Angiopep-2-engineered extracellular vesicles derived from spinal cord injury-responsive brown adipose tissue for targeted neuroinflammation modulation

Abstract

Spinal cord injury (SCI) affects over 15 million people globally with no curative treatments available. While primary mechanical trauma initiates tissue damage, secondary pathological cascades including neuroinflammation and mitochondrial dysfunction expand the injury. Current clinical interventions focus only on symptom management without restoring neural function. This study investigates the role of brown adipose tissue (BAT) in SCI repair through extracellular vesicles (EVs). Using ¹⁸F-FDG PET/CT imaging, this study discovered significant BAT activation post-SCI, peaking at 7 days, which was confirmed by histological analysis. Activated BAT increased EV secretion, with secreted EVs being selectively taken up by spinal microglia. miRNA sequencing identified miR-692 as the key EV cargo that silenced pro-inflammatory Spp1 gene in microglia, promoting their anti-inflammatory polarization and enhancing neuronal survival. Further development of a targeted delivery system using Angiopep2-modified BAT-EVs encapsulated in GelMA hydrogel for sustained release at injury sites significantly reduced lesion volume and improved functional recovery. The research establishes the BAT-EV-microglia axis as crucial for SCI repair and presents a promising biomaterial-enhanced EV therapy for SCI treatment, marking a significant advancement in regenerative medicine.

Graphic abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04163-1.

Introduction

Spinal cord injury (SCI) is a devastating neurological disorder affecting over 15,000,000 people worldwide, with no curative treatments available [1, 2]. Primary mechanical trauma initiates secondary pathological cascade-including neuroinflammation, glutamate excitotoxicity, mitochondrial dysfunction, and glial scar formation. These processes exacerbate tissue damage and inhibit endogenous repair mechanisms [3, 4]. Current clinical interventions (e.g., surgical stabilization, methylprednisolone pulse therapy) focus on symptom management but fail to restore lost neural function [5]. This therapeutic void underscores the critical need for strategies targeting the dynamic interplay between systemic metabolism, inflammation, and neural regeneration.

Conventional imaging (MRI, CT) provides structural insights but lacks sensitivity to real-time metabolic and inflammatory changes during secondary injury [6]. Positron emission tomography–computed tomography (PET/CT) revolutionizes this paradigm by quantifying glucose metabolism via ¹⁸F-fluorodeoxyglucose (FDG) uptake, enabling non-invasive mapping of hypermetabolic niches in neural and peripheral tissues [7, 8]. In current SCI study, FDG-PET/CT reveals elevated metabolic activity in spared spinal segments, brainstem nuclei, and visceral organs, reflecting neuroinflammation-driven energy demands and compensatory plasticity [9–11]. This technology offers unprecedented potential for prognostication, therapy monitoring, and mechanistic discovery, positioning it as a cornerstone for modern SCI research.

Beyond thermogenesis, brown adipose tissue (BAT) functions as a potent immunomodulatory organ [12, 13]. Activated BAT secretes anti-inflammatory cytokines (e.g., IL-10, TGF-β), catecholamines, and lipid mediators that suppress systemic inflammation—a role implicated in sepsis, rheumatoid arthritis, and cardiovascular disorders [14–17]. In SCI, sympathetic hyperactivity triggers BAT activation, suggesting an adaptive response to mitigate neural inflammation [18]. Parallels in Alzheimer’s disease, where BAT activation correlates with improved Alzheimer’s disease outcomes, further support its therapeutic relevance [19]. Despite this, BAT’s crosstalk with the injured spinal cord remains poorly defined, presenting a key knowledge gap.

Extracellular vesicles (EVs) are nanoscale lipid carriers released by all cell types, playing a pivotal role in orchestrating interorgan communication. These sophisticated biological entities achieve their function by transferring functional miRNAs, proteins, and metabolites, enabling intricate cross-talk between cells and tissues [20, 21]. In regenerative contexts, EVs from stem cells or macrophages promote repair in myocardial infarction, acute kidney injury, and liver fibrosis by modulating immune responses, angiogenesis, and extracellular matrix remodeling [22, 23]. For SCI, EVs from various cell and tissue sources have been proven to promote the recovery of SCI by transferring bioactive substances. The specific mechanisms include promoting nerve regeneration, reducing inflammation and other pathways [24–26]. These highly biocompatible vesicles have brought new breakthroughs to the field of regenerative medicine for SCI.

In this study, we unveil BAT-derived EVs as pivotal mediators of SCI repair and develop a targeted delivery platform for clinical translation. Using longitudinal ¹⁸FDG-PET/CT, we identified robust BAT activation post-SCI, evidenced by elevated FDG uptake peaking at 7 days post SCI. Upon activation, BAT concomitantly boosted EV secretion, and experimental evidence has illustrated that spinal microglia selectively take up BAT-EVs. Integrated miRNA sequencing and bioinformatics analyses identified miR-692 as the dominant EV cargo. This miRNA silences the pro-inflammatory gene Spp1 (secreted phosphoprotein 1) in microglia, driving their anti-inflammatory polarization and enhancing neuronal survival. Crucially, BAT ablation impaired locomotor recovery, validating its endogenous role. To translate this biology, we engineered Angiopep2-modified BAT-EVs encapsulated in GelMA hydrogel. This system achieved sustained, targeted EV delivery to SCI epicenters, reducing lesion volume and improving neural functional recovery. This work delineates a BAT-EV-microglia axis essential for neural repair and pioneers a biomaterial-enhanced EV therapy poised for preclinical development.

Results

SCI triggers BAT activation enhancing extracellular vesicles biogenesis

Based on the T10 contusion spinal cord injury mouse model, PET/CT technology was employed to detect the uptake and distribution of ¹⁸F-FDG at 1, 3, 7, 14, and 28 days post-injury (dpi). Notably, ¹⁸F-FDG PET/CT images demonstrated significant enrichment in the brown adipose tissue (BAT) of the mouse dorsum after injury, peaking at day 7, indicating that spinal cord injury induces BAT activation (Fig. 1A and S1A-C). To further clarify the effect of spinal cord injury on BAT, BAT tissues were harvested at specific time points post-injury, and histology was used to characterize BAT activation-related markers. HE staining revealed a significant reduction in lipid droplet size in BAT after spinal cord injury, a hallmark of BAT activation (Fig. 1C). Additionally, BAT activation is often accompanied by increased vascular density, sympathetic innervation, and elevated UCP1 expression. Immunofluorescence staining showed that spinal cord injury induced increases in vascular and nerve density in BAT (Fig. 1D-E and G-H), as well as heightened UCP1 expression (Fig. 1F and I), with a peak at day 7, consistent with the PET/CT observations. Studies have shown that the activation of BAT extends beyond its thermogenic effects and is often accompanied by increased synthesis of numerous bioactive factors which could be released via extracellular vesicles [27, 28]. Detection by Western blotting (WB) revealed that with the expression level of UCP1 (a marker of BAT activation) increased, the expression of markers related to EVs generation, such as RAB27, lysosome-associated membrane protein 2 (LAMP2), and CD63, also increased (Fig. 1J-K). This result suggests that the BAT activation phenomenon after SCI may be related to the increased synthesis and release of BAT-derived EVs, indicating that BAT may play a potential role in the pathophysiological process after spinal cord injury by releasing relevant mediators.

Fig. 1.

SCI triggers BAT activation enhancing extracellular vesicles biogenesis. (A) Represent PET/CT images showing the ¹⁸F-FDG distribution in BATs of naïve mice (NC), sham operated mice and SCI mice at 1, 3, 7, 14, and 28 days post-injury (dpi). (B) Represent H&E staining of BAT tissues from NC mice, sham operated mice and SCI mice at 1, 3, 7, 14, and 28 dpi. (C) Representative immunofluorescent stains of the endothelial cells (CD31, green fluoresce) images of the BATs from NC mice, sham operated mice and SCI mice at 1, 3, 7, 14, and 28 dpi. Scale bar, 100 μm. (D) Representative immunofluorescent stains of the sympathetic nerves (TH, red fluoresce) images of the BATs from NC mice, sham operated mice and SCI mice at 1, 3, 7, 14, and 28 dpi. Scale bar, 100 μm. (E) Representative immunofluorescent stains of the UCP1 (green fluoresce) images of the BATs from NC mice, sham operated mice and SCI mice at 1, 3, 7, 14, and 28 dpi. Scale bar, 100 μm. (F) Quantification of CD31⁺ area of groups in (C) (one-way ANOVA, F [6, 28] = 99.80, P < 0.001. Tukey’s post hoc test. n = 5 per group). (G) Quantification of TH⁺ area of groups in (D) (one-way ANOVA, F (6, 28) = 83.51, P < 0.001. Tukey’s post hoc test. n = 5 per group). (H) Quantification of UCP1⁺ signaling intensity of groups in (E) (one-way ANOVA, F (6, 28) = 324.8, P < 0.001. Tukey’s post hoc test. n = 5 per group). (I) Western blotting analysis of the expression of UCP1 and EVs biogenesis markers, including CD9, LAMP2, and RAB27 of the BATs from NC mice, sham operated mice and SCI mice at 1, 3, 7, 14, and 28 dpi. (J) Quantification of UCP1, CD9, LAMP2, and RAB27 expression in (I) (two-way ANOVA, F (6, 56) = 314.2, p < 0.001. Tukey’s post hoc test. n = 3 per group). Data are presented as mean ± SD, NS, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001

