Artificial mitochondria ameliorates osteoarthritis through restoring cellular energy metabolism homeostasis

Abstract

Mitochondrial dysfunction under pathological or aging conditions disrupts adenosine triphosphate (ATP) synthesis, exacerbating disease progression by skewing energy metabolism toward catabolism. Current strategies to restore metabolic balance remain limited by complexity or inefficiency. Inspired by the phosphocreatine-creatine kinase (CK) system-a mitochondrial-independent energy pathway, we developed chemotactic artificial mitochondria (CAMs) to address this challenge. CAMs consist of crosslinked phosphocreatine monomers (MPCr) and perfluorooctyl acrylate, designed to exploit CK's chemotactic properties for targeted delivery while resisting biofluid interference. CAMs entered degenerated chondrocytes and meniscus fibrochondrocytes via clathrin-mediated endocytosis, escaped lysosomal degradation, scavenged reactive oxygen species, and restored ATP production. Transcriptomic analysis revealed CAMs upregulated chondrogenic markers (COL2A1, ACAN, SOX9) and suppressed inflammatory pathways (MMP3, IL6), while enhancing extracellular matrix biosynthesis. In a murine knee osteoarthritis (OA) model, intra-articular CAM injections reduced synovial inflammation, preserved cartilage glycosaminoglycan content, and restored gait function by systemic metabolic reprogramming. Histological and radiographic assessments confirmed CAMs mitigated joint space narrowing and cartilage erosion. This study establishes CAMs as a robust, mitochondria-agnostic platform for treating degenerative diseases by rectifying cellular energy imbalance, with immediate translational potential for OA therapy.

Graphical abstract

The schematics diagram demonstrates the preparation of CAM, energy production and potential for knee osteoarthritis therapy. BAC represents N, N′-bis (acryloyl) cysteamine, PFA represents 1H, 1H-perfluorooctyl acrylate, DS represents diclofenac sodium.

Highlights

• •CAMs were developed to restore ATP production via mitochondria-independent phosphocreatine–creatine kinase energy pathway.
• •CAMs reprogram cellular metabolism, increasing ATP synthesis, scavenging ROS, and promoting anabolic pathways.
• •CK-mediated chemotaxis enables targeted delivery of CAMs to cartilage and meniscus in early osteoarthritis (OA).
• •Intra-articular CAMs administration alleviated OA, preserving joint structure and function.

1. Introduction

Aerobic respiration metabolism, including glycolysis, the tricarboxylic acid cycle (TCA cycle), and oxidative phosphorylation, is the main pathway for adenosine triphosphate (ATP) synthesis in mammalian cells [1]. However, under pathological or ageing conditions, mitochondrial dysfunction leads to a decrease in ATP synthesis. Energy metabolism is imbalanced, resulting in a decrease in anabolism and an increase in catabolism, which further exacerbates the progression of diseases or ageing [2]. Previous studies have attempted to increase ATP synthesis and metabolism by intervening in the TCA cycle. However, the entire glucose metabolic process is exceptionally complex, and reversing changes in mitochondrial energy metabolism by intervening in a single or a few targets is impossible. Excessive intervention can even lead to cell death [3]. Although transplanting natural mitochondria directly into cells can replace endogenous damaged mitochondria, the extraction, purification, and storage of natural mitochondria are complex and costly [4]. Another previous study demonstrated the plant thylakoids were encapsulated by cell membranes and produced ATP through photosynthesis under light when transplanted into mammalian cells [5]. This study is very interesting because mammals may synthesize ATP through photosynthesis. However, the entire system preparation process is complex, and light cannot be effectively transmitted to deep tissues, which limits its potential application in deep tissues.

Mammalian skeletal muscle and nerve tissues contain abundant phosphocreatine, which is a high-energy phosphate compound [6]. Approximately 43 kJ of free energy could be released after the hydrolysis of 1 mol of phosphocreatine by creatine kinase (CK), which is greater than the energy released by ATP hydrolysis (30.5 kJ per mole) [7]. Moreover, the released phosphate ions can combine with adenosine diphosphate (ADP) to form ATP [8]. Thus, the phosphocreatine–CK catalytic reaction system is an efficient way to supply and store energy. Moreover, the reaction process is independent of glycolysis, the TCA cycle, and oxidative phosphorylation. In this study, we hypothesize that energy metabolism could be improved by the addition of exogenous phosphocreatine, thus compensating for the damaged mitochondria in pathological or ageing cells. In this study, we found that CK has a chemotactic effect on phosphocreatine, which is similar to the binding process of chemokines and chemokine receptors in the body. Thus, the “chemotactic artificial mitochondria (CAMs)” were prepared in this study. Specifically, a phosphocreatine monomer (MPCr) containing a double-bond structure was obtained through an amide reaction, and a free radical polymerization reaction induced by ammonium persulfate (APS) was used to crosslink MPCr and 1H, 1H-perfluorooctyl acrylate (PFA) using the crosslinking agent N, N′-bis (acryloyl) cysteamine (BAC) to prepare the chemotactic artificial mitochondria (PFMPCr CAMs) nanoparticles. PFA belongs to the hydrophobic and lipophobic substance, thus reducing the blocking effect of body fluids during in vivo administration. BAC could be responsive to glutathione (GSH) degradation [9], thus causing disassembly of CAMs nanoparticles and subsequent release of phosphocreatine monomer.

Knee osteoarthritis (OA) is a very common degenerative disease. Knee joint swelling and joint pain are the main clinical symptoms that seriously affect the knee joint function and daily activities of patients [10]. Knee cartilage and the meniscus are the most severely affected weight-bearing tissues [11]. Osteoarthritic chondrocytes and meniscus fibrochondrocytes exhibit severe mitochondrial dysfunction and imbalanced energy metabolism, accompanied by elevated levels of reactive oxygen species (ROS), interleukins, and matrix metalloproteinase (MMP) enzymes [12]. At present, only early-stage OA may be reversed through external intervention because late-stage OA can be treated only by partial or total knee replacement through the implantation of joint prostheses [13]. Currently, pharmacological treatments for early OA, such as nonsteroidal anti-inflammatory drugs (NSAIDs), mostly aim to relieve the symptoms associated with inflammation or pain [14]. However, few studies have been performed to treat early-stage OA by correcting the imbalance in cellular energy metabolism. In the present study, we observed a concentration gradient of CK between the synovial fluid and cartilage/meniscus tissue in OA samples. The CK concentration in cartilage/meniscus tissue was significantly higher than that in synovial fluid. Therefore, CAMs can selectively penetrate into cartilage and meniscus tissue after an intra-articular injection to increase energy production. Moreover, the anti-inflammatory drug diclofenac sodium (DS) was loaded into the CAMs to reduce inflammation within the knee joint. In the present study, a mice early knee OA model was generated to preliminarily verify the ability of “CAMs” to improve cell energy metabolism, as well as their therapeutic effects on knee OA, and the results provide preliminary evidence for their future clinical application. The schematics diagram was illustrated in Fig. 1.

Fig. 1.

The schematics diagram demonstrates the preparation of CAM, energy production and potential for knee osteoarthritis therapy. BAC represents N, N′-bis (acryloyl) cysteamine, PFA represents 1H, 1H-perfluorooctyl acrylate, DS represents diclofenac sodium.

2. Results

2.1. CK levels in the cartilage and meniscus increase during early knee OA

Synovial fluid, articular cartilage and meniscus tissue were collected from OA patients during total knee replacement to evaluate the CK content within the knee joint. We first measured the CK content in the synovial fluid, articular cartilage and meniscus tissue of OA patients using an ELISA kit. The mean concentrations of CK in synovial fluid, articular cartilage and meniscus tissue were 16.06, 173.99 and 201.36 ng/g, respectively. The concentration of CK in the articular cartilage and meniscus tissue exceeded that in the knee synovial fluid, thus creating a chemoattractant gradient from synovial fluid to cartilage/meniscus (Fig. 2a). Next, synovial fluid, articular cartilage and meniscus tissue frozen sections were prepared. The results of immunofluorescence staining of frozen sections also revealed relatively higher CK contents in the cartilage and meniscus tissue than in synovial fluid (Fig. 2b and c). An in vivo mice OA model and in vitro cell OA model were used to comprehensively assess the effect of the stage of knee OA on the CK content. A mice model of early OA was generated via medial meniscectomy for 5 weeks. A mice model of late OA was generated via medial meniscectomy for 12 weeks. CK immunofluorescence staining was performed on paraffin sections of mice knees. In the mice early OA model, the surface of the articular cartilage was relatively smooth and intact, similar to that of normal articular cartilage. However, the mice model of late OA presented severe cartilage abrasion and defects. The subchondral bone was even exposed. Interestingly, we found that the CK content in the articular cartilage was significantly higher in the early OA model than in the normal cartilage. However, the CK content of the articular cartilage decreased significantly in late OA (Fig. 2d and e). In addition to the post-traumatic OA model, the MIA chemically induced inflammatory OA model was prepared. In the present study, 10 μl 1% (w/v) monoiodoacetic acid (MIA) was injected into mice knee joint. At 6 weeks after injection, the CK expression within articular cartilage and meniscus was evaluated using immunofluorescence. We found CK expression was upregulated significantly in chondrocytes and meniscus fibrochondrocytes (Fig. S1). Next, the CK expression between human OA and normal cartilage tissues was evaluated. In the present study, the tibial plateau cartilage was harvested from OA patients during total knee replacement. The relative normal and degenerated cartilage regions from macroscopic (Fig. S2a) and histology (Fig. S2b and c) were recognized. We found the CK expression was significantly upregulated in degraded human OA chondrocytes (Fig. S2d–f). In vitro rat chondrocyte and meniscus cell OA models were generated by stimulation with LPS. The relative CK content was characterized using SDS-PAGE. In chondrocytes, the expression of matrix metalloproteinase-13 (MMP13) increased constantly as the LPS stimulus time increased from 0.5 to 8 h, which also reflected an increased severity of OA [15]. However, the expression of CK increased from 0.5 to 2 h of LPS stimulation and then decreased at 4 and 8 h of stimulation (Fig. 2f–h). A similar trend was also observed in meniscus fibrochondrocytes. The expression of MMP13 increased constantly as the LPS stimulus time increased from 0.5 to 8 h. However, the expression of CK increased from 0.5 to 4 h of LPS stimulation and then decreased at 8 h of stimulation (Fig. 2i–k). In summary, the expression of CK in the cartilage and meniscus increased during early knee OA. Early knee OA was considered in our subsequent animal study.

Fig. 2.

The upregulation of CK in the cartilage and meniscus during early OA. (a) The contents of CK in the synovial fluid, articular cartilage and meniscus tissue of OA patients are measured with an ELISA kit. n = 20, one-way ANOVA, ∗∗∗∗ represents p < 0.0001. (b, c) Immunofluorescence staining (b) and semi-quantitative analysis (c) of CK levels within synovial fluid, articular cartilage and meniscus tissues of OA patients. CK represents creatine kinase. n = 6, one-way ANOVA, ∗∗ represents p < 0.01 and ∗∗∗∗ represents p < 0.0001. White scale bar: 50 μm; yellow scale bar: 10 μm. (d, e) Immunofluorescence staining (d) and semi-quantitative analysis (e) of CK levels in the knee joints of normal, early OA and late OA mice. F represents the femur, T represents the tibia, M represents the meniscus, AC represents the articular cartilage and SB represents the subchondral bone. n = 6, one-way ANOVA, ∗∗∗∗ represents p < 0.0001. White scale bar: 50 μm; yellow scale bar: 10 μm. (f–h) SDS-PAGE (f) results and semi-quantitative analysis of CK (g) and MMP13 (h) levels in rat chondrocytes after in vitro LPS stimulation for different durations. n = 3, one-way ANOVA, ∗ represents p < 0.05, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001 and ns represents no significant difference. MMP13 represents matrix metalloproteinase-13, and CK represents creatine kinase. (i–k) SDS‒PAGE results (i) and semi-quantitative analysis of CK (j) and MMP13 (k) levels in rat meniscus fibrochondrocytes after in vitro LPS stimulation for different durations. n = 3, one-way ANOVA, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001 and ns represents no significant difference.

To delineate the cellular origin of CK within the OA joint microenvironment and to assess the cell-type-specific targeting potential of our CAMs platform, we next analyzed the expression patterns of different CK isoforms across various joint-resident cell types using publicly available single-cell RNA sequencing (scRNA-seq) data from different mice OA models [16]. This analysis aimed to identify which cell populations predominantly express CK and its isoforms during OA, thereby informing the likelihood of CAMs accumulation in specific tissue compartments upon intra-articular delivery. Briefly, a total of six mice models were included: MN (male normal mice), MB (male bipedal mice), FN (female normal mice), FB (female bipedal mice), FO (female mice with bilateral ovariectomy), and FBO (female bipedal mice with bilateral ovariectomy). The forelimbs and tails were amputated to develop bipedal model at the age of 3 weeks. Bilateral ovariectomy was performed in half of the female mice at the age of 10 weeks. The bipedal model was selected to increase mechanical load on knees, which was validated to develop cartilage deterioration and subchondral bone alteration resembling knee OA patients. The femoral condyles were harvested at the age of 22 weeks for single cell RNAseq. A total of 82,083 cells were collected in all mice models (Fig. 3a and b). A total of 15 celltypes were identified (Fig. 3c), including Endothelial cell (EndC, Flt1, Cdh5, Cldn5), Chondrocyte (Cytl1, Chad, Ecrg4), Reticular_cell (Ghr, Cxcl12, Csmd1), Erythrocyte (Hba-a1, Hbb-bs, Hbb-bt), Vascular smooth muscle cell (VSMC, Tagln, Gm13889, Rgs5), Progenitor_cell (Cxcl1, Ccl7, Clec3b), Osteoblast (Bglap, Bglap2, Col1a1), Dendritic_cell (Bst2, Ccr9, Cox6a2), Monocyte (Ctla2a, Adgrl4, Cdk6), Macrophage (Pid1, Cxcl3, Slpi), B_cell (Bach2, Ebf1, Flt1), T_cell (Ccl5, Camk4, Itk), Granulocyte_1 (S100a8, S100a9, Retnlg), Granulocyte_2 (Elane, Prtn3, Mpo), Granulocyte_3 (Prss34, Mcpt8, Ifitm1). The CK family mainly contains 4 isozymes, including creatine kinase, M-type (Ckm), creatine kinase B (Ckb), creatine kinase, mitochondrial 1 (Ckmt1), creatine kinase, mitochondrial 2 (Ckmt2). The CK gene expression profiles were analyzed. Firstly, the UMAP plot demonstrated all celltypes expressed Ckb, Ckm, Ckmt1, and Ckmt2 (Fig. 3d–g). Fig. 3d demonstrated stronger Ckb expression (indicated by red color) in chondrocytes and progenitor cells. Next, we analyzed the relative CK isozyme expression in each mice model. Regardless of gender, osteoarthritis or ovariectomy, Ckb was mostly expressed among CK isozymes and mainly concentrated in Chondrocyte, Osteoblast, Progenitor cells, Reticular cells and VSMC. However, the CK gene expression level was relatively low in Macrophage, Monocyte, Erythrocyte, T_cell, B_cell and Granulocyte (Fig. 3h–m). We further analyzed the CK gene expression changes between control normal and OA status. Regardless of gender or ovariectomy, the CK gene expression was overall upregulated in OA status, especially in Chondrocyte, Osteoblast and VSMC (Fig. 3n–p).

