Extracellular vesicles derived from engineered BMSCs improve damaged cartilage in mice with osteoarthritis by delivering PBX1

Abstract

Background

The beneficial effects of bone marrow mesenchymal stem cells (BMSCs) have been linked to their secreted extracellular vesicles (EVs). These EVs can suppress cartilage degradation and prevent apoptosis in articular cells. We aimed to explore the therapeutic efficacy of engineered BMSCs-EVs in the treatment of osteoarthritis (OA).

Methods

The isolation of BMSC-derived EVs was conducted via ultracentrifugation techniques, followed by characterization. Engineered EVs were obtained by transfecting chondrocyte-affinity peptide (CAP) and Pre-B-cell leukemia transcription factor 1 (PBX1) expression plasmids into BMSCs. OA injury was induced by the treatment of chondrocytes with IL-1β, and the OA mouse model was established by DMM surgery. The efficacy of different EVs was examined in vitro and in vivo. EVs treatment and combined genetic interventions were performed in vitro and in vivo.

Results

Both CAP modification and increased protein loading of PBX1 strengthened the protective effect of EVs on chondrocytes and ameliorated OA-induced cartilage damage. PBX1 promoted extracellular signal-regulated kinase (ERK) signaling and the expression of the downstream molecule c-FOS in chondrocytes by activating the transcriptional expression of calcium channel voltage-dependent subunit beta 4 (CACNB4). The chondroprotective effect of engineered EVs was significantly attenuated by inhibiting CACNB4/ERK signaling in IL-1β-treated chondrocytes, as well as in the DMM-induced mouse model.

Conclusion

Engineering (CAP modification and increased PBX1 protein loading) enhances the therapeutic efficacy of BMSCs-EVs on chondrocytes, which depends on the activation of the CACNB4/ERK signaling. Engineered EVs represent a promising treatment for OA cartilage damage.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13287-026-05152-9.

Introduction

Osteoarthritis (OA), a multifaceted, heterogeneous disease that affects an entire joint, including the bone, cartilage, ligaments, and muscles, is a major cause of disability and premature work loss [1, 2]. In 2020, approximately 7.6% of the global population, a total of 595 million individuals, were living with knee osteoarthritis OA [3], and the primary risk factors include age, female sex, obesity, and articular trauma [4]. It can affect various joints, including those in the knees, hands, and hips, and a set of guidelines has been developed that advocates a multifaceted approach, including exercise regimens, dietary modifications, weight management, as well as pharmacological therapies to improve the quality of life of patients diagnosed with OA [5]. Cartilage demonstrates a limited intrinsic capacity for repair following injury or during disease, which can be partially attributed to the limited activity of chondrocytes and the avascular nature of the tissue [6]. The exceptional mechanical properties of articular cartilage are attributable to its highly organized extracellular matrix (ECM), the composition of which is principally collagen type II [7]. Therefore, the development of efficacious therapeutic interventions capable of facilitating cartilage regeneration may represent a novel therapeutic strategy for the management of OA.

Mesenchymal stem cells (MSCs) have garnered recognition for their capacity for differentiation and immunomodulatory effects, while challenges such as restricted post-injection survival and heterogeneous therapeutic outcomes impede the advancement of MSC clinical applications [8]. Extracellular vesicles (EVs) have been demonstrated to facilitate the transmission of genetic information from donor cells, thereby regulating intercellular communication and serving as a functional paracrine factor of stem cells [9]. It is acknowledged that EVs possess an array of advantageous characteristics, including high biocompatibility, enhanced permeability across biological barriers, reduced toxicity, and minimal immunogenicity, while naturally produced EVs inherently lack targeting specificity due to their extensive distribution throughout vital organs following systemic administration [10]. Various engineering techniques are employed to fulfill targeting functions in the context of EV engineering, encompassing genetic engineering, membrane fusion, chemical modification, and physical modification [11]. For instance, engineered EVs by chondrocyte affinity peptide (CAP) silencing MMP13 can decrease MMP13 expression and upregulate COL2A1 expression, resulting in reorganization of the cartilage matrix and alleviation of progression in the OA model [12]. Here, we sought to decipher new molecules that might be loaded into the engineered EVs derived from bone marrow MSCs (BMSCs) to strengthen the efficacy of BMSC-EVs. Members of the Pre-B cell leukemia (PBC) transcription factor family are evolutionarily conserved [13]. However, the specific role of Pre-B-cell leukemia transcription factor 1 (PBX1) in OA remains to be elucidated. In this study, we aimed to evaluate the chondroprotective effects of engineered EVs loaded with PBX1 in OA and elucidate the underlying molecular mechanism.

Materials and methods

Cell cultures and treatment

Primary human BMSCs (HUXMA-01001) and BMSC medium (HUXMA-90011) were purchased from Cyagen (Suzhou, Jiangsu, China). Human primary chondrocytes (Delf-10715), human synovial fibroblasts (Delf-10701), chondrocyte medium (Delf-26229), and synovial fibroblast culture medium (Delf-25663) were purchased from Wanwu Shengwu (Hefei, Anhui, China). Cells were cultured at 37℃ under 5% CO₂. Cells at P1-P2 were used in the experiment. Protein constructs consisting of CAP (sequence: DWRVIIPPRPSA) or a randomly scrambled peptide (SP, sequence: ARDWPIRPVPIS) fused to LAMP2B (CAP-LAMP2B, SP-LAMP2B), PBX1/MEK1 expression vectors, the shRNA plasmid of CACNB4, and the control plasmid were constructed by WZ Biosciences (Jinan, Shandong, China). Plasmids were transfected into cells using Lipofectamine 3000 (L3000015, Thermo Fisher Scientific Inc., Waltham, MA, USA), and subsequent experiments were performed 48 h later. The sequence of shRNAs targeting CACNB4 was as follows: shCACNB4 #1: CTAAGAGGTCTGTCCTAAATACTCGAGTATTTAGGACAGACCTCTTAG; shCACNB4 #2: CTTGAGTCTAGATGGATATTACTCGAGTAATATCCATCTAGACTCAAG.

Chondrocytes were treated with recombinant human IL-1β protein (HY-P70586, MedChemExpress, Monmouth Junction, NJ, USA) at different concentrations (2.5, 5, 10, and 20 ng/mL) for 24 h to induce OA-associated chondrocyte injury [14]; cells treated with PBS served as controls. Different doses of EVs (2.5, 5, 10 μg/mL) were supplemented to chondrocytes with IL-1β and incubated for 24 h.