BAT resection significantly inhibited the recovery of neurological function in mice with spinal cord injury

To elucidate the role and mechanism of BAT in neural regeneration and repair following SCI in mice, the BATs were removed on SCI mice (BAT-RM) (Figure S2). Systematical evaluation of post-injury neurological functional recovery and spinal cord regeneration were performed (Fig. 2A). The Basso Mouse Scale (BMS) scores further indicated that BAT-RM mice only recovered to approximately BMS 2, demonstrated that mice subjected to BAT resection exhibited significantly impaired motor function after SCI (Fig. 2B). Corresponding gait analysis revealed that BAT-RM mice showed markedly reduced weight support capability and severe loss/dysfunction of hindlimb motor control, which was characteristic by reduced range of motion in the hip, knee, and ankle joints, alongside decreased vertical amplitude of crest displacement (Fig. 2C-G). Additionally, motor evoked potential (MEP) measurements showed further decreased amplitude and prolonged latency in BAT-RM mice, indicating weakened neural connectivity in the injured spinal cord (Fig. 2H-J). Next, we obtained the spinal cord tissues from the previously grouped SCI mice and stained the spinal cord tissue sections using immunofluorescence technology. We quantified the injured area of the spinal cord injury tissue based on CSPG and Lamnin staining. The results showed that obvious fibrous scar tissue hyperplasia was observed after spinal cord injury, while the injured area was significantly enlarged after BAT resection (Fig. 2K-M). Meanwhile, we used neurofilament to label the nerve fiber regeneration in the injured area. The results showed that BAT resection significantly inhibited the nerve ingrowth into the center of the injured area (Fig. 2N-O). In summary, the above results indicate that BAT plays a crucial role in spinal cord injury repair, and BAT resection blocks the endogenous regeneration process of spinal cord injury tissue.

Fig. 2.

BAT removal significantly inhibited the recovery of neurological function in mice with spinal cord injury. (A) The Schematic of the brown adipose tissue removal process and SCI model establishment. (B) BMS scores in Sham, SCI and SCI + BAT-RM groups at specific time points post-SCI (two-way ANOVA, F (2, 120) = 11824, P < 0.001. Tukey’s post hoc test. n = 6 per group). * SCI vs SCI + BAT-RM group. (C) Color-coded stick views and angle degree curves from Sham, SCI and SCI + BAT-RM groups. (D) Quantification of crest height amplitude of mice in Sham, SCI and SCI + BAT-RM groups (one-way ANOVA, F (2, 15) = 41.89, P < 0.001. Tukey’s post hoc test. n = 6 per group). (E) Quantification of hip angle amplitude of mice in Sham, SCI and SCI + BAT-RM groups (one-way ANOVA, F (2, 15) = 130.0, P < 0.001. Tukey’s post hoc test. n = 6 per group). (F) Quantification of knee angle amplitude of mice in Sham, SCI and SCI + BAT-RM groups (one-way ANOVA, F (2, 15) = 579.8, P < 0.001. Tukey’s post hoc test. n = 6 per group). (G) Quantification of ankle angle amplitude of mice in Sham, SCI and SCI + BAT-RM groups (one-way ANOVA, F (2, 15) = 403.8, P < 0.001. Tukey’s post hoc test. n = 6 per group). (H) Representative electrophysiological trace images recorded in Sham, SCI and SCI + BAT-RM groups at 4 weeks post-SCI. (I) Measurement of the MEP latent period in (H) (one-way ANOVA, F (2, 15) = 337.1, P < 0.001. Tukey’s post hoc test. n = 6 per group). (J) Measurement of the MEP amplitude in (H) (one-way ANOVA, F (2, 15) = 972.6, P < 0.001. Tukey’s post hoc test. n = 6 per group). (K) Representative immunofluorescent stains of the scar tissue (CSPG, green, Laminin, Red) images of the spinal cord at 4 weeks post-injury. Scale bar, 500 μm. (L) Quantification of CSPG positive signals in the epicenter of Sham, SCI and SCI + BAT-RM groups in (K) (one-way ANOVA, F (2, 15) = 308.0, P < 0.001. Tukey’s post hoc test. n = 6 per group). (M) Quantification of Laminin positive signals in the epicenter of Sham, SCI and SCI + BAT-RM groups in (K) (one-way ANOVA, F (2, 15) = 300.5, P < 0.001. Tukey’s post hoc test. n = 6 per group). (N) Representative immunofluorescent stains of Neurofilament (NF, Red) images of the spinal cord at 4 weeks post-injury. Scale bar, 500μm. (O) Quantification of NF positive signals in the epicenter of Sham, SCI and SCI + BAT-RM groups in (N) (one-way ANOVA, F (2, 15) = 195.3, P < 0.001. Tukey’s post hoc test. n = 6 per group). Data are presented as mean ± SD, NS, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001

BAT-derived EVs could be internalized by microglia and exert anti-inflammatory effects after SCI

To delineate the communication pathway between BAT and the injured spinal cord via EVs SCI, we utilized an AAV-CD63-EGFP viral vector injected directly into murine BAT to label BAT-derived EVs. Following SCI induction, we tracked EV cellular uptake in the lesioned spinal cord (Fig. 3A). The results demonstrated successful labeling of adipocytes within brown adipose tissue (BAT) using AAV-CD63-EGFP (Fig. 3B). Notably, EGFP expression was predominantly observed in the adipocyte content of BAT, indicating specific and efficient targeting of these cells by the labeling technique (Figure S3A-B). And, at 7 days post-injury, the results showed that most of EVs were taken up by microglia at the injured spinal cord, while only a small amount of EVs were taken up by endothelial cells, astrocytes, and neurons at the injury site. (Fig. 3C and S4A-D). This indicates that post-SCI, BAT-EVs could mediate repair processes by influencing microglial pathophysiology. Given this BAT-EV-microglia interaction, we further examined neuroinflammation levels in BAT-resected mice. BAT ablation significantly increased CD68⁺ activated inflammatory cells and pro-inflammatory genes expression in spinal cord tissues, suggesting BAT-EVs attenuate neuroinflammation through microglial uptake (Fig. 3D-E and Figure S5A-D). Subsequently, to define the regulatory role of BAT-derived EVs in neuroinflammation, we isolated EVs from BAT tissues of both sham-operated and SCI mice. EV characterization was performed using transmission electron microscopy (TEM), nanoparticle tracking analysis (NTA) and western blot (Fig. 3F-G and S6). We then established a lipopolysaccharide (LPS)-induced primary microglial neuroinflammation model. Treatment with either Sham-EVs or SCI-EVs revealed that SCI-EVs significantly attenuated inflammatory responses in LPS-stimulated microglia, as evidenced by decreased CD86⁺ signaling (Fig. 3H-I). Western blot and flow cytometry analysis of pro-inflammatory (iNOS, CD86) and anti-inflammatory (CD163, Arg-1 and CD206) markers further corroborated these findings, demonstrating consistent suppression of neuroinflammation (Fig. 3J-K and S7A-D). Collectively, these results demonstrate that spinal cord injury triggers BAT to release EVs that are internalized by microglia, where they exert potent anti-neuroinflammatory effects.

Fig. 3.

