Therapeutic efficacy of extracellular vesicles from hiPSC-derived MSCs in serum-containing and xeno-free media for osteoarthritis treatment

Abstract

Background

Extracellular vesicles derived from human induced mesenchymal stromal cells (hiEVs) constitute a promising cell-free therapeutic option for osteoarthritis. To facilitate transition to the clinic we evaluated the therapeutic effects of hiEVs for osteoarthritis treatment. Specifically, we compared the efficacy of hiEVs collected from serum-containing and serum-free, PurStem (PS), media in an osteoarthritis mouse model.

Methods

hiEVs were administered via intra-articular injection in a destabilization of the medial meniscus (DMM) mouse model, with or without hydrogel to determine added value of localized application and controlled hiEV-release. Fluorescence imaging was used to monitor the retention of IR780-labeled hiEVs in the joint cavity. Therapeutic effects were evaluated by scoring of damage as well as expression of Mmp13 and Col2, catabolic and anabolic markers respectively, in joint tissues. Subchondral bone changes were assessed with Micro-CT.

Results

Fluorescence imaging confirmed that hiEVs remained localized at the injection site without systemic migration. HiEVs demonstrated significant protective effects against joint tissue degeneration in the osteoarthritis DMM mouse model as evidenced by reduced damage scores, decreased Mmp13 expression, and increased Col2 expression independent of the medium used for hiEV collection. The hydrogel alone also showed beneficial therapeutic effects, illustrated by reduced damage scores, increased Col2, and reduced Mmp13 expression. These effects, however, were notably smaller than those achieved with hiEV treatment. Micro-CT analysis further showed that hiEV treatment attenuated DMM-induced subchondral bone sclerosis as reflected by normalization of the bone volume fraction and trabecular structure.

Conclusions

Together, our findings demonstrate that hiEVs from xeno-free conditions effectively prevent cartilage degradation and promote its repair. This paves the way for future clinical translation of hiEV-based therapies as a safe, scalable, and effective approach to treat osteoarthritis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13287-025-04890-6.

Background

Osteoarthritis (OA) is a progressive and multifaceted joint disease which affects millions of individuals worldwide, especially in the aging population [1]. OA pathophysiology is marked by degeneration of cartilage, remodelling of subchondral bone, and downstream processes frequently involve inflammatory responses of the synovium and other joint structures [2]. Repairing damaged cartilage is particularly challenging due to its limited capacity for regeneration [3]. Current OA treatments are limited to lifestyle changes, physical therapy, and medications aimed at symptom management whereas surgical options concern total joint arthroplasty (TJA) at end stage OA. Although TJA is currently considered the only disease modifying treatment, approximately 30% of patients remain dissatisfied with the outcomes [4]. Consequently, there is an increasing interest in innovative therapeutic approaches to meet the growing needs of individuals affected by OA [5].

Human induced mesenchymal stromal cells (hiMSCs) derived from induced pluripotent stem cells (hiPSCs) offer a scalable and consistent source of stem cells for regenerative therapies. In addition to their scalability, hiMSCs exhibit greater proliferative potential and are less prone to donor-related variability as compared to isolated mesenchymal stromal cells (MSCs), making them particularly suitable for large-scale therapeutic applications [6]. Even more, MSC-derived extracellular vesicles (EVs) have emerged as a promising cell-free therapy with significant regenerative potential for the treatment of OA [7, 8]. EVs are small membrane-bound particles that are naturally released by cells and contain bioactive molecules such as proteins, lipids, RNA, and/or DNA. A proportion of EVs are involved in mediating intercellular communication, and if they originate from specific cells, such as MSCs, they can modulate pathophysiological processes to induce regeneration and tissue repair [6, 9]. The culture conditions of MSCs, however, are suggested to critically influence the molecular and functional characteristics of EVs. In particular, serum-free and serum-containing media differentially affect EV yield, composition, biological activity, and overall therapeutic potential [10, 11]. Nonetheless, EVs derived from MSCs have been shown to support cartilage regeneration by promoting chondrocyte proliferation, reducing inflammation, and enhancing production of extracellular matrix (ECM) [12, 13]. In a rat OA model, MSC-EVs improved cartilage repair in osteochondral defects. However, producing large quantities of EVs for clinical applications is challenging because of the aging processes of primary MSCs and the intradonor heterogeneity of EV releasing MSCs, resulting in batch-to-batch variations in MSC-EV products [14]. In this respect, hiMSC-derived EVs (hiEVs) can address these challenges and offer a more consistent, scalable source for cell-free OA therapy [15].

