Mesenchymal stromal cells therapy for remodeling the joint microenvironment: mechanisms, nanotechnology-enhanced strategies, and translation prospects

Abstract

Osteoarthritis (OA) is a complex, multifactorial whole-joint disease characterized by progressive articular cartilage degeneration, synovitis, and subchondral bone remodeling. Current clinical interventions primarily offer symptomatic management but do not halt or reverse disease progression. Recent advancements in regenerative medicine have emphasized mesenchymal stromal cell (MSC) therapy owing to its substantial potential in tissue repair and microenvironmental modulation. This review systematically evaluates the therapeutic efficacy and potential mechanisms of bone marrow-derived MSCs, adipose-derived stromal cells, umbilical cord-derived MSCs, synovium-derived MSCs, embryonic stem cell-derived MSCs and induced pluripotent stem cell-derived MSCs, as well as MSC-derived exosomes, in modulating the joint microenvironment. Furthermore, we discuss recent innovations in nanotechnology-enhanced strategies, designed to improve targeting specificity and therapeutic durability of MSC-based interventions. This review aims to establish a foundational framework and translational roadmap for the development of next-generation disease-modifying therapies for OA.

Graphical Abstract

Introduction

Osteoarthritis (OA) is a complex degenerative disease that can affect the whole joint, including bone, cartilage, synovium, ligaments, and muscles [1–3]. Globally, OA affected 595 million people in 2020 (7.6% of the global population), with a 132.2% increase in total cases since 1990, representing a significant burden on health expenditure [4]. Abnormal mechanical stress or injury is a common initiating factor of OA, subsequently disrupting joint tissue homeostasis [5]. These disturbances progress to anatomical and physiological impairments, such as cartilage degeneration, synovitis, and subchondral bone remodeling, driven by a complex interplay of genetic, aging, metabolic, hormonal, and inflammatory mechanisms [5–7]. The common clinical features of OA include joint pain, swelling, and functional impairment, which will reduce the patients’ quality of life [8]. Therefore, a comprehensive understanding of OA is essential for the development of targeted therapeutic interventions.

Current therapeutic modalities for OA predominantly include lifestyle interventions, physical rehabilitation, pharmacological management, and surgical arthroplasty [9–11]. Pharmacological treatment mainly consists of non-steroidal anti-inflammatory drugs and intra-articular hyaluronic acid (HA) injections [10]. While these interventions are commonly employed, they provide only temporary symptomatic relief and fail to address the fundamental degenerative processes within the joint microenvironment [12]. Total joint replacement remains the definitive surgical option for advanced OA; however, it is limited by prosthetic durability, risk of postoperative infection, and the potential need for revision procedures, especially in geriatric patients [13, 14]. These limitations highlight the critical need for novel therapeutic strategies aimed at microenvironmental remodeling to improve clinical outcomes.

Mesenchymal stromal cells (MSCs) have emerged as a promising strategy for modulating the joint microenvironment, offering disease-modifying capabilities with the potential to address the limitations of conventional therapies [15, 16]. The widespread recognition of this therapeutic potential is underscored by a 2025 bibliometric analysis of 2,341 studies, which indicates that MSC-based precision therapy has emerged as a global research trend. By resolving inflammatory tissue damage and restoring immune homeostasis, this therapy provides a promising approach for organ-protective regeneration [17]. With the deepening understanding of MSC biology, it is increasingly recognized that MSCs exert their therapeutic effects mainly through paracrine mechanisms, including the secretion of various bioactive components [18]. Among these, MSC-derived exosomes (MSC-Exos) have gained growing attention, as they can mitigate, to a certain extent, the limitations associated with MSC transplantation, such as poor in vivo survival and potential teratogenicity [18, 19]. Therefore, MSC-Exos hold great promise as a novel therapeutic strategy for OA, providing new research directions for delaying disease progression.

This comprehensive narrative review delineates the current applications of bone marrow-derived MSCs (BMMSCs), adipose-derived stromal cells (ADSCs), umbilical cord-derived MSCs (UCMSCs), synovium-derived MSCs (SMSCs), embryonic stem cell-derived MSCs (ESMSCs), induced pluripotent stem cell-derived MSCs (iMSCs), and MSC-Exos, elucidating their underlying mechanisms in remodeling the joint microenvironment. Furthermore, it highlights nanotechnology-enhanced strategies and evaluates recent technological advancements, aiming to serve as a valuable reference for future translational research and therapeutic strategies for OA. Studies were selected for inclusion in this review based on the following criteria: (1) original research articles focusing on OA, published in English between 2016 and 2026; (2) studies providing substantial in vivo evidence or clinical data regarding the regulation of the joint microenvironment; (3) research offering clear mechanistic insights into the therapeutic effects of MSCs or MSC-Exos; and (4) investigations featuring innovative nanotechnology-enhanced strategies. A comprehensive literature search was performed in the PubMed database to retrieve relevant studies. The search terms employed included: “bone marrow mesenchymal stromal cells”, “adipose-derived stromal cells”, “umbilical cord mesenchymal stromal cells”, “synovial mesenchymal stromal cells”, “infrapatellar fat pad-derived stromal cells”, “synovial fluid-derived mesenchymal stromal cells”, “embryonic stem cell-derived mesenchymal stromal cells”, “induced pluripotent stem cell-derived mesenchymal stromal cells”, “exosomes”, “engineered exosomes”, “hydrogel”, and “osteoarthritis”.

Pathological mechanisms of OA

OA is now widely recognized as a whole-joint disease involving coordinated pathological alterations across multiple anatomical compartments, including articular cartilage, synovium, subchondral bone, synovial fluid, and surrounding tissues [20–22]. The pathological mechanisms of OA are shown in Fig. 1.

Fig. 1.

The pathological mechanisms of OA

Articular cartilage degeneration

Articular cartilage degeneration constitutes a core pathological feature of OA [23]. In the initial phases of OA, alterations in chondrocyte phenotype precede the structural destruction of the extracellular matrix (ECM) [22]. A key pathogenic mechanism in osteoarthritic cartilage degeneration is hypertrophic differentiation of chondrocytes, a process typically involved in endochondral ossification but aberrantly reactivated in OA [24]. Hypertrophic chondrocytes display increased expression of type X collagen (COL10), Runt-related transcription factor 2 (RUNX2), matrix metalloproteinase-13 (MMP13), and a disintegrin and metalloproteinase with thrombospondin motifs 5 (ADAMTS5), concomitant with decreased expression of cartilage-specific markers such as COL2 and SRY-related high-mobility group-box gene 9 (SOX9) [25]. Although chondrocytes mount a compensatory response by upregulating TIMP1 to transiently counteract MMP-driven catabolism, this protective mechanism is ultimately overwhelmed during disease progression. This phenotypic shift facilitates ECM degradation, calcification, and chondrocyte apoptosis, characterized by the pathological upregulation of COL1 and the transition from hyaline cartilage to fibrocartilage. Simultaneously, the recruitment of additional proteases, such as MMP2 and cathepsin K, further accelerates cartilage catabolism [22, 26].

Additionally, cellular senescence emerges as a key factor of cartilage degeneration in OA [27]. While OA pathology encompasses diverse joint tissues and cell types, research on cellular senescence has primarily focused on chondrocytes [28]. Senescent chondrocytes are characterized by telomere attrition, elevated senescence-associated β-galactosidase (SA-β-gal) activity, and secretion of a senescence-associated secretory phenotype (SASP), comprising pro-inflammatory cytokines and mediators, including granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin-1 alpha (IL-1α), IL-1β, IL-6, IL-7, IL-8, monocyte chemoattractant protein-1 (MCP-1), MCP-2, MMP-1, MMP-3, MMP-10 and MMP-13 [29]. These factors promote autocrine signaling within the cartilage and paracrine signaling that propagates to other joint tissues, thereby amplifying the joint-wide pathological network [30]. Consistent with this, Jeon et al. demonstrated that the selective local clearance of senescent cells in a post-traumatic OA model significantly attenuates cartilage destruction and reduces pain [31].

Synovitis

During the initial phases of OA, synovitis is often initiated by cartilage debris or inflammatory mediators, leading to synovial hyperplasia, increased synovial membrane thickness, and compromised synovial fluid quality [32]. Under these conditions, the synovial membrane loses its physiological selective permeability, facilitating the enrichment of activated immune cells within the synovial fluid [33]. Synovial fibroblasts become activated in response to inflammatory stimuli, secreting pro-inflammatory cytokines such as IL-6, IL-8, C-C motif chemokine ligand 2 (CCL2), and CCL5, alongside matrix-degrading enzymes like MMP-3 [34–36]. These mediators not only amplify synovial inflammation but also propagate inflammatory signals to adjacent cartilage and subchondral bone, contributing to joint degeneration [37]. Elevated p16 expression serves as a reliable biomarker for cellular senescence within osteoarthritic synovial tissue, where these senescent cells actively sustain the inflammatory environment through the SASP [38]. Additionally, inflammatory mediators within synovial fluid can directly impair chondrocyte function by inhibiting the synthesis of lubricin and HA, establishing a deleterious feedback loop that exacerbates joint degeneration [32].

Subchondral bone remodeling

Subchondral bone alterations have traditionally been regarded as a hallmark of late-stage OA [39]. However, emerging evidence suggests that they may serve as initiating and actively pathogenic factors in OA development [40]. In early-stage OA, abnormal mechanical stress induces bone marrow lesions characterized by disrupted trabecular connectivity, increased porosity of the subchondral bone plate, and the formation of microcracks [5, 41]. These structural alterations compromise the shock-absorbing capacity of the subchondral bone, thereby accelerating cartilage degeneration [5]. Circulatory disturbances, specifically impaired venous drainage resulting in intraosseous hypertension, alongside the pathological invasion of sensory nerves and angiogenesis into the subchondral bone, are also critical drivers of OA-associated pain and progressive structural decay [42–44]. As the disease advances, osteoblastic activity escalates, leading to increased subchondral sclerosis and osteophyte formation [45]. Furthermore, senescence of osteoblasts and osteocytes within the subchondral bone has been shown to impair bone-cartilage crosstalk, further driving the progression of OA [38].

Interaction between cartilage degeneration, synovitis, and subchondral bone remodeling

As mentioned above, the pathophysiology of OA encompasses a synergistic and self-perpetuating vicious cycle involving articular cartilage, synovial membrane, synovial fluid, and subchondral bone [46].

Initial cartilage ECM degradation releases damage-associated molecular patterns (DAMPs) into the synovial fluid. These DAMPs diffuse to the synovial membrane, where they bind to pattern recognition receptors (e.g., TLR2 and TLR4) on synovial macrophages and fibroblasts, triggering the secretion of pro-inflammatory cytokines and chemokines [36, 47]. These mediators then diffuse back into the cartilage via the synovial fluid, suppressing chondrocyte anabolic activity while upregulating matrix-degrading enzymes (e.g., MMPs and ADAMTSs), thus accelerating a fluid-borne catabolic loop [36, 47].

Simultaneously, abnormal biomechanical loading and inflammatory mediators collaboratively induce subchondral bone remodeling. This remodeling disrupts the essential mechanical and nutritional support provided by the subchondral bone, while the secretion of vascular endothelial growth factors (VEGFs) further stimulates angiogenesis and intensifies the pathological crosstalk between cartilage and bone [33]. The progressive loss of cartilage results in joint space narrowing, increased intra-articular friction, and heightened mechanical stress on the subchondral bone, thereby perpetuating the cycle [43]. Moreover, this inflammatory state creates a hypoxic and acidic microenvironment, which activates matrix-degrading proteases such as cathepsin K and MMP-3 to further degrade COL1 and COL2, thereby aggravating the deterioration of the joint microenvironment [33].

Furthermore, emerging evidence identifies cellular senescence as a joint-wide phenomenon that orchestrates pathological crosstalk among all compartments [38]. Senescent chondrocytes release SASP factors into the ECM, which then diffuse into the synovial fluid. These mediators initiate pathological crosstalk with the synovial membrane, triggering inflammatory activation and SASP production in synovial fibroblasts [48, 49]. This interaction is further amplified by chondrocyte-derived extracellular vesicles (EVs), which carry pro-senescent factors that spread the SASP to the synovium [50]. In turn, the inflamed synovium reinforces cartilage decay by secreting additional proteases and cytokines into the joint space [48, 49]. Simultaneously, senescent osteocytes within the subchondral bone disrupt the osteochondral crosstalk, allowing bone-derived catabolic factors to further compromise cartilage stability [51].

These findings highlights the imperative for future therapeutic interventions to adopt a comprehensive, joint-organ-centric approach with coordinated strategies aimed at disrupting this complex pathological network.

Application of MSCs and MSC-Exos in joint microenvironment

Stem cells serve as the foundational biological substrate for tissue morphogenesis and regeneration [52]. The classification of stem cells is shown in Fig. 2 In the context of OA therapy, tissue-derived MSCs and pluripotent stem cell-derived MSCs, along with MSC-Exos, have garnered increasing attention for their potential to remodel the joint microenvironment [53, 54]. This section systematically reviews their specific applications and therapeutic efficacy.

Fig. 2.

Classification of stem cells. Based on their differentiation potency, stem cells are classified into four main categories: Totipotent stem cells (e.g., zygote) capable of forming a complete individual; Pluripotent stem cells that differentiate into all three germ layers, including embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs); Multipotent stem cells that have undergone lineage commitment to differentiate into cell types within a specific germ layer. They are most commonly represented by hematopoietic stem cells, and also include tissue-derived MSCs; and Unipotent stem cells, which are restricted to differentiating into a single type of terminally differentiated cell. From a developmental perspective, ESCs originate from the blastocyst’s inner cell mass; Adult stem cells (ASCs) are distributed across nearly all somatic tissues and serve as endogenous reservoirs for repair. Tissue-derived MSCs belong to this category; iPSCs are generated via somatic cell reprogramming, bypassing lineage restrictions to regain pluripotency similar to ESCs

Tissue-derived mesenchymal stromal cells

Bone marrow-derived mesenchymal stromal cells

BMMSCs are multipotent stromal progenitors isolated from the bone marrow hematopoietic niche, distinguished by their self-renewal capacity and ability to differentiate into osteogenic, chondrogenic, and adipogenic lineages [52]. They are primarily derived from the iliac crest, while they can also be harvested from the proximal humerus, tibia, and femur [55]. Current research suggests that, compared to direct differentiation, BMMSCs may promote tissue repair more through paracrine signaling, which mediates potent immunomodulatory and trophic effects [56, 57].

Clinical evidence demonstrates that both BMMSCs and bone marrow aspirate concentrate (BMAC) exhibit safety in the management of OA. Systematic reviews have confirmed that allogeneic BMMSC implantation ameliorates visual analogue scale (VAS) and Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores within a 12-month follow-up period, with adverse reactions limited to the local injection site and no reported long-term side effects [58, 59]. However, recent evidence has indicated that BMMSCs derived from either the posterior iliac crest or the proximal tibia do not demonstrate superiority over platelet-rich plasma (PRP) in terms of therapeutic efficacy for OA [60].

Currently, there is no consensus regarding the optimal injection dosage of BMMSCs. Meta-analyses suggest that a single injection of up to 2.5 × 10⁷ cells yields optimal clinical benefits, as higher doses have not demonstrated a clear dose-response relationship [61]. Similarly, a dose-escalation study (1 × 10⁷, 5 × 10⁷, and 1 × 10⁸ cells) by Muthu et al. noted that while the medium- and high-dose cohorts exhibited significantly improved radiological scores, no significant differences in clinical functional outcomes were observed among the three groups [62].

BMMSCs have also demonstrated substantial potential in the targeted management of focal chondral and osteochondral defects. Meta-analytical evidence from animal studies indicates that BMMSC-seeded implants can achieve high-quality histological integration with host tissues [63]. Furthermore, the incorporation of nerve growth factor into BMMSC-laden scaffolds has been reported to accelerate subchondral bone repair, thereby establishing a critical structural foundation for the regeneration of the overlying articular cartilage [64]. Compared to ADSCs, BMMSCs have also been reported to exhibit an enhanced potential for osteochondral repair [65].

BMMSCs possess potent immunomodulatory and regenerative properties that make them a promising modality for OA management [66, 67]. However, some evidence suggests that the repair tissue induced by BMMSCs may resemble fibrocartilage [68]. Unlike hyaline cartilage, this fibrocartilaginous tissue is characterized by inferior biomechanical properties and limited long-term durability [69]. Moreover, functional heterogeneity arises following in vitro expansion, with studies demonstrating that BMMSCs passaged beyond four doublings exhibit increased cytokine expression—including IL-6, IL-10, IL-1β, tumor necrosis factor-alpha (TNF-α), and vascular endothelial growth factor-C (VEGF-C)—resulting in altered immunoregulatory functions and reduced chondrogenic differentiation capacity [70]. Similarly, certain studies have indicated that BMMSCs may not exhibit clear clinical advantages over alternative therapeutic modalities, despite often entailing higher costs [60, 71].

