise to Pt@WM-PEI and reacted at 37 °C for 24 h to obtain Pt@WM-MAL. After washing with ethanol, Pt@WM-MAL was redispersed in 10 mL of ethanol and reacted with NHS–PEG₂₀₀₀–folic acid (NHS–PEG₂₀₀₀–FA, 60 mg, Xarxbio) dissolved in 40 mL of ethanol at 37 °C for 6 h to obtain Pt@WM-FA. Subsequently, rhodamine B isothiocyanate (RITC, 0.8 mg, Aladdin) dissolved in 40 mL of ethanol was added and reacted at room temperature for 6 h in the dark, yielding Pt@WM-FA/R. Finally, Pt@WM-FA/R was redispersed in 10 mL of DMF and reacted with poly(aspartic acid)₁₆₀₀–PEG₁₀₀₀–thiol (PASP₁₆₀₀–PEG₁₀₀₀–SH, 2.4 mg, Xarxbio) dissolved in 60 mL of DMF and supplemented with 10 mL of DI water at 37 °C for 4 h. After washing with ethanol three times, the final Pt@WM-PASP nanoparticles were stored in ethanol at room temperature.

5.3. Characterization

Scanning electron microscopy (SEM) images were acquired using a Regulus 8100 microscope (Hitachi) at 3.0 kV without coating. For transmission electron microscopy (TEM) measurements, the specimens were dispersed in ethanol and transferred to a copper grid. TEM images were obtained using a JEM-1200EX microscope (JEOL) operated at 100 kV. Elemental mapping was conducted using TEM-based energy dispersive spectroscopy (EDS) on a Tecnai G2 F30 microscope (FEI) operated at 300 kV. The Pt/silica ratio was determined by TEM-EDS. ζ-potential and dynamic light scattering (DLS) measurements were carried out in DI water at 25 °C using a Zetasizer Pro (Malvern). Nitrogen adsorption–desorption isotherms were recorded at −196 °C using a Micromeritics ASAP® 2020 analyzer. Samples were degassed at 180 °C for 12 h prior to measurement. The Brunauer–Emmett–Teller (BET) method was used to calculate the specific surface area, and pore size distributions were derived from the adsorption branch using the Barrett–Joyner–Halenda (BJH) method. The total pore volume was calculated from the amount adsorbed at maximum relative pressure (P/P₀) of 0.99. FT-IR spectra of Pt@WM, Pt@WM-PEI, and Pt@WM-MAL were acquired using the KBr pellet method on a Nicolet 6700 spectrometer (Thermo Fisher Scientific). Spectra were recorded from 400 to 4000 cm⁻¹ at a resolution of 4 cm⁻¹ by averaging 32 scans. The instrument signal-to-noise ratio was 50,000:1. UV–Vis absorption spectra of NHS–PEG₂₀₀₀–FA, Pt@WM-MAL, and Pt@WM-FA/R were recorded from 200 to 800 nm using a Cary 60 spectrophotometer (Agilent). Samples were dispersed in DI water by ultrasonication. The spectrum of Pt@WM-MAL served as the reference for identifying absorption features attributable to FA. PASP conjugation was assessed by high performance liquid chromatography (HPLC)-based amino acid analysis after hydrolysis of Pt@WM-PASP with 6 M HCl. Insoluble residues were removed by filtration, and the hydrolysate was analyzed using an L-8900 amino acid analyzer (Hitachi) equipped with a cation-exchange resin column (4.6 mm internal diameter × 60 mm length; 3 μm particle size) under the protein hydrolysate analysis program. Amino acids were separated by cation-exchange chromatography, followed by post-column derivatization with ninhydrin and detection at 570 nm. A mixed amino acid standard was analyzed to assign the retention time of aspartic acid (Asp). The appearance of an Asp peak in the hydrolysate at the retention time corresponding to the standard was used as qualitative evidence of PASP conjugation.