To explore and compare the therapeutic potential of hiEVs for OA treatment, either in serum-containing or serum-free (PurStem) media (hiEV_serum and hiEV_PS, respectively), we used a destabilizing medial meniscus (DMM) mouse model. Moreover, given that intra-articular (i.a.) therapies in OA have been suggested to be limited due to clearance of injected agents from the synovial cavity and often necessitating repeated injections [16], we employed a recently developed thermosensitive hydrogel based on Poloxamer (P407) with a self-assembling peptide as an effective delivery system of hiEVs for i.a. injection while offering potential advantages of extended retention at the target site [17].

Materials and methods

Extracellular vesicle isolation and characterization

Generation and characterization of the hiPSCs and there of derived hiMSCs that were used in this study have been described previously [18]. HiMSCs were cultured under two different conditions. For serum-containing medium, hiMSCs were cultured in DMEM-GlutaMAX medium (GiBCo, Waltham, MA, United States) supplemented with 10% FBS (Sigma‒Aldrich, St. Louis, MO, United States), 1 ng/ml fibroblast growth factor-2 (FGF-2, Peprotech, London, United Kingdom) and 100 U/ml penicillin-100 mg/ml streptomycin mixture (GiBCo). For the serum-free condition, the cells were continuously cultured in PurStem xeno-free medium [19] devoid of any serum contamination. The cells were maintained at 37 °C with 20% O₂ and 5% CO₂, and passaged upon reaching ~ 80% confluence.

To isolate hiEVs from serum-containing and serum-free media, conditioned medium (CM) was collected from cell cultures after 72 h and immediately centrifuged at 300 × g for 10 min at 4 °C to remove dead cells and debris. The supernatant was then centrifuged at 2000 × g for 20 min at 4 °C to eliminate apoptotic bodies and stored at -80 °C until hiEV-collection by ultracentrifugation. For hiEV isolation by ultracentrifugation, CM from passages 7 to 11 were pooled and subjected to ultracentrifugation as previously described [18, 19].

Characterization of hiEVs was performed according to the MISEV criteria [20] including standards for surface markers CD9, CD63, and CD81 (Supplementary Figure S1). Data acquisition was performed by ImageStreamX Flow Cytometry (IFCM) for hiEV_serum, or by FACS on an Amnis ImageStreamX MkII instrument for hiEV_PS. Analysis and gating were carried out via IDEAS Software 6.2 according to the established workflow, as previously described, including the quantification of single EV events via custom fluorescence and spot count masks [21].

To assess the site of injection and check retention of the hiEVs at the site of injection in the mice, they were labeled with IR780 as previously described [22]. Briefly, EVs were incubated with 100 µM IR780 at 4 °C for 30 min and subsequently passed through a size exclusion purification column (Exo-spin^®) to remove any unbound dye according to the manufacturer’s protocol. This step ensured that only labeled EVs were collected, increasing the specificity of the labeling.

Measurement of HiEV release with degradation of the thermosensitive hydrogel

The release kinetics of hiEVs from a recently developed thermosensitive hydrogel, based on Poloxamer (P407) with a self-assembling peptide [17], were determined via the direct release method [23, 24]. Briefly, 1 × 10⁹ hiEVs were mixed with 1 mL of 25% (w/w) hydrogels in 15 mL tubes at 4 °C. Afterwards, the tubes were incubated at 37 °C for 5 min to allow the formation of a stable gel. Then, 4 mL of pre-warmed PBS was gently added on top of the gel layer in the tubes, and the tubes were placed in a shaking water bath at 37 °C and 30 rpm. At specified time points (1, 2, 4, 8, 12, 24, 48, 72, 96, 120, 144, 168, 192, and 216 h), 1 mL of the supernatant was collected from the tube, and the volume was replaced with fresh 37 °C PBS. To determine the concentration of EVs in the supernatant, Nanoparticle tracking analysis (NTA) was performed in triplicate via a Nanosight^® NS300 (Malvern).

In addition, degradation of the hydrogels was evaluated. To this end, 1 mL of hydrogel was loaded in 15 mL Falcon tubes and placed at 37 °C for gel formation. The initial weight of the tube with the hydrogel was taken as 100% gel weight. Afterwards, 4 mL of prewarmed 37 °C PBS was added on top of the hydrogel. At predetermined time points (corresponding to those in the release study), the entire volume of PBS was removed, and the change in weight of the conical tube with respect to the remaining hydrogel was measured to calculate the amount of hydrogel degradation. Measurements were performend in triplicate.