Adipose-derived stromal cells

ADSCs constitute a population of multipotent stromal cells isolated from the stromal vascular fraction (SVF) of adipose tissue, distinguished by their self-renewal capacity and multi-lineage differentiation potential [72]. Owing to their relatively minimally invasive harvest, high cellular yield, and potent proliferative capacity [73], ADSCs have gained prominence as a promising cellular source for regenerative interventions in OA management [74].

Multiple randomized controlled trials have corroborated the efficacy and safety of intra-articular ADSC injections in OA patients, as evidenced by improvements in VAS, WOMAC scores, and Knee Injury and Osteoarthritis Outcome Score (KOOS) [75–81]. However, the ADIPOA2 phase 2b randomized clinical trial presented a contrasting perspective; this study demonstrated that a single intra-articular injection of autologous ADSCs at doses of either 2 × 10⁶ or 10 × 10⁶ cells failed to significantly improve pain or function compared to saline placebo at the 12-month follow-up [82]. Regarding cartilage regeneration, the evidence remains similarly inconsistent. Several studies reported measurable increases in cartilage thickness or volume within 6 to 12 months post-intervention [75, 79, 80]. In contrast, Kim et al. found no statistically significant reduction in cartilage defect size by MRI in both a large-scale Phase III RCT and a 5-year mid-term follow-up study [77, 78].

The optimization of dosing strategies also remains unresolved; meta-analyses indicate that although high-dose injections (> 50 × 10⁶ cells) surpass the therapeutic outcomes of moderate-dose (25–50 × 10⁶ cells) and low-dose (< 25 × 10⁶cells) groups in terms of early-stage pain relief and functional recovery, the incidence of adverse events increases in a dose-dependent manner [76]. Investigations into dosing frequency suggest that multiple intra-articular injections may provide more sustained anti-inflammatory effects and enhanced chondroprotective benefits compared to a single dose [83].

Collectively, these findings suggest that ADSCs hold therapeutic promise for OA, yet the inconsistency across trials likely reflects unresolved variables in dosing regimens, injection frequency, and patient selection. Optimization of these parameters may be essential to better enhance the clinical efficacy of ADSC-based therapies.

Umbilical cord-derived mesenchymal stromal cells

UCMSCs are multipotent progenitor cells located within neonatal umbilical cord tissue [15]. Relative to other tissue-derived MSCs, UCMSCs display more primitive stemness traits, evidenced by their enhanced proliferative potential and diminished immunogenicity [84, 85]. In OA therapeutics, UCMSCs have gained attention as a viable cell-based intervention owing to their ease of standardization, high yield, regenerative and immunomodulatory capabilities [85, 86].

To date, several clinical trials have substantiated the long-term safety and preliminary efficacy of UCMSCs in the treatment of OA [87, 88]. Regarding administration strategies, current evidence indicates that repeated semiannual injections (2.0 × 10⁷cells) exhibit more favorable outcomes compared to a single injection [89]. Additionally, moderate-to-low doses (2.0 × 10⁷ and 2.0 × 10⁶ cells) have been shown to be more effective than high-dosage regimens (8.0 × 10⁷ cells), with joint swelling observed in all subjects within the high-dosage group [90]. Furthermore, evidence suggests that therapeutic performance can be further enhanced through the utilization of UCMSC-EVs [91], the adoption of subchondral rather than traditional intra-articular injection [92] and the selection of early-passage cells [93].

Despite the therapeutic potential of UCMSCs in the management of OA, their clinical translation remains constrained by several challenges. A study evaluating the combined use of HA, hUCMSCs, and synthetic human growth hormone indicated that while the treatment could alleviate clinical symptoms, it did not significantly promote articular cartilage regeneration [94]. Furthermore, prolonged in vitro expansion leads to telomere shortening and loss of telomerase activity, which compromises the chondrogenic differentiation potential of UCMSCs as the passage number increases [93].

Synovium-derived mesenchymal stromal cells

SMSCs, isolated from synovial membrane tissue, constitute a population of adult progenitor cells characterized by their self-renewal capacity and multipotent differentiation potential [95, 96].

Regarding administration strategies, studies suggest that early and periodic intra-articular administration of SMSCs can sustain cellular viability within the joint environment, effectively inhibiting cartilage degeneration and osteophyte formation [97, 98]. The influence of donor disease status on SMSC function remains controversial. A study utilizing a canine model suggested that the OA environment may adversely affect the proliferation and chondrogenic differentiation potential of SMSCs, potentially leading to reduced reparative efficacy [99]. Furthermore, SMSCs derived from OA patients tend to exhibit senescent characteristics, potentially compromising their regenerative capacity [100]. In contrast, investigations involving early-stage OA patients indicated that their SMSCs showed no significant differences in clonogenicity, multi-lineage differentiation potential, or immunophenotype compared to non-OA controls [101]. These inconsistencies may stem from variations in species, disease severity, and donor age, suggesting that the pathological characteristics of donor tissue should be carefully considered when developing individualized cell-based therapies [99–101].

The tissue-specific homing capacity of SMSCs relies on the expression of multiple adhesion molecules that facilitate their targeted migration to synovial and cartilage injury sites [102, 103]. Furthermore, the stability of this therapeutic modality is supported by evidence that cryopreserved SMSCs demonstrate comparable efficacy to freshly isolated cells, thereby enabling the development of off-the-shelf regenerative products [104]. However, the necessity for invasive procurement via arthroscopic synovial harvesting elevates patient morbidity and procedural risks, thereby hindering the clinical translation of SMSCs [105].

Comparative efficacy of different tissue-derived mesenchymal stromal cells

In comparative studies of BMMSCs and ADSCs, both a prospective clinical trial and a pilot study demonstrated that the two sources exhibit comparable therapeutic efficacy in improving knee pain and function [106, 107]. At the 6-month follow-up, no statistically significant differences were observed in VAS scores or functional indices [107]. However, a meta-analysis by Han et al. suggested that ADSCs might possess greater therapeutic potential than BMMSCs in ameliorating VAS and WOMAC scores [108].

Regarding the comparison between BMMSCs and UCMSCs, the two cell types manifest distinct biological profiles and clinical focuses. BMMSCs appear more prominent in mitigating the pathological progression of OA in radiographic evaluations [109] and exhibit enhanced ECM synthesis [110]. Specifically, although UCMSCs may produce higher quantities of COL2, they might concurrently express markers associated with fibrocartilage, such as COL1, and generate an ECM with less uniform quality and distribution compared to that derived from BMMSCs [110]. In contrast, UCMSCs may offer a more favorable safety profile, triggering milder local inflammatory responses compared to BMMSCs [111]. In addition, studies suggest that Wharton’s Jelly-derived MSCs (WJMSCs) possess a higher mitochondrial transfer efficiency compared to BMMSCs [112].

To further delineate the specific characteristics of these MSC types, Chen et al. conducted a network meta-analysis, revealing that BMMSCs may offer better outcomes in VAS and range of motion (ROM), while ADSCs are more effective for WOMAC scores. Additionally, UCMSCs appear to be more advantageous for Whole-Organ Magnetic Resonance Imaging Score (WORMS) [113]. These findings suggest that MSCs from different sources may possess distinct advantages in symptom relief, functional improvement, and tissue repair, resulting in non-uniform therapeutic efficacies.

While some studies suggested that MSC injections offer more favorable pain relief compared to other non-surgical therapies [87, 114], the recent MILES trial presented contradictory findings. At 12 months post-injection, no significant differences in efficacy were observed among autologous BMAC, autologous SVF, and allogeneic UCMSCs. Moreover, none of these MSC-based injection regimens demonstrated a clinical advantage over corticosteroid injections [71].

As a more cartilage-targeted cell source, SMSCs exhibit unique therapeutic properties [115]. Research indicates that intra-articular injection of SMSCs leads to higher secretion levels of lubricin compared to ADSCs, which may enhance joint lubrication and protect the cartilage [116]. Additionally, a canine study suggested that SMSCs possess greater proliferative capacity and chondrogenic potential relative to BMMSCs and ADSCs [117]. Their lower tendency toward hypertrophic differentiation further underscores their substantial value for clinical translation in the field of cartilage regenerative medicine [118, 119].

Other tissue-derived mesenchymal stromal cells

Infrapatellar fat pad-derived stromal cells (IPFSCs), located intracapsularly, are distinct from subcutaneous ADSCs due to their enhanced chondrogenic potential and age-independent biological activity [120, 121]. While preliminary clinical data support their safety and efficacy in symptom alleviation [122], challenges regarding their migratory impairment in inflammatory environments and potential for hypertrophy require further resolution [123].

Synovial fluid-derived MSCs (SFMSCs) represent a novel, relatively minimally invasive source obtainable via arthrocentesis, characterized by efficient expansibility and inherent adaptability to the joint environment [124, 125]. Although SFMSC administration has been shown to be safe in early trials [126], the quality of regenerated cartilage often lacks native hyaline characteristics, necessitating optimization to ensure effective tissue integration [125].

MSCs have also been successfully identified and isolated from various other anatomical reservoirs, such as the tendon [127], dental pulp [128], peripheral blood [129], placenta [130], and endometrium [131]. These alternative sources are expanding the therapeutic landscape by offering diverse options for allogeneic or autologous transplantation.

Pluripotent stem cell-derived mesenchymal stromal cells

Given the limitations inherent to tissue-derived MSCs, such as donor-related heterogeneity, cellular senescence, and limited expansion capacity, pluripotent stem cells (PSCs) have emerged as promising candidates for generating next-generation MSCs due to their favorable self-renewal capabilities [132–134]. These sources mainly include ESCs and iPSCs. Although undifferentiated PSCs are unsuitable for direct clinical application owing to tumorigenic risks and ethical complexities, they serve as a viable in vitro platform for the derivation of ESMSCs and iMSCs [135, 136].

ESMSCs can be efficiently generated from human ESC (hESC) lines, facilitating quality control and safety assurance [137]. Multiple preclinical studies utilizing animal models have demonstrated that hESMSCs exert protective and regenerative effects within the joint microenvironment. For instance, the transplantation of hESMSC spheroids was observed to effectively alleviate spontaneous osteoarthritic pathology in rhesus monkeys, suggesting the therapeutic efficacy of hESC-based interventions in promoting cartilage regeneration [138]. In rat OA models, the administration of hESMSCs effectively attenuated disease progression, increased aggrecan (ACAN) synthesis, and decreased ECM degradation [139].

As another class of PSC-derived MSCs, iMSCs have garnered increasing attention in recent years [140]. Compared to somatic tissue-derived MSCs, iMSCs retain similar morphology, immunophenotype, and tri-lineage differentiation potential, while demonstrating enhanced regenerative capacity [134]. iMSCs exhibit potential in promoting cartilage regeneration and modulating the inflammatory microenvironment [141]. Compared to BMMSCs, iMSCs exhibit lower pro-angiogenic activity, contributing to the inhibition of pathological neovascularization during the progression of OA [142].

Despite these promising prospects, the application of PSC-derived MSCs faces several challenges. The inherent tumorigenicity of residual undifferentiated PSCs represents a critical safety concern that impedes their clinical translation [135]. While iMSCs mitigate inter-donor heterogeneity, batch-to-batch variability remains a persistent issue [140]. Moreover, the application of ESMSCs involves ethical controversies regarding embryonic procurement [133]. Another biological hurdle is immunogenic rejection; specifically, inflammatory cytokines within the pathological microenvironment can upregulate surface HLA-I molecules on these allogeneic cells, thereby triggering host CD8⁺ T cell-mediated surveillance and clearance [143]. To meet clinical-grade requirements, current research is shifting toward small-molecule-driven targeted differentiation and xeno-free culture systems to ensure the production of consistent, robust, and safe cell populations [144]. Figure 3 summarizes the distinct characteristics of different tissue-derived MSCs and PSC-derived MSCs.

Fig. 3.

The characteristics of different tissue-derived MSCs and PSC-derived MSCs

Mesenchymal stromal cell-derived exosomes

EVs are nanoscale, membrane-bound vesicles actively secreted by cells that mediate intercellular communication via the transfer of genetic information from donor cells [145]. They are classified into three primary categories: Exos, microvesicles, and apoptotic bodies, distinguished by their size, biogenesis mechanisms, and membrane marker expression profiles [146, 147]. Structurally, MSC-Exos range from approximately 30–150 nm in diameter, encapsulated by a lipid bilayer membrane, and enriched with bioactive cargo including proteins, lipids, DNA, microRNAs (miRNAs), messenger RNAs, and metabolites [148, 149]. Functionally, they act as intercellular messengers that mediate immunomodulatory and regenerative activities through the delivery of bioactive cargo [150, 151]. Notably, it has been reported that BMMSCs, BMMSC-MPs, and BMMSC-Exos could equally protect against joint damage in vivo and exhibit similar immunosuppressive capacities in vitro, suggesting that MSC-Exos retain comparable therapeutic efficacy to their parent MSCs [152]. Moreover, their lower immunogenicity and tumorigenic potential further position them as promising therapeutic agents [153, 154]. Figure 4 illustrates the biogenesis and release of MSC-Exos.

Fig. 4.

The biogenesis and release of MSC-Exos

Although MSC-Exos exhibit promising therapeutic potential, current evidence remains predominantly focused on preclinical in vitro and in vivo studies, with clinical investigations still limited. The following studies employ various EV preparations, including EVs, small EVs (sEV), and Exos, which share overlapping characteristics and are discussed collectively. A randomized, triple-blind, placebo-controlled clinical trial conducted by Bolandnazar et al. demonstrated that intra-articular injection of placental MSC-EVs (7 × 10⁹ particles/ml) exhibited a favorable safety profile for knee osteoarthritis management; however, no significant differences were observed in pain alleviation or functional improvement compared to the placebo group [155]. Conversely, a recent single-patient first-in-human study utilizing UCMSC-sEV indicated that intra-articular administration (2 × 10¹⁰±0.5 × 10⁹ particles/joint) effectively ameliorated pain and enhanced functional recovery [156]. Further investigation has suggested that intra-articular injection (3 × 10¹¹, 4 × 10¹¹ or 5 × 10¹¹ particles/joint) of hUCMSC-Exos can also promote increases in articular cartilage thickness [157]. Although findings remain inconsistent across studies, these preliminary investigations collectively support the safety and feasibility of intra-articular MSC-EV administration, while highlighting the need for larger, well-designed clinical trials to establish their therapeutic efficacy in OA management. Tables 1 and 2 summarize the in vivo and clinical studies of MSCs and MSC-Exos.

Table 1.