BAT-derived EVs could be internalized by microglia and exert anti-inflammatory effects after SCI. (A) The Schematic of the pAAV-CMV-Cd63-linker-EGFP-3xFLAG-WPRE transfect procedure by local injection in brown adipose tissue. (B) Representative immunofluorescent pictures showed AAV-CD63-EGFP exhibited labeling of BAT in the injection site. Scale bar, 100 μm. (C) Representative immunofluorescent pictures showed the CD63-EGFP positive EVs spot (green fluoresce) co-labeling with spinal microglia (IBA1+), Scale bar, 100 μm. The bottom right image showed the 3D viewer of the CD63-EGFP+ EVs co-labelling with the microglia in the injury site. (D) Representative immunofluorescent stains of the CD68+ cells in injury spinal cord at 4 weeks post-injury. Scale bar, 500 μm. (E) Quantification of CD68 positive signals in the epicenter of Sham, SCI and SCI + BAT-RM groups in (D) (one-way ANOVA, F (2, 15) = 442.8, P < 0.001. Tukey’s post hoc test. n = 6 per group). (F) TEM image of EVs derived from by the sham and SCI BATs. scale bar, 100 nm. (G) Nanoparticle tracking analysis of EVs derived from by the sham and SCI BATs. (H) Representative immunofluorescent images (IBA1, green fluoresce, CD86, red fluoresces) of the primary microglia of the naive (control), LPS, LPS + Sham-EVs and LPS + SCI-EVs treated groups. Scale bar, 100 μm. (I) Quantification of the relative CD86+ signaling intensity of groups in (H). (one-way ANOVA, F (3, 16) = 100.7, P < 0.001. Tukey’s post hoc test. n = 5 per group). (J) Western blot analysis of the inflammatory markers of iNOS, CD86 and the anti-inflammation markers of CD163, ARG1 in naive (control), LPS, LPS + Sham-EVs and LPS + SCI-EVs treated groups. (K) Quantification of iNOS, CD86, CD163, and ARG1 expression in (J). (two-way ANOVA, F (3, 12) = 29.24, P < 0.001. Tukey’s post hoc test. n = 4 per group). Data are presented as mean ± SD, NS, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001

miR-692 serves as the pivotal mediator for BAT-EVs in negatively regulating neuroinflammation

To elucidate the mechanism by which BAT-EVs cross-regulate microglial activation post- SCI, we isolated EVs from BAT of SCI mice and Sham mice as well as plasma derived EVs from SCI and SCI + BAT-RM mice (Fig. 4A). miRNA sequencing revealed 24 upregulated miRNAs in BAT-EVs from SCI mice and 122 downregulated miRNAs in plasma EVs from BAT-RM SCI mice (Fig. 4B). Among these miRNAs, the miR-455-5p, miR-692, miR-27a-5p, and miR-21a-3p were commonly regulated (Fig. 4C). To identify the key mediator of BAT-EV-driven microglial immunomodulation, we isolated microglia from sham and SCI mice via flow cytometry and assessed miRNA expression by RT-qPCR (Fig. 4D). This demonstrated significant upregulation of miR-692 in SCI-activated microglia, which was downregulated after BAT-RM (Fig. 4E). Critically, precursor miR-692 levels remained unchanged in these microglia, indicating that the elevated mature miR-692 originates exogenously (Figure S8A). Meanwhile, qRT-PCR detection of BATs derived from sham and SCI mice showed that the expression levels of both pri-miR-692 and miR-692 increased with the occurrence of spinal cord injury (Figure S8B). This indicated that SCI microglia internalize BAT-EV-derived miR-692 to orchestrate downstream molecular pathways governing neuroinflammation. Meanwhile, based on the prediction results of the target genes of miR-692, we found that miR-692 can bind to the mRNAs of various inflammation-related factors and regulate their transcription. GO enrichment analysis and KEGG pathway analysis of the target genes showed that the target genes of miR-692 were significantly enriched in some inflammation-related pathways, such as the NF-κB and MAPK pathway (Figure S9A-B). Then, we used miR-692 inhibitor (miR-692^IN) combined with SCI-EVs to intervene in LPS-induced microglial inflammation to evaluate the role of miR-692 in the anti-inflammatory effect mediated by SCI-EVs. Immunofluorescence results showed that the inhibition of miR-692 counteracted the anti-inflammatory effect of SCI-EVs on microglia, as evidenced by the increased expression of the pro-inflammatory microglial transformation marker CD86 (Fig. 4F-G). Western blot and flow cytometry results also showed that the inhibition of miR-692 increased the expression of the pro-inflammatory markers iNOS and CD86 in microglia treated with SCI-EVs, accompanied by a decrease in the expression of the anti-inflammatory transformation markers CD163, CD206 and ARG1 (Fig. 4H-M). The above results indicate that miR-692 is an important bioactive molecule in BAT-derived EVs that plays a key anti-inflammatory role.

Fig. 4.

miR-692 serves as the pivotal mediator for BAT-EVs in negatively regulating neuroinflammation. (A) The Schematic of the miRNA-seq analysis of the EVs derived from BATs and plasma of SCI mice. (B) Volcano plot showed the expression pattern of miRNAs in EVs derived from BATs of sham and SCI mice. (C) Volcano plot showed the expression pattern of miRNAs in EVs derived from BATs of sham and SCI mice. (D) Flow cytometry sorting of the CD11b+CD45low microglia from the sham, SCI and BAT removed SCI mice. (E) Quantification of miR692, miR-455-5p, miR-27a-5p, miR-21a-3p expression in sorted microglia using qRT-PCR (two-way ANOVA, F (2, 24) = 29.06, P < 0.001. Tukey’s post hoc test. n = 3 per group). (F) Representative immunofluorescent images (IBA1, green fluoresce, CD86, red fluoresces) of the primary microglia of the LPS, LPS + SCI-EVs, LPS + SCI-EVs + miR-NC and LPS + SCI-EVs + miR-692IN treated groups. Scale bar, 100 μm. (G) Quantification of the relative CD86+ signaling intensity of groups in (F). (one-way ANOVA, F (3, 16) = 111.5, P < 0.001. Tukey’s post hoc test. n = 5 per group). (H-I) Flow cytometry analysis of the inflammatory marker CD86 and the anti-inflammation marker CD206 in LPS, LPS + SCI-EVs, LPS + SCI-EVs + miR-NC and LPS + SCI-EVs + miR-692IN treated groups. (J) Quantification of CD206+ cells percentage in of groups in (H). (one-way ANOVA, F (3, 8) = 60.35, P < 0.001. Tukey’s post hoc test. n = 3 per group). (K) Quantification of CD86+ cells percentage in of groups in (I). (one-way ANOVA, F (3, 8) = 168.6, P < 0.001. Tukey’s post hoc test. n = 3 per group). (L) Western blot analysis of the inflammatory markers of iNOS, CD86 and the anti-inflammation markers of CD163, ARG1 in LPS, LPS + SCI-EVs, LPS + SCI-EVs + miR-NC and LPS + SCI-EVs + miR-692IN treated groups. (M) Quantification of iNOS, CD86, CD163, and ARG1 expression in (L). (two-way ANOVA, F (3, 12) = 35.07, P < 0.001. Tukey’s post hoc test. n = 4 per group). Data are presented as mean ± SD, NS, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001

BAT-derived EVs exert anti-inflammatory effects in microglia via the miR-692/Spp1 pathway

After spinal cord injury, the active microglia highly express genes related to pro-inflammation. These upregulated pro-inflammatory genes could potentially serve as the critical targets through which BAT-EVs exert their regulatory effects on inflammatory processes. To clarify the potential molecular pathways by which BAT-EVs affect the change in the inflammatory properties of microglia after SCI, we performed an integrative analysis. Specifically, based on the single-cell sequencing data of spinal cord microglia after SCI [29], we integrated the single cell sequencing data of microglia from sham mice and those 7 days after spinal cord injury (the time point when BAT-EVs are most actively secreted). Then, according to the gene expression profiles, we clustered the integrated microglia. The results showed that microglia could be divided into 5 cell clusters (Fig. 5A). Microglia from the sham group were mainly distributed in cluster 1, while those from SCI mice were mainly distributed in clusters 2–5 (Fig. 5B). Among them, microglia in cluster 1 mainly expressed microglia markers related to homeostasis, such as P2ry12 and Tmem119, while microglia in clusters 2–5 showed a decrease in the expression of homeostatic microglia markers. Microglia in cluster 2 were the most distinctive, showing an increase in the expression of genes such as Spp1, Lpl, and Lgals, which are markers of activated microglia (Fig. 5C-D). Meanwhile, the proportion of corresponding cells in this cluster increased most significantly after spinal cord injury (Fig. 5E). By comparing the gene expression differences between MG-Homeo and MG-active1, the results showed that the relevant differentially expressed genes were significantly enriched in neural death and neuroinflammatory response (Fig. 5F). To further clarify the potential molecular targets of miR-692, we obtained a dataset of target genes of miR-692 based on databases (Targetscan, microT and miRDB) [30–32] and intersected it with the gene-set upregulated in microglia after SCI (Fig. 5G). The results showed that there were 3 target genes that might be the potential sites through which BAT-EVs regulate microglia (Fig. 5H). Then, by constructing an in vitro cell model, we directly used miR-692 to intervene in LPS-induced microglia. The results showed that only the expression of Spp1 was inhibited by miR-692, suggesting that Spp1 is a key regulatory site of miR-692 (Fig. 5I). To clarify the binding relationship between Spp1 and miR-692, based on the dual-luciferase assay, we mutated the gene site where Spp1 binds to miR-692. The results showed that after mutation, miR-692 lost its effect on regulating the Rluc/Luc fluorescence ratio, suggesting that miR-692 has the ability to target and bind to Spp1 (Fig. 5J). In addition, based on the in vitro experiments, we also found that miR-692 has a certain anti-inflammatory effect, and the anti-inflammatory effect of SCI-EVs can be counteracted by the intervention of SPP1 recombinant protein (Fig. 5K-N and S10). To ensure that miR-692 also has an anti-inflammatory effect in vivo, we adopted local injection of miR-692 inhibitor combined with BAT-EVs to investigate its regulatory effect on SPP1. The results showed that the injection of miR-692 inhibitor further increased the expression of SPP1 in the injury area. Meanwhile, adding the injection of miR-692 inhibitor on the basis of BAT-EVs injection would counteract the inhibitory effect of BAT-EVs on SPP1 (Figure S11). These results collectively demonstrate that BAT-derived EVs deliver miR-692 to suppress Spp1 expression, thereby attenuating neuroinflammation.