Characterization of BMSCs

BMSCs (5 × 10⁵) were suspended in 50 μL of flow cytometry assay cell staining buffer and incubated with FITC-coupled anti-CD29 (AC0282), anti-CD44 (AC0303), anti-CD73 (AC0346), anti-CD105 (AC0360), anti-CD34 (AC0288), anti-CD45 (AC0305), or isotype control IgG, respectively, at 4ºC for 30 min in the dark. Relevant antibodies were obtained from Beyotime (Shanghai, China). Detection was performed by flow cytometry, and the expression of relevant antigens in the cells was visualized based on an isotype control.

BMSCs were plated in 48-well plates at 1 × 10⁵ cells/well. For osteogenic differentiation, BMSCs were cultured for 21 d using the osteogenic-induced differentiation kit (HUXMX-90021, Cyagen) and stained with Alizarin Red for 10 min to observe the mineralization of the cells under a light microscope. For adipogenic differentiation, BMSCs were incubated with the adipogenic induction kit (HUXMX-90031, Cyagen) for 14 days, followed by staining with Oil Red O solution for 30 min, and the accumulation of intracellular lipid droplets was visualized under a light microscope. Cell differentiation was also induced by chondrogenic induction of differentiation kit (HUXMX-90041, Cyagen) for 14 days, and chondrogenic differentiation of BMSCs was observed using Alcian blue solution staining for 30 min.

Western blot analysis

Samples were lysed with RIPA buffer (R0010, Solarbio, Beijing, China) supplemented with PMSF (1 mM), and the lysates were centrifuged at 10,000 g for 5 min to harvest the supernatant. Protein concentration was quantified by the BCA kit (23,225, Thermo Fisher). An equal amount (20 μg) of protein was then subjected to SDS-PAGE and transferred to a PVDF membrane. The unbound sites on the membrane were blocked with 5% BSA. Primary antibodies were used to target specific proteins, including LAMP2B (1:1000, ab18529, Abcam, Cambridge, MA, USA), GAPDH (1:10,000, K110496P, Solarbio), CD9 (1:5000, 20597-1-AP, ProteinTech Group, Chicago, IL, USA), CD81 (1:1000, 27855-1-AP, ProteinTech), TSG101 (1:2000, 28283-1-AP, ProteinTech), Calnexin (1:5000, 10,427-2-AP, ProteinTech), Cleaved-caspase3 (1:1000, 25128-1-AP, ProteinTech), Bax (1:5000, 50599-2-Ig, ProteinTech), IL-6 (1:2000, A11115, ABclonal, Wuhan, Hubei, China), TNF-α (1:1000, ab183218, Abcam), PBX1 (1:1000, 18204-1-AP, ProteinTech), NFIC (1:1000, 16399-1-AP, ProteinTech), CACNB4 (1:2000, 17770-1-AP, ProteinTech), ERK1/2 (1:10,000, 11,257-1-AP, ProteinTech), Phospho (p)-ERK1/2 (Thr202/Tyr185) (1:1000, ab201015, Abcam), and c-FOS (1:1000, ab222699, Abcam). The membranes were incubated with a specific secondary antibody HRP-labeled goat anti-rabbit IgG (1:1000, A0352, Beyotime) for 1 h at room temperature (RT). Protein blots were developed using an ECL Western blotting substrate (PE0010, Solarbio). For quantitative analysis, the gray values of the target protein blots were measured in ImageJ (v1.54 k) and normalized to the gray value of the internal reference GAPDH or the first group.

Isolation and characterization of EVs

Referring to a previous report [15], BMSC-derived EVs were separated by ultracentrifugation. The complete medium of BMSCs was ultracentrifuged at 100,000 g for 20 h to remove the EVs, and the BMSCs were incubated with this medium for 48 h. The medium was centrifuged at 300 g for 10 min, 2000 g for 10 min, and 10,000 g for 30 min at 4 °C. The supernatant was taken after each centrifugation to remove cells and cellular debris. Finally, the supernatant was filtered using a 0.22 μm pore size vacuum bottle top filter (S2GPT02RE, Millipore Corp, Billerica, MA, USA) and centrifuged at 100,000 g for 70 min to precipitate the EVs. The precipitated EVs were resuspended in PBS, and the EVs were quantified by protein concentration using the BCA method.

The EVs dissolved in PBS were used to measure the particle size distribution by Nanoparticle Tracking Analysis (NTA), and the morphological structure of EVs was observed by transmission electron microscopy (TEM, HT7700, Hitachi, Tokyo, Japan). Additionally, 20 μg of EVs were subjected to Western blot analysis to evaluate the expression of EV marker proteins CD63, TSG101, and CD9, as well as the endoplasmic reticulum marker protein Calnexin.

Labeling of EVs and uptake assay

Isolated EVs were suspended in PBS containing the green fluorescent dye PKH67 (HY-D1421, MedChemExpress) and incubated at 4 °C for 30 min. To remove excess dye, the solution containing EVs was centrifuged three times at 4000 g for 3 min each at RT. The supernatant was diluted with PBS and ultracentrifuged at 100,000 g for 90 min at 4 °C to precipitate the PKH67-labeled EVs. EV precipitates were diluted in PBS and added to chondrocytes or synovial fibroblasts cultured in a 96-well plate. After a 24-h incubation, unbound EVs were rinsed off with PBS. The uptake of EVs by chondrocytes or synovial fibroblasts was assessed by quantifying the fluorescence intensity values of PKH67 in three random fields of view using ImageJ software after labeling the nuclei of the cells with DAPI staining solution (C1006, Beyotime).

PKH67-labeled EVs were injected into the joint cavities of mice (40 μg/mouse). Twenty-four h later, the mice were euthanized by intraperitoneal injection of an overdose of sodium pentobarbital (150 mg/kg), and knee joint and synovial tissues were collected. After labeling cell nuclei with DAPI staining solution (C1006, Beyotime) on tissue sections, the uptake of EVs by different tissues was assessed by quantifying the PKH67 fluorescence intensity in three random fields of view using ImageJ software.

Cell counting assay

Cell counting was carried out using the Cell Counting Kit-8 (CCK-8). Chondrocytes were plated into 96-well plates at 1 × 10⁴ cells/well. After treatment according to the grouping, 10 μL of CCK-8 solution (HY-K0301, MedChemExpress) was added to each well, and the cells continued to be incubated for 2 h. Cell viability was assessed by measuring absorbance at 450 nm with a microplate reader, and data from each independent replicate experiment were normalized according to the control.