In vivo experiments

Mouse osteoarthritis model and experimental design

For the in vivo experiment, 40 male 12 week-old C57BL/6J mice were purchased from Charles River Laboratories (Charles River, Chatillon-sur-Chalaronne, France). The animal procedures were all conducted at the Leiden University Medical Center and were approved by the Animal Welfare Committee (IvD) under number AVD1160020171405- PE.18.101.006 and in line with ARRIVE guidelines 2.0. All mice were housed in groups in polypropylene cages (4 animals/ cage) on a 12-hour light/dark cycle with unrestricted access to standard mouse food and water. The first group (N = 4 mice) served as a sham (positive) control. The remaining 36 mice underwent surgery for destabilization of the medial meniscus (DMM) to establish a knee OA model as described elsewhere [25, 26]. Thirty min before surgery, pre-operative analgesia was administered by sub-cutaneous injection of Buprenorphine HCl (Vetergesic; Alstoe Animal Health, York, UK). The animal was then placed under isoflurane anaesthesia (4–5% upon induction, 1–2% for maintenance) and the incision for suregery was started on the right knee when the animal no longer displayed reflexes while the breathing was constant. A 1-cm longitudinal medial para-patellar incision was made to expose the knee joint. Subsequently, the knee joint was opened gently through lateral dislocation of the patella and patellar ligament and the medial meniscotibial ligament which anchors medial meniscus to the tibial plateau was cut. Successful destabilization of the medial meniscus was confirmed during surgery. After transection, the knee joint capsule was closed with a 6 − 0 absorbable suture and the skin-incision was closed with biological glue. Mice were immediately transferred to a warm post-operative recovery room. Within 48 h post-surgery all animals received buprenorphine HCl sub-cutaneously every 8 h. The mice were monitored daily to confirm their general health indicators according to their body weight and knee diameter Twenty-one days after surgery, the DMM mice were treated with a one-time intra-articular injection of 5 µL PBS (30 G needle), either or not with hydrogel and 10⁷ labeled hiEVs from serum-containing media or 10⁷ hiEVs from PurStem (PS) media as illustrated in Fig. 1. Animals were randomly allocated to the experimental groups (n = 6) using a dice. Moreover, to ensure blinded treatments, intra-articular (i.a.) injections were performed by a colleague not involved in randomization while the injection samples were prepared and coded by another colleague not involved in the animal experiment. At 35 days post-treatment, mice were euthanized by carbon dioxide (CO₂) inhalation in accordance with the EU Directive 2010/63/EU, Annex IV, on the protection of animals used for scientific purposes and as approved by the IvD (number AVD1160020171405- PE.18.101.006).

Fig. 1.

Schematic overview of the experimental timeline for in vivo treatment. 21 days prior to the treatment, OA was induced by performing DMM surgery followed by a single i.a. injection of hiEV_serum or hiEV_PS (± hydrogel), hydrogel alone, or PBS. Mice were monitored along the trajectory and sacrificed on day 35. DMM: Destabilization of the Medial Meniscus; i.a.: intra articular; OA: Osteoarthritis

A Pearl Impulse Imaging System (LI-COR, Lincoln, NE, USA) was used to assess hiEV retention in the knee joint based on fluorescence. The mice were imaged at days 1, 7, 14, 21, 28, and 35 after i.a. injection under isoflurane anesthesia (4–5% induction, 1–2% for maintenance). Imaging was performed in the 800 nm NIR fluorescence channel (IR780 emission). The fluorescence intensity was quantified via Image Studio Software (LI-COR Biosciences), with a region of interest (ROI) around the knee joint. Background fluorescence was corrected by using a nonfluorescent ROI.

Micro computed tomography measurements

Thirty-five days after i.a. administration of the treatment groups, mice were euthanized by continuous CO₂ inhalation, the right knee joints were harvested, and fixed with 4% paraformaldehyde. After fixation for 48 h, the specimens were transferred to 70% ethanol for high-resolution micro computed tomography (Micro-CT) (Skyscan 1072, Bruker, Kontich, Belgium). The scanner was set at a gamma-ray voltage of 50 kV and a current of 200 µA, 0.5 mm Al filter, and a resolution of 9 μm per pixel. Subchondral bone morphology were analyzed using 3D data analysis software (CTAnalyzer, Bruker). The region of interest (ROI) was defined within the subchondral bone of the medial tibial plateau and comprised 50 consecutive cross-sections, corresponding to a total thickness of 450 μm. Quantitative parameters included bone volume fraction (BV/TV%), trabecular thickness (Tb.Th), trabecular number (Tb.N), and trabecular spacing (Tb.Sp).

Histological evaluation

Knee joints were fixed for two days in 4% buffered paraformaldehyde, followed by 5 days of decalcification using a commercial mol-decalcifier (Milestone; pH 7.4) at 37 °C. The joints were then embedded in paraffin and sectioned at 5 μm thickness. Sections were stained with Safranin O/Fast Green and hematoxylin–eosin (H&E) and examined by light microscopy to evaluate cartilage damage in the medial compartment of the knee.