Fig. 5.

BAT-derived EVs exert anti-inflammatory effects in microglia via the miR-692/Spp1 pathway. (A) Uniform Manifold Approximation and Projection (UMAP) map of the cells from mice spinal cord. Colors indicate different cell types. (B) UMAP map of the cells from mice spinal cord. Colors indicate the different original of cells. (C) The expression level of markers of cell type annotations in different clusters. (D) The expression level of pro-inflammatory or anti-inflammatory genes in different clusters. (E) The cell fraction of either the sham or SCI group as a proportion of the total analyzed cells. (F) GSEA analysis of the upregulated genes in MG-active1 cluster. (G) Volcano plot showed the expression pattern of genes in MG-active1 and MG-homeo cluster. (H) Venn diagram shows the potential target gene for miR-692. (I) Quantification of Spp1 expression in microglia treated with PBS, LPS or LPS+miR-692. (two-way ANOVA, F (2, 18) = 149.3, P < 0.001. Tukey’s post hoc test. n = 3 per group) (J) Complementary sequences between miR-692 and the 3′UTR of Spp1 and relative luciferase activities of the Spp1-wild type (WT) + negative control (NC) group, Spp1-WT + miR-692 group, Spp1-MUT + NC group, and Spp1-MUT + miR-692 group. (one-way ANOVA, F (3, 8) = 25.60, P < 0.001. Tukey’s post hoc test. n = 3 per group). (K) Western blot analysis of the inflammatory markers of iNOS, CD86 and the anti-inflammation markers of CD163, ARG1 in naive (NC), LPS, LPS + SCI-EVs, LPS + miR-692, LPS + SCI-EVs + SPP1 treated groups. (L) Quantification of iNOS, CD86, CD163, and ARG1 expression in (K). (two-way ANOVA, F (4, 10) = 154.4, P < 0.001. Tukey’s post hoc test. n = 3 per group). (M) Representative immunofluorescent images (IBA1, green fluoresce, CD86, red fluoresces) of the primary microglia of the naive (NC), LPS, LPS + SCI-EVs, LPS + miR-692, LPS + SCI-EVs + SPP1 treated groups. Scale bar, 100 μm. (N) Quantification of the relative CD86+ signaling intensity of groups in (M). (one-way ANOVA, F (4, 20) = 193.4, P < 0.001. Tukey’s post hoc test. n = 5 per group). Data are presented as mean ± SD, NS, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001

Construction of the hydrogel based targeted-delivery strategy for BAT derived EVs therapy

Above all, we have uncovered a novel mechanism rooted in the communication between BAT and SCI repair, which holds the potential to offer innovative strategies for SCI treatment. However, the abundance of adult BAT is exceedingly low [33]. Based on current clinical observations and experimental data, although BAT is activated following SCI, its therapeutic efficacy remains limited [18]. Moreover, the efficiency of utilizing circulating EVs to cross the blood-spinal cord barrier is relatively modest [34]. Consequently, we propose that transforming this inter-organ communication mechanism into a definitive and clinically viable treatment modality will necessitate the integration of specific tissue engineering approaches for targeted delivery, thereby enhancing therapeutic efficacy. Therefore, we employed a GelMA-based hydrogel EVs sustained-release system for the treatment of spinal cord injury. Specifically, by covering the injured spinal cord of mice with a hydrogel containing a photoinitiator and irradiating it with a 400 nm laser for curing, the hydrogel was uniformly attached to the spinal cord injury area to achieve a sustained-release effect (Fig. 6A). Electron microscopy results showed that the hydrogel constructed based on GelMA cross-linking had a loose and porous microstructure, which was conducive to the attachment of EVs and the achievement of the EVs sustained-release effect (Fig. 6B). PKH67 was used to label SCI-EVs, and an EVs-hydrogel sustained-release system was constructed. The results showed that PKH67-labeled EVs were uniformly present in the hydrogel (Fig. 6C). Through the swelling coefficient test, we found that the addition of EVs had no significant effect on the swelling of the hydrogel (Fig. 6D). Further evaluation of the mechanical properties of the hydrogel system revealed that the G’ of the 10 wt% GelMA group was approximately 1 kPa, which was close to the modulus of the natural spinal cord (Fig. 6E). From the elastic modulus results, although the addition of EVs reduced the strength of the scaffold, the scaffold still showed good ability to resist external force deformation (Fig. 6F). The degradation process of the EVs hydrogel showed an obvious time - dependent pattern, and the degradation rate could reach approximately 60% at 4 weeks (Fig. 6G). The EVs hydrogel system exhibited typical two-stage continuous sustained-release characteristics: within 0–3 days, 30% of the cumulative release of EVs was achieved through rapid diffusion in the pores, and after 3 days, it entered the slow- release stage (Fig. 6H and S12A). To further clarify that this hydrogel encapsulation method has no impact on the structural integrity of EVs, we performed TEM analysis on these slowly released EVs. The results showed that these slowly released EVs could still maintain the intact classic cup-shaped vesicle structure (Figure S12B). Based on the co-culture system of EVs hydrogel and neuronal cells, the viability of neuronal cells under co-culture conditions was detected. The results showed that neither the addition of EVs nor GelMA had an impact on the viability of nerve cells, indicating that the sustained-release system had good biocompatibility (Fig. 6I). To clarify the in vivo release characteristics of this hydrogel sustained-release system, in vivo imaging analysis showed that: The freely injected EVs, administered above the injured spinal cord, exhibited extensive diffusion into the surrounding lesion areas within 24 h. And the fluorescence signal in the injured region gradually diminished, eventually returning to background levels after 14 days. In contrast, the hydrogel-encapsulated EVs demonstrated a sustained and robust fluorescence signal within the injury site over the same period. This persistent signal retention provided compelling evidence of their long-acting, controlled-release capability. (Fig. 6J).

Fig. 6.

Construction of the hydrogel based targeted-delivery strategy for BAT derived EVs therapy. (A) Schematic diagram of the establishment of a hydrogel-EVs sustained-release system based on photocuring and its application in the treatment of spinal cord injury. (B) Scanning electron microscope (SEM) images showed the microstructures of the hydrogel. Scale bar: 20 μm. (C) Confocal multilayer scanning and 3D reconstruction of PKH67 labelled EVs retained within the hydrogel. Scale bar, 20 μm. (D) Swelling ratios of hydrogels over time. N = 4 per group. (E) Rheological curves of 10 wt% GelMA and 10 wt% GelMA + EVs scaffold. G' represents the storage modulus, and G" represents the loss modulus. (F) The figure shows the stress-strain curves of the 10 wt% GelMA and 10 wt% GelMA + EVs scaffold samples. (G) In vitro degradation curves of hydrogels over time based on gravimetric measurements. N = 3 per group. (H) Release curves of Exosomes at different time points. N = 3 per group. (I) The survival and cell viability of neurons co-incubated with hydrogel, which were detected by the CCK-8 assay. N = 3 per group. (J) In vivo tracing of the distribution of DiR-labeled EVs embedded in hydrogels in the injured spinal cord at 1 day and 14 days post-SCI. (K) The schematic diagram showing the construction of engineered extracellular vesicles with the abilities of crossing the blood-brain barrier by co-incubating DSPE-PEG2000-Angiopep2 with EVs. (L) Representative immunofluorescence image of uptake of EV (PKH67 marked, green) by microglia (IBA1 marked, red) at 7 dpi. The nucleus was stained with DAPI (blue). Scale bar, 500 μm. Data are presented as mean ± SD, NS, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001

