Biomimetic M2 exosome-based dual-drug nanoplatform for combined immunomodulation and bone protection in rheumatoid arthritis

Abstract

The current treatment of rheumatoid arthritis (RA) remains limited by severe drug-associated side effects and poor suppression of bone erosion. Herein, we report the development of biomimetic nanoparticles (CEC NPs) that co-deliver celecoxib (CXB) and the lysine-specific demethylase 1 (LSD1) inhibitor CC-90011 via M2 macrophage-derived exosomes (M2 Exos), thereby integrating targeted delivery with innovative multi-mechanistic therapeutic strategies. The M2 Exos enable innate homing to inflamed synovium and osteoclast-rich lesions, while ensuring efficient intracellular delivery. CEC NPs combined repolarize macrophages from the M1 to M2 phenotype, suppress fibroblast-like synoviocyte activation, and critically inhibit bone erosion by blocking the LSD1–NFATc1 signaling pathway in the osteoclast. Collectively, these effects result in potent anti-inflammatory and osteoprotective outcomes. Notably, we identified CC-90011 as a previously unrecognized anti-erosive agent to directly suppress bone erosion in RA treatment. In a collagen-induced arthritis (CIA) model, CEC NPs markedly outperform monotherapies, with pronounced reduction in joint swelling, cartilage degradation, and bone erosion, without systemic toxicity. This work introduces an M2 Exo-based dual-drug delivery system as a versatile strategy for multi-mechanistic modulation of inflammation and bone destruction, offering a promising paradigm that could be extended to other inflammatory and autoimmune bone disorders.

Graphical abstract

Highlights

• •Biomimetic dual-drug system enables anti-inflammation and anti-bone erosion in RA.
• •First evidence CC-90011 directly inhibits osteoclasts to suppress RA bone erosion.
• •First use of M2 Exos as CXB carriers to target inflammation and reprogram M1 to M2.
• •CXB suppresses RA-FLS activation, improving safety and therapeutic precision.

1. Introduction

Rheumatoid arthritis (RA) is a chronic inflammatory autoimmune disease characterized by inflammatory cell infiltration, persistent synovial inflammation, and progressive bone erosion [1,2]. It affects approximately 0.5% to 1.0% of the global population, and without timely and adequate intervention, up to 75% of patients may develop joint deformities and disabilities within three years [3]. Current clinical treatments primarily rely on non-steroidal anti-inflammatory drugs (NSAIDs), glucocorticoids (GCs), conventional disease-modifying anti-rheumatic drugs (DMARDs), and biologic agents [4]. However, each of these medications presents certain limitations, including adverse effects, insufficient accumulation at pathological sites, and diminished therapeutic response over time [5]. More importantly, although these agents can alleviate inflammation to some extent, they fail to effectively prevent bone destruction and restore joint architecture [6,7]. Long-term administration may even lead to complications such as osteoporosis and accelerated joint damage [8]. These shortcomings highlight the urgent need for novel therapeutic approaches that not only suppress inflammation but also concurrently prevent bone erosion and promote joint repair.

The pathogenesis of RA is characterized by immune-mediated inflammation, in which macrophages play a crucial role [9]. In RA, the balance between M1 and M2 macrophages is markedly disrupted, with a predominance of the pro-inflammatory M1 phenotype. Activated M1 secretes large amounts of pro-inflammatory cytokines, which further activate various immune cells and amplify the inflammatory cascade, leading to cartilage destruction and bone erosion [10]. Therefore, modulating the dynamic equilibrium between M1 and M2 macrophages represents a promising strategy for suppressing RA-associated inflammation. Furthermore, the inflammatory microenvironment promotes the activation and abnormal proliferation of fibroblast-like synoviocytes (FLS), thereby enhancing their invasive behavior and contributing to progressive joint destruction. In turn, activated FLS cells exacerbate the inflammation by producing abundant pro-inflammatory cytokines [11,12]. The activation of FLSs is closely related to the progression of arthritis, cartilage damage, and the severity of macrophage infiltration. Celecoxib (CXB), a selective cyclooxygenase-2 (COX-2) inhibitor, is widely used in the clinical setting to help alleviate inflammation and pain in RA patients [13]. Studies have shown that CXB promotes the polarization of macrophages from the pro-inflammatory M1 toward the anti-inflammatory M2 phenotype [14]. Furthermore, it can reduce FLS viability in a dose-dependent manner, thereby contributing to the amelioration of disease pathology [15,16]. However, CXB as a Biopharmaceutics Classification System (BCS) class II drug with low aqueous solubility and high hydrophobicity, may exhibit suboptimal pharmacokinetic behavior following systemic administration, including nonspecific distribution and reduced effective drug exposure, thereby limiting its accumulation at inflamed sites [17,18]. Moreover, prolonged use of CXB leads to gastrointestinal side effects, limiting its long-term clinical application [[19], [20], [21]]. Consequently, improving the solubility and bioavailability of CXB has become a major focus in the development of more effective therapeutic strategies for RA.

Bone destruction is another hallmark and debilitating consequence of RA, substantially contributing to the risk of long-term disability. Although the anti-inflammatory agents which effectively control inflammation provide partial protection against localized bone erosion, strategies that directly target osteoclastogenesis are poorly studied [22]. This pathological bone destruction is primarily driven by excessive osteoclast (OC)-mediated bone resorption [23,24]. Mature osteoclastogenesis heavily relies on glycolysis and oxidative phosphorylation for sufficient energy supply [25]. Emerging evidence has identified lysine-specific demethylase 1 (LSD1) as a critical regulator of glycolytic metabolism [26]. In murine models of osteoporosis, RA, and inflammation-induced bone erosion, blockade of LSD1 has been shown to suppress osteoclastogenesis and attenuate pathological bone resorption [27]. However, despite its therapeutic potential, LSD1-targeted therapy has not yet been explored for the treatment of RA. CC-90011 is a newly identified potent reversible LSD1 inhibitor that has demonstrated strong antitumor activity in small cell lung cancer, neuroendocrine tumors, and neuroendocrine carcinomas [28]. Clinical trials have shown that CC-90011 possesses favorable tolerability, clinical activity, and durable responses [29]. Therefore, CC-90011 may represent a novel and promising strategy to inhibit osteoclast-mediated bone destruction in RA. However, its therapeutic potential in RA has not yet been explored.

Exosomes, as lipid-based carriers, effectively protect encapsulated drugs from enzymatic degradation and immune clearance, thereby enhancing their bioavailability and prolonging their half-life [30]. Exosome-mediated drug delivery primarily depends on cellular uptake followed by intracellular release via membrane fusion or endocytic pathways, rather than passive extracellular diffusion [31]. In recent years, macrophage-derived exosomes have attracted increasing attention due to their intrinsic homing ability toward inflammatory microenvironments and their capacity to transfer bioactive molecules that regulate immune responses [32,33]. Notably, exosomes not only reflect the polarization state of their parent cells but also actively modulate macrophage phenotypes [34,35]. In particular, M2 Exos possess inherent inflammation-targeting capability and exhibit immunomodulatory properties by promoting macrophage polarization toward the anti-inflammatory M2 phenotype [36]. Therefore, M2 macrophage-derived exosomes were selected as the drug delivery vehicle in our design to enhance targeted anti-inflammatory therapy in RA, providing a mechanistic basis for the observed therapeutic effects.

In this study, we developed M2 exosome-coated nanoparticles (CEC NPs) co-loaded with CXB and CC-90011, aiming to establish a dual drug delivery system that targets the inflamed sites in RA and enables controlled drug release within the inflammatory microenvironment (Scheme 1). M2 exosomes (M2 Exos) can not only specifically home to inflammatory cells and exert anti-inflammatory biological effects, but also improve stability and bioavailability of drugs. CXB alleviates arthritis inflammation and inhibits synovial activation of FLSs. Meanwhile, CC-90011 suppresses pathological bone resorption and alleviates bone erosion by disrupting OC energy metabolism and reprogramming OC differentiation. In summary, our M2 Exo-based therapeutic nanoplatform achieves the dual-target effects by simultaneously reducing inflammation and preventing bone erosion, thereby providing a promising new strategy for RA treatment.

Scheme 1.