The commercial decalcification method was found to markedly reduce glycosaminoglycan (GAG) staining intensity compared with the standard 10% EDTA (pH 7.3) protocol, as observed with Safranin O, Toluidine Blue (TB), and Alcian Blue (AB) staining (Supplementary Figure S2A, compare upper and lower panel). To enable damage scoring despite this limitation, we developed a modified OARSI scoring system (damage OARSI score) based on direct visualization of structural cartilage damage in the cartilage zones using H&E and Safranin O staining which is described here below in ‘Grading of cartilage‘. This approach emphasizes parameters such as surface integrity, cartilage thickness, and chondrocyte organization [27] and was validated by scoring sections decalcified with 10% EDTA both with the modified damage OARSI scoring system and with conventional OARSI scoring, and comparing the results with Spearmann correlation analysis (Supplementary Figure S2B).

Grading of cartilage

In DMM mouse models of OA, lesions are most severe in the medial compartment, which includes the medial femoral condyle (MFC) and medial tibial plateau (MTP). Joint degeneration was assessed using a cartilage damage score based on histological evaluation, following principles similar to the OARSI scoring system [25].

Representative grading criteria for the damage OARSI scoring system are shown in Supplementary Figure S3. Grading of the MFC and MTP in each section was performed by three independent researchers, blinded to experimental conditions and to each other’s scores. The results were averaged, and the mean damage OA score for each section was taken as the representative measure of knee joint damage, as described elsewhere [26].

Immunohistochemical staining

Immunohistochemical (ICH) staining was performed on knee joint sections for collagen type II (Col2) and for matrix metalloproteinase (Mmp13). In brief, paraffin Sect. (5 μm) were blocked for endogenous peroxidase activity via incubation with 0.3% H₂O₂ for 10 min at room temperature. Antigen retrieval was performed with 25 µg/ml proteinase K (Prot K) prepared in 0.1 M Tris/HCL, pH 5.0, for 10 min at 37 °C, followed by 30 min of treatment with hyaluronidase (5 mg/ml in Tris/HCL, pH 5.0). All the sections were blocked in 5% PBS-BSA for 30 min at room temperature. The following primary antibodies were used: anti-type II collagen (Col2) (Abcam, ab34712, Abcam, Cambridge, MA, dilution 1:200) and anti-matrix metalloproteinase (Mmp)-13 (sc-515284, Santa Cruz Biotechnology, Santa Cruz, Dallas, TX, USA, dilution 1:200). Primary antibody incubation was performed overnight at 4 °C. Normal mouse IgG1 (2 µg/mL) served as an isotype control (sc-3877; Santa Cruz, Dallas, TX, USA). All the antibodies were diluted in 5% PBS-BSA. The next day, the slides were incubated with anti-mouse HRP (Envision, Dako, CA, USA) for 30 min at room temperature and subsequently incubated with a liquid DAB + 2-component system (Agilent, Santa Clara, CA, USA) for 5 min. The sections were counterstained with hematoxylin, dehydrated, cleared in xylene and covered with glass using Eukitt mounting medium (Sigma‒Aldrich, Saint Louis, USA). Quantification of the IHC intensity was performed via Fiji/ImageJ (version 1.54p) software as described elsewhere [28]. In brief, negative high-resolution images of stained sections were converted to 8-bit grayscale, and identical thresholds were applied across all images to ensure consistency. Regions of interest (ROIs) encompassing the entire tissue area were manually outlined to exclude background. Mean gray values, representing staining intensity, were measured for each ROI.

Statistical analysis

Optimal sample size was determined based on statistical analysis of the results of similar DMM studies. For statistical analysis, GraphPad Prism 8.1.1 software (GraphPad Software, San Diego, CA, USA) was utilized. All data are expressed as the mean ± standard deviation (SD) of 4–6 independent experiments as indicated. Statistical significance was determined using two-way analysis of variance (ANOVA) or mixed model for the in vitro experiments as indicated per analysis. P-values and estimation of the independent effects of covariables of the histology and IHC were determined by applying a linear generalized estimating equation (GEE) analysis via IBM SPSS statistics 25. Statistical significance is expressed as: * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001, **** P ≤ 0.0001.

Results

Hydrogel enhances sustained EV release

hiEVs were collected from serum-containing and from serum-free conditioned medium to compare their therapeutic efficacy in a mouse DMM model as indicated in Fig. 1. Upon concentration via ultracentrifugation, hiEVs were quantified by nanoparticle tracking analysis (NTA) and characterized (Supplementary Figure S1) before storage until use. First, the release kinetics of hiEVs from a recently established thermosensitive hydrogel [17] were assessed in vitro. The results showed that, within the first 12 h, there was no considerable hiEV-release (Fig. 2A). From 12 h onwards, hiEV-counts in the supernatant increased progressively closely paralleling hydrogel degradation over time, with 50% release of the hiEVs around day 5 (120 h). These findings indicate that the hydrogel could serve as a useful carrier for intra-articular delivery and sustained release of hiEVs.