In vivo studies on the treatment of OA with MSCs and MSC-Exos

Source	Model	Duration	Dose and route	Curative effect	Refs
ESMSCs (human)	Rhesus macaques, natural aging model	9 months	5 × 106 cells/joint, IA	1.Reduce articular swelling; 2.Increase knee joint extension degree; 3.Chondrogenic effect	[138]
BMMSCs (macaque)
ESMSCs (human)	SD rats, ACLT model	6 or 10 weeks	1 × 106 or 3 × 106 cells/joint, IA	1.Prevent the OA progression; 2.Multiple injections demonstrated superior therapeutic effects	[139]
BMMSCs (rat)	male Wistar rats, MIA model	4 weeks	5 × 106 cells/joint, IA	1.Anti-inflammatory effect; 2.Anti-oxidant effect	[158]
BMMSC-Exos (mouse)	C57BL/6 mice, collagenase model	6 weeks	0.25 µg/joint, IA	1.Chondroprotective effect; 2.Inhibit the degradation of cartilage ECM; 3.BMMSCs, MPs, Exos equally protected the joint in vivo; 4.BMMSCs, MPs and Exos exerted similar chondroprotective and anti-inflammatory function in vitro	[152]
BMMSC-MPs (mouse)			0.5 µg/joint, IA
BMMSCs (mouse)			2.5 × 105 cells/joint, IA
BMMSC-Exos (human)	Female BALB/c mice, ciprofloxacin model	36 days	2.5 µg/joint, IA	Chondrogenic effect	[159]
ADSC-Exos (human)
BMMSC-Exos	Male SD rats, ACLT + DMM model	8 weeks	100 µg/joint, IA	1.Reduce the senescence and apoptosis of chondrocytes; 2.Maintain the chondrocyte phenotype	[160]
BMMSCs			1 × 106 cells/joint, IA
BMMSC-Exos (rat)	Male SD rats, ACLT + MCLT model	7 weeks	NA, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect	[161]
BMMSC-Exos (human)	Male C57BL/6 mice, traumatic OA model	1 h	1010 particles/joint, IA	1.Promote chondrocyte migration; 2.Chondroprotective effect	[162]
BMMSC-Exos (mouse)	C57BL/6 mice, CIA or ACLT models	8 weeks	1010 particles/joint, IA	1.Promote osteoblast proliferation and differentiation; 2.Anti-inflammatory effect	[163]
BMMSC-Exos (human)	Male C57BL/6 mice, traumatic OA model	1 h	1010 particles/joint, IA	1.Enhance chondrocyte migration; 2.Inhibit cartilage degeneration	[164]
BMMSC-Exos (human)	Female C57B/L10 mice, collagenase model	4 weeks	7.5 µg/joint, IA	1.Chondrogenic effect; 2.Regulate cartilage ECM homeostasis	[165]
BMMSC-Exos (human)	Wistar rats, ACLT model	8 weeks	0.25 µg/joint, IA	1.Anti-inflammatory effect; 2.Alleviate OA damage	[166]
BMMSCs (horse)	French Standardbred horses, post-traumatic OA model	12 weeks	1 × 10⁷ cells/joint, IA	BMMSCs exhibit greater efficacy than UCMSCs	[109]
UCMSCs (horse)
ADSCs (rat)	Male SD rats, collagenase model	4 weeks	1 × 106 cells/joint, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect	[167]
ADSCs (human)	Male SD rats, ACLT model	7 weeks	1 × 106 cells/joint, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect	[168]
ADSC-Exos (human)			3 × 108 particles/joint, IA
ADSC-CM (rat)	SD rats, ACLT + MMx model	6 weeks	100 or 200 µl, IA	1.Anti-inflammatory effect; 2.Chondrogenic effect	[169]
ADSCs (mouse)	Male C57BL/6 mice, diabetes model	8 weeks	1 × 106 cells/joint, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect	[170]
ADSCs (Beagle dog)	Female C57 mice, DMM model	4 weeks	1 × 105 cells/joint, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect	[83]
	Female Chinese garden dogs, CrCLt model	8 weeks	1 × 107 cells/joint, IA
ADSCs (human)	Female New Zealand white rabbits, bilateral medial anterior hemimeniscectomies model	8 weeks	2 × 106, 6 × 106 or 12 × 106 cells/joint, IA	1.Immunomodulatory functions; 2.Favorable safety profile	[171]
ADSC-Exos (human)	C57BL/6J mice, ACLT model	4 weeks	100 µg/joint, IA	1.Enhance mitophagy; 2.Chondrogenic effect	[172]
ADSCs (rat)	Male SD rats, MIA model	1 or 2 weeks	1 × 106 cells/joint, IA	Chondrogenic effect	[173]
SVF (rat)			5 × 106 cells/joint, IA
UCMSC-Exos (human)	C57BL/6 mice, collagenase model	4 weeks	5 × 105 particles/joint, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect	[174]
UCMSC-Exos (human)	Male Wistar rats, MMx model	1 or 3 weeks	6 µg/joint, IA	1.Anti-inflammatory effect; 2.Enhance macrophages M2 polarization	[175]
UCMSC-MBVs (human)
UCMSC-EVs (human)	Male SD rats, ACLT model	4 or 8 weeks	NA, IA	1.Enhance macrophages M2 polarization; 2.Chondroprotective effect	[176]
UCMSCs (human)	Male Wistar rats, ACLT model	12 weeks	0.4 × 105 or 2 × 105cells/joint, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect	[177]
UCMSCs (human)	Male SD rats, MIA model	4 weeks	2 × 106 cells/joint, IA	1.Improve joint pathology; 2.Anti-inflammatory effect; 3.Chondroprotective effect	[178]
UCMSCs (human)	Male SD rats, ACLT model	8 weeks	5 × 105cells/joint, IA or SC	1.Regulate aberrant vascularization; 2.Regulate osteoblast apoptosis; 3.Regulate differentiation imbalance	[92]
UCMSC-Exos (human)	Male SD rats, ACLT + MCLT + DMM or collagenase models	4 weeks	5 × 109 particles/joint, IA	1.Anti-inflammatory effect; 2.Chondrogenic effect	[179]
UCMSCs (human)	Male Wistar rats, MIA model	5 weeks	2.5 × 105 cells/joint, IA	Chondroprotective effect	[180]
UCMSCs (human)	Female SD rats, MIA model	12 weeks	NA, IA	1.Anti-inflammatory effect; 2.Chondroprotective effect; 3.P3 and P8 UCMSCs exhibit superior therapeutic potential compared to P13	[93]
UCMSCs (human)	Male SD rats, MIA model	4 weeks	2.5 × 105 cells/joint, IA	1.Chondrogenic effect; 2.Enhance cartilage ECM synthesis and homeostasis; 3.Anti-inflammatory effect	[84]
UCMSCs (human)	Male SD rats, ACLT model	9 weeks	5 × 105 cells/joint, IA	1.Chondrogenic effect; 2.Enhance macrophages M2 polarization;	[91]
UCMSC-sEV (human)			30 µg/joint, IA
UCMSC-sEV (human)	C57BL/6j mice, collagenase model	6 weeks	2 × 108 particles/joint, IA	1.Enhance macrophages M2 polarization; 2.Chondrogenic effect; 3.Anti-inflammatory effect; 4.Immunosuppresive effect	[156]
UCMSC-Exos (human)	C57BL/6JNifdc mice, MIA model	5 weeks	1 × 108 particles/joint, IA	1.Chondrogenic effect; 2.Enhance cartilage ECM synthesis and homeostasis;	[157]
UCMSC-dECM (human)	Male Wistar rats, ACLT model	8 weeks	1 µg/joint, IA	1.Anti-inflammatory effect; 2.Attenuate chondrocyte senescence	[181]
UCMSC-EVs (human)	Male C57BL/6 rats, ACLT + DMM model	6 weeks	109 particles /joint, IA	1.Anti-inflammatory effect; 2.Inhibit the degradation of cartilage ECM	[182]
SMSC-Exos (human)	Male C57BL/6 rats, ACLT + DMM model	6 weeks	109 particles /joint, IA	Chondroprotective effect	[183]
SMSCs (rat)	Female Lewis rats, MMx model	8 weeks	1 × 106 cells/joint, IA	Thawed SMSCs showed comparable inhibitory effects to cultured SMSCs	[104]
IPFSCs (human)	SD rats, Hulth method	20 weeks	3.0 × 105 or 1.5 × 105 cells/joint, IA	1.Reduce chondrocyte ROS; 2.Anti-inflammatory effect	[184]
SMSCs (human)
SMSC-Exos (mouse)	Male C57BL/6J mice, DMM model	4 weeks	1 µg/joint, IA	1.Inhibit the degradation of cartilage ECM; 2.Chondroprotective effect	[185]
SMSCs (human)	Male Lewis rats, MMx model	1 day	2 × 106 cells/joint, IA	SMSCs express more lubricin than ADSCs	[186]
ADSCs (human)
SMSCs (rats)	Male Lewis rats, ACLT model	4 weeks	1 × 106 cells/joint, IA	1.Chondroprotective effect; 2.Anti-inflammatory effect; 3.Not single but periodic injections of SMSCs maintain viable cells	[98]
SMSC-Exos	Male SD rats, MIA model	8 weeks	1 × 1010 particles/joint, IA	1.Chondroprotective effect; 2.Enhance chondrocyte migration; 3.Enhance macrophages M2 polarization; 4.Anti-inflammatory effect	[187]
SMSC-Exos (human)	Male SD rats, ACLT + MCLT + MMx model	12 weeks	1 × 1010 particles/joint, IA	1.Chondroprotective effect; 2.Enhance chondrocyte migration	[188]
SFMSCs (human)	Male SD rats, DMM model	5 or 6 weeks	2 × 106 cells/joint, IA	Chondroprotective effect	[125]
DPSC-Exos (human)	Male C57BL/6 mice, MIA model	6 weeks	2 × 108 particles/joint, IA	1.Alleviate the abnormal subchondral bone remodeling; 2.Chondroprotective effect; 3.Anti-inflammatory effect	[189]

OA, osteoarthritis; ESMSCs, embryonic stem cell-derived mesenchymal stromal cells; BMMSCs, bone marrow-derived mesenchymal stromal cells; ADSCs, adipose-derived stromal cells; UCMSCs, umbilical cord-derived mesenchymal stromal cells; SMSCs, synovial membrane-derived mesenchymal stromal cells; iMSCs, induced pluripotent stem cell-derived mesenchymal stromal cells; ADSC-CM, adipose-derived stromal cell conditioned medium; IPFSCs, infrapatellar fat pad-derived stromal cells; SVF, stromal vascular fraction; SFMSCs, synovial fluid-derived mesenchymal stromal cells; DPSCs, dental pulp-derived mesenchymal stromal cells; Exos, exosomes; MBVs, matrix-bound nanovesicles; ABs, apoptotic bodies; EVs, extracellular vesicles; sEV, small extracellular vesicles; dECM, decellularized extracellular matrix; DMM, destabilization of the medial meniscus; ACLT, anterior cruciate ligament transection; MCLT, medial collateral ligament transection; MIA, monosodium iodoacetate; MMx, medial meniscectomy; CIA, collagen-induced arthritis; PTCD, partial thickness cartilage defect; CFA, complete Freund’s adjuvant; CrCLt, cranial cruciate ligamentectomy; IA, intra-articular injection; SC, subchondral injection; ROS, reactive oxygen species; ECM, extracellular matrix; MPs, microparticles

Table 2.

Clinical studies on the treatment of OA with MSCs and MSC-Exos

Source	Study design	Groups	K-L grade	Intervention	Duration	Therapeutic effects	Refs
BMMSCs	Phase I prospective clinical trial	13 knee OA patients	2 or 3	28–35 × 106 and 26–33 × 106 cells/joint	24 months	Knee cartilage thickness improved	[190]
BMMSCs	Nonrandomized, open-label, dose‐escalation phase I/II clinical trial	12 knee OA patients	3 or 4	1 × 106, 10 × 106, or 50 × 106 cells/joint	12 months	Safe and result in significant improvements in PROMs at 12 months	[191]
BMMSCs	Prospective randomized study	140 knee OA patients	2–4	Average 1.56 × 105 cells/joint	15 years	Have sufficient effect on pain to postpone or avoid the TKA in the contra lateral joint of patients with bilateral osteoarthritis	[192]
ADSCs	Phase I clinical trial	11 knee OA patients low-dose cohort: (n = 5) high-dose cohort: (n = 6)	2–4	6.7 × 106 or 4 × 107 cells/joint	48 weeks	Safe and well-tolerated	[193]
ADSCs	Phase I/II, randomized, active-control, single-blind, multiple-center clinical trial	57 knee OA patients HA group: (n = 8); 16 M group: (n = 17); 32 M group: (n = 17); 64 M group: (n = 15)	2 or 3	16 × 106, 32 × 106 or 64 × 106 cells/joint	96 weeks	Effective, safe, and well-tolerated	[194]
ADSCs	A randomized controlled trial	30 knee OA patients control group: (n = 10); one-injection group: (n = 10); two-injection group: (n = 10)	2 or 3	100 × 106 cells/joint	12 months	Effective and safe	[195]
SVF	Double-blinded prospective randomized controlled clinical trial	39 knee OA patients control group: (n = 13); low dose group: (n = 13); high dose group: (n = 13)	2 or 3	3 × 107 or 1.5 × 107 cells/joint	12 months	Significantly decrease knee OA symptoms and pain for at least 12 months	[196]
ADSCs	Prospective, randomized, open-label, blind end-point, 2-arm parallel, controlled trial	26 knee OA patients control group: (n = 13) ADSC group: (n = 13)	2–4	1 × 108 cells/joint	24 months	Reduce cartilage degeneration	[197]
ADSCs	Phase III, randomized, double-blind, placebo-controlled trial	261 knee OA patients ADSC group:131 patients; control group:130 patients	3	1 × 108 cells/joint	6 months	Significant pain relief and functional improvements	[77]
ADSCs	Prospective, randomized, double-blinded, placebo‐controlled trial	24 knee OA patients MSC group:12 patients; control group:12 patients	2–4	1 × 108 cells/joint	6 months	Satisfactory functional improvement and pain relief	[198]
ADSCs	Phase II, triple-blinded, placebo controlled, randomized trial	40 knee OA patients control group: (n = 20); ADSC group: (n = 20)	2 or 3	100 × 106 cells/joint	12 months	Slight increase in the thickness of tibial and femoral articular cartilages	[79]
ADSCs	Interventional triple-blind clinical trial	20 knee OA patients control group: (n = 10); ADSCs group: (n = 10)	2–4	0.5 × 108 cells/joint	6 months	Safe and effective in reducing pain and increasing cartilage thickness	[75]
UCMSCs	Phase I clinical trial	40 knee OA patients low dose group: (n = 16); medium dose group: (n = 16) high dose group: (n = 8)	1–3	2 × 106, 20 × 106 or 80 × 106 cells/joint	24 weeks	Low and middle doses were more efficient in OA patients	[90]
UCMSCs	Phase I/II randomized, double-blind, controlled trial	29 knee OA patients HA group: (n = 9); MSC-1 group: (n = 10) MSC-2 group: (n = 10)	1–3	20 × 106 cells/joint	12 months	Repeat UCMSCs treatment is safe and superior to active comparator in knee OA at 1-year follow‐up	[89]
UCMSCs	Open-label, single‐arm, single‐center, phase I/II clinical trial	7 knee OA patients low dose group: (n = 4); high dose group: (n = 3)	3	1.15–1.25 × 107 or 1.65-2.00 × 107 cells/joint	7 years	Safe and effective for the regeneration of durable articular cartilage	[88]
UCMSCs	Single‑arm, open‑label study	29 knee OA patients	1–4	10 × 106 cells/joint	12 months	Effective and the maximum effect was achieved after 6 months of injection	[199]
PMSC-EVs	Randomized, triple-blind, placebo-controlled clinical trial	31 knee OA patients (62 knees) normal saline group: (n = 31); PMSC-EVs group: (n = 31)	2 or 3	7 × 109 particles/ml	6 months	Safe but does not improve clinical symptoms	[155]
BMMSCs	Randomized, double-blind, placebo-controlled study	24 knee OA patients control group: (n = 12); MSCs group: (n = 12)	1–4	1 × 108 cells/ joint	12 months	Allogeneic BMMSCs have a preventive effect on OA progression	[59]
BMAC	Prospective, multi-center	29 knee OA patients (37 knees)	3 or 4	5 ml/joint	4 years	1.Improve IKDC and WOMAC scores; 2.Improve walking distance	[200]
BMAC	Single-centre, parallel, randomised controlled unblinded study	90 knee OA patients crest Group: (n = 30); tibia Group: (n = 30); PRP Group: (n = 30)	1–4	5 ml/joint	6 months	Both sources of BMAC have a beneficial clinical outcome but not superior over PRP	[60]
PRP				0.8-1 × 109/ml
BMAC	Phase I controlled dose-escalation study	75 knee OA patients low-dose group: (n = 25); medium-dose group: (n = 25); high-dose group: (n = 25)	1 or 2	10 × 106 or 50 × 106 or 100 × 106 cells/joint	12 months	Safe and effective for pain relief and functional improvement, with no significant differences among dose groups	[62]
BMAC	Phase 2/3, four-arm parallel, multicenter, single-blind, randomized, controlled clinical trial	480 knee OA patients BMAC group: (n = 119); SVF group: (n = 121); UCMSCs group: (n = 119) CSI group: (n = 121)	2–4	NA	12 months	BMAC, SVF and UCMSCs have no superior effects to CSI	[71]
SVF
UCMSCs
CSI
ADSCs	Retrospective analysis of prospectively collected data	24 knee OA patients study group: (n = 12); control group: (n = 12)	2 or 3	1 × 108 cells/joint	5 years	1.VAS and WOMAC scores improved significantly at 6 months; 2.the chondral defect size on MRI or other radiologic evaluations did not change significantly	[78]
ADSCs	Phase I, prospective, bicentric, single-arm, open-label, dose-escalating clinical trial	18 knee OA patients low-dose group: (n = 6); medium-dose group: (n = 6); high-dose group: (n = 6)	3 or 4	2 × 106 or 10 × 106 or 50 × 106 cells/joint	6 months	Safe and effective in pain levels and function improvement	[81]
ADSCs	Prospective, 3-arm phase 2b, multicentric, randomised, double-blind controlled trial	135 knee OA patients control group: (n = 45); low-dose group: (n = 45); high-dose group: (n = 45)	2 or 3	2 × 106 or 10 × 106 cells/joint	12 months	Safe but no significant improvement in pain and function compared with the placebo group	[82]
UCMSCs	Randomized, prospective, double-blind, controlled, and parallel-group pilot study	30 knee OA patients Triamcinolone group: (n = 15); UCMSCs group: (n = 15)	2 or 3	5 × 106 ± 5 × 105 cells/joint	12 months	Safe and effective in pain levels and function improvement	[87]
UCMSCs	Four-arm, double-blind randomized trial	28 knee OA patients (51 knees) HA group: (n = 13 knees); HA+UCMSCs group: (n = 12 knees); HA+somatotropin group: (n = 12 knees) HA+somatotropin+UCMSC group: (n = 14 knees)	1 or 2	1 × 107 cells/joint	12 months	The combined use of HA, hUCMSCs, and somatotropin improved symptoms of knee OA, but did not enhance cartilage regeneration significantly	[94]
BMAC	Prospective comparative clinical trial	102 knee OA patients BMAC group: (n = 51); ADSCs group: (n = 51)	2–4	10 ml/joint	6 months	1. Safe and effective in pain levels and function improvement; 2. No significant clinical and functional difference between BMAC and ADSCs	[107]
ADSCs
IPFSCs	Phase 1, open-label, non-comparative and single-center clinical trial	12 knee OA patients	2 or 3	5 × 107 cells/joint	48 weeks	Safe and effective in pain levels and function improvement	[122]
PMSCs	Double-blind, placebo-controlled clinical trial	20 knee OA patients control group: (n = 10); hPMSCs group: (n = 10)	2–4	0.5–0.6 × 108 cells/joint	24 weeks	Safe and effective in pain levels and function improvement at 24 weeks follow-up	[201]
PMSCs	Non-randomized, open-label study	26 knee OA patients control group: (n = 11); hPMSCs group: (n = 15)	2 or 3	6 × 107 cells/joint	12 months	Cryopreserved hPMSCs in combination with HA are safe and effective for treating KOA	[130]
UCMSC-sEV	First-in-human case report	1 knee OA patients sEV group: (n = 1)	2	2 × 1010 ± 0.5 × 109 particles/joint	12 months	Safe with no severe adverse effects after a 12 month follow-up period	[156]
UCMSC-Exos	Prospective, randomized, double-blind, ascending dose study	41 knee OA patients (45 knees) low-dose group: (n = 15); medium-dose group: (n = 15); high-dose group: (n = 15)	2 or 3	3 × 1011, 4 × 1011 or 5 × 1011 particles/joint	9 months	Safe and effective in pain levels and function improvement	[157]