Schematic illustration of M2 Exo-based CEC NPs delivering CXB and CC-90011 to achieve combinational anti-inflammatory and osteoclast-suppressive effects in RA. (A) Schematic illustration of CEC NP synthesis. (B) Mechanistic roadmap of CEC NPs achieving dual anti-inflammatory and osteoclast-inhibitory effects in RA through combination therapy. In this study, a CIA model is established using DBA/1 mice. Following intravenous administration via tail vein injection, CEC NPs are specifically recruited to arthritic inflammatory sites through M2 Exo-mediated targeting. The acidic inflammatory microenvironment triggers the release of CC-90011 from the CEC NPs, leading to LSD1 inhibition and consequent suppression of osteoclastogenesis, ultimately preventing bone erosion. Concurrently, EC is selectively internalized by M1 macrophages and activated FLSs due to the inflammatory targeting capability of M2 Exo, exerting potent anti-inflammatory effects.

2. Results and discussion

2.1. Synthesis and characterization of CC-Exo@CXB nanoparticles

In this study, we developed a novel nanoparticle system utilizing M2 Exos as a biocompatible carrier to co-deliver CXB and CC-90011. This nanoplatform targets inflamed sites in RA and exerts dual therapeutic effects by attenuating inflammation and inhibiting bone erosion. Although lipid carriers have been reported for CXB delivery, the use of M2 Exos for CXB loading in RA treatment has not been reported [37]. As shown in Fig. 1A, we first incorporated CXB into M2 Exos to prepare uniform Exo@CXB (EC) NPs using a liposome extrusion technique. Subsequently, CC-90011 was conjugated to DSPE-Hyd-PEG₂₀₀₀-NHS (linker) via an amidation reaction, yielding the intermediate product DSPE-Hyd-PEG₂₀₀₀-CC-90011 (CC-linker). Finally, the DSPE moieties on the linker were then inserted into the exosomal lipid bilayer, successfully anchoring CC-90011 to the Exo surface and forming the final nanocomposite CC-Exo@CXB (CEC). Within CEC NPs, CXB alleviates joint inflammation by repolarizing macrophages to M2 and suppressing FLS activation. Concurrently, CC-90011 suppresses pathological bone resorption by disrupting the metabolic pathways essential for osteoclast (OC) differentiation and function, thereby effectively mitigating bone erosion.

Fig. 1.

Synthesis and Characterization of CEC NPs. (A) Diagram of the preparation process of CEC NPs. (B) FCM analysis shows the proportion of CC-linker-FITC modified EC NPs after co-incubation with different concentrations of CC-linker-FITC (10, 50, and 100 μM) with EC. (C) Representative CLSM images of CEC NPs showing signals for EC NPs (red, labeled with Dil) and CC-linker (green, labeled with FITC). Scale bar = 1 μm. (D) Representative TEM images of M2 Exo, EC NPs, and CEC NPs. Scale bar = 100 nm. (E) Western blot analysis of exosome markers (TSG101, CD81), and macrophage markers (Arg-1) in different material groups (M2 Exo, EC NPs, and CEC NPs), n = 3. (F) Zeta potential of CC-linker, EC NPs, and CEC NPs. (G) Hydrodynamic size distribution of Exo, EC NPs, and CEC NPs measured by DLS. (H) UV-Vis absorption spectra of CC-linker, EC NPs, and CEC NPs. (I) Hydrodynamic size of CEC NPs dispersed in PBS for different storage periods. (J) Cumulative release of CC-90011 from CEC NPs are measured in buffers with different pH values (Red: pH 6.5, Blue: pH 7.4). n = 3. All the representative images are shown from three independent experiments.

To prepare M2 Exos, RAW 264.7 cells were first polarized to the M2 phenotype by IL-4 stimulation (Fig. S1A). The phenotypic transition was investigated by measuring the expression levels of surface markers CD206 and CD86 via flow cytometry (FCM). The results showed that the proportion of CD206⁺ cells reached approximately 90% after IL-4 treatment at concentrations of 20 and 100 ng/mL, demonstrating the successful polarization of RAW 264.7 into M2 (Fig. S1B). M2 Exos were subsequently collected from culture supernatant by gradient centrifugation. To verify the successful extraction and purity of exosomes, exosome-associated protein expression in macrophages and exosomes was analyzed by Western blot. The results showed that M2 Exos highly expressed exosome markers TSG101, CD9, and lacked the endoplasmic reticulum marker Calnexin (Fig. S1C), confirming the purity of the exosomal preparation. We then fabricated EC NPs by loading CXB into M2 Exos using a liposome extruder (Fig. S2A). The successful encapsulation of EC NPs was confirmed by UV-Vis absorption spectra (Fig. S2B). High-performance liquid chromatography (HPLC) analysis revealed that the loading efficiency (LE) of CXB into EC was approximately 43.8% (Fig. S2C–E).

To load CC-90011 onto EC, CC-90011 was covalently conjugated to the amphiphilic linker DSPE-Hyd-PEG₂₀₀₀-NHS via an amidation reaction, forming the CC-linker conjugate (Fig. S3A). The successful synthesis of the conjugate was confirmed by ¹H NMR and Raman spectroscopy. Compared to CC-90011 and linker, the appearance of the amide hydrogen peak (δ 8.33) in the CC-linker indicated the formation of the new amide bond, confirming the successful synthesis of CC-linker (Fig. S3B–D). The Raman spectra further confirmed the successful conjugation, showing the presence of the characteristic peaks of the linker at approximately 2874 cm⁻¹ for the -CH₃ vibration and the characteristic peak of CC-90011 at approximately 2222 cm⁻¹ for the -C ≡ N vibration (Fig. S3E) [38]. The UV-Vis absorption spectrum of the CC-linker exhibited a characteristic absorption peak at 219 nm, consistent with CC-90011 (Fig. S4A) and HPLC determined the LE of CC-90011 in the CC-linker conjugate to be ∼34% (Fig. S4B–D).

To generate CEC NPs, the CC-linker was integrated into the membrane of EC NPs via co-incubation, leveraging the insertion of DSPE moieties into the exosomal lipid bilayer. The results of FCM indicated FITC-labeled CC-linker was efficiently inserted into EC. After incubation with 50 μM and 100 μM CC-linker-FITC at 37 °C for 2 h, 62.0% and 62.2% of EC were FITC-positive, respectively (Fig. 1B). The images of Confocal laser scanning microscopy (CLSM) showed the co-localization of Dil-labeled EC with FITC-labeled CC-linker, further confirming that CC-90011 was successfully loaded onto EC (Fig. 1C). Notably, the CEC maintained its exosomal morphology (Fig. 1D), the expression of exosomal surface markers (Fig. 1E), zeta potential (Fig. 1F) and particle size (Fig. 1G), indicating that the structural and surface integrity of the exosomes remained intact during the CEC synthesis. The larger particle size observed in CLSM compared to TEM is attributed to differences in imaging conditions, with CLSM reflecting hydrated states with fluorescence diffusion and diffraction effects, and TEM reflecting dehydrated structures; the DLS result (∼100 nm) is consistent with the TEM measurement [39,40]. The UV-Vis absorption spectra of CEC exhibited characteristic absorption peaks attributable to EC and the CC-linker, further validating the successful synthesis of CEC (Fig. 1H).

Additionally, our nanoparticles also exhibit excellent stability in PBS, with no significant change in particle size over 7 days (Fig. 1I). To further evaluate the stability of the nanoformulation under physiologically relevant conditions, CEC were incubated in serum-containing medium. As shown in Fig. S5, the hydrodynamic size of CEC remained relatively stable over time, with no significant aggregation or disassembly observed. The hemolysis assay further demonstrated that, compared with free drugs, CEC markedly reduced hemolytic activity, thereby significantly enhancing biosafety (Fig.S6A and B). Finally, UV-Vis measurements were used to evaluate the release profile of CC-90011 in solutions with different pH conditions. The results demonstrated a pH-responsive release profile, with markedly accelerated drug release under acidic conditions that mimic the inflammatory microenvironment of RA (Fig. 1J).