Fig. 2.

Sustained release of hiEVs in vitro and in vivo within thermosensitive hydrogel (A) In vitro release profile of IR780-labeled hiEVs determined with NTA and concurrent hydrogel degradation as determined by the change in weight over 9 days at 37 °C (n = 3 per time point; mean ± SD). (B) Schematic overview for the application of in vivo imaging. Concentrated hiEVs were labeled with IR780, mixed with or without hydrogel (50% labeled, 50% unlabeled), and 10⁷ hiEVs were injected intra-articularly into DMM-operated mouse knees. Near-infrared (NIR) fluorescence imaging was performed weekly to monitor retention of IR780-labeled hiEVs. (C) Representative longitudinal fluorescence images of mouse knees at days 1, 7, 14, 21, 28, and 35 post i.a. injection, across treatment groups. (D) Quantification of fluorescence intensity over time (n = 4–6 mice per group; mean ± SD analysed using a two-way ANOVA mixed model). * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001, **** P ≤ 0.0001; NTA: nanoparticle tracking analysis; i.a.: intra articular

Next, we assessed the in vivo behaviour of IR780-labeled hiEVs following intra-articular (i.a.) injection into the knee joints of OA mice (Fig. 2B). Fluorescent signals were monitored weekly over a 5-week-period (Fig. 2C) and signal intensities were compared accross the different qroups by two-way ANOVA with multiple comparisons (Fig. 2D). One day post-injection, the signal intensity was significantly decreased in the groups without hydrogel, indicating rapid clearance of free hiEVs from the joint cavity. In contrast, mice injected with hydrogel-embedded hiEVs retained strong fluorescence on day 1, suggesting improved short-term retention. In line with the findings in vitro, by day 7 the signal was reduced to less than half, and by day 14 fluorescence intensities had decreased to comparably low levels across all groups without significant differences. This indicates that the hydrogel prolongs hiEV-release with 5 to 7 days after injection as compared to the free hiEVs (Fig. 2D).

Administration of HiEVs slows down cartilage degeneration in a DMM mouse model

Histological analysis was performed to evaluate joint damage across the experimental groups. To account for reduced proteoglycan staining observed with the commercial decalcification method, a modified damage OARSI score was applied, focusing on structural cartilage changes. Validation against standard OARSI scores from previous EDTA-based datasets demonstrated a strong positive correlation (r = 0.763, P ≤ 0.01; Supplementary Figure S2) substantiating the alternative approach.

Histological evaluation revealed that sham-operated mice exhibited minimal joint damage, whereas PBS-treated DMM mice presented pronounced cartilage degradation and damaged structure (Fig. 3A-B). In contrast, mice treated with hydrogel or with hiEVs (+/- hydrogel) regardless of the culture medium, exhibited significantly reduced damage (Fig. 3A-B).

Fig. 3.

hiEVs reduce OA-associated joint damage in vivo. (A) Representative histological images of knee joints stained with H&E and Safranin-O/Fast Green from sham, PBS-injected DMM, and treated groups. (B) Quantification of cartilage damage scores in the meniscus area across treatment groups. Data are shown as the mean ± SD (n = 6). (C) Multivariate analyses of damage scoring using GEE to understand the independent therapeutic effect of hiEVs derived from both serum and PS and hydrogel in the meniscus lesions. All data are shown as means ± standard deviations (n = 4–6). DMM: Destabilization of the Medial Meniscus, H&E: Hematoxylin and Eosin; i.a.: Intra-articular; GEE: Generalized Estimating Equation

To independently evaluate the effects of, hiEV_serum, hiEV_PS, and hydrogel, a multivariate regression analysis using Generalized Estimating Equations (GEE) was performed. As shown in Fig. 3C, all treatment groups exhibited significantly reduced damage scores relative to PBS-treated DMM mice. Specifically, treatment with hiEV_serum and hiEV_PS both resulted in significant reductions in cartilage damage (Beta= − 1.34, P = 6.0 × 10⁻³ and Beta= − 1.51, P = 2.2 × 10⁻³ respectively). Hydrogel alone was also found to exert a significant chondroprotective effect, though this effect was smaller than that observed for hiEV-based treatments (Beta= -0.89, P = 7.7 × 10⁻³). Together, our results demonstrate that hiEVs exhibit robust protective effects against joint tissue degeneration, with hydrogel delivery offering an additional, albeit smaller, therapeutic benefit.

HiEV treatment reduces catabolic activity and enhances anabolic responses in a DMM mouse model

Following our observation that hiEVs with or without hydrogel exert therapeutic effects in the DMM model, we next evaluated their impact on catabolic and anabolic processes. We focused on matrix metalloproteinase 13 (Mmp13), a catabolic enzyme associated with cartilage degradation, and type II collagen (Col2), a key anabolic ECM component (Fig. 4A).