Although SCI is often accompanied by disruption of the blood-spinal cord barrier (BSCB), the remaining barrier structure still restricts the penetration of macromolecules. Given that simple dressing-based drug delivery makes it difficult to effectively deliver EVs to the injury area, this study introduced an Angiopep2 penetrating peptide modification strategy to enhance the barrier penetration ability of EVs [35, 36]. The modification mediated by DSPE-PEG2000 was validated through flow cytometry, utilizing DSPE-PEG2000 labeled with FITC to assess the binding efficiency to EVs. Our findings indicate that when DSPE-PEG2000-FITC is mixed with EV particles at a weight ratio of 1 : 10, approximately 90% of the EV particles were successfully labeled with DSPE-PEG2000-FITC, exhibiting clear FITC-positive signals (Figure S13A). By synthesizing the amphiphilic block copolymer DSPE-PEG2000-Angiopep2 and co-incubating it with SCI-EVs (ANG-SCI-EVs), Angiopep2-EVs were prepared and integrated into a hydrogel delivery system (Fig. 6K). The analysis results based on TEM, NTA, and Zeta potential indicated that, except for a slight increase in the median particle size of the vesicles, the modification of Angiopep-2 did not have a significant impact on the basic properties of EVs (Figure S13B-D). From a macroscopic perspective, although the sustained-release effect of the hydrogel and the strong positive signals within the hydrogel have interfered with our observation of the uptake of EVs by the injured spinal cord, and in vivo imaging data indicate that modification with Angiopep-2 did not enhance the enrichment of EVs in the injured area (Figure S14A-B). However, at the cellular level, a large number of Angiopep2-modified EVs were taken up by microglia, which fully demonstrates the excellent targeting performance of angiopep2-modified EVs (Fig. 6L and S14C). To ensure the biosafety of this hydrogel loaded with angiopep2-modified EVs, we implanted GelMA loaded with ANG-SCI-EVs into the subcutaneous tissue of normal mice. HE staining results confirmed that ANG-SCI-EVs did not induce cardiotoxicity, hepatotoxicity, spleen cell toxicity, pulmonary toxicity, or nephrotoxicity after 28-day intravenous injection of ANG-SCI-EVs, indicating favorable biosafety (Figure S15A-E). These results indicated that this hydrogel-based targeted delivery strategy for EVs derived from BATs could provide an efficient and safe drug delivery method for SCI.

ANG-SCI-EVs significantly mitigate neuroinflammation and promote functional recovery post-spinal cord injury

By constructing a spinal cord injury model, EVs derived from BAT and EVs modified with Angiopep-2 were used to intervene in the mouse model of SCI respectively based on the GelMA embedded strategy, and their therapeutic effects were evaluated. The results showed that the intervention of SCI-EVs significantly promoted the recovery of hind-limb motor function in mice with SCI, manifested by significant recovery in the coordination of hind-limb movement and weight-bearing ability (Fig. 7A-E). Meanwhile, the results of the BMS score also showed that the intervention of SCI-EVs promoted the recovery of motor function in mice with spinal cord injury (final BMS ≥ 4), and the modification of Angiopep2 further promoted the recovery of motor function in mice with SCI (Fig. 7F). Electrophysiological results also indicated that the modification of Angiopep2 more significantly promoted the connectivity of spinal nerve connections in mice with SCI, manifested by a higher voltage peak and a shorter latency (Fig. 7G-I). The spinal cord tissues of each group of mice were collected for immunofluorescence detection to quantitatively evaluate the inflammation level, scar area, and nerve regeneration. The results showed that SCI-EVs intervention significantly reduced the neuroinflammation in mice with SCI (decreased CD68⁺ expression), while ANG-SCI-EVs further inhibited the inflammation level in the injured area (Fig. 7J and M and S16A-D). Similarly, the scar area in the injured area of the ANG-SCI-EVs group was significantly reduced (Fig. 7K and N-O), accompanied by a significant increase in regenerated nerve axons in the injury center (increased NF fluorescence signal) (Fig. 7L and P). In conclusion, the Angiopep2 modification combined with the BAT-EVs hydrogel sustained-release system synergistically alleviates neuroinflammation and promotes nerve regeneration and repair by enhancing targeted delivery, providing a new targeted treatment strategy for SCI.

Fig. 7.

ANG-SCI-EVs significantly mitigate neuroinflammation and promote functional recovery post-spinal cord injury. (A) Color-coded stick views and angle degree curves from Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups. (B) Quantification of crest height amplitude of mice in Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups (one-way ANOVA, F (4, 25) = 55.37, P < 0.001. Tukey’s post hoc test. n = 6 per group). (C) Quantification of hip angle amplitude of mice in Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups (one-way ANOVA, F (4, 25) = 113.9, P < 0.001. Tukey’s post hoc test. n = 6 per group). (D) Quantification of knee angle amplitude of mice in Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups (one-way ANOVA, F (4, 25) = 113.9, P < 0.001. Tukey’s post hoc test. n = 6 per group). (E) Quantification of ankle angle amplitude of mice in Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups (one-way ANOVA, F (4, 25) = 50.42, P < 0.001. Tukey’s post hoc test. n = 6 per group). (F) BMS scores in Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups at specific time points post-SCI (two-way ANOVA, F (4, 192) = 4806, P < 0.001. Tukey’s post hoc test. n = 6 per group). (G) Representative electrophysiological trace images recorded in Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups at 4 weeks post-SCI. (H) Measurement of the MEP latent period in (G) (one-way ANOVA, F (4, 25) = 97.37, P < 0.001. Tukey’s post hoc test. n = 6 per group). (I) Measurement of the MEP amplitude in (G) (one-way ANOVA, F (4, 25) = 561.0, P < 0.001. Tukey’s post hoc test. n = 6 per group). (J) Representative immunofluorescent stains of the CD68+ cells in injury spinal cord at 4 weeks post-injury. Scale bar, 500 μm. (K) Representative immunofluorescent stains of the scar tissue (CSPG, green, Laminin, Red) images of the spinal cord at 4 weeks post-injury. Scale bar, 500 μm. (L) Representative immunofluorescent stains of Neurofilament (NF, Red) images of the spinal cord at 4 weeks post-injury. Scale bar, 500 μm. (M) Quantification of CD68 positive signals in the epicenter of Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups in (J) (one-way ANOVA, F (4, 25) = 227.5, P < 0.001. Tukey’s post hoc test. n = 6 per group). (N) Quantification of Laminin positive signals in the epicenter of Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups in (K) (one-way ANOVA, F (4, 25) = 40.19, P < 0.001. Tukey’s post hoc test. n = 6 per group). (O) Quantification of CSPG positive signals in the epicenter of Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups in (K) (one-way ANOVA, F (4, 25) = 75.34, P < 0.001. Tukey’s post hoc test. n = 6 per group). (P) Quantification of NF positive signals in the epicenter of Sham, SCI, SCI + GelMA, SCI-EVs and ANG-SCI-EVs groups in (L) (one-way ANOVA, F (4, 25) = 1829, P < 0.001. Tukey’s post hoc test. n = 6 per group). Data are presented as mean ± SD, NS, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001

Discussion

This study establishes a fundamental paradigm in spinal cord injury (SCI) repair: the activation of brown adipose tissue (BAT) and its secreted extracellular vesicles (EVs) constitute an endogenous neuroprotective axis that orchestrates microglial reprogramming and functional recovery. By leveraging multimodal approaches—from PET/CT metabolic imaging to engineered EV therapeutics—we demonstrate that SCI triggers profound BAT activation, which in turn releases miR-692-enriched EVs to suppress pro-inflammatory signaling in spinal microglia via the Spp1 pathway. The therapeutic translation of this mechanism using Angiopep2-functionalized, GelMA-encapsulated BAT-EVs significantly enhanced neuroregeneration, offering a blueprint for biologically inspired nanotherapies.

Our discovery that SCI induces BAT hypermetabolism (peaking at 7 days) and subsequent EV secretion expands the understanding of systemic responses to neural injury. BAT activation, quantified via FDG-PET/CT and corroborated by histochemical analyses, aligns with emerging evidence of adipose-neural crosstalk. Recent works revealed that dysregulated sympathetic outflow, driving β3-adrenergic receptor-mediated BAT thermogenesis—a compensatory mechanism that may mitigate secondary inflammation [37]. This parallels our finding that BAT ablation exacerbates functional deficits, underscoring its non-redundant role. Notably, BAT’s temporal activation profile coincides with the transition from acute neuroinflammation to chronic fibrotic scarring, suggesting its EVs may interrupt this maladaptive progression. It indicated that BAT not merely as an energy reservoir but as an active immunomodulator, akin to its role in dampening systemic inflammation in atherosclerosis, where BAT-derived factors inhibit NLRP3 inflammasome activation [38].