In summary, we successfully synthesized CEC NPs via M2 Exos as biocompatible carriers for the co-delivery of CXB and CC-90011. This exosome-based delivery system significantly improved the bioavailability of both lipophilic agents, thereby providing a robust foundation for subsequent cellular and in vivo investigation.

2.2. Cell-type-specific targeting of CEC components

The CEC NPs were cleaved into EC and CC-90011 components under acidic conditions. To determine whether EC retains the inflammation-targeting capacity of M2 Exos, we assessed its uptake by non-inflammatory cells (RAW 264.7; FLS) and inflammatory cells (M1; RA-FLS) at various time points, respectively. The cellular endocytosis of Dil-labeled EC was observed using CLSM. The results demonstrated a time-dependent accumulation of EC in all cell types. Notably, LPS-activated macrophages (Fig. 2A and C) and FLSs (Fig. 2B and D) exhibited markedly stronger EC-derived fluorescence compared with their non-activated counterparts, indicating enhanced EC accumulation in inflammatory cells. These findings were further confirmed by FCM, which showed consistent trends (Fig. 2E and F). Combined with our previous data on the effects of standalone M2 Exo treatment in LPS-treated macrophages and FLSs (Fig. S7A–D), all these results confirm that M2 Exos maintain their intrinsic inflammation-targeting properties even after drug encapsulation [32].

Fig. 2.

The targeting profiles of EC NPs and CC-90011. (A-B) Representative CLSM images showing uptake of Dil-labeled EC NPs (red) by RAW 264.7 (A) and FLS (B) cells, with or without LPS stimulation, at indicated time points. Scale bar = 25 μm. (C-D) Quantification of EC NPs intensity in RAW 264.7 (C) and FLS (D) cells corresponding to panels A and B (n = 3). (E-F) FCM analysis of Dio-labeled EC NPs (green) in RAW 264.7 (E) and FLS (F) cells with or without LPS at indicated time points. (G-I) The targeting ability of EC NPs and CC-90011 towards inflammatory cells and OCs, respectively. ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. All the representative images are shown from three independent experiments.

To further investigate whether these released components exhibit cell-type-specific targeting (EC toward inflammatory cells; CC-90011 toward OCs), we conducted the following experiments. Acid-pretreated CEC were introduced into a co-culture system of inflammatory cells (M1 macrophages and RA-FLS) and OCs, in which the exosomes were Dil-labeled and CC-90011 was FITC-labeled (Fig. S7E). CLSM revealed that EC were predominantly internalized by inflammatory cells, not the OCs, while CC-90011 was mainly engulfed by OCs (Fig. 2G–I). This cell-type-specific uptake pattern supports the design of our dual-target strategy. These results confirm the selective targeting of the CEC components following acid-triggered release. Based on the cell-specific targeting profiles, subsequent functional assays were designed to evaluate the anti-inflammatory effects of EC on inflammatory cells and the anti-resorptive effects of CC-90011 on OCs, respectively.

2.3. EC promotion of M1-to-M2 repolarization for anti-inflammatory effects

Before determining the effective concentration of CXB in macrophages in vitro, we first evaluated its cytotoxicity using a CCK-8 assay. The results showed that CXB exhibited negligible cytotoxic effects on macrophages at concentrations up to 16 μg/mL (Fig. S8A). We then investigated the repolarization and anti-inflammatory effects of CXB on M1 via RT-qPCR (Fig. S8B–F) and Western blot (Fig. S8G–I) analysis. Both sets of results revealed that CXB significantly downregulated the expression of M1 markers (iNOS, IL-6, and IL-1β) while simultaneously increasing the expression of M2 markers (Arg-1 and IL-10) in a concentration-dependent manner even at 2 μg/mL.

Macrophage polarization plays a crucial role in regulating inflammatory responses during the progression of RA [41]. It is well established that during the pathogenesis of RA, the balance between M1 and M2 macrophages is severely disrupted, with M1 macrophages playing a pivotal role in disease initiation and progression [42]. In contrast, M2 macrophages exhibit anti-inflammatory and tissue-repairing properties, making the modulation of macrophage polarization toward the M2 phenotype a promising therapeutic strategy for controlling chronic inflammation [43]. Therefore, it is particularly important to explore whether EC can induce M1-to-M2 repolarization and exert anti-inflammatory effects by evaluating the expression levels of M1 and M2 markers under different treatment conditions. As shown in the CLSM images, Exo treatment caused limited repolarization of M1 macrophages with a slight decrease in iNOS fluorescence intensity and increase in Arg-1. In contrast, CXB treatment resulted in a more pronounced promotion of the M1 to M2 polarization. Notably, EC treatment resulted in the most marked repolarization of M1 into M2 compared with other groups (Fig. 3A and B). We further assessed the changes in mRNA expression levels of M1 markers (Inos, Il1b) and M2 markers (Arg1, Il10) via RT-qPCR. Compared to the LPS-induced group, CXB significantly decreased the mRNA expression of iNOS and IL-1β, while simultaneously increasing that of Arg-1 and IL-10. Although the differences were not statistically significant compared with the LPS-induced group, Exos still exhibited a discernible anti-inflammatory effect. The mRNA expressions of iNOS and IL-1β in the EC group decreased by ∼76.69% and ∼81.83%, respectively, while the expressions of Arg-1 and IL-10 significantly increased by 2.97-fold and 3.55-fold, compared to the LPS-induced group (Fig. 3C). FCM analysis showed that the proportion of CD206⁺ macrophages was approximately 11.1% in LPS-treated cells, 11.7% in the M2 Exos group, 18.1% in the CXB group, and 30.7% in the EC group. The CD206⁺/CD86⁺ ratio exhibited a consistent trend, indicating that EC most effectively promoted M2 polarization (Fig. 3D and E). In addition, ELISA analysis revealed that EC most strongly suppressed IL-6 and increased IL-10 compared with CXB or M2 Exos (Fig. 3F). Western blot further confirmed that EC upregulated CD206 while downregulating iNOS, in agreement with the other results (Fig. S9A). Taken together, these results demonstrate that CXB exerts anti-inflammatory effects, which is consistent with previous reports [44], while EC at the same concentration exhibits enhanced anti-inflammatory properties by efficiently repolarizing pro-inflammatory M1 macrophages toward the M2 phenotype.

Fig. 3.

M1-to-M2 repolarization by EC NPs in vitro. (A-B) Representative CLSM images of the polarization markers of macrophages iNOS (green) and Arg-1 (red) after treatment with CXB, Exo, and EC NPs. DAPI: blue. Scale bar = 25 μm. (C-D) RT-qPCR (C) and FCM (D) verification of the expression of Arg-1, iNOS, CD86, CD206, IL-1β, IL-6, and IL-10 in macrophages after different treatment. (E) Quantification of the CD206⁺/CD86⁺ ratio in RAW 264.7 macrophages corresponding to the FCM data in (D). (F) Concentrations of IL-6 and IL-10 in cell culture supernatants were measured by ELISA. (G-J) The representative CLSM images (G) and FCM (J) results of ROS (green) generation in M1 macrophages treated with different conditions (CXB, Exo, EC NPs) detected using the DCFH-DA probe. Quantitative analysis of the fluorescence intensity of DCFH-DA in the CLSM images (H) and in FCM (J), corresponding to (G) and (I), respectively. n = 3, Scale bar = 25 μm ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns: no significance. All the representative images are shown from three independent experiments.

Reactive oxygen species (ROS) play a crucial role in the pathogenesis of RA [45,46]. Excessive ROS not only induce synovial inflammation and prompt the polarization of macrophages into M1, but also promote the generation of OCs, ultimately contributing to bone erosion [47]. Previous studies have demonstrated that reducing ROS levels facilitates the repolarization of M1 to M2, and suppresses the inflammatory response [48,49]. Therefore, efficient ROS scavenging is crucial for RA treatment. To evaluate the ROS-scavenging capability of EC, the intracellular ROS levels in M1 following EC treatment were examined using 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) [50]. CLSM imaging showed that CXB alone moderately reduced intracellular ROS, whereas EC achieved a substantially stronger decrease than free CXB and M2 Exos (Fig. 3G and H). This result was further supported by FCM analysis (Fig. 3I and J). Moreover, it is well-documented that nitric oxide (NO) and ROS exhibit a complementary interplay under inflammatory conditions, wherein elevated NO levels can further promote ROS production [51]. Compared with RAW264.7 cells, LPS-induced M1 macrophages exhibited significantly elevated NO production (Fig. S9B). NO production was measured in macrophages under different treatments (Fig. S9C). The results revealed that NO levels in M1 were significantly reduced after EC treatment compared with those in the CXB and M2 Exos groups. Collectively, these findings suggest that EC markedly reduces intracellular ROS and NO levels, thereby creating a microenvironment favorable for M2 polarization and enhancing its overall anti-inflammatory effect.