Fig. 3.

The CK gene expression profiles in OA mice knee through single cell RNAseq analysis. (a) The UMAP plot showing the cells from six mice models, including MN (male normal mice), MB (male bipedal mice), FN (female normal mice), FB (female bipedal mice), FO (female mice with bilateral ovariectomy), and FBO (female bipedal mice with bilateral ovariectomy). (b) The UMAP plot showing the cells of control normal and OA mice models. (c) The UMAP plot showing the 15 different celltypes. (d–g) The UMAP plot showing the CK isozymes gene expression patterns including Ckb (d), Ckm (e), Ckmt1 (f), Ckmt2 (g). (h–m) The dot plot showing the CK isozymes gene expression levels of various celltypes in each mice model, including MN (h), MB (i), FN (j), FB (k), FO (l), FBO (m). (n–p) The dot plot showing the relative CK gene expression levels of various celltypes between Ctrl and OA groups in male (n), female (o), and female ovariectomy (p) models.

2.2. Preparation and characterization of CAMs

First, a MPCr containing a double-bond structure was obtained through an amide reaction, and a free radical polymerization reaction induced by APS was used to crosslink MPCr and PFA using the crosslinking agent BAC to prepare chemotactic artificial mitochondria (PFMPCr CAMs). PFA belongs to the hydrophobic and lipophobic substance, thus reducing the blocking effect of body fluids during in vivo administration. BAC could be responsive to glutathione (GSH) degradation [9], thus causing disassembly of CAMs nanoparticles and subsequent release of phosphocreatine monomer. The anti-inflammatory drug diclofenac sodium (DS) was loaded onto PFMPCr CAMs through electrostatic adsorption to obtain PFMPCr/DS CAMs. Furthermore, PF nanoparticles (PF NPs) were synthesized by polymerizing BAC and PFA without the addition MPCr, which served as control samples lacking mobility. As shown in Fig. S3, the synthesis of MPCr was verified by ¹H NMR spectroscopy. Specifically, the ¹H NMR spectrum in Fig. S3 shows that PCr lacking any olefinic bond gives clear characteristic signals for its N-methyl (marked b in the left panel) and methylene (marked a) protons. After amidation with 2-methylallylamine, the resulting double-bond-containing PCr retains these two sets of peaks while exhibiting a new methyl signal from the 2-methylallyl moiety (marked e in the right panel) and additional olefinic proton signals attributable to the C=C bond (marked a and b in the right panel), confirming the successful synthesis of MPCr bearing the desired vinyl group. TEM images revealed that PF, PFMPCr CAMs, and PFMPCr/DS CAMs were uniformly spherical nanoparticles of approximate 100 nm, and the morphology of PFMPCr CAMs remained unchanged after loading with DS (Fig. 4a). DLS revealed that the hydrodynamic diameters of PF, PFMPCr CAMs, and PFMPCr/DS CAMs were slightly larger than the particle size results obtained using TEM (Fig. 4b). The PDI values of the three nanoparticles indicate that they are all well-dispersed. Moreover, after PFMPCr/DS was stored in water for one week, its dispersity remained essentially unchanged, demonstrating the good stability of CAMs (Fig. S4). The zeta potential indicated that both PF and PFMPCr CAMs exhibited negative potentials, which were attributed to the strongly electronegative F atoms in PFA and the negatively charged phosphate groups in MPCr. After the loading of negatively charged DS, the zeta potential of PFMPCr/DS CAMs decreased slightly (Fig. 4c). In addition, we screened the optimal ratios of the DS mass and PFMPCr mass for the preparation of PFMPCr/DS CAMs (Fig. 4d). Our research revealed that as the mass ratio of DS to PFMPCr increased, the drug loading of PFMPCr/DS CAMs gradually increased. When the mass ratio of DS to PFMPCr reached 5:1, the drug loading reached 170 mg/g. When the dosage of DS was increased at this ratio, almost no significant improvement in the drug loading of the PFMPCr/DS CAMs was observed. Therefore, we prepared PFMPCr/DS CAMs at a ratio of 5:1 between the DS mass and PFMPCr mass. It was proposed that a controlled release behavior for both PCr and DS was attributed to GSH-sensitive degradation of CAMs nanoparticles. Then, the release kinetics of DS and PCr from CAMs in vitro under relevant knee joint physiological condition was completed. Previous studies demonstrated the GSH concentration within knee joint synovial fluid was approximate 100 μM [17]. PFMPCr/DS CAMs were dispersed in solutions with or without GSH (100 μM). At 1, 3, 6, 12, 24, 48, 72, 96, 120, and 144 h, the degradation of CAMs was reflected by the relative turbidity, which was measured by UV spectrophotometer at 660 nm. The release of DS and PCr was simultaneously measured. As shown in Fig. S5, the results showed that GSH treatment facilitated CAMs degradation and the release kinetics of DS and PCr was in a coordinated manner.

Fig. 4.

Characterization of CAMs. (a) TEM images of PF (PFA-only control nanoparticles) (left panel), PFMPCr CAMs (middle panel), and PFMPC/DS CAMs (right panel). Scale bar: 100 nm. (b) DLS particle size maps of PF, PFMPCr CAMs, and PFMPCr/DS CAMs. (c) Zeta potential maps of PF, PFMPCr CAMs, and PFMPCr/DS CAMs. (d) Maps of drug loading with different DS and PFMPCr mass ratios. (e) Motion trajectory of 10 randomly selected PF particles and their average velocity distribution in LPS-stimulated chondrocyte lysate. (f) Motion trajectory of 10 randomly selected PFMPCr/DS CAMs particles and their average velocity distribution in PBS. (g) Motion trajectory of 10 randomly selected PFMPCr CAMs particles and their average velocity distribution in LPS-stimulated chondrocyte lysate. (h) Motion trajectory of 10 randomly selected PFMPCr/DS CAMs particles and their average velocity distribution in LPS-stimulated chondrocyte lysate.

Next, the autonomous motion performance of the PFMPCr/DS CAMs was studied. In the present study, we detected an increase of CK content within cartilage and meniscus of early OA patients. Therefore, we investigated the motility of nanomaterials containing the endogenous reactive substrate PCr in a CK environment. First, we investigated the movement of PF without endogenous reactants in the presence of CK environment and the movement of PFMPCr/DS CAMs with endogenous reactant PCr in the absence of CK environment. In the chondrocyte lysis buffer prestimulated with LPS containing CK, the movement trajectory of Cy5-labeled PF exhibited irregular Brownian motion, with a movement rate ranging from 0 to 2 μm/s (Fig. 4e; Movie S1). In addition, in PBS without CK, the Cy5-labeled PFMPCr/DS CAMs also exhibited irregular Brownian motion at speeds of 0 to 2 μm/s (Fig. 4f; Movie S2). In the chondrocyte lysis buffer prestimulated with LPS containing CK, both the Cy5-labeled PFMPCr CAMs (Fig. 4g; Movie S3) and the Cy5-labeled PFMPCr/DS CAMs (Fig. 4h; Movie S4) moved in different directions, with velocities ranging from 2 to 4 μm/s. These findings indicate that nanomaterials containing endogenous reaction substrates have autonomous mobility in the CK environment. Next, regarding the issue of synthesis reproducibility, we additionally synthesized different batches of PFMPCr CAMs. The particle diameter, zeta potential, and movement behavior within chondrocyte lysate demonstrated good batch consistency, indicating superior reproducibility (Fig. S6; Movie S5; Movie S6).

2.3. Chemotactic performance of CAM

A static chemotaxis study was subsequently conducted on the PFMPCr/DS CAMs using a Y channel (Fig. 5a). First, the agarose gel containing CK was fixed in Chamber II, and the agarose gel containing PBS was fixed in Chamber III. A small amount of PBS was slowly added from Chamber I, and CK was dispersed from the agarose gel in Chamber II, forming a concentration gradient of the chemoattractant CK in the Y channel. After 20 μL of Cy5-labeled PCr was added to Chamber I, the fluorescence intensities in Chamber II and Chamber III were observed at 0, 30, and 60 min, respectively (Fig. S7). The fluorescence intensity in Chamber II was significantly higher than that in Chamber III at 30 and 60 min, indicating that PCr exhibited chemotactic behaviour towards the CK concentration gradient. Further research was conducted on the chemotactic behaviour of materials containing PCr towards environments containing CK. We placed LPS-prestimulated chondrocyte lysate mixed with the agarose gel in Chamber II and artificial synovial fluid mixed with the agarose gel in Chamber III. PBS was slowly added to Chamber I, CK in the lysate was dispersed from the agarose gel in Chamber II, and a concentration gradient of the chemoattractant CK was formed in the Y channel. We separately added 20 μL of Cy5-labeled PF, Cy5-labeled PFMPCr CAMs, or Cy5-labeled PFMPCr/DS CAMs to Chamber Ⅰ and observed the fluorescence intensities in Chamber Ⅱ and Chamber Ⅲ at 0, 30, and 60 min. We did not observe a significant difference in fluorescence intensity between Chamber I and Chamber II, to which PF was added at each time point. The groups treated with PFMPCr CAMs and PFMPCr/DS CAMs presented an increased fluorescence intensity in Chamber II at 30 and 60 min (Fig. 5b–d). This phenomenon indicated that materials containing PCr tend to migrate towards higher concentrations of CK along the CK concentration gradient.

Fig. 5.

Chemotactic performance of CAMs. (a) Schematic diagram of the Y channel. (b), (c), and (d) Fluorescence images and normalized fluorescence quantification at the bottom ends of the Y channel (Chamber Ⅱ and Chamber Ⅲ) at 0, 30, and 60 min after Cy5-labeled PF, PFMPCr CAMs, and PFMPCr/DS CAMs, respectively, were added to the Y channel. Scale bar: 1 mm, n = 3; two-way ANOVA; ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. (e) Schematic diagram of the microfluidic channel. (f), (g) and (h) Fluorescence images and normalized fluorescence quantification near the outlet of the microfluidic channel when different fluorescently labeled materials (PF, PFMPCr CAMs, and PFMPCr/DS CAMs) were added. Scale bar: 500 μm. (i) The motion trajectory of 5 randomly selected PFMPCr particles in the single-channel linear chamber. (j) The motion trajectory of 5 randomly selected PF particles in the single-channel linear chamber. (k) The chemotactic index of PF and PFMPCr. n = 5, unpaired t-test, ∗∗∗∗ represents p < 0.0001. (l) The movement speed of PF and PFMPCr. n = 5, unpaired t-test, ∗∗∗∗ represents p < 0.0001.

Furthermore, we examined whether inflammatory factors could interfere with this chemotaxis. As shown in Fig. S8, the fluorescence intensities in Chambers II and III did not differ significantly after the 0.5, 1.0 and 1.5-h assays, indicating that CAMs retain their CK-directed migratory capacity even in the presence of inflammatory cytokine TNF-α. Moreover, Single-channel assays (Fig. S9) revealed that when a CK gradient was present, PF nanoparticles slowly diffused from the top toward the bottom; re-supplementation after 1.5 h still produced only modest further migration. In contrast, PFMPCr/DS rapidly chemotaxed to the basal end in large numbers. Although CK at the bottom may have been partially consumed after 1.5 h, repeated additions every 1.5 h yielded no significant change in the incremental fluorescence at the bottom, indicating that CAM chemotaxis is insensitive to local fluctuations in the CK gradient and remains essentially unaffected.

Dynamic chemotaxis studies on PFMPCr/DS CAMs were subsequently conducted using the dynamic microfluidic channel shown in Fig. 5e. We filled the two sides of the microfluidic channel (Buffer I and Buffer III) with LPS-prestimulated chondrocyte lysate and artificial synovial fluid, respectively, and passed solutions containing different Cy5 labelling materials through the middle: PF, PFMPCr CAMs, and PFMPCr/DS CAMs. A CK concentration gradient was present perpendicular to the direction of fluid motion, and we studied the chemotactic ability of the material by comparing the range of fluorescence perpendicular to the direction of fluid motion. As shown in Fig. 5f, PF without PCr did not exhibit chemotactic behaviour in the microfluidic channels. Both PFMPCr CAMs and PFMPCr/DS CAMs containing PCr tended to significantly shift in the Buffer I environment with an increasing CK concentration (Fig. 5g and h). This phenomenon further confirms the significant chemotaxis of PFMPCr CAMs and PFMPCr/DS CAMs, which can respond to the CK concentration gradient and exhibit chemotactic behaviour.

The chemotaxis index was used as an indicator to evaluate the directed movement ability of nanomotors under chemical stimulation [18]. The lysate of LPS-stimulated chondrocytes mixed with agarose was filled in the right reservoir of a 2-cm single-channel linear chamber to construct a concentration gradient in the horizontal direction. PF or PFMPCr was filled on the left side of the single-channel linear chamber, and the movement trajectory of nanoparticles in the channel was captured. The ratio of nanoparticles displacement to path length represents the chemotaxis index. The speed of nanoparticles motion was recorded. The live tracking showed that PFMPCr particles demonstrated directional movement towards chondrocyte lysates (Fig. 5i; Movie S7), while this phenomenon was absent in PF particles (Fig. 5j; Movie S8). The chemotaxis index of PFMPCr was significantly higher than that of PF (Fig. 5k), and the movement speed was also much greater than that of PF (Fig. 5l), further supporting the claim of CK-mediated PFMPCr CAM chemotaxis.

2.4. Biocompatibility, cellular uptake and lysosomal escape of CAMs

First, the cellular safety of PF, PFMPCr CAMs, and PFMPCr/DS CAMs in chondrocytes and meniscus cells was evaluated using the MTT assay. As shown in Fig. S10, after three materials were incubated with chondrocytes and meniscus cells at different doses (50, 200, 500 μg/mL) for 1, 2, or 3 days, the cell viability of each group remained high, indicating that these three materials did not have significant cytotoxic effects on chondrocytes or meniscus cells. The MTT results revealed that the viability of the cells treated with PFMPCr CAMs or PFMPCr/DS CAMs was not only not inhibited but also higher than that of the cells in the untreated group. This result may be because PCr in the material and intracellular ADP produced under the catalytic action of CK promote the generation of ATP, thereby increasing the metabolic activity of the cells and improving their overall viability.