EVs derived from BAT play pivotal roles in diverse diseases. Upon cold exposure, BAT-EVs enriched with miR-378a-3p enhance hepatic gluconeogenesis by targeting p110α in the liver, preventing cold-induced hypoglycemia [28]. Meanwhile, accumulating evidence has demonstrated that exercise can expand BAT mass and stimulate BAT to secrete EVs enriched with miR-125b-5p, miR-128-3p, and miR-30d-5p into the circulation. These EVs then target and inhibit the mitogen-activated protein kinase (MAPK) pathway in cardiomyocytes, which significantly reduces cardiomyocyte apoptosis following ischemia-reperfusion injury [39]. The core of this work lies in identifying BAT-EVs as regulators for spinal microglia polarization. To elucidate the specific miRNA mediators underlying the regulatory function of BAT-EVs in spinal microglia, we first performed miRNA sequencing. Through this analysis, we identified multiple miRNAs that were synchronously upregulated in both peripheral blood and BAT-EVs. Subsequent brown fat ablation experiments were conducted to validate these candidate molecules, revealing that only miR-692 exhibited significant expression fluctuations in microglia. Furthermore, detection of miRNA precursors demonstrated that the endogenous synthesis of miR-692 in microglia remained unchanged regardless of spinal cord injury or BAT ablation. Collectively, these findings indicate that miR-692 in spinal microglia is primarily derived from BAT-EVs, rather than being produced via autonomous transcriptional regulation within the target cells. This observation thus justifies our focused investigation on miR-692 in the present study. Meanwhile it is worth noting that other co-upregulated miRNAs may exert distinct regulatory functions at alternative levels, such as modulating the activity of other immune cell subsets or neuronal populations, which merits further in-depth exploration in future studies.

miRNAs are small non-coding regulatory RNAs that bind to the complementary sequence of the 3-untranslated region (3′UTR) of target mRNAs and can inhibit translation or lead to degradation of target mRNAs [40]. This study confirms that miR-692 serves as a key miRNA molecule exerting core regulatory functions in BAT-EVs. To identify the downstream therapeutic targets mediating its anti-inflammatory effects, we integrated the inherent anti-inflammatory properties of miR-692 with a single-cell sequencing database of spinal cord injury models. Specifically, we first screened for microglial subpopulations that undergo a pro-inflammatory phenotypic transition post-injury, and further extracted the gene set highly expressed in this specific subpopulation following injury-these genes represent the core targets governing inflammatory activation of microglia. On this basis, we performed an intersection analysis of the predicted target genes of miR-692 and the prementioned microglial high-expression gene set, ultimately identifying Spp1 as the critical downstream target through which miR-692. Given that SPP1 is a classic pro-inflammatory marker that functions in multiple canonical pro-inflammatory signaling pathways-such as activating the NF-κB pathway and upregulating the expression of prototypical pro-inflammatory proteins including CD86 and iNOS [41, 42]. Notably, the elevated expression of Spp1-associated pro-inflammatory genes drives pro-inflammatory microglia to secrete a repertoire of pro-inflammatory cytokines and chemokines (e.g., CCL2, CX3CL1, IL-1β, TNF-α) [43, 44]. These factors further recruit and activate myeloid-derived inflammatory cells (e.g., monocytes and macrophages), amplifying the inflammatory cascade and exacerbating central nervous system inflammation. Moreover, the identification of SPP1 as a therapeutic target of BAT-EVs sheds light on another key observation of our study: BAT-EVs exert a potent anti-scarring effect. Since SPP1 has been well-documented to promote scar formation and tissue fibrosis [45, 46], the miR-692/Spp1 regulatory axis establishes a mechanistic link that bridges the observed phenotypic changes to the underlying molecular events.

Currently, native EVs face challenges such as rapid clearance, off-target biodistribution, and inconsistent cargo loading. Systemically delivered EVs (via intravenous injection, oral administration, intraperitoneal injection etc.) may lead to the accumulation of EVs in non-injured areas, or be rapidly excreted from the body through fluids [34]. Direct injection of EVs can be rapidly cleared and cause secondary spinal cord injury [47]. Currently, the engineered modification of EVs has significantly improved their therapeutic effects, opening up new avenues for the translational application of EVs [48]. Constructing a sustained-release system by encapsulating EVs with hydrogels can prolong their residence time at the injury site [49, 50]. Alternatively, content loading engineering can be carried out by pre-loading neuroprotective microRNAs (miRNAs) or CRISPR components into EVs to enhance the precision of treatment [51]. In addition, through targeted modification, EVs can be combined with peptides capable of crossing the blood - spinal cord barrier (BSCB) (such as Angiopep-2 and RVG) to achieve specific delivery to the central nervous system (CNS) [35, 52]. To overcoming two critical barriers for EVs delivery to injured spinal cord: inefficient CNS biodistribution and rapid EV clearance, our solution is to modify BAT-derived EVs with Angiopep2 and embed them in GelMA hydrogel. Angiopep2 is a 19-amino-acid peptide (TFFYGGSRGKRNNFKTEEY) that specifically binds low-density lipoprotein receptor-related protein-1 (LRP-1), which is highly expressed on blood-brain barrier (BBB) endothelial cells and brain tumor cells [53, 54]. Via LRP-1-mediated receptor transcytosis, Angiopep-2 efficiently delivers conjugated drugs, nanoparticles, or gene vectors into the brain parenchyma, enabling precise treatment of cerebral disorders such as glioma, Parkinson’s disease, and Alzheimer’s disease [55–57]. Simultaneously, GelMA’s shear-thinning properties enabled minimally invasive injection and sustained EV release over > 14 days, directly countering short half-life of native EVs [47, 58]. This combinatorial strategy yielded a reduction in lesion volume and significant BMS score improvement, underscored the power of “biology-first” engineering, where endogenous repair mechanisms inform therapeutic design.

Our findings redefine BAT as a therapeutically exploitable organ beyond metabolic disease. While most EV therapies focus on stem cells, BAT-EVs offer intrinsic advantages: their activation is endogenously amplified by SCI, enabling autologous sourcing; they carry mitochondria-regulating miRNAs (e.g., miR-21) that may rescue bioenergetic failure in neurons; and their anti-inflammatory effects likely extend to other neuroinflammatory disorders. For instance, Spp1 is upregulated in Alzheimer’s-associated microglia and BAT activation improves outcomes in neural degenerative disease [59, 60]. These findings collectively indicated the potential of EVs derived from BAT in the treatment of central nervous system diseases. However, the clinical translation of brown adipose tissue is still impeded by multiple challenges. BAT activity in humans is significantly downregulated with aging and obesity [61], which necessitates the development of alternative strategies, such as the utilization of EVs derived from specific adipocytes or the optimization of organoid functions via CRISPR gene-editing technology, to enhance EV yields. Given the limited reserve of BAT in adults, the therapeutic application of autologous BAT-derived EVs often requires supplementation with exogenous preparations. This limitation has prompted us to develop a Angipep-2-EVs-hydrogel delivery system, aiming to optimize the delivery efficiency and therapeutic efficacy of BAT-derived EVs. This strategy is designed to address the clinical application bottlenecks associated with BAT scarcity, thereby maximizing the therapeutic potential of BAT-derived EVs.

In summary, this work illuminates the BAT-EV-microglia axis as a central pathway in SCI repair, mechanistically driven by miR-692/Spp1 silencing. By emulating this natural process with engineered EVs, we achieved superior functional recovery—a testament to the promise of decoding tissue cross-talk for neurological therapeutics. As the EV engineering landscape accelerates, our study exemplifies how bridging fundamental biology and advanced materials science can yield transformative therapies for conditions long deemed untreatable.

Method and materials

Animals

All animal experiments performed in this study were previously authorized by the Ethics Committee of Xiangya Hospital Central South University for Scientific Research. Mice were bred in Animals Central Xiangya Hospital South University under SPF (specific pathogen-free) conditions, with a 12 h light-dark cycle, room temperature of 22–24℃ and libitum access to food and water. All procedures were approved by the Xiangya Hospital Medical Ethics Committee of Central South University (XY20240926005).

Spinal cord injury construction

Eight-week-old male C57BL/6J mice (20–25 g) were acclimated for 1 week in an SPF facility and then anesthetized with 50 mg/kg sodium pentobarbital. After shaving the back and disinfecting with iodine and alcohol, a 1 cm midline incision was made over the T10 vertebra to expose the spinal cord. A modified Allen device was used to drop a 10 g weight from 2.5 cm, inducing injury confirmed by local hemorrhage and hind-limb spasms. The incision was closed in layers and re-sterilized. Sham-operated animals only had the cord exposed without impact. For three days post-surgery, penicillin (3000 IU/mL) was added to the drinking water; bladders were manually expressed twice daily until spontaneous urination returned. Buprenorphine (0.03 mg/kg, subcutaneous injection) was administered daily during the first week for analgesia. Mice were housed 4–5 per cage with ad libitum food and water.

Tracing of BAT-derived EVs

To investigate the impact of BAT-derived extracellular vesicles (EVs) on neuroinflammation, we performed local injections of pAAV-CMV-Cd63-linker-EGFP-3xFLAG-WPRE (AAV-CD63-EGFP) (0.5 µL, 1 × 10¹² vg/mL, OBiO Technology (Shanghai) Corp., Ltd.) into the BAT tissue of 5-week-old male mice. Age-matched naive male mice received equivalent PBS injections as negative controls (NC). SCI was established 3 weeks post-transfection. 1-week post-SCI, BAT and spinal cord tissues were collected for immunofluorescence analysis to validate both transfection efficiency and spinal cord EV localization. At the same time, to clarify the specific transfection of the virus in brown adipocytes, we used collagenase & dispase to digest the brown adipose tissue fragments for lysis. After terminating the digestion, the density centrifugation method (20% bovine serum albumin, BSA, 3000 g, 15 min) was used to separate adipocytes from non-adipocytes. Then, we obtained the floating cells in the upper layer of the solution (brown adipocytes) and non-adipocytes (precipitated cells, including vascular cells, inflammatory cells, stromal cells, etc.). Subsequently, we performed qRT-PCR on these two components respectively. The detection results showed that the expression of EGF was relatively high in the floating adipocytes in the upper layer, indicating that our transfection strategy has good affinity and high transfection efficiency for brown adipocytes.