OA, osteoarthritis; BMMSCs, bone marrow-derived mesenchymal stromal cells; ADSCs, adipose-derived stromal cells; UCMSCs, umbilical cord-derived mesenchymal stromal cells; SMSCs, synovial membrane-derived mesenchymal stromal cells; Exos, exosomes; EVs, extracellular vesicles; sEV, small extracellular vesicles; BMAC, bone marrow aspirate concentrate; SVF, stromal vascular fraction; CSI, corticosteroid injection; IPFSCs, infrapatellar fat pad-derived mesenchymal stromal cells; PMSCs, placenta-derived mesenchymal stromal cells; PRP, platelet-rich plasma; KOA, knee osteoarthritis; HA, hyaluronic acid; PROMs, patient-reported outcome measures; TKA, total knee arthroplasty; VAS, visual analogue scale; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index; IKDC, International Knee Documentation Committee

Mechanisms of MSCs and MSC-Exos in remodeling the joint microenvironment

Given their multifaceted biological functions, including regenerative repair, immunomodulation, and microenvironment remodeling, MSCs and MSC-Exos demonstrate broad application prospects in the treatment of early-to-mid-stage OA [53, 54]. Figure 5 highlights the key mechanisms of MSCs and MSC-Exos in remodeling the joint microenvironment and Table 3 summarizes these mechanisms. Specifically, within the articular cartilage microenvironment, MSCs and MSC-Exos balance ECM anabolism and catabolism, promote chondrocyte proliferation and migration while suppressing senescence and programmed cell death, and modulate energy metabolism and oxidative stress. In the synovial microenvironment, they regulate macrophage polarization and suppress inflammatory hyperplasia and fibrosis of synovial fibroblasts. In the subchondral bone microenvironment, they modulate the coupling of osteoblasts and osteoclasts, and inhibit the formation of H-type vessels and pathological innervation.

Fig. 5.

The key mechanisms of MSCs and MSC-Exos in remodeling the joint microenvironment

Table 3.

The interaction and mechanism of MSCs and MSC-Exos in remodeling the joint microenvironment

Interaction (source→target)	Experimental model	Key cargo and mechanism	The expression of related factors	Outcome	Ref
IPFSC-CM→chondrocytes	In vitro, chondrocytes from OA patients, IL-1β induced	modulate PI3K/AKT/NF-κB signaling pathway	COL2↑, ACAN↑, MMP13↓, ADAMTS5↓, clv-PARP↓, Caspase-3↓, BAX↓, BCL-2↑, TNF-α↓, IL-6↓, iNOS↓, COX-2↓	1.Restore ECM homeostasis; 2.Anti-inflammatory effect; 3.Suppress apoptosis	[202]
ADSC-Exos→chondrocytes	In vitro, Chondrocytes, IL-1β induced	miR-93-5p, target ADAMTS9, activate PI3K/AKT/mTOR signaling pathway	ADAMTS9↓, IL-6↓, IL-1β↓, TNF-α↓, iNOS↓, PI3K↑, p-mTOR↑, p-AKT↑, BCL-2/BAX↑, LC3-II/I↓	Inhibit autophagy and apoptosis	[203]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes from OA patients or healthy controls; In vivo, C57BL/6J mice, DMM models	circRNA_0001236, modulates miR-3677-3p/Sox9 axis	COL2↑, SOX9↑, MMP13↓	Restore ECM homeostasis	[204]
SMSC-Exos→chondrocytes	In vivo, SD rats, DMM models	miR-320c, target ADAM19-dependent Wnt signalling	IL-6↓, IL-1β↓, TNF-α↓, MMP3↓, MMP13↓, COL2↑, ACAN↑ clv-Caspase-3↓, BCL-2↑, ADAM19↓, β-catenin↓, MYC↓	1.Restore ECM homeostasis; 2.Anti-inflammatory effect; 3.Suppress apoptosis	[205]
ADSC-Exos→chondrocytes	In vitro, chondrocytes from C57/B6J mice, IL-1β induced; In vivo, C57B/6J mice, DMM models	miR-99b-3p, inhibit ADAMTS4 expression	ADAMTS4↓, ACAN↑, COMP↑, COL2↑	Restore ECM homeostasis	[206]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes; In vivo, C57BL/6 mice, traumatic OA models	miR-125a-5p, inhibit E2F2 expression	COL2↑, ACAN↑, SOX9↑, MMP13↓, E2F2↓	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation and migration	[162]
ADSC-Exos→chondrocytes	In vitro, primary chondrocytes from SD rats, IL-1β induced; In vivo, SD rats, DMM models	miR-574-3p, modulates CRIM1/BMPs signaling pathway	BMP2↓, BMP6↓, BMP9↓, ALP↓, RUNX2↓, MMP13↓, IL-1β↓, TNF-α↓, CRIM1↑	Restrain chondrocytes hypertrophy and inflammatory response	[207]
BMMSC-Exos→chondrocytes and synovial tissue	In vitro, chondrocytes from Wistar rats, IL-1β induced; In vivo, Wistar rats, ACLT models	miR-361-5p, target DDX20, inhibit NF-κB signaling pathway	iNOS↓, MMP3↓, MMP13↓, IL-18↓, IL-6↓, TNF-α↓, p-IκBα↓, nuclear p65↓, cytoplasm p65↑	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation; 3.Alleviates synovial inflammation and hyperplasia	[166]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes from OA patients; In vivo, C57B/L10 mice, collagenase models	miR-92a-3p, inhibit WNT5A expression	COL2↑, COL9↑, SOX9↑, ACAN↑, COMP↑, COL10↓, RUNX2↓, MMP13↓, WNT5A↓	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation and migration	[165]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes from SD rats, IL-1β induced	miR-127-3p, target CDH11, inhibit Wnt/β-catenin pathway	COL2↑, MMP13↓, CDH11↓, β-catenin↓	Restore ECM homeostasis	[208]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes from mice, LPS induced	miR-210, target Tnfrsf21, inhibit NF-κB pathway	COL2↑, SOX9↑, ACAN↑, clv-Caspase-3↓, IL-1β↓, TNF-α↓, IL-6↓, p-p65/p65↓, IκBα↑, Tnfrsf21↓	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation; 3.Inhibit apoptosis	[209]
BMMSC-Exos→chondrocytes	In vitro, chondrocytes, IL-1β induced	Inhibit lncRNA LYRM4-AS1 expression, modulates miR-6515-5p/GRPR signaling pathway	MMP13↓, AKT1↓, PTEN↑, GRPR↓, LYRM4-AS1↓, miR-6515-5p↑	1.Restore ECM homeostasis; 2.Inhibit apoptosis; 3.Promote chondrocyte proliferation	[210]
SMSC-Exos→chondrocytes	In vitro, primary chondrocytes from OA patients, IL-1β induced	miR-212-5p, inhibit ELF3 expression	ELF3↓, COL2↑, ACAN↑, SOX9↑, MMP1↓, MMP3↓, MMP13↓, IL-6↓, MCP-1↓, TNF-α↓	1.Restore ECM homeostasis; 2.Anti-inflammatory effect	[211]
ADSC-Exos→chondrocytes	In vitro, chondrocytes from SD rats, IL-1β induced; In vivo, SD rats, MIA models	miR-429, inhibit FEZ2 expression	FEZ2↓, Beclin 1↑, COL2↑, LC3-II/I↑	Promote chondrocyte autophagy and proliferation	[212]
BMMSC-Exos→the whole joint	In vivo, SD rats, ACLT+MCLT models	miR-9-5p, inhibit SDC1 expression	IL-1↓, IL-6↓, TNF-α↓, CRP↓, NO↓, iNOS↓, SOD↑, MDA↓, COX2↓, MMP13↓, OCN↓, COMP↓, SDC1↓	1.Restore ECM homeostasis; 2.Anti-inflammatory effect; 3.Suppress oxidative stress	[161]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes; In vivo, C57B/L6 mice, post-traumatic models	miR-136-5p, inhibit ELF3 expression	SOX9↑, ACAN↑, COL2↑, MMP13↓, ELF3↓	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation and migration	[164]
UCMSC-dECM→chondrocytes	In vitro, primary chondrocytes from Wistar rats, H2O2 induced; In vivo, Wistar rats, ACLT models	Inhibit STING expression, inhibit NF-κB signaling pathway	IL-6↓, COX-2↓, iNOS↓, MMP3↓, MMP13↓, ADAMTS5↓, ACAN↑, COL2↑, p16↓, p21↓, p-IKKβ↓, p-p65↓, p-IκBα↓, IκBα↑	1.Inhibit chondrocyte senescence; 2.Restore ECM homeostasis; 3.Anti-inflammatory effect	[181]
BMMSC-Exos→chondrocytes	In vitro, chondrocytes from rats, IL-1β induced; In vivo, SD rats, MIA models	miR-326, target HDAC3, activate STAT1/NF-κB p65 signaling pathway	COL2↑, SOX9↑, ACAN↑, PRG4↑, IL-1β↓, IL-18↓, IL-6↓, TNF-α↓, GSDMD↓, Caspase-1↓, NLRP3↓, ASC↓, p65↓, HDAC3↓	1.Inhibit pyroptosis; 2.Restore ECM homeostasis 3.Anti-inflammatory effect	[213]
SMSC-Exos→chondrocytes	In vitro, chondrocytes, IL-1β induced	miR-129-5p, inhibit HMGB1 expression	COX2↓, iNOS↓, MMP13↓, COL2↑, HMGB1↓, TLR4↓, p-NF-κB/NF-κB↓, clv-Caspase-3↓, BCL-2↑	1.Anti-inflammatory effect; 2.Inhibit apoptosis; 3.Restore ECM homeostasis	[214]
MSC-Exos→chondrocytes	In vitro, chondrocytes from C57BL/6 mice, IL-1β induced; In vivo, C57BL/6 mice, collagenase models	lncRNA-KLF3-AS1, modulates miR-206/GIT1 axis	COL2↑, ACAN↑, MMP13↓, RUNX2↓, GIT1↑	1.Promote chondrocyte proliferation; 2.Inhibit apoptosis	[215]
MSC-Exos→chondrocytes	In vitro, chondrocytes from C57BL/6 mice, IL-1β induced	lncRNA-KLF3-AS1, target YBX1, activate PI3K/Akt/mTOR signaling pathway	LC3-II/I↓, p62↑, PI3K↑, p-Akt↑, p-mTOR↑	Inhibit autophagy and apoptosis;	[216]
DPSC-Exos→chondrocytes	In vitro, primary chondrocytes, IL-1β induced; In vivo, SD rats, MIA models	miR-140-5p	COL2↑, ACAN↑, SOX9↑, BCL-2↑, BAX↓, BAD↓	1.Restore ECM homeostasis; 2.Inhibit apoptosis	[217]
ADSC-Exos→chondrocytes	In vitro, ATDC5 cell line, IL-1β induced	miR-338-3p, inhibit RUNX2 expression	IL-6↓, IL-1β↓, PGE2↓, TNF-α↓, MMP3↓, MMP13↓, COL2↑, ACAN↑, RUNX2↓	1.Anti-inflammatory effect; 2.Restore ECM homeostasis; 3.Promote chondrocyte proliferation; 4.Inhibit apoptosis	[218]
ADSCs and SMSCs→chondrocytes	In vitro, chondrocytes from OA patients, IL-1β induced; In vivo, nude rats, MMx models	Activate FoxO1 signaling pathway	COL2↑, MMP13↓, IL-1β↓, TNF-α↓, IL-10↑, LC3-II/I↑, p62↓, FoxO1↑	1.Promote autophagy; 2.Promote chondrocyte proliferation	[219]
ADSCs→chondrocytes	In vitro, primary chondrocytes from SD rats, IL-1β induced; In vivo, SD rats, ACLT + DMM models	Downregulate miR-7-5p, promote ATG4A expression, inhibit mTORC1 signaling pathway	COL2↑, MMP13↓, SOX9↑, Beclin1↑, ATG3↑, LAMP1↑, LC3-II/I↑, ATG4A↑	1.Promote autophagy; 2.Anti-inflammatory effect; 3.Restore ECM homeostasis	[220]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes from trauma patients, IL-1β induced	lncRNA SNHG7, sponging miR-485-5p, promote FSP1 expression	FSP1↑, LDH↓, TNF-α↓, IL-6↓, GSH↑, MDA↓, ROS↓, GPX4↑, PTGS2↓	1.Inhibit ferroptosis; 2.Anti-inflammatory effect	[221]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes from rats, IL-1β induced; In vivo, SD rats, DMM models	Modulate METTL3-m6A-ACSL4 signaling pathway	Fe2+↓, MDA↓, GSH↑, ROS↓, ACSL4↓, METTL3↓	Inhibit ferroptosis	[222]
BMMSC-Exos→chondrocytes	In vitro, primary chondrocytes from SD rats, AGE induced	Inhibit Drp1 expression	MMP3↑, MMP13↑, Drp1↓, LC3-II/I↑, Beclin1↑	1.Promote mitophagy; 2.Inhibit apoptosis; 3.Restore ECM homeostasis	[223]
UCMSC-Exos→chondrocytes	In vitro, primary chondrocytes from OA patients	miR-100-5p, inhibit NOX4 expression	ROS↓, NOX4↓	Inhibit ROS production and apoptosis of chondrocytes	[224]
ADSC-Exos→chondrocytes	In vitro, primary chondrocytes from SD rats, IL-1β induced; In vivo, SD rats, ACLT + DMM model and C57BL/6J mice, DMM model	miR-199a-3p, inhibit mTOR signaling pathway	COL2↑, MMP13↓, LC3B↑, mTOR↓, p-p70S6K↓	1.Promote autophagy; 2.Restore ECM homeostasis	[225]
IPFSC-Exos→chondrocytes	In vitro, primary chondrocytes from OA patients, IL-1β induced; In vivo, C57BL/6 mice, DMM model	miR-100-5p, inhibit mTOR signaling pathway	COL2↑, MMP13↓, ADAMTS5↓, p62↓, LC3-II/I↑, p-mTOR↓, p-p70S6K↓	1.Promote autophagy; 2.Inhibit apoptosis; 3.Restore ECM homeostasis	[226]
SMSC-Exos→chondrocytes	In vitro, primary chondrocytes from C57BL/6J mice, IL-1β and IL-17 A induced; In vivo, C57BL/6J mice, DMM model	MATN3, interact with IL-17A, inhibit PI3K/AKT/mTOR signaling pathway	IL-6↓, TNF-α↓, ADAMTS5↓, MMP13↓, ACAN↑, COL2↑, LC3-II/I↑, Beclin1↑, MATN3↑, p-PI3K/PI3K↓, p-Akt/Akt↓, p-mTOR/mTOR↓	1.Promote autophagy; 2.Restore ECM homeostasis 3.Anti-inflammatory effect	[185]
DPSCs→ macrophages	In vitro, RAW264.7 cell line, LPS and IFN-γ induced; In vivo, SD rats, MIA and CFA model	Activate PINK1/Parkin signaling pathway	CD86↓, CD206↑, LC3-II/LC3-I↑, Beclin1↑, P62↓, PINK1↑, Parkin↑, Arg-1↑	1.Promote macrophage mitophagy and efferocytosis; 2.Polarize macrophage from M1 to M2	[227]
ShK-modified UCMSCs→macrophages and chondrocytes	In vitro, ATDC5 cell line, IL-1β induced; and RAW264.7 cell lines, LPS induced; In vivo, C57BL/6 mice, DMM models	ShK, inhibit PI3K/Akt signaling pathway	Kv1.3↓, iNOS↓, COX-2↓, ARG-1↑, IL-1β↓, IL-6↓, TNF-α↓, p-PI3K/PI3K↓, p-Akt/Akt↓, MMP3↓, MMP13↓, COL2↑, SOX9↑	1.Inhibit M1 macrophage polarization; 2.Restore ECM homeostasis; 3.Anti-inflammatory effect	[228]
CXCL9- preconditioned BMMSCs→macrophages	In vitro, macrophages from rat bone marrow; In vivo, SD rats, DMM models	CXCL9, activate TLR2/TGF-β2 signaling pathway	CD86↓, iNOS↓, IL-6↓, TNF-α↓, CD206↑, ARG-1↑, IL-10↑, TGF-β↑, TLR2↑, TGF-β2↑	Polarize macrophage from M1 to M2;	[229]
3D-cultured UCMSC-Exos→macrophages and chondrocytes	In vitro, C28/I2 cell lines; and THP-1 cell lines, PMA induced; In vivo, SD rats, ACLT models	miR-365a-5p, inhibit TLR2/Myd88/NF-κB signaling pathway	COL2↑, ACAN↑, MMP3↓, MMP13↓, CD163↑, CD206↑, ARG-1↑, TLR2↓, MyD88↓, NF-κB↓	1.Promote macrophage M2 polarization; 2.Restore ECM homeostasis	[230]
BMMSC-Exos→macrophages and chondrocytes	In vitro, chondrocytes from OA SD rats; In vivo, SD rats, Hulth models	lncRNA TUC339	TUC339↑, CD163↑, CD206↑, ARG-1↑, CD86↓, iNOS↓, HLA-DR↓, BAX↓, BCL-2↑, COL2↑, SOX9↑, COL10↓, RUNX2↓	1.Polarize macrophage from M1 to M2; 2.Inhibit chondrocytes apoptosis; 3.Restore ECM homeostasis	[231]
SMSC-Exos→macrophage and chondrocytes	In vitro, primary mononuclear cells from SD rats, LPS induced; Chondrocytes from SD rats, M1-CM induced; In vivo, SD rats, ACLT models	miR-3614-5p, target ANXA2, inhibit TLR4/MyD88/NF-κB signaling pathway	CD86↓, iNOS↓, ARG-1↑, CD206↑, IL-1β↓, IL-6↓, TNF-α↓, ANXA2↓, TLR4↓, MyD88↓, NF-κB↓, SOX9↑, COL2↑, ADAMTS5↓, MMP13↓	1.Polarize macrophage from M1 to M2; 2.Restore ECM homeostasis	[232]
TGF-β1 preconditioned BMMSC-Exos→macrophages	In vitro, macrophages from rat synovial tissues, LPS induced; In vivo, SD rats, ACLT + DMM models	miR-135b, inhibit MAPK6 expression	IL-1β↓, PGE2↓, COX-2↓, COX-1↓, NO↓, SOX9↑, ACAN↑, CD86↓, iNOS↓, ARG-1↑, CD163↑, MAPK6↓	1.Polarize macrophage from M1 to M2; 2.Restore ECM homeostasis	[233]
BMMSC-Exos→macrophage	In vitro, RAW264.7 cell line, LPS induced; In vivo, SD rats, ACLT + DMM models	Inhibit PINK1/Parkin signaling pathway	iNOS↓, CD86↓, CD206↑, ARG-1↑, IL-6↓, IL-1β↓, TNF-α↓, IL-10↑, PINK1↓, Parkin↓, p-Akt/Akt↑	1.Polarize macrophage from M1 to M2; 2.Anti-inflammatory effect;	[234]
ADSC-Exos→fibroblasts and chondrocytes	In vitro, CP-H096 cell line, IL-1β induced; primary synovial fibroblasts from rats, IL-1β or LPS induced; In vivo, SD rats, MIA models	miR-376c-3p, targeting WNT3 or WNT9a, inhibit Wnt/β-catenin signaling pathway	COL2↑, MMP13↓, ADAMTS5↓, ACAN↑, TNF-α↓, IL-6↓, IFN-γ↓, α-SMA↓, COL3↓, WNT3↓, WNT9a↓, β-catenin↓	1.Inhibit synovial fibrosis; 2.Anti-inflammatory effect;	[235]
ADSCs→osteoclasts	In vitro, CD11b+ osteoclasts from mice; In vivo, DBA/1J mice, collagenase models	CD39, inhibit RANKL/NF-κB signaling pathway	p65↓, p50↓, INF-γ↓, IL-17↓, TNF-α↓, RANKL↓, IgG↓	Inhibit osteoclastogenesis and bone erosion	[236]
3D-Cultured ADSC-sEV→osteoclasts and macrophages	In vitro, osteoclasts, THP1 cell line, RANKL and M-CSF induced; macrophages, THP1 cell line, LPS and IFN-γ induced; In vivo, mice, MIA models	miR-27b-3p, target M-CSF, inhibit PI3K/AKT signaling pathway	COL2↑, MMP13↓, IL-1β↓, TRAP↓, M-CSF↓, p-PI3K/PI3K↓, p-Akt/Akt↓, NFATc1↓, TNF-α↓, ROS↓	1.Inhibit osteoclastogenesis and bone erosion; 2.Inhibit M1 macrophage polarization	[237]
UCMSC-EVs→osteoclasts	In vitro, osteoclasts from primary bone marrow-derived macrophages induced; In vivo, SD rats, MIA models	miR let-7a-5p, inhibit ITGβ3 expression	CTSK↓, TRAP↓, CTR↓, NFATc1↓, DC-STAMP↓, ITGβ3↓	Inhibit osteoclastogenesis and bone erosion	[238]
BMMSC-Exos→osteoblasts	In vitro, primary osteoblasts from OA mice; In vivo, C57BL/6 mice, ACLT model	miR-206, inhibit Elf3 expression	ELF3↓, ALP↑, OCN↑, BMP2↑, TNF-α↓	Promote osteoblast proliferation and differentiation	[163]
DPSC-Exos→osteoclasts, cartilage and synovial tissue	In vitro, osteoclasts from RAW 264.7 cell line induced; In vivo, C57BL/6 mice, MIA models	Inhibit TRPV4 expression	COL2↑, IL-1β↓, TNF-α↓, TRAP↓, TRPV4↓	1.Inhibit osteoclastogenesis and bone erosion; 2.Restore ECM homeostasis; 3.Anti-inflammatory effect	[189]
BMMSC-Exos→sensory nerves, blood vessels, osteoclasts and chondrocytes	In vivo, C57BL/6 mice, resect L3-L5 spinous processes and ligaments	Inhibit RANKL-RANK-TRAF6 signaling pathway	ACAN↑, MMP13↓, CGRP↓, CD31↓, TRAF6↓, RANKL↓, TRAP↓, Endomucin↓	1.Inhibit sensory nerves invasion and angiogenesis; 2.Restore ECM homeostasis; 3.Inhibit osteoclastogenesis and bone erosion	[239]