In theory, the chemotactic abilities of PFMPCr CAMs and PFMPCr/DS CAMs are expected to promote their cellular uptake. Based on these findings, the ability of chondrocytes and meniscus cells to take up PFMPCr/DS CAMs and the corresponding control material were further evaluated. Cy5-labeled PF, PFMPCr CAMs, and PFMPCr/DS CAMs were incubated with LPS-stimulated chondrocytes for 12 h. The fluorescence intensity of each group of samples was subsequently observed and analyzed using confocal laser scanning microscopy (CLSM). As shown in Fig. 6a, PFMPCr CAMs and PFMPCr/DS CAMs, which have chemotactic abilities, could be taken up more efficiently by chondrocytes than PF without a chemotactic ability. Specifically, the normalized fluorescence quantification via CLSM revealed that the amounts of PFMPCr/DS taken up by chondrocytes was approximately 2.7 times higher than that of PF. The results of the quantification of the cellular uptake efficiency revealed that the uptake efficiency of PF by chondrocytes was 10.7%, whereas the uptake efficiency of PFMPCr/DS CAMs increased to 27.2%. As shown in Fig. 6b, when PF, PFMPCr CAMs, and PFMPCr/DS CAMs were cocultured with LPS-stimulated meniscus cells, the results were consistent with the findings of chondrocyte uptake, indicating that meniscus cells had a significantly better ability to take up PFMPCr CAMs and PFMPCr/DS CAMs than PF. Specifically, the normalized fluorescence quantification via CLSM revealed that the amounts of PFMPCr/DS taken up by meniscus cells was approximately 2.9 times higher than that of PF. The results of the quantification of the cellular uptake efficiency revealed that the uptake efficiency of PF by meniscus cells was 9.5%, whereas the uptake efficiency of PFMPCr/DS CAMs increased to 25.6%. This result confirms that the chemotactic abilities of PFMPCr CAMs and PFMPCr/DS CAMs play a crucial role in promoting cellular uptake.

Fig. 6.

Cellular uptake and lysosomal escape of CAMs. (a), (b) Fluorescence images and normalized fluorescence quantification of chondrocytes and meniscus cells after the uptake of different materials, as well as the cellular uptake efficiency results. red: fluorescently labeled materials; green: cell membrane; scale bar: 10 μm; n = 3; one-way ANOVA; ∗∗ represents p < 0.01; ∗∗∗ represents p < 0.0005; and ns represents no significant difference. (c, d) Flow cytometry analysis and normalized fluorescence quantification of chondrocytes (c) and meniscus cells (d) taking up CAMs after exposure to different inhibitors. n = 3; one-way ANOVA; ∗ represents p < 0.05; ∗∗ represents p < 0.01; ∗∗∗ represents p < 0.0005; ∗∗∗∗ represents p < 0.0001; and ns represents no significant difference. (e) Lysosomal escape of chondrocytes incubated with PF or PFMPCr/DS CAMs for 5 min, 1 h, or 6 h. Blue: nuclei; red: fluorescently labeled materials; green: lysosomes; scale bar: 10 μm. (f) Lysosomal escape of meniscus cells incubated with PF or PFMPCr/DS CAMs for 5 min, 1 h, or 6 h. Blue: nuclei; red: fluorescently labeled materials; green: lysosomes; scale bar: 10 μm.

Next, we investigated the possible pathways through which chondrocytes and meniscus cells take up CAMs. The inhibition methods used in the study were as follows: (1) rottlerin, which blocks the macropinocytosis pathway; (2) dynasore, which blocks the clathrin-mediated endocytosis pathway; (3) nystatin, which blocks the caveolae-mediated endocytosis; (4) cytochalasin D (cyto D), which blocks the phagocytosis; and (5) 4 °C, which blocks the energy-dependent endocytic pathway. The chondrocytes and meniscus cells were pretreated with 10 μg mL⁻¹ LPS for 2 h to develop osteoarthritic model. The corresponding inhibitors were subsequently added and incubated for 30 min. Then, the Cy5 fluorescence labeled CAMs were added and incubated for 2 h, followed by flow cytometry analysis. As shown in Fig. 6c, the flow cytometry analysis demonstrated chondrocytes remained high Cy5 fluorescence positive rates even in the condition of rottlerin, dynasore, nystatin, cyto D. While, 4 °C treatment significantly reduced the chondrocytes Cy5 positive rate. The relative mean Cy5 fluorescence intensity of chondrocytes were also analyzed. We found the mean fluorescence intensity reduced to 58.3% and 16.7% of the initial level after rottlerin and 4 °C treatment, respectively. These findings suggest that chondrocytes may primarily take up CAMs through macropinocytosis and energy-dependent pathways. Similarly, in the presence of rottlerin and 4 °C, the uptake capacity of meniscus cells on CAMs was significantly reduced (Fig. 6d). Specifically, the ability of meniscus cells to take up CAMs was observed after treatment with rottlerin. The mean fluorescence intensity of meniscus cells decreased to 51% of the original value, and at 4 °C, it decreased to 28.7% of the original value. This decrease in uptake may be attributed mainly to CAMs uptake by meniscus cells through macropinocytosis and energy-dependent pathways.

As organelles that are primarily responsible for cellular breakdown, lysosomes can engulf and degrade most of the nanoparticles that enter the cell [19]. Therefore, whether nanomaterials can effectively escape lysosomal degradation is a key factor affecting their expected function within cells. Subsequently, the escape from lysosomes in chondrocytes and meniscus cells of PF without a chemotactic ability and PFMPCr/DS CAMs with a chemotactic ability was studied. Fluorescence images of PF, PFMPCr/DS CAMs with chondrocytes and meniscus cells were obtained through CLSM after incubations for 5 min, 1 h, and 6 h, and the degree of colocalization between the material and lysosomes was further analyzed. At different time points, the red peak with surface PF completely overlapped with the green peak representing lysosomes, and PF colocalized with lysosomes, with no observed escape from lysosomes. However, PFMPCr/DS CAMs, with chemotactic ability, showed a slight displacement of the red and green fluorescence peaks when coincubated with chondrocytes for 5 min, and over time, more material escaped from the lysosomes of chondrocytes (Fig. 6e). Similarly, PF did not have the ability to escape from lysosomes in meniscus cells, whereas PFMPCr/DS CAMs could escape from lysosomes, thereby preventing their lysosomal degradation (Fig. 6f). Next, we co-stained mitochondria with MitoTracker and CAMs-Cy5 to perform subcellular colocalization analysis (Fig. S11). The results revealed that CAMs are predominantly localized in the cytoplasm, aligning with the cytoplasmic distribution of CK, further supporting the role of CAMs as a cytoplasmic ATP-generating system.

2.5. Effects of CAMs on ATP production, ROS scavenging and the amelioration of mitochondrial dysfunction

Mitochondria in normal cells provide ATP to chondrocytes and meniscus cells to maintain their energy needs, whereas damaged cells experience a decrease in ATP synthesis and metabolism due to mitochondrial dysfunction. Therefore, we conducted research on the core function of the PFMPCr/DS CAMs, which is the ability to generate ATP. Under the catalysis of CK, the high-energy phosphate bond of PCr is transferred to ADP to generate ATP. Therefore, PFMPCr CAM and PFMPCr/DS CAM containing PCr have the ability to provide ATP to cells. As shown in Fig. 7a, after 12 h of exposure to the different treatments, significant differences in the ATP content of the chondrocytes and meniscus cells were observed for each component. Specifically, after LPS-injured cells were co-cultured with PF or free PCr for 12 h, their ATP levels showed no significant difference compared with the untreated damaged group. In contrast, damaged chondrocytes and meniscus cells cocultured with PFMPCr CAMs or PFMPCr/DS CAMs presented a significant increase in the intracellular ATP content. Especially after 12 h of treatment with PFMPCr/DS CAMs, the ATP content in damaged chondrocytes and meniscus cells approached normal levels. Additionally, treating TNF-α-injured chondrocytes and meniscus cells with PFMPCr CAMs or PFMPCr/DS CAMs for 12 h also elevated their intracellular ATP levels (Fig. S12). Next, in order to confirm CAMs enhanced cellular ATP production directly through PCr-CK system, a small-molecule Ompenaclid (RGX-202) was applied, which was the SLC6A8 transporter inhibitor [20]. Ompenaclid robustly inhibited creatine import in vitro and in vivo. We found the addition of Ompenaclid significantly reduced ATP content within chondrocyte even in the condition of PFMPCr/DS CAMs, thus confirming that PFMPCr/DS CAMs enhanced cellular ATP production directly through PCr-CK system (Fig. S13).

Fig. 7.

Effects of CAMs on ATP production, ROS scavenging and the amelioration of mitochondrial dysfunction. (a) ATP production by chondrocytes (left panel) and meniscus cells (right panel) after exposure to the different treatments for 12 h n = 3, one-way ANOVA; ∗ represents p < 0.05, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. I: normal group; II: LPS-stimulated group; III: LPS + PF group; IV: LPS + PCr group; V: LPS + PFMPCr CAM group; VI: LPS + PFMPCr/DS CAM group. (b) Effects of CAMs on ATP production in chondrocytes (left panel) and meniscus cells (right panel) stimulated with LPS at different time points. (c–e) Fluorescence images (c) and normalized fluorescence quantification of the changes in mitochondrial mtROS levels in chondrocytes (d) and meniscus cells (e) after treatment with different materials. Red: ROS stained with MitoSOX Red; scale bar: 10 μm; n = 3; one-way ANOVA; ∗ represents p < 0.05, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. (f, g) Normalized fluorescence quantification of ROS levels in chondrocytes (f) and meniscus cells (g) after treatment with different materials. ROS were stained with DCFH-DA, n = 3; one-way ANOVA, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. (h, i) Full-spectrum fluorescence scan and quantitative results for chondrocytes (h) and meniscus cells (i) stained with JC-1 after excitation at 490 nm and treatment with different materials; n = 3, one-way ANOVA; ∗ represents p < 0.05, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001. I: normal group; II: LPS-stimulated group; III: LPS + PF group; IV: LPS + PCr group; V: LPS + PFMPCr CAM group; VI: LPS + PFMPCr/DS CAM group.

As shown in Fig. 7b, we specifically studied the changes in ATP production of chondrocytes and meniscus cells within 48 h in different conditions. We found that in normal chondrocytes and meniscus cells, the intracellular ATP content tended to increase, while LPS treatment caused reduced ATP production. PFMPCr CAM addition apparently increased ATP production. Moreover, the benefits of PFMPCr CAM could be further enhanced by the addition of DS. Phosphocreatine and creatine kinase together constitute a very important cellular energy buffer and transport system. The core of this system is the reversible reaction catalyzed by creatine kinase. Phosphocreatine (PCr) + ADP + H⁺ ⇌ Creatine + ATP. In our study, we confirmed the CK expression was upregulated in OA chondrocytes and meniscus fibrochondrocytes, which could facilitate the reaction. In the environment of CK upregulation, the released PCr from CAMs system could produce ATP. As shown, ATP production was prolonged increased. This means the injured status of cells recovered gradually and produced excessive ATP. Due to the reversible feature, CK can catalyze ATP to transfer its terminal high-energy phosphate group to creatine to synthesize creatine phosphate. In this way, the energy is stored to prepare for the next rapid energy demand. Thus, the PCr from CAMs was recyclable and responsible for cellular energy balance. Next, as aforementioned, GSH treatment facilitated CAMs degradation and the release of DS and PCr. The prolonged ATP elevation after CAMs treatment should also come from the gradual release of PCr and long-term particle retention in situ.

Under the inflammatory conditions of OA, excessive intracellular ROS levels can lead to mitochondrial dysfunction [21]. Therefore, a further evaluation of the mitochondrial reactive oxygen species (mtROS) content in damaged chondrocytes and meniscus cells treated with PFMPCr/DS CAMs was conducted. MitoSOX Red was used to stain mtROS in chondrocytes and meniscus cells to evaluate the ability of PFMPCr/DS CAMs to clear mtROS. As shown in Fig. 7c, the mitochondrial mtROS content was significantly increased in damaged chondrocytes and meniscus cells after LPS stimulation. After treatment with PFMPCr/DS CAMs, the mtROS content was significantly reduced. Specifically, after treatment with PFMPCr/DS CAMs, the mtROS content in damaged chondrocytes decreased to 0.28 times that of the untreated group (Fig. 7d); after treatment with PFMPCr/DS CAMs, the mtROS content in meniscus cells decreased to 0.29 times that of the untreated group (Fig. 7e). In contrast, PFMPCr CAMs also partially cleared mtROS in cells, but PCr and PF had minimal effects on the clearance of mtROS in cells. In addition, we also stained ROS in cells with a DCFH-DA fluorescent probe and detected the specific fluorescence intensity under excitation at 480 nm and emission at 525 nm. The overall ROS levels in damaged chondrocytes (Fig. 7f) and meniscus cells (Fig. 7g) were significantly increased; PFMPCr/DS CAMs significantly reduced the ROS levels in both types of cells. These results indicate that a PFMPCr/DS CAM treatment can effectively reduce oxidative stress levels in damaged cells.

The mitochondrial membrane potential plays an important role in the normal function of cells [22]. Therefore, we used the JC-1 fluorescent probe to study the changes in the mitochondrial membrane potential of chondrocytes (Fig. 7h) and meniscus cells (Fig. 7i) after treatment with PFMPCr/DS CAMs. JC-1 forms aggregate in normal mitochondria with a high membrane potential. When mitochondria are damaged and the membrane potential decreases, JC-1 exists in the cytoplasm in the form of monomers. The fluorescence intensity of JC-1 aggregates (598 nm) and JC-1 monomers (542 nm) in cells was obtained by scanning the full fluorescence spectrum upon excitation at 490 nm, thus revealing changes in the mitochondrial membrane potential. Chondrocytes and meniscus cells were damaged, and the ratio of JC-1 aggregates to JC-1 monomers in the cells decreased, indicating depolarization of the mitochondrial membrane. After treatment with PFMPCr/DS CAMs, the JC-1 aggregate/JC-1 monomer ratio returned to near normal cellular levels, indicating that CAMs improved the mitochondrial functional status. The above results indicate that PFMPCr/DS CAMs can effectively eliminate ROS in cells, repair mitochondrial function, and have significantly better therapeutic effects than the other treatments.