Cell culture

Cortices from neonatal mice were aseptically dissected, digested with 0.05% trypsin, and seeded into PDL-coated flasks. After 10–14 days of culture in 10% FBS-DMEM until 90% confluence, flasks were shaken at 200 rpm, 37 °C for 8 h to detach and collect the overlying microglia. The collected microglia were cultured in DMEM containing 10% FBS and 1% Penicillin/Streptomycin under 37℃ and 5% CO2 environment. To establish an in vitro model of the inflammatory microenvironment after SCI, the BV2 microglial cell line was stimulated with 100 µg/mL lipopolysaccharide (LPS) to induce an inflammatory activation state [62–64]. Based on this inflammatory model, the experimental group adopted the following intervention strategy such as BAT-EVs intervention (100 µg/mL), miR-692 or miR-692 inhibitor administration (100 µg/mL) and SPP1 recombined protein treatment (100 µg/mL). And 24 h incubation, the cells were harvest for further analysis.

Extracellular vesicles (EVs) extraction

At 7 days post-surgery, mice from the sham and SCI groups were euthanized and the interscapular BAT tissue was collected. During the collection process, we operated on a 4 ℃ ice platform to reduce changes in the metabolic activity of BAT and simultaneously inhibit the activity of lipase. We separated the subcutaneous tissue layer by layer. After exposing the target area, we only dissected the BAT tissue and appropriately trimmed the obtained brown adipose tissue to avoid including white adipose tissue or fascia. We ensured that the obtained BAT tissue was brownish-tan and had a soft texture. After dissection, BATs were excised and cut into ~ 1 mm³ fragments. Fragments were incubated (37 °C, 30 min) in medium containing 0.1% Collagenase A (Roche, Switzerland), 0.1% Dispase (Sigma-Aldrich, US) and 0.05% DNase (Sigma-Aldrich, US). Digestion was stopped by adding an equal volume of PBS containing 2 mM EDTA (Invitrogen, US). The mixture was centrifuged at 4 °C, 300 × g for 15 min to remove residual cells, then the supernatant was transferred and spun at 3, 000 × g for 30 min to sediment residual debris. The cleared supernatant was filtered through a 0.22 μm syringe filter, and the filtrate was ultracentrifuged at 120, 000 × g for 3 h. EV pellets were gently resuspended in fresh PBS to yield purified preparations for downstream assays. The EVs were characterized by transmission electron microscopy (TEM, FEI company, USA) for morphological analysis. Nanoparticle tracking analysis (NTA) was performed on Nanoparticle Tracking Analyzer (ZetaView, Particle Metrix, Germany).

Immunofluorescence

After anesthesia, mice were transcardially perfused with ice-cold saline followed by 4% paraformaldehyde. Brown adipose tissue and the spinal cord segment were dissected, dehydrated in graded sucrose (20%, then 30%, 4 °C overnight), embedded in OCT compound (Sakura, Torrance, CA, USA), and snap-frozen. Serial 20 μm cryosections were collected and stored at − 80 °C until use. Sections were encircled with a hydrophobic pen, permeabilized with 0.1% Triton X-100 in PBST (0.05% Tween-20) for 20 min, and blocked with 3% BSA for 30 min at room temperature. After three brief PBS rinses, slides were incubated overnight at 4 °C with the appropriate primary antibodies (Table S1) diluted in antibody diluent. Following two 20 min washes in PBST and two 5 min washes in PBS, sections were incubated for 2 h at room temperature in the dark with fluorophore-conjugated secondary antibodies. After identical wash cycles, nuclei were counterstained with DAPI in an anti-fade reagent, coverslipped, and imaged on a Zeiss APTOME 3.0 fluorescence microscope.

Flow cytometry

Microglia were recovered from injured mouse spinal cord by fluorescence-activated cell sorting (FACS). Following cervical dislocation, spinal cord segments were immediately harvested and minced on ice. Tissue was enzymatically dissociated for 30 min at 37 °C in HBSS containing 0.25% Collagenase A (Roche, Switzerland), 0.1% Dispase (Sigma-Aldrich, US), 0.05% DNase (Sigma-Aldrich, US) and 10 mM HEPES. Single-cell suspensions were washed twice in ice-cold FACS buffer (PBS + 1% FBS), then incubated with anti-mouse CD16/32 (Fc Receptor Blocking Solution, 1:50, BD) for 15 min at 4 °C to block Fc receptors. Cells were stained with antibodies (CD45-FITC, CD11b-APC; as showing in Table S1) and 1 µg/mL DAPI in PBS for 30 min at 4 °C. After a final wash, events were acquired on a FACS Aria II SORP (BD Biosciences). CD45^lowCD11b⁺DAPI⁻ cells were gated and sorted as microglia. Data were analyzed using FlowJo v10 (TreeStar, US).

qRT-PCR

Total RNA was isolated from EVs released by BAT of sham or SCI mice, BAT tissue harvested at the indicated time points after SCI, and microglia purified from the spinal cord at 7 days post-surgery—using the Quick-DNA/RNA Microprep Plus Kit (Zymo Research, US). First-strand cDNA for mRNA was generated with the PrimeScript™ RT Reagent Kit (Promega, US). Mature miRNA and pri-miRNA were reverse-transcribed with the miRNA First Strand cDNA Synthesis Kit (tailing reaction) (Sangon Biotech, CN). Quantitative PCR was performed on an ABI system with GoTaq qPCR Master Mix (Promega, US). Relative expression of mRNA or miRNA was calculated by the 2^-ΔΔCT method using GAPDH and U6 as endogenous controls; primer sequences are provided in Table S2.

Western blot

Proteins were extracted from EVs, BAT (harvested at the indicated time points) and cultured microglia with ice-cold RIPA buffer (Beyotime, CN) and quantified using a BCA kit (Multisciences, CN). Equal amounts of protein were separated on 10% SDS–PAGE gels and electro-transferred to PVDF membranes. After blocking with 5% non-fat milk for 2 h at room temperature, membranes were incubated overnight at 4 °C with primary antibodies (Table S1), washed three times with TBST (0.1% Tween-20), and then exposed to HRP-conjugated secondary antibodies for 90 min at room temperature. Following three additional TBST washes (10 min each), immunoreactive bands were visualized with enhanced chemiluminescence (ECL) reagent.

microRNA-seq

EVs were purified from interscapular BAT and plasma of sham or SCI mice on day 7 post-injury. Total RNA was extracted with the miRNeasy Serum/Plasma Advanced Kit (Qiagen, US) and checked on a NanoDrop ND-1000 and an Agilent 2100 Bioanalyzer. Libraries were prepared with a small-RNA prep kit that exploits unique 3′ and 5′ ends for efficient adapter ligation. After reverse transcription and limited-cycle PCR, cDNA fragments were size-selected on an acrylamide gel and re-assessed on the Bioanalyzer. Bar-coded libraries were pooled and sequenced on an Illumina HiSeq/MiSeq platform. All miRNA-seq procedures were performed by Wuhan Seqhealth Co., Ltd., China.

Luciferase reporter assay

Plasmid constructs were generated by Hanbio Biology (Shanghai, China). A full-length Spp1 3′-UTR harboring the wild-type miR-692 target site (Spp1-WT) and a mutant version (Spp1-MUT) were sub-cloned into the psi-CHECK-2 reporter backbone. 293T cells were seeded in 24-well plates and co-transfected with either Spp1-WT or Spp1-MUT together with 50 nM miR-692 mimic or negative-control RNA (miR-692-NC) using Lipofectamine 3000 (Thermo Fisher Scientific, US). Forty-eight hours later, luciferase activities were measured with the Dual-Luciferase Reporter Assay System (Promega, US) on a GloMax-96 microplate luminometer (Promega, US). Sequences: miR-692, 5’-AUCUCUUUGAGCGCCUCACUC-3’; miR-692-NC, 5’-UCACAACCUCCUGAAAGAGUAGA-3’.

Engineering EVs construction and characteristic

DSPE-PEG2000-Angiopep2 was custom-synthesized by ShanXi BeiOu Biological Technology Co., Ltd. (Xi’an, China). The functional DSPE conjugates were then integrated into extracellular vesicles using established modification protocols [65]. Briefly, a 500 µM stock was prepared in dimethyl sulfoxide (DMSO, Sigma-Aldrich, USA) and diluted dropwise into 1× PBS under stirring to give a 5 µM working solution. Extracellular vesicles (100 µg, 2.5 × 10¹¹ particles) in 1 mL PBS were mixed with 500 µL of the working DSPE conjugate and incubated 1 h at 37 °C with gentle agitation to promote hydrophobic insertion into the EV membrane. Unincorporated conjugate was removed by ultrafiltration through a 100 kDa centrifugal filter. To confirm that DSPE-PEG2000 can be successfully inserted into the EVs, DSPE‐PEG2000‐FITC was added to the diluted EV solution. Insertion of DSPE‐PEG2000‐FITC on the EV surface was confirmed by visualizing labelled EVs using flow cytometry. We carried out the process above using various weight ratios of DSPE‐PEG2000‐FITC: EVs at 1:500, 1:250, 1:100, 1:50, 1:20, 1:10 and 1:5, followed by ultrafiltration for purification.

Preparation and characterization of the hydrogel

The hydrogel was prepared by mixing gelatin methacrylate (GelMA) and photoinitiator (LAPs) in a specific ratio according to the synthesis method described in the previous report [49]. For the preparation of the GelMA/LAP hydrogel, the lyophilized GelMA was dissolved in phosphate-buffered saline (PBS) at a concentration of 20% (w/v) at 40 °C to obtain GelMA at a final concentration of 10% (w/v) and LAP at 0.1% (w/v). Then, the EVs liquid (20 µg/µL) was mixed with the hydrogel liquid at a volume ratio of 1:1 to obtain an EVs-hydrogel mixed solution. The hydrogel solution mixed with EVs was irradiated with a 365–405 nm ultraviolet lamp at 37 °C for 15 s to transform it from a soluble state to a gel state. A Mira3-TESCAN scanning electron microscope (SEM) was used to study the structural and morphological changes of the hydrogel. The in-vitro degradation of the hydrogel was evaluated by preparing 100 µL of EVs-GelMA hydrogel, and the initial weight was recorded as Md0. Then, the hydrogel was immersed in a PBS solution at pH 7.4 and incubated in a shaker at 4 °C for 28 days. At specific time points (0 days, 3 days, 7 days, 14 days, 21 days, and 28 days), the supernatant was aspirated, and the weight of the hydrogel was measured and recorded as Mdt. The degradation rate of the hydrogel was calculated using the formula DR = ((Mdt - Md0)/Md0) × 100%. A rheometer equipped with a temperature control unit (Anton Paar, Austria) was used to measure the rheological properties of the hydrogel with a parallel-plate configuration (plate diameter: 20 mm). A frequency-sweep test in the frequency range of 0.1–1 Hz was performed at 37 °C and 1% strain. Hydrogel samples (100 µL, 10 mg GelMA) were photopolymerized in a 1.5 mL tube and 1 mL PBS was gently added. Samples were incubated at 37 °C. At specific time point (0, 3 h, 6 h, 12 h, 24 h, 3 d, 5 d, and 7 d), the PBS was removed and the swollen weight of the hydrogel sample was recorded (Wwet). The swelling ratio was calculated as Wwet/10 mg.

EVs loading and sustained releasing test in vitro

To generate the release profile of EV-loaded hydrogel, we performed a systematic in vitro analysis. Specifically, 100 µL of EV-loaded hydrogel (10 mg/mL, post UV irradiation) was immersed in 500 µL of PBS dissolution medium at 37 ℃. At predetermined time intervals, 500 µL of the sample was collected, and the same volume of pre-heated PBS dissolution medium was replenished to maintain a consistent dissolution environment. For the quantitative assessment of EV release, two methods were employed. First, the Micro BCA Protein Assay Kit (ThermoFisher, Rockford, USA) was utilized to measure the total protein content as an indicator of EV release. Second, nanoparticle tracking analysis (NTA) was conducted to quantify the concentration and size distribution of released EVs. Additionally, transmission electron microscopy (TEM) analysis was performed to evaluate the structural integrity of the EVs released from the hydrogel, ensuring their preservation during the release process.

Biocompatibility evaluation of hydrogel

The biocompatibility of EVs-loaded hydrogels was evaluated by in vitro determination of neural cell viability using the CCK-8 method, and in vivo histological examination of multiple organs was performed by hematoxylin-eosin (H&E) staining. For in vitro evaluation, neuronal cells were seeded in 96-well plates pre-coated with phosphate-buffered saline (PBS), hydrogel, and SCI-EVs + hydrogel, respectively. After culturing for 1 h, 3 h, 6 h, 12 h, 24 h, and 48 h, the absorbance (OD) values of cells in different groups were measured using a microplate reader. For in vivo evaluation, adult female C57BL/6 mice were anesthetized with sodium pentobarbital, and hydrogels with or without EVs were carefully implanted subcutaneously into the mice. Two months later, the mice were euthanized, and their hearts, livers, spleens, lungs, and kidneys were collected for histological analysis by H&E staining.

EVs labeling and in vivo tracking

EVs were re-suspended in ultrapure water at 1.0 µg/µL. For membrane labeling, an aliquot was mixed with 5 µg/mL PKH67 Red Fluorescent Cell Linker (Unmibio, CN) or DiR Iodide (DiIC18 [7]; Yeason, CN) and incubated for 30 min at room temp

erature. Unbound dye was removed by three washes through 100-kDa Ultra centrifugal filters (Millipore, USA) in PBS, followed by a final ultracentrifugation step. For in-vivo tracking, DiR or PKH67 labeled EVs were embedded in 10% GelMA and applied directly to the contused spinal cord immediately after injury. After 7 days, mice receiving PKH67-EVs were sacrificed; the spinal cord was dissected, frozen-sectioned at 20 μm, and counter-stained to visualize EV uptake by microglia. Animals injected with DiR-EVs were imaged live using the Xenogen IVIS Spectrum (Caliper Life Sciences).

Locomotor function and neuroelectrophysiological evaluation

Motor function was monitored by two investigators blinded to group identity. The Basso Mouse Scale (BMS) was applied before injury, immediately after, and at 1, 3, 7, 14, 21 and 28 dpi. Briefly, in a 4-min open-field test, mice were scored 0–9 for ankle movement, weight-bearing, paw placement, coordination and trunk stability (0 = complete paralysis, 9 = normal gait). For motor-evoked potentials (MEPs), animals were anesthetized with 0.3% pentobarbital sodium (70 mg/kg), shaved, disinfected and secured in a stereotactic frame on a 37 °C heating pad. A craniotomy exposed the motor cortex; a bipolar stimulating electrode was advanced 700–1000 μm below the pial surface to activate corticospinal neurons. Recording electrode was placed on the gastrocnemius. Signals were amplified and digitized with a Cereplex Direct system (Blackrock, US). Stimulation consisted of single 10 V pulses.

PET/CT imaging

For in-vivo visualization of mouse BAT activity, conscious mice were acclimated at 28 °C. Each animal received 7.4 ± 0.2 MBq ¹⁸F-FDG (in 0.1 mL saline) via tail-vein injection under 1.5% isoflurane/O₂ anesthesia. After an uptake period of 45 min at 28 °C, mice were re-anesthetized and positioned prone on the heated bed (30 °C) of a Micro PET/CT scanner (Mediso, Hungary). A 3-min low-dose CT (80 kVp, 500 µA, spatial resolution 100 μm) was acquired for attenuation correction, immediately followed by a 10-min static PET list-mode acquisition. Images were reconstructed with 3D-OSEM (4 iterations, 16 subsets) and CT-based scatter/attenuation correction. BAT uptake was quantified as %ID/g within CT-defined regions, excluding skeletal muscle. Respiratory gating and continuous warming minimized motion and cold-induced activation artifacts. The calculation of %ID/g can be expressed as follows: %ID/g = (Radioactivity Count (Bq/mL) * Tissue Weight(g) /(Injected Dose (Bq) ​×1000).

Statistical analysis

Statistical analyses were conducted using GraphPad Prism version 10. For pairwise comparisons, Student’s t-test was applied, and one-way ANOVA was used to evaluate differences across multiple groups. Tukey’s multiple comparisons post-test was subsequently implemented for post-hoc analysis. Data are expressed as the mean ± standard deviation (Mean ± SD), with statistical significance defined as P < 0.05.

Author contributions

TQ, SH and PL designed the study. SX performed most of the experiments and data analysis. YQ, YS, MZ and YT assisted in sample collection and data analysis. SX and TQ drafted the manuscript. TQ and SH revised the manuscript. All authors contributed to this manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (grant No. 82272045), Science and Technology Innovation Program of Hunan Province (grant No. 2021RC4056), Clinical Research Foundation of the National Clinical Research Center for Geriatric Diseases (XIANGYA) (grant No. 2020LNJJ01), Key Program of Ministry of Industry and Information Technology of China (grant No. CEIEC-2022-ZM02-021).

Data availability

All data in the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.