In summary, our findings demonstrate that M2 Exo-based EC NPs substantially promote M1 repolarization toward M2 to achieve potent anti-inflammatory effects. Notably, the role of the exosome carrier itself was also investigated. Although Exos exhibited modest anti-inflammatory activity in M1, they were insufficient to effectively induce M1 repolarization toward the M2 phenotype. These observations indicate that the enhanced immunomodulatory effects are mainly attributed to the drug-loaded exosome formulation. Our NPs demonstrate improved therapeutic performance relative to previously reported CXB delivery strategies, which can be ascribed to the following advantages [52]. The chemokine receptors (CCR2/CX3CR1) and CD163 on the surface of M2 Exos facilitate active targeting of inflammatory sites, resulting in enhanced CXB accumulation in inflamed tissues [30]. More importantly, the combined action of CXB and the inherent anti-inflammatory mediators in M2 Exos achieve an ∼80% reduction in key inflammatory factors such as IL-1β, thereby resulting in substantially enhanced therapeutic efficacy relative to free CXB or liposome-loaded CXB [53].

2.4. EC inhibition of FLS inflammation and migration

In addition to macrophages, FLSs also play a crucial role in the pathogenesis of RA [54]. Upon stimulation by inflammatory mediators, FLSs acquire an activation phenotype characterized by enhanced migration, secretion of pro-inflammatory cytokines, ultimately contributing to cartilage degradation and bone erosion [55]. Therefore, it is critical to evaluate the therapeutic effects of EC on FLSs. After isolating mouse primary FLSs [56], their identity was confirmed via CLSM for the synovial cell-specific protein vimentin [57]. The results revealed that the majority of the cells expressed vimentin and displayed typical spindle morphology with centrally located oval nuclei (Fig. S10A). To establish an inflammatory model of mouse FLSs (RA-FLS), we stimulated FLSs with 1 μg/mL of LPS. RT-qPCR analysis showed that the mRNA levels of Il1b and Tnf increased as expected, peaking at 3 h post stimulation, indicating successful induction of the inflammatory phenotype (Fig. S10B). Subsequently, cell viability assays demonstrated that CXB showed no cytotoxicity toward RA-FLS within the tested concentration range (Fig. S11A). Furthermore, CXB exerted a concentration-dependent inhibitory effect on pro-inflammatory cytokine (Il1b and Tnf) expression in FLSs, as assessed by RT-qPCR (Fig. S11B). Moreover, in contrast to previous studies, our work provided more comprehensive evidence demonstrating the inhibitory effect of CXB on the expression of inflammatory factors in FLSs, beyond its documented impact on IL-6 [58].

We next evaluated the therapeutic effects of EC on RA-FLS. Given that activated FLSs drive cartilage invasion and bone destruction through cell migration [59], we firstly investigated whether EC could impair their migratory capacity. The results of the scratch wound assay revealed that CXB inhibited cell migration significantly, but EC exerted a more pronounced inhibitory effect on the migration of RA-FLS compared to CXB or M2 Exos alone, reducing the cell migration rate to ∼10% after 24 h, indicating a strong suppressive effect on RA-FLS motility (Fig. 4A and B). Furthermore, the activation-related marker (collagen I) expression in RA-FLS was significantly downregulated after all the treatments. Both CXB and M2 Exos exhibited inhibitory effects, whereas EC treatment led to the most pronounced suppression (Fig. 4C, Fig. S12). The anti-inflammatory effects of EC on RA-FLS were further validated by RT-qPCR and ELISA. The RT-qPCR results demonstrated that Exo showed limited effects on downregulating the mRNA expression of Il1b and Il6, while CXB exhibited a stronger inhibitory effect. Notably, EC most significantly suppressed the gene expression of both inflammatory cytokines (Fig. 4D and E). These findings were corroborated at the protein level by ELISA (Fig. 4F and G), which showed that EC markedly suppressed the secretion of the pro-inflammatory cytokine IL-1β and enhanced the production of the anti-inflammatory cytokine IL-10 compared to the other groups.

Fig. 4.

EC NP-mediated suppression of RA-FLS activation. (A) Representative images from scratch assay images of RA-FLS treated with LPS, CXB, Exo, and EC at 0 h and 24 h. (B) Quantification of cell migration in scratch assay. (C) Western blot analysis of activation markers (collagen I) in RA-FLS after indicated treatments. (D-E) Relative Il1b (D) and Il6 (E) mRNA expression in RA-FLS across different treatment groups as detected by RT-qPCR. (F-G) Quantification of IL-1β (F) and IL-10 (G) concentrations in the supernatant of RA-FLS after indicated treatments by ELISA. n = 3; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns: no significance. All the representative images are shown from three independent experiments.

In summary, these results demonstrate that EC not only inhibits the migratory capacity of RA-FLS but also suppresses their production of inflammatory cytokines, highlighting its therapeutic potential in mitigating joint inflammation and preventing cartilage invasion in RA. In contrast to previous studies on the effects of CXB on FLSs for RA treatment, which primarily focused on the induction of FLS apoptosis by high concentrations of CXB, our study demonstrates for the first time that CXB at non-cytotoxic concentrations effectively inhibits both cell migration and the secretion of pro-inflammatory factors in RA-FLS [60]. This suppression may further contribute to reducing immune cell infiltration and attenuating RA-FLS-mediated invasion. Moreover, given the pharmacokinetic properties of CXB, achieving high drug concentrations locally at the lesion site remains challenging. In this context, our NPs enable anti-inflammatory and anti-migratory effects at lower, clinically relevant doses of CXB, highlighting a more practical approach for FLS-driven pathology.

2.5. Inhibitory effects of CEC on osteoclastogenesis

Previous studies have established that in RA, OCs are aberrantly differentiated and exhibit excessive proliferation, thereby disrupting bone homeostasis and leading to progressive bone erosion and joint damage [61]. LSD1 is a pivotal regulator of glycolytic metabolism during osteoclastogenesis, and its inhibition has been shown to attenuate bone resorption in RA [27,62]. CC-90011, a newly identified inhibitor of LSD1, has demonstrated potent anti-tumor activity in several malignancies [63]. However, its therapeutic potential in RA and the underlying mechanisms remain unexplored. In this study, we are the first to explore the inhibitory effects of CC-90011 on osteoclastogenesis in the setting of RA. We employed an acid-responsive CEC NP which enables site-specific release of CC-90011 within the acidic milieu of inflamed RA joints [64]. Firstly, we investigated the effects of CC-90011 on osteoclastogenesis. As shown in Fig. 5A, bone marrow macrophages (BMMs) were isolated from mice and cultured in a medium containing 30 ng/mL M-CSF and 50 ng/mL RANKL to induce OCs (Fig. S13A). Cell viability was assessed via CCK-8 assay, confirming that CC-90011 treatment was non-cytotoxic to BMMs at concentrations below 8 μM for 24 or 48 h (Fig. S13B and C). Then we added different concentrations of CC-90011 to investigate its effect on OC formation. Following treatment with varying concentrations of CC-90011, a marked reduction in the F-actin ring area, a key morphological indicator of OCs, was observed (Fig. 5B, Fig. S14A). In addition, tartrate-resistant acid phosphatase (TRAP) staining results demonstrated that treatment with 1 μM CC-90011 markedly reduced both the number and size of OCs (Fig. 5C, Fig. S14B). These findings collectively indicate that CC-90011 exerts a potent inhibitory effect on osteoclastogenesis in a dose-dependent manner. c-Src, a central signaling hub in the OC signaling network, is commonly used as a molecular indicator of osteoclast formation [65]. As anticipated, RT-qPCR analysis demonstrated that CC-90011 significantly suppressed the mRNA expression of both c-Src and LSD1, further supporting its inhibitory effect on osteoclastogenesis (Fig. 5D). As NFATc1 is a master transcription factor required for OC differentiation, and previous studies have demonstrated that LSD1 inhibitors can downregulate the protein expression of osteoclast-specific markers, such as NFATc1 during osteoclastogenesis [62] [66]. We further examined the impact of CEC on LSD1 and NFATc1 expression. Western blot (Fig. 5E and F) revealed a concentration-dependent downregulation at protein levels of NFATc1 upon CC-90011 treatment, accompanied by a parallel reduction in LSD1 expression. These data indicate that CC-90011 attenuates osteoclastogenesis by modulating the LSD1-NFATc1 signaling axis, highlighting the therapeutic potential of CC-90011 as an anti-resorptive agent in RA.

