Selenized neural stem cell-derived exosomes: A neotype therapeutic agent for traumatic injuries of the central nervous system

Summary

Oxidative damage and neuroinflammation are the key features of central nervous system (CNS) injury. Inspired by the neuroprotective properties of neural stem cell-derived exosomes (NExo) and the reactive oxygen species (ROS) scavenging ability of selenium, we develop an advanced NExo bearing ultrasmall nano-selenium (∼3.5 nm) via lipid-mediated nucleation (SeNExo). In addition to maintaining the biological components of NExo, the resulting SeNExo exhibits a Se–O bond that dramatically enhances its ROS-scavenging performance. SeNExo penetrates the blood-brain barrier (BBB) via the apolipoprotein E and prolow-density lipoprotein receptor-related protein 1 (APOE_LRP-1) interaction. Through proteomics, microRNA (miRNA) omics, and single-nucleus RNA sequencing, we find that SeNExo can alleviate neuronal apoptosis, restore glia homeostasis, and remodel glia-neuron networks. Therefore, SeNExo confers potent therapeutic benefits, significantly reducing cerebral lesions in a murine traumatic brain injury model. Even extending to a murine spinal cord injury model, SeNExo promotes locomotory recovery, further supporting SeNExo as a neotype and a promising therapeutic agent for treating traumatic CNS injury.

Graphical abstract

Highlights

• •NSC-derived exosomes are functionalized with ultrasmall nano-selenium
• •SeNExo crosses the BBB by utilizing APOE_LRP-1 interaction
• •SeNExo exhibits ROS scavenging and neuroprotection abilities
• •Therapeutic efficacies are demonstrated in TBI and SCI model mice

Wang et al. develop SeNExo, an advanced NExo bearing ultrasmall nano-selenium, exhibiting both ROS scavenging and neuroprotection abilities. SeNExo alleviates neuronal apoptosis, restores glia homeostasis, and remodels glia-neuron networks in TBI and SCI model mice.

Introduction

Traumatic injuries to the central nervous system (CNS), including traumatic brain injury (TBI) and spinal cord injury (SCI), often lead to long-term disability across all age groups.^1,2,3 CNS injury pathology is complex, consisting of primary and secondary injuries.^4,5 Primary injury causes immediate mechanical disruption and focal tissue damage, initiating a cascade of molecular and biochemical events that lead to secondary injury. Typically, immune cells, such as microglia, are activated by the release of potential cytotoxic molecules, such as reactive oxygen species (ROS).^6,7,8,9 This well-known secondary messenger engages in positive feedback to exacerbate neuroinflammation, ultimately resulting in neuronal loss and neurological dysfunction.^10,11,12

Motivated by the potential for cell-based therapeutics, researchers have explored neural stem cells (NSCs) to mitigate neuronal damage.^13,14,15,16 Despite promising results, the therapeutic efficacy of NSCs faces limitations.^{17,18,19,20,21} For instance, pathological microenvironments negatively impact NSC survival and directed differentiation, compromising therapeutic outcomes and causing side effects. Furthermore, the isolation and growth of NSCs takes a long time, making them unsuitable as off-the-shelf products for acute CNS injury treatment. Given that the therapeutic benefits of stem cells are known to involve the paracrine of exosomes and the good stability of exosomes raises the possibility of remaining medically active after lyophilization and in the pathological microenvironment,^22,23,24 NSC-derived exosomes (NExo) seem to be ideal candidates for preserving the benefits of NSCs while overcoming the aforementioned limitations for treating CNS injury. With regard to delivery, exosomes have shown the potential for penetrating biological barriers, facilitating them to access CNS injury site upon intravenous (i.v.) injection.^25,26,27

While certainly exciting, the complex pathological microenvironments of CNS injury highlight the need to further enhance the therapeutic potential of NExo. Given the aforementioned central role of ROS in exacerbating CNS injury, we propose developing advanced NExo with ROS scavenging ability to boost their therapeutic efficacy. Selenium, a trace element known to reduce oxidative damage,^28,29 is the natural choice for NExo modification. Previous studies show nano-selenium’s superiority in ROS scavenging and biocompatibility.^30,31 However, even state-of-the-art fabrication techniques only produce nano-selenium particles with a size of 40–400 nm,³² and the relatively low surface-to-volume ratio restricts the yet-higher antioxidant activity. Based on our nanocrystal growth experience, we anticipate that an in situ approach for growing nano-selenium at the membrane of NExo might yield two highly desirable outcomes: ultrasmall particle size (utilizing a high surface-to-volume ratio to promote ROS scavenging) and blood-brain barrier (BBB) penetration capacity (as an integrated component to hitchhike exosomes).

Keeping these considerations in mind, we developed a mild and facile one-pot approach to decorate NExo with nano-selenium (SeNExo) through lipid-mediated crystallization (Figure 1A). Owing to its ultrasmall size (∼3.5 nm), nano-selenium at the membrane exhibited a high ROS scavenging capacity. We elucidated that the BBB penetration of SeNExo was mediated by the APOE_LRP-1 interaction. Extensive proteomics, microRNA (miRNA) omics, and single-nucleus RNA sequencing (snRNA-seq) revealed that SeNExo alleviated neuronal apoptosis, restored glia homeostasis, and remodeled glia-neuron networks (Figure 1B). In both murine TBI and SCI models, SeNExo thus elicited significant therapeutic benefits, highlighting its potential as a neotype therapeutic agent for treating CNS injury.

Figure 1.

Design, construction, and characterizations of SeNExo

(A) Schematic illustration showing the construction of SeNExo.

(B) Schematic illustration showing the BBB penetration of SeNExo and subsequent efficient therapy for CNS injury.

(C) Representative TEM images of SeNExo and the magnified (211) facet with 0.308 nm interplanar spacing of the ultrasmall nano-selenium. Low-magnification image scale bars, 100 nm. High-magnification image scale bars, 2 nm.

(D) Representative cryo-TEM image of SeNExo, showing the ultrasmall nano-selenium at the membrane (indicated by white arrows). Scale bars, 20 nm.

(E) Elemental mapping images of SeNExo and the merged view. Scale bars, 50 nm.

(F) Electron spin resonance (ESR) spectra for assessing the free radical scavenging activity of SeNExo compared with Se NPs toward ⋅OH radicals. The Se content in both entities remained uniform.

(G) Left: illustration of the model for Se and Se with partial Se–O bond insertion, showing different adsorption levels for hydroxyl radicals, and the corresponding difference charge density for the Se/⋅OH and Se–O/⋅OH systems. Right: calculated adsorption energies of Se/⋅OH or Se–O/⋅OH systems, and the Bader charges of the predicted gained electrons for ⋅OH in the Se/⋅OH or Se–O/⋅OH systems.

(H) Stability characterization of SeNExo in terms of particle size, zeta potential, and Se content before and after a lyophilization/rehydration (Lyo/Reh) cycle (n = 3 independent replicates).

(I and J) Proteomics and miRNA omics analysis of fresh NExo and SeNExo after Lyo/Reh.

Data in (H) are presented as the mean ± SD and were compared by two-tailed unpaired Student’s t test.

Results

Construction and characterization of SeNExo

Inspired by the abundant phosphoryl groups on the membrane and their potential to interact with precursors and regulate subsequent nucleation, we developed a mild and facile method for in situ crystallized nano-selenium at the exosome membrane. Briefly, NExo was harvested from the supernatant secreted by NSC spheroids (Figure S1A) and incubated with a sodium selenite solution (0.75 mM, pH = 7.4) to absorb selenite at the NExo membrane. The selenite-adsorbed NExo was then exposed to an ascorbic acid solution (4.5 mM, pH = 6.5), enabling in situ growth of nano-selenium at the membrane. Transmission electron microscopy (TEM) imaging showed that the resulting SeNExo retained a similar size (∼120 nm) and cup-like morphology as pristine NExo (Figures 1C and S1B). However, ultrasmall crystals (∼3.5 nm) were observed on SeNExo, showing a (211) facet with an interplanar spacing of 0.308 nm. Cryo-TEM imaging verified that ultrasmall crystals grew in situ at the phospholipid bilayer of NExo (Figure 1D). Elemental mapping (Figure 1E), energy-dispersive spectroscopy (Figure S1C), and X-ray diffraction (Figure S1D) confirmed that these crystals were nano-selenium. After optimization and calculation, the Se content in SeNExo was up to 0.132 nmol/10⁸ particles with good reproducibility (Figures S1E and S1F).

To validate the exceptional ROS scavenging capacity of ultrasmall nano-selenium, we conducted electron spin resonance (ESR) spectroscopy. As shown in Figures 1F and S1G, SeNExo could fully quench high-energy species like hydroxyl radicals (⋅OH), superoxide anions (O₂^⋅-), and nitroxide radicals (⋅NO), with spectra approximating the baseline. In contrast, an equal dose of traditional selenium nanoparticles (Se NPs) (∼80 nm) showed only a modest decrease in amplitude relative to the control solution without an antioxidant agent. For quantitative comparison, we used a commercial free radical detection kit. Reinforcing our findings from the ESR analysis, SeNExo exhibited a nearly 4-fold ROS scavenging ability than traditional Se NPs (Figure S1H). Such an excellent ROS-scavenging performance of SeNExo was also verified in cultured cells treated with H₂O₂ (Figures S1I and S1J). For the underlying mechanism, we discovered that phospholipids in NExo were essential for in situ formation of ultrasmall nano-selenium and contributed to the formation of Se–O bond in SeNExo (Figures S2A–S2H). Upon this, we conducted density functional theory calculations and selected ⋅OH as the representative ROS. The Se–O bond increased adsorption energy between Se and ⋅OH from 0.77 to 0.99 eV and promoted the charge transfer from 1.02 to 1.09 e (Figures 1G, S2I, and S2J), thereby enhancing ⋅OH adsorption and subsequent ⋅OH scavenging, respectively.