Thermogravimetric analysis (TGA) was performed in air at a heating rate of 10 K/min using a Q600 simultaneous thermal analyzer (TA Instruments). Mass loss below 100 °C was excluded to minimize contributions from adsorbed water and residual solvent. The mass remaining at 100 °C was used as the dry sample mass. For each formulation, the mass-retention factor between 100 and 900 °C was calculated as:

$$R_i=\frac{W_{900,i}}{W_{100,i}}$$

where $W_{100,i}$ and $W_{900,i}$ represent the residual mass percentages of formulation $i$ at 100 and 900 °C, respectively. Pt@WM was used as the reference Pt@silica core–shell material. The mass of each modified formulation corresponding to 100 μg of Pt@WM was calculated as:

$$M_i=100\times \frac{R_{Pt@WM}}{R_i}$$

The difference between $M_i$ and 100 μg was taken as the estimated mass added during surface modification. Unless otherwise specified, nanoparticle doses were normalized to equivalent amounts of the Pt@WM core–shell material using this calculation. Residual CTAC was assessed by comparing pure CTAC and Pt@WM under the same TGA conditions. DTG curves were generated from the first derivative of the corresponding TGA curves and compared over the principal CTAC decomposition range.

5.4. Degradation of nanoparticles

To evaluate nanoparticle degradability, Pt@WM and Pt@WM-PASP were dispersed in 10 mL of phosphate-buffered saline (PBS, pH 7.4) at a concentration of 1 mg/mL and incubated at room temperature under constant stirring for up to 28 days. At predetermined time points (1, 3, 7, 14, and 28 days), samples were collected for TEM observation. Additionally, 1 mL of supernatant was collected, centrifuged, and analyzed using an iCAP 7400 inductively coupled plasma optical emission spectrometry (ICP-OES, Thermo Scientific) instrument to determine the concentration of dissolved Si released from the mesoporous silica framework. The precipitate was redispersed in fresh PBS to continue the degradation process.

5.5. Preparation of siP@WM

Small interfering RNA targeting interferon regulatory factor 5 (siIRF5) was loaded into the pores of Pt@WM-PASP to form siP@WM. To determine the optimal loading ratio, nucleic acid binding and agarose gel electrophoresis assays were performed. For preparation, 112 pmol of siIRF5 (1.5 μg, GenePharma) was mixed with 75 μg of Pt@WM-PASP dispersed in 200 μL of diethyl pyrocarbonate (DEPC)-treated water (Beyotime), vortexed for 30 s, and incubated at 4 °C overnight. Due to the instability of siRNA, siP@WM was freshly prepared prior to use. Detailed sequences of siRNAs are provided in Table S3.

5.6. Nucleic acid binding and protection

For nucleic acid binding, 150 pmol of siIRF5 (2 μg) was mixed with Pt@WM-PASP (12.5, 25, 50, 100, and 200 μg, respectively) in DEPC-treated water. The total volume was adjusted to 40 μL. After vortexing and incubation at 4 °C overnight, the mixtures were centrifuged at 20,000 × g, and the supernatants were collected. The nucleic acid concentration was quantified using a NanoDrop™ 8000 spectrophotometer (Thermo Scientific).

For agarose gel electrophoresis, the siIRF5–nanoparticle complexes were loaded onto a 2% agarose gel (Biowest) prepared in Tris–acetate–EDTA (TAE) buffer containing 0.01% (v/v) SYBR Safe DNA gel stain (Thermo Scientific). Electrophoresis was conducted at 100 V for 15 min. Gels were visualized using a Fusion FX imaging system (Vilber), and band intensities were quantified using ImageJ with Fiji software [62]. For nucleic acid protection, 18.75 pmol (0.25 μg) of free siIRF5 or siIRF5 bound to 12.5 μg of Pt@WM-PASP was incubated with RNase A (0.01 μg/μL, Biosharp) at 4 °C for 10 min. Dithiothreitol (DTT, 100 mM, Aladdin) was then added, followed by incubation at 60 °C for 30 min to inactivate RNase A. Excess heparin sodium (Solarbio) was added to displace siIRF5 from PEI, and the released siRNA was analyzed by agarose gel electrophoresis as described above.

To assess whether H₂O₂ pretreatment affected siRNA protection and gene-silencing activity, siP@WM prepared at a siIRF5-to-nanocarrier mass ratio of 1:50 was incubated in PBS containing 0, 1, or 100 μM H₂O₂ at room temperature for 24 h. The complexes were collected by centrifugation at 20,000 × g and washed three times with DEPC-treated water to remove residual H₂O₂. For the protection assay, recovered complexes were incubated with or without RNase A, followed by RNase A inactivation with DTT and siRNA displacement with heparin as described above. Released siRNA was analyzed by agarose gel electrophoresis, and band intensities were normalized to those of PBS-pretreated siP@WM. For the gene-silencing assay, recovered complexes were added to RAW264.7 cells at an equivalent nanocarrier concentration of 37.5 μg/mL. Untreated cells served as the blank control. IRF5 mRNA expression was measured by RT-qPCR after 24 h.