2.6. CAM reprogrammed the transcriptome of chondrocytes and meniscus fibrochondrocytes

First, the chondrocytes were treated with lipopolysaccharide (LPS) for 2 h, followed by an incubation with artificial mitochondria for 12 h. The qPCR results demonstrated that CAMs could maintain chondrogenic phenotypes (Col2a1, Acan, and Sox9) and decrease the inflammatory phenotypes (Mmp3 and Il6) of chondrocytes after LPS stimulation (Fig. 8a). Next, we performed a transcriptomic analysis of chondrocytes from the blank, LPS, and LPS + PFMPCr groups. The gene expression profiles of chondrocytes were significantly different among the groups (Fig. 8b and c; Fig. S14). The changes in the transcriptome of chondrocytes after LPS stimulation are described in the Supplemental Information. Overall, the inflammation-related gene sets were significantly upregulated after LPS stimulation. The gene sets associated with the cartilage phenotype (cartilage condensation, such as Sox9, Col2a1, and Acan), energy metabolism (response to ATP), glycogen metabolism (glycogen metabolic process) and lipid metabolism (cholesterol metabolic process) were downregulated after LPS stimulus. Next, the transcriptomes of the chondrocytes from the LPS + PFMPCr and LPS groups were further evaluated. The GO analysis revealed that the following biological processes were enriched in chondrocytes after PFMPCr treatment: negative regulation of the apoptotic process, positive regulation of cell proliferation, positive regulation of cell migration, and pyruvate oxidation (Fig. 8d). The KEGG analysis revealed that the MAPK signalling pathway and the HIF-1 signalling pathway were enriched after PFMPCr treatment (Fig. 8e). The GSEA_GO results indicated that the gene sets related to positive regulation of cell proliferation and positive regulation of the cell cycle were upregulated after PFMPCr treatment. The GSEA_GO results revealed that genes related to the negative regulation of the mitochondrial membrane potential, such as Bnip3l, Bax, and Trpv1, were downregulated after PFMPCr treatment, which also represented an increased mitochondrial membrane potential after PFMPCr treatment. This finding was consistent with our in vitro results for JC-1 staining in cells. Moreover, the expression of genes involved in the extracellular matrix disassembly process, such as Adamts15 and Adamts5, was downregulated after PFMPCr treatment, which represented reduced ECM degradation (Fig. 8f). Chondroitin sulfate is the main glycosaminoglycan (GAG) in the articular cartilage extracellular matrix (ECM), which is critical for maintaining cartilage mechanics and load-bearing functions [23]. The GSEA_KEGG results revealed that genes related to glycosaminoglycan biosynthesis and the chondroitin sulfate/dermatan sulfate pathway, such as Chst15, Chsy1, Chst7, Chst11, Chst13, and Chst3, were upregulated after PFMPCr treatment. Moreover, the expression of genes related to glycosaminoglycan degradation, such as Arsb and Galns, was downregulated after PFMPCr treatment. The TGF-beta signalling pathway plays a pivotal role in cartilage biology and maintenance [24]. We found that genes in the TGF-beta signalling pathway, such as Acvr1c, Ppp2ca, Tnf, Gdf5, and Inhba, were upregulated after PFMPCr treatment. The gene sets related to protein anabolism (protein digestion and absorption and protein processing in the endoplasmic reticulum) and amino acid anabolism (arginine biosynthesis) were upregulated after PFMPCr treatment. The gene sets related to lipid metabolism were also affected, with the upregulation of fatty acid biosynthesis and the downregulation of fatty acid degradation. Interestingly, genes involved in the citrate cycle (TCA cycle), such as Idh1, Idh2, and Sdhc, and pyruvate metabolism, such as Adh1, Adh7, and Acacb, were downregulated after PFMPCr treatment (Fig. 8g). Our in vitro results showed that PFMPCr facilitated ATP production in chondrocytes, which increased the ATP/ADP ratio. An increased ATP/ADP ratio could negatively affect the activity of enzymes related to ATP production through negative feedback, thus inhibiting the TCA cycle [25]. Thus, after CAM treatment, genes related to inflammation and extracellular matrix degradation were downregulated in chondrocytes. The gene sets related to cell proliferation, glycosaminoglycan biosynthesis, the mitochondrial membrane potential and protein anabolism were upregulated after CAM treatment.

Fig. 8.

CAMs reprogram the transcriptome of chondrocytes. (a) qPCR results for the chondrogenic phenotypes (Col2a1, Acan, and Sox9) and inflammatory phenotypes (Mmp3 and Il6) of chondrocytes. n = 3, one-way ANOVA, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. (b) Clustering heatmap (left panel) and volcano plot (right panel) of differentially expressed genes in chondrocytes between the LPS and blank groups. (c) Clustering heatmap (left panel) and volcano plot (right panel) of differentially expressed genes in chondrocytes between the LPS + PFMPCr and LPS groups. (d) Enriched GO biological process (BP) terms between the LPS + PFMPCr and LPS groups. (e) KEGG pathway enrichment analysis between the LPS + PFMPCr and LPS groups. (f) GSEA of GO biological process terms between the LPS + PFMPCr and LPS groups. (g) GSEA of KEGG pathways between the LPS + PFMPCr and LPS groups.

Second, meniscus fibrochondrocytes were treated with LPS for 2 h to induce inflammation, followed by an incubation with CAMs for 12 h. The qPCR results demonstrated that CAM treatment could maintain the expression of chondrogenic markers (Col2a1, Acan, and Sox9) and reduce the expression of inflammatory markers (Mmp3 and Il6) in meniscus fibrochondrocytes after LPS stimulation. The addition of DS synergistically enhanced the function of CAMs (Fig. 9a). Next, a transcriptomic analysis was performed on the meniscal fibrochondrocytes of the blank, LPS, and LPS + PFMPCr groups. The gene expression profiles of meniscus fibrochondrocytes were significantly different among the groups (Fig. 9b and c; Fig. S15). The changes in the transcriptome of meniscus fibrochondrocytes after LPS stimulation are described in the Supplemental Information. Overall, the gene sets related to inflammatory responses were upregulated in meniscus fibrochondrocytes after LPS stimulation. The gene sets related to glucose metabolism, including glycolysis/gluconeogenesis, the citrate cycle (TCA cycle), oxidative phosphorylation and pyruvate metabolism, were downregulated in meniscus fibrochondrocytes after LPS stimulation, which indicated disordered energy production. Next, the transcriptomes of the meniscus fibrochondrocytes from the LPS + PFMPCr and LPS groups were further evaluated. The GO analysis revealed that positive regulation of cell proliferation was enriched in meniscus fibrochondrocytes after PFMPCr treatment, with the upregulation of Il11, Atf3, and Areg. Hyaluronan is a large but simple glycosaminoglycan that is composed of repeating D-glucuronic acid units with a beta1-3 linkage to N-acetyl-D-glucosamine beta1-4. Hyaluronan can bind to aggrecan to immobilize it, thus maintaining it at the high concentrations required for compressive resilience [26]. Moreover, hyaluronan within synovial fluid is critical for joint lubrication [27]. We found that the hyaluronan biosynthetic process was enriched in meniscus fibrochondrocytes after PFMPCr treatment, with the upregulation of Has1 and Has2. We also found that the negative regulation of hydrogen peroxide-induced neuronal death and the negative regulation of oxidative stress-induced neuronal death were enriched in meniscus fibrochondrocytes after PFMPCr treatment, with the upregulation of Nr4a2 and Nr4a3. The biological process of positive regulation of glucose transmembrane transport was enriched after PFMPCr treatment, which represented increased glucose intake. The ECM is critical for meniscus cell survival and phenotype maintenance [28]. We found that extracellular matrix assembly was enriched after PFMPCr treatment, which represented increased aggregation, arrangement and bonding together of the ECM. Moreover, energy homeostasis was enriched after PFMPCr treatment (Fig. 9d). The KEGG analysis revealed that the MAPK signalling pathway was enriched in meniscus fibrochondrocytes after PFMPCr treatment (Fig. 9e). The GSEA_GO analysis revealed that the gene sets associated with mitochondrial respiratory chain complex I assembly (Ndufs2, Tmem126a, Ndufab1, Acad9, Ndufb3, and Ndufa10l1) and mitochondrial electron transport, cytochrome c to oxygen (Cox6a2, Cox5a, Cycs, Proca1, and Lace1) were downregulated in meniscus fibrochondrocytes after PFMPCr treatment. Mitochondrial respiratory chain complex I assembly represents the assembly of the mitochondrial NADH dehydrogenase complex (ubiquinone). Mitochondrial electron transport, cytochrome c to oxygen, is defined as the transfer of electrons from cytochrome c to the oxygen that occurs during oxidative phosphorylation. Thus, PFMPCr treatment decreased ATP production through oxidative phosphorylation. This finding was identical to that observed for the chondrocytes after PFMPCr treatment as a result of the increased ATP/ADP ratio. Gene sets related to gluconeogenesis (Pgam2, Atf3, and Per2) were upregulated in meniscus cells after PFMPCr treatment, indicating the increased formation of glucose from noncarbohydrate precursors, such as pyruvate, amino acids and glycerol. The gene sets associated with extracellular matrix disassembly (Mmp13, Adamts15, Adamts5, Eng, and Plg) were downregulated after PFMPCr treatment, which represented decreased ECM degradation. Moreover, the gene sets related to the negative regulation of the proteasomal ubiquitin-dependent protein catabolic process (Fhit and Klhl40) were upregulated after PFMPCr treatment, which indicated decreased protein catabolism. The gene sets related to the regulation of mitotic spindle assembly (Hspa1a and LOC108348108) and the transforming growth factor beta receptor signalling pathway (Serpina1, Apoa1, and Fos) were upregulated in meniscus fibrochondrocytes after PFMPCr treatment (Fig. 9f). GSEA_KEGG analysis revealed that the gene sets related to pyruvate metabolism (Adh5, Ldhb, Ldhc, Pck1, and Grhpr) and the citrate cycle (TCA cycle) (Suclg1, Acly, Fh, Idh1, Mdh1, and Ogdhl) were downregulated in meniscus fibrochondrocytes after PFMPCr treatment. The results of our in vitro cellular experiment indicated that PFMPCr facilitated ATP production in meniscus fibrochondrocytes, which increased the ATP/ADP ratio. An increased ATP/ADP ratio could negatively affect the activity of enzymes related to ATP production through negative feedback, thus inhibiting the TCA cycle [25]. The gene sets related to fatty acid degradation (Acsl6, Echs1, Ehhadh, Acsbg2, Cpt1c, and Acaa1b) were downregulated after PFMPCr treatment. Moreover, genes related to carbohydrate digestion and absorption (Amy1a, Atp1b2, Hk2, Slc5a1, and Plcb2) were upregulated in meniscus fibrochondrocytes after PFMPCr treatment (Fig. 9g). Thus, CAM treatment decreased the level of oxidative stress in meniscus fibrochondrocytes. The gene sets related to cell proliferation, the hyaluronan biosynthetic process, extracellular matrix assembly and gluconeogenesis were upregulated after CAM treatment. CAM treatment also facilitated the anabolism of proteins, lipids and carbohydrates. However, ATP production through the citrate cycle (TCA cycle) and oxidative phosphorylation decreased because of the negative feedback of the increased ATP/ADP ratio after PFMPCr treatment.

Fig. 9.

CAMs reprogram the transcriptome of meniscus fibrochondrocytes. (a) qPCR results for chondrogenic markers (Col2a1, Acan, and Sox9) and inflammatory markers (Mmp3 and Il6) in meniscus fibrochondrocytes. n = 3, one-way ANOVA, ∗ represents p < 0.05, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. (b) Clustering heatmap (left panel) and volcano plot (right panel) of differentially expressed genes in meniscus fibrochondrocytes between the LPS and blank groups. (c) Clustering heatmap (left panel) and volcano plot (right panel) of differentially expressed genes in meniscus fibrochondrocytes between the LPS + PFMPCr and LPS groups. (d) Enriched GO biological process (BP) terms between the LPS + PFMPCr and LPS groups. (e) KEGG pathway enrichment analysis between the LPS + PFMPCr and LPS groups. (f) GSEA of GO biological process terms between the LPS + PFMPCr and LPS groups. (g) GSEA of KEGG pathways between the LPS + PFMPCr and LPS groups.

2.7. CAMs alleviated early knee OA

We first evaluated the tissue penetration capacity of PF and PFMPCr. PF and PFMPCr were labeled with Cy5 fluorescence and suspended in culture medium to a concentration of 200 μg/mL. Fresh cartilage and meniscus explants were first incubated with LPS for 4 h to induce inflammation, simulating the in vivo OA model. Then, the explants were incubated with the prepared mixture for 12 h. Frozen sections of cartilage and meniscus explants were prepared and scanned by confocal microscopy. Compared with that of PF, the tissue penetration capacity of PFMPCr was superior. Moreover, LPS stimulation facilitated the penetration of PFMPCr into cartilage and meniscus tissue (Fig. 10a and b). Furthermore, the affinity of artificial mitochondria for human osteoarthritic cartilage and meniscus tissue was evaluated. The frozen sections of human osteoarthritic cartilage and meniscus tissue were incubated with 200 μg/mL PF or PFMPCr for 4 h. After thorough irrigation, the frozen sections were scanned with a confocal microscope. We found that PFMPCr had more robust affinity for human osteoarthritic cartilage and meniscus tissue (Fig. 10c and d). A mixture of Cy5 fluorescence-labeled PF or PFMPCr was injected into the mice knee joint. The IVIS spectrum was used to detect the fluorescence signal intensity at different time points after a single injection. We found that the PF signal could persist for approximately 2 weeks within the knee joint. However, the PFMPCr signal could persist for 1 week within the knee joint (Fig. 10e). The relative decreased retention period of PFMPCr within knee joint compared to PF was likely the result of its high motility and metabolic efficiency. Thus, the frequency of artificial mitochondria injection in the subsequent mice study was set as 1 week. Next, the in vivo pharmacokinetic of artificial mitochondria were evaluated in mice. 10 μl of Cy5-labeled PFMPCr was injected into the right knee joint. The fluorescence intensity was detected by measuring the IVIS spectrum at predetermined time points post-intervention: 0 (baseline), 1, 2, 3, and 4 days (Fig. S16a). The in vivo clearance of Cy5-labeled PFMPCr followed a mono-exponential decay pattern (Fig. S16b). Nonlinear regression analysis using a one-phase decay model demonstrated an excellent fit to the experimental data (R² = 0.9615), indicating the model accounted for over 96% of the variance in the signal. The fitted curve yielded an apparent elimination half-life of 1.82 days (95% CI: 1.22 to 2.99 days). The corresponding elimination rate constant (K) was 0.381 day⁻¹ (95% CI: 0.232 to 0.569 day⁻¹). The detailed results of the curve fitting are summarized in Table S1. We further evaluated the penetration and retention of artificial mitochondria within the mice knee joint using immunofluorescence (Fig. 10f). The negative control of immunofluorescence was provided in Fig. S17. The mice knee joints were harvested 2 weeks after a single intra-articular injection of biotin-labeled PFMPCr and then prepared into paraffin sections. The penetration of PFMPCr into cartilage and meniscus tissues could be confirmed via histology. Moreover, we found PFMPCr could penetrate the deep zone of articular cartilage and meniscus tissue, which provided a solid in vivo basis for subsequent animal studies.

Fig. 10.

Penetration and affinity of artificial mitochondria for cartilage and meniscus tissue. (a) Penetration of artificial mitochondria into cartilage explant from SD rat. (b) Penetration of artificial mitochondria into meniscus explant from SD rat. (c) The affinity of artificial mitochondria for human osteoarthritic cartilage frozen sections. n = 14, one-way ANOVA, ∗∗∗∗ represents p < 0.0001. (d) The affinity of artificial mitochondria for human osteoarthritic meniscus frozen sections. n = 14, one-way ANOVA, ∗∗∗∗ represents p < 0.0001. (e) The IVIS spectrum was measured to evaluate the retention of artificial mitochondria within the mice knee joint after an intra-articular injection. (f) Histological evaluation of the in vivo penetration and retention of artificial mitochondria within the mice knee joint 2 weeks after a single intra-articular injection. The green fluorescence represents biotin-labeled PFMPCr. F represents the femur, T represents the tibia, M represents the meniscus, and AC represents the articular cartilage. All the scale bars are 100 μm.

In the present study, 1% (w/v) monoiodoacetic acid (MIA) was injected into the knee joint for 2 weeks to establish an early OA mice model according to a previous study [29]. The mice were subjected to sham, PBS, PF, PFMPCr, or PFMPCr/DS treatment. The mixture was injected into the knee joint every week according to the aforementioned IVIS results. All knee samples were collected after 4 weeks (Fig. 11a). Histology (safranin O staining) revealed apparent erosion and destruction of the articular cartilage and meniscus. Compared with PBS or PF treatment, PFMPCr treatment reduced the destruction of cartilage and meniscus tissue to some degree. PFMPCr/DS treatment significantly attenuated cartilage and meniscus tissue destruction and erosion. The GAG content within the cartilage and meniscus tissue was also retained prominently after PFMPCr/DS treatment (Fig. 11b). An immunohistochemical analysis of knee sections also revealed that PFMPCr/DS treatment could retain COL II and aggrecan (Fig. 11c and d). Synovial membrane inflammation is an important feature of knee OA [30]. The results of HE staining showed that PFMPCr/DS treatment alleviated hyperplasia of the synovial membrane, which represented decreased synovial inflammation (Fig. 11e). To assess the anti-inflammatory effects of CAMs, we performed IL-6 immunofluorescence on mice knee sections. PFMPCr/DS treatment significantly reduced IL-6 expression in the synovium, cartilage, and meniscus compared to the PBS- or PF-treated groups (Fig. S18). Furthermore, immunofluorescence analysis of synovial tissue revealed a shift in macrophage polarization. In the PFMPCr/DS group, anti-inflammatory M2-type macrophages (CD206⁺) predominated, a profile similar to that of the native sham group. In contrast, pro-inflammatory M1-type macrophages (CD86⁺) were the primary subset in the PBS- or PF-treated control groups (Fig. S19). Notably, pathological assessment of the CAMs-treated joints did not detect local adverse reactions, such as foreign body response or fibrosis, supporting the in vivo biosafety of the formulation. Narrowing of the knee joint space is another important characteristic of OA [31]. X-ray images revealed severe joint space narrowing in the medial and lateral knee compartments after PBS treatment. Similar to the sham group, the PFMPCr/DS treatment group maintained a normal joint space (Fig. 11f). OA Research Society International (OARSI) scoring further confirmed the ability of PFMPCr/DS treatment to alleviate knee OA (Fig. 11g). Sham group (no OA induction) serves as the baseline for normal behavior. We subsequently performed a hot plate test and a catwalk test to evaluate pain and gait. The hot plate results indicated that PFMPCr/DS treatment significantly relieved MIA-induced OA pain (Fig. 11h). The gait analysis revealed that the paw area and maximal contact intensity of the left hind limb were significantly lower in the PBS and PF treatment groups than in the sham group. However, PFMPCr/DS treatment significantly restored the paw area and maximal contact intensity (Fig. 11i–k).

Fig. 11.

Artificial mitochondria alleviate early knee OA in mice. (a) Schematic diagram of the in vivo mice study. (b) Safranin O staining of mice knee joints. Black scale bar: 0.2 mm, blue scale bar: 0.05 mm. (c) COL II immunohistochemical staining of mice knee joints. Black scale bar: 0.2 mm, blue scale bar: 0.05 mm. (d) Aggrecan immunohistochemical staining of mice knee joints. Black scale bar: 0.2 mm, blue scale bar: 0.05 mm. (e) HE staining of the synovial membrane within the mice knee joint. Scale bar: 0.1 mm. (f) X-ray of a mice knee joint. The white circles represent the medial knee compartment. (g) OARSI scoring of mice knee joint degeneration. n = 6, one-way ANOVA, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. Sham group (no OA induction) serves as the baseline for normal behavior. (h) Results of the hot plate test. n = 6, one-way ANOVA, ∗ represents p < 0.05, ∗∗∗ represents p < 0.0005, and ∗∗∗∗ represents p < 0.0001. (i) Schematic of the gait analysis of the mice. LF represents the left forelimb, LH represents the left hind limb, RF represents the right forelimb, and RH represents the right hindlimb. (j) The mean paw contact area of the left hind limb of the mice. n = 6, one-way ANOVA, ∗ represents p < 0.05, ∗∗ represents p < 0.01, ∗∗∗ represents p < 0.0005, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference. (k) Maximal contact intensity of the mice left hind limb. n = 6, one-way ANOVA, ∗ represents p < 0.05, ∗∗ represents p < 0.01, ∗∗∗∗ represents p < 0.0001, and ns represents no significant difference.

We have performed additional fluorescence imaging of major organs (liver, spleen, lung, kidney, heart, and skeletal muscle) at different time points post intra-articular injection of Cy5-labeled CAMs. The results show negligible fluorescence signals in non-target organs, confirming minimal systemic leakage (Fig. S20). These results demonstrated effective articular localization with minimal off-target accumulation in vital organs over the observed period. The histological analysis of primary organs further confirmed the biosafety of artificial mitochondria in the body (Fig. S21).

3. Discussion

This study introduces chemotactic artificial mitochondria (CAMs) as a transformative platform for correcting energy metabolic imbalances in degenerative diseases like OA. By leveraging the phosphocreatine-CK system-a potent, mitochondria-independent ATP pathway, CAMs address critical limitations of prior therapies, including mitochondrial transplantation's immunogenicity and photosynthetic systems' tissue-penetration barriers. Our study demonstrates that CAMs effectively restore cellular ATP production, scavenge reactive oxygen species (ROS), and mitigate mitochondrial dysfunction in osteoarthritic chondrocytes and meniscus fibrochondrocytes. The CK-driven chemotaxis of CAMs enables precise targeting to OA-affected cartilage and meniscus, where CK concentrations are significantly higher than in synovial fluid, ensuring localized therapeutic efficacy without systemic off-target effects. Importantly, intra-articular administration of CAMs alleviated structural and functional degeneration in a mice model of early-stage knee OA, highlighting their therapeutic potential.

Mitochondrial dysfunction underlies numerous age-related and degenerative conditions, including OA, where impaired ATP synthesis exacerbates catabolic processes and inflammation [30]. Traditional strategies to enhance ATP production, such as modulating the tricarboxylic acid (TCA) cycle or transplanting natural mitochondria, face challenges ranging from metabolic complexity to logistical impracticality [3,4]. The CAM system circumvents these hurdles by utilizing the PCr-CK pathway. By crosslinking phosphocreatine monomers with fluorinated polymers, CAMs achieve efficient ATP generation independent of mitochondrial oxidative phosphorylation. This design mirrors recent advances in bioinspired energy systems, such as photosynthetic ATP production in mammalian cells using plant thylakoids [5]. The chemotactic properties of CAMs toward CK-rich environments are particularly noteworthy. In OA, CK levels rise in cartilage and meniscus during early disease stages, creating a gradient that guides CAMs to sites of mitochondrial damage. This targeted delivery contrasts with conventional nanoparticles, which often rely on passive diffusion or non-specific uptake [32]. This CK-driven directional movement is fundamentally rooted in the specific and well-characterized enzyme–substrate interaction between creatine kinase and phosphocreatine (PCr). The reversible reaction PCr + ADP ⇌ creatine + ATP, central to cellular energy buffering, requires high-affinity binding of PCr to the active site of CK. Classical biochemical and kinetic studies-including detailed analyses of substrate synergism, competitive product inhibition, and the structural role of the active-site cysteine-have long established that CK binds its guanidino substrate (creatine/PCr) with high specificity [33]. In particular, PCr acts as a competitive inhibitor with respect to creatine, confirming that both molecules occupy the same specific binding pocket. Thus, the CK concentration gradient in early OA joints essentially establishes a chemoattractant field for PCr, analogous to chemokine-receptor systems. Our observation that PCr-containing CAMs migrate along this gradient represents a direct macroscopic manifestation of this precise biochemical recognition, leveraging CK's intrinsic substrate specificity to achieve spatially targeted delivery. However, in late OA, CK expression declines, which may reduce CAMs' targeting efficiency. This underscores the potential of CAMs for early intervention. The ability of CAMs to penetrate dense cartilage, a tissue notoriously resistant to drug delivery [34], highlights their potential to address a major unmet need in OA therapy. Moreover, the incorporation of diclofenac sodium (DS) into CAMs synergistically combines metabolic support with anti-inflammatory action, offering a dual therapeutic modality that surpasses the symptomatic relief provided by current NSAIDs.

The intracellular behavior of CAMs reveals a multifaceted mechanism of action. Upon clathrin-mediated endocytosis, CAMs escape lysosomal degradation, a feat rarely achieved by conventional nanoparticles due to lysosomal entrapment [35]. This lysosomal evasion ensures sustained release of PCr, which regenerates ATP via CK-mediated phosphate transfer. The resultant increase in ATP production not only restores anabolic processes but also downregulates energy-intensive pathways like the TCA cycle through negative feedback [25]. Transcriptomic analyses further corroborate this metabolic reprogramming, showing upregulation of glycosaminoglycan biosynthesis (e.g., Chst11, Chsy1) and suppression of matrix-degrading enzymes (e.g., MMP13, ADAMTS5) in CAM-treated cells. These changes align with the observed preservation of collagen II and aggrecan in OA joints, underscoring CAMs’ capacity to halt cartilage degeneration at the molecular level. CAMs also mitigate oxidative stress, a hallmark of OA progression. By scavenging mitochondrial ROS (mtROS) and stabilizing the mitochondrial membrane potential, CAMs counteract the vicious cycle of oxidative damage and inflammation perpetuated by dysfunctional mitochondria [36]. This dual antioxidant and bioenergetic support mirrors the effects of mitochondrial-targeted nanozymes [21], yet CAMs achieve this without relying on exogenous catalysts, enhancing their biocompatibility. The observed restoration of mitochondrial membrane potential and reduction in mtROS following CAMs treatment can be mechanistically linked to the provision of an alternative ATP source via the PCr-CK system. In OA chondrocytes and meniscal cells, mitochondrial dysfunction leads to a deficit in ATP synthesis, forcing the electron transport chain (ETC) to operate under a heightened thermodynamic burden, which promotes electron leakage and excessive ROS generation b), [36]. By directly supplying ATP independently of oxidative phosphorylation, CAMs reduce the metabolic load on compromised mitochondria. This alleviation of ETC overwork diminishes electron leakage, thereby directly lowering mtROS production. Furthermore, an increased cytosolic ATP/ADP ratio, resulting from CAMs activity, can contribute to the stabilization of the mitochondrial membrane potential (ΔΨm) by reducing the demand for proton pumping to drive ATP synthesis [25]. This “energetic respite” may allow stressed mitochondria to engage in essential quality-control processes, such as the activation of mitophagy to remove damaged components and the initiation of biogenesis for repair [2a]. Thus, CAMs not only compensate for bioenergetic failure but also create a metabolic context that favors the recovery of mitochondrial homeostasis. In addition to restoring mitochondrial bioenergetics, CAMs treatment significantly reduced both mitochondrial ROS (mtROS) and total cellular ROS levels in OA chondrocytes and meniscal fibrochondrocytes (Fig. 7). It is worth noting that the decrease in mtROS was particularly pronounced. This can be attributed to the direct role of the PCr-CK system in mitigating mitochondrial stress. By supplying ATP independently of the electron transport chain (ETC), CAMs reduce the thermodynamic burden on compromised mitochondria, thereby decreasing electron leakage-a primary source of mtROS. In contrast, total cellular ROS, as detected by the broad-spectrum probe DCFH-DA, originates from multiple sources including not only mitochondria but also NADPH oxidases (NOX) and other inflammatory enzymes upregulated in OA. The observed reduction in total ROS (Fig. 7f and g) likely reflects an indirect consequence of improved cellular energy homeostasis and the anti-inflammatory action of diclofenac sodium released from CAMs, which together downregulate pro-oxidative inflammatory signaling pathways. Thus, while CAMs exert a direct and potent effect on mtROS through bioenergetic substitution, their impact on overall oxidative stress is mediated through a combination of mitochondrial unloading and integrated anti-inflammatory modulation.

The GSH-responsive degradation of CAMs is central to their sustained release profile. It is worth noting that synovial fluid GSH concentrations may vary in clinical OA populations, influenced by factors such as disease stage, inflammation severity, and individual metabolic differences. Our in vitro release studies, conducted at a physiologically relevant GSH concentration of 100 μM, confirm the fundamental operability of this mechanism, showing coordinated and sustained release of both diclofenac sodium and phosphocreatine (Fig. S5). Importantly, the consistent therapeutic outcomes observed in our in vivo OA model-where the joint presents its own dynamic GSH microenvironment-support the robustness of CAMs under biologically variable conditions. Future pharmacokinetic studies characterizing release kinetics across a wider range of GSH concentrations could further refine the formulation for clinical translation.

The preclinical efficacy of CAMs in a murine OA model underscores their translational potential. Intra-articular CAM administration attenuated cartilage erosion, synovial hyperplasia, and joint space narrowing. Notably, CAMs restored gait function and pain thresholds, metrics directly relevant to patient quality of life. These results position CAMs as a potential candidate therapeutic drug for early-stage OA, a condition for which disease-modifying treatments remain elusive. The selective accumulation of CAMs in CK-rich tissues also minimizes off-target effects, a critical advantage over systemic therapies. The transient retention of CAMs within the joint (∼1 week) suggests that periodic injections could sustain therapeutic benefits without inducing chronic immune responses, a concern associated with protein-based therapies. Furthermore, the absence of cytotoxic effects in primary organs underscores the biosafety of this platform, a prerequisite for clinical translation. The therapeutic evaluation in our study did not include a positive control group with clinically approved symptomatic drugs (e.g., NSAIDs like celecoxib), as current first-line pharmacological interventions for OA primarily aim at pain relief and inflammation reduction without halting or reversing the underlying structural degeneration of cartilage and meniscus [13,14]. For instance, while NSAIDs can alleviate pain, they do not address the core bioenergetic deficit in OA cells and may even exacerbate cartilage breakdown upon long-term use. Therefore, they are not suitable as a disease-modifying positive control for evaluating a therapy like CAMs, which is designed to correct the fundamental metabolic dysfunction driving OA progression. Instead, our study employed the internal control PFMPCr/DS, which contains both the energy-modulating component (PCr) and the anti-inflammatory drug (DS), allowing us to dissect the contribution of bioenergetic restoration versus pure anti-inflammatory action. This comparison clearly demonstrates that the combination strategy yields superior structural and functional outcomes than either component alone, highlighting the necessity of targeting cellular energy homeostasis for effective OA therapy.

While promising, several challenges warrant further investigation. i. although BAC crosslinkers enable glutathione-responsive disassembly, residual fluorinated polymers (PFA-derived components) may pose accumulation risks in prolonged use. We note that PFA is a well-characterized fluoropolymer with established biocompatibility in biomedical applications, and its hydrophobic nature contributes to joint retention without evidence of acute systemic toxicity in our 4-week study. To address concerns regarding local reactions, we performed extended histological evaluations of injected joints. No signs of foreign body response, fibrosis, or abnormal synovial hyperplasia were observed beyond the levels seen in PBS-injected controls. Furthermore, immunofluorescent analysis showed no significant increase in pro-inflammatory macrophage infiltration (M1-type) or IL-6 expression compared to sham groups, indicating minimal immunogenic reaction to CAMs. Nevertheless, the long-term fate, potential metabolic pathways, and excretion profiles of PFA components beyond 8 weeks warrant dedicated pharmacokinetic and toxicological studies in future translational development. ii. the CAMs contained several key components, including PF, PCr and crosslinkers. Each component had corresponding genetic regulatory effects, which was difficult to be defined. iii. the current model focuses on early-stage OA, whether CAMs can reverse advanced disease, marked by extensive cartilage loss and subchondral bone remodeling, requires validation. iv. the reliance on CK gradients for targeting may limit efficacy in tissues with low CK expression, necessitating adaptations for broader applicability. Future studies should explore combinatorial strategies, such as coupling CAMs with growth factors (e.g., TGF-β) to enhance tissue repair or integrating real-time imaging agents to monitor therapeutic delivery. Scaling up production while maintaining nanoparticle uniformity will also be critical for clinical adoption. Additionally, optimizing CAM formulations for sustained release or stimuli-responsive degradation could improve therapeutic efficacy. v. comparative studies against emerging therapies, such as mitochondrial transplantation, gene editing or DS solution intervention, will clarify CAMs’ niche in the evolving OA treatment landscape. vi. the transcriptomic analyses provided a global molecular narrative consistent with the observed in vivo functional recovery (e.g., anabolism upregulation vs. cartilage preservation; inflammation downregulation vs. synovitis reduction). Future targeted proteomic studies were warranted to validate key nodal proteins within these pathways.

In conclusion, CAMs represent a breakthrough in metabolic therapy for degenerative diseases. By mimicking the PCr-CK system, they provide a simple, efficient, and targeted solution to mitochondrial dysfunction. Their ability to restore energy homeostasis, reduce oxidative stress, and modulate inflammation positions CAMs as a versatile platform not only for OA but also for other conditions linked to metabolic failure, such as neurodegenerative disorders or ischemic injuries. The administration methods could be optimized to achieve non-invasive administration, such as oral or patch. The CAM could also be modified to achieve cell-specific targeting, such as peptide or small molecule modifications. This study lays the groundwork for translational development, with the potential to redefine treatment paradigms for age-related and degenerative diseases.

4. Materials and methods

4.1. Materials

3-(4,5-dimethylthiazol-2-thiazolyl)-2,5-diphenyltetrazolium bromide (MTT) was purchased from Sigma Aldrich (475989). Rabbit muscle creatine kinase and NHS Biotin were purchased from Shanghai Yuanye Biotechnology Co., Ltd (S10076, S13005). Creatine Phosphate Sodium Salt Hydrate (PCr), 2-methylallylamine, N,N′-bis (acryloyl) cysteamine (BAC), ammonium persulfate (APS) and 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (EDC) were all purchased from Shanghai McLean Biochemical Technology Co., Ltd (C804629, D853383, N836631, A801035, N808856). Pentafluorooctyl acrylate (PFA) was purchased from Adamas Chemical Reagent Co., Ltd (013568103). N,N,N′,N′-tetramethylethylenediamine (TEMED) and N-hydroxysuccinimide (NHS) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd.(T105497, H109330). Cy5 NHS, Chlorpromazine (Chlor.), Nystatin, Cytochalasin D. (Cyto. D) were purchased from APEX BIO Technology LLC (A8108, C6410, B1993, B6645). Hoechst 33342 (nuclear dye) and 2′,7′- dichlorofluorescein diacetate (DCFH-DA, ROS probe) were purchased from Beijing Solaibao Technology Co., Ltd (C0031, D6470). 3,3′-Dioctadecyloxacarbocyanine perchlorate (Dio, cell membrane dye) and lysosome dye were purchased from Jiangsu Kaiji Biotechnology Co., Ltd (KGE2604-10, KGE2209-200). JC-1 was purchased from Nanjing Fumeisi Biotechnology Co., Ltd (FMS-FZ006). MitoSOX Red was purchased from MedChemexpress Biotechnology (HY-D1055). Diclofenac Sodium (DS) was purchased from SIGMA-ALDRICH, Co (D6899). Mito-Tracker Red CMXRos was purchased from Biyuntian Biotechnology Co., Ltd. (C1049B). The rabbit Anti-Creatine kinase B type antibody was purchased from Abcam (ab92452). The rabbit anti MMP13 antibody was purchased from Proteintech (18165-1-AP). FITC-labeled anti-biotin antibody was purchased from MilliporeSigma (F6762). TNF-α was purchased from Sangon Biotech (Shanghai) Co., Ltd. (C600158-0005). The mice anti COL II antibody was purchased from Invitrogen (MA5-13026). The mice anti Aggrecan antibody was purchased from MilliporeSigma (C8035). The CD86 antibody was purchased from Abcam (ab119857). The CD206 antibody was purchased from CST (24595S). The IL-6 antibody was purchased from Proteintech (21865-1-AP). The Goat anti-mice IgG H&L (HRP) was purchased from Abcam (ab6789). The Goat anti-rabbit IgG H&L (HRP) was purchased from Abcam (ab6721). The Donkey anti-rabbit IgG H&L (Alexa Fluor® 647) was purchased from Abcam (ab150075). The Donkey anti-rat IgG H&L (Alexa Fluor® 594) was purchased from Abcam (ab150156).

4.2. Ethics statement

All surgical procedures and postoperative care of the animals were performed according to the guidelines of Institutional Animal Care and Use Committee and was approved by the Ethics Committee of Peking University (LA2021007) and Peking University Health Science Center (DLASBD0633).

4.3. Synthesis of MPCr

PCr (255 mg, 1 mmol) was dissolved in 10 mL of PBS (pH = 7.4), and EDC (229.1 mg, 1.2 mmol) and NHS (127.5 mg, 1.2 mmol) were added to activate it. After 3 h, 2-methylallylamine (89 μL, 1 mmol) was introduced, and the reaction was incubated at room temperature for 24 h. The mixture was then purified by dialysis and lyophilized to obtain MPCr (kDa = 200).

4.4. Synthesis of PFMPCr CAMs

BAC (18.6 mg, 0.0715 mmol) was dissolved in 10 mL of PBS (pH = 7.4). MPCr (39.4 mg, 0.149 mmol) was then added and thoroughly mixed. PFA (91.9 mg, 0.21 mmol) was subsequently introduced, and the mixture was degassed under nitrogen for 30 min. The initiators APS (3 mg, 2 wt%) and TEMED (6 mg, 4 wt%) were added. After the reaction proceeded for 2 h, the product was centrifuged and washed three times (8000 rpm, 10 min each). Finally, the product was lyophilized to obtain PFMPCr CAMs.

4.5. Synthesis of PFMPCr/DS CAMs

Ten milligrams of PFMPCr CAM was weighed and dissolved in 10 mL of a PBS solution containing 50 mg of DS. After stirring for 24 h, the mixture was centrifuged and washed twice to remove excess DS, yielding PFMPCr/DS CAMs.

4.6. Synthesis of PF (PFA-only control nanoparticles)

The PFA-only nanoparticles (designated as PF), serving as the carrier control group without bioactive components (MPCr or DS), were synthesized as follows: BAC (18.6 mg, 0.0715 mmol) was dissolved in 10 mL of PBS (pH = 7.4). PFA (153.1 mg, 0.35 mmol) was added, and the mixture was degassed under nitrogen for 30 min. The initiators APS (3.5 mg, 2 wt%) and TEMED (7 mg, 4 wt%) were then introduced. After the reaction proceeded for 2 h, the product was centrifuged, washed three times (8000 rpm, 10 min each), and then lyophilized to obtain PF nanoparticles. This formulation contains only the fluorinated polymer (PFA) and crosslinker (BAC), allowing for the assessment of any non-specific effects attributable to the carrier material itself.

4.7. Synthesis of Biotin-NHS labeled samples

Ten milligrams of PFMPCr was dissolved in 10 mL of PBS, followed by the addition of EDC (10 mg, 0.052 mmol) and NHS (5.5 mg, 0.052 mmol). After 3 h, Biotin-NHS (20 μL, 1 mg/mL in DMSO) was added, and the mixture was reacted in the dark for 24 h. After the reaction was complete, the mixture was centrifuged and washed three times (8000 rpm, 10 min each) to remove excess fluorescent material, and then lyophilized to obtain Biotin-labeled PFMPCr.

4.8. Synthesis of Cy5-NHS-labeled samples

Ten milligrams of different samples were dissolved in 10 mL of PBS, followed by the addition of EDC (10 mg, 0.052 mmol) and NHS (5.5 mg, 0.052 mmol). After 3 h, Cy5-NHS (20 μL, 1 mg/mL in DMSO) was added, and the mixture was reacted in the dark for 24 h. After the reaction was complete, the mixture was centrifuged and washed three times (8000 rpm, 10 min each) to remove the excess fluorescent dye and then lyophilized to obtain Cy5-labeled samples.

4.9. Characterization

The TEM images of the samples were captured using a JEOL JEM-2100 transmission electron microscope. The zeta potential and hydrodynamic particle size of the samples were measured using a Nano-Z Zetasizer (Malvern Instruments). The ¹HNMR spectra of the samples were recorded using a Bruker Advance 400 spectrometer. Fluorescence images of the cells were obtained using a confocal laser scanning microscope (CLSM, HP Apo TIRF 100X N.A.1.49, Nikon, Ti-E-A1R, Japan). The absorbance and fluorescence intensity were detected using a multimode microplate reader (TECAN, INFINITE E PLEX, Switzerland). The motility of the CAMs was captured using an inverted fluorescence microscope (MF53-N, Guangzhou Microshot Technology Co., Ltd., China).

4.10. Drug loading capacity

Ten milligrams of PFMPCr CAM was dispersed in 10 mL of a PBS solution containing DS (with DS to PFMPCr mass ratios of 0.5:1, 1:1, 2:1, 3:1, 4:1, and 5:1). After stirring for 24 h, the mixture was centrifuged and washed, and the supernatant was collected to determine the amounts of unloaded DS by measuring the absorbance at 276 nm, from which the DS loading capacity was calculated.

4.11. Drug release and CAMs degradation characterization

10 mg of PFMPCr/DS CAMs were dispersed in 10 mL solution with or without GSH (100 μM) and placed in static mixer. The relative turbidity of the solution mixture was detected by measuring the absorbance at 660 nm using a UV spectrophotometer at different time points (1, 3, 6, 12, 24, 48, 72, 96, 120, and 144 h) to reflect the degradation of CAMs. At each time point, the supernatant was collected by centrifugation (8000 rpm, 10 min), and the DS drug release at each time point was determined by measuring the absorbance at 276 nm. Meanwhile, the concentration of phosphate in the supernatant was detected by the molybdenum blue method. The solution was replenished to original volume after each collection.

4.12. Motion capture and analysis

Chondrocytes were cultured in 10-cm dishes until they reached a density of 1 × 10⁵ cells·mL⁻¹ and were then stimulated with LPS (10 μg mL⁻¹) for 4 h to obtain prestimulated chondrocytes. The prestimulated chondrocytes were lysed and diluted with PBS (lysis buffer: PBS = 1:4, v/v) and transferred to confocal dishes. Cy5-labeled PFMPCr/DS CAM (20 μL, 1 mg mL⁻¹) was added to the PBS, while Cy5-labeled PF, PFMPCr CAM, and PFMPCr/DS CAM (20 μL, 1 mg mL⁻¹) were added to the diluted prestimulated chondrocyte lysate. The motility of the samples was recorded using an inverted fluorescence microscope (100 × objective lens), and the trajectories were manually tracked using ImageJ software. The trajectories of 10 randomly selected particles were analyzed, and their average velocity were calculated in each group.

4.13. Static chemotactic behaviour of CAMs in a Y-shaped channel

The main channel of the Y-shaped channel is 1 cm in length and 0.4 cm in width, whereas the branch channels are 0.7 cm in length and 0.3 cm in width. Agarose (10 mg) was completely dissolved in PBS (1 mL) at 90 °C. CK (1 mg mL⁻¹, 150 U/mg) was mixed with 50 μL of the still-warm but not-yet-solidified agarose and then placed at 4 °C to solidify, forming Chamber II. Agarose was mixed with 50 μL of PBS and allowed to solidify, forming Chamber III. The Y-shaped channel was prefilled with a small amount of PBS, and 20 μL of Cy5-labeled PCr (1 mg mL⁻¹) was added to Chamber I. Fluorescence images of Chambers II and III were captured at 0, 30, and 60 min using an inverted fluorescence microscope, and the normalized fluorescence intensity was calculated using ImageJ software to evaluate the chemotactic behaviour of PCr towards CK. Agarose was mixed with 50 μL of LPS (10 μg mL⁻¹)-prestimulated chondrocyte lysate and allowed to solidify, forming Chamber II. Agarose was also mixed with 50 μL of artificial synovial fluid and allowed to solidify, forming Chamber III. The Y-shaped channel was prefilled with a small amount of PBS, and 20 μL of Cy5-labeled PF, PFMPCr CAMs, or PFMPCr/DS CAMs (1 mg mL⁻¹) was added to Chamber I. Fluorescence images of Chambers II and III were captured at 0, 30, and 60 min using an inverted fluorescence microscope, and the normalized fluorescence intensity was calculated using ImageJ software to evaluate the chemotactic behaviour of PCr-containing materials towards CK-containing environments.

The effect of inflammatory cytokine TNF-α on PCr-CK chemotaxis was evaluated. Agarose was mixed with 50 μL of artificial synovial fluid containing TNF-α (10 ng mL⁻¹) and CK (1 mg mL⁻¹), allowed to solidify, and used to form Chamber II. An equal volume of agarose was mixed with 50 μL of artificial synovial fluid containing only CK (1 mg mL⁻¹) and solidified to form Chamber III. After pre-filling the Y-shaped channel with PBS, 20 μL of Cy5-labeled PFMPCr/DS nanoparticles (1 mg mL⁻¹) was added to Chamber I. Fluorescence images of Chambers II and III were acquired at 0.5, 1, and 1.5 h, and normalized fluorescence intensities were calculated to evaluate the chemotactic ability of CAMs toward the CK environment in the presence of inflammatory cytokines.

4.14. Chemotactic behavior of CAMs in a single-channel device

50 μL of chondrocyte lysate pre-treated with 10 ng mL⁻¹ TNF-α for 24 h were immobilized with agarose at the basal end of a 2-cm straight single-channel chip, and the channel was pre-filled with PBS. Subsequently, 20 μL of Cy5-labeled PFMPCr/DS or PF nanoparticles (1 mg mL⁻¹) were applied to the apical end. After 1.5 h, fluorescence intensity at the basal end was recorded using an inverted fluorescence microscope; an identical volume of Cy5-labeled formulation was then replenished at the apical end and imaged again. This “replenishment–chemotaxis–imaging” cycle was repeated three times to evaluate the sustained chemotactic performance of the nanoparticles along the CK gradient.

4.15. Dynamic chemotactic behaviour of CAMs in a microfluidic channel

A three-inlet microfluidic channel with dimensions of 2.2 cm (length) × 1.5 mm (width) × 300 μm (height) was used to evaluate the dynamic chemotaxis of the CAMs. The lysate of LPS (10 mg mL⁻¹)-prestimulated chondrocytes diluted in PBS (lysate: PBS = 1:4, v/v) was used as Buffer I; the Cy5-labeled samples in PBS were used as Buffer II; and artificial synovial fluid was used as Buffer III. The flow rate was controlled at 0.6 mL h⁻¹. Images were captured near the outlet of the microfluidic channel using an inverted fluorescence microscope (10× objective lens). The fluorescence intensity perpendicular to the flow direction was measured using ImageJ software to analyse the chemotactic displacement of the CAMs.

4.16. Chemotaxis index experiment

The chemotaxis index experiment of CAM was completed according to a previous study [18]. Briefly, the lysate of LPS-stimulated chondrocytes mixed with agarose was filled in the right reservoir of a 2-cm single-channel linear chamber to construct a concentration gradient in the horizontal direction. PF or PFMPCr was filled on the left side of the single-channel linear chamber, and the movement trajectory of nanoparticles in the channel was captured using an inverted fluorescence microscope. The speed of nanoparticles motion was recorded. The ratio of nanoparticles displacement to path length represents the chemotaxis index.

4.17. Cell viability assays

Chondrocytes and meniscus cells (5 × 10⁴ cells·mL⁻¹) were seeded in 96-well plates and cultured overnight. The cells were then incubated with culture media containing different materials (200 μL, 200 μg mL⁻¹) for 1, 2, or 3 days. After the incubation, 20 μL of MTT (5 mg mL⁻¹) was added, and the mixture was incubated with the cells for 4 h. Cell viability was determined by measuring the absorbance at 570 nm using a multimode microplate reader.

4.18. Cellular uptake behaviour

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were seeded into 6-well plates and cultured overnight. The culture medium was then replaced with medium containing LPS (10 μg mL⁻¹) and incubated with the cells for 4 h to induce damage to the cells. The cells were subsequently coincubated with normal culture medium containing different Cy5-labeled samples (200 μg mL⁻¹) for 12 h. Afterwards, the cell membranes were labeled with DiO. Fluorescence imaging was performed using CLSM (100× objective lens). The fluorescence intensity was normalized using ImageJ software to analyse the uptake of the materials.

4.19. Cellular uptake pathways

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were seeded in 6-well plates and cultured overnight. After inducing cell damage by coculturing the cells with medium containing LPS (10 μg mL⁻¹) for 2 h, the original culture medium was replaced with medium containing corresponding inhibitors, including dynasore (15 μM), rottlerin (20 μM), cytochalasin D (10 μM) and nystatin (54 μM) or with normal culture medium. The inhibitors group was incubated at 37 °C for 30 min. While, the normal culture medium group was incubated at 4 °C for 30 min. Then, the Cy5 labeled CAMs were added (200 μg mL⁻¹) and incubated in original conditions for 2 h. Finally, the cells were collected for subsequent flow cytometry analysis.

4.20. Lysosomal escape

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were cultured overnight in confocal culture dishes and then coincubated with culture medium containing LPS (10 μg mL⁻¹) for 4 h. The cells were subsequently incubated with Cy5-labeled PF or PFMPCr/DS CAMs (20 μg mL⁻¹) for 5 min, 1 h, or 6 h. The lysosomes of the cells were labeled with a lysosomal fluorescent probe, and the nuclei were stained with Hoechst 33342. Fluorescence imaging was performed using CLSM, and the colocalization of the materials with lysosomes was assessed using ImageJ software.

4.21. Colocalization of the materials with mitochondria

Chondrocytes and meniscus cells were seeded in confocal dishes at 1 × 10⁵ cells·mL⁻¹ and cultured overnight, then co-incubated with medium containing LPS (10 μg mL⁻¹) for 4 h. After removing the medium, Cy5-labeled PFMPCr/DS CAMs (200  μg mL⁻¹) were added and incubated for 1 h. Mitochondria were stained with MitoTracker, and fluorescence images were acquired by CLSM. Colocalization of the materials with mitochondria was analyzed using ImageJ.

4.22. Measurement of intracellular ATP production in vitro

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were cultured overnight in 6-well plates and then coincubated with LPS (10 μg mL⁻¹) for 4 h. The culture medium was subsequently replaced with normal culture medium containing different samples (200 μg mL⁻¹) and coincubated for 12 h. The ATP content in the cells after different treatments was detected using an ATP Assay Kit (Beyotime, Cat. No. S0026).

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were cultured overnight in 6-well plates and then coincubated with TNF-α (10 ng mL⁻¹) for 24 h. The culture medium was subsequently replaced with normal culture medium containing different samples (200 μg mL⁻¹) and coincubated for 12 h. The ATP content in the cells after different treatments was detected using an ATP Assay Kit.

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were cultured overnight in 6-well plates and then coincubated with LPS (10 μg mL⁻¹) for 4 h. The culture medium was subsequently replaced with normal culture medium containing different samples (200 μg mL⁻¹) and coincubated for 48 h. The ATP content in the cells was measured at 3, 12, 24, and 48 h after treatment.

Chondrocytes (1 × 10 ⁵ cells · mL ⁻¹) were cultured overnight in a 6-well plate and then incubated with LPS (10 μg mL⁻¹) for 4 h. Subsequently, replace the culture medium with normal medium containing 50 μM Ompenaclid (HY-W015828, MCE) or different samples (200 μg mL⁻¹) or normal medium containing the inhibitor Ompenaclid and different samples (200 μg mL⁻¹) and incubate together for 48 h, and measure the ATP content in the cells.

4.23. In vitro ROS scavenging assay

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were cultured overnight in a confocal dish and then coincubated with LPS (10 μg mL⁻¹) for 4 h. After the chondrocytes and meniscus cells were incubated with different materials for 12 h, the cells were stained with MitoSOX-Red and DCHF-DA to detect mitochondrial and overall cellular reactive oxygen species (ROS), respectively. Fluorescence imaging of MitoSOX Red in the cells was performed using CLSM, and the fluorescence intensity was normalized using ImageJ software. The fluorescence intensity of DCHF-DA in the cells was quantified using a multimode microplate reader (λex = 480 nm, λem = 525 nm).

4.24. Amelioration of mitochondrial dysfunction in vitro

Chondrocytes and meniscus cells (1 × 10⁵ cells·mL⁻¹) were cultured overnight in 6-well plates and then coincubated with LPS (10 μg mL⁻¹) for 4 h. After the cells were incubated with different samples (200 μg mL⁻¹) for 12 h, they were stained with JC-1 for 30 min. The JC-1-stained cells were collected and scanned for fluorescence spectra using a multimode microplate reader at 490 nm excitation. The full fluorescence spectrum was obtained, with the fluorescence intensity at 598 nm representing the aggregate form of JC-1 in the cells and that at 542 nm representing the monomeric form of JC-1.

4.25. Human knee samples

The human knee samples were harvested from OA patients (female, 50–60 years old, race: Han nationality) during total knee replacement. The synovial fluid was collected using a syringe. The removed cartilage and meniscus tissue were collected. All the samples were temporarily stored at 4 °C. After the operation, the samples were transferred to the laboratory and stored at −80 °C for subsequent use. The use of the patients’ samples was approved by the Ethics Committee of Peking University (Approval number: LA2021007). The synovial fluid, articular cartilage and meniscus tissue were prepared for frozen sections. For cartilage and meniscus, the samples were frozen at −80 °C for subsequent use. The synovial fluid was injected into the embedding box (7∗7∗5 mm) and then frozen at −80 °C for subsequent use.

The CK expression between human OA and normal cartilage tissues was evaluated. In the present study, the tibial plateau cartilage was harvested from OA patients during total knee replacement. The relative normal and degenerated cartilage regions from macroscopic were harvested and fixed with 4% Paraformaldehyde. After decalcification, the cartilage paraffin sections of 4 μm were prepared. The histology (HE staining) was performed to confirm macroscopic findings. The CK immunofluorescence was performed followed by semi-quantitative analysis.

4.26. CK evaluation in mice OA models

In the present study, the post-traumatic mice OA model and MIA chemically induced inflammatory OA model were prepared. Firstly, the early post-traumatic OA model was generated via medial meniscectomy for 5 weeks. The late post-traumatic OA model was generated via medial meniscectomy for 12 weeks. CK immunofluorescence staining was performed on paraffin sections of mice knees. Secondly, the MIA chemically induced inflammatory OA model was prepared. 10 μl 1% (w/v) monoiodoacetic acid (MIA) was injected into mice knee joint. At 6 weeks after injection, the CK expression within articular cartilage and meniscus was evaluated using immunofluorescence.

4.27. ELISA

Preparation of articular cartilage and meniscus tissue lysate. i. Tissue grinding and powder collection: a. Place fresh frozen articular cartilage and meniscus pieces (cleaned of subchondral bone and synovium) into the grinding jar pre-cooled with liquid nitrogen. b. Operate the cryogenic mill according to the manufacturer's instructions to obtain a fine, homogeneous powder. c. Keep all tools and containers in contact with the powder pre-cooled. Rapidly transfer the tissue powder to pre-chilled microcentrifuge tubes using a cooled microspatula. Tubes can be stored at −80 °C if not processed immediately. ii. Tissue weighing: a. Working quickly on dry ice metal rack, briefly open the tube and accurately weigh the amount of tissue powder. Record the wet weight (W_tissue). b. Immediately return the tube to ice. iii. Protein extraction: a. Add ice-cold supplemented lysis buffer to the powder. b. Vortex vigorously for 15-20 s to thoroughly suspend the powder. c. Incubate the suspension on ice for 30 min, vortexing briefly every 10 min. d. Sonication. e. Centrifuge the lysate at 12,000×g for 15 min at 4 °C. f. Carefully collect the supernatant (the soluble protein fraction) into a new, pre-chilled microcentrifuge tube. Avoid the pellet (insoluble debris) and any lipid layer on top. Discard the pellet. g. The supernatant are the cartilage and meniscus tissue lysate. Proceed to quantification or aliquot and store at −80 °C. Avoid repeated freeze-thaw cycles.

Preparation of synovial fluid samples. Clarification: a. Thaw synovial fluid samples on ice. b. Centrifuge the synovial fluid at 2000×g for 10-15 min at 4 °C to remove cells, debris, and large particulates. c. Carefully transfer the clear supernatant to a new tube. This clarified synovial fluid can be used directly.

Protein concentration determination and sample normalization. i. Quantify total protein: a. Dilute the cartilage or meniscus tissue lysate in lysis buffer appropriately to fall within the linear range of protein assay (BCA assay). b. Perform the protein assay according to the kit's instructions, using BSA standards. c. Calculate the total protein concentration (C_total) of the undiluted lysate. ii. Normalize and prepare samples for ELISA: a. For Cartilage or meniscus lysates: Dilute all lysate samples with the ELISA kit's sample diluent to a uniform total protein concentration. Record the final dilution factor (DF_lysate). b. For synovial fluid: Dilute the clarified fluid with the ELISA kit's sample diluent. The dilution factor (DF_SF) should be determined by a pre-test to ensure the measured CK concentration falls mid-range of the ELISA standard curve. c. All samples are now ready for the ELISA procedure.

Briefly, the plates were stabilized at room temperature for 20 min and then removed from the aluminium foil bag. Standard wells and sample wells were included. A volume of 50 μl of standard samples of different concentrations was added to each standard well. First, 10 μl of the test sample was added to the sample well, and then 40 μl of sample mixture was added. Next, 100 μl of horseradish peroxidase (HRP)-labeled detection antibody was added to the standard and sample wells, followed by sealing with a membrane. The plates were incubated at 37 °C for 60 min. The reaction mixture was discarded, followed by vigorous patting. The well was filled with detergent and incubated for 1 min. The detergent was removed, and the plate was patted vigorously. The rinsing process was repeated 5 times. Fifty microlitres of substrates A and B were added to each well and incubated at 37 °C in the dark for 15 min. Fifty microlitres of termination solution was added to each well, and the OD value of each well was measured at a wavelength of 450 nm within 15 min. A linear regression curve of the standard sample was drawn according to the OD values and concentrations of the standard samples. Finally, the concentration of CK in each sample was calculated according to the equation for the standard curve. The dilution factor should be included to calculate the final concentration of CK in each sample.

4.28. Isolation and expansion of rat chondrocytes and meniscus fibrochondrocytes

The articular chondrocytes and meniscus fibrochondrocytes were harvested from the rats. The SD rats (male, 8 weeks) were sacrificed. The superficial articular cartilage layer of the femoral head and meniscus tissue in the knee joint were harvested using sterile instruments. The cartilage and meniscus were rinsed with cold sterile PBS to remove blood clots and then cut into approximately 0.1 mm pieces. The tissue pieces were digested with 0.2% type I collagenase and 0.2% type II collagenase (Gibco, USA) for 2 h at 37 °C. The mixture was filtered with an 80 μm strainer, after which a single-cell suspension was prepared. The isolated cells were seeded in culture dishes and cultured with complete medium (α-MEM + 10% FBS + 0.1% penicillin‒streptomycin). Expanded chondrocytes and meniscus fibrochondrocytes at passage 2 were used for subsequent experiments.

4.29. Western blot

SDS‒PAGE (sodium dodecyl sulfate‒polyacrylamide gel electrophoresis) was performed to evaluate the relative contents of CK and MMP13 in rat chondrocytes and meniscus fibrochondrocytes after stimulation with LPS for different durations (0.5, 1, 2, 4, and 8 h). Total protein was extracted using RIPA lysis buffer (C1053, Applygen, China) containing protease inhibitors (P1265-1, Applygen, China) and phosphatase inhibitors (P1260-1, Applygen, China). The concentration was measured using a NanoDrop spectrophotometer (Thermo Fisher, USA) at 280 nm. The proteins were electrophoretically separated (Bio-Rad, USA) on a 4–20% Bis-Tris polyacrylamide gel (M00930, GenScript, China) and then transferred onto polyvinylidene fluoride (PVDF) membranes (ISEQ00010, Millipore, USA) using the eBlot™ L1 fast wet transfer system (GenScript, USA) according to standard procedures. The membranes were blocked with 5% (w/v) bovine serum albumin (BSA) (P1621, Applygen, China) for 1 h at room temperature. The membranes were incubated with primary antibodies overnight at 4 °C. After being rinsed with TBST, the membranes were incubated with secondary antibodies for 1 h at room temperature. After thorough rinsing with TBST, the ECL ultrawestern horseradish peroxidase (HRP) substrate was used to develop the colour. Finally, the protein signals were captured with a ChemiDocXRS + Imaging System (Tanon, Shanghai, China).

4.30. Quantitative real-time polymerase chain reaction (qPCR)

Rat chondrocytes or meniscus fibrochondrocytes were treated with LPS (10 μg mL⁻¹) for 2 h, followed by an incubation with artificial mitochondria (200 μg mL⁻¹) for 12 h. Total RNA was extracted using TRIzol reagent (15596018, Invitrogen, USA) according to the manufacturer's instructions. The concentration was measured using a NanoDrop spectrophotometer (Thermo Fisher, USA). The reverse transcription process was performed using a commercial kit (R323-01, Vazyme, Nanjing, China). qPCR was performed by amplifying 20 μl of diluted complementary DNA with a SYBR Green Q-PCR Kit (Q141-03, Vazyme, Nanjing, China) using the Applied Biosystems StepOnePlus Real-Time PCR System (Foster City, CA, USA). The target mRNA level was determined using the 2^−ΔΔCT method. The primer sequences for the chondrogenic markers (COL2A1, ACAN, and SOX9) and inflammatory markers (MMP3 and IL6) are summarized in Table S3.

4.31. Single cell RNAseq analysis on OA mice knee

The single cell RNAseq data of different OA mice knee models were accessed from a previous study [16]. A total of six mice models were included: MN (male normal mice), MB (male bipedal mice), FN (female normal mice), FB (female bipedal mice), FO (female mice with bilateral ovariectomy), and FBO (female bipedal mice with bilateral ovariectomy). The forelimbs and tails were amputated to develop bipedal model at the age of 3 weeks. Bilateral ovariectomy was performed in half of the female mice at the age of 10 weeks. The bipedal model was selected to increase mechanical load on knees, which was validated to develop cartilage deterioration and subchondral bone alteration resembling knee OA patients. The femoral condyles of five mice in each model were harvested at the age of 22 weeks for single cell RNAseq. The single cell RNAseq data could be accessed under the accession number GSE267616 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE267616). A total of 82,083 cells were collected in all mice models. A total of 15 celltypes were identified, including Endothelial cell (EndC, Flt1, Cdh5, Cldn5), Chondrocyte (Cytl1, Chad, Ecrg4), Reticular_cell (Ghr, Cxcl12, Csmd1), Erythrocyte (Hba-a1, Hbb-bs, Hbb-bt), Vascular smooth muscle cell (VSMC, Tagln, Gm13889, Rgs5), Progenitor_cell (Cxcl1, Ccl7, Clec3b), Osteoblast (Bglap, Bglap2, Col1a1), Dendritic_cell (Bst2, Ccr9, Cox6a2), Monocyte (Ctla2a, Adgrl4, Cdk6), Macrophage (Pid1, Cxcl3, Slpi), B_cell (Bach2, Ebf1, Flt1), T_cell (Ccl5, Camk4, Itk), Granulocyte_1 (S100a8, S100a9, Retnlg), Granulocyte_2 (Elane, Prtn3, Mpo), Granulocyte_3 (Prss34, Mcpt8, Ifitm1).

4.32. RNA sequencing

Rat chondrocytes or meniscus fibrochondrocytes were treated with LPS (10 μg mL⁻¹) for 2 h, followed by an incubation with artificial mitochondria (200 μg mL⁻¹) for 12 h. Total RNA was extracted from the tissue using TRIzol (15596018, Invitrogen, USA) according to the manufacturer's instructions. The library was prepared using an Optimal Dual-mode mRNA Library Prep Kit (BGI-Shenzhen, China). A certain amount of RNA was denatured at a suitable temperature to open the secondary structure, and the mRNA was enriched with oligo (dT)-attached magnetic beads. After reacting at a suitable temperature for a fixed time period, the RNAs were fragmented with fragmentation reagents.

First-strand cDNA was subsequently generated using random hexamer-primed reverse transcription, followed by second-strand cDNA synthesis. The synthesized double-strand cDNA was subjected to an end repair reaction. After cDNA end repair, a single ‘A’ nucleotide was added to the 3′ ends of the blunt fragments through an A tailing reaction. The reaction system for adaptor ligation was then configured to ligate adaptors with the cDNAs, and finally, the library products were amplified through PCR and subjected to quality control. Next, the single-stranded library products were produced by denaturation. The reaction system for circularization was established to obtain single-stranded cyclized DNA products. Any uncyclized single-stranded linear DNA molecules were digested. The final single-strand circularized library was amplified with phi29 and rolling circle amplification (RCA) to generate a DNA nanoball (DNB), which carried more than 300 copies of the initial single-stranded circularized library molecule. The DNBs were loaded into the patterned nanoarray, and PE 100/150 base reads were generated with the G400/T7/T10 platform (BGI-Shenzhen, China).

The sequencing data were filtered with SOAPnuke [37] by (1) removing reads containing sequencing adapters; (2) removing reads whose low-quality base ratio (base quality less than or equal to 15) was greater than 20%; and (3) removing reads whose unknown base ('N' base) ratio was greater than 5%. Afterwards, clean reads were obtained and stored in FASTQ format. The subsequent analysis and data mining were performed using the Dr. Tom Multiomics Data mining system (https://biosys.bgi.com). The clean reads were mapped to the reference genome (Rattus norvegicus, GCF_000001895.5_Rnor_6.0) using HISAT2 [38]. Next, Ericscript (v0.5.5) [39] and rMATS (V4.1.2) [40] were used to detect fusion genes and differentially spliced genes (DSGs), respectively. Bowtie2 [41] was applied to align the clean reads to the gene set, in which known and novel, coding and noncoding transcripts were included. The expression levels of genes were calculated with RSEM (v1.3.1) [42]. A heatmap was drawn with pheatmap (v1.0.12) according to the differences in gene expression among different samples. Essentially, a differential expression analysis was performed using DESeq2 (v1.34.0) [43], with a Q value ≤ 0.05 (or FDR ≤0.001). GO (http://www.geneontology.org/) and KEGG (https://www.kegg.jp/) enrichment analyses of annotated differentially expressed genes were performed with Phyper using a hypergeometric test to obtain insights into the changes in phenotypes. The significance levels of terms and pathways were corrected by the Q value with a rigorous threshold (Q value ≤ 0.05). Gene set enrichment analysis (GSEA) is a computational method that determines whether an a priori defined set of genes shows statistically significant, concordant differences between two biological states (e.g., phenotypes). In the present study, GSEA was performed.

4.33. Incubation of artificial mitochondria with cartilage and meniscus tissues

First, the ability of artificial mitochondria to penetrate fresh cartilage or meniscus tissue was evaluated. Fresh cartilage explants were harvested from the superficial cartilage layer of the SD rat femoral head. Fresh meniscus explants were harvested from the knee joints of SD rats. Fresh cartilage and meniscus explants were incubated with 10 μg mL⁻¹ LPS for 4 h at 37 °C to simulate osteoarthritis. The control group was incubated with PBS. The explants were subsequently incubated with 200 μg mL⁻¹ Cy5-labeled PF or PFMPCr for 12 h at 37 °C. The explants were cut into 10 μm thick frozen sections and scanned with a confocal microscope.

Second, the affinity of artificial mitochondria for human osteoarthritic cartilage and meniscus tissue was evaluated. Human cartilage and meniscus tissue were prepared into 50 μm thick frozen sections and then incubated with 200 μg mL⁻¹ Cy5-labeled PF or PFMPCr for 4 h at 37 °C. After thorough irrigation with PBS, the frozen sections were scanned with a confocal microscope. The fluorescence intensity was calculated using ImageJ software. The relative affinities of artificial mitochondria for human osteoarthritic cartilage and meniscus tissue were subsequently evaluated.

4.34. In vivo penetration, retention and pharmacokinetic of artificial mitochondria

The in vivo penetration and retention of artificial mitochondria were evaluated in mice (8 weeks, male). First, 10 μl of Cy5-labeled PF or PFMPCr was injected into the knee joint. The fluorescence intensity in the knee joint was detected by measuring the IVIS spectrum at 1 d, 1 week, 2 weeks, and 7 weeks after the injection. Second, the penetration and retention capacity of artificial mitochondria within the mice knee joint was evaluated via histology. A volume of 10 μl of biotin-labeled PFMPCr was injected into the knee joints of the mice. Two weeks after a single injection, the knee joint was collected, and paraffin-embedded sections were prepared. The FITC-labeled antibiotin antibody was incubated with the paraffin-embedded sections, which were then scanned with a confocal microscope. The green fluorescence represents PFMPCr.

10 μl of Cy5-labeled PFMPCr was injected into the right knee joint. Fluorescence signals of right knee joint and major organs (quadriceps of left knee, liver, spleen, heart, kidney and lung) were quantified at 3 and 5 days post-injection using an IVIS imaging system.

The in vivo pharmacokinetic of artificial mitochondria were evaluated in mice (8 weeks, male). First, 10 μl of Cy5-labeled PFMPCr was injected into the right knee joint. The fluorescence intensity of right knee joint was detected by measuring the IVIS spectrum at predetermined time points post-intervention: 0 (baseline), 1, 2, 3, and 4 days. During each session, animals were anesthetized and placed in the imaging chamber. Identical imaging parameters (e.g., exposure time, binning, f/stop) were maintained across all time points and subjects to ensure consistency. Region-of-interest (ROI) analysis was applied to the resulting images to quantify the total radiant efficiency (photons/sec/cm²/steradian). For each animal, the signal at each time point was normalized to its respective baseline (Day 0) value and expressed as a percentage of the initial signal. The clearance kinetics were analyzed by nonlinear regression using GraphPad Prism software. The one-phase exponential decay model was selected. The model was defined by the equation: Y = Span × exp (-K × X), where Y is the signal percentage, X is the time in days, Span is the initial signal amplitude, and K is the apparent elimination rate constant. The plateau was constrained to 0%, reflecting the assumption of complete signal clearance over time, and K was constrained to be > 0. The goodness-of-fit was assessed by the coefficient of determination (R²). The model provided the best-fit value and the 95% confidence interval (CI), calculated using the profile likelihood method, for key parameters: the elimination rate constant (K), the apparent elimination half-life (t½, calculated as ln(2)/K), and the initial span.

4.35. Treatment with artificial mitochondria in mice of early knee osteoarthritis

In the present study, a total of 40 mice (8 weeks, male) were included. The mice were randomly divided into the following groups: the PBS, PF, PFMPCr, PFMPCr/DS and sham groups. Each group included 8 mice. In the PBS, PF, PFMPCr, and PFMPCr/DS groups, 10 μl of 1% (w/v) monoiodoacetic acid was injected into the left knee joint for 2 weeks to induce early osteoarthritis. A volume of 10 μl of PBS, PF, PFMPCr, or PFMPCr/DS was injected into the left knee joints of the corresponding groups every week for a total of 4 weeks. The hot plate test and gait analysis were performed before the animals were euthanized. The knee joint samples were subsequently harvested and scanned by X-ray to evaluate the knee joint space. Five-micron-thick paraffin sections were prepared for subsequent histological and immunohistochemical analysis. Safranin O staining was performed to evaluate the morphology and GAG content of the knee joint. Immunohistochemistry for COL II and aggrecan was performed. HE staining was performed to evaluate the synovial membrane within the knee joint. The Osteoarthritis Research Society International (OARSI) scoring system was used to evaluate knee joint degeneration.

4.36. In vivo evaluation of the biosafety of artificial mitochondria

After the mice were sacrificed, the main organs (liver, lung and kidney) were collected and fixed with 4% paraformaldehyde. HE staining was performed to evaluate pathological changes.

4.37. Tissue immunohistochemistry and immunofluorescence staining

For tissue immunohistochemistry, the paraffin sections were immersed in xylene and graded ethanol solutions to deparaffinize and hydrate the sections. Then, antigen retrieval was completed by incubating the sections with pepsin for 1 h at 37 °C. The intrinsic peroxidase activity was blocked with 3% hydrogen peroxide for 15 min. Goat serum (Boster, AR0009, China) was incubated with the tissue sections for 1 h at room temperature to block nonspecific binding. The tissue sections were incubated with primary antibodies overnight at 4 °C. The tissue sections were incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature after thorough rinsing with PBST. After thorough washing with PBST, the reagents from the diaminobenzidine (DAB) substrate kit were added to develop the colour. The nuclei were stained with haematoxylin. Finally, the slices were scanned with a digital scanner (NanoZoomer, Hamamatsu, Japan).

For tissue immunofluorescence staining, the steps before the secondary antibody incubation were mostly identical to those of tissue immunohistochemistry. The hydrogen peroxide incubation was omitted. After an incubation with fluorescent dye-conjugated secondary antibodies for 1 h at room temperature, the slices were sealed with an anti-fluorescence quenching agent containing DAPI. Finally, the slices were scanned with a confocal microscope (TCS-SP8 DIVE, Leica, Germany).

4.38. Mice behavioural tests

A hot plate test was performed to evaluate osteoarthritis-related pain. The mice were held on the hot surface of the plate using a glass cylinder, and the temperature of the plate was maintained at 55 ± 0.5 °C to measure the reaction latency in the hot plate test. The period (seconds) between the placement of the mice and the occurrence of paw shaking, licking, or jumping behaviour was recorded as an indicator of the response latency. The hot plate test was completed using an intelligent hot plate instrument (ZS-INP, Beijing Zhongshi Dichuang Technology Development Co., Ltd.). The mice gait analysis was completed using an automated gait analysis system (ZS-BT/S, Beijing Zhongshi Dichuang Technology Development Co., Ltd.). A video camera was used to record the process from below while each mice freely walked through the illuminated gate platform. The software completed the statistical analysis based on the footprints and weight distributions.

4.39. Statistical analysis

The data are represented with mean values ± standard deviation (SD). The individual data are also displayed. The Shapiro-Wilk test is used to evaluate data distribution. The equal variance of data is checked before analysis. The unpaired t-test is performed for the comparison of two groups. The ordinary one-way ANOVA or two-way ANOVA with Bonferroni multiple comparison test is performed for the comparison of multiple groups. The statistical significance is considered when P value < 0.05. The statistical analysis is completed by GraphPad Prism software (version 8.0.1, USA).

CRediT authorship contribution statement

Wenqiang Yan: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Yu Chen: Writing – review & editing, Writing – original draft, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation. Haoda Wu: Writing – review & editing, Writing – original draft, Validation, Software, Investigation, Formal analysis, Data curation. Zeyuan Gao: Data curation, Investigation, Visualization, Writing – original draft. Xiaoqing Hu: Writing – review & editing, Writing – original draft, Validation, Supervision, Data curation. Yingfang Ao: Writing – review & editing, Writing – original draft, Validation, Supervision, Data curation. Chun Mao: Writing – review & editing, Writing – original draft, Validation, Supervision, Project administration, Funding acquisition, Formal analysis. Mimi Wan: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Ethics approval and consent to participate

All surgical procedures and postoperative care of the animals were performed according to the guidelines of Institutional Animal Care and Use Committee and was approved by the Ethics Committee of Peking University (LA2021007) and Peking University Health Science Center (DLASBD0633).

Funding declaration

The work was supported by National Natural Science Foundation of China [No. 52422306 (the Excellent Young Scholars NSFC), 82502881, 22275095, 22175096, 22475103].

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:Reports a relationship with that includes:. Has patent pending to. If there are other authors, they 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 are the Supplementary data to this article:

Data availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.