Upon successful fabrication of SeNExo with excellent ROS-scavenging performance, we conducted a series of experiments to evaluate their stability. During 48 h of incubation with phosphate-buffered solution (PBS) containing 10% serum, SeNExo exhibited no changes in size, zeta potential, or Se content (Figure S2K), indicating their good structural stability. These characteristics remained unchanged after a lyophilization and rehydration cycle (Lyo/Reh; Figure 1H). To assess shelf life, lyophilized SeNExo was stored at room temperature for 28 days. The results showed minimal impact on SeNExo’s size and zeta potential after rehydration (Figure S2L).²² Moreover, there were no significant changes in Se content or 1,1-diphenyl-2-picrylhydrazyl free radical scavenging capacity, demonstrating favorable physicochemical stability during storage (Figure S2L). In addition, proteomic data in Figure 1I show that, compared with fresh NExo, SeNExo after a Lyo/Reh cycle had comparable intensities of representative proteins associated with exosome markers (e.g., TSG101 and CD63), transport and cell adhesion (e.g., MFGE8 and APOE), and nervous system development (e.g., YWHAE and NCAM1). MiRNA-omics data (Figure 1J) showed that miRNAs related to the inhibition of neuronal apoptosis (e.g., miR-30d-5p and miR-335-5p) and anti-inflammatory function (e.g., miR-16-5p and miR-125b-5p) were retained in SeNExo after a Lyo/Reh cycle, which could be attributed to the protected and controlled internal microenvironment provided by the vesicle membrane enclosure. Beyond the aforementioned physical stability, this good biological stability further supports SeNExo as an off-the-shelf therapeutic agent that can be used in acute CNS injury settings.

Evaluation and mechanism of SeNExo’s BBB penetration capacity

Because accumulation at the lesion site is a prerequisite for treating CNS injury, we investigated the BBB penetration capacity of SeNExo in TBI model mice induced by a controlled cortical impactor.³³ While TBI disrupts the BBB, leading to acute breakdown and increased permeability, the BBB typically self-repairs within hours to days after injury.^34,35 Therefore, we investigated whether SeNExo possess penetration capacity during both the acute injury and recovery stages of the BBB. We initially examined the accumulation performance in the acute stage of brain injury (0 days). Liposome (Lipo, a protein-free membrane control), NExo, and SeNExo were labeled with Al (III) phthalocyanine chloride tetrasulfonic acid (Alp), and fluorescence was monitored over time after i.v. injection (Figures 2A and 2B; Figures S3A–S3C). All groups shared a similar peak time (6 h) for fluorescence signals at the TBI lesion site. At this time, the NExo group exhibited nearly 2.7-fold signal intensity than the Lipo group, and similar results were observed in the SeNExo group, suggesting no compromise from the modified ultrasmall nano-selenium. Note that the nano-selenium remained stable during circulation and Se content in lesion tissues from SeNExo-treated mice significantly increased (Figures S3D and S3E), together indicating that nano-selenium hitchhiked the exosomes for accumulation at the lesion site.

Figure 2.

Mechanism of BBB penetration ability of SeNExo

(A and B) Schematic illustration of the administration of Alp-labeled Lipo, NExo, and SeNExo via i.v. injection for TBI model mice at 0 days, corresponding in vivo real-time fluorescence images of mice, and quantitative fluorescence analysis.

(C and D) Schematic illustration of the administration of Alp-labeled Lipo, NExo, and SeNExo via i.v. injection for TBI model mice at 7 days, corresponding in vivo real-time fluorescence images of mice, and quantitative fluorescence analysis.

(E) Dot plot showing the top 6 expressed ligands and receptors associated with cellular recognition and transport between SeNExo and bEnd.3 cells.

(F) Schematic illustration of the administration of Alp-labeled SeNExo and anti-APOE-blocked SeNExo via i.v. injection for TBI model mice (7 days, left), corresponding in vivo fluorescence images of mice (middle), and quantitative fluorescence analysis (right).

(G) Representative fluorescence images of frozen TBI brain section (7 days) at 6 h post injection with 3,3′-dioctadecyloxacarbocyanine perchlorate (DIO)-labeled SeNExo. Regions in the white squares (f1–f3, f3i–f3iii) were enlarged to show the colocalization of DIO-labeled SeNExo (green) with microglia (IBA1+, cyan), astrocytes (GFAP+, yellow), and neurons (NEUN/β3-tubulin+, red/gray) near the injury. Nuclei: Hoechst 33342 (blue). Low-magnification image scale bars, 500 μm. Scale bars in f1–f3, 50 μm. Scale bars in f3i–f3iii, 10 μm.

Data in (B), (D), and (F) are presented as the mean ± SD from 3 biological replicates and were assessed via one-way ANOVA (B and D) and two-tailed unpaired Student’s t test (F).

Subsequently, we investigated the accumulation performance of the samples 7 days after TBI. Unlike the acute stage, the Lipo signal was hardly detected in the recovery stage of the BBB, while both NExo and SeNExo still showed abundant accumulation at the lesion site, exhibiting a signal intensity nearly 4.5-fold than that in the Lipo group (Figures 2C and 2D). These results indicated that unlike Lipo, which relied on the integrity of the BBB for penetration, NExo and SeNExo still had BBB penetration capacity even in the recovery stage, probably because of the proteins on the membrane.

To gain deeper insight at the molecular level, we used a well-established Transwell model (Figures S3F–S3I) and conducted proteomics analysis of SeNExo and bEnd.3 cells.³⁶ Based on the top six expressed ligands associated with cellular recognition and transport on SeNExo, we observed the expression of the corresponding receptors on bEnd.3 cells (Figure 2E). Among these ligand_receptor pairs, apolipoprotein E (APOE) from SeNExo and prolow-density lipoprotein receptor-related protein 1 (LRP-1) from bEnd.3 were in abundance, indicating the possibility of this pair deeply participating in the SeNExo’s BBB penetration. To verify this hypothesis, we conducted a blocking experiment using the Transwell model (Figure S3J). As expected, blocking APOE of SeNExo (SeNExo+anti-APOE) significantly compromised its penetration behavior. For verification, we also examined the APOE-blocked SeNExo 7 days after TBI and observed that the fluorescence signals at the TBI lesion site decreased to nearly 50% (Figures 2F and S3K), again supporting that the APOE_LRP-1 interaction contributed significantly to the BBB penetration of SeNExo.

For higher resolution, the frozen slides of brain tissue were immunostained with ionized calcium-binding adapter molecule 1 (IBA1), glial fibrillary acidic protein (GFAP), neuronal nuclear antigen (NEUN), and β3-tubulin to mark microglia, astrocytes, neuronal nuclei, and neuronal microtubules, respectively. As shown in Figure 2G, abundant SeNExo was observed in the cerebral cortex and hippocampal areas on the injured side. In the enlarged images, SeNExo co-localized with IBA1, GFAP, and β3-tubulin in the cerebral cortex and hippocampal areas. Additional sorting experiments, confocal laser scanning microscopy imaging, and TEM observation of microglia, astrocytes, and neurons further provided evidence that SeNExo indeed accumulated in these three cell types (Figures S3L–S3N).

Therapeutic effects of SeNExo in TBI model mice

Having found that SeNExo can cross the BBB and accumulate at the TBI lesion site, we evaluated the in vivo therapeutic effects of SeNExo. The TBI model mice were randomly divided into different treatment groups, including PBS, Se NPs, NExo, a mixture of Se NPs and NExo (Se NPs + NExo), SeNExo, and the sham group with PBS treatment was used as the control. Upon dose-escalation study (Figures S4A and S4B), mice were intravenously administered a dose of 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg. Considering the changes in BBB permeability following brain injury (Figures S4C and S4D), we set the administration time points as once every 2 days during the acute stage and once a week during the recovery stage (Figure 3A).^35,37 In the 5^th week, the standard Morris water maze (MWM) test was used to evaluate the spatial learning and memory of the various groups.³⁸ In the 6^th week, these mice were further subjected to two-photon microscopic imaging of ROS levels and histological analyses of the lesion site.

Figure 3.

Therapeutic effects of SeNExo in TBI model mice

(A) Experimental design for evaluating the therapeutic effects in TBI model mice.

(B) Representative MWM-swimming track. The platform was in the third quadrant, with the red and blue squares representing the entry and end positions, respectively.

(C) The number of platform crossings and the percentage of searching time in the target quadrant.

(D) Representative two-photon fluorescence images showing ROS levels at the lesion site and quantitative analysis. Scale bars, 200 μm.

(E) Representative immunofluorescence of neuroinflammation markers IBA1 and GFAP at the lesion site and quantitative analysis. Scale bars, 50 μm.

(F) Representative immunofluorescence of neuronal survival marker NEUN at the lesion site and quantitative analysis. Scale bars, 50 μm.

(G) Representative fluorescence images of Hoechst 33342-stained brain sections with lesion area quantitation. Scale bars, 1 mm.

(H) Disparities of seven indices between the sham group and different treatment groups. The linear distance between two groups was calculated using the Euclidean distance method.

(I) Heatmap of seven indices among the six groups. Data were normalized for plotting.

Data in (C)–(G) are presented as the mean ± SD from 6 biological replicates and were assessed via one-way ANOVA. The images in (E), (F), and (G) were serial tissue sections.

As shown in Figures 3B and 3C, TBI model mice showed a sharply reduced number of platform crossings and time percentage in the target platform quadrant. Such spatial learning and memory disabilities were modestly ameliorated in mice treated with traditional Se NPs or NExo alone. Upon addition of their mixture (Se NPs + NExo), further improvement was observed owing to the cooperation of ROS scavenging and neuroprotection. Even so, it should be noted that the effect of ROS scavenging was still limited, since these Se NPs failed to efficiently enrich at the lesion site. Once the mice were treated with SeNExo, ultrasmall nano-selenium with higher ROS scavenging ability could be ferried by the NExo to the lesion site. Given the preferential local antioxidant activity of ultrasmall nano-selenium synergizing with its systemic effects (Figures S4E and S4F) and NExo-mediated neuroprotection, we detected the most prominent therapeutic effects in the SeNExo group, with 6.5-fold crossings and a 3.2-fold time percentage in the target quadrant compared with those in the PBS group. Note that these improvements in spatial learning and memory surpassed those achieved with a common clinical TBI intervention combining an ROS scavenger (edavarone) and a neuroprotective agent (GM1 ganglioside, GM1) (Figures S4G and S4H).^39,40

To gain a deeper insight into the excellent therapeutic benefits of SeNExo, we performed histological analyses. Using 2′, 7′-dichlorofluorescin diacetate as the ROS probe, the SeNExo group showed the lowest fluorescence intensity at the lesion site, similar to the sham group (Figure 3D). As representative indicators of neuroinflammation, the numbers of reactive microglia and astrocytes were evaluated by analyzing IBA1 and GFAP levels, respectively. Unlike the PBS group, the SeNExo group showed significantly fewer activated glial cells (Figure 3E), with the tumor necrosis factor alpha (TNF-α) and interleukin (IL)-1β production similar to the sham group (Figure S4I). Reduced ROS levels and neuroinflammation in the SeNExo group resulted in the lowest neuronal loss, indicated by the maximum fluorescence intensity signal of NEUN and minimum lesion area in coronal sections (Figures 3F and 3G), demonstrating excellent therapeutic effects at the histological level.

To provide a clearer visualization for the disparities of the aforementioned data, we calculated the linear distance between the sham group and different treatment groups across seven indices, including MWM behaviors (crossing number and time percentage in the target quadrant), ROS levels, activated microglia and astrocyte indicators (IBA1 and GFAP), and neuron number indicators (NEUN and lesion area). As shown in Figure 3H, the SeNExo group was closer to the sham group, confirming SeNExo’s excellent therapeutic effects on TBI. A heatmap of data from individual mice (Figure 3I) showed that SeNExo had the highest Z scores for crossing number, time percentage, and NEUN and the lowest Z scores for ROS, IBA1, GFAP, and lesion area. This further confirmed that SeNExo efficiently repaired TBI by reducing oxidative stress, inhibiting neuroinflammation, and attenuating neuronal loss. Notably, these effects were observed with few abnormalities in hematological parameters (Figure S4J), serum biochemical indicators (Figure S4K), and histology of the main organs (Figure S4L), indicating the safety of SeNExo.

Mechanism of SeNExo’s therapeutic effects using snRNA-seq

To better understand the underlying mechanism behind SeNExo’s therapeutic effects, we performed snRNA-seq analysis on lesion tissue from TBI model mice treated with SeNExo or PBS (Figure 4A), with brain tissue (same position) from sham mice as a healthy control. Based on known marker genes, all single cells were clustered into 10 major clusters: neurons, microglia, astrocytes, oligodendrocyte progenitor cells (OPCs), oligodendrocytes, vascular leptomeningeal cells, ependyma, endothelia, smooth muscle cells, and macrophages (Figure 4B). The cell frequencies of these clusters varied among the sham, PBS, and SeNExo groups (Figure 4C), with the cell frequencies of neurons and four types of glial cells (microglia, astrocytes, OPCs, and oligodendrocytes) prominently altered.

Figure 4.

Diversity of cell types identified by snRNA-seq analysis and the mechanisms of SeNExo’s effects for alleviating neuronal apoptosis

(A) Schematic of lesion tissues collected from mice brains and processing for snRNA-seq.

(B) t-distributed stochastic neighbor embedding (t-SNE) plot of brain tissue colored by main cell lineage (left), and a dot plot depicting the cell type-specific marker genes for individual clusters (right).

(C) Different cell type frequency changes in the sham, PBS, and SeNExo groups.

(D) Histogram of differentially expressed genes (DEGs) in neurons of different groups. Venn diagram of upregulated and downregulated DEGs in neurons was displayed on the right side. |Log2 (fold change)| > 0.36, q value < 0.05.

(E) Heatmap of representative DEGs selected from Venn intersection in (D).

(F) Violin plots of different gene-set scores in neurons across the sham, PBS, and SeNExo groups.

(G) Heatmap showing the similar trend between the ROS level and expression of representative genes for the sham, PBS, and SeNExo groups.

(H) Network plot (from PBS vs. SeNExo) illustrating that the proteins (circle) in the SeNExo interact with the neuronal upregulated DEGs (square, red edge) related to neuroprotection and neuronal downregulated DEGs (square, blue edge) related to apoptosis.

(I) Sankey plot (from PBS vs. SeNExo) showing the miRNAs in the SeNExo target the neuronal DEGs with known functions in the regulation of neuronal apoptosis.

Data in (F) were assessed via one-way ANOVA.

We initially focused on differentially expressed genes (DEGs) in neurons after SeNExo treatment (Figure 4D). Compared to the PBS group, SeNExo-treated neurons showed reduced levels of specific genes involved in oxidative stress (e.g., Cst3 and Acsl3) and apoptosis (e.g., Ubb and Bax) and increased levels of specific genes with neuroprotection functions (e.g., Tenm4 and Mef2c) (Figure 4E). We also inspected the gene expression signature and confirmed the functional regulation of SeNExo on neurons in the aspect of oxidative stress, apoptosis, and neuroprotection (Figure 4F). To understand the relationship between the SeNExo components and aforementioned DEGs, we conducted a deeper analysis. Given the well-demonstrated roles of ROS in neuronal cell death and the superior ROS scavenging ability of nano-selenium on SeNExo, we performed a heatmap, which showed a similar trend between the previously quantified ROS levels and the expression levels of representative genes related to oxidative stress and apoptosis in the sham, PBS, and SeNExo groups (Figure 4G). Considering the abundant protein and miRNA components in the vesicles of SeNExo, we explored their potential relationship with DEGs. As shown in Figure 4H, the proteins in the vesicles had network interactions with upregulated neuroprotective DEGs (e.g., Pparg and Mef2c) and downregulated apoptotic DEGs (e.g., Ubb and Xbp1). Meanwhile, SeNExo’s miRNAs could target genes involved in apoptosis regulation (e.g., Traf6 and Xbp1), contributing to alleviate neuronal apoptosis (Figure 4I).

Recalling the significant alterations in the proportions of the four glial cell types, we subsequently assessed DEGs in microglia, astrocytes, OPCs, and oligodendrocytes. Microglia, as resident macrophages in the brain, respond to injury by transitioning from a homeostatic to a proliferative state, producing cytokines and chemokines to activate immune cells.⁴¹ Given that microglia in TBI exhibit relatively greater heterogeneity than in the resting state, we performed pseudo-time analysis to characterize their “state changes.”⁴² As shown in Figure 5A, 3.8%, 89.3%, and 9.0% of the microglia in the sham, PBS, and SeNExo groups were categorized as “late” state, respectively. This supported the notion that SeNExo treatment profoundly altered the transcriptional program of inflammatory responses and promoted microglia to a homeostatic state, similar to the sham group. A heatmap along the trajectories of microglia revealed seven microglial gene clusters (MCs). Notably, MC1 (e.g., C1qa and Trem2) and MC2 (e.g., B2m and H2-K1) genes were upregulated in the “late” state (Figures 5B and S5A). Focusing on the representative MC1 and MC2 genes (Cst3, C1qa, Trem2, Cd9, B2m, and Fth1) during pseudo-time, significantly upregulated expression was observed in the PBS group (Figure 5C), consistent with previous studies on inflammation-stimulated microglia.⁴³ However, the SeNExo treatment downregulated these genes, which mostly overlapped with the sham group. We also inspected the gene expression signature and confirmed the functional regulation of SeNExo on microglia in the aspect of oxidative stress and inflammatory response (Figure S5B). Through the heatmap analysis, we found the similar trend between ROS levels and the expression levels of representative genes related to oxidative stress and inflammation in the sham, PBS, and SeNExo groups (Figure S5C), thus confirming the roles of scavenging ROS in alleviating neuroinflammation. Taken together, these data supported that SeNExo treatment successfully restored the homeostatic state and ameliorated the inflammatory response of microglia in TBI model mice.

Figure 5.

Mechanisms of SeNExo’s effects on restoring glial homeostasis and remodeling glia-neuron networks

(A) Pseudo-time analysis showing the “early” and “late” states of microglia along the main pseudo-time trajectories in the sham, PBS, and SeNExo groups. Each dot represented one cell.

(B) Heatmap showing clustering and expression of genes along the trajectories of microglia. Z score represented the gene expression abundance: a higher score indicated higher gene expression abundance at a given pseudo-time point.

(C) Scatterplot of the expression of representative MC1 and MC2 genes along pseudo-time.

(D) Overview of the significantly enriched ligand_receptor interactions involved in astrogliosis, inflammatory response, and CNS development between neurons and glial cells in the sham, PBS, and SeNExo groups. The pink dot represented p < 0.05, and the gray dot represented p ≥ 0.05.

(E) Diagram of representative ligand_receptor pairs in (D).

In the terms of astrocytes, we found that SeNExo downregulated the expression of DEGs related to inflammation (e.g., Clu and Gja1), oxidative stress (e.g., Cst3 and Acsl3), and apoptosis (e.g., Ctsd and Ubb, Figure S5D), sharing a similar trend with ROS levels (Figure S5E). Extending our attention to OPC and oligodendrocytes, we found that OPCs progressed toward oligodendrocytes before bifurcation (pre-branch) by pseudo-time analysis, which confirmed that oligodendrocytes were generated from a pool of OPCs. However, oligodendrocytes bifurcated into two diverse branches and revealed a decreased proportion of “immune state” of oligodendrocytes after SeNExo treatment, which significantly downregulated genes such as CD81 and CD9 (Figures S5F–S5I). The aforementioned genes in the four glial cell types are all known to be associated with inflammatory responses, and their downregulation in the therapeutic group could be attributed to the functional components of SeNExo. On the one hand, the efficient scavenging of ROS sourced from the ultrasmall nano-selenium on the vesicle downregulated the expressions of Traf6 and Trem2, which inhibited the nuclear factor κB signaling pathway in glial cells and further alleviated neuroinflammation.⁴⁴ On the other hand, a proportion of miRNAs in exosomes, such as mmu-miR-125b-5p and mmu-miR-26a-5p, could target genes such as B2m and Cd81 with known functions in the regulation of inflammatory responses (Figure S5J).

Dysfunctional crosstalk between neurons and glial cells has been considered as another essential driving force for exacerbating CNS injury.⁴⁵ To this end, we used CellphoneDB to investigate the intercellular communication between glial cells and neurons.⁴⁶ Significant ligand_receptor pairs (p < 0.05) associated with inflammation, astrogliosis, and CNS development were enriched (Figures 5D and 5E). Specifically, ligand_receptor pairs like Sema3a_Nrp1 and P2ry6_Nampt between microglia and neurons, and Pdgfc_Pdgfra and Ptprz1_ Ptn between OPCs and neurons, were highly expressed in the PBS group but were sharply downregulated after SeNExo treatment, reducing the pro-inflammatory signaling cascades to alleviate neuronal death. Furthermore, the known astrogliosis-related ligand_receptor pair Fgf1_Fgfr1 was downregulated between astrocytes and neurons after SeNExo treatment, which provided a genetic explanation for the reduced astrogliosis observed in our immunofluorescence data (GFAP level in Figure 3E). In addition, ligand_receptor pairs with neuronal repair functions between astrocytes and neurons (e.g., Lgr4_Nrg1) and oligodendrocytes and neurons (e.g., Nrg1_Erbb3) were upregulated after SeNExo treatment. These results collectively support that treatment with SeNExo induced cell-cell communication between neurons and glial cells, which was conducive to ameliorating the injured microenvironment by inhibiting inflammation and astrogliosis while promoting neuronal survival.