5.7. Amine reactivity

The amount of amine groups, representing reactive sites for surface modification, was monitored using the ninhydrin reaction. Due to non-stoichiometric behavior and potential interference from folic acid, the results reflect relative trends rather than absolute grafting amounts. Briefly, 400 μg of Pt@WM, Pt@WM-PEI, Pt@WM-MAL, Pt@WM-FA, Pt@WM-FA/R, and Pt@WM-PASP were dispersed in 100 μL of DI water, respectively. Then, 100 μL of ninhydrin solution (2% (w/v) in dimethyl sulfoxide (DMSO), Aladdin) was added, followed by incubation at 90 °C for 30 min. After cooling to room temperature, 800 μL of ethanol was added to terminate the reaction. Absorbance at 570 nm was recorded using a Cytation 7 microplate reader (BioTek). Glycine solutions (2.5–25 mM, Macklin) were used to generate a standard curve.

5.8. Maleimide quantification

Samples of Pt@WM-PEI, Pt@WM-MAL, Pt@WM-FA/R, and Pt@WM-PASP (400 μg) were dispersed in 800 μL of DI water and reacted with 50 μL of reduced glutathione (GSH, 10 mM, Macklin) at 37 °C for 1 h. Subsequently, 150 μL of 4 mg/mL 4,4′-dipyridyl disulfide (DTDP, Macklin) was added and incubated at 25 °C for 10 min. The absorbance at 324 nm was recorded using a Cytation 7 microplate reader (BioTek). Maleimide content was calculated based on the consumption of GSH. GSH solutions (1–10 mM) were used to construct a standard curve.

5.9. ROS-responsive thioketal cleavage

To evaluate reactive oxygen species (ROS)-responsive cleavage of the thioketal (TK) linker, 400 μg of Pt@WM-FA/R was dispersed in 800 μL of DI water and reacted with excess reduced GSH at 37 °C for 1 h. After washing three times with DI water, complete thiol-maleimide conjugation was confirmed using the maleimide quantification assay. The resulting Pt@WM-FA/R–GSH was treated with hydrogen peroxide (H₂O₂, 0.2–4 wt%) at 37 °C for 3 h. These concentrations provided accelerated oxidative conditions for assessing TK cleavage and were not intended to reproduce physiological extracellular H₂O₂ levels. After centrifugation, the precipitate was collected. TK cleavage was estimated indirectly using the described ninhydrin assay based on the decrease in GSH content. Untreated Pt@WM-FA/R and Pt@WM-FA/R–GSH without H₂O₂ treatment were used as positive and negative controls, respectively.

5.10. In vitro nanozyme activity

The catalase (CAT)-, peroxidase (POD)-, and superoxide dismutase (SOD)-like activities of PtNPs, Pt@WM, Pt@WM-PEI, Pt@WM-FA/R, Pt@WM-PASP, and WM-PASP were evaluated using commercial assay kits: CAT activity assay kit (BC4780, Solarbio), POD activity assay kit (PMK1038, Bioprimacy), and SOD activity assay kit (BC5160, Solarbio), respectively. All procedures and calculations were performed according to the manufacturers’ instructions.

5.11. Bone affinity

Circular bovine bone slices (15 mm diameter) were carefully polished to remove surface defects and incubated with Pt@WM-FA/R, Pt@WM-PASP, and H₂O₂-treated Pt@WM-PASP (500 μg/mL) at room temperature for 4 h under gentle shaking (75 rpm) in the dark, respectively. Bone slices incubated with PBS served as the negative control. After washing with PBS three times, nanoparticle retention was evaluated by SEM and a TCS SP8 DIVE confocal laser scanning microscope (CLSM, Leica). The mean fluorescence intensity was quantified using ImageJ with Fiji software [62].

For the PASP competition assay, bone slices were preincubated with PBS or sodium poly(aspartate) (PASP-Na, M.W 4000, Macklin) at 5 or 20 wt% for 24 h at room temperature with shaking at 75 rpm. The pretreatment solutions were removed, and the slices were incubated with Pt@WM-PASP (500 μg/mL) for 4 h under the conditions described above. Slices pretreated with PBS or PASP-Na without nanoparticle exposure served as fluorescence controls. After three washes with PBS, nanoparticle retention was assessed by gross imaging and CLSM.