Live-dead cell staining

Calcein/PI Cell Viability/Cytotoxicity Assay Kit (C2015, Beyotime) was applied to detect chondrocyte death. After chondrocytes were seeded in 96-well plates, treated as indicated, incubated with 100 μL of Calcein AM/PI assay working solution at 37 °C for 30 min in the dark, and observed under a fluorescence microscope. The green fluorescence of Calcein AM staining labeled the live cells, and the red fluorescence of PI staining labeled the dead cells. Dead cells were counted as a percentage of the total cells (live and dead).

Alcian blue staining

The acidic polysaccharide, glycosaminoglycan (GAG) content in chondrocytes was assessed by the Alcian blue stain kit (DR0005, G-CLONE, Beijing, China). The treated chondrocytes were fixed with 4% paraformaldehyde, incubated with Alcian acidification solution for 3 min, stained with Alcian staining solution for 30 min, washed twice with 3% acetic acid, and imaged by microscopy. To quantify the stained GAG content [16], 6% acetic acid was added to dissolve the dye, and the absorbance at 595 nm was measured. The data from each independent replicate of the experiments were normalized according to the control.

DNA synthesis assay

DNA synthesis in chondrocytes was evaluated using E-Click EdU Cell Proliferation Imaging Assay Kits (E-CK-A377, Elabscience Biotechnology Co., Ltd., Wuhan, Hubei, China). Chondrocytes were treated with 10 μM EdU at 37 °C for 2 h, fixed, permeabilized, incubated with the Click reaction solution for 30 min at RT in the dark, and incubated with DAPI for 10 min at RT in the dark to label the nuclei. Three unduplicated fields of view were randomly selected under a fluorescence microscope, and the percentage of EdU-positive cells to total cells was calculated.

Apoptosis assay

One-step TUNEL In Situ Apoptosis Kit (E-CK-A320, Elabscience) was used to analyze apoptosis in chondrocytes. Briefly, each group of chondrocytes was incubated with TdT Equilibration Buffer at 37 °C for 20 min after fixation and permeabilization, followed by incubation with labeling solution at 37 °C for 60 min and with DAPI for 5 min at RT (both in the dark). The percentage of TUNEL-positive cells to total cells was examined under the fluorescence microscope.

Animal OA model establishment and administration of EVs

The ethics committee of Harbin Medical University (approval no. hrbmuecdc20250304) approved animal experiments. Ten-week-old male C57BL/6 J mice were obtained from Beiyou Biology (Beijing, China). All the mice were housed in a standard SPF laboratory with controlled temperature (20–26 °C), humidity (ranging from 50 to 60%), a 12 h light/dark cycle, and fed chow and water ad libitum. Animals were assigned to groups using a random number table. Only the experimenters knew the group assignments. All outcome measurements were performed by researchers unaware of the group assignments, and data from all mice were included in the analysis.

After being anesthetized by inhalation of 2% isoflurane, the mice were immobilized in the supine position, and the right hind limb was flexed at the knee joint to fully expose the articular surface of the patella. We performed destabilization of the medial meniscus (DMM) surgery to induce the OA model [17]. Briefly, the medial meniscus ligament of the right knee in the hindlimb was surgically excised, and mice in the sham group underwent the same procedure, but did not excise the meniscus ligament after locating it. Mice were euthanized 9 weeks post-surgery via intraperitoneal injection of an overdose of sodium pentobarbital (150 mg/kg). The operated knees were collected from 10 mice per group.

Following the same procedure described above, the mice were anesthetized by inhalation of 2% isoflurane, and an intra-articular injection was administered into the right knee of the hind limb. Joint tissues from five randomly selected mice in each group were fixed with 4% paraformaldehyde for 24 h at 4 °C, followed by decalcification with 20% EDTA (pH = 7.3) solution at RT for 14 d, dehydration, and paraffin embedding. Cartilage tissues from the remaining five mice in each group were weighed, ground into fine fragments using liquid nitrogen, and placed in a grinder (70 HZ, 2 min, 3 times, − 20 °C) to be fully ground into powder for protein extraction. After 1 week of DMM surgery, OA mice were intra-articularly injected with 40 μg of EVs (including Ctr-EVs, CAP-EVs, CAP-NC-EVs, CAP-PBX1-EVs) once a week, and mice injected with an equal amount of PBS served as the control group [18].

AAV-mediated gene intervention

Two weeks before DMM surgery, 10 μL of AAV carrying CACNB4 shRNA (titer 1 × 10¹² vg/mL) was administered into the right knee joints of the hind limbs of mice, and two supplemental AAV injections were performed at 2 and 8 weeks after DMM surgery [19]. The AAVs used were provided by Hanbio (Shanghai, China) and contain the COL2A1 promoter, which can specifically target chondrocytes.

Immunofluorescence staining

Chondrocytes were processed, fixed with 4% paraformaldehyde at RT for 15 min, and permeabilized with PBS solution containing 0.25% Triton X-100 for 10 min. Paraffin-embedded sections (5 μm) of mouse knee cartilage were routinely dewaxed, hydrated, and heated at 95 °C for 15 min by adding EDTA antigen retrieval solution (pH = 9.0, E673003, Sangon Biotech, Shanghai, China). The samples were sealed for 1 h with 5% goat serum and incubated overnight with primary antibodies for MMP13 (1:500, 18,165-1-AP, ProteinTech), ADAMTS5 (1:200, ab246975, Abcam), COL2A1 (1:200, ab34712, Abcam), Aggrecan (ACAN, 1:100, A8536, ABclonal), and PBX1 (1:200, 18204-1-AP, ProteinTech) and with FITC-labeled goat anti-rabbit IgG (1:500, A0562, Beyotime) at 37 °C for 1 h in the dark. Nuclei were counterstained with DAPI. The samples were observed and photographed under a fluorescence microscope. The fluorescence intensity or percentage of positive cells was quantified by ImageJ software.

Safranin-O and fast green (SO/FG) staining

Paraffin-embedded sections (5 μm) of mouse knee joints were stained using the modified SO/FG stain kit (G1371, Solarbio), and all operations were performed. The stained sections were examined with a light microscope and photographed to analyze the thickness of articular cartilage. Stained sections were graded (grades 0–6) by two pathologists who were unaware of the grouping according to the OARSI cartilage OA histopathology grading system [20], and the average of the two assessed grades was taken.