Fig. 4.

Immunohistochemistry for Mmp13 and Col2 expression in the knee joint. (A) Representative images of mouse knees in the different treatment groups as indicated (Sham, DMM-PBS, Hydrogel, hiEV_serum, hiEV_serum + hydrogel, hiEV_PS, and hiEV_PS + hydrogel) stained for Mmp13 and Col2 as indicated. (B) Quantification of IHC staining intensity for Mmp13 in the meniscus across the different treatment groups. (C) Quantification of IHC staining intensity for Col2 in the meniscus across the different treatment groups as indicated. (D) GEE analysis for Mmp13 and Col2 to understand the independent effect of hiEVs_serum and hiEV_PS and hydrogel alone in the meniscus lesions. All data are shown as means ± standard deviations (n = 6 or 12). **P ≤ 0.01 (0: no hydrogel; 1: with hydrogel). GEE: Generalized Estimating Equation

Quantification of Mmp13 expression (Fig. 4B) revealed the highest intensity values in PBS-treated DMM mice. Both hiEV treatment groups, with or without hydrogel, showed visibly lower Mmp13 expression levels, while hydrogel alone resulted in intermediate values between PBS and hiEV-treated joints. GEE analysis confirmed the independent effect of each treatment (Fig. 4D). Mmp13 expression was significantly reduced in the hiEV_serum and hiEV_PS groups compared with PBS-treated DMM mice (Beta= − 13.5, P = 3.4 × 10⁻⁵ and Beta= − 9.64, P = 8.5 × 10⁻³ respectively). Hydrogel alone also reduced Mmp13 expression, but the effect was not statistically significant (Beta= − 4.77, P = 7.1 × 10⁻²).

Figure 4C shows quantification of Col2 expression across the different groups. PBS-treated DMM mice had the lowest values, whereas both hiEV treatment groups showed visibly higher Col2 levels. Hydrogel alone appeared to modestly increase Col2 expression relative to PBS. GEE analysis revealed significant increase in Col2 expression for hiEV_serum and hiEV_PS compared with PBS (Beta = 1.50, P = 4.0 × 10⁻⁷ and Beta = 1.06, P = 7.0 × 10⁻⁴ respectively). Again, although hydrogel alone showed a slight trend towards improved Col2 expression this was not significant (Beta = 0.53, P = 7.0 × 10⁻²).

In vivo Micro-CT analysis shows improved subchondral bone characteristics upon HiEV treatment of DMM-operated mice

Micro-CT analysis was conducted to quantitatively evaluate subchondral bone remodeling in the tibial plateau of mice in response to DMM and upon hiEV treatment. Figure 5A shows representative micro-CT images in epiphysis coronal and transaxial views. Images were analyzed using standard trabecular parameters (Fig. 5B). In line with previous observations [29] this showed increased BV/TV% and Tb.Th in DMM + PBS mice as compared to sham controls, with reduced trabecular spacing and increased numbers reflecting trabecular compaction.To determine the independent effects of hiEV_serum, hiEV_PS, and hydrogel treatment on DMM-induced subchondral bone remodeling, GEE analysis was performed across all trabecular parameters (Fig. 5C). This showed that treatment with both hiEV preparations exerted strong and highly significant reductions in trabecular bone volume fraction compared with untreated DMM joints (BV/TV% hiEV_serum: Beta = − 12.84, P = 1 × 10⁻⁴; hiEV_PS: Beta = − 9.94, P = 3.93 × 10⁻⁴). Likely, this effect is mostly driven by the increased trabecular spacing (Tb.Sp hiEV_serum: Beta = 21.70, P = 3.15 × 10⁻³; hiEV_PS: Beta = 23.65, P = 5.63 × 10⁻⁴), but also the number of trabeculae was reduced (Tb.N hiEV_serum: Beta = − 0.64, P = 9.07 × 10⁻³; hiEV_PS: Beta = − 0.93, P = 2.17 × 10⁻³). Notably, hydrogel alone did not affect the bone volume fraction (BV/TV% Beta = 1.54, P = 0.51) although the trabecular spacing was significantly reduced (Tb.Sp Beta = − 19.43, P = 2.17 × 10⁻³) and the number increased (Tb.N Beta = 0.45, P = 3.95 × 10⁻²).

Fig. 5.