Fig. 5.

Inhibitory effects of CEC NPs on osteoclastogenesis. (A) Schematic diagram of BMM isolation and OC differentiation. (B) Representative images of CLSM showing F-actin ring formation (red) and nuclei (blue) in OCs under CC-90011 (0, 1, 2, 4, 8 μM) treatment. Scale bar = 100 μm. (C) TRAP staining of multinucleated OCs (≥3 nuclei, purple) under CC-90011 treatment. Scale bar = 200 μm. (D) RT-qPCR analysis of mRNA expression levels of Lsd1 and osteoclast-specific genes Csrc. (E) Western blot analysis of NFATc1 and LSD1 protein expression (β-actin as loading control). (F) Densitometric quantification of LSD1 and NFATc1 protein levels. (G-H) Representative images of F-actin ring (G, Scale bar = 250 μm) and TRAP (H, Scale bar = 500 μm) staining of OCs after treated with CC-90011, EC NPs, and CEC NPs. (I) RT-qPCR analysis of Lsd1 and Nfatc1 mRNA expression levels in OCs following treatment with CC-90011, EC NPs, and CEC NPs. n = 3; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns: no significance. All the representative images are shown from three independent experiments.

Having confirmed the inhibitory effect of CC-90011 on OC formation, we further investigated the effects of CEC NPs on osteoclastogenesis. Based on the acidic inflammatory microenvironment characteristic of rheumatoid arthritis, we pre-incubated CEC NPs in an acidic PBS (pH 6.5) to simulate the inflammatory microenvironment for CC-90011 release and directly applied them to OCs. To exclude the potential influence of EC on OC formation, an EC treatment group was still included as a control. F-actin ring staining and TRAP assays revealed obvious morphological alterations and size reduction in OCs with CC-90011 (G3), EC (G4), and CEC (G5) treatments compared to the untreated OC group (Fig. 5G and H, Fig. S14C and D) and the inhibitory effects of CC-90011 and CEC are much more pronounced than EC, which indicates the potent inhibitory effect of CC-90011 on osteoclastogenesis. RT-qPCR results showed that the CEC group significantly suppressed the LSD1-NFATc1 signaling pathway, consistent with the CC-90011 treatment, thereby corroborating the above assay results (Fig. 5I). It should be noted that the EC group also exhibited partial inhibitory effects, which may be attributed to the interplay between bone destruction and inflammation, as suppression of the inflammatory response can indirectly alleviate bone erosion [11,67]. Notably, our earlier data presented above demonstrated EC's anti-inflammatory therapeutic efficacy. Moreover, the overall therapeutic effect of CEC was superior to that of CC-90011, which may be attributed to the additional anti-inflammatory activity contributed by EC. Therefore, the therapeutic agent developed in this study may exert its anti-resorptive effects through both direct inhibition of OC formation and indirect modulation of the inflammatory microenvironment.

Collectively, these results indicate that CEC NPs not only exert potent anti-inflammatory effects but also robustly suppress OC formation by directly blocking the LSD1 signaling pathway. This dual functionality highlights the therapeutic potential of CEC for RA treatment and provides a strong rationale for subsequent in vivo investigations. LSD1 has been implicated in the regulation of glycolytic metabolism, which is closely associated with inflammatory activation in RA. In the present study, CC-90011 showed limited effects on macrophages and RA-FLS, likely due to its low cellular uptake in these cells, suggesting that its primary therapeutic role is related to osteoclastogenesis. Future studies incorporating direct metabolic assays will be required to further elucidate the role of glycolytic regulation in this system.

2.6. Anti-inflammatory and anti-bone erosive effects of CEC NPs in CIA model

After confirming the potent anti-inflammatory and anti-erosion effects of CEC NPs in vitro, we further evaluated their therapeutic efficacy in vivo using a collagen-induced arthritis (CIA) murine model, which recapitulates key pathological features of human RA [68]. To assess the in vivo biodistribution and inflammation-targeting capability of CEC NPs, CY5.5-labeled CEC NPs were intravenously administered to both healthy and arthritic mice. Fluorescence imaging at 0, 2, 4, 8, 12, and 24 h post injection revealed a time-dependent accumulation of CEC NPs at the inflamed joints of CIA mice, whereas negligible fluorescence signals were observed in the joints of healthy controls. These results confirmed that CEC NPs retained their inflammation-targeting ability in vivo, which is primarily mediated by M2 macrophage-derived exosomal membranes (Fig. S15).

To evaluate the therapeutic efficacy of CEC, CIA mice were randomly divided into five groups and administered with PBS, free CXB, free CC-90011, EC NPs, or CEC NPs via tail vein injection every three days, starting from day 7 after booster immunization (Fig. 6A). In the in vitro studies, the anti-inflammatory effect of M2 Exos was much weaker than EC. Moreover, CC-90011 is released from under the inflammatory microenvironment of RA. Based on this design rationale, M2 Exo and Exo-CC-90011 control groups were not included in the in vivo experiments. Healthy mice served as the control group. Arthritis severity was evaluated using the arthritic index (AI) scoring based on swelling and erythema in the ankle and foot joint [69]. Compared to the PBS-treated group, all treatment groups showed gradual improvements in paw thickness and arthritis scores over time. Among them, mice treated with CC-90011 exhibited comparable efficacy to the CXB group, indicating that CC-90011 alone could achieve a substantial therapeutic effect during treatment. EC treatment demonstrated stronger efficacy than both CC-90011 and CXB. Notably, CEC NPs showed near-complete recovery by day 18, with AI scores returning to baseline levels comparable to those of the controls (Fig. 6B and C). In addition, body weight in the CEC NP-treated group exhibited a steady increase, comparable to that of healthy controls, suggesting favorable systemic tolerance and improvement in overall disease status (Fig. 6D). Thermal imaging conducted on day 18 further indicated pronounced inflammation in the PBS group, characterized by swelling and localized hyperthermia. CXB monotherapy yielded marked mitigation of joint swelling and hyperthermia, demonstrating the potent anti-inflammatory activity of CXB in vivo. Exosomal encapsulation of CXB (EC) further enhanced this effect, due to improved drug targeting and bioavailability [70]. The inflammation in the CEC group recovered to the level comparable to that of normal mice (Fig. 6E and F). In contrast, CC-90011 showed limited efficacy in alleviating soft tissue inflammation, which is in line with its predominant role in suppressing OC formation rather than modulating inflammatory responses. Accordingly, bone preservation in the ankle joints was further assessed using micro-computed tomography (micro-CT). In the pathological progression of arthritis, the formation of osteophytes represents a characteristic ectopic ossification phenomenon at joint margins, whose development is associated with the chronic inflammatory microenvironment [71]. Micro-CT analysis at day 18 demonstrated that CXB, CC-90011, EC NPs, and CEC NPs alleviated joint pathology in CIA mice to varying extents (Fig. 6G). With respect to inflammation-associated osteophyte formation, CXB markedly attenuated osteophyte development, whereas CC-90011 monotherapy exhibited only limited efficacy. EC NPs further enhanced its anti-inflammatory activity, leading to improved suppression of osteophyte formation. Notably, CEC NPs achieved the most pronounced inhibition of osteophyte formation among all treatment groups. Regarding bone erosion, CC-90011 monotherapy significantly inhibited bone loss, consistent with its anti-osteoclastogenic mechanism. EC NPs conferred partial protection against bone erosion, while CEC NPs further amplified this effect, resulting in the most substantial preservation of bone architecture compared with all other groups.