IPFSC-CM, infrapatellar fat pad-derived mesenchymal stromal cells conditioned medium; hSMSCs, human synovial membrane-derived mesenchymal stromal cells; ADSCs, adipose-derived stromal cells; Exos, exosomes; BMMSCs, bone marrow-derived mesenchymal stromal cells; dECM, decellularized extracellular matrix; iMSC, induced pluripotent stem cell-derived mesenchymal stromal cells; UCMSCs, umbilical cord-derived mesenchymal stromal cells; DPSCs, dental pulp-derived mesenchymal stromal cells; ShK, Stichodactyla toxin; CXCL9, C-X-C motif chemokine ligand 9; 3D, three-dimensional; sEV, small extracellular vesicles; TGF-β1, transforming growth factor beta 1; TNF-α, tumor necrosis factor alpha; EVs, extracellular vesicles; OA, osteoarthritis; IL-1β, interleukin-1 Beta; DMM, destabilization of the medial meniscus; ACLT, anterior cruciate ligament transection; MCLT, medial collateral ligament transection; MMx, medial meniscectomy; MIA, monosodium iodoacetate; LPS, lipopolysaccharide; AGEs, advanced glycation end products; IFN-γ, interferon gamma; CFA, complete Freund’s adjuvant; PMA, phorbol 12-myristate 13-acetate; OVX, ovariectomy; M-CSF, macrophage colony-stimulating factor; L3-L5, lumbar vertebrae 3 to 5; PI3K, phosphoinositide 3-kinase; AKT, protein kinase B; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; lncRNA, long non-coding RNA; miRNA, micro RNA; circRNA, circular RNA; MyD88, myeloid differentiation primary response 88; ADAMTS, a disintegrin and metalloproteinase with thrombospondin motifs; mTOR, mammalian target of rapamycin; SOX9, SRY-Box transcription factor 9; ADAM19, a disintegrin and metalloproteinase 19; Wnt, wingless-type MMTV integration site family; RalA, RAS-like proto-oncogene A; E2F2, E2F transcription factor 2; CRIM1, cysteine-rich motor neuron protein 1; BMPs, bone morphogenetic proteins; DDX20, Asp-Glu-Ala-Asp-box polypeptide 20; CDH11, cadherin 11; GRPR, gastrin releasing peptide receptor; ELF3, E74 Like ETS Transcription Factor 3; FEZ2, fasciculation and elongation protein zeta 2; SDC1, syndecan-1; SDF-1, stromal cell-derived factor 1; STING, stimulator of interferon genes; HDAC3, histone deacetylase 3; STAT1, signal transducer and activator of transcription 1; HMGB1, high mobility group box 1; GIT1, GRK-interacting transcript 1; YBX1, Y-box binding protein 1; RUNX2, Runt-related transcription factor 2; FoxO1, forkhead box O1; ATG4A, autophagy-related 4 A cysteine peptidase; FSP1, ferroptosis suppressor protein 1; ACSL4, Acyl-CoA synthetase long chain family member 4; METTL3, methyltransferase-like 3; m6A, N6-methyladenosine; Drp1, dynamin-related protein 1; NOX4, NADPH oxidase 4; MATN3, matrilin 3; PINK, PTEN-induced kinase; TLR2, Toll-like receptor 2; ANXA2, annexin A2; MAPK6, mitogen-activated protein kinase 6; GOT1, glutamic-oxaloacetic transaminase 1; CCR2, C-C motif chemokine receptor 2; p53, tumor protein p53; Nrf2, nuclear factor erythroid 2-related factor 2; HO-1, heme oxygenase 1; RANKL, receptor activator of NF-κB ligand; ITGβ3, Integrin beta 3; TRPV4, transient receptor potential cation channel subfamily V member 4; TRAF6, TNF receptor associated factor 6; COL1/2, collagen type I/collagen type II; ACAN, aggrecan; MMP13, matrix metallopeptidase 13; PARP, Poly (ADP-ribose) polymerase; BAX, BCL2 associated X; BCL-2, B-cell lymphoma 2; BAD, BCL2-associated death modifier; iNOS, inducible nitric oxide synthase; COX2, cyclooxygenase-2; LC3B, microtubule-associated protein 1 light chain 3; COMP, cartilage oligomeric matrix protein; ALP, alkaline phosphatase; CRP, C-reactive protein; NO, nitric oxide; SOD, superoxide dismutase; GSH, glutathione; MDA, malondialdehyde; OCN, osteocalcin; GSDMD, gasdermin D; PGE2, prostaglandin E2; PTGS2, prostaglandin-endoperoxide synthase 2; α-SMA, α-smooth muscle actin; NFATc1, nuclear factor of activated T cells 1; Arg-1, arginase-1; PRG4, proteoglycan 4; LC3-Ⅱ/Ⅰ, microtubule-associated protein 1 A/1B-light chain 3; MYC, myelocytomatosis oncogene; MCP-1, monocyte chemoattractant protein-1; NLRP3, NOD-like receptor thermal protein domain associated protein 3; LDH, lactate dehydrogenase; ROS, reactive oxygen species; GPX4, glutathione peroxidase 4; CD86, cluster of differentiation 86; Kv1.3, Voltage-gated potassium channel 1.3; THP-1, human monocytic leukemia cell line 1; CTSK, cathepsin K; p, phosphorylated; clv, cleaved

Articular cartilage microenvironment

Balancing the ECM anabolism and catabolism

In the pathological microenvironment of OA, inflammatory cytokines secreted by the synovium into the synovial fluid continuously stimulate chondrocytes, disrupting their metabolic homeostasis [25]. This disruption leads to a reduction in the synthesis of anabolic markers (e.g., COL2 and ACAN), coupled with the overexpression of catabolic markers (e.g., MMP13 and ADAMTS5), ultimately resulting in the degradation of the ECM [12]. Consequently, restoring the balance between ECM anabolism and catabolism represents a key therapeutic strategy for attenuating cartilage degeneration in OA.

MSCs counteract ECM degradation by secreting endogenous protease inhibitors and cytokines [240]. In vitro studies indicate that the conditioned medium from IPFSCs (IPFSCs-CM) promotes the expression of COL2 and ACAN by modulating the PI3K/AKT/NF-κB signaling pathway in chondrocytes [202]. Furthermore, evidence from both in vitro and in vivo studies demonstrates that MSCs exert paracrine effects that upregulate SOX6 and SOX9 while downregulating MMP-2/3/9/13 and ADAMTS-4/5/9, thereby restoring ECM homeostasis [177, 202, 241].

MSC-Exos may inhibit ECM degradation by delivering miRNAs that target metabolic enzymes and their upstream transcription factors. For instance, both in vitro and in vivo studies demonstrate that BMMSC-Exos deliver miR-136-5p and miR-125a-5p, which target ELF3 and E2F2 genes, respectively. This action suppresses the expression of MMP13, promoting chondrocyte migration and upregulating the expression of COL2, ACAN, and SOX9 [162, 164]. Additionally, ADSC-Exos deliver miR-99b-3p to target ADAMTS4, thereby suppressing ECM degradation [206].

Beyond the direct regulation of matrix metabolism-related molecules, MSC-Exos can also indirectly maintain ECM homeostasis by modulating key signaling pathways involved in inflammation and development. Among these, the NF-κB and Wnt/β-catenin signaling pathways are widely investigated. In vitro evidence indicates that hBMMSC-Exos deliver miR-210, which targets TNFSF21 and inhibits the activation of the NF-κB signaling pathway, thereby reducing chondrocyte apoptosis and ECM degradation [209]. Furthermore, both in vitro and in vivo evidence demonstrates that hBMMSC-Exos deliver miR-361-5p, which similarly blocks the activation of the NF-κB signaling pathway by targeting DDX20, thereby effectively suppressing the expression of MMP3, MMP13, and inflammatory cytokines [166]. Regarding the Wnt pathway, in vivo evidence demonstrates that SMSC-Exos deliver miR-320c to target ADAM19, further downregulating the expression of β-catenin and MYC. This cascade blocks the aberrant activation of the Wnt signaling pathway, thereby promoting chondrocyte proliferation and suppressing ECM degradation [205]. Furthermore, both in vitro and in vivo evidence reveals that hBMMSC-Exos deliver miR-92a-3p to target WNT5A, similarly suppressing its aberrant activation [165]. While Wnt5a/b carried by SMSC-Exos can activate the YAP pathway to promote chondrocyte proliferation and migration, this simultaneously inhibits ECM secretion. Exos derived from miR-140-5p-overexpressing SMSCs counteract this Wnt-induced ECM suppression by targeting RalA, thereby restoring ECM synthesis homeostasis while promoting proliferation [188].

Beyond miRNAs, circular RNAs (circRNAs) also participate in the regulation of ECM metabolism. BMMSC-Exos deliver circRNA_0001236, which functions as a competing endogenous RNA (ceRNA) for miR-3677-3p. By modulating the miR-3677-3p/Sox9 axis, it upregulates COL2 and SOX9 while inhibiting MMP13, promoting ECM repair [204].

Promote chondrocyte proliferation and migration while suppressing senescence and programmed cell death

In the pathological progression of OA, the synergistic effects of an inflammatory microenvironment and oxidative stress induce chondrocyte senescence and lead to programmed cell death (e.g., apoptosis and pyroptosis) [12]. This contributes to a progressive decline in chondrocyte viability and suppressed ECM synthesis, thereby accelerating the structural disintegration of articular cartilage [242]. Consequently, promoting chondrocyte proliferation and migration while suppressing senescence and programmed cell death represents a crucial therapeutic strategy to delay the progression of OA.

MSCs facilitate cartilage regeneration through multidimensional paracrine mechanisms. They secrete essential growth factors, including IGF-1 and FGF-2, which provide trophic support to drive cartilage regeneration [240]. UCMSCs secrete IL-10 and TGF-β1, which inhibit chondrocyte apoptosis by upregulating the anti-apoptotic protein BCL-2 while downregulating the pro-apoptotic proteins BAX and BAD [178]. Mechanistically, in vitro and in vivo evidence demonstrates that ADSCs restore autophagic activity by downregulating miR-7-5p levels in chondrocytes, thereby alleviating the suppression of ATG4A [220]. Moreover, the combined application of ADSCs and SMSCs exerts a synergistic effect, further enhancing chondrocyte autophagy via activation of the FoxO1 signaling pathway, which consequently suppresses inflammatory responses and ECM degradation [219]. Regarding cell recruitment and senescence inhibition, iMSCs recruit host BMMSCs via the SDF-1/CXCR4 axis to facilitate cartilage repair [243]. Furthermore, hUCMSC-dECM ameliorates chondrocyte senescence by moderating the STING-NF-κB pathway [181]. Regarding the inhibition of pyroptosis, hADSCs release sTNFR1 to block the TNFR1-mediated NLRP3/Caspase-1/GSDMD axis, thereby inhibiting chondrocyte pyroptosis at its source [244]. Beyond paracrine mechanisms, MSCs possess inherent chondrogenic differentiation potential, enabling them to differentiate into chondrocytes within specific microenvironments to repair tissue defects [240].