Micro-computed tomography imaging of the knee in a DMM mouse model following hiEV and/or hydrogel administration. (A) Representative micro-CT images in epiphysis coronal and transaxial views. (B) Quantitative assessment of trabecular bone parameters, including bone volume fraction (BV/TV%), trabecular thickness (Tb.Th), trabecular spacing (Tb.Sp), and trabecular number (Tb.N) across treatment groups. (C) Summary of generalized estimating equation (GEE) analyses evaluating the independent effects of hiEV_serum, hiEV_PS, and hydrogel on each trabecular parameter (n = 6 or 12)

Discussion

This study aimed to evaluate and compare the therapeutic potential of extracellular vesicles isolated from hiPSC-derived MSCs cultured in either serum-containing or serum-free medium in an in vivo DMM-induced OA mouse model. Our findings confirmed that hiEV treatment had protective effects as shown by the reduced cartilage degeneration and Mmp13 expression, and increased Col2 expression within the joint. Of note, hiEV treatment also improved subchondral bone characteristics. While the PBS-treated DMM group showed the previously reported increase in BV/TV% associated with OA-related subchondral sclerosis, the groups treated with hiEVs reversed this phenotype mostly via normalization of the trabecular spacing. Importantly, the similar beneficial effects of both hiEVs derived from serum-containing and from serum-free medium. This highlights the robustness and translational potential of hiMSC-derived EVs as cell-free products and provides proof-of-principle that hiEVs produced in a clinically compliant, xeno-free, system can be developed for translational applications as the therapeutic activity is retained even when hiMSCs are cultured under serum-free conditions. In addition, the hydrogel delivery system that we applied proved to sustain local hiEV availability within the joint cavity up to a week, thereby potentially enhancing their therapeutic effect although we did not observe significant differences between mice treated with or without hydrogel embedded hiEVs.

Our results demonstrating beneficial effects of hiMSC-derived EVs in OA models are consistent with previous reports. For example, Warmink and colleagues used bone marrow derived MSC-EVs to treat rats with a groove-surgery [30]. Although the therapeutic effects were not significant the treatment did result in decreased cartilage degeneration and less pain behaviour. Effects of EVs secreted by hiPSCs or their derivatives have not yet extensively been studied for application to treat OA [31]. Nonetheless, it was shown before that EVs derived directly from iPSC reduced cartilage degeneration through preservation of matrix integrity, lower MMP13 activity, and attenuation of inflammation [32] which is in line with our findings for the EVs derived from hiMSCs cultured in serum-free medium. The role of the culture environment in shaping EV properties is an ongoing subject of investigation. It has been suggested that serum components may adsorb onto EVs or become incorporated into their structure, potentially altering their composition and downstream bioactivity [13]. In this context, the use of xeno-free media becomes particularly important when transitioning toward clinical applications, as it eliminates the risk of xenogeneic contamination and reduces the potential for immunogenic reactions [18–20]. Moreover, xeno-free production clearly offers advantages for regulatory approval and patient safety, reinforcing its relevance for future clinical application [33].

A second key aspect of our study was the incorporation of hiEVs into a thermosensitive hydrogel, which served as a carrier to provide sustained release following intra-articular injection and to prolong their retention within the joint cavity. This hydrogel, previously developed and characterized by our group [17], demonstrated suitable thermosensitive properties, remaining injectable at room temperature while undergoing rapid gelation at physiological temperature. Our in vitro release experiments demonstrated that the hydrogel provided a sustained release of hiEVs over nine days, closely paralleling the degradation kinetics of the carrier. In vivo imaging confirmed prolonged intra-articular retention of IR780-labeled hiEVs when delivered in hydrogel compared with free injection, supporting its utility as a delivery vehicle. This is consistent with previous studies reporting that hydrogels can enhance the local bioavailability of EVs and improve therapeutic outcomes in OA and other joint diseases [34–36].

Compared to their parental hiMSCs, hiEVs offer notable advantages including improved storage stability, lower immunogenicity, and reduced tumorigenic risk. The hiEVs can be manufactured and standardized more readily, supporting their use as a scalable and off-the-shelf therapy [37, 38]. Although several studies have demonstrated that EVs have many of the same therapeutic effects as MSCs, some findings suggest that MSCs may exert broader or more sustained effects in certain contexts due to their ability to continuously secrete a wide range of bioactive factors, including cytokines, chemokines, and growth factors, not fully represented in EVs alone [39]. To enhance the efficacy of hiEVs, future strategies should focus on optimizing EV loading and cargo composition, improving targeting and uptake efficiency, and potentially combining EVs with complementary biologics or scaffolds to recapitulate the multifaceted benefits of live cell therapies.