ELISA

Mouse cartilage tissues were weighed, ground, and PBS containing protease inhibitors was added for repeated freezing and thawing. The homogenate was centrifuged at 4 °C, 5000 g for 10 min. Inflammatory cytokine levels in the supernatants were detected by mouse interleukin (IL)-6 (E-EL-M0044, Elabscience) and TNF-α (E-EL-M3063, Elabscience) ELISA Kits, and all operations were performed according to the manufacturer's protocol.

RT-qPCR

Total RNA was extracted from chondrocytes using Beyozol (R0011, Beyotime). PrimeScript FAST RT reagent Kit with gDNA Eraser (RR092A, Takara Biotechnology Ltd., Dalian, Liaoning, China) was used to generate cDNA from total RNA, and CACNB4 expression was detected by TB Green Fast qPCR Mix (RR430A, Takara). The 2^−ΔΔCT method was used to calculate its relative expression, and GAPDH was chosen for normalization. The primers were as follows: CACNB4 forward: 5′-GACGAGGATGTGCCTGTTCCAA-3′, reverse: 5′-CCTCTTTCACCAGCCTTCCTATC-3′; GAPDH forward: 5’-GTCTCCTCTGACTTCAACAGCG-3’, reverse: 5’-ACCACCCTGTTGCTGTAGCCAA-3’.

ChIP-qPCR

The Pierce Magnetic ChIP Kit (26,157, Thermo Fisher) was used to isolate chromatin bound to PBX1. Briefly, chondrocytes (1 × 10⁴ cells) were fixed using paraformaldehyde to cross-link proteins and DNA, and fragmented chromatin was prepared by Halt protease and phosphatase inhibitor mixtures and micrococcal nuclease. After that, 5% of the sample was harvested as Input, and 1 μg of anti-PBX1 (18204-1-AP, ProteinTech) or normal rabbit IgG was added to the remaining sample and incubated overnight at 4℃. Protein-DNA complexes were captured by Protein A/G magnetic beads, and DNA bound to PBX1 or IgG was eluted and purified. qPCR was performed to detect the CACNB4 promoter. The following ChIP-qPCR primers were used: CACNB4-promoter-F, 5’-TTCCTTTCCGCAATCCCTCT-3’; CACNB4-promoter-R, 5’-GTGCCTCCCATTTCTATGCG-3’.

Dual-luciferase reporter assay

A wild-type (WT) CACNB4 promoter fragment (chr2: 152,099,045–152,100,044) was cloned from human genomic DNA, and a mutant (MT) promoter fragment lacking the PBX1 binding site was designed. The WT/MT CACNB4 promoter was inserted upstream of the Firefly Luciferase Genes of pGL4.20 Luciferase Vectors (E6751, Promega Corporation, Madison, WI, USA) to construct the WT/MT CACNB4 promoter reporter. The promoter reporter and pRL Renilla luciferase control reporter vectors (E2231, Promega) were co-transfected into chondrocytes with the PBX1 expression vector or control vector, respectively, by Lipofectamine 3000 transfection reagent. After 48 h, the luciferase activity of the promoter reporter was assessed by a dual-luciferase reporter assay system (E1910, Promega).

Ca²⁺ influx detection

Chondrocytes (2 × 10⁶ cells/well) were seeded in a 6-well plate and treated according to the assigned groups. The cells were incubated with 5 μM of the calcium ion fluorescent probe Fluo-4AM (S1060, Beyotime) at 37 °C for 30 min and incubated for another 20 min to ensure that Fluo-4AM was completely converted to Fluo-4 within the cells. The intensity of green fluorescence was directly proportional to intracellular Ca²⁺ levels. Ca²⁺ fluorescence intensity was quantified using a fluorescence microplate reader at one-minute intervals over 15 min, and a deviation from the initial fluorescence intensity of greater than 30% (1.3-fold change) was defined as a Ca²⁺ peak.

Statistical analysis

Each set of data originated from five independent replicate experiments. Statistical power analysis was performed using the post hoc analysis feature in the GPower software (v3.1.9.7), which required the input of three parameters: α, sample size, and effect size. The α error probability was set to the default value (0.05), and the sample size was calculated as the number of groups multiplied by 5. The effect size was automatically calculated by the GPower software based on the mean and standard deviation of the data for each group. Using these three parameters, it was verified that all experimental data met statistical power requirements (β err prob > 0.8). All graphs were made using GraphPad Prism 10.6.1 software (GraphPad, San Diego, CA, USA). The results were presented as means with SEM. The area under the curve was used to assess changes in Ca²⁺ levels. The Kolmogorov–Smirnov test was used for normal distribution of the data. Parametric test based on the unpaired t-test for comparisons between two groups or one-way/two-way ANOVA followed by Tukey's or Šídák's post hoc test. For the nonparametric test of multigroup comparisons, the Kruskal–Wallis test is followed by Dunn's multiple comparisons test. In all experiments, a p-value < 0.05 was statistically significant.

Results

CAP modification promotes the uptake of BMSCs-derived EVs by chondrocytes

We identified the commercially purchased BMSCs using flow cytometry. CD29, CD44, CD73, and CD105 were positive, while CD34 and CD45 were negative in these source cells (Fig. S1A). Meanwhile, the cells had good differentiation ability and could differentiate into osteoblasts, adipocytes, and chondrocytes (Fig. S1B). We transfected BMSCs with plasmids expressing the LAMP2B protein fused to either a CAP or an SP, and Western blot detected enhanced LAMP2B expression, confirming successful transfection (Fig. 1A). The EVs derived from the above cells were isolated by ultracentrifugation and designated as Ctr-EVs, SP-EVs, and CAP-EVs, respectively. The peaks of the particle sizes of all groups of EVs were around 110 nm (Fig. S1C), and TEM observed that all EVs were typical disc-shaped vesicle structures (Fig. S1D). All EVs expressed the marker proteins CD9, CD81, and TSG101, but not the endoplasmic reticulum marker protein Calnexin. Furthermore, no expression of these proteins was observed in the EVs-depleted medium used for EV isolation (Fig S1E). There were no significant differences in the number of particles per unit protein concentration (/μg) among the various groups of EVs (Fig. 1B). When EVs with the same number of particles (10⁹ particles) were isolated, there were also no significant differences in the quantified protein content (Fig. 1C). This indicated that, under identical extraction conditions, BMSCs can serve as a good source for the standardized production of EVs.

Fig. 1.

CAP modification promotes the uptake of BMSCs-derived EVs by chondrocytes. A Detection of the effect of transfection of CAP-LAMP2B on LAMP2B protein expression in BMSCs by Western blot analysis. B Number of particles per unit protein content (/μg). C Protein content of EVs at a specific particle count (10⁹ particles). D The uptake of EVs (labeled by PKH67) by chondrocytes or synovial fibroblasts was observed. E Uptake of PKH67-labeled EVs by mouse cartilage and synovial tissues. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA (A–D) or two-way ANOVA (E)

When EVs were labeled with PKH67 and co-cultured with chondrocytes or synovial fibroblasts, it was found that the uptake of CAP-EVs by chondrocytes was significantly enhanced, and that CAP modification inhibited the uptake of EVs by synovial fibroblasts (Fig. 1D). Following intra-articular injection of fluorescently labeled EVs, it was observed that CAP-EVs were primarily taken up by cartilage tissues and were present in lower concentrations in synovial tissues, whereas the distribution of the remaining EVs in cartilage or synovial tissues showed no significant differences (Fig. 1E).

CAP modification enhances the therapeutic effect of EVs on chondrocyte injury

IL-1β induced damage to chondrocytes, while further increases in IL-1β concentration beyond 10 ng/mL did not significantly enhance cellular damage. Therefore, a concentration of 10 ng/mL was selected for subsequent experiments (Fig. 2A). The effects of IL-1β induction (10 ng/mL) and combined gradient dose (2.5, 5, and 10 μg/mL) EVs treatment on chondrocyte viability were examined by CCK-8 assay (Fig. 2B). IL-1β induced chondrocyte damage. Even though 10 μg/mL of Ctr-EVs significantly ameliorated the cellular damage induced by IL-1β, 5 μg/mL of CAP-EVs demonstrated significant therapeutic effects. At the same dose (5 or 10 μg/mL), the therapeutic effect of CAP-EVs was superior to that of Ctr-EVs, and the 10 μg/mL concentration for both Ctr-EVs and CAP-EVs was selected for subsequent experiments.

Fig. 2.

CAP modification enhances the therapeutic effect of EVs on chondrocyte injury. A Cell viability in chondrocytes treated with IL-1β at concentrations of 2.5, 5, 10, and 20 ng/mL was assessed using the CCK-8 assay. B Cell viability in chondrocytes treated with IL-1β (10 ng/mL) combined with graded doses of EVs (2.5, 5, and 10 μg/mL) for 24 h was assessed using CCK-8. C The effect of IL-1β treatment (10 ng/mL) and EVs (10 μg/mL) treatment on chondrocyte death was assessed using live-dead cell staining. D Expression of Cleaved-caspase3, Bax, IL-6, and TNF-α in chondrocytes was assessed using Western blot analysis. E MMP13 and ADAMTS5 expression in chondrocytes was assessed using immunofluorescence staining. F COL2A1 and ACAN expression in chondrocytes was assessed using immunofluorescence staining. G GAG content in chondrocytes was assessed using Alcian blue staining. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA

Live-dead cell staining observed that IL-1β treatment resulted in a decrease in the number of live cells and an increase in the number of dead cells; EVs, particularly CAP-EVs, significantly reduced cell death (Fig. 2C). The expression of apoptosis marker proteins Cleaved-caspase3 and Bax, and inflammation marker proteins IL-6 and TNF-α was significantly elevated in IL-1β-treated chondrocytes. By contrast, EV treatment reduced apoptosis and inflammatory damage in chondrocytes, and CAP-EVs were superior to Ctr-EVs (Fig. 2D).

As shown by immunofluorescence staining in Fig. 2E, F, IL-1β enhanced the expression of ECM catabolic markers (MMP13 and ADAMTS5) and diminished the expression of ECM anabolic markers (COL2A1 and ACAN) in chondrocytes, and treatment with EVs inhibited ECM catabolism and promoted anabolism. EVs significantly blocked IL-1β-induced reduction of GAG in chondrocytes and improved chondrolysis (Fig. 2G). The therapeutic effects of CAP-EVs on OA-associated chondrolysis were all significantly better than those of Ctr-EVs.

Screening of endogenous therapeutic molecules in EVs

The GSE179716 dataset (containing chondrocytes isolated from preserved cartilage and damaged cartilage) was used to analyze transcriptomic changes due to OA progression, and the GSE158875 dataset (human primary chondrocytes treated with or without inflammatory cytokines) was used to analyze chondrocyte transcriptomic changes due to pro-inflammatory cytokine treatment. Transcriptome difference analysis was performed in GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/), and differentially expressed genes were screened by Benjamini & Hochberg correction of p-value with adjusted p-value < 0.05 as the threshold (Fig. 3A). We downloaded known transcription factors from AnimalTFDB (https://guolab.wchscu.cn/AnimalTFDB4//?#/). In addition, proteins contained in EVS of BMSCs origin were downloaded from Vesiclepedia (http://microvesicles.org/index.html). The EVs proteins were cross-screened with genes with reduced expression (log₂FoldChange < 0) in OA from the GEO datasets and transcription factors, with two intersections: PBX1 and nuclear factor 1 A-type (NFIA) (Fig. 3B). The expression of PBX1 and NFIA in chondrocytes with different treatments was detected by Western blot analysis. Both PBX1 and NFIA expression were diminished in IL-1β-treated chondrocytes, whereas EVs, especially CAP-EVs, were effective in restoring intracellular PBX1 expression, but had no significant impact on NFIA expression (Fig. 3C). Western blot confirmed that PBX1 protein was carried in BMSCS and their derived EVs (Fig. 3D).

Fig. 3.

Screening of endogenous therapeutic molecules in EVs in OA. A DEGs associated with cartilage damage in OA were screened in the GSE179716 and GSE158875 datasets. B Cross-screening of transcription factors contained in BMSCs-derived EVs that are lowly expressed in OA cartilage damage. C The effect of IL-1β and EVs treatment on PBX1 and NFIA protein expression in chondrocytes by Western blot analysis. D PBX1 protein expression in BMSCs and EVs by Western blot analysis. E Efficiency of overexpression of PBX1 in BMSCs by Western blot analysis. F NTA analysis of particle size distribution of EVs derived from BMSCs transfected with NC or PBX1. G TEM observation of the morphological structure of EVs derived from BMSCs transfected with NC or PBX1. H Characterization of protein expression of CD9, CD81, TSG101, and Calnexin in EVs derived from BMSCs transfected with NC or PBX1 by Western blot analysis. I The uptake of EVs (labeled by PKH67) by chondrocytes was observed. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by unpaired t-tests (E, I) or one-way ANOVA (C)

We co-transfected the CAP-LAMP2B plasmid and the PBX1 or negative control (NC) plasmid in BMSCs, and the successful overexpression of PBX1 in BMSCs was verified (Fig. 3E). These two groups of BMSCs-derived EVs were collected and named CAP-NC-EVs and CAP-PBX1-EVs, respectively. A comparison of the two groups of EVs revealed no statistically significant disparities in terms of particle size or morphology (Fig. 3F, G). Both groups of EVs expressed specific molecular markers (CD9, CD81, and TSG101), but not the negative marker Calnexin, and CAP-PBX1-EVs contained more PBX1 protein (Fig. 3H). There was no significant difference in the uptake capacity of chondrocytes for these EVs (Fig. 3I).

PBX1 protein is a key molecule for the chondroprotective role of CAP-EVs

Treatment with IL-1β resulted in a significant inhibition of the proliferation of chondrocytes, as evidenced by a reduction in cellular DNA synthesis, as evidenced by EdU staining (Fig. 4A). EVs treatment improved the proliferation of chondrocytes; the therapeutic effect of EVs could be further enhanced by CAP modification and loading of PBX1. Consistently, EV treatment significantly reduced IL-1β-induced chondrocyte apoptosis, which was further suppressed by CAP modification of EVs and increased PBX1 content (Fig. 4B). Live/dead cell staining further confirmed that increased PBX1 levels more effectively inhibited cell death and promoted cell survival (Fig. 4C).

Fig. 4.

PBX1 protein is a key molecule for the chondroprotective role of CAP-EVs. A The effect of IL-1β treatment and combined EVs treatment on the DNA synthesis of chondrocytes was assessed using EdU staining. B The effect of IL-1β treatment and combined EVs treatment on chondrocyte apoptosis was assessed using TUNEL. C Chondrocyte death was assessed using live-dead cell staining. D Expression of Cleaved-caspase3, Bax, IL-6, and TNF-α in chondrocytes treated with PBX1-loaded EVs. E MMP13 and ADAMTS5 expression in chondrocytes was assessed using immunofluorescence staining. F COL2A1 and ACAN expression in chondrocytes was assessed using immunofluorescence staining. G GAG content in chondrocytes was assessed using Alcian blue staining. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA

Increased PBX1 content greatly downregulated the expression of apoptosis and inflammation-related proteins in chondrocytes (Fig. 4D), upregulated ECM anabolism in chondrocytes (Fig. 4E), and inhibited catabolism in chondrocytes (Fig. 4F). Meanwhile, the content of GAG in chondrocytes was further restored upon the loading of PBX1 protein (Fig. 4G).

Engineered EVs significantly improve DMM-induced OA cartilage damage

OA mouse model was induced by DMM and treated with EVs (Fig. 5A). Immunofluorescent staining of PBX1 in the knee joint showed (Fig. 5B) that DMM surgery resulted in the loss of PBX1 expression in the articular cartilage region. EVs mitigated the loss of PBX1 expression, but only engineered (CAP-modified and increased PBX1 protein content) EVs significantly increased PBX1 expression in articular cartilage. SO/FG staining was used to assess OA-related cartilage damage (Fig. 5C). It was observed that DMM surgery reduced cartilage thickness and enhanced OARSI grade in mice. Treatment with EVs ameliorated DMM-induced cartilage damage to varying degrees, but only EVs that underwent both CAP modification and increased PBX1 protein content significantly reduced OARSI grade. We hypothesized that this may be related to the multiple groups compared simultaneously.

Fig. 5.

Engineered EVs improve OA-related cartilage damage. A Schematic diagrams illustrating mouse modeling. B The effect of DMM modeling and EVs treatment on PBX1 expression in mouse articular cartilage was assessed using immunofluorescence staining. C Cartilage damage in mice was evaluated using SO/FG staining. D ECM anabolic marker protein COL2A1 expression in mouse articular cartilage was assessed using immunofluorescence staining. E ECM catabolic marker protein MMP13 expression in mouse articular cartilage was assessed using immunofluorescence staining. F Detection of inflammatory cytokines IL-6 and TNF-α in cartilage tissues by ELISA. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA (B, C, D, E, F) or Kruskal–Wallis test (C)

Immunofluorescence analysis of COL2A1 (Fig. 5D) and MMP13 (Fig. 5E) showed that DMM surgery reduced ECM synthesis and promoted catabolism, and EVs, especially engineered EVs, significantly inhibited cartilage ECM catabolism. Inflammatory cytokines IL-6 and TNF-α levels in cartilage tissues were increased after DMM surgery, while EVs treatment ameliorated the inflammatory damage. Intriguingly, CAP modification or enhancement of PBX1 did not further enhance the therapeutic effect of EVs (Fig. 5F). This suggests that OA-associated inflammatory injury is not only caused by chondrocyte damage, but may also involve other cells, such as immune cells.

Downstream signaling analysis of PBX1 in OA

We obtained 2097 target genes of PBX1 from Harmonizome 3.0 (https://maayanlab.cloud/Harmonizome/) from the Dataset: ChEA Transcription Factor Targets. The target genes of the PBX1 were cross-screened with the downregulated genes in the GEO datasets, and there were 21 intersecting genes (Fig. 6A). KEGG pathway enrichment analysis of the 21 intersecting genes revealed hsa04010: MAPK signaling pathway as the most significantly enriched pathway (Fig. 6B). Among the genes enriched to this pathway (CACNB4, IL1R1, and MAP3K12), CACNB4 (CACN) acted as an initiator by activating the classic MAPK pathway, the ERK pathway, through Ca²⁺ signaling, to promote the expression of downstream factors related to cell proliferation and differentiation (Fig. 6C). By Jaspar (https://jaspar.elixir.no/), we predicted the presence of a transcriptional binding site for PBX1 in the CACNB4 promoter (chr2: 152,099,045–152100044) (Fig. 6D). Therefore, we hypothesized that EV-delivered PBX1 mediates ERK pathway activation by promoting the transcriptional expression of CACNB4, thereby alleviating OA cartilage damage and promoting OA cartilage regeneration.

Fig. 6.

Downstream signaling analysis of PBX1. A Screening of potential target genes for the transcription factor PBX1. B KEGG pathway enrichment analysis of potential PBX1 target genes. C CACNB4 is involved in the regulation of the ERK signaling pathway. D Predicted binding sites for PBX1 on the CACNB4 promoter fragment. E Detection of CACNB4 transcription in chondrocytes by RT-qPCR. F CACNB4, p-ERK, and c-FOS protein expression in chondrocytes by Western blot analysis. G Effect of overexpression of PBX1 on CACNB4/ERK/c-FOS signaling in chondrocytes detected by Western blot analysis. H Detection of PBX1 binding ability to the CACNB4 promoter by ChIP-qPCR. I The effect of PBX1 on the transcriptional activity in the WT/MT CACNB4 promoter was analyzed using a dual-luciferase reporter assay. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA (E, F), unpaired t-test (G, H), or two-way ANOVA (I)

IL-1β attenuated CACNB4 expression in chondrocytes as detected by RT-qPCR, and CACNB4 expression was significantly restored in EV-treated chondrocytes. Both CAP modification and increased PBX1 protein loading further augmented the activation of CACNB4 expression by EVs (Fig. 6E). EVs exerted a certain ameliorative effect on the Ca²⁺ oscillation; both CAP modification and increased PBX1 protein loading further enhanced the ameliorative effect of EVs on Ca²⁺ signaling (Figs. S2A, S2B).

PBX1 protein changes in chondrocytes were consistent with its transcription changes (Fig. 6F). The expression of the ERK signaling and downstream c-FOS in chondrocytes was also evaluated (Fig. 6F). Interestingly, phosphorylation of ERK (ERK1/2) as well as c-FOS protein expression were significantly enhanced in IL-1β-treated chondrocytes, which may be related to stress response. Treatment with EVs differentially promoted ERK signaling in chondrocytes and protein expression of c-FOS, which was more pronounced in engineered EVs.

To exclude interference caused by other components of EVs, we transfected chondrocytes with an overexpression vector for PBX1. Forced expression of PBX1 in chondrocytes promoted CACNB4 expression and activated the ERK pathway and c-FOS expression (Fig. 6G). ChIP-qPCR detected that anti-PBX1 significantly enriched the promoter fragment of CACNB4 (Fig. 6H). Furthermore, overexpression of PBX1 significantly enhanced the luciferase activity of the CACNB4 WT-promoter reporter, but failed to enhance the luciferase activity of the MT-promoter reporter, which lacked the PBX1 binding site (Fig. 6I).

The protective effect of engineered EVs on chondrocytes is dependent on ERK signaling activated by CACNB4

Chondrocytes were transfected with plasmids encoding CACNB4 shRNA and MEK1, an upstream activator of ERK, followed by treatment with CAP-PBX1-EVs and IL-1β, and IL-1β-treated cells served as model controls. Knockdown of CACNB4 reversed the promotion of ERK phosphorylation by CAP-PBX1-EVs, leading to attenuated c-FOS expression, which could be rescued by the MEK1 overexpression plasmid (Fig. 7A). Calcium influx assays revealed that knockdown of CACNB4 impaired intracellular Ca²⁺ signaling, and MEK1-mediated ERK activation failed to restore intracellular Ca²⁺ signaling (Fig. S2C, 2D), indicating that ERK is located downstream in the signaling pathway. Knockdown of CACNB4 attenuated the chondroprotective effects of CAP-PBX1-EVs, leading to reduced DNA synthesis (Fig. 7B) and increased apoptosis rates (Fig. 7C), resulting in an overall increase in cell death (Fig. 7D). ECM anabolism was also inhibited after knockdown of CACNB4 (Fig. 7E, F), and GAG content was reduced (Fig. 7G). The cartilage damage caused by CACNB4 knockdown was improved by restoring ERK signaling.

Fig. 7.

The protective effect of engineered EVs on chondrocytes is dependent on ERK signaling activated by CACNB4. A The protein expression of CACNB4/ERK/c-FOS in chondrocytes treated with shRNAs targeting CACNB4 and the plasmid encoding MEK1. B The effect of shRNAs targeting CACNB4 and the plasmid encoding MEK1 on the DNA synthesis of chondrocytes was assessed using EdU staining. C The effect of shRNAs targeting CACNB4 and the plasmid encoding MEK1 on chondrocyte apoptosis was assessed using TUNEL. D Chondrocyte death was assessed using live-dead cell staining. E ECM anabolic marker proteins COL2A1 and ACAN expression in chondrocytes was assessed using immunofluorescence staining. F ECM catabolic marker proteins MMP13 and ADAMTS5 expression in chondrocytes was assessed using immunofluorescence staining. G GAG content in chondrocytes was assessed using Alcian blue staining. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA

The ameliorative effect of engineered EVs on OA-associated cartilage damage is dependent on CACNB4/ERK signaling

AAV containing a chondrocyte-specific promoter (COL2A1) was used to knock down CACNB4 expression in mouse cartilage, followed by DMM surgery and CAP-PBX1-EVs treatment (Fig. 8A). AAV-COL2A1-shCACNB4 significantly reversed CAP-PBX1-EVs-induced CACNB4 expression and ERK signaling as detected by Western blot analysis (Fig. 8B). SO/FG staining observed that knockdown of CACNB4 expression in chondrocytes significantly attenuated the chondroprotective effect of CAP-PBX1-EVs, with reduced cartilage thickness and elevated OARSI grade in OA mice (Fig. 8C). As shown by immunofluorescence staining, COL2A1 expression in cartilage tissues was reduced after knockdown of CACNB4, while MMP13 expression was elevated following administration of AAV-COL2A1-shCACNB4 (Fig. 8D, E). Inflammatory cytokine content in cartilage tissues was also significantly increased after the knockdown of CACNB4 (Fig. 8F).

Fig. 8.

The ameliorative effect of engineered EVs on OA-associated cartilage damage is dependent on CACNB4/ERK signaling. A Schematic diagrams illustrating mouse modeling. B Detection of CACNB4 expression, ERK phosphorylation, and expression of ERK downstream effector molecule c-FOS in cartilage tissues of mice by Western blot analysis. C Cartilage damage in mice was evaluated using SO/FG staining. D ECM anabolic marker protein COL2A1 expression in mouse articular cartilage was assessed using immunofluorescence staining. E ECM catabolic marker protein MMP13 expression in mouse articular cartilage was assessed using immunofluorescence staining. F Detection of inflammatory cytokines IL-6 and TNF-α in cartilage tissues by ELISA. N = 5/group. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA

Discussion

In this investigation, we sought to elucidate the impact of engineered BMSC-EVs on cartilage damage in a murine model of OA. The in vitro study showed that CAP-EVs treatment significantly attenuated IL-1β-induced human primary chondrocyte injury, as evidenced by upregulation of COL2A1 and ACAN and downregulation of MMP13 and ADAMTS5. In the mouse model of OA, loading of PBX1 in BMSC-EVs further alleviated DMM-induced thinning of cartilage thickness and lowered OARSI grade in mice.

The prevailing pharmaceutical interventions are directed towards the mitigation of symptoms, yielding only marginal outcomes [21]. In OA, there is a loss of normal cartilage ECM structure, which includes degradation of the functional matrix (most notably type II collagen and proteoglycans), decreased tissue hydration, and an inaccurate fibrous ECM, which is often accompanied by the overgrowth of chondrocytes and inflammation of the joint [22]. GAG, a critical component of articular cartilage ECM, functions as a vital detector during various physiological and pathological stages of articular cartilage, thereby contributing to a comprehensive understanding of tissue health and dysfunction [23]. Here, we revealed the superior effects of CAP-EVs on inhibiting ECM catabolism and promoting anabolism compared to Ctr-EVs. MSC-EVs have been demonstrated to play a significant role in regenerative medicine approaches for musculoskeletal regeneration [24]. Even though multiple strategies have been used to modify the EVs for the treatment of OA [15, 25, 26], the molecules involved remain to be further elucidated. Recent findings suggest that intra-articular injection of microRNA-7704-overexpressing EVs derived from human umbilical cord MSCs may enhance walking capacity, preserve cartilage morphology, and yield higher histological scores [27]. In the present study, we mainly focused on transcription factors since they are important players in chondrocyte development and maturation [28]. It has long been established that PBX1-deficient mice have significantly reduced chondrocyte proliferation and increased hypertrophic chondrocytes with premature skeletal ossification [29]. The engineered EVs derived from BMSCs transfected with CAP-LAMP2B and PBX1 have a consistent particle size distribution and marker proteins, similar to those from BMSCs transfected with CAP-LAMP2B and NC, but they have great potential in alleviating cartilage damage.

The activation of the ERK pathway has been demonstrated to play a pivotal role in facilitating the process of cartilage differentiation, as well as enhancing the production of ECM [30]. Here, we found that the most downstream targets of PBX1 in OA are enriched in the MAPK signaling pathway, and CACNB4, as the upstream gene in the classic MAPK pathway, i.e., ERK signaling, piqued our attention. Additionally, MSC-derived exosomes enhanced sulfated GAG synthesis and suppressed IL-1β-induced MMP13 production, which were partially abrogated by inhibitors of ERK phosphorylation [31]. As posited by Matsuoka et al., an increase in the presence of FOS-positive cells among human and murine OA cartilage has been observed during the progression of the disease, and mice with cartilage-specific c-Fos knockout exhibited exacerbated cartilage destruction induced by DMM [32]. Calcium channels, the principal mediators of calcium signaling transduction, have the capacity to facilitate the influx of extracellular calcium or the release of calcium from stores and play a pivotal role in the inflammatory response, particularly in rheumatoid arthritis [33]. Consistently, we found that the overexpression of PBX1 in chondrocytes contributed to the enhancement of CACNB4, p-ERK, and c-FOS, indicating that the activation of the ERK signaling in chondrocytes challenged by IL-1β might be a stress response. The ERK pathway has been revealed to play a protective role against chondrocyte apoptosis, and inhibition of the ERK pathway enhanced the apoptosis induced by nitric oxide [34]. The published reports of CACNB4 function have mainly concentrated on the brain [35, 36]. Our findings here that CACNB4-mediated ERK signaling is involved in the chondroprotective role of CAP-EVs might offer fresh insights into its function beyond the CNS.

The limitations of the present study are as follows. Although ERK activation can promote protective responses in chondrocytes, it is important to note that sustained ERK signaling has been implicated in chondrocyte hypertrophy [37]. In this investigation, molecular indicators of hypertrophic differentiation remained undetected. Nevertheless, further research incorporating extended time course analyses will be necessary to ensure that the treatment regimen does not induce excessive ERK overactivation. Furthermore, while CAP decoration has been demonstrated to enhance targeting, it may also introduce risks of immunogenicity, including complement activation or antibody formation upon repeated dosing. Future research will incorporate complement activation assays, anti-CAP antibody profiling, and repeat-dose safety studies. These studies will ensure that CAP-mediated targeting does not elicit unintended immune responses. Finally, while engineered EVs demonstrate potential for biological activity, their translation into a clinical therapy will necessitate extensive further work, including thorough safety assessments, scalable manufacturing strategies, and long-term efficacy studies. They will be crucial for the future development of the project.

Conclusion

In summary, the present study corroborated the notion that the expression of PBX1, a transcription factor, is diminished in IL-1β-induced chondrocytes. Furthermore, PBX1 loading in engineered EVs has the potential to promote proliferation and inhibit apoptosis in these cells. In addition, PBX1 loading in engineered EVs has been shown to reverse the decreased synthesis and increased degradation of ECM and inflammatory reaction in both in vivo and in vitro models. The study further elucidates the role of PBX1 in OA by demonstrating its influence on CACNB4 and the ERK signaling pathways in chondrocytes. In essence, our findings have yielded previously undocumented evidence of the role of CAP-EVs with PBX1 loading in maintaining cartilage homeostasis.

Abbreviations

EVs

Extracellular vesicles

BMSCs

Bone marrow mesenchymal stem cells

OA

Osteoarthritis

CAP

Chondrocyte-affinity peptide

PBX1

Pre-B-cell leukemia transcription factor 1

CACNB4

Calcium channel voltage-dependent subunit beta 4

Author contributions

YQZ designed the study. FSW and HY performed all the experiments and drafted the manuscript. CQJ and ZL participated in the analysis of the data. RQB performed the experiments and acquired the data. JY supervised the study and revised the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Renji Hospital, Shanghai Jiao Tong University School of Medicine, Pinnacle Program Project (No. 20240122).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Ethics approval and consent to participate

This manuscript includes the use of primary human BMSCs and human primary chondrocytes that are commercially available. The initial ethical approval and informed consent were obtained for Cyagen and Wanwu Shengwu. We granted institutional approval on March 24, 2025, by the ethics committee of Harbin Medical University (Title: Extracellular vesicles derived from engineered BMSCs improve damaged cartilage in mice with osteoarthritis by delivering PBX1; Approval Number: hrbmuecdc20250304). All the animal experiments were compliant with the ARRIVE guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Artificial intelligence

The authors declare that they have not used AI-generated work in this manuscript.