To assess H₂O₂-induced release after bone binding, bone slices were incubated with Pt@WM-PASP (500 μg/mL) for 4 h under the adsorption conditions described above and washed three times with PBS. The slices were then incubated in PBS containing 0, 1, or 100 μM H₂O₂ at room temperature with shaking at 75 rpm for 72 h. Solutions were replaced every 12 h. The H₂O₂ concentrations were selected to represent low and elevated levels based on previous studies [63], [64], and PBS alone served as the control for passive desorption during shaking. After three washes with PBS, nanoparticles remaining on the bone surfaces were examined by gross imaging and CLSM.

5.12. Cell culture

RAW264.7 cells were purchased from Servicebio Biotechnology Co., Ltd. (Wuhan, Hubei, China). MC3T3-E1 and bEnd.3 cells were purchased from Procell Biotechnology Co., Ltd. (Wuhan, Hubei, China). All cells were cultured at 37 °C in a humidified atmosphere containing 5% CO₂. RAW264.7 and bEnd.3 cells were maintained in Dulbecco's modified Eagle medium (DMEM, Gibco), whereas MC3T3-E1 cells were maintained in α-modified minimum essential medium (α-MEM, Gibco). Both culture media were supplemented with 10% (v/v) fetal bovine serum (FBS, Procell), 100 U/mL penicillin, and 0.1 mg/mL streptomycin (Sigma).

5.13. Cell viability

RAW264.7 cells were seeded in 96-well plates (3 × 10³ cells per well) and treated with nanoparticles (0–100 μg/mL). After incubation for 24, 48, or 72 h, 100 μL of Cell Counting Kit-8 (CCK-8, AQ308, Aoqing) solution was added to each well and incubated for 30 min at 37 °C. Absorbance at 450 nm was measured using a Cytation 7 microplate reader (BioTek). Cell viability was calculated relative to untreated controls.

5.14. Macrophage polarization and transfection

RAW264.7 cells were polarized toward the M1 phenotype using lipopolysaccharide (LPS, 100 ng/mL, Sigma) for 24 h, or toward the M2 phenotype using recombinant mouse interleukin-4 (IL-4, 20 ng/mL, Novaprotein) and interleukin-13 (IL-13, 20 ng/mL, Abclonal) for 24 h. For transfection, M1 macrophages were treated with siIRF5 or negative control siRNA (siCON) using jetPRIME® transfection reagent (Polyplus) for 24 h. For nanoparticle-mediated transfection, siIRF5 or siCON was loaded onto Pt@WM-PASP at the previously described ratio and incubated with cells for 24 h. The nanoparticle concentration was 37.5 μg/mL, corresponding to approximately 0.75 μg/mL siRNA at a siRNA-to-nanocarrier mass ratio of 1:50. The same siRNA concentration was used in the corresponding jetPRIME-mediated transfection groups.

5.15. Western blotting

Total cellular proteins were extracted using RIPA lysis buffer (Solarbio) supplemented with phenylmethylsulfonyl fluoride (PMSF, 2 mM) on ice, followed by centrifugation at 12000 × g for 15 min. Protein concentration was determined using a Pierce™ BCA protein assay kit (Thermo Scientific). Equal amounts of protein (30 μg) were denatured, separated by SDS–PAGE (4–12%, ACE FuturePAGE™), and transferred to polyvinylidene fluoride (PVDF) membranes (0.2 μm, Millipore). Membranes were blocked using a QuickBlock™ blocking buffer (P0269, Beyotime) and incubated with primary antibodies at 4 °C overnight, followed by incubation with secondary antibodies at room temperature for 1 h. Protein bands were visualized using an enhanced chemiluminescence (ECL) kit (Affinibody). Band intensities were quantified using ImageJ with Fiji software [62]. Antibody details are listed in Table S4.

5.16. Reverse transcription quantitative polymerase chain reaction (RT-qPCR)

Total RNA was extracted using TRIzol reagent (Takara). Complementary DNA (cDNA) was synthesized from 1 μg of RNA using a reverse transcription kit (Takara). Quantitative PCR was performed using SYBR Green Fast qPCR Mix (Abclonal) on an Applied Biosystems 7500 system (Thermo Scientific). Gene expression levels were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and analyzed using the 2^−ΔΔCt method. Primer sequences are listed in Table S5.