Some limitations of our study should be acknowledged. First, although we demonstrated improved cartilage protection, we did not assess long-term functional outcomes such as pain behaviour or joint biomechanics, which are critical for clinical relevance. Second, while we confirmed EV retention and histological effects, insights into the therapeutic mode of action by addressing transcriptomic or proteomic profiling of target tissues are important to further explore. For that matter, the microRNA cargo of bone marrow MSC-derived EVs in xeno-free medium was determined previously and showed, for example, upregulation of miR-145 and miR-214 that protected chondrocytes from IL-1α-induced inflammation while reducing upregulation of pro-inflammatory cytokines [19] Thirdly, the therapeutic effect of hiEVs observed in this study was modest. In this respect, we only applied a single dose of hiEVs while repeated injections over the course of the 35 days could have further improved the therapeutic effects. Finally, we used a modified OARSI scoring system because the commercial decalcification method that was applied markedly reduced proteoglycan staining. The modified scoring focused on structural cartilage parameters that are visible in H&E and Safranin O staining. However, it may underestimate early degenerative changes where proteoglycan loss precedes structural breakdown. Therefore, the damage scoring primarily reflects structural preservation, while subtle biochemical changes might be under-represented. Further evidence that hiEVs support cartilage ECM preservation was provided by the IHC analysis showing upregulation of the anabolic marker Col2 and downregulation of the catabolic marker Mmp13 as well as with the micro-Ct scans. This effect was comparable for hiEVs produced in serum-containing and in xeno-free media.

In conclusion, we provide evidence that hiEVs produced under xeno-free conditions retain full therapeutic efficacy and can modulate both catabolic and anabolic processes in osteoarthritic cartilage as well as the characteristics of osteoarthritic subchondral bone. Combined with the practical benefits of hydrogel-based delivery, these findings pave the way for future clinical translation of hiEV-based therapies as a safe, scalable, and effective approach to treat osteoarthritis. Looking ahead, an important next step is to determine whether specific EV-subpopulations contribute differently to the therapeutic effects observed. hiEVs are heterogeneous in size and cargo, and biological activity often correlates with vesicle diameter, with small EVs typically enriched in regulatory RNAs and signalling molecules, whereas larger vesicles may contain DNA or cell-derived debris and show distinct functional behaviour [40]. Therefore, in addition to application of multiple injections, future studies isolating specific size-based EVs [41] and their application may potentially identify fractions with enhanced therapeutic potency.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

OA

Osteoarthritis

MSCs

Mesenchymal stromal cells

TJA

Total joint arthroplasty

EVs

Extracellular vesicles

hiMSCs

Human-induced pluripotent stem cell derived mesenchymal stromal cells

hiPSCs

human induced pluripotent stem cells

hiEVs

Extracellular Vesicles derived from human-induced Mesenchymal stromal cells

ECM

Extracellular Matrix

PS

PurStem

DMM

Destabilization of the Medial Meniscus

BV/TV

bone volume fraction (Bone Volume / Tissue Volume)

Tb.Th

Trabecular thickness

Tb.N

Trabecular number

Tb.Sp

Trabecular spacing

CO₂

Carbon dioxide

IHC

Immunohistochemistry

Col2

Collagen type II

Mmp13

Matrix metalloproteinase 13

P407

Poloxamer 407

H&E

Hematoxylin and Eosin

NTA

Nanoparticle tracking analysis

I.A.

Intra-articular

GEE

Generalized estimating equations

Author contributions

Sana S. Sayedipour: Collection and/or assembly of data (in vivo experiments), data analysis and interpretation, manuscript writing, final approval of manuscript. Jelle Nikkels: Collection and/or assembly of data (in vivo experiments), final approval of manuscript. Tobias Tertel: Collection and/or assembly of data, final approval of manuscript. Eka Suchiman: Collection and/or assembly of data, final approval of manuscript. Marijke Koedam: Collection and/or assembly of data, final approval of manuscript. Matilda Balbi: Collection and/or assembly of data, final approval of manuscript. Georgina Shaw: Collection and/or assembly of data, final approval of manuscript. Luis Cruz: Collection and/or assembly of data, final approval of manuscript. Bram van der Eerden: Collection and/or assembly of data, final approval of manuscript. Louise van der Weerd: Collection and/or assembly of data, final approval of manuscript. Chiara Gentili: Collection and/or assembly of data, final approval of manuscript. Mary Murphy: Conception and design, final approval of manuscript. Bernd Giebel: Conception and design, final approval of manuscript. Ingrid Meulenbelt: Conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript. Yolande FM Ramos: Conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript.

Funding

The research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation program AutoCRAT under grant agreement No 874671. The material presented and views expressed here are the responsibility of the author(s) only. The EU Commission takes no responsibility for any use made of the information set out.

Declarations.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The project “Dose finding & potential adverse effects observations in the DMM model” was conducted at the Leiden University Medical Center and was approved by the Animal Welfare Committee (IvD) under number AVD1160020171405- PE.18.101.006 on Apr 21, 2023. The Medical Ethics Committee of the LUMC gave approval for generation of hiPSCs from skin fibroblasts of healthy donors under number P13.080 on July 2 2014 (Parapluprotocol: hiPSC). Informed consent was obtained from all donors.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.