Fig. 6.

Anti-inflammatory and Anti-Bone Erosive effects of CEC NPs in vivo. (A) Schematic diagram of CIA mouse model establishment and treatment. (B-D) Effects of different treatments on paw thicknesses (B), AI scores (C) and body weights (D) in CIA mice. n = 5. (E-F) Paw swelling and temperature changes in CIA mice after indicated treatments. Scale bar = 1 cm. (G) Micro-CT evaluation of bone erosion (yellow) and osteophyte formation (red) in ankle joints after indicated treatments. All images are representative of five mice per group.

Collectively, these findings suggest that CEC NPs confer superior joint protection through targeted delivery and multi-mechanistic therapeutic actions. Importantly, for the first time, CC-90011 was shown to exhibit significantly greater efficacy in mitigating bone erosion compared with CXB, highlighting its distinct therapeutic potential, particularly in late-stage RA characterized by severe bone destruction.

2.7. Histopathological evaluation of therapeutic efficacy

To further evaluate the histological outcomes of various treatment regimens in alleviating rheumatoid arthritis, we conducted a comprehensive pathological analysis of the ankle joints and spleen tissues in mice. Hematoxylin and eosin (H&E), tartrate-resistant acid phosphatase (TRAP) staining, and safranine O/Fast green cartilage staining (SOFG) were employed to assess synovial inflammation, OC formation, and cartilage degradation, respectively (Fig. 7A–C). Mice in the PBS group displayed pronounced synovial hyperplasia, severe cartilage degradation, and excessive OC formation. Treatment with CXB alone reduced inflammatory cell infiltration and synovial invasion, while slightly reduced OC formation and increased chondrocyte expression (orange staining). When CXB was delivered using EC NPs, its therapeutic efficacy was significantly enhanced due to the inflammatory targeting ability and anti-inflammatory factors of the carrier, but showed minimal improvement in bone destruction. Treatment with CC-90011 alone still resulted in mild synovial invasion, but markedly reduced OCs formation and increased chondrocyte expression. In contrast, CEC therapy provided remarkable protection of joint integrity comparable to controls, concurrently reducing inflammation and structural damage, underscoring its superior efficacy.

Fig. 7.

Histological analysis of therapeutic effects following CEC administration in vivo. (A) Histopathological evaluation of ankle joints in each group was performed by H&E staining. Scale bar = 200 μm. (B) Histochemical identification of OCs by TRAP staining in arthritic joints from different treatment groups. Scale bar = 200 μm. (C) Cartilage damage in murine ankle joints following different treatments was assessed by SOFG staining. Scale bar = 50 μm. (D) Immunofluorescence staining was performed to detect the expression of macrophage marker F4/80 (purple), M1 macrophage marker (iNOS, red) and M2 macrophage marker (Arg-1, green) in joint tissues of healthy mice and CIA mice receiving different treatments, scale bar = 100 μm. All images are representative of five mice per group.

To investigate the immunoregulatory effects, we examined macrophage polarization in the synovium via immunofluorescence staining (Fig. 7D). A high density of pro-inflammatory M1 macrophages (F4/80⁺/iNOS⁺) was observed in the CIA group. Following CXB treatment, inflammatory infiltration was significantly attenuated, with reduced M1 macrophage abundance and increased M2 (F4/80⁺/Arg-1⁺) macrophages, proving that CXB effectively repolarized the phenotype of macrophages in vivo. In contrast, CC-90011's inhibitory effect on M1 polarization was moderate. Whereas treatment with EC and CEC NPs led to a substantial reduction of M1 macrophages accompanied by a concomitant increase in anti-inflammatory M2 macrophages (F4/80⁺/Arg-1⁺), indicating a significant reversal of the M1/M2 polarization ratio and reprogramming of the immune microenvironment in vivo.

Given that rheumatoid arthritis is often accompanied by systemic immune dysregulation, we further assessed splenic histopathology to evaluate systemic immune involvement [72,73]. Firstly, from the size of the spleen, it was observed that compared to the healthy group, all other groups exhibited varying degrees of swelling, while the swelling in the mice treated with CEC was the least tumid among all the groups (Fig. S16A). In the histological analysis of splenic H&E staining, the PBS group showed significant tissue damage and immune cell infiltration, confirming the successful establishment of the CIA model. In the CXB group, partial mitigation of inflammation and modest white pulp restoration were observed. The EC group showed superior improved tissue recovery relative to CXB, suggesting carrier-mediated synergism. The CC-90011 group showed limited improvement in spleen, suggesting that CC-90011 may possess insufficient therapeutic efficacy against inflammatory processes. Remarkably, the CEC group exhibited a similar therapeutic effect to the EC group, and the staining results revealed nearly normal splenic architecture alongside a significant reduction in immune cell infiltration (Fig. S16B). Collectively, CEC contributed to the maintenance of immune homeostasis, demonstrating strong therapeutic potential and clinical translational value for the treatment of rheumatoid arthritis.

Finally, histological examination of major organs (hearts, livers, lungs, and kidneys) revealed no significant tissue damage or inflammatory infiltration in any treatment group (Fig. S17A), further supporting the favorable biocompatibility and in vivo safety of the CEC NPs. To further evaluate the systemic biosafety of the nanoplatform, serum biochemical parameters were analyzed. As shown in Fig. S17, liver function markers (ALT and AST, Fig. S17B) and the renal function indicator (BUN, Fig. S17C) showed no significant differences among the treatment groups compared with the control, indicating no evident systemic toxicity under the current experimental conditions.

Notably, a direct comparison with the co-administration of free CXB and CC-90011 was not included in this study. While exosome-based delivery improved the therapeutic efficacy of CXB, it is quite reasonable that CEC exhibit better therapeutic outcome than the free drug mixture. This limitation should be addressed in future studies to more comprehensively evaluate the advantage of the nanoplatform.

3. Conclusion

This study presents a novel therapeutic strategy to simultaneously address the dual pathological processes of chronic inflammation and bone erosion in RA. We developed biomimetic nanoparticles (CEC NPs) that innovatively integrate the anti-inflammatory agent CXB and LSD1 inhibitor CC-90011 with M2 macrophage-derived exosomes for targeted and combinational therapy of RA, by harnessing the inherent targeting and immunomodulatory properties of M2 Exos. CEC NPs not only retain potent anti-inflammatory effects, but also effectively suppress osteoclast formation through the LSD1-NFATC1 signaling axis. Both in vitro and in vivo studies consistently demonstrate superior therapeutic outcomes over monotherapies, including effective macrophage repolarization toward M2, inhibition of FLS activation, and substantial protection against cartilage and bone destruction. Moreover, comprehensive biosafety assessments confirmed the excellent biocompatibility of this nanoplatform. Collectively, this study provides a versatile therapeutic strategy for precise and combinational RA therapy by combining small-molecule inhibitors with M2 Exo-based nanocarriers, laying a solid foundation for clinical translation in the management of RA and potential application in other autoimmune and inflammatory bone disorders.

4. Experimental section

4.1. Cell culture

The RAW264.7 cells were purchased from Haixing Biosciences and cultured in RAW 264.7-specific medium (CM-0190, Procell, China) under conditions of 37 °C and 5% CO₂. Bone marrow macrophages (BMMs) and fibroblast-like synoviocytes were isolated from 6- to 8-week-old C57BL/6J mice and cultured in α-MEM medium (C12571500BT, Gibco, USA). The medium was supplemented with 10% fetal bovine serum (C0234, Beyotime, China) and 1% Penicillin-Streptomycin Solution (PSS, C0222, Beyotime, China).

4.2. Preparation of the M2 Exos

To obtain M2 Exos, RAW264.7 cells were seeded at a density of 1.0 × 10⁶ cells per well in a six-well plate. The cells were cultured in serum-free medium supplemented with 20 ng/mL IL-4 (PRP1137, Abbkine, China) for 24 h. Afterward, the culture supernatant was collected and subjected to centrifugation: firstly at 300 g for 15 min to remove cell debris and large particles, followed by centrifugation at 10,000 g for 15 min to remove dead cells. Finally, exosomes were enriched at 100,000 g under high-speed conditions for 120 min. The exosomes were then resuspended in Phosphate Buffered Saline (PBS, G4207, Servicebio, China) for further experimental use.