Extension of the therapeutic effects of SeNExo in SCI model mice

Given that CNS injury includes both TBI and SCI, we were finally interested in examining the in vivo therapeutic effects of our SeNExo to SCI model mice (Figure 6A). After establishing a severe T10 contusive SCI model using the weight-drop method,⁴⁷ we initially assessed the ability of SeNExo (labeled with Alp) to accumulate in SCI lesions after i.v. injection. Ex vivo imaging over a time course revealed that the fluorescence signal of SeNExo gradually increased in the spinal cord area, with a peak observed at 12 h, indicating the ability of SeNExo to cross the blood-spinal cord barrier (BSCB, Figures 6B and S6A). For higher resolution, frozen slides of the spinal cord were prepared to evaluate the distribution of SeNExo. Similar to the TBI, the results showed that SeNExo were colocalized with IBA1, GFAP, and β3-tubulin at the lesion site, suggesting the uptake of SeNExo by microglia, astrocytes, and neurons (Figure 6C).

Figure 6.

Distribution and therapeutic effects of SeNExo in SCI model mice

(A) Experimental design for ex vivo imaging to evaluate BSCB penetration of SeNExo and establishment of another batch of SCI mode mice to evaluate the therapeutic effects treated with either PBS or SeNExo.

(B) Representative ex vivo fluorescence images of dissected spinal cords from SCI model mice (at 7 days) at the indicated time points after i.v. injection with Alp-labeled SeNExo. Fluorescence intensity was quantified.

(C) Representative fluorescence images of frozen spinal cord from SCI model mice at 12 h post injection with DIO-labeled SeNExo. Regions in the white squares (f1–f2, f1i–f1ii) were enlarged to show the colocalization of DIO-labeled SeNExo (green) with microglia (IBA1+, cyan), astrocytes (GFAP+, yellow), and neurons (NEUN/β3-tubulin+, red/gray) near the injury. Nuclei: Hoechst 33342 (blue). Low-magnification image scale bars, 200 μm. Scale bars in f1i–f1ii, 20 μm. Scale bars in f2, 20 μm.

(D) BMS score for the left and right hindpaws at the indicated post-injury time points for different groups.

(E) Representative pictures of mice in different groups (left) and corresponding representative images of limb footprints (right). Cyan: right front (RF); magenta: right hind (RH); yellow: left front (LF); green: left hind (LH).

(F) Representative pictures of RH and LH footprints and corresponding 3D footprint contact intensities in different groups.

(G) Quantitative analysis of stride length of LF and RF forefeet (E), mean print area of LH and RH footprints (F), and mean contact intensity of LH and RH footprints (F) to evaluate the extent of locomotion recovery.

(H) Representative two-photon fluorescence images of ROS levels at the lesion site and the quantitative analysis. Scale bars, 200 μm.

(I) Representative immunofluorescence of neuroinflammation markers IBA1 and GFAP at the lesion site. Low-magnification image scale bars, 500 μm; high-magnification image scale bars, 200 μm.

(J) Representative immunofluorescence of neuronal survival markers NEUN and NF200 at the lesion site. Low-magnification image scale bars, 500 μm; high-magnification image scale bars, 200 μm.

(K) Quantitative analysis of IBA1, NEUN, and the lengths of the lesions following various treatments.

Data in (B), (H), and (K) are presented as the mean ± SD from 3 biological replicates. Data in (D) and (G) are presented as the mean ± SD from 6 biological replicates. Data in (D) were assessed by two-tailed unpaired Student’s t test. Data in (G), (H), and (K) were assessed via one-way ANOVA.

To assess the therapeutic effects on locomotor function, the SCI model mice received PBS or SeNExo treatment (1.32 × 10¹¹ particles/kg), with a sham group as the healthy control. From days 0 to 56, the Basso Mouse Scale (BMS) was used to monitor overall locomotor performance.⁴⁸ Over time, the BMS scores of the SeNExo group increased more quickly than those of the PBS group. Consequently, the BMS scores of the SeNExo group (left: 5.17, right: 5) were higher than those of the PBS group (left: 2.67, right: 2.83) on day 56 (Figure 6D). Further CatWalk gait analysis showed that SeNExo treatment significantly extended the forefoot stride length and increased both footprint area and contact intensity (Figures 6E–6G).

Recalling the pathological features of CNS injury, the spinal cord tissues from the SCI model mice were excised for histological evaluation after gait analysis. Similar to the results in the TBI model mice, the SeNExo group showed significantly reduced ROS levels compared to the PBS group (Figure 6H). Moreover, these mice also showed a significant reduction in the number of reactive microglia, accompanied by a significant decrease in the levels of TNF-α and IL-1β (Figures 6I and S6B). Consequently, the neuronal loss and length of the lesion site in the SeNExo group were also obviously reduced (Figures 6J and 6K), which demonstrated the SeNExo’s excellent therapeutic effects on recovering locomotor function in SCI. The aforementioned potent therapeutic benefits were also achieved with few abnormalities in hematological parameters (Figure S6C), serum biochemical indicators (Figure S6D), and histology of the main organs (Figure S6E). These data demonstrated that our nano-selenized NExo strategy could be extended for the safe and efficient repair of SCI.

Discussion

Owing to the barrier between the blood and CNS, and the complex pathological features of CNS injury, it is desirable to develop therapeutics that can efficiently accumulate in CNS lesions and simultaneously ameliorate the injury microenvironment. Herein, we established a mild and facile strategy for the in situ crystallization of ultrasmall nano-selenium at NExo. Upon efficiently penetrating the barriers between the blood and the CNS (BBB and BSCB) to reach the lesion site, SeNExo effectively alleviated neuronal apoptosis, restored glia homeostasis, and remodeled glia-neuron networks. In murine models of TBI and SCI, SeNExo significantly improved cognition and locomotion, respectively, highlighting its potential as a promising therapeutic agent for CNS injury repair.

Given the crucial role of ROS in exacerbating CNS injury, efficient ROS reduction is an indispensable requirement for therapeutics. However, traditional aqueous-synthesized Se NPs have a size range of 40–400 nm, and relatively low surface-to-volume ratios limit its antioxidant activity to a higher level. By exploiting the interaction between phosphoryl groups on the NExo membrane and sodium selenite precursor, we succeeded in in situ growth of ultrasmall nano-selenium (∼3.5 nm) at the membrane. On the one hand, it appeared that phosphoryl groups functioned as stabilizers for selenite crystallization with ultrasmall sizes, leading to a substantially increased surface-to-volume ratio. On the other hand, phospholipids facilitated Se–O bond formation, enhancing free radical adsorption and electron transfer. Therefore, the ROS scavenging ability of SeNExo was nearly 4-fold than that of traditional Se NPs, paving the way for efficient scavenging of ROS and alleviation of oxidative stress and neuroinflammation in the lesions. Regarding the underlying mechanisms of such excellent performance, we proposed two manners according to previous investigations.^49,50 On the one hand, nano-selenium might directly engage in redox reactions with ROS, working in a self-sacrificing manner.⁴⁹ On the other hand, it might also act as a GPx-like enzyme in a self-renewing manner when the microenvironment was enriched with glutathione, glutathione reductase, and its coenzyme nicotinamide adenine dinucleotide phosphate.⁵⁰

In addition to nano-selenium, exosome components also deserve further discussion. Proteomics analysis indicated that APOE on NExo could bind to the LRP-1 receptor on vascular endothelial cells to facilitate BBB penetration. Throughout this process, the ultrasmall nano-selenium at the membrane does not disturb the aforementioned binding and can hitchhike to the lesion site. Beyond the targeting mechanism, we utilized a multi-omics approach to reveal the contributions of proteins and miRNAs from NExo to the amelioration of neuroinflammation and the enhancement of neuroprotection to a certain extent. For example, mmu-miR-26a-5p, which has been reported to inhibit inflammatory responses, was strongly associated with a reduction in inflammation-related genes (e.g., Cd81and Ptn) in glial cells. With regard to neuroprotection, the proteins in the vesicle had network interactions with upregulated neuroprotective DEGs (e.g., Pparg and Mef2c), downregulated apoptotic DEGs (e.g., Ubb and Xbp1), and downregulated miRNA-targeted (e.g., mmu-miR-335-5p) neuronal apoptotic genes (e.g., Olfm1 and Xbp1). These findings raise the possibility that we can engineer NSC parental cells to selectively augment the expression of these therapeutic components to strengthen the yet-higher targeting performance and therapeutic effects.

Regarding clinical translation, SeNExo also holds several superiorities that deserve emphasis. Compared with traditional stem cell therapy, which is always compromised by pathological conditions, our NExo chassis with stable biological components excluded the possibility of phenotypic alterations, thus ensuring therapeutic functions at the lesion site. Furthermore, the phosphoryl groups enabled nano-selenium to stably anchor at the NExo membrane, preventing disassociation during in vivo circulation. In addition, SeNExo can be lyophilized and rehydrated, making it suitable for storage and flexible applications.^22,51 Beyond biological, structural, and storage stabilities, its biocompatibility sourced from the natural trace element (Se) and endogenous cellular component (NExo) has paved the way for safe use. Note that our SeNExo was prepared via a mild and facile one-pot approach without organic solvents, which not only ensured the biological activities of the NExo components but also facilitated scale-up production.

Limitations of the study

In this study, SeNExo efficacy was demonstrated solely through i.v. injection. Intranasal administration is also worth investigating, as this non-invasive approach facilitates the delivery of nanoformulation to the CNS, potentially reducing liver retention while enhancing targeting. Furthermore, establishing a non-human primate model to evaluate the therapeutic effects of SeNExo is more suitable for assessing its clinical translational potential. Given that our SeNExo comprises a mixture of inorganic elements, proteins, and diverse miRNAs, the specific contributions of the signaling pathways activated by these components in neural repair remain to be fully elucidated.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Wei Wei (weiwei@ipe.ac.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The raw sequence data of snRNA-seq and small RNA-seq data have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2021) in National Genomics Data Center (Nucleic Acids Res 2022), China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA024534 for snRNA-seq and GSA: CRA024600 for small RNA-seq) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa.
• •The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium (https://proteomecentral.proteomexchange.org) via the iProX partner repository with the dataset identifier PXD062901 and PXD062903.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (grant nos. T2225021, U2001224, 21821005, and 32101120), Beijing Natural Science Foundation (JQ21027), and Shenzhen Science and Technology Program (JCYJ20220818102804009). We thank Chuanjie Zhang (Third Affiliated Hospital, Jinzhou Medical University) for his help in establishing SCI models and Pengyan Xia for cryo-TEM sample preparation.

Author contributions

W. Wei, G.M., H.T., and W.L. conceived and designed the study; W. Wang constructed the therapeutic agent and performed in vivo experiments and snRNA-seq analysis; G.L., P.G., Y.W., and C.L. performed in vitro experiments; Y.W. and J.Z. helped with the preparation of liposomes; S.L. and F.L. helped with in vivo experiments; G.L. and P.G. helped with histological analyses; D.Z. discussed the results and gave advice; H.Z., D.W., and M.Q. have made some efforts on the revised manuscript; W. Wei and W. Wang wrote the manuscript; and W. Wei, G.M., H.T., and W.L. further revised the manuscript.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-IBA1	Abcam	Cat# ab283346; RRID: AB_3065282
Anti-GFAP	Abcam	Cat# ab207165; RRID: AB_2924325
Anti-NEUN	Abcam	Cat# ab177487; RRID: AB_2532109
Anti-β3 Tubulin	Abcam	Cat# ab78078; RRID: AB_2256751
Anti-IBA1	Abcam	Cat# ab178847; RRID: AB_2832244
Anti-GFAP	Abcam	Cat# ab279289; RRID: AB_3698194
Anti-NF200	Proteintech	Cat# 60331-1-Ig; RRID: AB_2881440
Anti-APOE	Abcam	Cat# ab183597; RRID: AB_3331650
Anti-TSG101	Affinity	Cat# DF8427; RRID: AB_2841675
Goat Anti-Rat IgG H&L (Alexa Fluor® 647)	Abcam	Cat# ab150159; RRID: AB_2566823
Goat Anti-Rabbit IgG H&L (Alexa Fluor® 555)	Abcam	Cat# ab150078; RRID: AB_2722519
Goat Anti-Mouse IgG H&L (Alexa Fluor® 647)	Abcam	Cat# ab150115; RRID: AB_2687948
Goat Anti-Mouse IgG H&L (Alexa Fluor® 488)	Abcam	Cat# ab150113; RRID: AB_2576208
Anti-SOX2	Abcam	Cat# ab92494; RRID: AB_10585428
Anti-Nestin	Invitrogen	Cat# MA1-110; RRID: AB_2536821
Anti-MFGE8	Invitrogen	Cat# PA5-109955; RRID: AB_2855366
Anti-ITGB1	Abcam	Cat# ab183666; RRID: AB_3698195
Chemicals, peptides, and recombinant proteins
Sodium Selenite	Sigma-Aldrich	Cat# S5261
Ascorbic Acid	Sigma-Aldrich	Cat# A4544
S-nitroso-N-acetylpenicillamine	Sigma-Aldrich	Cat# N3398
DOPC	Sigma-Aldrich	Cat# P6354
Cholesterol	Solarbio	Cat# C8280
DSPE-PEG2000	Xi’an ruixi Biological Technology	Cat# R-1028-2K
3,3′-dioctadecyloxacarbocyanine perchlorate (DIO)	Sigma-Aldrich	Cat# D4292
Edaravone	Aladdin	Cat# P109105
Ganglioside GM1	Aladdin	Cat# G130562
L-Selenocystine (SeCys)	Aladdin	Cat# S640012
2′,7′-dichlorofluorescin diacetate (DCFH-DA)	Sigma-Aldrich	Cat# 4091-99-0
Premo™ cellular hydrogen peroxide Sensor	Thermo Fisher	Cat# P36243
5-(Diethoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide (DEPMPO)	Santa Cruz Biotechnology	Cat# sc-202132
5,5-Dimethyl-1-pyrroline-N-oxide (DMPO)	Santa Cruz Biotechnology	Cat# sc-202587
1,1-Diphenyl-2-picrylhydrazyl Free Radical (DPPH)	Aladdin	Cat# D273092
Al (III) phthalocyanine chloride tetrasulfonic acid	Frontier	Cat# AIPcS-834
Phenylmethanesulfonyl fluoride (PMSF)	Solarbio	Cat# P0100
Hoechst 33342	Solarbio	Cat# C0031
Acridine orange	Abcam	Cat# ab270791
Triton X-100	Solarbio	Cat# T8200
RIPA Lysis Buffer	Thermo Fisher	Cat# 89901
Percoll	Cytiva	Cat# GE17-0891-02
Papain	MCE	Cat# HY-P1645
N, N dimethylformamide	Sigma-Aldrich	Cat# 319937
Evans blue	Sigma-Aldrich	Cat# E2129
Dulbecco’s Modified Eagle’s Medium (DMEM)	Gibco	Cat# 6124090
Fetal bovine serum (FBS)	Gibco	Cat# 10099141C
Trypsin EDTA Solution B	VivaCell	Cat# C3532-0100
Penicillin-Streptomycin Solution	VivaCell	Cat# C3420-0100
Mouse Neural Stem Cell Complete Medium (serum free)	Cyagen, OriCell®	Cat# MUXNF-90011
Critical commercial assays
Mouse IL-1β ELISA KIT	Solarbio	Cat# SEKM-0002
Mouse TNF-α ELISA KIT	Solarbio	Cat# SEKM-0034
Micro Hydroxyl Free Radical Scavenging Capacity Assay Kit	Solarbio	Cat# BC1325
Superoxide Anion Activity Content Assay Kit	Solarbio	Cat# BC1295
8-Hydroxydeoxyguanosine (8-OHdG) ELISA Kit	Solarbio	Cat# SEKSM-0038
ElaBoXTM Mouse IgG (Total) ELISA Kit	Solarbio	Cat# SEKM-0098
CD11b Microbeads	Miltenyi Biotec	Cat# 130-093-636
Anti-ACSA-2 Microbead Kit	Miltenyi Biotec	Cat# 130-097-679
Neuron Isolation Kit	Miltenyi Biotec	Cat# 130-115-390
NSC astrogenic differentiation kit	Cyagen, OriCell®	Cat# MUXNX-90091
Deposited data
snRNA-seq data	This study	GSA: CRA024534
Small RNA-seq data	This study	GSA: CRA024600
Proteomics data	This study	PXD062901, PXD062903
Experimental models: Cell lines
bEnd.3 Mouse brain microvascular endothelial cells	Cell Resource Center, Peking Union Medical College (PCRC)	Cat# 3101MOUTCM40
HT22 hippocampal neuronal cells	Cell Resource Center, Peking Union Medical College (PCRC)	Cat# 3101MOUGNM47
BV2 mouse microglial cells	Cell Resource Center, Peking Union Medical College (PCRC)	Cat# 3101MOUGNM45
Primary Neural Stem Cell derived from C57BL/6 mice	Cyagen, OriCell®	Cat# MUBNF-01001
Experimental models: Organisms/strains
Mouse: C57BL/6, female	Vital River Laboratories	N/A
Software and algorithms
GraphPad Prism 9.0.0	GraphPad Software Inc.	https://www.graphpad.com/
ImageJ	ImageJ Software Inc.	https://imagej.nih.gov/ij/index.html
CatWalk XT 10.6	Noldus	https://noldus.com/catwalk-xt
Vectra software (v.2.0.7.1 version)	AKOYA BIOSCIENCES	https://www.akoyabio.com/phenoimager/instruments/vectra-3-0/
Cell ranger 5.0.0	10x Genomics	https://www.10xgenomics.com/
Omicsmart platform	Gene Denovo Biotechnology Co. Ltd.	https://www.omicsmart.com/#/
R 3.5.1	R Project	https://www.r-project.org
Monocle2 2.8.0	Qiu et al.42	https://github.com/cole-trapnell-lab
CellphoneDB 2.1.2	Efremova M et al.46	https://www.cellphonedb.org
Cytoscape (v3.8.2)	Shannon P et al.52	https://cytoscape.org

Experimental model and study participant details

Cell lines and culture

bEnd.3 Mouse brain microvascular endothelial cells, HT22 hippocampal neuronal cells, and BV2 mouse microglia were obtained from Cell Resource Center, Peking Union Medical College (PCRC). Primary Neural Stem Cells derived from C57BL/6 mice were obtained from Cyagen Biosciences (Guangzhou) Inc. bEnd.3, HT22 and BV2 cells were cultured in DMEM supplemented with 10% FBS and 1% of penicillin-streptomycin in a 5% CO₂ environment at 37°C. NSCs were cultured in Mouse Neural Stem Cell Complete Medium (OriCell MUXNF-90011, serum free) in a 5% CO₂ environment at 37°C. Cell line identity was validated using STR analysis. All cell lines were regularly tested negative for mycoplasma by PCR.

Animals

C57BL/6 mice (6–8 weeks, female) mice were obtained from Vital River Laboratories (Beijing, China). All mice were raised in a standard environmentally controlled room with relatively constant room temperature (21°C–23°C) and humidity (55% ± 5%) under a 12-12 h light-dark cycle. The Animal Ethics Committee of the Institute of Process Engineering thoroughly reviewed and approved all animal experiments (approval ID, IPEAECA2020072). This study was conducted in strict compliance with the Regulations for the Care and Use of Laboratory Animals and the Guideline for Ethical Review of Animals (China, GB/T 35892-2018).

Method details

Extraction of exosomes

The culture supernatant from NSCs was harvested and subsequently subjected to centrifugation at 300g for 10 min to eliminate cells, followed by centrifugation at 2,000 g for 10 min to eliminate any remaining cell fragments. The resulting supernatant was then subjected to further centrifugation at 10,000 g for 30 min at 4°C to remove any remaining debris. Subsequently, the final supernatant underwent ultracentrifugation at 110,000 g for 70 min, repeated twice, to isolate a pellet containing exosomes. These obtained exosomes, denoted as NExo, were reconstituted in PBS for subsequent utilization.

Preparation and characterization of SeNExo, SeLipo and Se NP

For SeNExo preparation, NExo (4×10¹¹ particles/mL) was incubated with a sodium selenite solution (0.75 mM, pH = 7.4) in PBS buffer at 0°C for 4 h. Then, we exposed the selenite-adsorbed NExo to ascorbic acid solution (4.5 mM, pH = 6.5) at 37°C for 12 h. The purified SeNExo was obtained by ultracentrifuged at 110,000 g for 70 min and the selenium content is 0.132 nmol/10⁸ particles by ICP-MS (iCAP RQ, Thermo Fisher Scientific).

In order to explain the growth mechanism of ultrasmall-sized nano-selenium at NExo, selenium synthesized on different substrates was implemented. For SeLipo, liposomes were synthesized first and then crystallization of nano-selenium at liposomes. Briefly, 1 mg of DOPC, 0.33 mg of cholesterol, and 0.33 mg of DSPE-PEG2000 were blended in chloroform/methanol (3:1, v/v, 20 mL). Subsequently, the mixture was evaporated to create a thin film, dried overnight under vacuum, and rehydrated in PBS buffer (1 mL). To achieve nanoscale liposomes (Lipo), a MiniExtruder (Avanti) was employed to gradually extrude the liposomes through filter membranes at 800 nm, 400 nm, and 200 nm. Lipo crystallized with nano-selenium was prepared in the same protocol as that of SeNExo. For Se NP (BSA), the protocol was the same as for the synthesis of SeNExo except that NExo was changed to 4 mg/mL BSA solution. For Se NP (H₂O), the protocol was the same as for the synthesis of Se NP (BSA) except that BSA was discarded.

We deposited the aqueous solution directly onto an ultra-thin carbon film (Electron Microscopy China) to prepare the samples for electron microscopy. TEM imaging (negative staining using uranyl acetate), element mapping (unstained), and EDS analysis (unstained) were implemented by a JEM-2100F microscopy.

Cryo-TEM was implemented to provide further evidence of the in situ crystallization of ultrasmall-sized nano-selenium at NExo. The ultrathin grid of cryogenically immobilized SeNExo was prepared by using Vitrobot. Cryo-TEM observations were conducted utilizing a FEI Tecnai G2 F20 X-TWIN Transmission Electron Microscopy.

For XRD investigations, the SeNExo was lyophilized and then operated on Empyrean high resolution powder X-ray diffractometer.

For XPS investigations, the SeNExo was lyophilized and then characterized on a ESCALAB 250Xi instrument (Thermo Fisher Scientific).

For EXAFS analysis, Lipo-Se, Se NP (BSA) and Se NP (H₂O) were lyophilized and detected with the Se foil as the reference. For EXAFS parameter fitting, the obtained XAFS data underwent processing in Athena (version 0.9.26) for background, pre-edge line and post-edge line calibrations. Fourier transformed fitting was then carried out using Artemis (version 0.9.26).⁵³

Evaluation of ROS scavenging activity

ROS scavenging abilities of SeNExo and Se NP with equivalent of selenium were performed by electron spin resonance (ESR) spectrometer (Bruker ElexSys E500 spectrometer). Hydroxyl radicals (⋅OH) and Superoxide anions (O₂^⋅-) were produced through the Fenton reaction and trapped by DMPO and DEPMPO, respectively. The nitrogen free radicals (⋅NO) were produced by SNAP under UV irradiation and trapped by DEPMPO. Equal amounts of Se in SeNExo and Se NP were added to each mixture, and changes in amplitude of the ESR spectra of ⋅OH-DMPO, O₂^⋅--DEPMPO, and ⋅NO-DEPMPO adducts demonstrated their scavenging abilities. For quantitative comparison, commercial free radical detection kits of hydroxyl free radical and superoxide anion were used. Se NP and various concentrations of SeNExo were added to each kit, and absorbance intensities were tested according to the manufacturer’s instructions. DPPH was used to assess the nitrogen free radical scavenging activity. DPPH was diluted to 0.05 mM, Se NP and various concentrations of SeNExo were added and then incubated at 37°C for 30 min. The absorbance value wes tested at 526 nm.

For a quantitative comparison the ROS-scavenging performance of SeNExo in vitro, we utilized a genetically encoded optical H₂O₂ sensor to detect the intracellular H₂O₂ content after treatment with PBS, Se NP, or SeNExo. Briefly, 2 ×10⁴ HT22 cells were plated in a 30-mm Petri dish and transduced with 80 μL sensor. After 48 h of culture, cells were incubated with H₂O₂ (100 μM) and treated with PBS, Se NP (equal Se dose), or SeNExo (2 ×10⁹ particles per well). After 12 h of culture, time-lapse imaging was performed using Nikon A1 confocal microscope with 400 nm and 488 nm lasers for excitation and 505–550 nm for emission. Images were captured every 15 s for a duration of 5 min. The background of the images was subtracted. The fluorescence intensity values were used to calculate 400/488 nm excitation ratios. Each dot represented the average ratio over a 5-min period (n = 3 independent replicates).

For comparison of antioxidant capacity between SeNExo and SeCys in vitro, we utilized flow cytometry with DCFH-DA probe to detect the intracellular H₂O₂ levels. Briefly, 4 ×10⁴ HT22 cells were seeded in a 24-well plate. After 24 h of culture, cells were incubated with H₂O₂ (100 μM) and treated with PBS, SeNExo (2 ×10⁹ particles per well), or NExo with SeCys (equal Se dose and equal NExo dose). After 12 h of culture, cells were harvested and stained with DCFH-DA (10 μM) in serum-free DMEM medium at 37°C for 30 min. After washing, the intracellular H₂O₂ levels of different groups were quantified using flow cytometry via Beckman Coulter (CytoFLEX LX).

Stability evaluation

To supervise the structural stability, SeNExo were dispersed in PBS containing 10% serum and kept at 4°C. The size and zeta potential were tested via nanoparticle tracking analysis (NTA) every 12 h. The changes of Se content of SeNExo were also investigated during the incubation. At different time points, SeNExo was digested with aqua regia for 3 h and then diluted to measure the Se content by ICP-MS.

To evaluate the stability of SeNExo after a lyophilization/rehydration cycle (48 h, R.T.), size, zeta potential, and Se content of SeNExo were tested. To further evaluate the shelf life, lyophilized SeNExo was stored at R.T. for 28 days. Then, size, zeta potential, Se content, and DPPH scavenging capacity were tested. To evaluate the biological stability, proteomic analysis and miRNA omics were implemented for fresh NExo and rehydrated SeNExo. We sent the samples to Bio-Tech Pack Technology Company Ltd. for sample preparation and label-free quantitative proteomic analysis or small RNA-seq.

Proteomic analysis

Protein (from fresh NExo, rehydrated SeNExo, or bEnd.3 cells) was isolated by the RIPA lysis method. Through the following processes: enzymolysis with filter assisted proteome preparation, desalination with C18 Cartridge, and freeze-drying, peptides were obtained. We further analyzed the prepared peptide samples using LC-MS/MS system (Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer, Thermo Fisher Scientific). The raw MS files were analyzed and queried against protein database based on the species of the samples utilizing MaxQuant (1.6.2.10).

Small RNA-seq

Exosomes RNA was isolated using Trizol extraction from samples. Quantification of RNA was implemented by Agilent Bioanalyzer small RNA assay kit (Novogene). The small RNA-seq library followed the PE150 sequencing scheme, and the quality of the sequencing library was assessed using fastqc. N-base excision, q20 filtration, and adapter excision at both ends of the sequence were carried out using fastp. Using the Ensemble release 102 Homo sapiens genome as a reference, we utilized the Bowtie short sequence alignment tool to align the Rfam library, eliminating non-coding RNAs (ncRNAs) such as rRNA and tRNA. Small RNA quantification was performed using miRDeep2, and differential expression analysis was conducted using DESeq2 (with duplicate samples) or EdgeR (without duplicate samples).

DFT calculations

We utilized the Vienna Ab initio Simulation Package to conduct density functional theory (DFT) calculations, employing the generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) formulation.^54,55 The ionic cores were described using projected augmented wave (PAW) potentials, while valence electrons were considered with a plane wave basis set having a kinetic energy cutoff of 450 eV. We introduced partial occupancies of Kohn-Sham orbitals through Gaussian smearing with a width of 0.05 eV. Self-consistency in electronic energy was attained when the energy change was less than 10⁻⁵ eV. In terms of geometry optimization, convergence was deemed reached when the energy change per Ångström dropped below 0.03 eV/Å. To sample the Brillouin zone, we employed 2 × 2 × 1 Gamma kpoints.

The adsorption energy (E_adsorption) of adsorbate A was defined as:

E_adsorption = E_A/surf – E_surf – E_A(g)

Here, E_A/surf represents the energy of adsorbate A adsorbed on the surface, E_surf is the energy of clean surface, and E_A(g) is the energy of an isolated A molecule in a cubic periodic box.

TBI model establishment

The TBI model was established on C57BL/6 mice using controlled cortex impact method by a brain impact instrument (Beijing Zhongshi Technology).³³ Briefly, after anesthetized with 1% pentobarbital sodium, the soft tissue and periosteum were bluntly dissected and the skull was exposed. A 3 mm diameter circular bone window was opened using a skull drill. Brain injury in the right hemisphere was induced using this device, with a 2 mm diameter tip, to compress the cortex to a depth of 0.8 mm at a velocity of 4 m/s and a duration of 250 ms. After the injury, the scalp wound was meticulously sutured, then the mice were kept warm on a heating blanket until they recovered from anesthesia. The same procedure was conducted in sham animals as TBI model mice except for the head trauma. All mice were put back into cages and raised under SPF level environment.

SCI model establishment

The SCI model was established on C57BL/6 mice using a weight-drop method by spinal cord striker (Beijing Zhongshi Technology).⁴⁷ Briefly, after administering 1% pentobarbital sodium anesthesia, the mice were secured on a sterile operating table. We made an incision in the skin and subcutaneous tissue to fully expose the spine. The spinal cord injury was induced at the T9-T10 segment using this device, with a 2 mm diameter tip, reaching a depth of 0.5 mm at a velocity of 4 m/s and a duration of 250 ms. After injury, we carefully sutured the subcutaneous tissue and skin wound and placed the mice on a heating blanket to maintain warmth until they recovered from anesthesia. Sham animals underwent an identical procedure, excluding the induction of spinal cord injury.

Evaluation of SeNExo’s blood-CNS barrier penetration capacity in vivo

The in vivo fluorescence imaging was implemented to assess the penetration ability of SeNExo. The fluorescent-labeled particles were prepared firstly. We labeled NExo with Alp by incubating NSCs with free Alp, so that the NExo secreted by NSCs would be carrying Alp inside (Alp-labeled NExo) and avoiding the risk of Alp leakage. The details were as follows. Alp (0.1 mg/mL) was co-cultured with NSCs in Mouse Neural Stem Cell Complete Medium in a 5% CO2 environment at 37°C, and the Alp-labeled NExo was obtained after the exosomes’ extraction method mentioned above. Alp-labeled SeNExo was prepared by in situ crystallization of nano-selenium at Alp-labeled NExo as the same protocol mentioned above. For Lipo, the obtained liposomes (16 mg/mL) were mixed with 0.5 mg/mL Alp and then gradually extruded through 800 nm, 400 nm, and 200 nm filter membranes using a MiniExtruder (Avanti) to produce Alp-labeled Lipo. In order to remove the excess free Alp, Nap-5 Column Sephadex (Cytiva, UK) was used as manufacturer’s instructions.

For TBI model, mice in the acute stage (0 days after injury) or recovery stage (7 days after injury) were randomly divided into three groups (n = 3 biological replicates) and received i.v. injection of Alp-labeled Lipo, NExo, or SeNExo (200 μL, 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg). Fluorescence imaging was taken at different time points utilizing an IVIS imaging system (Revvity IVIS Spectrum). Because the black fur of C57BL/6 mice could largely absorb fluorescence signals, we removed the fur covering the head to ensure the detection of fluorescence signals in the head region.

For SCI model, the mice (7 days after injury) were received i.v. injection of Alp-labeled SeNExo (200 μL, 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg) and sacrificed at different time points (n = 3 biologically independent mice per time point). The dissected spinal cords were imaged utilizing an IVIS imaging system. For comparison, the mice (7 days after injury) were randomly divided into three groups (n = 3 biological replicates) and received i.v. injection of Alp-labeled Lipo, NExo, or SeNExo (200 μL, 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg). At the time point of 12 h, the mice were sacrificed and imaged.

Investigation of the ex vivo distribution of SeNExo

For ex vivo imaging, the heart, liver, spleen, lung, kidney, and brain of the Alp-labeled Lipo, NExo or SeNExo-treated TBI model mice (0 days after injury) were dissected at 6 h and imaged utilizing an IVIS imaging system (Revvity IVIS lumina III). For observing the changes in distribution of SeNExo at different time points, SeNExo-treated TBI model mice were sacrificed at different time points and performed utilizing an IVIS imaging system.

For calculating the clearance rate of SeNExo from blood, the mean plasma concentration-time profile of SeNExo was plotted. Blood (50 μL) was collected at 5 min, 30 min, 1 h, 3 h, 6 h, 9 h, 12 h, 24 h, and 48 h following the injection (200 μL, 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg), and then diluted with 100 μL PBS. After removing blood cells using centrifugation, the Se content of plasma was quantified using ICP-MS. Then the clearance rate was calculated using the equation⁵⁶: Cl = Dose/AUC. Here, Cl represents the clearance rate, Dose represents the amount of drug administered, Area Under the Curve (AUC) represents the total exposure of the bloodstream to the drug over time, which is the integral of the concentration-time curve.

To investigate the stability of Se content in exosomes during in vivo circulation, we administered Alp-labeled SeNExo via i.v. injection in TBI model mice. We calculated the ratio between the changes of fluorescence intensity and the changes of Se content of plasma over different time points. Briefly, blood (50 μL) was collected at 5 min, 30 min, 1 h, 3 h, 6 h, 9 h, 12 h, and 24 h following the injection of Alp-labeled SeNExo (200 μL, 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg), and then diluted with 100 μL PBS. After removing red blood cells using centrifugation, the fluorescence intensity of the plasma was measured by microplate reader (Infinite M200, TECAN) and then the Se content of the plasma was quantified using ICP-MS.

For confirming the Se accumulated at the lesion site, the lesion tissues of PBS or SeNExo-treated TBI mice (200 μL, 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg) were dissected, minced, and homogenized in RIPA assay buffer. All procedures were carried out on ice. The lysates were immersed in the aqua regia for 24 h. The Se content was then measured using ICP-MS.

Exploring the mechanisms of BBB penetration by SeNExo

For in vitro BBB penetration, a well-established Transwell model was conducted as previously reported.³⁶ bEnd.3 cells were plated in the upper chamber of transwell plates at a density of 1×10⁴ cells/well (0.4 μm pore size, Corning), and then HT22 cells were plated in the bottom chamber at a density of 2×10⁴ cells/well. After the trans-endothelial electrical resistance of this model reached 150 Ω⋅cm², DIO-labeled Lipo, NExo, and SeNExo (2×10⁹ particles) were added into the upper chamber (The DIO-labeled Lipo, NExo, and SeNExo were purified by removing excess free dyes through Nap-5 Column Sephadex). For antibody blocking experiments, DIO-labeled SeNExo and APOE blocked DIO-labeled SeNExo (pre-incubating with 0.2 μM anti-APOE overnight, 2×10⁹ particles) were added into the upper chamber. After 24 h, bEnd.3 cells (the cell membrane was stained with Cy5.5-NHS, 5μg/mL) and HT22 cells (the nuclei were stained with Hoechst 33342, 1:100) were individually captured using CLSM (Nikon A1, Japan) and equipped with associated software (NIS-Elements AR 5.20.00). The internalization of the DIO-labeled Lipo, NExo, and SeNExo by bEnd.3 and HT22 cells was quantified using flow cytometry via Beckman Coulter (CytoFLEX LX).

For APOE blocking experiments, mice in recovery stage (7 days after injury) were randomly divided into two groups (n = 3 biological replicates) and received i.v. injection of Alp-labeled SeNExo and APOE blocked Alp-labeled SeNExo (pre-incubating with 0.2 μM anti-APOE overnight). Fluorescence imaging was taken at 6 h utilizing an IVIS imaging system after removing the fur from the mice’s heads. For TSG101 blocking experiments, mice in recovery stage (7 days after injury) were randomly divided into two groups (n = 3 biological replicates) and received i.v. injection of Alp-labeled SeNExo and TSG101 blocked Alp-labeled SeNExo (pre-incubating with 0.01 mg/mL anti-TSG101 overnight). Fluorescence imaging was taken at 6 h utilizing an IVIS imaging system (Revvity IVIS lumina III).

Characterization of SeNExo uptake by microglia, astrocytes, and neurons

For investigating the colocalization of SeNExo with cells at lesion site, histological evaluation was implemented. In order to enhance the fluorescence intensity of SeNExo and avoid interference from blood stasis self-luminescence at the lesion site, we labeled SeNExo with DIO. The DIO-labeled SeNExo was purified by removing excess free dyes through Nap-5 Column Sephadex. TBI model mice (after 6 h of i.v. injection of SeNExo) and SCI model mice (after 12 h of i.v. injection of SeNExo) were euthanized by intraperitoneal injection of an overdose of pentobarbital sodium and transcardially perfused with cold PBS followed by paraformaldehyde. Brains and spinal cords were taken out and embedded in 4% paraformaldehyde overnight at 4°C. Following dehydration in 10%, 20%, and 30% sucrose solutions, dissected tissues were frozen in OCT tissue compound on dry ice and sectioned into 10-μm serial slices. Subsequently, the slices were permeabilized with 0.3% Triton X-100 and blocked with goat serum for 2 h. They were then incubated overnight at 4°C with primary antibodies to label microglia, astrocytes, and neurons. The following primary antibodies were utilized to immunostain mouse tissues: anti-IBA1 (1:100, rat), anti-GFAP (1:200, rabbit), anti-NEUN (1:200, rabbit), anti-β3-tubulin (1:200, mouse). Then the slices were washed with PBST and incubated with secondary antibodies for 1 h at room temperature. The following secondary antibodies were utilized: Goat anti-rat IgG H&L (1:400, Alexa Fluor 647), Goat anti-rabbit IgG H&L (1:400, Alexa Fluor 555), and Goat anti-mouse IgG H&L (1:400, Alexa Fluor 647). After washing with PBST, the slices were incubated with Hoechst 33342 to mark Nuclei (1:100). Fluorescence images of all slices were captured by CLSM (Nikon A1, Japan) and equipped with associated software (NIS-Elements AR 5.20.00).

For investigating the changes of Se content in microglia, astrocytes, and neurons of TBI lesion tissues after treatment with SeNExo, sorting experiments were implemented using Microbeads as previous reported.^57,58,59 Briefly, lesion tissues from TBI model mice treated with PBS or SeNExo (pooled from six mice each) were dissected, minced, and digested using papain (2 mg/mL). The resulting cells were passed through a 40-μm strainer and mixed with a Percoll gradient to remove myelin by centrifugation (500 g, 30 min, 18°C, with minimal deceleration). Afterward, we sequentially isolated microglia (CD11b Microbeads), astrocytes (Anti-ACSA-2 Microbead Kit), and neurons (Neuron Isolation Kit) using the corresponding kits according to the manufacturer’s instructions. Following cell quantification, the cells were dissolved in aqua regia and the intracellular selenium content was measured using ICP-MS.

To investigate the uptake of SeNExo by microglia, astrocytes, and neurons, we co-cultured SeNExo with BV2, NSC-induced astrocytes, and HT22, respectively. We then observed the co-localization of SeNExo and acid organelles by CLSM and TEM. Note that astrocytes were induced from mouse NSCs by culturing them with mouse NSC astrogenic differentiation kit.

For CLSM imaging, the SeNExo was labeled with Alp. BV2, NSC-induced astrocytes, and HT22 cells were seeded at 2 ×10⁴ cells per petri dishes, respectively. After 12 h of culture (48 h for astrocytes), the cells were incubated with Alp-labeled SeNExo (2 ×10⁹ particles per well). After 24 h of incubation, the culture medium was removed and cells were washed with PBS. The cells were then stained with acridine orange (1:1000) and Hoechst 33342 (1:100) for 30 min. Following washing with PBS, the fluorescence images were captured by CLSM (Nikon A1, Japan) and equipped with associated software (NIS-Elements AR 5.20.00). Note that the acid organelles marked with acridine orange were captured using an excitation filter of 550 nm with an emission filter of 620 nm^60,61.

For TEM, in order to distinguish the SeNExo from other vesicles within the cells, SeNExo was labeled with gold nanoparticles (15 μg/mL in culture medium, Au-labeled SeNExo) using the same protocol as for Alp labeling. The gold nanoparticles were prepared using the citrate reduction approach.⁶² BV2, NSC-induced astrocytes, and HT22 cells were seeded at 2 ×10⁵ cells in 6-well plates. After 12 h of culture (48 h for astrocytes), cells were incubated with Au-labeled SeNExo (4 ×10⁹ particles per well). After 24h of incubation, the culture medium was removed and cells were washed with PBS for 3 times. The cells were then collected and fixed with 2.5% glutaraldehyde. The fixed samples were sent to Center of Biomedical Analysis, Tsinghua University, for TEM sample preparation. The images for the uptake by cells were captured by JEM-1400Flash TEM.

Dose escalation study

For a dose escalation study, the TBI model mice were divided into five groups including PBS, 0.5×, 1×, 2×, and 5× SeNExo group and the sham group treated with PBS as the healthy control (n = 6 biological replicates). Mice were i.v. injection with PBS, 0.5×, 1×, 2×, and 5× dose of 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg (200 μL) at indicated time points shown in Figure 3A. After four weeks of treatment, the spatial learning and memory abilities of mice were evaluated using MWM method. Then the mice were euthanized by intraperitoneal injection of an overdose of pentobarbital sodium and transcardially perfused with cold PBS. The lesion tissues were collected, minced, and homogenized in RIPA assay buffer containing 1 mM PMSF. All procedures were carried out on ice. Lysates were centrifuged at 10000 g for 5 min at 4°C to remove debris. Subsequently, the levels of mouse TNF-α and IL-1β were measured using ELISA kits.

Evaluation of the therapeutic effects of SeNExo on BBB recovery

For evaluation of BBB integrity, Evans blue staining method and IgG level detection of lesion tissues were conducted as previously reported.^37,38 Briefly, after one week of treatment, mice (n = 6 biological replicates) from different groups were i.v. injection with Evans blue solution (2%). After 40 min, mice were euthanized by intraperitoneal injection of an overdose of pentobarbital sodium and transcardially perfused with cold PBS. The dissected brains were harvested and photographed. Then, the lesion tissues were harvested, minced, and homogenized in N,N-dimethylformamide and centrifuged at 9000 rpm for 20 min. The supernatant was collected and optical absorption was measured at 620 nm for quantification of Evans blue. For mouse IgG estimation, the lesion tissues (n = 6 biological replicates) were collected, minced, and homogenized in RIPA assay buffer containing 1 mM PMSF. All procedures were carried out on ice. Lysates were centrifuged at 10000 g for 5 min at 4°C to remove debris. Subsequently, the levels of mouse IgG were measured by the ELISA kits.

Evaluation of the therapeutic effects of SeNExo in TBI model mice

The TBI model mice were divided into six groups randomly (n = 6 biological replicates) including PBS, Se NP, NExo, Se NP + NExo (mixture of Se NP and NExo), SeNExo, and GM1-EDR, and the sham group with PBS treatment was utilized as the healthy control. Mice were i.v. injection with a dose of 1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg (200 μL) at serial time points in Figure 3A except for mice in GM1-EDR group, which were intraperitoneal injection with 30mg GM1 equiv./kg and 5mg EDR equiv./kg.

For spatial learning and memory abilities assessment, the Morris water maze (MWM) test was performed to evaluate the spatial learning and memory abilities of mice in the 5^th week. The training process lasted 5 days, with 4 trials per day from four different quadrants, with each mouse searching for the platform in the middle of the third quadrant. The time of each trial was set to 90 s. If mice failed to reach the platform within 90 s, they were placed on the platform for 10 s to remember the site. After training 5 days, we removed the platform and each mouse entered the water from the first quadrant for a total of 60 s of swimming time. During 60 s, the number of crossing platform times and the searching time in the quadrant of the platform were recorded.

Investigating the mechanism of action of SeNExo

To investigate the mechanism of action of SeNExo, we employed intracerebral (IC) injection or intravenous (i.v.) injection to treat TBI model mice to control the distribution of SeNExo. The TBI model mice were divided into five groups including PBS (IC injection), SeNExo (IC injection), NExo (IC injection), SeNExo (i.v. injection), and blocked-SeNExo (i.v. injection) (n = 6 biological replicates). The blocked-SeNExo was prepared by pre-incubating SeNExo with anti-MFGE8 (1:100, rabbit), anti-APOE (1:100, rabbit), anti-ITGB1 (1:100, rabbit) overnight. For IC injection, a nanomite syringe pump and Hamilton syringe were used. 4 μL of SeNExo (1.32 × 10¹⁰ particles/mL), NExo (1.32 × 10¹⁰ particles/mL), or PBS was infused directly into the lesion site at a rate of 0.8 μL/min at 7 days post-injury. For i.v. injection, 200 μL of blocked-SeNExo (1.32 × 10¹¹ particles/kg and 175 nmol Se equiv./kg) was implemented at 7 days post-injury. After 24 h of treatment, mice were euthanized by intraperitoneal injection of an overdose of pentobarbital sodium and transcardially perfused with cold PBS. The lesion tissues (n = 6 biological replicates) were collected, minced, and homogenized in RIPA assay buffer containing 1 mM PMSF. All procedures were carried out on ice. Lysates were centrifuged at 10000 g for 5 min at 4°C to remove debris. Subsequently, the levels of 8-OHdG (a biomarker of oxidative stress) were measured using ELISA kits.

Evaluation of the therapeutic effects of SeNExo in SCI model mice

SCI model mice were randomly divided into two groups (n = 6 biological replicates) including PBS and SeNExo, and the sham group with PBS treatment was utilized as the healthy control. Mice were i.v. injection with a dose of 1.32 × 10¹¹ particles/kg (200 μL) at serial time points in Figure 6A.

For behavior assessment, the locomotory behaviors of SCI model mice were assessed and scored on days 0, 7, 14, 21, 28, 35, 42, 49, and 56 according to guidelines of the Basso Mouse Scale (BMS), with scores ranging from 0 to 9 (9 indicating complete normalcy, 0 indicating complete paralysis). The animals were put in an open field for 4-min observation. Three examiners who were unaware of the group of mice inspected and scored the left and right ankle joints, the degree of tactility of the soles and dorsum of the feet, trunk stability and tail position, respectively. This experiment was repeated 3 times.

For gait analysis, SCI model mice were evaluated using the Noldus CatWalk XT Automated Gait Analysis System in the 9^th week,⁶³ with the sham group serving as the control. Valid runs were defined as four uninterrupted runs per animal. The Catwalk software (CatWalk XT 10.6) recorded and analyzed movement track, stride length, hindlimb footprints, print area, and 3D hindlimb footprint intensities.

Two-photon imaging

After behavioral testing (35 days post-TBI or 58 days post-SCI), the ROS levels of lesion tissues were estimated. The same batch of TBI (n = 6 biological replicates) or SCI model (n = 3 biological replicates) mice with different treatment were i.v. injection with DCFH-DA (100 μM, 100 μL). After 30 min, the mice were euthanized via intraperitoneal injection of a lethal dose of pentobarbital sodium. Subsequently, transcranial perfusion with cold PBS was performed, and the levels of reactive oxygen species (ROS) in the brains or spinal cords were measured using two-photon confocal scanning microscopy (Olympus, FV1200MPE).

Estimation of inflammatory cytokines

For inflammatory cytokines (IL-1β and TNF-α) estimation, lesion tissues from brains (within a 3 mm × 3 mm × 3 mm cube, n = 6 biological replicates) and from spinal cords (within 5 mm segment centered on the lesion site, n = 3 biological replicates) were collected, minced, and homogenized in RIPA assay buffer containing 1 mM PMSF. All procedures were carried out on ice. Lysates were centrifuged at 10000 g for 5 min at 4°C to remove debris. Subsequently, the levels of inflammatory cytokines were measured by the ELISA kits.

Histological evaluation

For histological evaluation, dissected brains (n = 6 biological replicates) and spinal cords (n = 3 biological replicates) were embedded in 4% paraformaldehyde overnight at 4°C. According to the protocol described above, the serial tissue sections were incubated with primary antibodies and secondary antibodies. The following antibodies were used to immunostain brain tissues: anti-IBA1 (1:200, rabbit), anti-GFAP (1:200, rabbit), anti-NEUN (1:200, rabbit), and Goat anti-rabbit IgG H&L (1:400, Alexa Fluor 555). The following antibodies were used to immunostain spinal cord tissues: anti-IBA1 (1:200, rabbit), anti-GFAP (1:200, mouse), anti-NEUN (1:200, rabbit), anti-NF 200 (1:100, mouse), Goat anti-rabbit IgG H&L (1:400, Alexa Fluor 555), and Goat anti-mouse IgG H&L (1:400, Alexa Fluor 488). Fluorescence images of all slices were captured using a confocal laser scanning microscopy. Statistical analysis of histological evaluation experiments was conducted using ImageJ software.

Brain snRNA-seq analysis

TBI model mice were divided into two groups randomly including PBS and SeNExo, and the sham group with PBS treatment was utilized as the healthy control. Mice were i.v. injection with a dose of 1.32 × 10¹¹ particles/kg (200 μL) at serial time points in Figure 3A. After 28 days of treatment, mice were euthanized by intraperitoneal injection of an overdose of pentobarbital sodium and transcardially perfused with cold PBS. The dissected brains were harvested and flash frozen in liquid nitrogen, and then were sent to Gene Denovo Biotechnology Co. Ltd. for nuclei extraction and snRNA-seq (a pool of 3 biological replicates). A yield of 1000 intact nuclei/μL of sample was supposed to meet the snRNA-seq standard. After nuclei extraction, 20000 nuclei per group were run on a 10x Chromium system (10x Genomics). Libraries were generated and sequenced from single-nucleus cDNAs with Chromium Next GEM Single Cell 3′ Reagent Kits v3.1. Libraries were run on the NovaSeq 6000 for Illumina sequencing. In this study, we entrusted the Gene Denovo Biotechnology Co. Ltd. to perform GEMs generation, cDNA amplification, library preparation, and sequencing. Overall, 15116 cells from the sham group, 13623 cells from the PBS group, and 13318 cells from the SeNExo group passed the quality control threshold of more than 200 genes identified in each cell.

For pseudo-time analysis, we utilized Monocle2 (version 2.8.0) to reconstruct the development trajectory of microglia, OPCs, and oligodendrocytes.⁴² We assigned states using the DDRTree method.⁴² Genes that changed along the identified trajectory were detected using the differentialGeneTest function from the Monocle2 package. The “early” or “late” state of microglia were based on transcriptional signatures of homeostatic microglia (e.g., P2ry12 and Ptprk) and inflammation-stimulated microglia (e.g., Fth1 and Trem2).^41,43 OPCs was determined as precursors based on differentiation relationships established in previous studies.⁶⁴ For cell-cell communication analysis, we used CellphoneDB (version 2.1.2) method following a previous study.⁴⁶ Briefly, the candidate ligand_receptor pairs were acquired from a public repository of ligands, receptors and their interactions (https://www.CellphoneDB.org/). The ligand or receptor included in the downstream analysis should be expressed by more than 10% of cells in the specific cell cluster and characterized by a p-value less than 0.05.

Biosafety analysis

Blood samples from sham and SeNExo-treated TBI model mice or SCI model mice were collected for hematology to assess the safety of SeNExo. Major organs (heart, liver, spleen, lung, and kidney) were harvested and cut into 10-μm slices. Following hematoxylin and eosin staining, the slices were examined by using an automatic multispectral imaging system (Vectra III, AKOYA) with Vectra software (v.2.0.7.1 version).

Quantification and statistical analysis

All results are expressed as a mean ± standard deviation unless specified otherwise. Statistical analyses were conducted with GraphPad Prism 9.0.0. Statistical analysis was performed using one-way analysis of variance (ANOVA, comparison among more than two groups) or two-tailed Student’s t test (comparison between two groups). For the snRNA-seq analysis, cell clustering, UMAP visualization, heatmaps, Venn diagrams, Sankey plots, scatterplots, and pseudo-time trajectory visualizations were performed using the Omicsmart platform (http://www.omicsmart.com). Gene-protein interaction network analysis was visualized using Cytoscape (v.3.8.2).⁵²

Published: August 28, 2025