5.17. Cell immunofluorescence staining

Cells cultured on coverslips were fixed with 4% paraformaldehyde (Biosharp), permeabilized with 0.1% Triton X-100 (Bioruler), and blocked with a QuickBlock™ blocking buffer (P0269, Beyotime). Cells were incubated with primary antibodies at 4 °C overnight, followed by secondary antibodies at room temperature for 1 h. Coverslips were mounted with Fluoroshield™ containing DAPI (Sigma) and observed using CLSM. Antibody details are listed in Table S4.

5.18. Flow cytometry

Cells were washed with PBS, resuspended at 5 × 10⁶ cells/mL in PBS containing 2% FBS, fixed with 4% paraformaldehyde, and permeabilized with 0.1% Triton X-100. Fc receptors were blocked using anti-CD16/32 antibody (Proteintech) prior to staining at 4 °C for 15 min in the dark. Then, cells were incubated with corresponding fluorophore-conjugated antibodies at 4 °C for 30 min in the dark. Samples were analyzed using a BD Accuri™ C6 flow cytometer. Antibody details are listed in Table S4.

5.19. Cellular ROS scavenging

M1-polarized RAW264.7 macrophages were incubated with Pt@WM-PASP or WM-PASP (25–100 μg/mL) for 24 h. Intracellular ROS levels were assessed using a ROS Assay Kit (S0033S, Beyotime) following the manufacturer's instructions. Within the cells, ROS oxidize the non-fluorescent 2′,7′-dichlorodihydrofluorescein (DCFH) to generate fluorescent 2′,7′-dichlorofluorescein (DCF). Fluorescence intensity of DCF was analyzed using flow cytometry. Untreated M1 macrophages served as the positive control.

5.20. Cellular uptake and targeting

To evaluate cellular uptake and folic acid-mediated targeting, RITC-labeled Pt@WM-MAL, Pt@WM-PASP, and H₂O₂-treated Pt@WM-PASP (6.25–100 μg/mL) were incubated with M1-polarized RAW264.7 macrophages for 24 h, respectively. Cells treated with PBS alone served as the negative control. Intracellular fluorescence was analyzed by CLSM and flow cytometry.

To assess the roles of FA–FRβ binding and endogenous ROS in nanoparticle uptake, M1-polarized RAW264.7 cells were pretreated with free FA, an anti-FRβ antibody, or N-acetylcysteine (NAC). For FA competition, cells were incubated with FA (1 mM, Aladdin) for 1 h. For FRβ blockade, cells were incubated with an anti-CD16/32 antibody (1 μg/mL) for 10 min to block Fc receptors, followed by an anti-FRβ antibody (5 μg/mL) for 50 min. For ROS scavenging, cells were incubated with NAC (10 mM, Aladdin) for 1 h. Pt@WM-MAL, intact Pt@WM-PASP, or H₂O₂-treated Pt@WM-PASP was then added at 12.5 μg/mL. After 4 h, cells were washed three times with PBS, fixed with 4% paraformaldehyde (Biosharp), and permeabilized with 0.1% Triton X-100 (Bioruler). F-actin was stained with FITC-conjugated phalloidin (Abclonal), and cells were mounted with Fluoroshield™ containing DAPI (Sigma). Nanoparticle uptake was examined by CLSM. Antibody details are provided in Table S4.

5.21. Cellular uptake in Co-culture

Before co-culture, M1-polarized RAW264.7 cells were incubated in serum-free DMEM (Gibco) containing CellTracker™ Deep Red (1 μM, Thermo Fisher Scientific) at 37 °C for 30 min in the dark. Cells were washed three times with PBS to remove unbound dye and residual LPS and detached by gentle pipetting. MC3T3-E1 and bEnd.3 cells were labeled separately in the corresponding serum-free medium containing CellTracker™ Green CMFDA (1×, Beyotime) at 37 °C for 30 min in the dark. Cells were then washed three times with PBS and detached with 0.25% trypsin–EDTA (Gibco) to obtain single-cell suspensions. Labeled RAW264.7 cells were mixed with labeled MC3T3-E1 or bEnd.3 cells at a 4:1 ratio and seeded for co-culture. After 12 h for attachment, Pt@WM-MAL, intact Pt@WM-PASP, or H₂O₂-treated Pt@WM-PASP was added at 12.5 μg/mL. After 4 h, cells were washed three times with PBS, fixed with 4% paraformaldehyde (Biosharp), and mounted with Fluoroshield™ containing DAPI (Sigma). Nanoparticle uptake by each cell population was examined by CLSM.

5.22. Endosomal escape

siP@WM (25 μg/mL) with Cy5-labeled siIRF5 (GenePharma) was incubated with M1-polarized RAW264.7 macrophages for 8, 24, and 48 h, respectively. Cells grown on coverslips were fixed, permeabilized, and blocked. Lysosomes were stained with anti-lysosomal-associated membrane protein 2 primary antibody (LAMP2, Proteintech) at 4 °C overnight, followed by incubation with secondary antibody at room temperature for 1 h. Coverslips were mounted with Fluoroshield™ with DAPI (Sigma) and observed using CLSM. For better resolution, confocal images were processed with the sparse deconvolution algorithm [65]. Manders’ coefficient was calculated using ImageJ with JACoP extension [66]. Detailed antibody information is provided in Table S4.

5.23. Macrophage-conditioned medium (CM) preparation and osteoclast differentiation

CM was collected from RAW264.7 cells after the M1, M2, siCON (NP), or siIRF5 (NP) treatments described above. After treatment, the original culture medium was removed. Cells were washed three times with PBS and incubated in fresh complete α-MEM for 12 h. Culture supernatants were centrifuged at 300 × g for 5 min and then at 2000 × g for 10 min, both at 4 °C, to remove cells and debris. The resulting M1 CM, M2 CM, siCON (NP) CM, and siIRF5 (NP) CM were stored at −80 °C until use.

Bone marrow-derived macrophages (BMMs) were collected from the femora and tibiae of 6–8-week-old female ICR mice. After euthanasia, bones were harvested aseptically, surrounding soft tissues were removed, and both ends were excised. Marrow was flushed with PBS containing 200 U/mL penicillin and 0.2 mg/mL streptomycin (Sigma), passed through a 70 μm cell strainer (Novbio), and centrifuged at 300 × g for 5 min. Cells were incubated in ACK erythrocyte lysis buffer (Beyotime) for 5 min, centrifuged at 300 × g for 5 min, washed once with complete α-MEM, and seeded in 10 cm culture dishes. After 24 h, nonadherent cells were collected, centrifuged at 300 × g for 5 min, and replated in complete α-MEM containing macrophage colony-stimulating factor (M-CSF, 20 ng/mL, Abclonal). After 48 h, the medium was replaced to remove nonadherent cells, and the adherent BMMs were used for osteoclast differentiation. BMMs were assigned to seven groups: Control, RANKL, M1 CM, M1 CM + T-5224, siCON (NP) CM, siIRF5 (NP) CM, and M2 CM. All groups received M-CSF (20 ng/mL). The Control group received neither RANKL nor CM; all other groups received RANKL (50 ng/mL, Thermo Fisher Scientific). For CM treatment, CM was mixed with fresh complete α-MEM at a 1:3 vol ratio and included throughout osteoclast induction. T-5224 (10 μM) was added to the designated group at the start of RANKL stimulation. Medium containing the appropriate cytokines, CM, and T-5224 was replaced every 2 days. Proteins were extracted 30 min after RANKL stimulation for Western blot analysis of phosphorylated and total JNK and c-Jun, with GAPDH as the loading control. Nuclear c-Fos localization was examined by immunofluorescence at 24 h, and NFATc1, Acp5, Mmp9, Oscar, Dcstamp, and Ocstamp mRNA levels were measured by RT-qPCR at 48 h. On day 6, osteoclast differentiation and maturation were evaluated using a TRAP staining kit (Amizona) and fluorescent staining of F-actin rings. TRAP⁺ multinucleated cells (≥3 nuclei) were quantified as the number per unit culture area and the percentage of the culture area occupied by these cells.

5.24. Animals

Female ICR mice (6–8 weeks old) were purchased from the Department of Laboratory Animal Science, Peking University Health Science Center. All animal experiments were conducted in accordance with the ARRIVE guidelines and approved by the Ethics Committee of Peking University Health Science Center (Approval No. DLASBD0254). All mice were maintained under specific pathogen-free conditions with a 12-h light/12-h dark cycle. The animals were allowed access to free food pellets and tap water ad libitum.