MSC-Exos are proposed to function by delivering specific miRNAs, lncRNAs, and enzymatic cargoes that interact with target receptors on chondrocytes to inhibit senescence and apoptosis, while promoting proliferation. Among the involved mechanisms, the PI3K/AKT/mTOR signaling pathway is widely investigated. In vitro evidence demonstrates that ADSC-Exos deliver miR-93-5p to target ADAMTS9, while MSC-Exos deliver lncRNA KLF3-AS1 to target YBX1. Both cargoes further activate the PI3K/AKT/mTOR signaling pathway, thereby preventing excessive autophagy and suppressing apoptosis in chondrocytes [203, 216]. In vivo evidence further reveals that ADSC-Exos deliver miR-429 to inhibit FEZ2, consequently activating AKT signaling and preventing the overactivation of autophagy, thereby exerting chondroprotective effects [212]. In contrast, other studies highlight the necessity of suppressing this pathway to restore homeostasis. Specifically, SMSC-Exos deliver MATN3 to interact with IL-17 A, inhibiting PI3K/AKT/mTOR signaling [185]. In addition, IPFSC-Exos delivering miR-100-5p and ADSC-Exos delivering miR-199a-3p can both inhibit the overactivation of mTOR, thereby promoting autophagy and suppressing apoptosis [225, 226].

Beyond the PI3K/AKT/mTOR signaling pathway, MSC-Exos also maintain chondrocyte homeostasis through other molecular mechanisms. In vitro and in vivo evidence demonstrates that UCMSC-Exos deliver essential functional miRNAs that suppress the p53 signaling pathway, effectively reversing the senescent phenotype [245]. Moreover, DPSC-Exos delivering miR-140-5p target apoptosis-related proteins, and SMSC-Exos delivering miR-129-5p target HMGB1, both of which are effective in inhibiting chondrocyte apoptosis [214, 217].

Regarding the inhibition of pyroptosis, BMMSC-Exos deliver miR-326 to target HDAC3; this modulation increases the acetylation levels of STAT1/NF-κB p65, thereby blocking inflammatory signal transduction and inhibiting chondrocyte pyroptosis [213].

Modulating energy metabolism and oxidative stress

Existing evidence indicates that mitochondrial dysfunction and mitochondrial DNA (mtDNA) variation are not merely pathological consequences of OA but are pivotal factors driving disease progression [246]. Within the inflammatory microenvironment of OA, chondrocytes exhibit reduced activity of mitochondrial respiratory chain complexes (particularly complexes II and III) and dissipation of membrane potential, leading to insufficient ATP synthesis [247]. This bioenergetic exhaustion not only compromises the energy supply required for ECM synthesis but also precipitates an explosive accumulation of reactive oxygen species (ROS), disrupting intracellular redox homeostasis [184]. Sustained oxidative stress further attacks mtDNA and membrane lipids, perpetuating a vicious cycle of mitochondrial damage and ROS amplification, which ultimately culminates in the metabolic impairment of cartilage tissue [248, 249].

Studies indicate that MSCs can improve chondrocyte mitochondrial function and alleviate oxidative stress-induced damage through multiple mechanisms. One prominent mechanism involves the direct transfer of healthy mitochondria from MSCs to damaged chondrocytes via tunneling nanotubes, F-actin-based intercellular tubular channels, thereby effectively reducing ROS levels and improving cellular energy metabolism [250, 251]. Furthermore, MSCs can encapsulate functional mitochondria into vesicles [252] or facilitate their transfer via cell fusion [248], thereby further exerting protective effects on damaged chondrocytes.

MSC-Exos can modulate chondrocyte function by delivering specific miRNAs, thereby activating antioxidant pathways. In vitro evidence indicates that hUCMSC-Exos deliver miR-100-5p to target and inhibit the expression of NOX4 in chondrocytes, thereby attenuating ROS production [224]. Beyond general ROS scavenging, MSC-Exos also specifically target lipid peroxidation-driven ferroptosis. In vitro evidence indicates that lncRNA SNHG7 delivered by BMMSC-Exos functions as a ceRNA to miR-485-5p, relieving the suppression of FSP1 and alleviating chondrocyte ferroptosis [221]. Furthermore, in vivo studies demonstrate that BMMSC-Exos downregulate METTL3, reducing m6A modification levels of ACSL4 mRNA, thereby blocking the ferroptosis process [222].

Synovial microenvironment

Modulating macrophage polarization

In the progression of OA, damaged chondrocytes release DAMPs that activate TLRs, thereby inducing the polarization of macrophages toward the M1 phenotype [36, 47]. M1 macrophages, which rely on glycolytic metabolism, secrete substantial amounts of IL-1β, TNF-α, and ROS, resulting in synovial inflammation and the degradation of the cartilage ECM [253]. MSCs and MSC-Exos have been demonstrated to effectively suppress the expression of M1 markers (iNOS, CD86) while concurrently upregulating M2 markers (ARG1, CD206, CD163), thereby restoring synovial immune homeostasis [253].

MSCs possess the capacity to inhibit macrophage M1 polarization via the paracrine secretion of specific bioactive proteins or metabolites. For instance, hDPSCs effectively suppress the activation of OA macrophages and their secretion of inflammatory factors, which is largely mediated by the paracrine secretion of HGF and TGF-β1, and involves the attenuation of the MAPK signaling pathway [254]. In vitro evidence indicates that DPSCs can activate the PINK1/Parkin signaling pathway, thereby promoting mitophagy to inhibit M1 polarization [227]. Moreover, evidence indicates that ADSCs secrete high levels of lactate, which inhibits the expression of pro-inflammatory genes in M1-like macrophages via epigenetic reprogramming induced by histone H3K27 acetylation [255].

MSC-Exos promote the macrophage M1-to-M2 phenotypic switch via the delivery of bioactive cargoes, including miRNAs and lncRNAs. Both in vitro and in vivo evidence indicates that SMSC-Exos deliver miR-3614-5p to specifically inhibit ANXA2, thereby blocking the TLR4/MyD88/NF-κB signaling pathway and promoting the polarization of macrophages from the M1 to the M2 phenotype [232]. Similarly, Exos derived from three-dimensional (3D) cultured hUCMSCs deliver miR-365a-5p to inhibit the TLR2/MyD88/NF-κB pathway [230]. BMMSC-Exos delivering lncRNA TUC339 have been confirmed to effectively promote the M1-to-M2 phenotypic switch, thereby reducing inflammation levels and enhancing chondrocyte activity [231]. Additionally, Exos derived from TGF-β1-preconditioned BMMSCs are enriched with miR-135b, which drives synovial macrophage polarization toward the M2 phenotype by targeting and downregulating MAPK6 expression [233].

Furthermore, MSCs and MSC-Exos can modulate the functional status of macrophages by influencing their metabolic reprogramming and oxidative stress states. DPSCs induce M2 polarization by enhancing PINK1/Parkin-dependent mitophagy to facilitate the clearance of damaged mitochondria [227]. Conversely, Li et al. observed that BMMSC-Exos inhibit LPS-induced overexpression of PINK1 and Parkin proteins, ultimately suppressing the macrophage inflammatory response by preserving mitochondrial integrity and reducing ROS production [234].

Suppressing inflammatory hyperplasia and fibrosis of synovial fibroblasts

Beyond inflammatory infiltration, synovial fibroblast-mediated abnormal proliferation and fibrosis represent another pivotal pathological basis driving the progression of OA. Under chronic inflammatory stimulation, fibroblast-like synoviocytes undergo aberrant proliferation and fibrotic transformation, culminating in synovial hyperplasia and stiffening [256]. This pathological synovial fibrosis not only compromises the physiological low-friction properties of the articular cavity but also impedes the diffusion and infiltration of nutrients from the synovial fluid to the cartilage, thereby exacerbating chondrocyte injury [257].

Both in vitro and in vivo studies indicate that UCMSCs can remodel the phenotype of synoviocytes via paracrine mechanisms, significantly downregulating the expression of MMP-1, MMP-3, and MMP-13, thereby mitigating the degradation of the cartilage ECM [258]. Furthermore, hADSC-Exos deliver miR-376c-3p, which targets WNT3 and WNT9a in synovial fibroblasts to suppress the aberrant activation of the Wnt/β-catenin signaling pathway, consequently attenuating the severity of synovial hyperplasia and fibrosis [235].

Subchondral bone microenvironment

In the early stages of OA, aberrantly elevated osteoclast activity disrupts the remodeling equilibrium with osteoblasts, leading to a predominance of bone resorption over bone formation [43]. This pathological process manifests as thinning of the subchondral bone plate, trabecular rarefaction, and microstructural degeneration, collectively indicating bone loss [43]. As the disease progresses to advanced stages, aberrant alterations in mechanical loading induce excessive osteoblastic activity, which conversely manifests as subchondral bone sclerosis, thickening, cyst formation, and osteophytosis [5]. Furthermore, the aberrant angiogenesis of H-type vessels and the invasion of nerve fibers not only disrupt the integrity of the osteochondral interface but also exacerbate cartilage degeneration and provoke pain through the secretion of algogenic factors and proteases [38].

MSCs regulate the osteoblast-osteoclast coupling mechanism via paracrine signaling or direct cell-to-cell contact, thereby restoring the homeostasis of subchondral bone remodeling [259]. Mechanistically, MSCs constitutively secrete osteoprotegerin, which competitively binds to RANKL, consequently downregulating the expression of NFATc1 and Cathepsin K in osteoclasts [259, 260]. Concurrently, MSCs target and antagonize the RANKL signaling pathway via the CD200 and CD39 axes, effectively suppressing osteoclastogenesis [236, 259].

Regarding the modulation of osteoclast activity, DPSC-Exos inhibit osteoclast differentiation and subchondral bone sclerosis both in vivo and in vitro by suppressing TRPV4 activation and blocking calcium influx signals [189]. Similarly, hUCMSC-EVs deliver let-7a-5p to inhibit ITGβ3 on the osteoclast surface, thereby attenuating their bone resorptive function [238]. Addressing the synovium-bone crosstalk mechanism, MSC-sEV engineered to overexpress miR-27b-3p inhibit CSF-1 to suppress the PI3K/AKT signaling pathway; this action simultaneously ameliorates the synovial microenvironment and blocks aberrant osteoclast activation, thereby mitigating excessive bone resorption [237].

In terms of improving bone microstructure, BMMSC-Exos deliver miR-206 to suppress ELF3 expression, thereby increasing OCN and BMP2 levels and promoting osteoblast proliferation, differentiation, and calcium deposition [163]. Furthermore, in vivo studies demonstrate that BMMSC-Exos alleviate pain by suppressing the aberrant CGRP⁺ nerves invasion and H-type vessel formation in subchondral bone, which is accompanied by inhibition of the RANKL-RANK-TRAF6 signaling pathway [239].

Engineering strategies for MSC-Exos

Although MSC-Exos have demonstrated clear therapeutic potential in OA treatment, natural MSC-Exos still face challenges such as poor targeting, rapid in vivo clearance, and a lack of standardized protocols for their preparation, isolation, and storage [261, 262]. To some extent, these limitations hinder their further clinical translation. Consequently, bioengineering of MSC-Exos has become an important direction in current research. Engineered MSC-Exos represent an innovative bioengineering strategy involving the precise reconstitution of MSC-Exos via physical, chemical, or biological techniques to optimize their targeting specificity, payload encapsulation efficiency, physicochemical stability, and therapeutic potency [148, 263]. Currently, engineering strategies for MSC-Exos are predominantly classified into three main categories: parental cell preconditioning strategies, cargo loading strategies and surface modification strategies [264]. Furthermore, MSC-Exos-based hydrogel composite delivery systems have been attracting increasing attention in recent years due to their advantages of sustained release and prolonged local therapy. A comprehensive overview of these engineering strategies for MSC-Exos is illustrated in Fig. 6 and summarized in Table 4.

Fig. 6.

The engineering strategies of MSC-Exos

Table 4.

The summary of engineering strategies and mechanism of MSC-Exos

Engineering strategy	Sources	Engineering methods and cargo	Evidence level	Mechanism	Outcomes	Refs
Parental cell preconditioning
Cytokines preconditioning	BMMSCs	IL-1β preconditioned	In vitro	Upregulated miR-147b inhibit NF-κB pathway	Anti-inflammatory effect	[265]
	IPFSCs	TNF-α preconditioned	In vitro and in vivo	TNF-α preconditioning active PI3K/AKT signaling pathways and promote ATG16L1 expression, promote Exos secretion and upregulated LRP1 in Exos	1.Restore ECM homeostasis; 2.Anti-inflammatory effect	[266]
	BMMSCs	TGF-β1 preconditioned	In vitro and in vivo	Upregulated miR-135b inhibit sp1 and MAPK6 expression	1.Promote chondrocyte proliferation and migration; 2.Polarize macrophage from M1 to M2; 3.Restore ECM homeostasis	[233, 267]
Microenvironment preconditioning	BMMSCs	Hypoxia preconditioned	In vitro and in vivo	-	1.Anti-inflammatory effect; 2.Restore ECM homeostasis; 3.Inhibit senescence; 4.Alleviate pain	[268]
	DPSCs	Hypoxia preconditioned	In vitro	-	1.Anti-inflammatory effect; 2.Restore ECM homeostasis; 3.Promote apoptosis	[269]
	hMSCs	3D Culture	In vitro	Upregulate Wnt, TNF, Hippo and MAPK signalling pathways and downregulate of cell cycle, DNA replication and cellular senescence associated pathways	1.Restore ECM homeostasis; 2.Inhibit apoptosis	[270]
	BMMSCs	dECM preconditioning	In vitro and in vivo	Upregulated miR-3473b target PTEN, activate PTEN/AKT signaling pathway	1.Inhibit apoptosis 2.Promote chondrocyte proliferation and migration	[271]
	BMMSCs	Co-culture with chondrocytes	In vitro	-	Co-culture the adult MSCs with juvenile chondrocytes could promote functional differentiation and improved ECM production	[272]
	BMMSCs	Co-culture with chondrocytes	In vitro and in vivo	-	Promote matrix production and functional differentiation	[273]
Chemical preconditioning	BMMSCs	Curcumin preconditioned	In vitro and in vivo	Upregulate miR-143 and miR-124 expression by reducing the DNA methylation, inhibit NF-κB and ROCK1/TLR9 signaling pathways	Inhibit apoptosis	[274]
	BMMSCs	Quercetin preconditioned	In vitro and in vivo	Modulates PI3K-AKT signaling pathway	1.Anti-inflammatory effect; 2.Restore ECM homeostasis	[275]
	BMMSCs	Fucoidan preconditioned	In vitro and in vivo	Upregulated miR-146b-5p target TRAF6, inhibit PI3K/AKT/mTOR signaling pathway	1.Anti-inflammatory effect; 2.Inhibit M1 polarization; 3.Restore ECM homeostasis; 4.Promote autophagy	[276]
Loading strategies
Endogenous cargo loading methods	BMMSCs	Transfection, miR-210	In vitro	Overexpressed miR-210, inhibit NF-κB pathway	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation; 3.Inhibit apoptosis	[209]
	SMSCs	Transfection, miR-212-5p	In vitro	Overexpressed miR-212-5p inhibit ELF3 expression	1.Restore ECM homeostasis; 2.Anti-inflammatory effect	[211]
	SMSCs	Lentivirus transfection, miR-140-5p	In vitro and in vivo	Wnt5a and Wnt5b in SMSC-Exos active YAP/TAZ signaling pathway; Overexpressed miR-140-5p inhibit RalA	1.Promote chondrocyte proliferation and migration; 2.Restore ECM homeostasis	[188]
	BMMSCs	Transfection, miR-92a-3p	In vitro and in vivo	Overexpressed miR-92a-3p inhibit WNT5A expression	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation and migration	[165]
	BMMSCs	Adenovirus transfection, RUNX2	In vitro and in vivo	-	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation and migration	[277]
	ADSCs	Lipofectamine 2000 transfection, USP15	In vitro and in vivo	Upregulated USP15 promote FOXC1 expression	1.Polarize macrophage from M1 to M2; 2.Restore ECM homeostasis; 3.Anti-inflammatory effect	[278]
Hydrogel strategies
Natural hydrogel	BMMSCs	Photo-crosslinking, GelMA/OCS	In vitro and in vivo	GMOCS-Exos active Nrf2 signaling pathway, inhibit NETs expression	1.Promote chondrocyte proliferation and migration; 2.Restore ECM homeostasis; 3.Suppress oxidative stress	[279]
	BMMSCs	Enzymatic crosslinking, AD/CS/RSF	In vitro and in vivo	-	Promote MSC recruitment and chondrogenic differentiation	[280]
	BMMSCs	Photo-crosslinking, KGMMA+EGCG	In vitro and in vivo	-	1.Suppress oxidative stress; 2.Inhibit apoptosis; 3.Promote chondrocyte proliferation and migration; 4.Anti-inflammatory effect	[281]
	BMMSCs	Photo-crosslinking, GelMA/CSMA/HAMA	In vitro and in vivo	-	1.Promote chondrocyte proliferation and migration; 2.Restore ECM homeostasis; 3.Promote MSC chondrogenic	[282]
	UCMSCs	3D-printing, physical crosslinking and photo-crosslinking, upper, GelMA + BP; lower, GelMA + BP + β-TCP	In vitro and in vivo	-	1.Restore ECM homeostasis; 2.Promote bone and cartilage regeneration	[283]
Hydrogel microspheres	UCMSCs	Microfluidic electrospray and photo-crosslinking, GelMA/HAMA grafted with SKPPGTSS peptide	In vitro and in vivo	-	1.Promote MSC recruitment and chondrogenic differentiation; 2.Restore ECM homeostasis	[284]
	BMMSCs	Freeze-drying microfluidic technology and photo-crosslinking, porous GelMA+PDGF-BB	In vitro and in vivo	-	1.Promote cell adhesion and proliferation; 2.Promote MSC recruitment and chondrogenic differentiation; 3.Restore ECM homeostasis	[285]
	UCMSCs	Microfluidic electrospray and cryogelation, HAMA/CSMA+spiny MNPs + DS	In vitro and in vivo	-	1.Promote chondrocyte proliferation and migration; 2.Restore ECM homeostasis	[286]
Composite hydrogel	BMMSCs	Upper, chemical crosslinking, PVA/TSPBA + DC; Lower, Ionic crosslinking, SA/HA+Exos	In vitro and in vivo	Modulate chemokine signaling pathways to promote the differentiation of BMMSCs into chondrocytes	1.Anti-inflammatory effect; 2.Reduce oxidative stress; 3.Promote MSC recruitment and chondrogenic differentiation	[287]
	MenSCs	Ionic crosslinking, SA+CEFFE-NFs	In vitro and in vivo	-	Promote chondrocyte proliferation and migration	[288]
Combinatorial engineering strategies
Exogenous cargo loading methods + genetic manipulation	UCMSCs	Electroporation, miR-223; lentiviral vector transfection, CTP	In vitro and in vivo	Upregulated miR-223 inhibit NLRP3 expression	1.Inhibit pyroptosis; 2.Restore ECM homeostasis; 3.Anti-inflammatory effect	[289]
Exogenous cargo loading methods + genetic manipulation	ADSCs	Electroporation, miR-199a-3p; plasmid transfection, CAP	In vitro and in vivo	Upregulated miR-199a-3p inhibit mTOR expression	1.Promote autophagy; 2.Restore ECM homeostasis	[225]
Microenvironment preconditioning + exogenous cargo loading methods	ADSCs	3D Culture; Electroporation, miR-27b-3p	In vitro and in vivo	Upregulated miR-27b-3p target M-CSF, inhibit PI3K/AKT signaling pathway	1.Inhibit osteoclastogenesis and bone erosion; 2.Inhibit M1 macrophage polarization; 3.Inhibit oxidative stress	[237]
Endogenous cargo loading methods + composite hydrogel microspheres	ADSCs	Lipofectamine 2000 transfection, miR-99a-3p; microfluidic technology, HB-PEG/HA	In vitro and in vivo	Upregulated miR-99a-3p inhibit ADAMTS4 expression	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation and migration; 3.Inhibit apoptosis	[206]
Exogenous cargo loading methods + composite hydrogel	BMMSCs	freeze-thawing, ICA; Thermosensitive physical gelation and Chemical crosslinking, CS/CAT-C/β-GP/DF-PEG	In vitro and in vivo	-	1.Promote chondrocyte proliferation and migration; 2.Inhibit ECM degradation	[290]
Exogenous cargo loading methods + Chemical modification + hydrogel microspheres	BMMSCs	Sonication, LRRK2-IN-1; Covalent conjugation, CAP; Emulsification and photo-crosslinking, GelMA	In vitro and in vivo	-	1.Restore ECM homeostasis; 2.Anti-inflammatory effect	[291]
Physical preconditioning + natural hydrogel	BMMSCs	Hypoxia preconditioned; Sonication-induced gelation, SF	In vitro and in vivo	Upregulated miR-205-5p, modulates PTEN/AKT signaling pathway	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation and migration; 3.Anti-inflammatory effect; 4.Inhibit apoptosis	[292]
Endogenous cargo loading methods + synthetic hydrogel	BMMSCs	Lipofectamine, modATF5; Thermosensitive, PLGA-PEG-PLGA	In vitro and in vivo	Upregulated ATF5 target UPRmt, modulate ClpP/mTOR/Ulk1 signaling pathway	1.Anti-inflammatory effect; 2.Restore ECM homeostasis; 3.Alleviate mitochondrial dysfunction; 4.Promote autophagy	[293]
Cytokines preconditioning + hydrogel microspheres	UCMSCs	IL-1β preconditioned; Microfluidic technology and photo-crosslinking, HAMA	In vitro and in vivo	-	1.Promote chondrocyte proliferation and migration; 2.Polarize macrophage from M1 to M2 3.Restore ECM homeostasis; 4.Anti-inflammatory and anti-apoptotic effects	[294]
Exogenous cargo loading methods + chemical modification + hydrogel microspheres	UCMSCs	Co-incubation, miR-148a; Sortase A Ligation, CAP; Microfluidic technology and photo-crosslinking, GelMA	In vitro and in vivo	Upregulated miR-148a target Robo2, inhibit MAPK signaling pathway	Restore ECM homeostasis	[295]
Endogenous cargo loading methods + hydrogel microspheres	BMMSCs	Lentiviruses transfection, miR-874-3p; Microfluidic technology and photo-crosslinking, GelMA	In vitro and in vivo	Overexpressed miR-874-3p inhibit NF-κB signaling pathways	1.Restore ECM homeostasis; 2.Suppress oxidative stress	[296]
Chemical modification + hydrogel microspheres	UCMSCs	Lipid anchor, CAP and CWYRGRL; Microfluidic technology and photo-crosslinking, HAMA+dECM	In vitro and in vivo	Modulate MAPK signaling pathway, PI3K/AKT pathway, and pathways related to cell adhesion and communication	1.Restore ECM homeostasis; 2.Promote chondrocyte proliferation 3.Inhibit chondrocyte hypertrophy	[297]
Chemical modification + composite hydrogel	BMMSCs	Lipid anchor, CAP; Physical and ionic crosslinking, PVA + SA	In vitro and in vivo	miR-210-3p, inhibit HIF-3α expression	1.Inhibit senescence; 2.Restore ECM homeostasis	[298]
Chemical modification + microgel	UCMSCs	Hydrophobic bonding, CAP; Emulsification, HA + SH	In vitro and in vivo	Inhibit p53 signaling pathway	1.Inhibit senescence; 2.Restore ECM homeostasis	[245]

BMMSCs, bone marrow-derived mesenchymal stromal cells; IPFSCs, infrapatellar fat pad-derived mesenchymal stromal cells; DPSCs, dental pulp-derived mesenchymal stromal cells; SFMSCs, synovial fluid-derived mesenchymal stromal cells; SMSCs, synovial membrane-derived mesenchymal stromal cells; ADSCs, adipose-derived stromal cells; Exos, exosomes; UCMSCs, umbilical cord-derived mesenchymal stromal cells; MenSCs, menstrual blood-derived mesenchymal stromal cells; IL-1β, interleukin-1 beta; TNF-α, tumor necrosis factor alpha; TGF-β1, transforming growth factor beta 1; 3D, three-dimensional; dECM, decellularized extracellular matrix; RUNX2, Runt-related transcription factor 2; USP15, ubiquitin-specific protease 15; CAP, cartilage affinity peptide; GelMA, gelatin methacryloyl; OCS, oxidized chondroitin sulfate; SA, sodium alginate; CEFFE-NFs, cell-free fat extract electrospun nanofibers; AD/CS/RSF, alginate-dopamine, chondroitin sulfate, and regenerated silk fibroin; EGCG, epigallocatechin gallate; KGMMA, konjac glucomannan methacryloyl; HAMA, hyaluronic acid methacryloyl; BP, black phosphorus; β-TCP, β-tricalcium phosphate; PDGF-BB, platelet-derived growth factor-BB; CSMA, chondroitin sulfate methacrylate; DS, diclofenac sodium; MNPs, magnetic nanoparticles with spiny; TSPBA, N1-(4-boronobenzyl)-N3-(4-boronophenyl)-N1, N1, N3, N3tetramethyl-1,3-propanediaminium; PVA, polyvinyl alcohol; CTP, collagen II-targeting peptides; PEG, polyethylene glycol; CAT-C, catechol-modified chitosan; SF, silk fibroin; PLGA, poly(D, l-lactide-co-glycolide); NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; PI3K, phosphoinositide 3-kinase; AKT, protein kinase B; ATG16L1, autophagy-related 16-like protein 1; LRP1, low density lipoprotein receptor-related protein 1; Sp1, specificity protein 1; MAPK6, mitogen-activated protein kinase 6; PTEN, phosphatase and tensin homolog; ROCK1, Rho-associated coiled-coil containing protein kinase 1; TLR9, toll-like receptor 9; TRAF6, TNF receptor associated factor 6; ELF3, E74 Like ETS transcription factor 3; YAP/TAZ, yes-associated protein/transcriptional coactivator with PDZ-binding motif; FOXC1, forkhead box C1; Nrf2, nuclear factor erythroid 2-related factor 2; NETs, neutrophil extracellular traps; NLRP3, NLR family pyrin domain containing 3; mTOR, mammalian target of rapamycin; ATF5, activating transcription factor 5; ClpP, caseinolytic protease; Ulk1, Unc-51 Like autophagy activating kinase 1; Robo2, roundabout homolog 2; DC, diclofenac sodium; ICA, icariin; CS, chitosan; β-GP, β-glycerophosphate

Parental cell preconditioning strategies

Accumulating evidence suggests that preconditioning MSCs enhances their activity and augments both the yield and bioactivity of MSC-Exos []. Strategies for preconditioning parental cells are primarily categorized into microenvironment, cytokine, and chemical preconditioning [299].

Microenvironment preconditioning

The application of traditional 2D culture is constrained by its inability to maintain the MSC phenotype. Prolonged culture in 2D monolayers often leads to morphological changes and cellular senescence, resulting in diminished exosome production and impaired therapeutic efficacy [300]. 3D culture strategies offer an effectively alternative by providing a microenvironment defined by porosity, surface activity, mechanical strength, and biocompatibility that closely resembles the native physiological state of MSCs [301]. Research by Yuan et al. revealed that 3D aggregates facilitate the activation of ESCRT-dependent and -independent biogenesis pathways. This modulation not only augmented MSC-Exos secretion by approximately twofold compared to 2D controls but also potentiated their anti-inflammatory and immunomodulatory functions [270].

Co-culture systems can partially recapitulate the complex in vivo intercellular communication environment [302]. Studies indicate that co-culturing with chondrocytes enhances the chondrogenic differentiation and ECM-generating capacity of MSCs, a process primarily mediated by EVs [272]. Furthermore, in vivo investigations suggest that EVs derived from co-culture systems, particularly those utilizing a higher proportion of chondrocytes, exhibit improved potential for cartilage repair and functional recovery compared to EVs from monocultures [273].

Moreover, given that MSCs naturally reside in a hypoxic niche, hypoxic culture conditions more accurately recapitulate their in vivo physiological microenvironment, thereby potentiating the reparative capacity of MSC-Exos [292]. Research indicates that Exos derived from hypoxia-preconditioned BMMSCs (Hypo-BMMSC-Exos) effectively attenuate apoptosis and senescence in OA chondrocytes. They also facilitate cartilage regeneration by downregulating the expression of pro-inflammatory cytokines and matrix-degrading enzymes [268]. Mechanistically, hypoxic preconditioning upregulates miR-205-5p levels, which subsequently promotes chondrocyte anabolism and migration via the PTEN/AKT signaling pathway [292]. Additionally, Hypo-DPSC-Exos have been verified to augment glycosaminoglycan content in chondrocytes while suppressing the production of pro-inflammatory cytokines [269].

In addition to the aforementioned strategies, dECM preconditioning has also been demonstrated to enhance the biological functions of MSC-Exos. A recent study indicates that Exos derived from dECM-preconditioned BMMSCs possess enhanced cellular repair and anti-apoptotic capabilities [271].

Cytokine preconditioning

Studies indicate that preconditioning MSCs with inflammatory cytokines or growth factors enhances their anti-inflammatory, immunomodulatory, and pro-angiogenic capabilities, while concurrently augmenting the yield of MSC-Exos [303, 304]. Specifically, TNF-α preconditioning of IPFSCs activates the PI3K/AKT signaling pathway and upregulates ATG16L1 expression, which in turn enriches exosomal LRP1 levels and significantly enhances their chondroprotective efficacy [266]. In addition to TNF-α, IL-1β is frequently employed as a priming agent; research suggests that Exos derived from IL-1β-preconditioned hBMMSCs (IL-hBMMSC-Exos) exhibit upregulated levels of miR-147b, which mitigates inflammatory responses by inhibiting NF-κB pathway activation [265]. Furthermore, IL-hUCMSC-Exos have been shown to facilitate macrophage polarization toward the M2 phenotype and enhance ECM synthesis [294]. Similarly, preconditioning with the growth factor TGF-β1 results in the upregulation of miR-135b in BMMSC-Exos, which promotes chondrocyte proliferation and macrophage M2 polarization by targeting and inhibiting the expression of Sp1 and MAPK6 [233, 267].

Chemical preconditioning

In addition to microenvironment and cytokine preconditioning, priming MSCs with bioactive small molecules or natural compounds represents an effective strategy to optimize exosomal cargo and functionality [299]. Research indicates that Exos derived from curcumin-preconditioned BMMSCs exhibit upregulated levels of miR-143 and miR-124, which suppress chondrocyte apoptosis by inhibiting the NF-κB and ROCK1/TLR9 signaling pathways [274]. Furthermore, MSC-Exos preconditioned with quercetin or fucoidan have also been demonstrated to exert enhanced chondroprotective effects [275, 276].

Loading strategies

Exosomal cargo loading strategies can be classified into two primary categories: endogenous cargo loading methods and exogenous cargo loading methods [299].

Endogenous cargo loading methods involve the genetic modification of MSCs through molecular techniques such as viral vector transduction or plasmid transfection to upregulate the expression of intrinsic bioactive molecules. These molecules are subsequently incorporated into MSC-Exos via the cell’s endogenous biosynthetic pathways, thereby potentiating their therapeutic potential [264]. For instance, the transfection of miR-210 [209] or miR-874-3p [296] into BMMSCs, as well as miR-212-5p [211] or miR-140-5p [188] into SMSCs, has been suggested to effectively augment their chondroprotective effects. In addition to miRNAs, functional proteins can also be loaded. A recent study has shown that enhancing USP15 levels in ADSC-Exos promotes the deubiquitination and stabilization of FOXC1, thereby modulating macrophage M2 polarization and maintaining chondrocyte ECM homeostasis in both in vitro and in vivo models [278].

Common exogenous cargo loading methods encompass sonication, electroporation, freeze-thawing, co-incubation, and extrusion [305]. These strategies are distinguished by their operational simplicity, improved stability, and scalability for manufacturing processes [306]. For instance, the introduction of miR-223 into UCMSC-Exos via electroporation has been shown to inhibit chondrocyte pyroptosis by targeting NLRP3 in both in vivo and in vitro models [289]. Similarly, loading miR-199a-3p into ADSC-Exos suppresses mTOR expression, thereby promoting chondrocyte autophagy [225]. Furthermore, the incorporation of icariin into BMMSC-Exos via freeze-thawing cycles [290], and LRRK2-IN-1 via sonication [291], has been demonstrated to effectively protect chondrocytes against inflammatory microenvironments and preserve ECM homeostasis. It is worth noting that these methods may compromise the integrity of the exosomal membrane, leading to cargo leakage [307].

Surface modification strategies

Surface engineering of MSC-Exos is generally classified into three modalities: chemical modification, genetic manipulation, and membrane fusion [308]. The principal aim is to augment the homing and targeting specificity of MSC-Exos, thereby enabling accurate delivery of therapeutic payloads to pathological microenvironments and enhancing therapeutic outcomes [299].

Chemical modification is broadly categorized into covalent and non-covalent conjugation [264]. Among covalent approaches, click chemistry is frequently utilized; for instance, the conjugation of cartilage affinity peptide (CAP) onto the surface of BMMSC-Exos has been shown to enhance cartilage targeting efficacy [291]. Alternatively, Sortase A-mediated ligation offers high site-specificity for attaching CAP to the surface of UCMSC-Exos [295]. Beyond covalent methods, non-covalent strategies are also extensively employed. For example, conjugation of CAP onto the surface of UCMSC-Exos via hydrophobic insertion, facilitates precise recognition and targeted delivery to senescent chondrocytes, thereby markedly delaying OA progression and promoting cartilage regeneration [245]. Furthermore, a dual-targeting strategy employing lipid anchors to co-display both CAP and the collagen II-binding peptide (CWYRGRL) on UCMSC-Exos has been demonstrated to improve cartilage-targeting capabilities [297].

Genetic manipulation of MSC-Exos augments their functional efficacy by engineering them into hybrid vesicles enriched with CAP and a CRISPR/Cas9 payload for FGF18 activation, thereby enabling precise gene editing and enhancing chondrogenic regeneration [309]. Additionally, plasmid transfection strategies can also be employed to display CAP on the Exo surface [225].

Exos are frequently hybridized with liposomes, which are vesicular nanostructures composed of phospholipid bilayers with diameters ranging from 50 to 5,000 nanometers [310]. As nanocarriers, liposomes effectively encapsulate a broad spectrum of bioactive agents, including hydrophobic pharmaceuticals, hydrophilic drug molecules, and nucleic acids [311]. For instance, Liang et al. engineered hybrid nanovesicles by fusing CAP-Exos with liposomes encapsulating CRISPR/Cas9 plasmids targeting MMP13. These hybrids not only retained cartilage-targeting capabilities but also enhanced plasmid loading efficiency. In vivo studies demonstrated that this system specifically ablated the MMP13 gene within chondrocytes, thereby preserving ECM homeostasis [312].

MSC-Exos combined with hydrogel

Hydrogels, as hydrophilic biomaterials characterized by a 3D polymeric network, have emerged as promising carriers for intra-articular drug delivery owing to their exceptional biocompatibility and injectability [313]. Natural hydrogels are typically composed of polysaccharides and proteins, such as HA, alginate (ALG), chitosan (CS), gelatin (Gel), and silk fibroin (SF) [314]. Their enhanced mechanical properties and capacity to emulate the native ECM of osteochondral tissue create a favorable microenvironment for endogenous cell proliferation, underscoring the potential of MSC-Exos encapsulation for osteochondral regenerative applications [315].

To accurately recapitulate the native ECM microenvironment, Pei et al. developed a composite biomimetic hydrogel scaffold comprising gelatin methacryloyl (GelMA), CSMA, and HAMA; this system not only facilitates the delivery of BMMSC-Exos but also effectively recruits endogenous MSCs [282]. Similarly, the GMOCS hydrogel synthesized by Zhou et al. via photocrosslinking of GelMA and oxidized chondroitin sulfate (OCS) was shown to effectively simulate the in vivo ECM microenvironment [279]. Furthermore, addressing the challenges posed by the lubricious nature of articular cartilage and the difficulty of in situ fixation for natural hydrogels, Zhang et al. engineered a highly adhesive injectable hydrogel utilizing ALG-dopamine, CS, and regenerated SF, which exhibited superior binding strength in wet physiological environments [280]. Regarding structural architecture, Sun et al. employed 3D printing technology to fabricate a GelMA-based bilayer scaffold, in which the upper layer encapsulates Exos and black phosphorus for cartilage repair, while the lower layer incorporates β-tricalcium phosphate to promote subchondral bone regeneration, thus achieving integrated osteochondral repair [283].

Distinct from bulk hydrogels, hydrogel microspheres fabricated via microfluidics or electrospray techniques have garnered increasing attention due to their advantages in relatively minimally invasive delivery and precise targeting capabilities. For instance, hydrogel microspheres composed of GelMA, HAMA, and the SKPPGTSS peptide, engineered via microfluidic electrospray and photocrosslinking, were shown to not only deliver Exos but also facilitate endogenous stem cell recruitment [284]. Similarly, Yang et al. developed magnetic polysaccharide microcarriers (HAMA/CSMA) utilizing microfluidic electrospray technology; by co-encapsulating UCMSC-Exos and diclofenac sodium (DC), they demonstrated that an external magnetic field could precisely localize the microcarriers at the lesion site, thereby effectively prolonging their intra-articular retention time [286].

Beyond natural hydrogels, synthetic polymers, such as polyethylene glycol (PEG), polyvinyl alcohol (PVA), and polyacrylamide, are also extensively utilized for hydrogel fabrication, offering distinct advantages including extended shelf stability and reduced immunogenicity [314]. For instance, Ma et al. employed PLGA-PEG-PLGA as a thermosensitive hydrogel carrier; this system undergoes rapid in situ gelation upon intra-articular injection, thereby effectively mitigating the rapid clearance of Exos [293].

Composite hydrogel delivery systems, fabricated by integrating natural hydrogel components with synthetic polymers, effectively circumvent the intrinsic limitations of natural hydrogels, such as poor stability and suboptimal mechanical properties [314]. For instance, researchers engineered a smart bilayer composite hydrogel scaffold featuring an upper layer composed of TSPBA and PVA. Possessing a specific ROS-responsive degradation capability, this PVA hydrogel undergoes triggered degradation under conditions of high oxidative stress to release the anti-inflammatory agent DC. This mechanism not only scavenges deleterious ROS but also ameliorates the inflammatory microenvironment via macrophage reprogramming, thereby creating a conducive niche for the chondrogenic reparative effects of BMMSC-Exos encapsulated within the underlying HA hydrogel [287].

Challenges and future perspectives

The escalating global prevalence of OA imposes a substantial burden on healthcare systems and society. Despite notable advancements in OA research and disease management, numerous challenges persist. A major obstacle is the lack of consensus regarding OA phenotypes and molecular endotypes [316]. Given the heterogeneous therapeutic responses observed across different endotypes, establishing a precise stratification strategy based on molecular profiles is essential for optimizing clinical research. Looking forward, the integrated application of multi-omics technologies will be instrumental in elucidating the complex endotypes of OA [317].

Furthermore, accurate disease staging, particularly the identification of early OA, is critically important. Traditional therapeutic interventions are predominantly initiated upon the onset of clinical symptoms, by which stage structural joint damage is often irreversible [316]. In recent years, however, the discovery of early diagnostic biomarkers via big data, machine learning, and artificial intelligence (AI) has made the detection of early-stage OA feasible, thereby laying the foundation for developing effective preventive interventions and therapeutic strategies [318].

MSC-based therapeutic strategies have emerged as a promising avenue for OA management, owing to their capacity to effectively intervene in disease progression. Notably, clinical outcomes are often better in patients with early-to-mid-stage OA compared to those with advanced disease. However, several critical challenges must be addressed to further optimize the safety and efficacy of MSC therapies. A primary obstacle remains the post-injection survival and functional persistence of MSCs. Although a preponderance of prior studies supported the therapeutic benefits of MSC injections, recent findings from the MILES trial and the ADIPOA-2 Phase IIb randomized controlled trial have yielded contradictory results. These studies reported that autologous BMAC, ADSCs and allogeneic UCMSCs showed no statistically significant therapeutic benefit over placebo at one-year follow-up [71, 82]. Beyond the placebo effect, a more fundamental limitation likely lies in the heterogeneity of OA endotypes. Therefore, future research should further promote the precise matching of MSC intervention strategies with specific OA pathological endotypes to achieve individualized treatment. In addition to efficacy concerns, while serious adverse events following MSC administration are rare, transient local complications persist, including joint swelling, pain, and restricted range of motion [90]. Furthermore, despite their chondroprotective effects, MSCs may also promote osteophyte formation [319], and current clinical research remains predominantly concentrated in Phase I and Phase II trials, with a scarcity of Phase III studies.

The selection of cell source also influences therapeutic outcomes. Although BMMSCs are the most extensively characterized, their yield and proliferation potential tend to decline with increasing donor age; notably, BMMSCs from elderly donors may exhibit reduced proliferative capacity and an increased propensity for adipogenic differentiation [320]. SMSCs, owing to their unique anatomical niche, have been reported to possess superior chondrogenic capacity and differentiation potential [117]. However, current evidence is predominantly preclinical, with limited validation in clinical settings [321, 322]. Similarly, while ESMSCs and iMSCs have demonstrated promising chondrogenic potential in preclinical models, universally accepted clinical protocols for their application remain unestablished. Future research must prioritize resolving the challenge of cellular heterogeneity, focusing on optimizing dosing regimens, establishing optimal administration timing, and exploring combinatorial therapeutic strategies to advance the standardization and precision of MSC-based therapies.

While the therapeutic efficacy of MSCs was traditionally attributed to their chondrogenic differentiation potential, accumulating evidence indicates that their reparative effects are mainly mediated through paracrine signaling and immunomodulatory pathways. As pivotal paracrine effectors, MSC-Exos facilitate intercellular communication by delivering bioactive cargo. However, similar to their parent cells, free MSC-Exos exhibit limited targeted homing capability and a short in vivo half-life following systemic administration. Therefore, future research should employ engineering strategies, such as surface modification and biomaterial-based delivery, to further enhance the targeted delivery efficiency and local therapeutic persistence of MSC-Exos, thereby facilitating their clinical translation [264].

Regarding efficacy enhancement, while preconditioning and endogenous cargo loading strategies are widely utilized, their application is constrained by MSC heterogeneity; consequently, the optimal duration and intensity of distinct preconditioning strategies require further standardization. Furthermore, the complex processing required for engineering MSC-Exos, involving transfection, chemical modification, or physical intervention, impedes clinical-grade scalability. In recent years, the adoption of dynamic 3D culture systems (e.g., bioreactors) has effectively augmented both the yield and potency of MSC-Exos, offering a viable pathway toward large-scale, standardized manufacturing [323]. Regarding delivery systems, diverse materials such as hydrogels, liposomes, inorganic nanoparticles, and microneedles have been employed for the sustained-release delivery of MSC-Exos [262, 312, 324]. However, future research must comprehensively balance several critical issues, including the protective effects of delivery systems on exosome bioactivity, their impact on cellular uptake and in vivo biodistribution, as well as the optimization of release kinetics. To achieve this, it is essential to integrate MSC-Exo source and dosage with the specific degradation rates of scaffolds and administration frequency to establish optimized therapeutic strategies.

Notably, OA is increasingly recognized as a complex whole-joint disease, driven by the multifaceted interplay among distinct cellular compartments, soluble/biochemical factors, the ECM, and biophysical factors (such as mechanical loading, fluid shear stress, hydrostatic pressure, and hypoxia). Traditional animal models often fail to recapitulate the complex mechanical loading and unique microenvironment of the human joint. Consequently, in response to the global trend to minimize animal suffering through the “3Rs” (Replacement, Reduction, and Refinement) alongside the enactment of the FDA Modernization Act 2.0, regulatory bodies have explicitly signaled a transition toward phasing out mandatory animal testing [325]. Instead, priority is now given to “New Approach Methodologies”, including organoids, organs-on-a-chip, and computational modeling (e.g., AI-driven predictive models) [325]. This regulatory shift has catalyzed the rapid evolution of in vitro models. While organoid technology successfully replicates native tissue microarchitecture via cellular self-assembly, thereby facilitating the elucidation of intercellular interactions and disease pathogenesis [318], its standardization is limited by inherent batch-to-batch functional variability and genetic instability during long-term culture [326]. Coinciding with rapid advancements in microfluidics, Joint-on-a-Chip (JOC) platforms and micro-physiological systems have garnered extensive research attention. JOC platforms not only enable dynamic crosstalk between multiple joint tissues but also allow for the real-time monitoring and precise regulation of microenvironmental constituents, including physicochemical properties (e.g., oxygen, pH, temperature, CO₂, and osmolarity), physical cues (e.g., mechanical cell-cell and cell-ECM interactions), and biochemical cues (e.g., soluble metabolites, cytokines, and growth factors) [318, 327]. Furthermore, in a multicenter Phase III clinical trial, researchers demonstrated that a microfluidic on-chip 3D system maintained higher levels of bioactive secretion profiles compared to 2D cultures and provided improved prediction of patient WOMAC scores [328]. Collectively, the emergence of next-generation in vitro models effectively reduces the reliance on animal experimentation. This progress represents not merely a scientific inevitability but a pivotal response to evolving regulatory mandates and ethical responsibilities.

Finally, it is important to acknowledge the inherent methodological limitations of this review. First, unlike a strict systematic review, the literature selection process inevitably carries a degree of subjectivity, which may introduce selection bias. Second, our literature search primarily relied on the PubMed database, potentially leading to the omission of relevant studies indexed in other major databases. Third, the inclusion criteria were limited to English-language publications between 2016 and 2026, which may have inadvertently excluded valuable preclinical and clinical research published in other languages or in regional journals.

Conclusions

Current clinical interventions for OA primarily focus on symptomatic relief but fail to halt or reverse the progressive degeneration of the joint. MSC-based therapies exhibit broad therapeutic prospects in tissue repair, immunomodulation, and the restoration of joint homeostasis, offering the potential to enhance existing OA treatment strategies. This review summarizes the applications of MSCs derived from diverse tissue sources and MSC-Exos in OA treatment, elucidating their potential regulatory mechanisms within the articular cartilage, synovium, and subchondral bone microenvironments. Additionally, it provides a comprehensive overview of emerging bioengineering and nanotechnology-enhanced strategies for MSC-Exos. This review aims to provide a valuable reference for future OA therapeutic strategies and translational research.

Abbreviations

OA

Osteoarthritis

MSCs

Mesenchymal stromal cells

TNF-α

Tumor necrosis factor-alpha

VEGF-C

Vascular endothelial growth factor-C

NF-κB

Nuclear factor kappa-light-chain-enhancer of activated B cells

MMPs

Matrix metalloproteinases

ADAMTS

A disintegrin and metalloproteinase with thrombospondin motifs

COL10

Type X collagen

MMP13

Matrix metalloproteinase-13

IL-1α

Interleukin-1 alpha

CCL2

C-C motif chemokine ligand 2

ACAN

Aggrecan

HA

Hyaluronic acid

MSC-Exos

mesenchymal stromal cell-derived exosomes

ECM

Extracellular matrix

RUNX2

Runt-related transcription factor 2

SOX9

SRY-related high-mobility group-box gene 9

SA-β-gal

Senescence-associated β-galactosidase

SASP

Senescence-associated secretory phenotype

GM-CSF

Granulocyte-macrophage colony-stimulating factor

MCP1

Monocyte chemoattractant protein-1

TGF-β

Transforming growth factor-beta

ESCs

Embryonic stem cells

iPSCs

Induced pluripotent stem cells

BMMSCs

Bone marrow-derived mesenchymal stromal cells

BMAC

Bone marrow aspirate concentrate

SMSCs

Synovium-derived mesenchymal stromal cells

ESMSCs

Embryonic stem cell-derived mesenchymal stromal cells

iMSCs

Induced pluripotent stem cell-derived mesenchymal stromal cells

iNOS

Inducible nitric oxide synthase

ADSCs

Adipose-derived stromal cells

UCMSCs

Umbilical cord-derived mesenchymal stromal cells

ROS

Reactive oxygen species

Exos

Exosomes

EVs

Extracellular vesicles

miRNAs

microRNAs

VEGF

Vascular endothelial growth factor

CAP

Cartilage-targeting peptides

DAMPs

Damage-associated molecular patterns

ASCs

Adult stem cells

VAS

Visual analogue scale

WOMAC

Western Ontario and McMaster Universities Osteoarthritis Index

KL

Kellgren-Lawrence

PRP

Platelet-rich plasma

SVF

Stromal vascular fraction

KOOS

Knee Injury and Osteoarthritis Outcome Score

hESC

Human ESC

ICRS

International Cartilage Regeneration & Joint Preservation Society

WJMSCs

Wharton’s Jelly-derived mesenchymal stromal cells

ROM

Range of motion

MRI

Magnetic resonance imaging

IPFSCs

Infrapatellar fat pad-derived mesenchymal stromal cells

SFMSCs

Synovial fluid-derived mesenchymal stromal cells

PSCs

Pluripotent stem cells

UCMSC-sEV

UCMSC-derived small extracellular vesicles

IPFSCs-CM

Conditioned medium from IPFSCs

mtDNA

Mitochondrial DNA

circRNAs

Circular RNAs

ceRNA

Competing endogenous RNA

3D

Three-dimensional

Hypo-BMMSC-Exos

Exos derived from hypoxia-preconditioned BMMSCs

IL-hUCMSC-Exos

Exos derived from IL-1β-preconditioned hUCMSCs

ALG

Alginate

CS

Chitosan

Gel

Gelatin

SF

Silk fibroin

GelMA

Gelatin methacryloyl

OCS

Oxidized chondroitin sulfate

DC

Diclofenac sodium

PEG

Polyethylene glycol

PVA

Polyvinyl alcohol

JOC

JOINT-on-a-Chip

WORMS

Whole-Organ magnetic resonance imaging score

Author contributions

All the authors listed have approved the manuscript that is enclosed. YS, ZN and FW drafted the manuscript. ZL, SL, KL and XC designed the tables and figures. CX and FW supervised and revised the work.

Funding

Our work was supported by the National Natural Science Foundation of China (81873983), Natural Science Foundation of Hebei Province (H2022206534, H2021206162) and Graduate Innovation Project of Hebei Province (CXZZBS2024128).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.