4.3. Western blot

After differential treatment, cells were lysed using Western and IP lysis buffer (P0013, Beyotime, China). Total protein concentration was determined using a BCA Protein Assay Kit (979-88-4, GlpBio, USA). Proteins were denatured in sample loading buffer at 95 °C for 10 min, separated by electrophoresis at 80 V for 100 min, and transferred onto a 0.45 μm PVDF membrane (IPVH00010, Merck Millipore, Germany). The membrane was blocked with 5% BSA for 2 h at room temperature and then incubated with primary antibody overnight at 4 °C. After washing with TBST, the membrane was probed with an HRP-conjugated secondary antibody for 1 h at room temperature. Protein bands were visualized using a chemiluminescent substrate and quantified with ImageJ software.

4.4. Preparation and characterization of CC-linker

After dissolving 100 μL (200 μg) of CC-90011 (T11882, TargetMol, USA) in 1000 μL of sodium bicarbonate buffer (NaHCO₃, pH 8.5), 100 μg of DSPE-Hyd-PEG₂₀₀₀-NHS (Yusi Pharmaceutical Technology Co, China) was added and thoroughly mixed. The reaction proceeded at room temperature for 24 h. The resulting mixture was then dialyzed using a membrane with a 500 Da molecular weight cutoff (YA1069, Solarbio, China) to remove unreacted CC-90011 and other small-molecule impurities. Finally, CC-90011, DSPE-Hyd-PEG₂₀₀₀-NHS, and the conjugate DSPE-Hyd-PEG₂₀₀₀-CC-90011 (CC-linker) were characterized by Raman spectroscopy (XploRA PLUS, HORIBA, Japan) and ¹H NMR spectroscopy (600 MHz, JNM-ECZ600R, JEOL, Japan).

4.5. Synthesis and Characterization of Exo@CXB

After mixing 100 μg of CXB (C129279, Aladdin, China) with 50 μg of M2 Exos, CXB-encapsulated M2 Exos (Exo@CXB, EC) were prepared by extrusion using a liposome extruder (Avanti® 610000, USA). The mixture was then centrifuged with a 50 kDa ultrafiltration membrane (Amicon Ultra-0.5, Millipore Co, USA) at 14,000×g for 10 min to remove unencapsulated CXB. The encapsulation efficiency and drug loading efficiency of CXB were subsequently quantified by high-performance liquid chromatography (HPLC).

4.6. Synthesis and Characterization of CC-Exo@CXB

The purified EC and CC-linker were dissolved in PBS solution in a mass ratio of 1:2, and subsequently underwent a reaction at room temperature for 2 h. After the reaction was completed, the free drugs were removed by a 50 kDa ultrafiltration tube to obtain the CC-90011-modified exosomes (CC-Exo@CXB), namely CEC. The successful synthesis of CEC was verified by UV–Vis spectroscopy (UV-2600, Techcomp, China) and HPLC.

4.7. Characterization of different exosomes

Using transmission electron microscopy (JEM-1200EX, JEOL, Japan), the morphology of different exosome structures was observed. The sizes and zeta potentials of these structures were measured using dynamic light scattering (Nano ZS90, Malvern Panalytical, England). Protein expression levels of CD81 (1:1000, 66866-1-Ig, Proteintech, China), TSG101 (1:1000, 28283-1-AP, Proteintech, China), Calnexin (1:1000, 10427-2-AP, Proteintech, China), CD206 (1:1000, 83485-1-RR, Proteintech, China), CD86 (1:1000, 30691-1-AP, Proteintech, China) and Arg-1 (1:1000, A4923, Abclonal, China) in M2 Exo, EC, and CEC were assessed using Western blot.

4.8. Hemolysis test

1 ml of blood was collected from C57/BL6 mice using cardiac blood sampling, and serum was removed by centrifugation three times at 500 g. Subsequently, the red blood cells were dispersed in PBS to prepare a 2% erythrocyte suspension, and 1 ml was subsequently added to the different nanoparticles. After 2 h of incubation at room temperature, the supernatant was removed by centrifugation at 500g for 5 min, and the nanoparticles were subsequently removed by centrifugation at 12,000 g, and the absorbance at 540 nm was measured using a microplate reader.

4.9. Drug acid-responsive release experiment in vitro

Equal amounts of CEC were dissolved in 200 mL of sodium acetate solution (pH 6.5) and PBS solution (pH 7.4), respectively. The solutions were then dialyzed using a 3000 Da dialysis bag to release the acid-responsive CC-90011, and the absorbance of CC-90011 at 219 nm in the solution at different time points was measured using UV-Vis spectroscopy.

4.10. Cell viability

For macrophages and synovial fibroblasts, RAW264.7 and FLS cells were inoculated on 96-well plates at 1 × 10⁴ and then stimulated with 1 μg/mL LPS after adhesion, the cells were treated with CXB (0, 2, 4, 8, 16 μg/mL), M2 Exo and EC at different concentration gradients for 24 h. For osteoclasts, BMMs with a density of 8 × 10³ per well were inoculated into two 96-well plates, and cultured for 1 day, then treated with 30 ng/mL M-CSF and CC-90011 (0, 1, 2, 4, 8 μM) with different concentration gradients for 24 h and 48 h, respectively. 10 μL of a cell counting kit-8 (CCK-8) detection solution was dropped into each well after incubation, and then the absorbance of each well was tested according to the instructions.

4.11. Cellular uptake study

Firstly, M2 Exo and EC were modified using Dil. RAW264.7 and FLS cells were seeded in a 24-well plate and cultured overnight. The cells were then stimulated with LPS (1 μg/mL, S11060, Shanghai Yuanye Bio-Technology Co., Ltd, China) for a specified period. Subsequently, Dil-labeled M2 Exo and EC were introduced into the culture medium and co-incubated with the cells for 1–4 h. After incubation, the cells were washed with PBS, and the nuclei were stained with DAPI-containing mounting medium (S2110, Solarbio, China). Cellular uptake and localization were visualized using confocal laser scanning microscopy (CLSM, Leica TCS SP8, Germany).

4.12. Detection of reactive oxygen species (ROS) in vitro

The RAW264.7 cells were seeded into confocal culture dishes at a density of 2 × 10⁴ cells per well. After stimulation with LPS (S11060, Shanghai Yuanye Biotechnology, China) for 12 h, CXB, M2 Exo, and EC were added and incubated for 24 h. Subsequently, 10 μM DCFH-DA (S0033S, Beyotime, China) was added and incubated for 30 min, and the cells were imaged using CLSM. Next, the cells were seeded into 6-well plates at a density of 1 × 10⁶ cells per well, and the same treatment was repeated. ROS levels were measured using flow cytometry, and the collected cells were subjected to statistical analysis of ROS intensity based on fluorescence intensity.

4.13. Immunofluorescence staining

Cells were seeded into a 24-well plate containing coverslips and treated with the designated drugs. The cells were then fixed at room temperature with 4% paraformaldehyde for 30 min. After washing with PBS, 400 μL of 0.3% Triton-X100 solution was added to each well for permeabilization on ice for 10 min. The cells were subsequently blocked with 5% BSA for 60 min. Following this, the cells were incubated overnight at 4 °C with primary antibodies, then incubated with the corresponding fluorescent secondary antibodies for 60 min. Finally, the nuclei were stained with DAPI, and images were captured using CLSM.

4.14. RT-qPCR

For RT-qPCR analysis, cells from the wells were collected, and total RNA was extracted using the TRIzol method. For reverse transcription, 1 μg of total RNA was reverse-transcribed into cDNA using the ABclonal Reverse Transcription Kit (RK20433, ABclonal, China) in a 20 μL reaction volume containing 4 × ABScript Neo RT Master Mix and 20 × gDNA Remover Mix, following the thermal cycler protocol: 37 °C for 2 min, 55 °C for 15 min, and 85 °C for 5 min. Gene expression levels were detected using 2X Universal SYBR Green Fast qPCR Mix (RK21203, ABclonal, China), with GAPDH as the internal reference gene. All samples were run in triplicate to ensure technical reproducibility, and the relative gene expression was calculated using the comparative 2^−ΔΔCT method.