5.25. Ovariectomy model and treatments

Bilateral ovariectomy (OVX) or sham surgery was performed under general anesthesia using pentobarbital sodium (50 mg/kg). OVX mice were randomly divided into groups receiving saline, nanoparticles (WM-PASP or Pt@WM-PASP, 200 μg), siIRF5 (300 pmol), or siP@WM (200 μg of Pt@WM-PASP loaded with 300 pmol of siIRF5). Treatment was initiated 8 weeks after surgery to evaluate therapeutic efficacy after the development of estrogen-deficiency-induced bone loss [67], [68], [69], [70]. Treatments were administered via tail vein injection once a week for 4 weeks. At the endpoint, all mice were euthanized. The serum, internal organs (heart, liver, lungs, spleen, kidneys), and femora were collected.

5.26. Hemolysis

The hemolysis assay was carried out to evaluate the toxicity of nanoparticles to erythrocytes. Fresh mouse blood was collected in ethylenediaminetetraacetic acid (EDTA) anticoagulant tubes and centrifuged to isolate erythrocytes, which were washed three times with PBS. Erythrocytes were incubated with nanoparticle suspensions (6.25–200 μg/mL) at 37 °C for 1 h. After centrifugation, hemoglobin release was quantified by measuring absorbance at 540 nm. PBS and DI water were used as negative and positive controls, respectively.

5.27. Micro-CT scanning and analysis

The mouse femora were fixed in 70% ethanol and scanned using an IMAGING 100 micro-CT system (Raycision) at a resolution of 10 μm. Images were reconstructed and analyzed using 3D Slicer with Bone Texture Extension (Kitware) [71]. The region of interest was defined as 1.5–2.0 mm proximal to the distal femoral epiphysis. Trabecular bone parameters, including BMD, BV/TV, Tb.Th, Tb.Sp, and Tb.N, were quantified.

5.28. Bulk RNA sequencing

The mouse femora were harvested immediately after euthanasia and snap-frozen in liquid nitrogen. RNA extraction, library preparation, and sequencing were performed by Majorbio Biopharm Biotechnology Co., Ltd. (Shanghai, China) according to the manufacturer's instructions. The RNA-seq transcriptome library was prepared using the Illumina® Stranded mRNA Prep, Ligation Kit with 1 μg of total RNA. Briefly, messenger RNA was isolated using a poly(A) selection method by oligo(dT) beads and then fragmented with fragmentation buffer. Double-stranded cDNA was synthesized with random hexamer primers. Then the synthesized cDNA was subjected to end-repair, phosphorylation, and adapter addition according to the library construction protocol. Libraries were size-selected for cDNA target fragments of 300–400 bp using magnetic beads, followed by PCR amplification for 10–15 cycles. After quantification with Qubit 4.0, sequencing was performed on a NovaSeq X Plus platform (PE150) using NovaSeq Reagent Kit. The expression level of each transcript was calculated according to the transcripts per million reads (TPM) method. RSEM [72] was used to quantify gene abundances. Differentially expressed genes (DEGs) were identified using DESeq2 [73] with |log₂FC| ≥ 0.263 and FDR < 0.05. Further analyses were performed on Majorbio Cloud Platform (https://cloud.majorbio.com/) [74].

5.29. Bone histomorphology and immunofluorescence staining

Hard tissue femoral sections were prepared through dehydration, embedding in resin, and sectioning into 5–8 μm slices using a tungsten steel blade, followed by deplasticization. Bone mineralization was visualized using Von Kossa staining. To assess bone formation rate (BFR), sequential in vivo fluorescent labeling was performed: calcein (10 mg/kg, Aladdin) and tetracycline (60 mg/kg, Macklin) were intraperitoneally administered on weeks 3 and 4 of treatment, respectively. The inter-label distance was measured to calculate mineral apposition.

For conventional histology, femora were fixed in 4% paraformaldehyde for 24 h and decalcified in 10% EDTA solution (Servicebio) at room temperature. After dehydration, samples were embedded in paraffin and sectioned into 5 μm slices. Hematoxylin and eosin (H&E), Masson trichrome, and tartrate-resistant acid phosphatase (TRAP) staining were performed using commercial kits (Solarbio) following the manufacturers’ protocols. The slices were scanned using an Axioscan 7 digital slide scanner (Zeiss).

For immunofluorescence staining, femora underwent the same fixation, dehydration, and embedding procedures as above. Antigen retrieval was performed by incubating slides in Tris–EDTA buffer (pH 9.0, Beyotime) at 95 °C for 15 min. Sections were then incubated overnight at 4 °C with primary antibodies. After washing, sections were incubated with appropriate secondary antibodies for 1 h at room temperature and mounted with Fluoroshield™ containing DAPI (Sigma). Images were acquired using CLSM. Detailed information on primary and secondary antibodies is provided in Table S4.

5.30. Mechanical testing

Three-point bending tests were carried out to evaluate the mechanical performance of mouse femora, using an Exceed® E43.104 universal mechanical testing machine (MTS). The femora were centered on the supporting platform at 25 mm apart, with the anterior surface facing up. The load was applied at the center of the femur at a rate of 5 mm/min until bone fracture. Maximum loading, fracture energy, and Young's modulus were calculated using the built-in software.

5.31. Biodistribution

To assess the distribution and retention of nanoparticles, and the contribution of PASP modification to bone enrichment, female ICR mice (6–8 weeks old) were used. Pt@WM-FA/R or Pt@WM-PASP (1000 μg per mouse, dispersed in 200 μL saline) were administered via tail vein injection. A control group received 200 μL of saline to account for background fluorescence. At predetermined time points (1 h, 1 day, 7 days, and 14 days post-injection), three mice per group were euthanized, and organs (heart, liver, spleen, lungs, and kidneys) as well as femora and tibiae were collected. The tissues were immediately subjected to ex vivo fluorescence imaging using an IVIS spectrum in vivo imaging system (PerkinElmer) for visualization and quantitative analysis of nanoparticle-associated fluorescence.

5.32. Serum marker analysis

In order to assess potential effects of nanoparticles on bone metabolism, serum levels of mouse bone turnover markers, including procollagen type Ⅰ N-terminal propeptide (P1NP, Cusabio) and C-terminal telopeptide of type Ⅰ collagen (CTX-Ⅰ, Elabscience), were measured using enzyme-linked immunosorbent assay (ELISA) kits according to the manufacturers’ protocols.

5.33. Toxicity evaluation

To evaluate in vivo biocompatibility and potential organ toxicity, excised organs were fixed in 4% paraformaldehyde, embedded in paraffin, sectioned, and subjected to H&E staining. Histological sections were examined under a BX51 light microscope (Olympus) to identify any morphological alterations.

At the experimental endpoint, serum biochemical markers, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), total protein (TP), albumin (ALB), total bilirubin (TBIL), blood urea nitrogen (BUN), uric acid (UA), and creatinine (CREA), were measured using a BS-360E automated chemistry analyzer (Mindray) and the corresponding commercial reagent kits (Dirui Industrial Co., Ltd.) according to the manufacturers’ instructions.

For quantification of tissue silica content, collected lungs and livers were carefully weighed and stored in phosphate-buffered saline (PBS). Lungs and livers were mechanically homogenized using a high-throughput tissue grinder with steel beads (Qiagen), followed by ultrasonic disruption to ensure complete tissue dissociation. NaOH was then added to reach a final concentration of 4 M, and samples were incubated at room temperature for 24 h to fully dissolve silica nanoparticles. After centrifugation, the supernatants were filtered and analyzed by ICP-OES to determine the silica content. Data were normalized to tissue weight to quantify nanoparticle accumulation in each organ.

5.34. Statistical analysis

Statistical analyses were performed using GraphPad Prism (version 10.4.1, GraphPad Software, Boston, Massachusetts USA). Data are presented as mean ± standard deviation (SD). Two independent groups were compared using two-tailed unpaired Student's t-tests. Before one-way analysis of variance (ANOVA), residual normality and homogeneity of variances were assessed using the Shapiro–Wilk and Brown–Forsythe tests, respectively. Data meeting these assumptions were analyzed by one-way ANOVA followed by Tukey's multiple-comparisons test. P < 0.05 was considered statistically significant.

CRediT authorship contribution statement

Zining Chen: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation. Jinjin Wu: Methodology, Investigation. Yunyan Zhu: Resources. Anqi Tan: Investigation. Junhao Deng: Resources. Yuting Niu: Writing – review & editing, Supervision, Funding acquisition, Conceptualization. Weiran Li: Writing – review & editing, Supervision, Project administration, Funding acquisition. Yunfei Zheng: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Ethics approval and consent to participate

All animal experiments were conducted in accordance with the ARRIVE guidelines and approved by the Ethics Committee of Peking University Health Science Center (Approval No. DLASBD0254).

Funding

This project was supported by Beijing Natural Science Foundation (L244084, YN), Hainan Provincial Natural Science Foundation of China (825MS195, YZ), National Natural Science Foundation of China (82471028, YN; 82671242, YZ), and National Key Research and Development Program of China (2023YFC2413604, WL).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: