Athlete-derived extracellular vesicles protect against spinal cord injury via inhibition of neuronal ferroptosis

Abstract

Spinal cord injury (SCI) causes high morbidity, disability, and mortality, while current surgical and pharmacological treatments provide limited benefit. Ferroptosis, a newly recognized form of regulated cell death, contributes critically to SCI pathology, and targeting this process may enhance neuronal survival. Extracellular vesicles, key mediators of intercellular communication, are emerging as promising therapeutic agents for central nervous system injury. Here, we examined the role of athlete-derived plasma extracellular vesicles (AEVs) in neuronal ferroptosis and motor function recovery after SCI. In a murine model, AEVs markedly inhibited ferroptosis and improved motor outcomes. Mechanistically, AEVs delivered RNF216, which promoted ubiquitination and degradation of NOX1, thereby reducing ferroptotic damage and facilitating recovery. Moreover, RNF216-enriched vesicles enhanced synaptic plasticity, supporting neuronal regeneration and network reestablishment. These findings reveal a previously unrecognized RNF216-NOX1 axis in SCI and highlight AEVs as a previously unidentified therapeutic strategy.

Athlete-derived extracellular vesicles can reduce ferroptosis and promote motor recovery after spinal cord injury.

INTRODUCTION

Spinal cord injury (SCI) is associated with poor clinical outcomes and frequently results in either transient or irreversible functional deficits. Owing to its high incidence, severe disability, elevated mortality rates, and substantial health care costs, SCI imposes a substantial socioeconomic burden (1). Etiologically, SCI is classified as traumatic or nontraumatic, and pathophysiologically, it comprises primary and secondary phases. The primary phase involves immediate mechanical disruption of neural structures, while the secondary phase includes a cascade of deleterious processes that enhance tissue damage and neuronal loss (2). As a result, neuroprotective strategies are of paramount importance. Despite the availability of surgical, pharmacologic, and other therapeutic interventions, their clinical efficacy in enhancing neurological recovery remains limited, with few approaches yielding meaningful outcomes in patients with SCI (3). This highlights an urgent need for the development of safe and efficacious neuroprotective therapies.

Ferroptosis, distinct from apoptosis, necrosis, and autophagy, has been identified as a novel form of programmed cell death. It is driven by oxidative damage resulting from intracellular iron-dependent reactive oxygen species (ROS) accumulation and lipid peroxidation (4, 5). This cell death pathway is implicated in various pathological conditions and contributes significantly to central nervous system (CNS) injuries, including traumatic SCI (6, 7). Prior findings demonstrated that Ubiquitin specific protease 11 (USP11) modulated autophagy-dependent ferroptosis in neuronal cells following spinal cord ischemia-reperfusion injury through the deubiquitination of Beclin 1 (8). Silencing USP11 substantially diminished ferroptotic neuronal death and enhanced motor function recovery in murine models. Notably, neurons, in contrast to astrocytes, exhibit limited lipid metabolic capacity. Their phospholipid membranes are enriched in fatty acid residues, have insufficient ROS scavenging ability, and are thus particularly vulnerable to oxidative stress (9). Current research on ferroptosis predominantly emphasizes systemic regulation of iron metabolism and redox balance, offering limited translational prospects. Targeting neuronal ferroptosis specifically may constitute a viable therapeutic approach to enhancing neuronal viability post-SCI.

Extracellular vesicles (EVs), membrane-bound vesiculo-vacuolar structures measuring approximately 50 to 150 nm in diameter, represent essential mediators of paracrine signaling. By transporting diverse biomolecules such as proteins, cytokines, mRNAs, and microRNAs, EVs serve as key vehicles for intercellular communication (10). Owing to these properties, EVs have been explored as therapeutic candidates for CNS injuries. Previous research confirmed that EVs could traverse the blood-brain barrier and contribute to functional improvement following SCI (11, 12). Moreover, EV-based interventions demonstrate therapeutic efficacy and functional parallels to direct stem cell transplantation, without the associated risks (13). Their therapeutic potential is further reinforced by low cytotoxicity, high biocompatibility, and immunological inertness (14, 15). Accumulated evidence from earlier investigations has established that stem cell–derived EVs enhance neural regeneration in SCI models. Nevertheless, therapeutic optimization may require alignment of EV profiles with specific stem cell sources and lineages. A persistent limitation involves the scalable production of stem cell–derived EVs with consistent quality and purity, which remains a major technical bottleneck.

EVs are widely distributed in the circulatory system, with plasma (or serum) concentrations far exceeding those produced by any single cell type (16). Emerging evidence has identified multiple therapeutic properties of blood-derived EVs, particularly emphasizing the benefits of plasma EVs from young, healthy donors. Wang et al. (17) reported that serum-derived EVs from young individuals mitigated inflammation by partially restoring immune tolerance in senescent T cells. Similarly, Chen et al. (18) demonstrated that small EVs from young plasma reversed age-associated degeneration and dysfunction through Peroxisome proliferator-activated receptor-gamma coactivator 1-alpha (PGC-1α) activation and enhancement of mitochondrial energy metabolism. Additional studies have shown that EVs from young healthy human plasma promoted functional recovery after cerebral hemorrhage by modulating the P53/SLC7A11/GPX4 (glutathione peroxidase 4) signaling axis to suppress ferroptosis-induced damage (19). Long-term clinical observations indicate that athletes tend to exhibit superior motor recovery following SCI compared to the general population. Nevertheless, whether plasma-derived EVs from athletes (AEVs) provide enhanced therapeutic effects over those from nonathletic general individuals (GEVs) in SCI treatment remains to be determined.

Ubiquitination, a posttranslational protein modification involving the covalent attachment of one or more ubiquitin molecules, modulates diverse cellular processes by regulating protein stability and signal transduction pathways (20). Ring finger protein 216 (RNF216), also termed Triad3, functions as a canonical E3 ubiquitin ligase. Xu et al. (21) reported that RNF216 impeded autophagic activity in macrophages by promoting the ubiquitination and degradation of Beclin1. Husain et al. (22) found that missense mutations in RNF216 disrupted the clearance of cytoskeletal-associated proteins under neuronal activity–dependent regulation, ultimately impairing synaptic integrity and cognitive performance, thus linking RNF216 dysfunction to neurological pathologies. Although RNF216 has been extensively investigated across multiple pathological contexts, its functional relevance in traumatic CNS injury—particularly in traumatic SCI—remains undefined.

This study delineates the molecular mechanism through which RNF216 attenuates ferroptosis and contributes to motor function restoration via plasma EVs derived from healthy athletes. Elevated RNF216 expression in these EVs emerges as a key determinant in suppressing neuronal ferroptosis through intercellular transfer mechanisms. Mechanistically, RNF216 targets nicotinamide adenine dinucleotide phosphate (NADPH) oxidase 1 (NOX1) for ubiquitination-mediated degradation, thereby mitigating oxidative stress and subsequent ferroptotic damage in neuronal cells. The present findings suggest that EVs from healthy athletes modulate post-SCI ferroptosis dynamics through RNF216-dependent pathways, offering a potential therapeutic avenue for SCI intervention.

RESULTS

Identification and uptake of AEVs

Plasma samples collected from human volunteers were cleared of cellular debris before EV isolation by ultracentrifugation and density-gradient centrifugation. The isolated EVs were characterized in accordance with MISEV guidelines. Transmission electron microscopy (TEM) imaging identified spherical vesicles measuring 50 to 150 nm in diameter (Fig. 1A), consistent with the expected morphological profile. Nanoparticle tracking analysis (NTA) confirmed comparable size distributions across both experimental groups (Fig. 1B). No statistically significant differences were noted in particle size, concentration, or zeta potential between GEVs and AEVs (Fig. 1, C to F). Western blot analysis indicated that CD9, CD63, CD81, ALIX, and TSG101 were consistently enriched in both GEV and AEV populations (Fig. 1G). Furthermore, particle-to-protein ratios remained similar between the two EV subtypes (fig. S1). AEVs labeled with PKH26 dye were cocultured with neuronal cells, and neuronal identity was validated by immunofluorescence (Fig. 1H). Confocal microscopy demonstrated cytoplasmic localization of AEVs within neuronal cells, with signal intensity increasing over time, indicative of progressive uptake (Fig. 1, I and J).

Fig. 1. Identification and uptake of AEVs.

(A) TEM images showing the morphology of GEVs and AEVs. (B) NTA of particle size distribution for GEVs and AEVs. (C) Comparison of mean particle diameters between GEVs and AEVs (n = 6 per group). (D) Particle concentration (particles per ml) in GEVs and AEVs (n = 6 per group). (E and F) Surface zeta potential measurements for GEVs and AEVs (n = 6 per group). (G) Detection of EV-associated protein markers via Western blotting. (H) Identification of neuronal cells using MAP2/NeuN immunofluorescence staining. (I and J) Quantitative analysis of PKH26-labeled AEV internalization by neuronal cells. Note that n denotes the number of independent biological replicates. ns, not significant.

To assess the in vivo biodistribution and persistence of DIO-labeled AEVs, fluorescence signals were monitored both in vivo and ex vivo after tail vein injection. In vivo imaging revealed clear fluorescence accumulation at the SCI site as early as 6 hours postinjection, with signal intensity persisting at 12 hours (fig. S2A). Ex vivo imaging of dissected spinal cords confirmed these findings, showing robust red fluorescence at 6 and 12 hours, which was still detectable at 24 hours, albeit slightly reduced (fig. S2B). These observations confirm the targeted accumulation and temporal retention of AEVs at the lesion site.

AEVs promote motor function recovery after SCI

The therapeutic potential of plasma-derived EVs in promoting functional recovery postinjury was evaluated, with a specific focus on the comparative efficacy of AEVs. Behavioral assessments were conducted at predefined intervals (Fig. 2A). Over the 4-week recovery period following SCI, mice receiving AEVs exhibited significantly enhanced motor function—including hind paw placement, ambulation, and hindlimb coordination—as reflected by higher Basso Mouse Scale (BMS) scores compared to those treated with GEVs (Fig. 2B and fig. S3A). Superior recovery in hindlimb function and tail balance in the AEV group was also evident in the rotarod test (Fig. 2C and fig. S3B) and corroborated by the von Frey filament test results (Fig. 2D and fig. S3C). Footprint analysis further revealed accelerated gait restoration and improved motor coordination in AEV-treated mice relative to GEV counterparts (Fig. 2, E and F, and fig. S3D). Tissue damage in the thoracic spinal cord was subsequently assessed in the murine SCI model. Macroscopic examination clearly delineated the injury region (Fig. 2G and fig. S3E). Notably, treatment with AEVs resulted in a marked reduction in the lesion area compared to both the phosphate-buffered saline (PBS) and GEV groups (Fig. 2H and fig. S3F). To corroborate motor function recovery, pathological alterations in the injured spinal cord across treatment groups were assessed via magnetic resonance imaging (MRI). Leveraging its noninvasive capability and diagnostic precision in SCI evaluation, MRI was conducted 28 days postinjury to monitor tissue repair. As shown in Fig. 2I, low signal intensity at the lesion site corresponded to fluid-filled cyst formation. Notably, AEV administration markedly improved this pathological phenotype. Compared to PBS and GEV groups, AEV-treated spinal cords displayed enhanced structural preservation, reduced cystic cavitation, and superior reparative outcomes (Fig. 2, I and J), in agreement with the results presented in Fig. 2H. Behavioral data further confirmed that AEV treatment promoted functional recovery in SCI mice.

Fig. 2. AEVs promote motor function recovery in male mice after SCI.

(A) Schematic representation of the animal experimental design. (B) BMS score assessment over 28 days post-SCI (n = 6 per group). (C) Quantitative analysis of motor coordination using the rotarod test (n = 6 per group). (D) Mechanical allodynia evaluation via von Frey filament testing (n = 6 per group). (E and F) Gait analysis using footprint tracking (n = 6 per group). (G and H) Quantification of spinal cord lesion size and gross morphology (n = 6 per group). (I and J) Representative sagittal and axial MRI scans of spinal cord tissue (n = 6 per group). (K and L) NeuN immunofluorescence staining and corresponding quantification (n = 3 per group). (M and N) NF200 immunofluorescence staining and quantitative analysis (n = 3 per group). (O and P) Western blot detection and densitometric analysis of NF200, MAP2, and GFAP (glial fibrillary acidic protein) proteins in spinal cord tissue at day 28 postinjury (n = 3 per group). Note that n denotes the number of independent biological replicates. d, days; GAPDH, glyceraldehyde phosphate dehydrogenase.

Hindlimb motor recovery following SCI is tightly linked to the preservation of spinal cord motor neurons. To quantify neuronal survival, sagittal sections of the spinal cord were analyzed at defined distances from the lesion epicenter. Neuronal viability in discrete zones (Z1 to Z4), demarcated relative to the lesion margins, was assessed via Neuronal Nuclei (NeuN) immunolabeling. A markedly higher density of NeuN⁺ neurons was observed in the Z1 to Z3 regions in the AEV group compared to the GEV group (Fig. 2, K and L).

To further elucidate the anatomical correlates of functional restoration, the axonal structure within the lesion core was examined. NF200⁺ axons served as indicators of axonal regeneration. At 28 days postinjury, a substantially greater number of neurofilament-positive (NF200⁺) axons was present within the lesion sites of AEV-treated mice relative to GEV-treated counterparts (Fig. 2, M and N). This observation was consistent with Western blot results for Neurofilament 200 (NF200) and Microtubule-Associated Protein 2 (MAP2), which mirrored the immunofluorescence data (Fig. 2, O and P). Collectively, the data demonstrate superior neuroprotective efficacy and enhanced motor recovery following AEV administration compared to GEV treatment.

AEV treatment attenuates oxidative stress–induced ferroptosis after SCI

Previous studies have shown that plasma-derived EVs released following endurance exercise exhibit antioxidant capacity (23). This study evaluated the potential of AEVs to counteract oxidative stress–induced ferroptosis in injured spinal cord tissue. Western blot analysis revealed up-regulation of ferroptosis-associated markers ACSL4, 4HNE, and PTGS2, along with a concomitant reduction in GPX4 expression, when comparing post-SCI tissue to normal spinal cord. Administration of GEVs and AEVs markedly attenuated the expression of ACSL4, 4HNE, and PTGS2 while restoring GPX4 levels relative to the PBS control group (Fig. 3, A and B, and fig. S3, M and N). Notably, AEV treatment demonstrated superior efficacy over GEVs alone (Fig. 3, A and B, and fig. S3, M and N). Quantitative reverse transcription polymerase chain reaction (RT-qPCR) analysis corroborated the preceding results (Fig. 3, C to E). Ferroptosis is characterized by Fe²⁺ accumulation, glutathione (GSH) depletion, and lipid peroxidation; accordingly, the relative levels of Fe²⁺, malondialdehyde (MDA), and GSH were quantitatively assessed. Compared to the PBS group, both GEVs and AEVs significantly increased GSH levels while reducing MDA content and Fe²⁺ accumulation (Fig. 3, F to H, and fig. S3, J to L). Notably, the AEV group exhibited higher GSH levels and lower MDA and Fe²⁺ concentrations than the GEV group, indicating enhanced regulatory effects. To further assess oxidative stress within the injured spinal cord, dihydroethidium (DHE) staining was performed. SCI led to a marked increase in DHE fluorescence intensity, indicating elevated ROS levels. GEV treatment mitigated ROS accumulation in the lesion area, with AEVs exerting a more pronounced suppressive effect (Fig. 3, I to K, and fig. S3, H and I). Neuronal apoptosis was evaluated via terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining on day 7 post-SCI. A substantial increase in TUNEL-positive cells was observed in the SCI group (Fig. 3, L and M, and fig. S3G). Administration of GEVs and AEVs significantly reduced neuronal loss, with AEVs producing the most substantial neuroprotective outcome, as evidenced by the lowest number of TUNEL-positive cells. Collectively, these data indicate that AEVs attenuate lipid peroxidation and mitigate neuronal ferroptosis following SCI.

Fig. 3. AEV treatment attenuates oxidative stress–induced ferroptosis in male mice following SCI.

(A and B) Western blot analysis of ferroptosis-associated protein expression in spinal cord tissue on day 7 postinjury, with corresponding quantification (n = 3 per group). (C to E) RT-qPCR analysis of mRNA expression levels for ferroptosis-related genes in injured spinal cords (n = 3 per group). (F to H) Quantification of Fe²⁺ levels, along with GSH and MDA concentrations in spinal cord tissue (n = 6 per group). (I to K) Assessment of ROS production using DHE staining and fluorescence intensity analysis in spinal cord sections (n = 6 per group). (L and M) Evaluation of apoptotic activity via TUNEL staining and quantification of TUNEL-positive cells (n = 6 per group). Note that n denotes the number of independent biological replicates.

To verify the mechanistic association between functional recovery and ferroptosis suppression, Ferrostatin-1 (Fer-1; 1 mg/kg per day, MedChemExpress (MCE, USA) was administered intraperitoneally for 7 consecutive days post-SCI. Fer-1 treatment led to significant improvements in locomotor performance, as indicated by elevated BMS scores, improved rotarod outcomes, and increased grip strength relative to vehicle-treated controls (fig. S4, A to D). Histological assessment revealed a reduction in lesion volume (fig. S4E) and decreased neuronal apoptosis (fig. S4, F and G). Moreover, Fer-1 treatment mitigated ferroptosis in SCI mice, as evidenced by decreased ACSL4, 4HNE, PTGS2, MDA, and Fe²⁺ levels, increased GPX4 and GSH levels, and reduced ROS accumulation (fig. S4, H to M). The results substantiate the involvement of ferroptosis in SCI pathogenesis and indicate that its inhibition contributes significantly to the neuroprotective actions attributed to AEVs.

AEVs inhibit oxidative stress–induced ferroptosis in neuronal cells in vitro

To elucidate the inhibitory effects of AEVs on ferroptosis, neuronal cells were exposed to erastin (10 μM, 6 hours) followed by coculture with GEVs or AEVs. Western blot analysis revealed that both GEVs and AEVs attenuated erastin-induced up-regulation of ACSL4, 4HNE, and PTGS2 while restoring GPX4 expression (fig. S5, A and B). These protein-level changes were further supported by RT-qPCR analysis, which showed consistent modulation at the transcript level (fig. S5C). Hallmarks of ferroptosis, including lipid ROS accumulation, GSH depletion, lipid peroxidation, and Fe²⁺ overload, were evaluated posttreatment. GEVs and AEVs effectively reduced erastin-induced increases in lipid ROS, MDA, and Fe²⁺ while mitigating GSH depletion (fig. S5, D to H). Cell viability assays demonstrated that GEVs and AEVs alleviated erastin-induced cytotoxicity in neuronal cells, with AEVs exerting a more pronounced protective effect (fig. S5I). Calcein-AM/propidium iodide (PI) staining further confirmed these observations, as erastin-treated neurons exhibited substantial loss of viability and elevated cell death, whereas cotreatment with GEVs or AEVs significantly improved neuronal survival, particularly in the AEV group (fig. S5J). Extensive research has established a link between ferroptosis and mitochondrial dysfunction. To further delineate this relationship, erastin-induced alterations in mitochondrial function were examined following AEV intervention. AEVs markedly reduced mitochondrial superoxide accumulation in neuronal cells, an effect partially attenuated by GEVs (fig. S5, K and L). JC-1 staining revealed substantial recovery of mitochondrial membrane potential after exposure to both GEVs and AEVs, with AEVs exhibiting a more robust effect (fig. S5, M and N). Morphological assessment showed that erastin induced disruption of mitochondrial structure, including cristae loss and deformation of the outer membrane. Administration of GEVs and AEVs mitigated these structural abnormalities, with AEVs producing greater restoration (fig. S5O). Collectively, the data indicate that AEVs improve mitochondrial integrity and suppress ferroptosis driven by oxidative stress under in vitro conditions.

AEVs preferentially target neurons and modulate ferroptosis-related gene expression

To delineate the CNS cell types responsible for mediating AEVs’ effects, uptake assays were conducted using neurons, astrocytes, microglia, and endothelial cells. Cell isolation and culture followed protocols established in previous studies (24–26). Coculture with PKH26-labeled AEVs revealed that neuronal uptake was markedly higher than in other cell types (fig. S6), identifying neurons as the principal targets of AEVs.

To evaluate the influence of AEVs on neuronal ferroptosis in vivo, single-nucleus RNA sequencing (snRNA-seq) was performed on spinal cord tissue harvested 7 days postinjury. Analysis of 32,913 high-quality nuclei identified eight transcriptionally distinct cell clusters, including neurons, astrocytes, microglia, and others (Fig. 4, A and B). AEV treatment led to an increased proportion of surviving neurons following SCI (Fig. 4, C and D). Within the neuronal cluster, the expression of the proferroptotic gene Acsl4 was reduced, while the antiferroptotic genes Gpx4 and Fth1 were up-regulated in response to AEVs (Fig. 4, E to G). These transcriptional changes suggest heightened susceptibility of neurons to ferroptosis after SCI and indicate that AEVs confer a neuroprotective effect by modulating ferroptosis-associated gene expression. Collectively, the results support the conclusion that AEVs exert their protective action predominantly through neuronal regulation of ferroptotic pathways.

Fig. 4. snRNA-seq of mouse spinal cord after AEV treatment.

(A) UMAP plot of the cells collected from injured spinal cord at 7 dpi (days post injury) with or without AEV treatment. (B) Dot plot of marker genes used to annotate cell clusters in (A). (C) UMAP plots of cells collected from spinal cord of indicated groups. (D) Cell frequency of each cluster in (A). NK, natural killer. (E) Violin plot showing the expression of neuronal Gpx4 in indicated groups. (F) Violin plot showing the expression of neuronal Fth1 in indicated groups. (G) Violin plot showing the expression of neuronal Acsl4 in indicated groups. OGD, Oligodendrocyte.

AEVs promote recovery of motor function and inhibit the occurrence of ferroptosis after SCI by transferring RNF216 in vivo

In vivo and in vitro findings indicate that AEVs exert greater efficacy than GEVs in mitigating oxidative stress–induced ferroptosis and promoting functional recovery. Previous research has demonstrated that endurance exercise alters the protein composition of circulating GEVs, increasing the abundance of antioxidant proteins with cytoprotective roles. This raised the possibility that AEVs have a distinct proteomic profile. Comparative proteomic analysis revealed that RNF216 was the most significantly up-regulated protein in AEVs relative to GEVs. To validate the proteomic results, RNF216 expression was assessed in vitro, revealing markedly elevated protein levels in both AEVs and recipient neuronal cells (Fig. 5, A and B).

Fig. 5. AEVs promote recovery of motor function and inhibit ferroptosis after SCI by in vivo transfer of RNF216.

(A) Western blot analysis of RNF216 protein expression in GEVs and AEVs. (B) Western blot and quantitative analysis of RNF216 expression in neuronal cells (n = 3 per group). (C) Lentiviral-mediated overexpression of RNF216 in neural stem cells. NSCs, neural stem cells. (D) RNF216 expression in Ad-Vec-NEVs and Ad-RNF216-NEVs assessed by Western blotting. (E) RNF216 levels in neuronal cells treated with Ad-Vec-NEVs or Ad-RNF216-NEVs, analyzed by Western blotting and quantification (n = 3 per group). (F) Experimental grouping scheme. (G) BMS score evaluation of motor function in mice administered Ad-Vec-NEVs or Ad-RNF216-NEVs at 28 days postinjury (n = 6 per group). (H) Rotarod performance comparison between the two groups with statistical analysis (n = 6 per group). (I and J) Gait assessment via footprint analysis in both groups (n = 6 per group). (K) Quantification of mechanical sensitivity using the von Frey filament test (n = 6 per group). (L) NeuN immunofluorescence and statistical evaluation of neuronal survival (n = 3 per group). (M) NF200 immunofluorescence and corresponding quantification of axonal integrity (n = 3 per group). (N and O) Representative Western blots and quantitative analysis of ferroptosis-related proteins in the injured spinal cord (n = 3 per group). (P to R) RT-qPCR quantification of ferroptosis-associated gene expression in the injured spinal cord (n = 3 per group). (S to U) Assessment of Fe²⁺ concentration, GSH content, and MDA levels in spinal cord tissue (n = 6 per group). (V and W) ROS detection and fluorescence intensity analysis in spinal cord sections via DHE staining (n = 6 per group). (X) TUNEL staining and quantification of TUNEL-positive cells indicating cell death (n = 6 per group). Note that n denotes the number of independent biological replicates.

To further clarify the function of plasma EV–associated RNF216, neural stem cell–derived EVs (NEVs) were selected for RNF216 loading in vitro, in accordance with previous studies (27, 28). Neural stem cells were first isolated and cultured and then transduced with lentiviral vectors to overexpress RNF216 (Fig. 5C). Subsequently, NEVs were harvested and cocultured with neuronal cells. RNF216 expression was significantly higher in Ad-RNF216-NEVs compared to Ad-Vec-NEVs (Fig. 5D). Neuronal cells exposed to Ad-RNF216-NEVs similarly exhibited elevated RNF216 levels relative to those treated with control NEVs (Fig. 5E), indicating successful transfer of RNF216 from EVs to recipient cells.

To delineate the role of RNF216 in AEV-mediated suppression of ferroptosis and enhancement of functional recovery in vivo, Ad-Vec-NEVs and Ad-RNF216-NEVs were administered to mice post-SCI. Motor and sensory functions were evaluated through BMS scoring, rotarod performance, von Frey filament testing, and footprint analysis. Results indicated that Ad-RNF216-NEV treatment produced superior functional outcomes relative to Ad-Vec-NEVs (Fig. 5, F to K). Histological evaluation using NeuN and NF200 immunostaining revealed a greater increase in NeuN⁺ neuronal density and NF200⁺ axonal content in Ad-RNF216-NEV–treated spinal cords compared to Ad-Vec-NEVs (Fig. 5, L and M).

To further assess RNF216-mediated regulation of ferroptosis, Western blot and RT-qPCR analyses were performed. Expression levels of ferroptosis-promoting markers (ACSL4, 4HNE, and PTGS2) were further reduced, while the ferroptosis-repressing marker GPX4 was elevated following Ad-RNF216-NEV treatment (Fig. 5, N to R). Biochemical assays corroborated these molecular changes, demonstrating diminished GSH depletion, lower MDA generation, and reduced Fe²⁺ accumulation (Fig. 5, S to U). In addition, DHE and TUNEL staining confirmed attenuation of oxidative stress and cell death (Fig. 5, V to X). Collectively, these data indicate that the therapeutic efficacy of AEVs in promoting functional recovery and restraining ferroptosis partially relies on RNF216 activity.

AEVs stabilize mitochondrial function and inhibit oxidative stress–induced ferroptosis via RNF216

Neuronal cells were incubated with Ad-Vec-NEVs or Ad-RNF216-NEVs to assess the function of RNF216 in modulating cellular responses to erastin in vitro. Western blot and RT-qPCR analyses demonstrated that RNF216 overexpression within NEVs led to reduced expression of ACSL4, 4HNE, and PTGS2, accompanied by increased GPX4 levels during ferroptosis (fig. S7, A to D). Compared to the negative control, RNF216-enriched NEVs more effectively suppressed erastin-induced lipid ROS accumulation, GSH depletion, MDA production, and Fe²⁺ overload, resulting in enhanced cell viability (fig. S7, E to K). Mitochondrial function was also more effectively preserved by Ad-RNF216-NEV treatment, as indicated by lower mitochondrial superoxide levels, improved membrane potential, and reversal of erastin-induced morphological alterations (fig. S7, L to O).

To validate the gain-of-function results, NEVs derived from RNF216-deficient neural stem cells were examined (fig. S8A). In contrast to wild type (WT)–NEVs, RNF216-KO NEVs failed to suppress ferroptosis-related alterations, as reflected by elevated MDA, reduced GSH, increased ACSL4, 4HNE, and PTGS2 expression, and diminished GPX4 levels (fig. S8, B to E, I, and J). Mitochondrial dysfunction was further evidenced by depolarized membrane potential, increased mitochondrial ROS, and reduced neuronal survival (fig. S8, F to H, K, and L). These results identify RNF216 as a necessary component for mediating the ferroptosis-suppressive and neuroprotective effects of NEVs. Collectively, the in vitro data delineate a role for RNF216 in preserving mitochondrial integrity and mitigating oxidative stress–induced ferroptotic damage in neuronal cells.

RNF216 promotes NOX1 ubiquitination and degradation

To investigate the mechanism by which RNF216 modulates ferroptosis in neuronal cells, immunoprecipitation–mass spectrometry (IP-MS) was used to identify RNF216-associated proteins (Fig. 6A). Coimmunoprecipitation (co-IP) subsequently validated this interaction, revealing that RNF216 pulled down NOX1, and reciprocal co-IP confirmed NOX1 also precipitated RNF216 (Fig. 6B). In human embryonic kidney (HEK) 293T cells, Flag-tagged RNF216 coprecipitated with Myc-tagged NOX1, and the reciprocal interaction was similarly detected (Fig. 6, C and D), establishing a physical association between the two proteins. To determine the functional relevance of this interaction, the effects of RNF216 knockdown on NOX1 expression were examined. Compared to shCtrl, silencing RNF216 significantly increased NOX1 protein levels without altering its mRNA expression (Fig. 6E), indicating a posttranscriptional regulatory mechanism potentially involving protein stability. Treatment with the proteasome inhibitor MG132 further increased NOX1 protein levels, which were already elevated due to RNF216 knockdown (Fig. 6, F and G), supporting the role of RNF216 in the proteasomal degradation of NOX1. Further analysis assessed whether RNF216’s E3 ligase activity contributes to NOX1 regulation. Overexpression of WT RNF216 reduced NOX1 protein levels, whereas the catalytically inactive C688A mutant failed to do so (Fig. 6H), implicating ubiquitin ligase activity in NOX1 degradation. To assess the effect on protein stability, cycloheximide chase assays were conducted. RNF216 overexpression accelerated NOX1 degradation, while RNF216 knockdown significantly delayed its turnover (Fig. 6, I to L), confirming RNF216-mediated destabilization of NOX1 in a proteasome-dependent manner.

Fig. 6. RNF216 binds and regulates NOX1 protein expression.

(A) Representative peptide sequences. (B) Co-IP analysis confirmed endogenous RNF216-NOX1 interaction in neuronal cells. IgG, immunoglobulin G. (C and D) Exogenous expression of RNF216 and NOX1 in HEK 293T cells also demonstrated reciprocal binding under overexpression conditions. (E) Quantification of NOX1 mRNA levels in neuronal cells transfected with shCtrl or shRNF216 (n = 3 per group). (F and G) Western blot analysis of RNF216 and NOX1 protein levels in shRNF216-transfected neuronal cells with or without MG132 treatment (n = 3 per group). (H) Western blot detection of NOX1 and Flag-tagged RNF216 in HEK 293T cells transfected with increasing doses of WT or C688A RNF216 constructs. (I and J) Assessment of NOX1 protein stability following RNF216 overexpression in HEK 293T cells. (K and L) Evaluation of NOX1 protein stability after RNF216 knockdown in HEK 293T cells. Note that n denotes the number of independent biological replicates. CHX, cycloheximide. IB, Immunoblotting.

RNF216, an E3 ubiquitin ligase, plays a regulatory role in modulating NOX1 protein stability. This study examined whether RNF216 governs NOX1 ubiquitination and subsequent proteasomal degradation. RNF216 silencing markedly reduced NOX1 ubiquitination while elevating NOX1 protein levels relative to shCtrl (Fig. 7, A and B). To further assess the functional contribution of RNF216, HEK 293T cells were cotransfected with Flag-RNF216 (either WT or the catalytically inactive C688A mutant), Myc-NOX1, and HA-Ub. Overexpression of WT RNF216 enhanced NOX1 ubiquitination, whereas the C688A mutant exerted negligible influence (Fig. 7, C and D). In vivo, endogenous RNF216-mediated regulation of NOX1 ubiquitination was evaluated in spinal cord tissues post-SCI. AEV administration led to a pronounced increase in NOX1 ubiquitination and degradation compared to GEVs (Fig. 7, E and F). This effect was further intensified in mice treated with Ad-RNF216-NEVs versus Ad-Vec-NEVs (Fig. 7, G and H). Among the two predominant polyubiquitin linkage types—Lys⁴⁸- and Lys⁶³-linked chains—RNF216 selectively removed Lys⁶³-linked chains from NOX1, with a limited effect on Lys⁴⁸ linkages (Fig. 7I). To verify the functional relevance of Lys⁶³-linked ubiquitination in RNF216-dependent NOX1 turnover, a ubiquitin mutant resistant to Lys⁶³ linkage (Lys63R) was expressed in RNF216-deficient HEK 293T cells. Introduction of Lys63R ubiquitin neutralized the RNF216 knockdown–induced accumulation of NOX1 (Fig. 7J). To delineate the molecular basis of RNF216-mediated NOX1 ubiquitination, six lysine residues (K38, K44, K92, K500, K554, and K561) were identified through sequence analysis and mass spectrometry. Among them, K500 exhibited strong evolutionary conservation across species (fig. S9A). Site-directed mutagenesis revealed that K500 substitution markedly impaired RNF216-induced NOX1 ubiquitination, identifying it as a key regulatory site (fig. S9B). These data suggest that RNF216 regulates NOX1 stability in neuronal cells.

Fig. 7. RNF216 promotes NOX1 ubiquitination.

(A) Neuronal cell lysates transfected with shCtrl or shRNF216 and treated with MG132 were subjected to immunoprecipitation with anti-NOX1, followed by immunoblotting using the indicated antibodies. (B) Quantification of Ub-conjugated NOX1 (Ub-NOX1) levels (n = 3 per group). (C) Analysis of NOX1 ubiquitination in neuronal cells overexpressing either WT RNF216 or its C688A mutant. (D) Quantification of Ub-NOX1 in response to RNF216 or C688A overexpression (n = 3 per group). (E) Immunoprecipitation of spinal cord lysates postinjury with anti-NOX1, followed by immunoblotting for ubiquitin and NOX1 in the indicated mouse groups. (F) Quantification of relative Ub-NOX1 levels from (E) (n = 3 per group). (G) Comparative analysis of NOX1 ubiquitination in spinal cord lysates from both mouse groups postinjury via immunoprecipitation and Western blotting. (H) Quantification of Ub-NOX1 from (G) (n = 3 per group). (I) HEK 293T cells were cotransfected with Flag-RNF216, Myc-NOX1, and HA-tagged ubiquitin variants (WT, K48 only, and K63 only) to evaluate linkage specificity in NOX1 ubiquitination. (J) HEK 293T cells transfected with Ub WT or Ub K63R were cultured with either shCtrl or shRNF216. Lysates were analyzed by Western blotting with anti-NOX1 and anti-RNF216. Note that n denotes the number of independent biological replicates.

Knockdown of NOX1 alleviates oxidative stress and neuronal injury under erastin-induced conditions

To elucidate the functional role of NOX1 in neuronal ferroptosis, loss-of-function analyses were conducted in erastin-treated neurons using Ad-shNOX1. Neurons transduced with shNOX1 displayed markedly reduced MitoSOX and total intracellular ROS levels compared to controls (fig. S10, A and B). Restoration of mitochondrial membrane potential was evidenced by JC-1 staining in shNOX1-treated neurons (fig. S10, C and D), while Calcein-AM/PI staining demonstrated improved neuronal survival (fig. S10, E and F). These results establish NOX1 as a key mediator of oxidative stress–related injury in SCI and support the neuroprotective effect conferred by its degradation via RNF216.

RNF216 in AEVs inhibits oxidative stress–induced ferroptosis in neuronal cells by promoting NOX1 degradation

To investigate the mechanistic link between RNF216-enriched EVs and NOX1 in modulating ferroptosis, NOX1 was overexpressed in neuronal cells, followed by in vitro rescue experiments. Overexpression of NOX1 substantially reversed the suppression of ferroptosis and cell death induced by Ad-RNF216-NEVs (fig. S11, A to K). Furthermore, the improvements in mitochondrial membrane potential, reduction in mitochondrial superoxide levels, and enhancement of neuronal viability observed with Ad-RNF216-NEVs were abolished upon NOX1 up-regulation (fig. S11, L to O). These data indicate that RNF216-containing AEVs exert their protective effects by modulating mitochondrial function and ferroptotic activity through targeted degradation of NOX1.

AEVs promote recovery of motor function and inhibit ferroptosis in mice via NOX1 ubiquitination and degradation

To delineate the role of NOX1 in AEV-mediated neuroprotection, adeno-associated virus (AAV)–mediated overexpression of NOX1 in neuronal cells was induced in mice following SCI, followed by administration of Ad-RNF216-NEVs. Behavioral assessments, including BMS scoring, footprint analysis, von Frey filament testing, and rotarod performance, indicated markedly impaired functional recovery in the Ad-RNF216-NEVs + Ad-NOX1 group compared to the Ad-RNF216-NEVs + Ad-Vec group (Fig. 8, A to F). Immunostaining for NeuN and NF200 corroborated the behavioral outcomes, demonstrating reduced neuronal integrity in the Ad-NOX1 overexpression group (Fig. 8, G and H). Moreover, evaluation of ferroptosis markers and immunofluorescence analyses revealed that NOX1 overexpression exacerbated neuronal ferroptosis within the injured spinal cord despite AEV treatment (Fig. 8, I to Q). Collectively, the results suggest that the neuroprotective effects of AEVs on functional recovery are mediated, at least in part, through NOX1 ubiquitination and degradation, thereby mitigating oxidative stress–induced ferroptosis in vivo.

Fig. 8. AEVs promote recovery of motor function and inhibit ferroptosis in mice via NOX1 ubiquitination and degradation.

(A) Experimental subgrouping information. (B) BMS scores assessed 28 days postinjury in mice treated with Ad-RNF216-NEVs + Ad-Vec or Ad-RNF216-NEVs + Ad-NOX1 (n = 6 per group). (C) Quantitative analysis of motor coordination using the rotarod test (n = 6 per group). (D and E) Gait evaluation via footprint analysis in both treatment groups (n = 6 per group). (F) Mechanical sensitivity assessment using the von Frey filament test (n = 6 per group). (G) NeuN immunofluorescence staining and corresponding quantification of neuronal survival (n = 3 per group). (H) NF200 immunofluorescence staining and quantification of axonal integrity (n = 3 per group). (I and J) Representative Western blot images and statistical quantification of ferroptosis-related proteins in the injured spinal cord (n = 3 per group). (K) RT-qPCR quantification of mRNA expression levels of ferroptosis markers in spinal cord tissue (n = 3 per group). (L to N) Quantification of Fe²⁺ concentration, GSH content, and MDA levels in spinal cord homogenates (n = 6 per group). (O and P) ROS detection and fluorescence quantification using DHE staining in spinal cord sections (n = 6 per group). (Q and R) Cell damage assessment and quantification of TUNEL-positive cells using TUNEL staining (n = 6 per group). Note that n denotes the number of independent biological replicates.

DISCUSSION

The pathophysiology of SCI remains difficult to manage, with the mechanisms underlying neuronal loss and functional restoration still insufficiently defined. This study used a combination of in vivo and in vitro approaches to demonstrating that RNF216 carried by AEVs suppresses neuronal ferroptosis post-SCI, thereby promoting recovery of function (Fig. 9). This regulatory mechanism has not been previously reported. Mechanistically, RNF216 mitigated oxidative stress–induced ferroptosis in neuronal cells by enhancing the ubiquitination and subsequent degradation of NOX1. Both in vivo and in vitro NOX1 overexpression models, alongside functional recovery assessments, corroborated this regulatory axis. This study constitutes the first evidence that AEVs contribute to functional restoration following SCI, highlighting their therapeutic relevance in translational neuroregeneration research.

Fig. 9. Schematic diagram of plasma EVs in athletes inhibiting neuronal ferroptosis and promoting neurological recovery following SCI.

AEVs reduce mitochondrial damage and ferroptosis in neuronal cells by modulating ferroptosis-related pathways. RNF216 in AEVs enhances the ubiquitination and degradation of NOX1, reducing oxidative stress and ferroptotic cell death, thereby supporting functional recovery after SCI.

Extensive evidence supports the diagnostic and therapeutic relevance of EVs in a broad spectrum of pathologies, including neurodegenerative disorders, cardiovascular conditions, and malignancies (10, 29, 30). Their nanoscale dimensions and membrane permeability enable EVs to traverse the blood-brain barrier and deliver bioactive cargos, rendering them promising candidates for CNS-targeted interventions (31). Zhang et al. (32) reported that bone marrow mesenchymal stromal cell–derived EVs contributed to myelin repair and mitigated neuroinflammatory responses in demyelinated CNS tissues. In another study, ultrasound-stimulated astrocytes increased EV secretion, which alleviated amyloid-β–induced cytotoxicity (33). Liu et al. (11) found that EVs derived from melatonin-treated MSCs enhanced NRF2 stabilization via USP29 transfer, thereby promoting neurological recovery posttraumatic SCI. Despite these advances, the reliance on EVs produced from in vitro–cultured cells introduces limitations in both yield and consistency. In contrast, plasma-derived EVs—naturally present in human blood—may offer broader applicability across diverse disease contexts and patient cohorts. Notably, EVs isolated from the plasma of young, healthy donors have demonstrated potential in modulating inflammatory processes (17) and attenuating disease severity in COVID-19 models (34). Furthermore, EVs released into circulation following endurance exercise exhibit cardioprotective properties, attributed to their activation of antioxidant signaling pathways (23). Of particular importance, the current study is the first to demonstrate that plasma EVs sourced from healthy athletes have superior efficacy in enhancing neurological recovery after SCI in mice, relative to those derived from the general population.

Secondary SCI induces programmed neuronal death, contributing to neurological dysfunction. This form of cell death includes multiple pathways, including apoptosis, autophagy, and ferroptosis. Recent research has increasingly focused on the regulatory mechanisms of ferroptosis, which has been associated with various CNS disorders such as neurodegenerative diseases (35), traumatic brain injury (7), SCI (36, 37), and hypoxic-ischemic brain injury (38). Previous work confirmed that ferroptosis constituted a primary mode of neuronal loss following SCI, driven by intracellular iron-dependent oxidative stress (8). In the current study, plasma-derived EVs attenuated oxidative stress–induced ferroptosis in both in vitro and in vivo models. Among them, AEVs demonstrated greater therapeutic benefit compared to GEVs. Given that EV-mediated effects are largely attributed to their protein cargo, mechanistic investigation focused on identifying specific molecular mediators. Proteomic profiling revealed significant up-regulation of the E3 ubiquitin ligase RNF216 in AEVs, which acted as an antioxidant modulator in neuronal cells. Subsequent experiments involving NEVs overexpressing RNF216 confirmed the inhibitory effects on erastin-induced lipid ROS production, GSH depletion, MDA accumulation, and Fe²⁺ overload. Although numerous proteins were found to be up-regulated in AEVs, the current study concentrated on elucidating the function of RNF216. Nevertheless, the potential contribution of additional EV-associated proteins in ferroptosis regulation and post-SCI neurofunctional recovery remains unresolved. Future research is needed to explore the broader spectrum of biomolecules delivered by EVs that may influence SCI outcomes.

Ubiquitination, a widespread posttranslational modification, entails the covalent attachment of ubiquitin molecules to substrate proteins via isopeptide bonds through the coordinated action of E1 activating, E2 conjugating, and E3 ligating enzymes. This enzymatic cascade governs a range of biological functions, including proteasomal degradation and intracellular signaling pathways (39). Increasing evidence indicates that ubiquitination plays a regulatory role in ferroptosis. Nguyen et al. (40) reported that MARCHF6, an E3 ubiquitin ligase, recognized NADPH through its C-terminal regulatory domain, thereby enhancing its ligase activity to target and degrade key ferroptosis mediators such as ACSL4 and p53, ultimately down-regulating ferroptosis. Similarly, Wu et al. (41) demonstrated that HECT, UBA and WWE domain containing 1 (HUWE1)/ Mdm2 ubiquitin ligase, E3 (MULE) ubiquitinated and degraded the transferrin receptor, mitigating iron overload and suppressing ferroptosis in acute liver injury. In the present study, RNF216, an E3 ubiquitin ligase, was markedly enriched in AEVs and implicated in modulating ferroptosis in the context of neuronal oxidative stress. Previous research has associated RNF216 with various physiological and pathological processes, including spermatogenesis (42), glioblastoma (43), Gordon-Holmes syndrome (44), and regulation of the hypothalamic-pituitary-gonadal axis (45). Nonetheless, the precise mechanism by which RNF216 influences neuronal ferroptosis following traumatic CNS injury—particularly in SCI—remains undefined. To investigate its functional relevance in SCI, RNF216 was overexpressed in neural stem cells, and the NEVs were used as a delivery vehicle. NEVs were chosen on the basis of their intrinsic neurotropic properties, biocompatibility, and scalability for in vitro manipulation, rendering them well suited for CNS-targeted interventions. Previous work confirmed that NEVs efficiently delivered therapeutic agents such as 14-3-3τ and small interfering RNA to injured spinal tissues, attenuating oxidative damage and inflammation while supporting neurological recovery (27, 28). Collectively, these findings reinforce the strategic rationale and translational feasibility of using NEVs for RNF216 delivery in the treatment of SCI.

To further elucidate the mechanism by which RNF216 modulates ferroptosis in neuronal cells, IP-MS analysis was conducted, identifying NOX1 as a potential interacting partner. NOX1, a member of the NOX family, generates superoxide by transferring electrons from NADPH, thereby contributing to intracellular ROS production (46). Its expression is regulated at multiple levels, including transcriptional and posttranscriptional processes (47, 48). Posttranslational modification via SUMOylation also affects NOX1 activity; SUMO1 has been shown to inhibit NOX1-NOX5 function, while suppression of SUMO1 or global SUMOylation increases ROS output (48). Joo et al. (49) reported that interaction between NoxO1 and Grb2 enhanced the recruitment of the Cbl E3 ligase, reducing NoxO1 stability and indirectly modulating NOX1 activity. However, no E3 ubiquitin ligase had previously been confirmed to directly mediate NOX1 ubiquitination and degradation. The current study identified RNF216 as an E3 ligase that binds to NOX1, enhancing its ubiquitination and subsequent proteasomal degradation. This regulatory mechanism attenuated ROS accumulation in neuronal cells and suppressed ferroptotic cell death. In vivo, NOX1 overexpression via AAV neutralized the ferroptosis-suppressive effect of RNF216 and impaired motor function recovery, supporting the regulatory role of RNF216 in post-SCI neuronal repair. Collectively, these results indicate that RNF216 governs ferroptosis onset and motor recovery by targeting NOX1 for degradation.

Although the present study primarily centered on the RNF216-NOX1-ferroptosis axis in neurons, AEVs are known to contain a broad spectrum of bioactive molecules and may influence multiple cell types within the injured spinal cord microenvironment. To clarify cell type specificity, coculture uptake assays demonstrated preferential internalization of AEVs by neurons. Consistently, snRNA-seq analysis revealed that AEV exposure significantly altered the expression of ferroptosis-associated genes in neuronal populations, with negligible effects observed in glial subsets. These results delineate a neuron-specific ferroptotic regulatory pathway. Nonetheless, the involvement of glial activation, immune modulation, and axonal regrowth cannot be excluded from contributing to the therapeutic effects of AEVs. Comprehensive characterization of these multicellular interactions and the role of nucleic acid cargo within AEVs remain an essential direction for subsequent investigations.

In addition, the data indicate that exercise training markedly enhances RNF216 expression in plasma-derived EVs, offering a novel paradigm for noninvasive, exercise-responsive therapeutic development. Compared to conventional approaches such as drug administration or viral vector systems, exercise-induced endogenous EV secretion presents benefits in terms of biosafety, physiological regulation, and long-term patient adherence—features particularly suited for individuals undergoing extended spinal cord rehabilitation. Carefully structured exercise regimens not only support systemic metabolic balance and neurological recovery but may also promote the secretion of EVs enriched with RNF216 and other reparative components. These EVs could contribute to neuroprotection and structural repair at the injury site by modulating ferroptotic cascades, attenuating oxidative damage, and influencing the immune milieu. Further investigation is warranted to delineate the relationship between exercise and the molecular composition of circulating EVs.

In summary, AEVs enhance functional motor recovery in mice following SCI and contribute to the degradation of NOX1 through elevated ubiquitination mediated by RNF216 delivery. This mechanism mitigates oxidative stress–induced ferroptosis in neuronal tissue. Considering the involvement of ferroptosis in neuronal loss across various CNS pathologies, the present study expands current understanding of AEVs as a therapeutic strategy and supports their potential application in SCI intervention.

MATERIALS AND METHODS

Human donor selection and sample collection

Donors were healthy young adults aged 20 to 35 years, nonsmokers, and free of chronic conditions. The athlete cohort comprised national or provincial competitive athletes engaged in swimming, track and field, or weightlifting, all with over 5 years of structured training and a minimum weekly exercise duration of 10 hours. In contrast, individuals in the general population group lacked any history of regular or systematic physical activity.

Table S1 summarizes donor characteristics, including sex, age, body mass index (BMI), training duration, weekly training hours, sport discipline, and recent training intensity. To control for interindividual variability and minimize potential confounding factors, blood samples were collected under standardized conditions: in the morning following an overnight fast and at least 24 hours postexercise. EV quality control included free hemoglobin concentration to assess plasma hemolysis and platelet counts. No statistically significant differences were observed between groups regarding age, BMI, hemolysis, or platelet levels.

Spinal cord neuronal cell culture

Spinal cords were harvested via cervical dislocation under sterile conditions using a somatic microscope. Following removal of the dorsal root ganglia, the tissue was dissected into ~1-mm³ fragments with ophthalmic scissors. Digestion was performed by incubating the fragments in 0.25% trypsin (Thermo Fisher Scientific) at 37°C in a 5% CO₂ environment for 20 min. Enzymatic activity was halted by the addition of 2 ml of horse serum (Sigma-Aldrich). The resulting cell suspensions were filtered through a 70-μm cell strainer and centrifuged at 300g for 5 min to discard the supernatant. The pellet was resuspended in a seeding solution containing 10% horse serum and gently agitated to ensure homogeneity. After cell counting, the suspension was incubated for 4 hours to permit complete adhesion, followed by replacement of the medium with neuronal culture medium. Thereafter, half-medium changes were carried out every other day.

Extraction and identification of plasma EVs

Peripheral blood samples were collected from two cohorts comprising healthy individuals aged 20 to 35: one from the general population and the other consisting of athletes. The blood collection protocol received ethical clearance from the Ethics Committee of the First Affiliated Hospital of the University of Science and Technology, China (ethical approval no. 2024KY172), and written informed consent was obtained from all participants. EVs were isolated through a series of differential centrifugation steps. Initially, whole blood was centrifuged at 4°C, 2000g for 10 min to separate plasma. The plasma was then subjected to sequential centrifugation at 3000g and 10,000g for 20 min to remove cellular debris and macromolecular contaminants. Subsequent ultracentrifugation at 140,000g for 70 min at 4°C was performed, and the pellet was resuspended in PBS. To further improve purity, the resuspended pellet was loaded onto a discontinuous iodixanol density gradient and centrifuged at 140,000g for 16 hours at 4°C. The EV-enriched fraction was collected, diluted in PBS, and subjected to an additional ultracentrifugation under identical conditions to obtain highly purified EVs.

The morphological features of isolated EVs were examined using TEM. Particle size and distribution profiles were quantified by NTA, while protein concentration was determined using the Bicinchoninic Acid (BCA) assay. Western blotting was used to detect characteristic EV protein markers, including CD63, CD9, CD81, ALIX, and TSG101.

Cellular uptake assay

To determine the cellular specificity of AEVs, in vitro coculture assays were conducted using neurons, astrocytes, microglia, and brain microvascular endothelial cells. Each cell type was independently incubated with PKH26-labeled AEVs (10 μg/ml) for 48 hours. Following PBS rinsing and fixation, nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI), and cellular uptake of AEVs was visualized via confocal laser scanning microscopy (Zeiss LSM 880). Quantification of PKH26 fluorescence intensity was performed using ImageJ to assess differential uptake among cell populations.

CCK-8 assay

Cells were seeded into 96-well plates at a density of 1 × 10⁴ cells per well. After the designated treatments, culture medium was replaced with fresh medium. Cell Counting Kit-8 (CCK-8) reagent (Dojindo, Japan) was then added according to the manufacturer’s instructions, followed by a 3-hour incubation at 37°C. Absorbance at 450 nm was measured using a microplate reader to evaluate cell viability.

Calcein-AM/PI staining

Cells were seeded into 24-well plates at an appropriate density. After treatment in accordance with experimental protocols, 250 μl of Calcein-AM/PI working solution (Dojindo, Japan) was added to each well, followed by incubation at 37°C for 30 min in the dark. Upon completion, fluorescent signals were observed using a fluorescence microscope (Calcein-AM: green; PI: red). Quantitative fluorescence intensity was subsequently analyzed using ImageJ.

Determination of MDA, GSH, and iron content

MDA and GSH levels were determined using commercial assay kits (MDA: Beyotime, Shanghai, China; GSH: Sigma-Aldrich), following the manufacturers’ instructions. For Fe²⁺ quantification in murine spinal cord tissues, samples from each group were homogenized and processed using a specific detection kit (Elabscience, Wuhan, China) according to the standard protocol. Absorbance readings were obtained via microplate reader, and Fe²⁺ concentrations were calculated on the basis of a standard calibration curve.

Measurement of unstable intracellular iron

Labile iron (Fe²⁺) was quantified using the fluorescent probe SiRhoNox-1 (FerroFarRed, Goryo Chemical, Japan) following established protocols (50). Cells were cultured on confocal dishes and incubated with 5 μM SiRhoNox-1 for 1 hour at 37°C. After staining, the medium was substituted with observation buffer, and fluorescence intensity was evaluated via confocal microscopy. Image analysis was performed using ImageJ/Fiji.

JC-1 measurement

JC-1 dye (Beyotime, Shanghai, China) was applied to cells for 20 min at 37°C in the dark, followed by two rinses with JC-1 buffer. JC-1 aggregates and JC-1 monomers emitted red and green fluorescence, respectively. The red/green fluorescence ratio, indicative of mitochondrial membrane potential, was quantified through fluorescence microscopy.

ROS detection

Mitochondrial ROS levels were visualized using MitoSOX Red (Molecular Probes, USA) under confocal microscopy. For total cellular ROS, DCFH-DA (Beyotime, Shanghai, China) was applied according to the manufacturer’s instructions. Following treatment, cells were incubated with the probe at 37°C for 20 min in darkness. After adequate washing, fluorescence intensity was measured by flow cytometry and analyzed using FlowJo software.

Mitochondrial morphology observed by TEM

Tissue specimens were fixed overnight in 2.5% glutaraldehyde, followed by postfixation with 1% osmium tetroxide for 1 hour. Samples were subsequently embedded in resin, sectioned into 85-nm-ultrathin slices, and stained with 2% lead citrate and uranyl acetate. Imaging was performed using TEM to visualize mitochondrial ultrastructure.

Plasmid and cell transfection

Flag-tagged RNF216 and RNF216 C688A, along with Myc-tagged NOX1 expression plasmids, were generated by inserting the respective coding sequences with N-terminal tags (Genebay Biotech). HA-tagged ubiquitin K63 and K48 plasmids were commercially obtained (Genebay Biotech). Adenoviral vectors for RNF216 overexpression (Ad-RNF216) and knockdown (shRNF216) were engineered by Hanbio Biotechnology, with corresponding random constructs used as negative controls. Transfection procedures used Lipofectamine 3000 (Invitrogen) according to the manufacturer’s protocol. For NOX1 overexpression, AAV vectors (AAV-Syn-MCS-NOX1-3xFlag, designated AAV-NOX1) and control vectors (AAV-Syn-MCS-3xFlag, designated AAV-Vec) were prepared by Hanbio Biotechnology. Following SCI, mice were secured in a stereotactic apparatus and injected intrathecally with 1 × 10⁹ vg of viral vectors into the spinal cord at an infusion rate of 0.25 μl/min. For NOX1 knockdown, primary neurons were infected with adenoviral vectors carrying NOX1 short hairpin RNA (shRNA; Ad-shNOX1) or control shRNA (Ad-shCtrl) vectors per vendor instructions.

RNA isolation and RT-qPCR

Total RNA was extracted from cultured cells or spinal cord tissues using TRIzol reagent, followed by quantification with a NanoDrop microspectrophotometer. cDNA synthesis was performed according to the manufacturer’s instructions using a reverse transcription kit. RT-qPCR was conducted with 2× RealStar Green Power Mixed (GenStar), using glyceraldehyde phosphate dehydrogenase as the internal reference. Primer sequences for all target genes are listed in table S2.

Western blotting analysis

Cellular protein samples were prepared by placing culture dishes on ice, rinsing cells with PBS, and adding an appropriate volume of lysis buffer. For tissue samples, 1 ml of lysis buffer was applied per 100 mg of tissue, followed by thorough homogenization. All lysates were incubated on ice for 30 min and then centrifuged at 12,000g for 15 min at 4°C. The supernatants were collected, and pellets were discarded. Protein concentrations were measured using the BCA assay. Equal amounts of protein (25 μg per sample) were separated by 10% SDS–polyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad) and transferred onto 0.2-μm polyvinylidene difluoride membranes (Millipore). Membranes were blocked with 5% skim milk for 2 hours and subsequently incubated with primary antibodies (table S3), followed by horseradish peroxidase–conjugated secondary antibodies (1:10,000; Jackson ImmunoResearch). Detection was performed by applying chemiluminescent substrate to the membrane at the location of target proteins. Bands were visualized using an imaging system, and grayscale intensities were quantified using ImageJ software.

Immunoprecipitation–mass spectrometry

Following a 30-min lysis on ice using radioimmunoprecipitation assay buffer (Beyotime), cell lysates were centrifuged at 12,000g for 25 min to clear the supernatant. An aliquot of the supernatant was processed for Western blotting. The target antibody was incubated with Protein A+G magnetic beads for 15 min to 1 hour at room temperature on a Flip Mixer, after which the remaining lysate was added and the mixture was incubated overnight at 4°C. Magnetic separation was performed, followed by three washes with 1× tris-buffered saline. Bound proteins were eluted using SDS-PAGE loading buffer and subjected to Western blotting. To profile the protein cargo of EVs, liquid chromatography–tandem mass spectrometry was conducted. Differential expression was determined on the basis of a fold change cutoff of >1.5 or <0.67 and P < 0.05.

Immunofluorescence staining

Spinal cord tissues collected at designated time points post-SCI were cryosectioned at 8-μm thickness. Neuronal cells or tissue sections were fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.3% Triton X-100, and blocked with 10% goat serum for 1 hour at room temperature. Samples were incubated overnight at 4°C with primary antibodies (table S4), followed by a 1-hour room temperature incubation with secondary antibodies (Alexa Fluor 488 or Alexa Fluor 594; 1:200; Jackson ImmunoResearch). Nuclear counterstaining was performed using DAPI (1:1000; Invitrogen, Carlsbad, CA). Imaging was carried out by fluorescence microscopy, and quantitative analysis was conducted using ImageJ.

RNF216⁻/⁻ mice

RNF216⁻/⁻ mice (C57BL/6NCya background) were generated by Cyagen Biosciences (Guangzhou, China) through CRISPR/Cas9-mediated deletion of exons 3 to 5 in the Rnf216 gene (NCBI RefSeq: NM_207110; Ensembl: ENSMUSG00000045078), which constitute ~37.3% of the coding sequence. This gene, located on mouse chromosome 5, encodes a 910–amino acid protein (ENSMUST00000200607). A 3.4-kb genomic fragment including the targeted exons was excised, and genotyping was performed via PCR followed by sequence verification. Homozygous mutants were used for neural stem cell isolation and subsequent NEV production for downstream functional assays.

SCI model

All experimental procedures were conducted in accordance with institutional guidelines for the care and use of laboratory animals and protocols, which were approved by the Animal Care and Use Committee of the First Affiliated Hospital of the University of Science and Technology, China [approval no. 2024-N(A)-174]. The SCI model was generated in 6- to 8-week-old male C57BL/6 mice according to established procedures (24, 28). Following induction with isoflurane, anesthesia was maintained via intraperitoneal administration of sodium pentobarbital. Mice were positioned prone for precise localization of the T10 vertebra. After skin preparation, the dorsal surgical site was disinfected with 75% ethanol and draped. A midline skin incision was made, and paraspinal muscles were bluntly dissected to expose the T10 vertebra. Laminectomy was performed to access the spinal dural sac by removing the T10 spinous process and vertebral plate. A contusion injury was induced by dropping a 5-g rod (1 mm in diameter, Reward, Shenzhen, China) from a height of 6 cm directly onto the exposed cord. Prophylactic antibiotics were administered preoperatively and continued once daily for 5 days postsurgery. Manual bladder expression was conducted daily until the resumption of spontaneous voiding was observed. Animals were randomly assigned to five groups, each receiving daily tail vein injections for 5 consecutive days. Treatment regimens included GEVs, AEVs, Ad-Vec-NEVs, Ad-RNF216-NEVs (each containing 100 μg of the total EV protein in 100 μl of PBS), or PBS alone (100 μl) as the control. To examine treatment applicability across sexes, female mice underwent identical SCI modeling and received GEV or AEV injections (containing 100 μg of the total protein from EVs in 100 μl of PBS) using the same dosage and protocol.

Functional behavior analysis

BMS scoring was used to evaluate hindlimb reflexes and locomotor function following SCI. Mice were acclimated to the testing environment before assessment. Motor performance was subsequently recorded and independently rated by three trained evaluators based on the extent of motor impairment (score range: 0 = complete motor paralysis; 9 = normal function). Evaluations were conducted at predetermined postinjury time points.

Footprint analysis was conducted on day 28 post-SCI. Forepaws and hindpaws were coated with distinct ink colors, and mice were guided to traverse a runway lined with absorbent paper. The resulting footprints were analyzed to assess gait patterns and functional recovery.

Rotarod testing was carried out 28 days postinjury to assess sensorimotor coordination. Mice were placed on an accelerating rotarod (0 to 40 rpm), and both latency to fall and corresponding rotation speed were recorded. Each mouse underwent one acclimatization session followed by two test trials, with average values calculated for analysis.

The von Frey filament assay was used to evaluate mechanical sensitivity of the plantar surface. Mice were positioned on a metal mesh platform to allow access to the hind paws. A series of calibrated nylon filaments with ascending force gradients were applied until a withdrawal reflex was observed, and the mechanical threshold was recorded as the minimal force inducing a positive response.

In vivo biodistribution of AEVs

DiO-labeled AEVs (5 μM, MCE) were prepared by incubating vesicles at 37°C for 20 min, followed by ultracentrifugation to eliminate unbound dye. The labeled AEVs were administered intravenously via tail vein into SCI mice. Spinal cord tissues were collected at 6, 12, and 24 hours postinjection and subjected to ex vivo fluorescence imaging using the IVIS system (PerkinElmer). Signal intensities were quantified using Living Image software to evaluate biodistribution dynamics.

Spinal cord MRI

On day 28 post-SCI, MRI was conducted to assess spinal cord integrity. From each treatment group, mice were randomly selected, and sagittal and axial T2-weighted images were acquired using a Bruker BioSpec 7T/20USR small animal MRI system (Germany), following previously established imaging parameters and protocols (28).

DHE staining

Oxidative stress within spinal cord tissue was evaluated using DHE staining. Fresh frozen sections were incubated with 2 μM DHE (Thermo Fisher Scientific, USA) in a humidified chamber at 37°C for 30 min in the dark. Fluorescent images were captured by fluorescence microscopy to assess ROS levels.

TUNEL staining

Spinal cord sections were obtained from mice 7 days post-SCI and fixed in 4% paraformaldehyde for 15 min. After thorough rinsing, permeabilization was performed using 0.3% Triton X-100 for 15 min. Following PBS washes, sections were processed with the TUNEL staining kit (Beyotime, Shanghai, China) in accordance with the manufacturer’s protocol. Fluorescent signals were captured using a fluorescence microscope, and TUNEL-positive cells were quantified using ImageJ.

snRNA-seq and analysis

Spinal cord tissues from SCI mice, with or without AEV treatment, were harvested on day 7 postinjury. Nuclei were isolated via IGEPAL CA-630 (Sigma-Aldrich) lysis and filtration, followed by viability assessment using fluorescent AO/PI staining. Subsequent encapsulation and reverse transcription were performed using the SeekOne DD Single Cell 3′ Library Kit (SeekGene) according to the standard protocol. Library construction and sequencing were carried out on the Illumina NovaSeq X Plus platform with Paired-End 150 bp (PE150) read length.

Sequencing data were processed using the Seurat R package. Nuclei with <200 detected genes or >20% mitochondrial gene content, as well as doublets identified through DoubletFinder, were excluded. Gene expression matrices underwent log normalization and scaling based on the top 2000 variable genes, followed by data integration using anchor-based methods. Clustering was conducted using the top 30 principal components with Uniform Manifold Approximation and Projection (UMAP) at a resolution of 0.3. Cell type annotation relied on canonical markers and the SingleR package. Differential expression of ferroptosis-related genes within the neuronal cluster was assessed using the FindMarkers function.

Statistical analysis

Statistical analyses were conducted using GraphPad Prism (version 8.0.2). Data are expressed as mean ± SD. Group comparisons were evaluated using an unpaired two-tailed Student’s t test for two-group comparisons. One-way analysis of variance (ANOVA) with Tukey’s post hoc test was applied for analyses involving three or more groups. In experiments assessing variables across multiple conditions (e.g., BMS scores), two-way ANOVA followed by Bonferroni correction was used. Statistical significance was defined as P < 0.05.

Supplementary Materials

The PDF file includes:

Figs. S1 to S11

Tables S1 to S4

Legend for data S1

Raw western blot images

Other Supplementary Material for this manuscript includes the following:

Data S1