4.15. Macrophage repolarization study

For immunofluorescence staining, RAW264.7 cells were seeded into a well plate, and after LPS stimulation for 12 h, M2 Exo, EC, and CEC were added for treatment for 24 h. Subsequently, the expression of iNOS (1:200, A3774, ABclonal, China) and Arg-1 (1:200, A4923, ABclonal, China) in macrophages was detected using CLSM. For RT-qPCR, the Il1b, Il6, Inos, Il10, Arg1 mRNA expression levels were measured using RT-qPCR. For Western blot analysis, after treating the cells as described above, Western blot was used to detect the expression of iNOS and Arg-1 proteins. For ELISA, the cell culture supernatant was collected, and the concentration of IL-6 and IL-10 in the supernatant was measured using a corresponding mouse ELISA kit (E-EL-M0037, E-EL-M0046, Elabscience, China).

4.16. The purity identification of mouse FLS cells

Third-generation FLS cells were seeded into a 24-well plate containing coverslips. After incubating the cells with the Vimentin antibody (1∶200, 10366-1-AP, Proteintech, China), they were combined with Fluorescein (FITC)-conjugated Affinipure Goat Anti-Rabbit IgG (H + L) (1:300, SA00003-2, Proteintech, China), and the Vimentin staining was observed using CLSM to assess cell purity. Synovial cell populations are considered to be FLS when the proportion of Vimentin-positive cells exceeds 98%.

4.17. Establishment and validation of the FLS cells inflammation model

FLS cells with a purity of over 98% were seeded at a density of 1 × 10⁶ cells per well. LPS (1 μg/mL) was added to stimulate the FLS cells for 24 h. At 0, 1, 3, 6, and 12 h, the expression levels of the inflammatory factors Il1b and Tnf in FLS cells after LPS stimulation at different time points were detected using RT-qPCR.

4.18. The evaluation of different materials on inflammatory FLS cells

FLS cells (1 × 10⁶) were seeded into 6-well plates and treated with LPS (1 μg/mL) for 3 h. After washing three times with PBS, different concentrations of CXB (2, 4, 8, 16 μg/mL), M2 Exo and EC were added to the cells. For RT-qPCR analysis, the expression levels of Il1b, Tnf and Il6 were detected using RT-qPCR. For ELISA analysis, the cell culture supernatant was collected, and the concentrations of IL-1β and IL-10 in the supernatant were measured using the corresponding mouse ELISA kits.

4.19. The detection of CEC on osteoclast differentiation

Fresh bone marrow macrophages (BMMs) were isolated from mice. The procedure is as follows: under sterile conditions, 8-week-old mice were euthanized by cervical dislocation. The tibiae and femora were then extracted, and BMMs were flushed out from the bone marrow cavity using a 1 mL sterile PBS syringe. After lysing red blood cells and centrifugation, the cell pellet was resuspended in complete medium containing 30 ng/mL M-CSF (315-02, PeproTech, USA). The medium used was α-MEM supplemented with 10% FBS and 1% penicillin-streptomycin. To induce osteoclast differentiation, BMMs were seeded onto culture plates according to experimental needs. After 24 h, cells were treated with complete medium containing 30 ng/mL M-CSF and 50 ng/mL RANKL (462-TEC, R&D, USA) to promote differentiation. The medium was changed every other day until the cells matured into osteoclasts. BMMs were seeded into 24-well plates, and the culture medium was supplemented with 50 ng/mL RANKL and 30 ng/mL M-CSF to induce the formation of osteoclasts. In the treatment groups, different concentrations of CC-90011 (0, 1, 2, 4, 8 μM), EC and CEC were added. Once mature multinucleated osteoclasts had formed in the control group, the cells were fixed and subjected to TRAP staining. TRAP-positive osteoclast areas were quantified using ImageJ. Subsequently, the osteoclasts generated during the culture process were stained with DAPI and TRITC-phalloidin to visualize the nuclei and filamentous actin, and quantification was performed using ImageJ. Additionally, the expression levels of c-Src, NFATc1, and LSD1 in osteoclasts were detected by RT-qPCR and Western blot.

4.20. Model of CIA established in DBA mice

Mice were immunized by intradermal injection at the tail base with an emulsion of bovine type II collagen solution (2 mg/mL) and complete Freund's adjuvant (4 mg/mL, F5881, Sigma, USA). On day 21, the mice were boosted with an emulsion of bovine type II collagen solution (2 mg/mL) emulsified with incomplete Freund's adjuvant (IFA, F5506, Sigma, USA).

4.21. Assessment of foot swelling and arthritis score

The CIA mice were randomly divided into 5 groups, which were administered with the same dose of PBS, CXB, CC-90011, EC, and CEC, once every 3 days for a total of 5 injections. Healthy mice were injected intravenously with PBS as the control group. The diameter of the hind paws of the mice was measured using a caliper. Mice were evaluated every 3 days using the following scoring criteria: 0: normal; 1: slight redness and swelling confined to the tarsus or ankle; 2: mild redness and swelling from the ankle to the tarsus; 3: moderate redness and swelling from the ankle to the metatarsus; 4: severe swelling of the ankle, foot, and toes. The total score for each CIA mouse was obtained by adding the scores of all paws.

4.22. Micro-CT bone analysis

The mice in each group were sacrificed, and their ankle joints were collected for Micro-CT scanning. All samples were fixed in 4% paraformaldehyde. These data were used for three-dimensional reconstruction, and 3D-rendered images of the knee joints were generated to analyze the treatment effects in each group.

4.23. Histological analysis of mouse ankle joints

The ankle joint tissue samples were fixed in 4% paraformaldehyde solution, decalcified for two weeks, and then sectioned into knee joint tissue blocks. Sections were subjected to H&E staining, SOFG staining, and TRAP staining, followed by detection of macrophage repolarization-related markers of the M1 marker iNOS and the M2 marker Arg-1.

4.24. Biosafety evaluation

To assess the in vivo safety of CEC, 25 CIA mice were randomly divided into five groups and administered intravenous injections of PBS, CXB, CC-90011, EC, and CEC, respectively. On day 18, histopathological analysis of major organs (hearts, livers, lungs, kidneys) was performed.

4.25. Statistical analysis

Statistical analysis was performed using GraphPad Prism 9.0.0 software, and the data are expressed as mean ± standard deviation. Comparisons between two groups were conducted using a two-tailed unpaired Student's t-test, and comparisons among multiple groups were performed using one-way ANOVA. P < 0.05 was considered statistically significant.

CRediT authorship contribution statement

Qing Xu: Writing – original draft, Visualization, Validation, Methodology, Investigation, Data curation, Conceptualization. Yuan Jiang: Writing – original draft, Methodology, Data curation. Yi Luo: Methodology, Data curation. Lu Li: Methodology, Data curation. Bin Jiang: Methodology, Data curation. Cong Ren: Methodology, Formal analysis. Xiaofeng Liu: Methodology, Data curation. Runtong Liu: Methodology, Data curation. La Zhang: Methodology. HongQing Liu: Methodology. Yafang Chai: Methodology. Yue Li: Software, Methodology. Ning Jiang: Writing – review & editing, Formal analysis. Jianwei Wang: Writing – review & editing, Writing – original draft, Validation, Supervision, Resources, Project administration, Funding acquisition, Formal analysis, Conceptualization.

Ethics approval and consent to participate

All animal experimental protocols were reviewed and approved by Institutional Animal Care and Use of Chongqing Medical University (Approval No. IACUC-CQMU-2025-0577) and were performed in strict compliance with the National Institutes of Health (NIH) guidelines for laboratory animal welfare and experimental practices.

Data availability statement

The authors declare that all data supporting the findings of this study are available in the main text and its supplementary material, and are available from the corresponding author upon reasonable request.

Declaration of competing interest

The authors declare the following personal relationships which may be considered as potential competing interests: Bin Jiang is currently employed by R&D Division, Eureka Biotech Inc.

Appendix A. Supplementary data

The following is the Supplementary data to this article: