Abstract Oxidative phosphorylation (OXPHOS) in the mitochondrial inner membrane is a therapeutic target in many diseases. Neural stem cells (NSCs) show progress in improving mitochondrial dysfunction in the central nervous system (CNS). However, translating neural stem cell-based therapies to the clinic is challenged by uncontrollable biological variability or heterogeneity, hindering uniform clinical safety and efficacy evaluations. We propose a systematic top-down design based on membrane self-assembly to develop neural stem cell-derived oxidative phosphorylating artificial organelles (SAOs) for targeting the central nervous system as an alternative to NSCs. We construct human conditionally immortal clone neural stem cells (iNSCs) as parent cells and use a streamlined closed operation system to prepare neural stem cell-derived highly homogenous oxidative phosphorylating artificial organelles. These artificial organelles act as biomimetic organelles to mimic respiration chain function and perform oxidative phosphorylation, thus improving ATP synthesis deficiency and rectifying excessive mitochondrial reactive oxygen species production. Conclusively, we provide a framework for a generalizable manufacturing procedure that opens promising prospects for disease treatment. Subject terms: Mitochondria, Neural stem cells, Biotechnology __________________________________________________________________ Regulating oxidative phosphorylation and restoring redox homeostasis are crucial in neurological disorders. Here, the authors develop a top-down membrane self-assembly strategy to develop stem cell-derived artificial organelles (SAOs) that mimic mitochondrial oxidative phosphorylation without the risks associated with stem cell therapy. Introduction The mitochondrial oxidative phosphorylation (OXPHOS) system is the cornerstone of energy production in eukaryotic cells and comprises five enzymatic complexes and two mobile electron carriers within the mitochondrial inner membrane (MIM)^[42]1,[43]2. In terms of neurological function, the brain heavily relies on OXPHOS-generated adenosine triphosphate (ATP) to establish crucial electrochemical gradients and facilitate reliable synaptic transmission, which are essential for maintaining normal physiological functions. However, when mitochondrial metabolism is compromised, such as in cases of impaired OXPHOS activity, a cascade of detrimental effects ensues^[44]3–[45]6. Such detrimental effects include an imbalance in ATP production, heightened burden of reactive oxygen species (ROS), and compromised calcium buffering, ultimately leading to neuronal loss, characteristic of both acute and chronic degenerative neurological disorders^[46]7,[47]8. Insufficient OXPHOS and the resultant energy deficiency represent critical steps leading to neuronal death^[48]9. Addressing such energy metabolism imbalances poses a potential therapeutic approach for treating neurodegenerative conditions stemming from abnormal mitochondrial activity^[49]10,[50]11. Effective OXPHOS modulation ensures prompt energy provision crucial for neuronal recovery and mitigates excessive ROS generation, thereby shielding neurons against oxidative stress. Moreover, mitochondrial metabolism plays a pivotal role in driving cell fate transitions in the brain for certain crucial processes, including neurodevelopment and regeneration^[51]12. Boosting OXPHOS activity promotes neuronal differentiation in the developing cortex^[52]13,[53]14 and reduces neuronal death during acute brain injury. Similarly, enhancing energy metabolism improves neural tissue survival in severe spinal cord injury^[54]15, highlighting the potential of targeting mitochondrial metabolism to bolster regenerative capacity in central nervous system diseases. Despite these promising avenues, therapeutic interventions targeting OXPHOS still face complexities regarding its metabolic networks and precise activity regulation. Delivering specific factors that alter the intrinsic pathways of OXPHOS carries the risk of unintended consequences, including cell death^[55]16. Additionally, direct exogenous ATP or antioxidant provision has limited efficacy in restoring cellular metabolism homeostasis, as ATP does not permeate the cytomembrane well^[56]17, and antioxidants only provide temporary relief from excessive ROS. Hence, novel therapies that comprehensively regulate OXPHOS activity and restore redox homeostasis are urgently required to effectively address neurological disorders at their metabolic core. Organelle level regulation presents a potential therapeutic strategy as a candidate for natural and precise managers of diverse processes in cells. Hayakawa et al. demonstrated that astrocytes in the brain can transfer mitochondria to damaged neurons to accurately regulate oxidative stress-induced brain homeostasis imbalance^[57]18 Transfer of mitochondria from astrocytes to neurons after stroke. Palivan et al. created biomimetic artificial organelles capable of executing peroxidase activity triggered by reduction in microenvironment in vitro and in vivo, contributing to cellular homeostasis maintenance by regulating the redox equilibrium^[58]19. Miller et al. further developed an artificial organelle for environmental carbon sequestration via light-powered CO[2] fixation in a chloroplast mimic system, which could accurately perform complex biosynthetic tasks^[59]20. Chen et al. isolated a photosynthetic system, as a mimic organelle, which could precisely regulate the supply of key energy and metabolism in cells and exhibits promising clinical potential for degenerative disease treatment^[60]21. Several biochemical functions have been successfully reproduced by artificial organelles (functional unit) that may present an elegant strategy for precise regulation of cell activity in organelle level. Biomimetic membrane systems have previously been used as bioreactors, enabling the investigation of specific intracellular metabolic reactions using lipid vesicles in a simplified and isolated in vitro environment^[61]22–[62]24. Reconstituting a photoconverter and an ATP synthase into proteoliposomes enables ATP synthesis^[63]25. However, even the most sophisticated biomimetic structures struggle to rival the natural mitochondrial structure, and they still inherently lack the component variety and responsiveness found in the most basic mitochondrial structures. After elaborate evolution, the MIM perfectly houses the OXPHOS complexes and is constructed as a precise regulatory system to perform the function of respiratory chain electron transfer, ATP synthesis and participate in various core metabolic reactions in the body. A photosynthetic functional unit has been constructed by reforming natural chloroplast as a mimic organelle, which could independently precisely regulate the supply of key energy in cells^[64]21. Therefore, using MIM as a raw material to synthesize OXPHOS functional units is highly beneficial. Here, we proposed a systematic top-down design based on membrane self-assembly that could be used to develop neural stem-cell-derived oxidative phosphorylating artificial organelles (SAOs) for central nervous system (CNS), which has combinatorial characteristics of OXPHOS-like organelles and the intrinsic effectiveness of neural stem cells, offering as an alternative to neural stem cells. To overcome product heterogeneity and increase production process control, we constructed human conditionally immortal clone neural stem cells (iNSCs with c-mycER^TAM modification) as parent cells and used a streamlined closed operation system to prepare iNSC-derived highly homogenous SAOs and finally, we tried to prepare RVG29-modified SAOs which can specifically target CNS. We demonstrated that SAOs, through enrichment of OXPHOS complex proteins, carry out the function of OXPHOS in cells, which could comprehensively regulate levels of ATP to optimum concentrations, reduce ROS-induced oxidative stress-induced damage, maintain mitochondrial morphology, and promote damaged cell self-repair. This strategy can address key issues in the translation of stem cells into clinical applications, such as ethics, origin, and heterogeneity, while reducing the safety concerns of tumorigenicity and immune rejection caused by stem cells as living cells. The present work is generalizable and it opens up new avenues for organelle-based stem cell therapies, which could trigger a market transition: from molecular therapy to organelle therapy. Results Synthesis and characterization of SAOs A schematic illustration of SAO generation is shown in (Fig. [65]1a). We established three conditionally immortal clonal human NSC (iNSC) lines with a conditional immortalizing gene, c-mycER^TAM, to rapidly generate NSCs and provide a homogeneous cell population with consistent characteristics, to facilitate further cell engineering. After transfection of c-mycER^TAM, single viable NSCs were isolated using flow sorting and cloned to ensure that the cell population was genetically identical, thereby significantly reducing the heterogeneity of expression (Fig. [66]1b and Supplementary Fig. [67]1a). The iNSC characteristics were analyzed by detecting the presence of the c-mycER^TAM transgene and the neuroepithelial stem cell markers Nestin and SOX2 (Fig. [68]1c and Supplementary Fig. [69]1b). The iNSCs had normal human karyotypes (Supplementary Fig. [70]1c) and expanded more rapidly with 4-hydroxytamoxifen than without it (Fig. [71]1d, e). In the absence of growth factors and 4-hydroxytamoxifen, the cells underwent growth arrest and differentiated into neurons and neuroglia (Supplementary Fig. [72]1d). Fig. 1. Synthesis and characterization of SAOs. [73]Fig. 1 [74]Open in a new tab a Schematic of SAOs synthesis. b Morphology of iNSC during clone cell generation. c Immunoblot of c-MycER, Nestin, and SOX2 in NSC and iNSC, with GAPDH as a control. Quantification of proteins (n = 3). d EdU assay showing the effect of 4-OHT on iNSC (n = 3). e Cell confluence assay showing the effect of 4-OHT on iNSC (n = 4). ATP synthesis in control (non-mitochondrial NSC components), NSC mitochondria, and SAOs after stimulating pathways with complexes I, III, IV (f) and complexes II, III, IV (g) (n = 3). h ATP levels in HT22 cells treated with PBS (Control), low-dose SAOs (SAOs^L), or high-dose SAOs (SAOs^H) (n = 3). i Schematic of SAOs formation. Diameter (j) and zeta potential (k) of SAOs derived from different filters (n = 4). l Predictive model for protein yield based on cell input and SAOs output (n = 50). m Electron microscopy of SAOs. n Pie chart showing the distribution of membrane proteins in SAOs from various membranes. o Bar chart depicting the subcellular origin of total and high-abundance membrane proteins in SAOs. p Rank abundance curve of SAOs membrane proteins; black dots represent all proteins, red dots indicate proteins from specific subcellular locations. Proteomics analysis was performed using a sample size of three (n = 3) in (n–p). Experiments were in triplicate unless noted. Statistical analysis: two-sided t-test (c, d), and two-way ANOVA (e, f, g), one-way ANOVA (h). Error bars: mean ± s.e.m. Scale bars: 20 μm (b), 100 μm (d), 200 nm (m). d, e “4-OHT(-)” indicates iNSC without 4-OHT; “4-OHT(+)” indicates iNSC with 4-OHT. f, g “Control” indicates non-mitochondrial NSC components, “SAOs” indicates SAOs, and “Mitochondria” indicates NSC-isolated mitochondria. h “Control” represents untreated HT22 cells, “SAOs^L” represents HT22 with low-dose SAOs (5 μg per 1 × 10^6 cells), and “SAOs^H” represents HT22 with high-dose SAOs (15 μg per 1 × 10^6 cells). p “All” represents all detected proteins, and “High” represents high-abundance proteins detected via proteomics. Source data are provided as a [75]Source Data file. The first forms of life consisted of self-assembled molecular systems with specific sets of chemical and physical properties and relied on the self-assembly of membrane components for reproduction^[76]26,[77]27. Here, we disrupted the biological membranes structure of NSCs (particularly the IMM) using a mechanical extrusion in a high-pressure homogenizer and used the self-assembly property of the membrane components to promote the formation of stable SAOs abundantly enriched with OXPHOS complexes. We detected the activity of OXPHOS complexes in the SAOs and found that complexes I–V exhibited good biological activity (Supplementary Fig. [78]2a). Additionally, our findings demonstrate that SAOs contain cytochrome C (Supplementary Fig. [79]2b), a pivotal constituent of the electron transport chain. Subsequent examination of the synergistic action of OXPHOS complexes in SAOs upon the addition of exogenous reduction equivalents NADH and succinate illustrated the development of a membrane potential (Supplementary Fig. [80]2c, d) and successful ATP synthesis (Fig. [81]1f, g). To assess the functional relevance of SAOs within cells, we administered them to mouse hippocampal neurons (HT22) and observed a dose-dependent augmentation in neuronal ATP production (Fig. [82]1h). This observation supports that some SAOs may operate as functional units, utilizing host-provided substrates to execute OXPHOS processes. To validate this hypothesis further, we investigated the impact of exogenously supplemented SAOs on cellular function in the presence of inhibited mitochondrial OXPHOS. Notably, upon rotenone-induced inhibition of mitochondrial OXPHOS, we observed an increase in cellular oxygen consumption rate (OCR) (Supplementary Fig. [83]2e) and ATP synthesis (Supplementary Fig. [84]2f) following SAOs addition. These findings provide compelling evidence for the effective execution of OXPHOS functions by SAOs within host cells, thereby enhancing intracellular ATP synthesis. During the process of the membrane curling into spheres (membrane self-assembly), large amounts of NSC-derived content are encapsulated inside, eventually forming SAOs containing large amounts of proteins, long and short RNAs, lipids, and/or metabolites that are specific to the parent cells (Fig. [85]1i). Nanoparticle tracking analysis revealed that the self-assembled SAOs were homogenous with diameters between ~100 and 120 nm due to sequential extrusion through different size filters (Fig. [86]1j). The zeta-potential of the SAOs was approximately between –40 and –30 mV (Fig. [87]1k). The model for predicting the protein yield based on the amount of cell input and final number of SAOs was developed using multiple linear regression (Fig. [88]1l). Transmission electron microscopy revealed a “cup-shaped” morphology of the SAOs (Fig. [89]1m). Because SAOs consist of membranous components derived from parent cells, a cellular sub-localization analysis was conducted to examine the source of membrane proteins in SAOs from two distinct perspectives: structural and functional. From a structural standpoint, the analysis of proportion data regarding all detected membrane protein types within SAOs not only delineates their respective sources but also sheds light on the origin ratios of the identified membrane proteins. This dataset encompasses all protein types detected through proteomics. Notably, a hierarchical distribution of protein proportions within SAOs was observed, with the nucleus, mitochondria, cell membrane, and endoplasmic reticulum ranking from highest to lowest in abundance (Fig. [90]1n). Subsequently, the focus shifted to the functional aspect, where the types of high-abundance membrane proteins were examined. High-abundance proteins, defined by a relative Label-Free Quantification (LFQ) intensity exceeding 10^6, constitute the core functional clusters within SAOs, forming the basis of their functional protein composition. The analysis of high-abundance proteins within SAOs revealed that these proteins mainly originate from the mitochondria (Fig. [91]1o, p and Supplementary Fig. [92]2g). Overall, SAOs represent a mixture of fragmented biological membranes sourced from various origins, including the cell, mitochondrial, nuclear, endoplasmic reticulum, Golgi apparatus, lysosome, and peroxisome membranes. This mixture may arise either from the amalgamations of various fragmented membranes post-disruption or the fragmentation of a singular source membrane. Consequently, plasma membrane vesicles entrapping intracellular content, such as mitochondrial fragments, could represent a manifestation of the self-assembly process of SAOs. This process may also involve additional potential forms of self-assembly following membrane fragmentation, such as the generation of spherical vesicles through individual self-assembly subsequent to mitochondrial membrane disruption, or the fusion of fragmented membranes sourced from various origins, culminating in the formation of spherical vesicles through self-assembly. To confirm that the SAOs can be internalized by cells, we labeled the SAOs with a red fluorescent PKH-26 dye and incubated them with different target cells, including human umbilical vein endothelial cells, mouse hippocampal neuronal cells (HT22), and mouse microglial cells (BV2), and then confirmed the internalization of the fluorescently labeled SAOs by these cells using microscopic analysis (Supplementary Fig. [93]2h). There was a time- and dose-dependent uptake of SAOs in HT22 cells, with the process reaching saturation after 12 h (Supplementary Fig. [94]2i). SAO composition assessment using proteomic and miRNA analyses We systematically analyzed the proteins and miRNAs in the SAOs (Fig. [95]2a) to understand their biological characteristics, which are important determinants of the composition and function of the SAOs. Proteomics was used to analyze the protein composition of the SAOs, including the subcellular location, type, and function of the proteins, to clarify their formation mechanism. We compared SAOs with NSCs because SAOs are NSC-derived, and we wanted to determine the unique characteristics of SAOs compared to those of NSCs. Principal component analysis (PCA) and correlation analysis carried out to examine the global changes of proteins in SAOs and NSCs revealed an evident difference between SAOs and NSCs, and only a small difference in protein composition among SAOs of different origins, probably due to the membrane self-assembly preparation method (Supplementary Fig. [96]3a, b). Fig. 2. Omics analysis of SAO contents. [97]Fig. 2 [98]Open in a new tab a Schematic overview of the omics analysis process. b Protein profiles of SAOs and NSCs were categorized into five groups based on expression levels: NSCs-high, NSCs-only, SAOs-high, SAOs-only, and no difference. Pie charts display the distribution of these proteins across the plasma membrane, nucleus, cytoplasm, extracellular space, and other locations. c Bar chart showing the number of protein species from the nucleus, cytoplasm, and plasma membrane within the NSCs-high, NSCs-only, SAOs-high, and SAOs-only groups. d Left: Pie chart illustrating the distribution of cytoplasmic proteins in SAOs and NSCs among lysosomes, peroxisomes, Golgi apparatus, cytoskeleton, endoplasmic reticulum, mitochondria, and cytosol. Right: Bar chart depicting the number of protein species from these organelles in SAOs and NSCs. e Venn diagram comparing proteins identified in SAOs, NSCs, and the human mitochondrial protein databases. f Protein profiles of SAOs and NSCs were classified into twelve functional types, including transporter, cytokine, enzyme, G-protein coupled receptor, growth factor, ion channel, kinase, peptidase, phosphatase, transcription regulator, translation regulator, and transmembrane receptor. The radar map highlights the protein characteristics of SAOs and NSCs. g Bar chart showing the subcellular location of four types of proteins—ion channels, transporters, transcription regulators, and translation regulators—in SAOs and NSCs. h, Gene Ontology (GO) enrichment analysis of SAOs and NSCs, highlighting the top 8 significantly enriched biological process GO terms for target proteins. i Rank abundance curve of proteins in SAOs and NSCs; orange dots indicate all proteins, black dots represent those involved in oxidative reduction processes. j Heatmap displaying oxidative reduction process proteins in SAOs and NSCs. k Bar chart showing the distribution of oxidative reduction process proteins from mitochondrial and non-mitochondrial origins in SAOs and NSCs. l Functional analysis of proteins in SAOs. m Gene Set Enrichment Analysis (GSEA) showing enrichment of SAOs proteins in the OXPHOS pathway. n Rank abundance curve of SAOs and NSCs proteins; orange dots represent all proteins, black dots represent OXPHOS pathway proteins. Proteomic analysis was conducted with three samples per group (n = 3) in (b–n). Statistical analysis: one-sided Hypergeometric Test and Benjamini-Hochberg correction (l). To identify the subcellular origin of the proteins in the NSCs and SAOs, the protein location was divided into five categories: membrane, cytoplasm, nucleus, extracellular compartment, and other sites. Compared to NSCs, SAOs were enriched with more membrane and cytoplasmic proteins and fewer nuclear proteins (Fig. [99]2b, c). We performed a finer analysis of SAO components by further subdividing the cytoplasmic proteins into cytoplasmic free proteins and organelle proteins. Compared to NSCs, SAOs were enriched with more mitochondrial organelle proteins (Fig. [100]2d and Supplementary Fig. [101]3c); as shown in the Venn plot, 223 mitochondrial proteins were unique to the SAOs, far more than the 35 mitochondrial proteins that were unique to the NSCs (Fig. [102]2e). To identify the SAO-enriched protein types, the proteins were divided into 12 categories: cytokines, enzymes, G-protein coupled receptors, growth factors, ion channels, kinases, peptidases, phosphatases, transcription regulators, translation regulators, transmembrane receptors, and transporters. Compared to the NSCs, the SAOs contained more ion channels and transporter proteins and fewer transcriptional and translation regulatory proteins (Fig. [103]2f). We further verified whether this difference was due to subcellular protein origin by comparing the subcellular locations of the four protein types. Most of the ion channel and transporter proteins enriched in the SAOs were derived from cell membranes and mitochondria, while most of the transcriptional and translation regulatory proteins enriched in the NSCs originated from the nucleus and cytoplasm (free proteins) (Fig. [104]2g). Therefore, the differences in the enriched protein types between the SAOs and NSCs are due to SAOs enriching more mitochondrial and cell membrane proteins. Upon analyzing the function of the proteins within SAOs and NSCs, we found that although the main function of SAO and NSC proteins was oxidative-reduction processes (Fig. [105]2h), there were differences between them in the content of proteins involved in oxidation-reduction reactions (Supplementary Fig. [106]3d). Compared to NSCs, SAOs had a higher ratio of proteins participating in oxidative reduction and cell redox homeostasis (Fig. [107]2i, j and Supplementary Fig. [108]3e, f). We further verified whether this difference was caused due to protein subcellular origin, by comparing the subcellular location of the oxidative reduction proteins enriched in the SAOs and NSCs, and found that most of the SAO-enriched oxidative reduction proteins were mitochondria-derived, as compared to the NSC-enriched ones (Fig. [109]2k). Functional analysis of the SAO proteins revealed that their protein function mainly focused on energy metabolism, aerobic respiration, and nucleic acid metabolism (Fig. [110]2l). Gene set enrichment analysis identified that OXPHOS was the main enriched protein function pathway in the SAOs (Fig. [111]2m), and SAOs had a higher ratio of proteins involved in OXPHOS than NSCs (Fig. [112]2n). Based on these results, we confirmed that SAOs enriched with OXPHOS complexes could be prepared by membrane self-assembly preparation method. Considering that this preparation method enriches substances originating from mitochondria within SAOs, and recognizing mtDNA as a primary constituent of mitochondria, we conducted a whole exome sequencing (WES) to assess the presence of mtDNA within SAOs. The results of WES unveiled the presence of whole mtDNA sequences within SAOs (Supplementary Fig. [113]3g). Subsequently, we compared the highly expressed miRNAs enriched in the SAOs and NSCs using PCA and observed little difference in the enriched miRNA species (Supplementary Fig. [114]3h), which could be because miRNAs are primarily cytoplasmic and are less affected by the preparation method. Using clustering analysis of miRNAs in the SAOs, we found that the miRNA enriched by SAOs was mainly from the let-7 family (Supplementary Fig. [115]3i). Gene Ontology (GO) enrichment analysis of the high-expression miRNAs enriched in the SAOs found them to be mainly involved in the regulation of nervous system development (Supplementary Fig. [116]3j). Correlation analysis between miRNA and central nervous system diseases showed the high-expression miRNAs in the SAOs to be closely involved in the regulation of neurological diseases (Supplementary Fig. [117]3k). These results suggest that the SAOs are enriched with NSC-derived components involved in the regulation of the nervous system development and diseases. Validation of SAO efficacy in vitro After successfully fabricating the SAOs, we evaluated the effectiveness of SAOs against oxidative stress in vitro. We investigated the ability of SAOs to promote anastasis (recovery from the brink of cell death) using three models of oxidative stress-induced HT22 apoptosis, including the H[2]O[2], oxygen-glucose deprivation (OGD), and succinate-driven ROS models^[118]28(Fig. [119]3a). In the H[2]O[2] model, SAOs restored cell morphological alterations (Fig. [120]3b), reduced cell damage (Fig. [121]3c), and prevented H[2]O[2]-induced HT22 apoptosis (Fig. [122]3d). In the succinate-driven ROS (Supplementary Fig. [123]4a–d) and OGD models (Supplementary Fig. [124]4e, f), the SAOs reduced damage to HT22 cells and increased cell viability. We further observed SAOs could repair H[2]O[2]-induced human neuronal injury, including enhancements in neuronal morphology (Supplementary Fig. [125]4g), reduction in neurite damage (Supplementary Fig. [126]4h), and prevention of neuronal apoptosis (Supplementary Fig. [127]4i). Due to the abundance of mtDNA in SAOs and its therapeutic potential, we investigated whether efficacy of SAOs relies on their role as OXPHOS functional units or via mtDNA delivery. We used PCR and whole-genome sequencing (WGS) to evaluate the delivery and replication potential of mtDNA in cells. Our findings revealed that although SAOs delivered mtDNA to host cells, the concentration of mtDNA decreased over time and became undetectable after 36 h (Supplementary Fig. [128]5a–c). Subsequently, we observed that sustained therapeutic effects persisted even after using DNase I to degrade DNA within SAOs (Supplementary Fig. [129]5d–h), suggesting that the reparative efficacy of SAOs might function independently of mtDNA presence. Fig. 3. SAOs promote anastasis by enhancing mitochondrial function. [130]Fig. 3 [131]Open in a new tab a Schematic of in vitro experiments. b Morphology of HT22 cells. c Cell viability was assessed via lactate dehydrogenase (LDH) release assay (n = 3). d Apoptosis analysis of HT22 cells, with the percentage of cells in each quadrant indicated (n = 3). e Representative flow cytometry histograms of MitoSOX Green fluorescence in HT22 cells (n = 3). f Flow cytometry histograms of mitochondrial membrane potential measured by TMRM (n = 3). g Left: Oxygen consumption rate (OCR) after adding oligomycin (oligo), FCCP, and rotenone plus antimycin A (Rot + AA). Right: Baseline OCR, ATP production, and maximal respiration (n = 15 in normal and control group, n = 12 in SAOs group). h Electron microscopy (left) and proportion of class II mitochondria (right) (n = 12). i Cytoplasmic cytochrome c content measured by ELISA (n = 3). j Immunoblot of caspase 9, cleaved caspase 9, caspase 3, and cleaved caspase 3 in HT22 cells, with GAPDH as a loading control. k Quantitative analysis of caspase 9, cleaved caspase 9, caspase 3, and cleaved caspase 3 (n = 3). l Heatmap of differentially expressed genes in HT22. m PCA of gene expression in HT22. n Volcano plots representing OXPHOS-related genes between the SAOs group and the control group. o Gene Set Enrichment Analysis (GSEA) comparing gene sets involved in the TCA cycle, OXPHOS, mitochondrial translation, and fatty acid degradation between the SAOs and control groups. Transcriptomic analysis was performed using three samples per group (n = 3) in (l–o). All experiments were conducted in triplicate unless otherwise indicated. Statistical analysis: one-way ANOVA (c, d, e–i, k), two-sided Wald test and Benjamini-Hochberg correction (n). Error bars: mean ± s.e.m. Scale bar, 100 μm (b), 2 μm (h). b, o “Normal” represents untreated HT22 cells, “Control” represents HT22 cells exposed to 200 μM H[2]O[2], and “SAOs” represents HT22 cells treated with SAOs following 200 μM H[2]O[2] exposure. Source data are provided as a [132]Source Data file. Subsequently, we explored the impact of SAOs, acting as OXPHOS functional units, on enhancing mitochondrial function under oxidative stress conditions. The nanovesicle tracking results indicate that SAOs are efficiently internalized by HT22, with a fraction also entering the mitochondria (Supplementary Fig. [133]5i). Through assessment of mitochondrial function, we found that SAOs significantly reduced ROS levels in the mitochondria (Fig. [134]3e), restored the mitochondrial membrane potential (Fig. [135]3f). Oxygen consumption rate (OCR) assay in HT22 cells indicated that the SAOs could restore mitochondrial OXPHOS function (Fig. [136]3g). To explore the protective effect of the SAOs on mitochondrial structure, we examined mitochondrial morphology using transmission electron microscopy. Compared to those in the normal group, the mitochondria in the injured group showed noticeable swelling, with the disappearance of the mitochondrial cristae. In the SAO group, there was a significant reduction in mitochondrial swelling and restoration of the morphology of the mitochondrial cristae (Fig. [137]3h). A large amount of cytochrome c is involved in the OXPHOS reaction in mitochondrial cristae. Mitochondrial cristae damage causes the cytoplasmic release of cytochrome c to activate proapoptotic molecules and induce cell apoptosis. Therefore, we measured the cytoplasmic cytochrome c level and observed that it was reduced in the cells treated with SAOs compared to that in the control group (Fig. [138]3i). Since cytoplasmic cytochrome c activates the downstream apoptosis signal by cleaving caspase 9 and caspase 3, we detected the cleaved caspase 9 and caspase 3 levels and found them to be lower in the SAO group than in the control group (Fig. [139]3j, k). These results showed that SAOs can reduce the release of cytochrome c into the cytoplasm by maintaining the morphology of the mitochondrial cristae, thereby reducing the activation of apoptosis-promoting factors and, hence, reducing apoptosis. To explore the effects of SAOs on the transcriptome of neurons, transcriptomics analysis was performed on H[2]O[2]-induced HT22 cells treated with or without SAOs, to comprehensively determine the changes in the cells (Fig. [140]3l). Multivariate analysis (PCA) showed that genes in the normal and SAO groups were nearer in space and highly distinct from those in the control group (Fig. [141]3m). Compared to the control group, the SAO group showed upregulated expression of genes involved in OXPHOS (Fig. [142]3n). Consistently, gene set enrichment analysis identified that the SAOs group exhibited downregulated expression of genes associated with neuronal apoptosis and upregulated expression of genes involved in OXPHOS, tricarboxylic acid cycle, mitochondrial translation, fatty acid degradation and neural development (Fig. [143]3o and Supplementary Fig. [144]5j). Therefore, these results indicated that SAOs could reduce ROS-induced damage, improve OXPHOS, maintain mitochondrial morphology, and promote anastasis of impaired cells. Promoting neural regeneration is a crucial target for evaluating the effectiveness of SAOs in repairing neuronal damage. Consequently, we conducted comprehensive assessments to explore regulatory impact of SAOs on neural development. We found that SAOs can promote neural differentiation, maturation and function in control and oxygen-glucose deprivation (OGD) conditions. Morphological and LDH studies demonstrate that 24 h of SAOs treatment significantly restored neuronal morphology (Fig. [145]4a) and reduced neuronal damage (Fig. [146]4b). We also conducted transcriptome sequencing of neurons treated with SAOs, uncovering upregulation of genes linked to neurodevelopment and synaptic formation. Notably, NEUROD family genes and the synaptic-related gene SYN1 were among those upregulated (Fig. [147]4c), indicating that SAOs promote neuronal differentiation and axon formation. Additionally, pathways associated with neuronal differentiation, development, and axon formation showed significant upregulation (Fig. [148]4d). Subsequently, we assessed neuronal development at different time points post-treatment. We observed that the OGD SAO treated group exhibited elevated expression levels of TUJ1 and MAP2 (neuronal markers) at 21 days and PSD95 and synaptophysin (synaptic-related markers) co-localization at 42 days compared to the OGD group (Fig. [149]4e, f and Supplementary Fig. [150]5k). These findings suggest that neurons exhibit a greater degree of maturity in OGD SAO treated group compared to the OGD group. Furthermore, to evaluate neuronal maturation, we conducted an analysis of [Ca^2+][i] fluctuation in neurons. And calcium imaging revealed a higher frequency of spontaneous transients per unit of time in the OGD SAO treated group compared to the OGD group at 42 days (Fig. [151]4g), indicating enhanced neuronal maturation. Brain organoids, as pivotal models in the study of neural development, hold significant importance in neuroscience research. Following the construction of the brain organoids (Fig. [152]4h), we compared the uptake efficiency of SAOs by different cell types in brain organoids. The efficiency of SAOs penetrating neurons, neural progenitor cells (NPCs), and glial progenitor cells (GPCs) was assessed according to a previously published flow cytometry protocol^[153]29,[154]30. The results revealed a slightly higher uptake efficiency of PKH67-labeled SAOs by neurons compared to NPCs and GPCs (Fig. [155]4i). Subsequently, we utilized OGD brain organoids to investigate the impact of SAOs on brain development. Compared to the OGD group, the OGD SAO treated group exhibited an increased presence of well-defined layered structures, indicative of more typical processes of global brain development (Fig. [156]4j). The results of RNA-seq from the brain organoids OGD models further corroborated this finding, revealing significant upregulation of genes associated with neurogenesis and axon development in the OGD SAO treated group (Fig. [157]4k). Additionally, GSEA unveiled upregulation of gene sets associated with neuron generation and synaptic plasticity in the OGD SAO treated group (Fig. [158]4l). These finding reinforce the notion that SAOs facilitate brain development after OGD. In summary, SAOs not only prevent neuronal apoptosis but also promote the neural development and regeneration process. Fig. 4. SAOs promote neural regeneration. [159]Fig. 4 [160]Open in a new tab a Morphology of neuron cells. b Cell viability determined by lactate dehydrogenase (LDH) release assay (n = 3). c Volcano plots of genes related to neuron development, and synaptogenesis between OGD SAO-treated and OGD groups. d Gene set enrichment analysis (GSEA) comparing neuron development and axon development gene sets between OGD SAO-treated and OGD groups. e Immunostaining of TUJ1 in day 21 neurons (n = 12). f Immunostaining of MAP2 in day 21 neurons (n = 12). g Calcium transients recorded in day 42 neurons. Each panel shows [Ca^2+][i] traces from individual cells. X-axis: time in seconds (n = 10). h Morphology of hiPSC and brain organoid. i Evaluation of PKH67-labeled SAOs uptake efficiency in day 30 brain organoids analyzed by flow cytometry (n = 3). j Immunostaining of TUJ1 and SOX2 in day 30 brain organoids (n = 6). k Heatmap of differentially expressed genes in brain organoids. l GSEA comparing gene sets involved in neuron maturation and synaptic function between OGD SAO-treated and OGD groups. Transcriptomic analysis was performed with three samples per group (n = 3) in (c, d), and six samples per group (n = 6) in (k, l). All experiments were conducted in triplicate unless otherwise indicated. Statistical analysis: one-way ANOVA (b, e–g, i, j), two-sided Wald test and Benjamini-Hochberg correction (c). Error bars: mean ± s.e.m. Scale bars: 200 μm (a), 100 μm (e, f), 100 μm (h), 50 μm (j). a–g “Control” represents untreated neural cells, “OGD” represents neurons exposed to OGD, and “OGD SAO treated” represents neurons treated with SAOs following OGD exposure. j–l “Control” represents untreated brain organoids, “OGD” represents brain organoids exposed to OGD, and “OGD SAO treated” represents brain organoids treated with SAOs following OGD exposure. Source data are provided as a [161]Source Data file. SAOs perform anastasis by regulating the MAPK pathway To mimic the complex neural microenvironment in the brain, we selected a population of neural cells differentiated from hNSCs, including neural progenitor cells, neurons, and neural glial cells, to establish a model of H[2]O[2]-mediated oxidative stress. Subsequently, untreated neural cells were labeled as the normal group, while neural cells under oxidative stress were divided into two groups: one treated with SAOs (SAO group) and the other with PBS (control group). We employed unbiased high-throughput single-nucleus RNA sequencing (snRNA-seq) to explore the mechanisms underlying anastasis effects of SAOs by analyzing the transcriptional profiles of neural cells in different groups. We used an unsupervised clustering strategy to analyze the snRNA-seq data, resulting in cell segregation into ten distinct clusters (Fig. [162]5a). Considering apoptosis as the primary mechanism of oxidative stress-induced cell damage, we observed that SAOs could mitigate ROS-induced oxidative stress damage and reverse apoptosis in impaired cells. Consequently, we deemed it essential to annotate and analyze the apoptosis population separately in the snRNA-seq data analysis process. After analyzing the characteristics of different clusters, cluster 9 exhibited higher expression levels associated with stress response and programmed cell death than the other nine clusters (Fig. [163]5b). Consequently, cluster 9 was considered the apoptosis cluster. To better understand the SAO-specific repair mechanism for oxidative stress damage, we further categorized these ten clusters into five cell populations based on their respective states across different groups: stable state, H[2]O[2]-sensitive, SAO-sensitive, apoptosis, and others (Fig. [164]5c). Cells exclusively present in the normal group were designated as the stable state population, while those specifically emerging after H[2]O[2] treatment were labeled as the H[2]O[2]-sensitive population. Similarly, cells solely appearing after SAO treatment were categorized as the SAO-sensitive population, and those present across all three groups were defined as others. Fig. 5. SAOs perform anastasis via regulating the MAPK pathway. [165]Fig. 5 [166]Open in a new tab a Dimensionality reduction plot for visualizing single-nucleus RNA sequencing (SNS) data using UMAP of neural cells after exposure to H[2]O[2] with or without SAOs. b Gene set enrichment analysis was performed in cluster 9 and other clusters. c Dimensionality reduction for visualizing SNS data using UMAP of neural cells. d Gene set enrichment analysis was performed in apoptosis cluster and other clusters. e The morphology of human mix neural cells. f Representative images and quantification of TUNEL+ cells in neural cells (n = 12). g Cell viability was determined by lactate dehydrogenase (LDH) release assay (n = 3). h Immunoblot of ERK, p-ERK, P38, p-P38, JNK, p-JNK in neural cells. GAPDH was used as a loading control. i Quantitative analysis of ERK, p-ERK, P38, p-P38, JNK, p-JNK (n = 3). Transcriptomic analysis was performed using a sample size of three (n = 3) for each group in (a–d). All experiments were conducted in triplicate unless otherwise indicated. Statistical analysis, one-way ANOVA (f, g, i). Error bars: mean ± s.e.m. Scale bar, 200 μm (e), 100 μm (f). b The “Cluster 9” group represents Cluster 9 in snRNA-seq dataset, and the “Other clusters” group represents cluster 0-8 in snRNA-seq dataset. c The “Normal” group represents untreated neural cells, the “Control” group represents neural cells exposed to 200 μM H[2]O[2], and the “SAOs” group represents neural cells treated with SAOs following exposure to 200 μM H[2]O[2]. d The “Apoptosis” group represents apoptosis cluster in snRNA-seq dataset, and the “Other clusters” group represents stable state, H[2]O[2] sensitive, SAOs sensitive, and others clusters in snRNA-seq dataset. e–i The “Normal” group represents untreated neural cells, the “Control” group comprises neural cells exposed to 200 μM H[2]O[2], and the “SAOs” group consists of neural cells treated with SAOs following exposure to 200 μM H[2]O[2]. The “mSIRK” group consists of neural cells treated with SAOs following exposure to 200 μM H[2]O[2] and 10 μM mSIRK (ERK1/2 activators). Source data are provided as a [167]Source Data file. To elucidate the mechanism by which SAOs reverse apoptosis, we conducted GSEA of apoptosis clusters. Our findings unveiled a pronounced activation of both the MAPK pathway and phosphokinase pathway within apoptosis clusters, starkly distinct from the other four clusters (stable state, H[2]O[2] sensitive, SAOs sensitive, and others) (Fig. [168]5d). Specifically, the upregulation of the MAPK pathway was unique to the apoptosis cluster (Supplementary Fig. [169]6). A previous study linked MAPK to neuronal apoptosis in an oxidative stress model and found that H[2]O[2] significantly activated and upregulated the expression of phosphorylated (p)-extracellular signal-regulated kinase 1/2, p-c-Jun N-terminal kinase, and p-p38^[170]31–[171]33. To validate this hypothesis, we designated the untreated neural cells group as the normal group and subjected the H[2]O[2]-mediated oxidative stress model of neural cells to treatment with either PBS (control group) or SAOs (SAOs group). We observed that SAOs effectively attenuated oxidative stress-induced cell injury (Fig. [172]5e–g) and downregulated the expression of p-extracellular signal-regulated kinase 1/2, p-c-Jun N-terminal kinase, and p-p38 (Fig. [173]5h, i) compared to the control group. This indicates that SAOs have a protective effect against oxidative stress in neural cells, as evidenced by reduced cellular damage and decreased phosphorylation levels of key MAPK pathway components. The therapeutic effect of SAOs was reversed by mSIRK (ERK1/2 activators), leading to the increase of neural cell damage and apoptosis, further verifying the specificity of SAOs in regulating the ERK-JNK-P38 pathway (Fig. [174]5e-i). These results suggested that SAOs protect neural cells against H[2]O[2]-induced oxidative injury by inhibiting the phosphorylation of ERK-JNK-P38 pathway. Validation of SAO efficacy in vivo We developed a strategy to modify SAOs with RVG29 for brain targeting (Supplementary Fig. [175]7a). Upon injection into the tail vein of mice, RVG29-modified SAOs displayed significantly greater distribution in the brain compared to unmodified SAOs, confirming the success of our targeting approach (Supplementary Fig. [176]7b). Consequently, all subsequent in vivo experiments utilized RVG29-modified SAOs. Following intravenous injection, the distribution of SAOs across various organs revealed the brain as the site with the highest concentration, displaying a characteristic behavior typical of SAOs (Supplementary Fig. [177]7c). Additionally, when injected into the mouse striatum, SAOs demonstrated a distinct outward granular expansion (Supplementary Fig. [178]7d), with uptake observed not only by neurons but also by astrocytes, oligodendrocytes, microglia, and endothelial cells (Supplementary Fig. [179]7e). In our in vivo investigation, we investigated the therapeutic efficacy of SAOs in an ischemic stroke model of middle cerebral artery occlusion (MCAO)^[180]34 (Fig. [181]6a). The regional cerebral blood flow was used to detect the degree of ischemia before reperfusion (Supplementary Fig. [182]7f). We performed single-nucleus sequencing analysis to evaluate the reparative capacity of SAOs on neural cells in the MCAO model, with the model serving as the control and the control group treated with SAOs as the SAO group. We analyzed the transcriptomes of 23,499 single-cell nuclei (11,150 and 12,349 from the control and SAO groups, respectively) derived from the brains of three rats in each group. Following initial quality control steps, including removal of cells likely of low quality due to debris, doublets/multiplets, and death, we refined our analysis and identified a total of 20,502 single-cell nuclei (i.e., 9802 and 10,700 from the control and SAO groups, respectively) for further analysis. Next, we performed a principal component analysis of normalized read counts followed by dimensionality reduction using t-distributed stochastic neighbor embedding (t-SNE). Subsequently, we annotated the cell types using SingleR and canonical marker genes for neural cells. Specifically, we categorized the cells into six major lineages: neurons (Syt1, Map2, Gad1, and Ahcyl2), astrocytes (Gja1, Aqp4, and Bag3), oligodendrocytes (Mobp, Plp1, Ugt8, and Apod), microglia (Lyz2, Cd74, Csf1r, Cx3cr1, and P2ry12), endothelial Cells (Eng, Vwf, and Flt1), and fibroblasts (Dcn, Col1a1, and Pdgfrb). Our findings showed that compared to the control group, SAOs significantly promoted neuronal and brain development, myelination capacity of oligodendrocytes, support role of astrocytes in synaptic function, and endotheliocyte regeneration (Supplementary Fig. [183]7g–i). This finding indicates that SAOs promote regeneration and repair of different cell types in the brain after MCAO. Meanwhile, snRNA-Seq analysis revealed that SAOs enhanced OXPHOS levels in neurons, astrocytes, oligodendrocytes, microglia, and fibroblast and restored cellular metabolic equilibrium (Fig. [184]6b and Supplementary Fig. [185]7j). This was further validated by the OCR assay and suggested that the SAOs could successfully improve mitochondrial OXPHOS in the MCAO rat brain tissue (Fig. [186]6c). Magnetic resonance imaging (MRI) revealed that the SAOs reduced edema volume (Fig. [187]6d). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining, carried out to assess the neural cell apoptosis in brain, revealed that the SAO group had a significantly reduced number of TUNEL^+ cells compared with that in the control group (Fig. [188]6e, f). Furthermore, our findings revealed that SAOs decreased neuronal apoptosis (Fig. [189]6g) and astrocyte activation while inhibiting the formation of glial scars within the peri-infarct cortex region (Fig. [190]6h). Immunoblot assay indicated that the SAOs significantly inhibited activation of caspase 3 (Supplementary Fig. [191]7k). TUNEL assay and immunoblot both suggested that the SAOs attenuated neural cell apoptosis. Triphenyl tetrazolium chloride staining, carried out to evaluate the extent of infarct lesions (Fig. [192]6i), showed that the SAOs significantly decreased the brain infarct volume. We subsequently conducted further assessment of MCAO rat neurological function, revealing a significant reduction in the modified neurological severity scores of the SAO group on days 1, 2, and 3 compared to those of the control group (Fig. [193]6j). These findings suggest that the SAOs alleviated the neurological deficit symptoms in rats. Fig. 6. SAOs improve MCAO rat brain injury. [194]Fig. 6 [195]Open in a new tab a Schematic of in vivo experiments. b Gene set enrichment analysis of OXPHOS-related gene set. c Oxygen consumption rate in MCAO rat brain tissue (n = 15). d MRI images and quantification of brain infarction (n = 9 in control group, n = 11 in SAOs group). e Representative images and quantification of TUNEL+ cells in the infarct area, including cortex and f striatum (n = 12). g Immunostaining of TUNEL+ neurons (n = 12). h Immunostaining of GFAP in the peri-infarct cortex (n = 12). i Representative images of TTC staining (n = 13 in control group, n = 10 in SAOs group). j Quantification of neurological scores (n = 21 in control group, n = 17 in SAOs group). k Open field test and trajectories (n = 6 in normal group, n = 12 in control group, n = 12 in SAOs group). l Grip strength test (n = 3 in normal group, n = 11 in control group, n = 14 in SAOs group). m Adhesive test (n = 3 in normal group, n = 11 in control group, n = 11 in SAOs group). n Rotarod test (n = 3 in normal group, n = 11 in control group, n = 11 in SAOs group). o Cylinder test (n = 3 in normal group, n = 11 in control group, n = 13 in SAOs group). p Morris water maze. Upper panel: escape latency and time ratio in target quadrant. Lower panel: representative swim paths during the probe trial (n = 4 in normal group, n = 11 in control group, n = 13 in SAOs group). q Gross appearance of rat brain immediately after extraction (n = 3). r HE staining of rat brain (n = 3). Scale bars: 3 mm (d, i, q, r), 100 μm (e–h). Transcriptomic analysis was performed with three samples per group (n = 3) in (b). Experiments were conducted in triplicate unless indicated. Statistical analysis: one-way ANOVA (c, k, p), two-sided t-test (d–i, r), two-way ANOVA (j, l–o). Error bars: mean ± s.e.m. b–q “Sham” represents sham surgery rats, “Control” represents MCAO rats, and “SAOs” represents MCAO rats treated with SAOs. Source data are provided as a [196]Source Data file. To further assess the long-term therapeutic effects of SAOs on the MCAO model, we performed motor behavior tests, including open-field, grip-strength, rotarod, adhesive, and cylinder tests, for up to 14 days after treatment of the MCAO model with SAOs. We observed that SAO treatment significantly improved the motor ability of the rats, evaluated by means of the open-field test on day 1, as compared to that observed in the control group (Fig. [197]6k). SAOs also improved forelimb strength based on the grip-strength test (Fig. [198]6l), and the rats in the SAO group displayed significant improvement in the adhesive test at as early as 3 days after treatment, thus indicating early protective effect against MCAO-induced brain injury (Fig. [199]6m). Rats administered SAOs showed improved performance in the rotarod test compared to the performance of those in the control group (Fig. [200]6n). Similar results were recorded in the cylinder test, in which animals injected with the SAOs showed reduced bias (Fig. [201]6o). To study the reparative effect of the SAOs on the cognitive ability of rats, the Morris water maze test was conducted on days 23–28. The SAO-treated group displayed shorter escape latency and distance moved and spent longer time in the target quadrant (Fig. [202]6p). Consistent with the behavioral results, SAO treatment also significantly reduced brain tissue loss at 29 days after treatment of the MCAO model (Fig. [203]6q), as examined using hematoxylin and eosin staining (Fig. [204]6r). Collectively, these data indicated that the SAOs elicited a long-term protective effect against MCAO. Validation of SAO safety We conducted a safety assessment of the SAOs, including pro-tumorigenicity, immunoreactivity, and toxicity tests, to promote their clinical translation (Fig. [205]7a). We evaluated effect of SAOs on glioma by administrating a single dose to glioma-bearing mice, while control mice received PBS instead of SAOs. Stereotactic administration of high-dose (5 times the therapeutic dose) SAOs to mice for 4 weeks did not alter the glioma volume between the SAO and control groups, as determined using MRI (Fig. [206]7b). We also investigated the effects of SAOs on the in vitro proliferation, migration, invasion, and transcriptomic changes with three levels of malignancy (T98G, HS683, and U87). While we observed successful uptake of SAOs by glioblastoma cells (Supplementary Fig. [207]8a), there was no indication that SAOs facilitated tumor cell migration, invasion, or proliferation (Fig. [208]7c–e and Supplementary Fig. [209]8b, c). Additionally, SAOs did not significantly modulate anaerobic glycolysis, a primary energy production pathway in glioblastoma cells, across the cell lines, despite observed OXPHOS activation in T98G cells during incubation with SAOs (Supplementary Fig. [210]8d). Fig. 7. Validation of SAOs safety. [211]Fig. 7 [212]Open in a new tab a Schematic of in vitro and in vivo experiments. b, MRI images and quantification of tumors before and after SAOs administration (n = 6). c Migration ability of HS683 cells, determined by a scratch test (n = 3). d Invasive ability of T98G, U87, and HS683 cells, determined by transwell assay (n = 3). e Cell proliferation of T98G, U87, and HS683 cells, determined by CCK-8 assay (n = 3). f Immunotyping analysis of mouse peripheral blood (n = 16). g Serum Luminex multiplex assay showing CXCL1, IFN-γ, and TNF-α levels in mice (n = 8). h Proliferation of human PBMCs, analyzed via CCK-8 assay (n = 3). i Organ coefficients (liver, kidney, spleen, heart, lung, brain) of mice (n = 8). j Mouse weight (n = 8). k Routine blood and chemistry tests of mice (n = 8). l Representative H&E-stained tissue sections of mice injected with SAOs (n = 8). All experiments were conducted in triplicate unless otherwise indicated. Statistical analysis: two-way ANOVA (b, c, e, f, h, j), two-sided Mann-Whitney U test (g), two-sided t-test (d, i). For data in (k), a two-sided t-test was used if normality criteria were met; otherwise, the two-sided Mann-Whitney U test was applied. Error bars: mean ± s.e.m. Box plot: minima (lower whisker), maxima (upper whisker), center (median), bounds of the box (25th and 75th percentiles), whiskers (range from minima to maxima) in (f–h). Scale bars: 1 mm (b), 200 μm (l). b “Control” represents glioma-bearing mice with PBS, and “SAOs” represents glioma-bearing mice with SAOs. c–e “Control” represents untreated glioma cells, and “SAOs” represents glioma cells treated with SAOs. h “Control” represents untreated PBMCs, “PHA” represents PBMCs with PHA, and “SAOs” represents PBMCs with SAOs. f, g, i–k “Control” represents mice injected with PBS, and “SAOs” represents mice injected with SAOs. Source data are provided as a [213]Source Data file. We evaluated the immunoreactivity associated with intravenous administration of high-dose SAOs to C57 mice, on days 1, 3, and 7, through immunotyping analysis of peripheral blood and immune organs. The control mice were administered PBS. Differences in lymphocyte subtypes in the peripheral blood were observed between the two groups on days 1 and 3, which could have resulted from xenotransplantation of the SAOs, but these differences disappeared on day 7 (Fig. [214]7f). Apart from an increased number of CD3^+CD4^+ T cells in the spleen and decreased number of natural killer cells in the thymus of the SAO group, no evident differences in immunotyping were observed in the spleen, thymus, or bone marrow between the two groups, on day 7 (Supplementary Fig. [215]9). Circulating (serum) ICXCL1, interferon-γ, and interferon-α levels in mice on day 7, evaluated using Luminex® Multiplex Assays, were not significantly different between the two groups (Fig. [216]7g). We evaluated the potential side effects associated with grafting SAOs at a therapeutic dose into the left striatum of MCAO rats through immunotyping analysis of peripheral blood and immune organs on days 1 and 3. The control rats were administered PBS. Differences in lymphocyte subtypes in the peripheral blood were found between the two groups on day 1, which may have resulted from the SAO xenotransplantation and/or the disease state, but these differences disappeared on day 3 (Supplementary Fig. [217]10). In addition to the decreased number of CD4^+CD8^+ T cells and NKT cells in the bone marrow, decreased number of CD4^+CD8^+ T cells in the spleen, and decreased number of CD3^+ T cells in the thymus observed in the SAO group on day 3, no evident differences in immunotyping were observed in the spleen, thymus, or bone marrow between the two groups (Supplementary Fig. [218]11). We also examined the potential impact of the SAOs on human immune cells and their effects on leukocyte proliferation. Differences in leukocyte proliferation were observed between the two groups on day 3, but these differences disappeared on day 7 (Fig. [219]7h). We evaluated the potential toxicity associated with the intravenous administration of high-dose SAOs to mice. Control group mice were administered PBS. We initially assessed the metabolism of vital organs in mice, including heart, liver, spleen, lung, kidney and brain, 24 h after SAOs administration. The results of RNA-seq revealed that in brain tissue, SAOs notably enhanced metabolic processes, such as OXPHOS, compared to that in the control group (Supplementary Fig. [220]12a). Conversely, no significant activation of OXPHOS was observed in the other tissues following SAO administration. Subsequently, we conducted experiments on mice 2 months toxicity evaluation after the caudal intravenous administration of SAOs. Hematological and chemical analyses of the blood, weight, and organ coefficients and an in-depth histopathological evaluation of several different tissues did not reveal abnormalities upon treatment with SAOs. Weight and organ coefficients, including those of the heart, lungs, kidney, testis, liver, and spleen, were similar in both groups (Fig. [221]7i, j). No obvious differences in albumin, alkaline phosphatase, alanine aminotransferase, aspartate aminotransferase, blood urea nitrogen, globulin, and total protein levels or white and red blood cell counts were found between the groups (Fig. [222]7k and Supplementary Fig. [223]12b). Histopathological findings revealed no obvious abnormalities in the liver, kidneys, lungs, brain, mesentery, and spleen (Fig. [224]7l). We evaluated the potential side effects associated with grafting SAOs at a therapeutic dose into the left striatum of MCAO rats for 29 days. Control MCAO rats were administered PBS. No obvious potential side effects in organ coefficients were observed in the SAO group (Supplementary Fig. [225]12c). Upon hematological and chemical analyses of the blood in the SAO group, in addition to white blood cell, lymphocyte, and monocyte count analyses, no obvious potential side effects were observed (Supplementary Fig. [226]13a, b). In-depth histopathological evaluation of several different tissues also did not reveal abnormalities upon treatment with the SAOs (Supplementary Fig. [227]14). Overall, although c-mycER^TAM-immortal hNSCs were the parent cells, we noted that SAOs have a good safety profile in vivo based on an assessment of their tumorigenicity, associated immune response, and toxicity. Discussion Mimicking biological processes by engineering biomimetic organelles presents an elegant strategy for addressing several cell dysfunctions^[228]35. Here, we proposed a systematic top-down design strategy based on membrane self-assembly that can help prepare iNSC derived-SAOs (as a bioactive unit), which has combinatorial characteristics of OXPHOS artificial organelles and the intrinsic effectiveness of neural stem cells. Our study highlights the potential of SAOs as attractive cell-free therapeutic strategies for improving energy synthesis metabolism and unbalanced redox homeostasis with remarkable therapeutic efficacy. We are exceptionally excited about the fact that SAOs are functional in the ischemia-reperfusion model because it proves that the concept of SAOs as biomimetic organelles to correct OXPHOS dysfunction and restore mitochondrial bioenergetics is feasible in vivo. We speculated that the therapeutic effect of SAOs mainly depends on the following two aspects: i) SAOs improve ATP synthesis efficiency by executing OXPHOS function and preventing a supply shortage; and ii) SAOs restore mitochondrial dysfunction by rectifying mtROS generation abnormalities and reducing mtROS-induced damage to mitochondria. After oxidative stress occurs, cells are prone to mitochondrial homeostasis imbalance. Upon mitochondrial dysfunction, abnormal OXPHOS leads to insufficient ATP synthesis, which triggers a cascade of events, such as energy deprivation, ion gradient disruption, and calcium homeostasis imbalance, leading to cell damage or apoptosis^[229]36,[230]37. SAOs contain abundant respiratory chain complex and ATP synthase complex proteins, which are necessary for ATP synthesis. When SAOs are added to cells, these complexes can utilize substrates to perform OXPHOS, thereby improving ATP synthesis efficiency and increasing bioenergy synthesis capacity. In addition, abnormal OXPHOS also leads to electron transport chain leakage within the mitochondria, increasing mtROS production. Excessive mtROS may cause oxidative damage to the cell membrane, proteins, and nucleic acids, consequently aggravating mitochondrial OXPHOS dysfunction, forming a mtROS generation loop, and eventually triggering an avalanche release of mtROS^[231]38. The exogenous addition of SAOs can disperse the oxidative metabolic pressure of intracellular mitochondria, provide additional electron transport chains, and share the electron load. In turn, the electron transport chain in each mitochondrion within the cell becomes relatively less burdened by electrons, reducing the chance of incomplete oxygen molecule reduction during oxidation and reducing the likelihood of electrons leaking into oxygen to produce superoxide, giving rise to other forms of ROS. Building upon our demonstration of the pivotal role of proteins enriched in SAOs and elucidation of their functions, we further analyzed SAO-enriched miRNAs and their potential impact on the central nervous system (CNS). Through RNA sequencing (RNA-seq) analysis, we speculated that miRNAs found within SAOs may play a significant role in regulating the pathogenesis of several OXPHOS-related neurological disorders, such as Alzheimer’s disease, amyotrophic lateral sclerosis, epilepsy, schizophrenia, and multiple sclerosis^[232]39,[233]40. These findings not only emphasize the regulatory potential of miRNAs contained within SAOs but also pave the way for further functional studies and the development of innovative therapeutic strategies for treating various neurological conditions characterized by OXPHOS dysfunction. Compelling evidence suggests that NSC-derived nanovesicles inherit excellent neuroprotective and neuroregenerative capabilities from parent cells while mitigating the risks associated with stem cell therapy, including ethical issues, immune rejection, tumor formation, and other unpredictable risks^[234]41–[235]43. Meanwhile, to overcome product heterogeneity and guarantee SAO uniformity, rigorous process management and quality control procedures were executed from upstream production (conditions of cell culture, passage and cryopreservation) to downstream preparation (technical parameters of equipment), effectively minimizing batch-to-batch variations. This involves engineering human conditionally immortalized clone neural stem cells (iNSCs with c-mycER^TAM modification) as parent cells and employing a fully automatic device for producing highly homogeneous SAOs. Stem cells are used as producers of therapeutics and not as therapeutics per se, enhancing the clinical feasibility of the therapy and opening frontiers for stem cell therapy research. Our model offers the advantages of cost-effective production and low batch-to-batch variations, which are necessary for successful clinical translation. SAO contain large amounts of effective constituents specific to the parent cells. The self-assembly process and specific components of SAOs determine the uniqueness of its function. The parameters related to the formation and function of SAOs include the source of the parent cell, standardized preparation process (including the culture condition of the upstream parent cell, and the control parameters of the downstream high-pressure homogenizer), and specific modification of SAOs after preparation. These parameters all determine the formation and functional characteristics of SAOs. To expedite the clinical translation of SAOs, several strategies should be considered: First, selecting the appropriate parental cell sources based on the characteristics of specific diseases, such as umbilical cord mesenchymal stem cells, adipose mesenchymal stem cells, or bone marrow mesenchymal stem cells. Second, capitalizing on the amenability of the nanovesicles to modifications using genetic engineering or specific alterations to enhance the targeting of SAOs to particular systems and lesions, improve biocompatibility, and reduce immunogenicity, thereby broadening their clinical prospects, offering therapeutic options for a variety of diseases. Third, a comprehensive investigation into the in vivo distribution and metabolic processes of SAOs is imperative. This involves examining their distribution across various tissues and cell types, as well as elucidating their biotransformation, degradation pathways, and potential interactions with other drugs. Last, future evaluations of SAOs should be conducted in large animals to bridge the gap between rodent models and humans to accelerate the translation of stem cell technology to clinical practice. We anticipate that it would ultimately lead to the development of next-generation cell-free therapy and prompt a market shift from traditional stem cell therapies to cell-free approaches. Methods Cell culture The HT22, U87, T98G, and HS683 cell lines were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China) and cultured in DMEM (Gibco) supplemented with 10% FBS (Gibco) and 1% penicillin–streptomycin (Gibco). When cells reached 80% confluency, they were washed with PBS (Thermo Fisher Scientific) and passaged using 0.05% trypsin for 3 min. After incubation, the dissociation reagent was stopped by DMEM (Gibco) supplemented with 10% FBS (Gibco). The cell suspension was centrifuged (300 × g for 5 min, RT), and the cells were resuspended in a known volume of media. Cells were counted by a haemocytometer, plated at a seeding density of 10,000 cells/cm^2. The cells were maintained under standard culture conditions at 37 °C and 5% CO[2], with media replenished every 48 h. Human neural stem cells (NSCs) were obtained from the Stem Cell Clinical Research Institute of the First Affiliated Hospital of Dalian Medical University (LCKY2016-60) and grown in a NeuroCult™ NS-A Proliferation Kit (Stem Cell Technologies). When NSCs (NeuroCult™ NS-A Proliferation Kit) or iNSC (NeuroCult™ NS-A Proliferation Kit + 100 nM 4-OHT) reached 80% confluency in laminin (Sigma-Aldrich)-coated flasks, they were washed with PBS (Thermo Fisher Scientific) and passaged using Accutase (Stem Cell Technologies) for 6 min at 37 °C. After incubation, the dissociation reagent was stopped by dilution in proliferation media. The cell suspension was centrifuged (400 × g for 5 min, RT), and the cells were resuspended in a known volume of media. Cells were counted by a haemocytometer, plated at a seeding density of 50,000 cells/cm^2, and maintained in standard culture conditions at 37 °C and 5% CO[2], with feeding every 72 h. Human NSC-derived neural cells are obtained through the differentiation of NSCs using the NeuroCult™ Differentiation Medium Kit (Stem Cell Technologies). Prepare NSCs by resuspending them in Complete NeuroCult™ Differentiation Medium to attain a plating cell density of 4.2–5.3 × 10^4 cells/cm^2, then seed them onto Matrigel®-coated plates. Incubate the cultures in a humidified incubator at 37 °C with 5% CO[2]. Daily monitoring of the cultures is essential to evaluate the necessity for medium changes during differentiation. When the medium become acidic (turning yellow), conduct a half-medium change by substituting approximately half of the medium with fresh Complete NeuroCult™ Differentiation Medium. After 5–10 days, utilize an inverted light microscope to inspect the cultures for differentiation. To obtain human neurons, the STEMdiff™ Forebrain Neuron Differentiation Kit and STEMdiff™ Forebrain Neuron Maturation Kit (Stem Cell Technologies) were used to promote the differentiation of NSCs into neurons. NSCs are passaged as single cells using ACCUTASE™. The cells are then plated onto coated wells of a six-well plate at a density of 80–125,000 cells/cm^2 in 2 mL of STEMdiff™ Neural Induction Medium + SMADi and incubated at 37 °C with 5% CO[2] for 24 h. Starting from day 1, the medium is aspirated and replaced with 2 mL of STEMdiff™ Forebrain Neuron Differentiation Medium, with daily full-medium changes. Between days 7, when cells reach 80–90% confluence, they are passaged into STEMdiff™ Forebrain Neuron Maturation Medium using ACCUTASE™. For neuron maturation, neuronal precursors are seeded onto warm coated cultureware at a density of 5 × 10^4–1.5 × 10^5 cells/cm^2 in STEMdiff™ Forebrain Neuron Maturation Medium, with full-medium changes every 2–3 days. Maturation continues for a minimum of 7 days, with the option to culture neurons for up to 12 weeks. Human induced pluripotent stem cells (iPSCs) were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The iPSC line (DYR0100, catalog no. SCSP-1301) underwent rigorous validation for pluripotency and were confirmed to be free of mycoplasma contamination. Cultures were maintained on Matrigel (Corning, 354277) in mTeSR1 medium (STEMCELL Technologies, 85850), and maintained in standard culture conditions at 37 °C and 5% CO[2] with daily medium changes. Passage of iPSCs was performed at approximately 70% confluence using ReLeSR (STEMCELL Technologies, 05872). The cells were maintained under standard culture conditions at 37 °C and 5% CO[2]. Establishing a genetically engineered clonal NSC (iNSC) line The hNSC was obtained with informed consent from a legally constituted stem cell bank of Stem Cell Clinical Research Institute in the First Affiliated Hospital of Dalian Medical University, and in accordance with nationally (CHINA) approved ethical and legal guidelines. NSCs from three different parent cells were transduced with c-mycER^TAM lentivirus, which contains EGFP and a puro-selectable marker, to generate a fusion protein that stimulates cell proliferation in the presence of the synthetic drug 4-hydroxy-tamoxifen, to establish transfected pools of cells. In brief, NSCs with good growth and 50% cell density were selected in advance, and concentrated virus particles were added to the NSC supernatant, and the culture medium containing the virus was discarded after 24 h of transfection and replaced with fresh culture medium. Six days after transfection, the fluorescence expression of cells was observed under a fluorescence microscope. Puromycin (2 µg/ml) was added for cell screening of stable strains. Flow sorting was used to construct a single clone by Sony SH800 (Sony). A single-cell suspension of c-MycER^TAM NSCs was prepared for sorting, and cells were sorted according to their fluorescence signal. Single cells were sorted into a well in a 96-well plate containing complete medium and expanded for 3-4 weeks. Cerebral organoid generation and maintenance Human iPS cells (DYR0100, National Collection of Authenticated Cell Cultures, Shanghai, China) were dissociated into single cells using Stempro accutase (Thermo Fisher, catalog no. A1110501). Subsequently, 9000 cells were seeded per well in a 96-well ultra-low-attachment U-bottom plate in EB Seeding Medium supplemented with 10 mM Y-27632. Throughout days 2 and 4, 100 μl of EB Formation Medium were added to each well, followed by a complete medium change to Induction Medium on day 5. By day 7, each organoid was encapsulated in a 15 μL droplet of Matrigel using an Organoid Embedding Sheet and polymerized at 37 °C for 30 min. After embedding, 12–16 organoids were transferred to wells of a six-well plate pre-treated with Anti-Adherence Rinsing Solution and cultured in Expansion Medium. On day 10, the medium was switched to Maturation Medium, and the plates were placed on an orbital shaker. Maturation medium changes were subsequently performed every 3–4 days in accordance with the STEMdiff™ Cerebral Organoid Kit protocol. Cultures were routinely screened for mycoplasma on a monthly basis and consistently yielded negative results. Digestion of brain organoids After co-culturing brain organoids with PKH67-labeled SAOs for 24 h, they were subjected to five washes with PBS. Subsequently, the organoids were treated with a digestion buffer containing Stempro Accutase, Collagenase, and dNase for 30 min at 37 °C. Subsequently, the organoids were gently agitated by pipetting up and down ten times using a 1 ml pipette tip to release the cells. Any remaining cells were liberated from the organoids by agitation at 500 × g on an Eppendorf Thermomixer C for 10 min at 37 °C, followed by ten gentle pipetting motions with a 1 ml pipette tip. After allowing the cell debris to settle, the supernatant was carefully collected and filtered through a 70 μm filter paper. The resulting cell suspension was washed with FACS buffer, centrifuged, and the cell pellet collected for downstream analysis. Flow cytometry Digested cells were incubated in Fetal Bovine Serum Stain Buffer (554656, BD Pharmingen) containing the specified antibodies for 30 min at 4 °C. The flow cytometry gating strategy aims to identify neural progenitor cells (NPCs) exhibiting CD184^+/CD271^−/CD44^−/CD24^+, glial progenitor cells (GPCs) displaying CD184^+/CD44^+, and neurons characterized by CD184^−/CD44^−/CD15^−/CD24^+. Antibody staining was performed depending on the experimental design, utilizing specified antibodies including CD24 (555428, BD Pharmingen), CD15 (560828, BD Pharmingen), CD271 (560326, BD Pharmingen), CD44 (560533, BD Pharmingen), and CD184 (562448, BD Pharmingen). The cells were subsequently washed three times with PBS prior to detection and analysis. Samples were analyzed using a Sony SH800 for data acquisition. Flow cytometry data were processed using FlowJo software. Evaluation of brain organoid organization Cryo-sectioning of brain organoids Brain organoids were washed five times in PBS and fixed in 4% PFA for 15 min at 4 °C. Organoids were washed with PBS thrice and left in 30% sucrose overnight at 4 °C before embedding in a Tissue Freezing Medium at room temperature and finally moved at 4 °C to allow for solution polymerization. Small organoid-containing blocks were cut out using a scalpel blade followed by snap freezing in dry ice-containing isopentane for storage at −80 °C. The blocks were cut into 20-μm thick slices using a cryostat (Leica) and the sections were placed onto poly-L-lysine-coated slides for immunolabelling. Immunofluorescence of brain organoids Staining was performed on the largest cross-section slice of each brain organoid. Brain organoid slices on poly-L-lysine-coated slides were blocked with blocking buffer (0.5% Triton X-100 and 3% BSA) for 1 h at room temperature. After blocking, the sections were washed three times in PBS before incubation with the specified primary antibodies (in PBS containing 1% donkey serum) overnight at 4 °C. The primary antibodies used were TUJ1 (ab78078, Abcam) and Sox2 (ab92494, Abcam). The sections were then washed three times with PBS and labeled with secondary antibodies (ab150077, ab150113, ab150080, or ab150116, Abcam) for 1 h at room temperature. The secondary antibody was discarded, and the samples were washed with PBS. The samples were sealed with anti-fluorescence attenuating tablets containing DAPI (ab285390, Abcam). Imaging and analysis of brain organoids The brain organoid slides were observed and photographed under a fluorescence confocal microscope (Leica Biosystems) and analyzed using the ImageJ software. Organized stratification of neuroepithelial rosette structure is a crucial parameter for assessing the organization of brain organoids. We conducted imaging of all neuroepithelial rosette structures per slice and analyzed the distribution patterns of neurons (TUJ1+ cells) and neural progenitor cells (SOX2+ cells) to evaluate the organized stratification in neuroepithelial rosette structures and structure of brain organoids. During image analysis, regions were delineated based on the distribution of SOX2+ cells, with demarcations indicated by white dashed lines. The region within the white dashed lines (inner region) represented the layer of neural progenitor cells, while the area outside the white dashed lines (outer region) indicated the layer of neurons. We assessed the stratification of neuroepithelial rosette structures by evaluation of the distribution of neurons (TUJ1+ cells) across various regions. Thus, the relative fluorescence intensity of TUJ1 within the inner region (calculated as the total fluorescence intensity of TUJ1 within the inner region divided by the total area of the inner region) was used as the primary parameter to evaluate the organization of brain organoids. The fluorescence intensity was quantified by first splitting the images into their respective channels to enable the quantification of the corresponding markers. Subsequently, the polygon tool was used to delineate the inner region and measure its area. Finally, the total gray value within the inner region was measured to represent the total integrated fluorescence intensity of TUJ1, and the relative fluorescence intensity of TUJ1 was calculated by normalizing the total integrated fluorescence intensity by the inner region area, ensuring a standardized comparison across different samples or experimental conditions. Oxidative stress models and SAOs treatment In models inducing oxidative stress by hydrogen peroxide (H[2]O[2]), HT22 cells, human neurons, and mixed neural cells underwent a 24 h exposure to 200 μM H[2]O[2], with or without SAOs. Cells cultured under normal conditions constituted the normal group, while those exposed solely to H[2]O[2] formed the control group. The SAOs group comprised cells cultured with both H[2]O[2] and SAOs (10 μg per 1 × 10^6 cells). Similarly, in succinate-driven ROS models of HT22 cells, cultures were exposed to 50 mM succinate for 24 h with or without SAOs. The normal group included HT22 cells cultured under standard conditions, while the control group comprised cells treated solely with succinate. The SAOs group consisted of cells exposed to succinate along with SAOs (10 μg per 1 × 10^6 cells). In the oxygen-glucose deprivation (OGD) models, cells and brain organoids were initially placed in glucose-free medium and exposed to hypoxic conditions (1% O[2]) at 37 °C. After specific incubation periods (4 h for HT22 cells, 2 hours for human neurons, and 8 h for brain organoids), the medium was replaced with glucose-containing normal medium. Subsequently, cells or brain organoids were exposed to normoxic conditions at 37 °C, either with or without SAOs (10 μg per 1 × 10^6 cells, 10 μg per three brain organoid). These groups were designated as control and treatment groups, respectively. In the OGD model, the normal group comprises cells or brain organoids cultured under standard normoxic conditions. Additionally, the SAOs are administered continuously every 4 days until a specified time point for evaluation. Generation of SAOs In order to ensure the homogeneity of SAOs, we have implemented stringent process management from upstream production and downstream preparation. For downstream nanovesicle production, meticulous process management is crucial. During downstream preparation, we adhere strictly to engineering management practices for processing and utilize high-pressure homogenizers (AVESTIN) to produce SAOs. Technical parameters for each preparation are meticulously defined, including maintaining a pressure of 110 MPa, a temperature of 4 °C, a cell injection concentration of 1 × 10^6 cells/ml, and a physical state of cell suspension. Additionally, extrusion membrane apertures are set at 10 µm, 5 µm, 1 µm and 0.22 µm. For the preparation process, a cell population of 1 × 10^9 cells are suspended in 100 mL of phosphate-buffered saline (PBS), and the cell suspension is processed through a polycarbonate filter via a high-pressure homogenizer. During operation, the collected iNSCs are sequentially passed through a polycarbonate membrane (Whatman) with pore sizes of 10 µm, 5 µm, 1 µm and 0.22 µm, each subjected to five passes. To remove intact cells or cell fragments, the resulting mixture undergoes centrifugation at 2000 × g for 15 min. Subsequently, the supernatant is centrifuged at 10,0000 × g at 4 °C for 2 h to isolate SAOs. Micro bicinchoninic acid (BCA) protein assay SAOs underwent total protein content measurement using the microBCA Protein Assay Reagent Kit (23235, Thermo Fisher Scientific) as per the manufacturer’s instructions. To quantify the protein content of SAOs through the microBCA assay, initially lyse the SAOs to release their proteins using RIPA lysis buffer (89901, Thermo Fisher Scientific) supplemented with protease and phosphatase inhibitors (78430 and 78420, Thermo Fisher Scientific). Then, proceed to prepare standards and samples in the microBCA assay buffer, incubate them with the working reagent, measure absorbance at 562 nm, and subsequently calculate protein concentrations using a standard curve, followed by result analysis. Nanoparticle tracking analysis (NTA) NTA was used to determine the particle size and concentration distribution. SAOs were thawed and diluted in deionized distilled water at a dilution to achieve an approximated concentration between 10^6 and 10^9 particles/mL to be analyzed using the ZetaView PMX120.The analysis settings were optimized and kept constant between samples. Instrument settings: Zeta potential: accuracy 5 mV, precision 4 mV, reproducibility 5 mV; particle size test: accuracy 6 nm, precision 4 nm, reproducibility 4 nm; concentration test: accuracy 0.8 Mio/mL, precision 0.5 Mio/mL, reproducibility 1 Mio/mL. Western blotting The protein extraction process commenced with the use of 200 μl RIPA lysis buffer supplemented with protease and phosphatase inhibitors on cell/SAOs. Following centrifugation at 12,000 g at 4 °C for 15 min, the resulting supernatant was retained for subsequent analysis. Total protein concentration was assessed using BCA protein assays. Subsequently, 20 μg protein samples were subjected to separation via 4–12% SDS–PAGE (Thermo Fisher Scientific) and then transferred onto 0.45 µm polyvinylidene fluoride membranes (Millipore). Following a 1-h blocking step with 5% skim milk (BD Bioscience), membranes were subjected to overnight incubation at 4 °C with the primary antibody, followed by three washes with TBST, and then a 1-h incubation at room temperature with the corresponding HRP-conjugated secondary antibody. Antibody include C-myc (18583S, Cell Signaling Technology), Nestin (ab105389, Abcam), Sox2 (3579S, Cell Signaling Technology), Caspase 9 (9508S, Cell Signaling Technology), Caspase 3(9662S, Cell Signaling Technology), p44/42 MAPK (Erk1/2) (4695S, Cell Signaling Technology), Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (4370S, Cell Signaling Technology), p38 MAPK (8690S, Cell Signaling Technology), Phospho-p38 MAPK (Thr180/Tyr182) (4511S, Cell Signaling Technology), SAPK/JNK Antibody (9252S, Cell Signaling Technology), Phospho-SAPK/JNK (Thr183/Tyr185) (9255S, Cell Signaling Technology), GAPDH (ab125247, Abcam;), Anti-mouse IgG, HRP-linked Antibody (7076S, Cell Signaling Technology), Anti-rabbit IgG, HRP-linked Antibody (7074S, Cell Signaling Technology). Gray scan analysis was conducted using ImageJ software (National Institute of Health). Source data contains the full memebrane data. Transmission electron microscopy (TEM) A 20 μL suspension of SAOs was carefully pipetted onto a carbon-film copper mesh and allowed to settle for 5 min. Excess liquid was gently absorbed using filter paper, followed by the application of 2% phosphotungstic acid (Sinotech Genomics, China) onto the carbon-film copper mesh for 1–2 min. After absorbing the excess liquid, the sample was air-dried at room temperature. Subsequently, observation and image collection were performed using a transmission electron microscope (HT-7700, Hitachi, Japan). Quantification of ATP content The ATP content in cells was quantified using a commercial bioluminescent assay kit (ATP assay kit, Merck), following the manufacturer’s instructions and as previously described. Briefly, ATP was extracted by lysing cells in ATP Assay Buffer and deproteinized using a 10 kDa MWCO spin filter. Bioluminescence measurements were performed using a Multi-Mode Microplate Reader (FlexStation 3, Molecular Devices), where 50 µL of supernatant was mixed with 50 µL of Reaction Mix solution. A standard curve of ATP was generated by serial dilution of a 1 mM ATP solution. Data represent the mean of three independent experiments performed in duplicate. ATP content in SAOs was also measured by a commercial bioluminescent assay (ATP assay kit, Merck). Initially, samples were incubated at 37 °C in a buffer consisting of 100 mM Tris-HCl (pH 7.4), 100 mM KCl, 1 mM EGTA, 2.5 mM EDTA, 5 mM MgCl2, 0.2 mM di(adenosine-59) penta-phosphate, 0.6 mM ouabain, ampicillin (25 mg/ml), and 5 mM KH[2]PO[4]. The initiation of ATP synthesis was achieved by introducing 0.1 mM ADP into the mixture. To activate the pathway involving complexes I, III, and IV, mitochondria were supplied with 5 mM pyruvate and 2.5 mM malate. In the case of SAOs for ATP synthesis analyses, 0.7 mM NADH, a respiring substrate impermeable to mitochondria, was utilized. Additionally, for the activation of the pathway comprising complexes II, III, and IV, a uniform application of 20 mM succinate was incorporated across all samples. Mitochondria isolated from NSCs were utilized as a positive control following the protocol outlined in the mitochondrial extraction kit instructions (Thermo Fisher scientific, 89874). Briefly, the method involves traditional Dounce homogenization and utilizes a bench-top microcentrifuge for the differential centrifugation step to isolate mitochondrial and cytosolic components. After mitochondrial extraction from NSCs, the remaining components were processed through a high-pressure homogenizer to form nanovesicles (non-mitochondrial components of NSCs), which were then collected by centrifugation as the non-mitochondrial components for use as the negative control. The preparation of nanovesicle (non-mitochondrial components of NSCs) follow the SAOs preparation protocol. Membrane potential assay To assess the presence of membrane potential in SAOs, samples were labeled with 20 nM TMRM, a dye that fluoresces only in actively respiring membranes. Samples were then incubated at 37 °C in a buffer consisting of 100 mM Tris-HCl (pH 7.4), 100 mM KCl, 1 mM EGTA, 2.5 mM EDTA, 5 mM MgCl[2], 0.2 mM di(adenosine-59) penta-phosphate, 0.6 mM ouabain, ampicillin (25 mg/ml), and 5 mM KH[2]PO[4]. To activate the pathway involving complexes I, III, and IV, mitochondria were supplied with 5 mM pyruvate and 2.5 mM malate, while SAOs were supplied with 0.7 mM NADH. Additionally, for the pathway comprising complexes II, III, and IV, a uniform application of 20 mM succinate was added to all samples. As a negative control, the fluorescence of MT633 was assessed without the sample/substrate. Fluorescence intensity was monitored at 30 min using confocal fluorescence microscopy (CQ1, Yokogawa) for subsequent analysis. Mitochondrial membrane potential assay Cells were utilized for detecting mitochondrial membrane potential using the MitoProbe™ TMRM Kit for Flow Cytometry (Thermo [236]M20036). The cells were harvested and resuspended in 200 µl of TMRM staining working solution (20 nM), then incubated for 30 min at 37 °C with 5% CO[2]. Following the incubation period, the cells were washed twice with PBS. After discarding the supernatant, the cells were resuspended in 500 µl of PBS and immediately analyzed by flow cytometry with 561-nm excitation using emission filters suitable for R-phycoerythrin. Enzymatic activities of mitochondrial respiratory complexes The enzymatic activities of the mitochondrial complex I, complex II, complex III, complex IV, and complex V of the SAOs were determined by respiratory complexes enzymatic activities kits (Solarbio) following the manufacturer’s instructions. Briefly, reagent 1 from the kit was added to the SAOs particles, followed by sonication (20% power, 5 s on, 10 s off, repeated 12 times), to prepare samples for enzymatic activity and protein content determination. Enzymatic activity of mitochondrial complexes was assessed according to the kit instructions, involving the preparation of assay reagents and analysis of sample enzyme activity at specific wavelengths using a UV-visible spectrophotometer/enzyme analyzer. The control group comprised Reagent 1 without SAOs, while the experimental group consisted of samples containing SAOs. Three independent replicates were performed for each treatment. Immunofluorescence Cell samples were fixed with 4% paraformaldehyde for 15 min. The fixing solution was discarded, and 0.5% permeable solution was added for 10 min after washing with PBS. The permeable solution was discarded, the samples were washed with PBS, and 3% BSA solution was added and blocked for 30 min at room temperature. The blocking solution was discarded, and the primary antibody diluent was added after washing with PBS and incubated overnight at 4 °C. Primary antibodies used were Sox2 (ab92494, Abcam), Nestin (ab105389, Abcam), TUJ1 (ab78078, Abcam), MAP2 (ab32454, Abcam), and GFAP (ab68428, Abcam). After discarding the primary antibody, the samples were washed with PBS and then incubated with a diluted fluorescent secondary antibody (ab150077 or ab150113 or ab150080 or ab150116, Abcam) targeting the primary antibody for 1 hour at room temperature. The secondary antibody was discarded, and the samples were washed with PBS. The samples were sealed with anti-fluorescence attenuating tablets containing DAPI (ab285390, Abcam), observed and photographed under a fluorescence confocal microscope (Leica Biosystems), and analyzed using ImageJ software. The fluorescence intensities of TUJ1/MAP2 were quantified by first splitting the images into their respective channels to enable the quantification of the corresponding markers. Subsequently, the total gray value was measured to represent the total integrated fluorescence intensity of TUJ1/MAP2 within the field of view. Finally, to account for variations in cell density, the relative intensity of TUJ1/MAP2 was calculated by normalizing the total integrated fluorescence intensity by the number of nuclei in the same field, ensuring a standardized comparison across different samples or experimental conditions. Brain tissue sections on polysine-coated slides were blocked with blocking buffer (0.5% Triton X-100, 3% BSA) for 1 h at room temperature. After blocking, the sections were washed three times in PBS before they were incubated with the specified primary antibodies (in PBS containing 1% donkey serum) overnight at 4 °C. Primary antibodies used were TUJ1 (ab78078, Abcam), Sox2 (ab92494, Abcam), NeuN (#24307, Cell Signaling Technology), Myelin Basic Protein (#78896, Cell Signaling Technology), PSD95 (#36233, Cell Signaling Technology), CD31 (ab64543, Abcam), Iba1 (ab289370, Abcam),Synaptophysin (#36406, Cell Signaling Technology) and GFAP (#12389, Cell Signaling Technology). The sections were then washed three times with PBS and labeled with secondary antibodies (ab150077 or ab150113 or ab150080 or ab150116, Abcam) for 1 h at room temperature. The secondary antibody was discarded, and the samples were washed with PBS. The samples were sealed with anti-fluorescence attenuating tablets containing DAPI (ab285390, Abcam), observed and photographed under a fluorescence confocal microscope (Leica Biosystems), and analyzed using ImageJ software. The fluorescence intensity of GFAP was quantified by first splitting the images into their respective channels to enable the quantification of the corresponding markers. Subsequently, the total gray value was measured to represent the total integrated fluorescence intensity of GFAP within the field of view. Finally, to account for variations in cell density, the relative intensity of GFAP was calculated by normalizing the total integrated fluorescence intensity by the corresponding area assessed for fluorescence in the same field, ensuring a standardized comparison across different samples or experimental conditions. Neurites length measurement Neurons were cultured in STEMdiff™ Forebrain Neuron Maturation Kit (Stem cell technologies) supplemented with 1% penicillin‒streptomycin (Gibco) for 7 days, followed by the establishment of the oxidative stress model. In the H[2]O[2]-induced oxidative stress model of human neurons, 200 μm of H[2]O[2] was added to cells with or without SAOs (10 μg per 1 × 10^6 cells). After 24 h, the cells were labeled with anti-human TUJ1 antibody and stained with DAPI for detection. Subsequently, the number and length of neurites were measured using ImageJ software. Mitochondrial ROS detection Mitochondrial ROS (mtROS) levels in cells were measured using MitoSOX™ Green Mitochondrial Superoxide Indicators (ThermoFisher [237]M36006). Following centrifugation, the cell culture supernatant was removed and discarded, and 1 × 10^6 cells were collected. Then, the cells were treated with 2.5 μM MitoSOX Green working solution and incubated at 37 °C for 15 min in dark. After incubation, the cells were washed three times with PBS. The stained cells were then loaded onto the flow cytometer (Sony SH800), with excitation provided by a 488 nm laser, and fluorescence data collected using a 530/30 bandpass optical filter. All FACS data were processed with FlowJo software. Mitochondrial DNA assay The mtDNA copy number was estimated by real-time PCR. Total DNA was isolated from samples with a genomic DNA kit (AG, 21009), and 10 ng of DNA was used for PCR. For HT22, mtDNA copy number was calculated by normalizing the mitochondria-encoded cytochrome c oxidase subunit 2 (COX2) levels to nuclear ribosomal protein s18 levels. The primer sequences for the target genes are given below. COX2, Forwards: ATAACCGAGTCGTTCTGCCAAT Reverse: TTTCAGAGCATTGGCCATAGAA. Rsp18, Forwards: TGTGTTAGGGGACTGGTGGACA Reverse: CATCACCCACTTACCCCCAAAA). The Human Mitochondrial DNA (mtDNA) Monitoring Primer Set kit (7246, Takara) was utilized to assess mtDNA copy number in human cells. This comprehensive kit contains primer sets targeting two mtDNA regions (ND1, ND5) and two nuclear DNA (nDNA) regions (SLCO2B1, SERPINA1). The aim was to quantify the relative abundance of human mitochondrial DNA (mtDNA) using real-time PCR, with nuclear DNA (nDNA) content serving as a standard reference. By employing the four primer pairs provided in the set for real-time PCR, the relative quantification of mtDNA is determined based on the difference in Ct values between mtDNA and nDNA. For detailed analysis, please refer to the manual accompanying the kit. The experiment and subsequent analysis were conducted in accordance with the manufacturer’s instructions. Analysis of the oxygen consumption rate (OCR) Cellular respiration measurement Mitochondrial OCR analysis was carried out on a Seahorse XF-24 flux analyzer (Agilent) using a mitochondrial stress test (MST) assay kit (Agilent). HT22 cells were plated on poly-D-lysine-coated XFe-24 cell Seahorse culture microplates and subjected to various treatments, with untreated cells serving as the normal group. In rotenone-induced mitochondrial dysfunction model, HT22 cells were treated with 100 nM rotenone added to the culture media for 2 h, with or without SAOs. The HT22 cells without SAOs constituted the control group, while those treated with SAOs comprised the SAOs group. In the H[2]O[2]-induced oxidative stress model, we induced oxidative stress by exposing the cells to 200 μM H[2]O[2] for 24 hours, with or without the presence of SAOs. The HT22 cells without SAOs served as the control group, while those treated with SAOs constituted the SAOs group. After different treatments, the cells were further changed to MST buffer and equilibrated for 1 h at 37 °C in the absence of CO[2] before the assay. OCR was measured under basal conditions and in response to 1 μM oligomycin, 1 μM FCCP and 100 nM rotenone + 1 μM antimycin (all from Agilent). Basal mitochondrial respiration was determined from baseline measurements. Values of OCR (pmol O[2]/min) for each well were recorded and averaged for data analysis. ATP production respiration was determined by the subtraction of oligomycin A values from basal respiration. Maximal mitochondrial respiration was determined by subtracting the OCR after treatment with rotenone and antimycin A from the OCR measured following treatment with FCCP. Tissue respiration measurement Preparation of brain slices Brain slices were prepared following the euthanasia of rats using carbon dioxide (CO[2]), as per established laboratory protocols and previously documented literature. Within 30 s of decapitation, brains were swiftly removed and submerged in ice-cold artificial cerebrospinal fluid (aCSF; pH 7.4) oxygenated for 1 h with 95% O2 and 5% CO2. Using Vibrating Blade Microtomes (leica), coronal sections (220 μm) were then generated on a chilled stage and blade. These sections were subsequently transferred to a holding chamber filled with continuously oxygenated aCSF at room temperature (~23 °C). Tissue punches and respiration measurements Brain sections were placed individually in a biopsy chamber with fresh oxygenated aCSF. Tissue punches were extracted using biopsy punch needles (diameters ranging 1.0 mm) from left cortical sections. The punches were then directly transferred into an XFe24 Islet Capture Microplate with each well containing 700 µL of room temperature assay media (aCSF supplemented with 0.6 mM pyruvate and 4 mg/ml BSA). After ensuring the punches were submerged and centered at the bottom of each well, the microplate was incubated at 37 °C incubator without CO[2] for 1 h to allow temperature and pH equilibration. Meanwhile, a 10× concentration of assay drugs (prepared in aCSF) was loaded into the injection ports of a calibrated hydrated sensor cartridge, as previously reported. Once calibration was completed, the cartridge containing the study drugs was replaced by the microplate containing the tissue punches, and the assay protocol was initiated. OCR was measured under basal conditions and in response to 20 μg/mL oligomycin, 10 μM FCCP and 5 μM rotenone + 20 μM antimycin (all from Agilent). Basal mitochondrial respiration was determined from baseline measurements. Values of OCR (pmol O[2]/min) for each well were recorded and averaged for data analysis. ATP production respiration was determined by the subtraction of oligomycin A values from basal respiration. Maximal mitochondrial respiration was determined by subtracting the OCR after treatment with rotenone and antimycin A from the OCR measured following treatment with FCCP. Five replicates for each condition in any given experiment were used. TUNEL staining Rats were anaesthetized (10% isoflurane) and transcardially perfused with 200 ml of 5 mM sodium phosphate buffered 0.9% (w/v) saline (PBS, pH 7.2–7.4) followed by 500 ml of 4% paraformaldehyde in phosphate buffer. Then, the brains were rapidly removed. After being dehydrated in sucrose, the brain was embedded in OCT, and 20 μm thick horizontal slices were obtained by using a Freezing Microtome (Leica Biosystems). A TUNEL kit (Thermo Fisher Scientific) was used to detect apoptotic cells, and the experiment was performed according to the manufacturer’s instructions. Briefly, slice samples were fixed in 4% paraformaldehyde for 15 min and then permeabilized for 20 min in permeabilization reagent (0.25% Triton™X-100 in PBS). For the TUNEL reaction, TdT reaction cocktail were used for each sample, all of which were incubated for 60 min at 37 °C in a humidified chamber. Nuclei were visualized with DAPI (Abcam) staining. Three pictures were taken from randomly selected fields (basal ganglia or cortex) of each rat using a Leica DMI4000B. The rate (%) of TUNEL positive apoptotic cells was determined using the ImageJ software. EdU detection Cell proliferation was assessed using the 5-ethynyl-2’-deoxyuridine (EDU) assay. Briefly, NSCs and iNSCs were seeded into confocal dishes at a density of 30,000 cells/cm^2 and allowed to adhere overnight. Following treatment, cells were incubated with 10 μM EDU for 48 hours at 37 °C. To fluorescently label the incorporated EdU in samples, cells were fixed using 4% paraformaldehyde solution for 10 min. After the washing step with 3% BSA in PBS, cells were permeabilized using 0.5% Triton-X 100 in PBS for 20 min, and they were washed again. Then, the cells were incubated for 30 min in the dark with a reaction cocktail prepared according to the manufacturer’s instructions, which contained the compounds necessary for the bonding of azide-modified Alexa Fluor® dye with EdU. After an additional washing step using PBS, cell nuclei were stained with Hoechst 33342 (1:1000 in PBS) for 10 min and then washed. For the imaging procedure, a confocal quantitative image cytometer (CQ1, Yokogawa) was used. Laser lines at 405 nm and 594 nm were used to collect images from cell nuclei and incorporated EdU, respectively. Quantification of cell proliferation was performed by determining the percentage of EDU-positive cells relative to the total number of Hoechst-stained cells using Image J analysis software. Scratch wound assay Cells were placed in each well of a 12-well plate and cultured in complete medium (DMEM supplemented with 10% fetal bovine serum, FBS) until reaching 90% confluency. At this point, the growth media was replaced with basal medium lacking FBS for 12 hours to synchronize cells at the quiescent stage (G0) of the cell cycle. Subsequently, a full-thickness scratch was made in the middle of the cell monolayer using a 200 µL pipette tip, and the plates were washed several times with PBS to remove cell debris. PBS was then added to each well, and digital images were captured to measure the initial scratch width and area (0 h) using an inverted microscope. The cells were then supplemented with DMEM containing 1% FBS, with either SAOs (50 μg per 1 × 10^6 cells) or PBS. The cells were returned to the incubator and maintained for 24 h and 48 h for the evaluation of scratch closure progression. For the purpose of quantifying wound closure rates, digital images were obtained at 0 h and 24 h, 48 h, and analyzed using ImageJ software (National Institute of Health). Scratch wound healing was determined by comparing the percent wound closure for each time point relative to the scratch area at 0 h. Proteomics Protein extraction Samples were weighed and transferred into 2 mL centrifuge tubes. Steel balls, lysis solution containing 8 M Urea/50 mM Tris-HCL, and Roche cocktail (1X final concentration) were added, followed by a 5-min incubation on ice. After homogenization (60 Hz, 2 min) and centrifugation (20,000 × g, 15 min, 4 °C), the supernatant was collected. DTT (10 mM) was added and incubated at 37 °C for 1 h, followed by the addition of IAA (20 mM) and incubation for 30 minutes away from light. Quality control of protein extraction Protein quantitation was performed using the Bradford method. Standard protein solution (0.2 μg/μL BSA) was added to a 96-well plate along with pure water in a series of volumes. After thorough mixing, Coomas bright Blue G-250 quantitative working solution was added, and the absorption value at 595 nm (OD595) was measured using a microplate reader to create a standard curve. The OD595 of sample protein solutions was then measured, and their concentrations were calculated accordingly. Additionally, SDS-PAGE analysis was conducted with 10 μg protein per sample. After heating and centrifugation, the supernatant was loaded onto a 4–12% SDS polyacrylamide gel and subjected to constant pressure electrophoresis at 80 V for 20 min followed by 120 V for 60 min. The gel was subsequently stained, destained, and photographed. Protein digestion 150 μg of protein was taken from each sample and mixed with 3 μg of trypsin at a ratio of 50:1. Samples were then incubated at 37 °C for 14–16 h. The resulting digested peptides were desalted using Waters solid phase extraction cartridges, vacuum dried, and stored at -20 °C after redissolving in pure water. Quantitative detection by nano-LC-MS/MS Dried peptide samples were dissolved in 0.1% FA and centrifuged. The supernatant was injected into a self-loading C18 column (100 μm I.D., 1.8 μm particle size, ~35 cm length) at a flow rate of 300 nL/min. Separation was achieved using a Thermo Scientific EASY-nLC™ 1200 system with a gradient of solvent B (98% ACN, 0.1% FA). Peptides were ionized by nano-ElectroSpray Ionization and analyzed by Orbitrap Exploris™ 480 mass spectrometer in DDA mode. Protein identification and quantification MaxQuant (version 2.1.4.0) software analyzed the label-free MS/MS data with standard settings. Trypsin/P enzyme was used with a maximum of two missed cleavages. Carbamidomethyl (C) was set as a fixed modification, while oxidation (M) and acetyl (protein N-term) were variable modifications. Data were searched against protein sequences from the Uniprot database with an FDR threshold of 1% at both PSM and protein levels, removing contaminants and reverse proteins. Bioinformatics analysis Statistical analysis was conducted in R (version 4.0.0). Raw protein intensities were normalized using the “medium” method. Hierarchical clustering was performed using the pheatmap package, while PCA was executed with the metaX package. Correlation analysis utilized Pearson correlation coefficients from the cor package. Differential analysis employed the T-test, with a significance threshold of p value ≤ 0.05 and fold change ≥1.2 to identify statistically differentially expressed proteins. For annotation, hypergeometric-based enrichment analyses were performed individually for KEGG Pathway, Gene Ontology, Reactome Pathway, and InterPro domain annotations of protein sequences. Gene set enrichment analysis (GSEA) was conducted using GSEA software (v4.1.0) and MSigDB to identify significant enrichment of gene sets associated with specific GO terms, KEGG pathways, DO terms (human), and Reactome pathways (some model animals). Enrichment was considered significant if |NES | > 1, NOM p-val <0.05, and FDR q-val <0.25. Subcellular localization analysis was performed using WoLF PSORT. Detection of cell confluence In this study, neural stem cells (NSCs) and induced neural stem cells (iNSCs) were observed and tracked for a duration of 96 hours. The cell confluence was assessed using an automated confluence mask generated by the CellPlayer™ Kinetic Proliferation Assay, integrated with the IncuCyte™ Live-Cell Imaging Systems. This assay allowed for the automatic highlighting of cell positions, enabling precise monitoring of cell growth and proliferation dynamics over the specified time frame. Cell, mitochondrion and SAOs labeling Cell membranes were labeled with CellMask™ Green Plasma Membrane stain ([238]C37608, Thermo Fisher Scientific), known for its strong affinity to the plasma membrane, facilitating clear visualization of cell boundaries. Cells were incubated with a 1X working solution for 5–10 minutes at 37 °C, followed by three rinses with PBS to remove excess stain. Mitochondria were labeled using MitoTracker® probes (M7514, Thermo Fisher Scientific), which passively diffuse across the plasma membrane and accumulate within active mitochondria, emitting a green fluorescence signal. Cells were treated with a 100 nM working solution for 30 minutes at 37 °C, followed by three washes with PBS. SAOs were labeled with PKH-26 (Sigma-Aldrich), a lipophilic dye integrating into the lipid bilayer of membranes. SAOs underwent incubation with a 1X working solution for 15 minutes at 37 °C, followed by three rinses with PBS. Each labeling procedure strictly followed manufacturer’s instructions regarding concentrations and incubation times, ensuring accuracy and consistency of results. Apoptosis detection Quantification of apoptosis was performed using annexin V-FITC and propidium iodide (PI) staining kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, samples were washed and resuspended in Annexin V-binding buffer and stained with Annexin V (5 µL) on ice for 10 minutes. For double stained samples, propidium iodide (PI) (10 µL) was added directly. The stained samples were detected by flow cytometry (SONY SH800). Distinct stages of cell apoptosis were studied with this assay where early apoptotic cells are Annexin V positive and PI negative (AnnexinVFITC +/PI-), whereas late apoptotic cells are AnnexinV/PI-double-positive (AnnexinV-FITC + /PI + .) Viable cells are unstained (AnnexinV-FITC-/PI-). Cell viability and cytotoxicity assay Cell viability was evaluated by an enhanced cell counting kit-8 (CCK-8) assay following the manufacturer’s instructions (Dojindo). Briefly, cells were plated in 96-well plates at a concentration of 1 × 10^4 (200 μl/well). After treatments, 10 μl CCK-8 reagent was added to each well and incubated at 37 °C for 2 h. The optical density (OD) values of the samples were measured at 450 nm with a microplate reader. Cytotoxicity was evaluated with a Lactate Dehydrogenase Assay Kit (Dojindo) following the manufacturer’s instructions. In brief, cells were plated in 96-well plates at a concentration of 1 × 10^4 (200 μl/well). After treatments, 10 μl LDH reagent was added to each well and incubated at RT for 20 min. The OD values of the samples were measured at 490 nm with a microplate reader. DNase digestion of SAOs SAOs were incubated with 1 U/ml DNase I at 37 °C for 1 hour to remove DNA (90083, Thermo Fisher Scientific). The reaction was stopped by ultrafiltration at 4 °C, with a centrifugation at 100,000 × g for 2 hours, and the resulting SAO pellets were collected. DNA extraction from both co-incubated and non-co-incubated SAOs was carried out with a Universal Genomic DNA kit (56304, Qigen). Briefly, SAO samples were lysed to release genomic DNA, followed by protein precipitation to remove contaminants. The DNA was then bound to a membrane, washed to remove impurities, and eluted in a low-salt buffer. Subsequently, the concentrations and purity of the extracted DNA (double-stranded and single-stranded) were determined using either a spectrophotometer or Qubit 4. Gel electrophoresis was performed to evaluate the extent of degradation. Agarose gel electrophoresis Agarose gel electrophoresis was used to detect DNA fragments in SAOs. Agarose gel was prepared by dissolving agarose powder in either Tris-acetate-EDTA (TAE) or Tris-borate-EDTA (TBE) buffer to achieve a final concentration of 0.8-1.5%. The solution was heated until fully dissolved, poured into a casting tray with a comb to create wells, and allowed to solidify. The comb was then removed, and the gel was placed in an electrophoresis tank with fresh TAE or TBE buffer. DNA samples, mixed with 6x loading dye, were loaded into the wells using precision micropipettes. Electrophoresis was performed at a constant voltage of 125 V. After electrophoresis, the gel was stained with a DNA-specific dye for 20 minutes and visualized under UV light using a gel documentation system (Bio-Rad, Hercules, CA). Gel images were captured for analysis and documentation. Intracellular Ca^2+ flux assay Intracellular calcium ion levels were quantified employing the Fluo-4 Calcium Imaging Kit ([239]F10489, Thermo Fisher Scientific). Initially, neurons were exposed to the Fluo-4 AM loading solution and underwent a 15-minute incubation at 37 °C, followed by an additional 15-minute incubation at room temperature. Subsequently, the cells underwent three washes with calcium- and magnesium-free HBSS, each lasting 2 minutes. Following the washes, treatment with the Neuro Background Suppressor provided in the kit was conducted to further optimize imaging conditions. Calcium imaging was conducted over a duration of 7.5 minutes with a time interval of 3 seconds (150 frames) utilizing a Confocal microscope (CQ1, Yokogawa). Laser lines at 488 nm were used to collect images. Image analysis was performed using the integrated software within CQ1. A region of interest (ROI) was delineated around each cell, and the time-lapse acquisition mode of the software was utilized to track fluorescence changes over time. Subsequently, the raw fluorescence intensity values from each neuron were normalized to the initial fluorescence intensity signal of the baseline recording. The peak frequency graph was plotted and analyzed using prism analysis software. ELISA Enzyme-linked immunosorbent assay (ELISA) was used to detect cytochrome c in the cell supernatant (Elabscience). Cell lysates were prepared by homogenizing cells in lysis buffer containing protease inhibitors. The lysates were then centrifuged to remove cellular debris, and the protein concentration in the supernatant was determined using a Bradford protein assay. Equal amounts of protein were loaded onto ELISA plates coated with anti- cytochrome c antibody and incubated for 2 hours at room temperature. After washing, a secondary antibody conjugated to HRP was added and incubated for 1 hour. Following another wash step, the HRP substrate was added, and the absorbance was measured at 450 nm using a microplate reader. cytochrome c concentrations were determined by comparing the absorbance values to a standard curve generated with known concentrations of cytochrome c. Transwell invasion assay Invasion assays were performed in Transwell^TM plates (Corning) according to the manufacturer’s protocol. Briefly, cells were plated in the top Matrigel (Corning)-coated chambers of 24-well plates with 8-µm inserts at a density of 5×10^4 cells/well in serum-free medium. The cells were then treated with either SAOs (50 μg per 1× 10^6 cells) or PBS. Culture medium (800 µl) containing 10% FBS was added to the bottom chambers. After a 24 h incubation, the lower side of each transwell membrane was fixed and stained with 0.05% crystal violet. Invaded cells were observed under an inverted light microscope (Olympus) and the number of invaded cells were analyzed using ImageJ software (National Institute of Health). Preparation of RVG29 modificated SAOs RVG29-modified nanovesicles were prepared according to a previously established protocol^[240]44. Briefly, a lipid film of DSPE-PEG[2000]-RVG29 was prepared by rotary evaporation and was further dried under vacuum for 24 h. The dried lipid film was subsequently hydrated with PBS (pH 7.4) to the formation of micelles at 37 °C. For the RVG-SAOs preparations, a micelle solution of DSPE- PEG[2000]-RVG29 was added into preformed RVG29-SAOs and was incubated at 37 °C in PBS (pH 7.4), respectively. Detecting the content of SAOs in the brain Our research group has developed a method based on ddPCR to detect the distribution of SAOs in the brain. Due to the small RNAs contained in the nanovesicles, detecting the species-specific small RNA (sRNA) sequences carried by the SAOs can evaluate the SAOs content in the target tissue. All animal procedures were approved by the Institutional Animal Care and Use Committee of Dalian Medical University and conducted according to the Guidelines for the Care and Use of Laboratory Animals. Male C57BL/6 mice (weighing18-22 g) were purchased from the SPF Animal Laboratory Center, Dalian Medical University. The animals were given ad libitum access to food and water and housed in cages (fewer than five animals per cage) on a 12-h light/dark cycle with a room temperature set at 22 °C ± 1 °C and a humidity of 60–70%. All efforts were made to minimize animal suffering and to reduce the number of animals used. Mice were divided into four groups: positive control group (brain tissue directly mixed with SAOs), negative control group (tail vein injection of 200 μl PBS / each), vesicle group (tail vein injection of 70 μg SAOs in 200ul PBS / each), and targeted vesicle group (tail vein injection of targeted modified SAOs in 200 μl PBS / each). The mice were euthanized 60 minutes after injection of the drug into the tail vein, and brain tissue was taken. The brain tissue sRNA was isolated by miRNeasy Micro Kit (QIAGEN Cat#: 217084, Valencia, CA) following the manufacturer’s protocol. The quantity of sRNAs were measured by the Qubit^TM microRNA Assay Kit (Invitrogen Cat# [241]Q32881) on the Qubit® 2.0 Fluorometer. Droplet digital PCR The expression levels of sRNAs were measured by ddPCR. Briefly, primers of sRNAs were synthesized by Sangon Biotech (primer: GGCTGGTCCGATGGTAGTGGGTTATCAGAACT). Reverse transcription of sRNA was performed using sRNA First Strand cDNA Synthesis (Tailing Reaction) (Sangon Biotech Cat#: B532451-0020, Shanghai, China). EvaGreen Supermix (Biorad Cat#: 186-4035) 10 μl, forward and reverse primers 0.2 μl (10 μM), cDNA, and RNase free water were mixed in 20 μl solution. The automated Droplet Generator (Bio-Rad) was used for eumulsification, and the ddPCR cycle conditions were 95°C for 10 min, followed by 40 cycles of a 2-step thermal 4 profile of 94°C denaturation for 15 sec, 57°C annealing for 60 sec, and the followed by a 4°C hold. After the 96-well plate was loaded on to and read by a QX200 Droplet Reader (Bio-Rad), results were analyzed using QuantaSoft™ Analysis Pro (Bio-Rad). Transcriptomics For the transcriptomics study, RNA samples were prepared using 2 μg RNA per sample and the NEBNext Ultra RNA Library Prep kit for Illumina (E7530L, NEB) following the manufacturer’s instructions. Index codes were added to attribute sequences to each sample during library generation. Differential gene expression analysis was conducted, identifying genes with P < 0.05 and absolute log2(fold changes) ≥ 1. GO enrichment analysis of differentially expressed genes was performed using the hypergeometric test. Additionally, gene set enrichment analysis (GSEA) was utilized to assess the level of metabolic pathway enrichment between different groups, utilizing the complete transcriptome of all samples. Significant gene sets were determined based on nominal P < 0.05 and false discovery rate q values < 0.06. Single-cell nuclei RNA sequencing (sn-seq) Sample preparation for single-nucleus RNA sequencing For the scRNA-seq experiments, single-cell suspensions were meticulously prepared from the brains of three rats each in the control (MCAO) and SAOs (MCAO treated with SAOs) groups. Following isolation, the single-cell suspensions from individual rats within the same group were combined to create a single pooled sample for subsequent sequencing analysis. Briefly, the sample was placed in 1 ml of pre-chilled lysate and sheared into small pieces approximately 1 mm in size. Subsequently, 2 ml of pre-chilled lysate was added, transferred to a tissue homogenizer (885300-0007; Kimble Dounce Homogenizer), and gently ground five times. After incubating the tissue on ice for 5 minutes, 4 ml of pre-cooled 2% BSA dilution was added, and the reaction was stopped by pipetting with a barrel pipette. The samples were then passed through a 30 μm cell sieve, and the filtrate was collected and centrifuged at 300 g for 5 minutes at 4 °C. The supernatant was discarded, and the cells were resuspended in 5 ml of 2% BSA, followed by centrifugation at 4 °C for 5 minutes at 300 g. The supernatant was discarded again, and the cells were resuspended in 500 μl of resuspension solution (1xPBS, 2% BSA, 0.1% RNase inhibitor). Propidium iodide (PI) staining was performed, and debris impurities were removed by sorting using a BD FACSAria II flow cytometer. Finally, cells were collected and counted. Chromium 10x Genomics library and sequencing The single-cell nuclear suspension was introduced into the 10x Chromium chip following the prescribed guidelines for the 10X Genomics Chromium Single-Cell 3’ kit (V3) to capture approximately 8,000 cells. Subsequently, standard protocols were employed for cDNA amplification and library construction. The libraries were sequenced using the Illumina NovaSeq 6000 sequencing system, utilizing a double-end sequencing approach with 150 bp read length, ensuring a minimum sequencing depth of 20,000 reads per cell. Bioinformatics analysis The results from offline Illumina sequencing were transformed into FASTQ format using bcl2fastq software (version 5.0.1). The scRNA-seq sequencing data was aligned to a reference genome using CellRanger software, enabling the identification and quantification of cellular and individual cellular 3’ end transcripts in the sequenced samples ([242]https://support.10xgenomics.com/single-cell-gene expression/software/pipelines/latest/what-is cell-ranger, version 7.0.0). The output CellRanger expression profile matrix was imported into Seurat (version 4.1.0) for removing low-quality cells from the scRNA-seq data, followed by data downsampling and clustering. The criteria for filtering low cell quality included the expression of genes per cell <500, number of UMI per cell <500, mitochondrial genes expressed in >25% of the cells, and doublets (assuming 7.5% of doublet formation rate). Next, the cells were projected into the 2D space using t-SNE or UMAP. The subsequent steps were as follows: 1) applying a global-scaling normalization method, “LogNormalize,” that normalizes the feature expression measurements for each cell by the total expression and log-transforms the results. 2) Using a highly variable-features identification method, “FindVariableFeatures,” to calculate a subset of features with high cell-to-cell variation in the dataset. 3) Applying a linear transformation (“ScaleData”), a standard pre-processing step prior to PCA. 4) Executing PCA with the normalized expression values, applying the top 20 principal components for FindNeighbors and FindCluster analyses. 5) Analyzing marker genes for each cluster based on Findallmarker, with marker genes chosen according to the following criteria: expression in more than 10% of the cells within each cluster and P values and fold changes of ≤0.01 and ≥0.26, respectively. Hypergeometric testing was used to analyze the differential genes of each cluster obtained in the Findallmarker analysis relative to other clusters. 6) SingleR was used to identify each cluster cell type, and some cells were re-clustered based on the identification results of the canonical neural cell marker genes. 7) DEGs between the cell types of interest were identified using the Wilcoxon rank-sum test implemented in the FindMarkers function of Seurat. The adjusted P value was calculated by Bonferroni correction based on all tested genes, and only DEGs with adjusted P values of p < 0.05 were used for downstream interpretations such as pathway analysis and drawing the plots. Gene ontology (GO) analysis Differentially expressed genes (DEG) analysis was performed using R (version 4.3.3), where DEGs were identified with adjusted P values < 0.05. The clusterProfiler package in R was utilized for GO analysis, focusing on biological processes associated with these DEGs. The analysis employed the Benjamini-Hochberg method to adjust for multiple testing. KEGG pathway analysis Pathway enrichment analysis utilized the clusterProfiler package in R to assess the distribution of DEGs across KEGG pathways. Significantly enriched pathways, defined by adjusted P values < 0.05, were visualized using dot plots. Gene set enrichment analysis (GSEA) The fgsea package in R facilitated the GSEA to explore the biological pathways associated with DEGs. Analysis involved predefined gene sets, with the significance of enrichment determined through 10,000 permutations. The visualization of results was achieved using GSEA plots. Animal model and SAOs administration All animal procedures were approved by the Institutional Animal Care and Use Committee of Dalian Medical University and conducted according to the Guidelines for the Care and Use of Laboratory Animals ([243]AEE23127). Male Sprague‒Dawley (SD) rats (weighing 240–270 g) were purchased from the SPF Animal Laboratory Center, Dalian Medical University. The animals were given ad libitum access to food and water and housed in cages (fewer than five animals per cage) on a 12-h light/dark cycle with a room temperature set at 22 °C ± 1 °C and a humidity of 60–70%. All efforts were made to minimize animal suffering and to reduce the number of animals used. Middle cerebral artery occlusion (MCAO) was established in rats by transient occlusion of the left middle cerebral artery. In brief, before the MCAO surgery, 8% isoflurane was used, and anesthesia was maintained during the whole procedure under 2% isoflurane. A 2 cm incision was made on the midline of the neck, and blunt dissection exposed the left common carotid artery (CCA), external carotid artery (ECA), and internal carotid artery (ICA) bifurcation. A microvaSAOlar clip temporarily clamped the CCA and ICA, and a small cut was made on the ECA. A 4-0 silicon-coated nylon monofilament (Doccol) was advanced to the ICA along the ECA and advanced until the origin of the middle cerebral artery. The suture was removed slowly after a 60 min occlusion of the MCA to restore blood flow. Rats in the sham group received no filament insertion and otherwise underwent the same surgical procedures as the MCAO group. Within 30 min after restoring the blood flow, rats were implanted with SAOs (70 μg in 9 μL of PBS) or PBS (9 μL) into two sites within the left putamen (coordinates: anterior −1.3 mm, lateral −3.5 mm, ventral −6.5 mm, and anterior −1.8 mm, lateral −4.0 mm, ventral −6.0 mm; anterior and lateral measurements were relative to bregma and ventral relative to dura). The sham group did not receive any treatment. The operations were conducted in compliance with the double-blind principle. All rats in the study were recoded after implantation to allow for blinded behavioral testing. Modified neurological severity score (mNSS) In all animals, a modified neurological severity score (mNSS) test was performed to assess neurological deficits daily for 3 days starting 24 h after MCAO. The mNSS was used to assess the procedures, and neurological function was graded on a scale of 0 to 14 (normal score 0, maximal deficit score 14) and was determined as a composite of motor, sensory, reflexes, and balance tests^[244]45. MRI MRI was utilized to detect brain edema after MCAO^[245]46. Briefly, anesthesia induction was carried out using 8% isoflurane. Then, mice were placed into the MRI machine with a head holder supplied with 2% isoflurane to maintain anesthesia status during the whole procedure. MRI scanning was conducted on a 7.0 T CG NOVILA spectrometer (Shanghai Chenguang Medical Technologies Co., Ltd). MRI data were viewed and analyzed by a Radiant DICOM viewer and ImageJ software, respectively. TTC staining TTC is a redox indicator to distinguish infarct tissues from normal tissue. The brains of the rats were rapidly removed under deep anesthesia (10% isoflurane) and frozen at −20 °C for 15 min. Then, the brains were sliced coronally into 2-mm thick sections using a special brain matrix for rats (RWD). After that, the sections were stained with 2% TTC (Sigma‒Aldrich) at 37 °C for 30 min. Finally, the brain slices were photographed by a digital camera and analyzed using ImageJ software. Brain tissue clarity The brain tissue clarity protocol was adapted from pervious studies^[246]47,[247]48. The left hemisphere of mouse was rinsed once with PBS, then fixed with freshly prepared 4% PFA in PBS at room temperature for 30 min on a shaker. Organoids were rinsed three times in PBS, transferred into ice cold supernatant of 2% polyglycerol 3-polyglycidylether (wt/v) in 0.1 M phosphate buffer pH 7.2 and incubated for 2 days at 4 °C. Brain tissue was subsequently transferred into pre-warmed 0.1 M sodium carbonate buffer (pH 10) and incubated at 37 °C for 24 h. Brain tissue was washed extensively with PBS for 8 h, cleared in 0.2 M SDS buffer for 48 h at 55 °C while shaking in EasyClear system (LifeCanvas Technologies), and washed extensively in PBST (PBS, 0.1% Triton X-100, 0.02% sodium azide) for 24 h. Brain tissue was mounted and imaged with a light-sheet microscope (ZEISS) equipped with two lasers (488 nm, 561 nm) and a 10× objective (ZEISS). Behavioral assessment Rat behavioral tests were performed by an independent investigator who was blinded to the experimental groups, and the data were analyzed by a separate investigator. For the open field test, general locomotor activity was evaluated in an open field apparatus (round arena, 98 cm in diameter, with 40 cm high walls) using a video-tracking program with PanLab Smart Version 2.5 software (PanLab). The arena was illuminated by a light (60 W) fixed 100 cm above the center of the arena. Rats were familiarized with the test environment before the experiment. Each rat was placed individually in the experimental box in turn, and at the beginning of the test, the rat was gently placed into the center zone of the round arena. Horizontal locomotor activity was quantified as the distance walked in cm and analyzed for the whole arena (total distance). The operation was repeated three times for each experimental animal for 15 min with an interval of 15 min. For the cylinder test, rats were placed inside a plastic cylinder (50 cm tall with a diameter of 20 cm) and videotaped for 5 min. The score was calculated as follows: (number of right hands −  number of left hands)/(number of right hands + number of left hands + number of both hands). For the adhesive removal somatosensory test, a sticker with a size of 6 × 6 (mm^2) was placed gently onto the paralyzed forepaw using forceps. The time for rats to remove each stimulus from the forelimb was recorded for three trials per day. Individual trials were separated by at least 5 min. Before surgery, animals were trained for 3 days. Once rats were able to remove the dots within 10 s, they were subjected to MCAO. The duration from the start to the time the mouse started to contact the sticker was recorded as the time to touch. The duration from the start to the time the rat successfully tears down the sticker was recorded as the time to removal. For the rotarod test, the rat was placed onto the rotating rod with an accelerating speed starting from 5 rounds per minute to 30 rounds per minute within 4 min as high speed and 5 rounds per minute to 20 rounds per minute within 5 min as low speed. Rats were forced to run three times a day for 5 days with 2 days of low speed and 3 days of high speed before the MCAO procedure until a baseline of ~250 s of staying on the rod was reached. The mean duration of staying on the rod was recorded as the latency to fall. For the grip strength test, a rat grip strength meter was used to assess forelimb strength. Animals were positioned by facing the T bar of the grip strength meter, and the forelimbs of rats were placed on the tension bar. When the rat grasped the bar, the animal was gently pulled steadily by the root of the tail away from the T bar. The grip strength meter automatically determined and recorded the maximum force displayed by each animal in grams. The mean value of five consecutive measurements for each animal was calculated. Rats were allowed to recover for 30 s between the measurements. For the Morris water maze test, the water tank was 120 cm in diameter and 50 cm in height, and both the inner walls and underwater platform in black color were used to assess the cognitive function of the rats. On the inner wall of the tank, various symbols were marked to divide the water surface into four equal quadrants. A hidden platform (12 cm diameter, 30 cm height) was placed 35 cm away from the inner wall and 2 cm beneath the water face. The water temperature was maintained at 22 ± 2 °C during the whole experiment. A video camera was placed above the water tank and was connected to the computer for analyzing movement traces. The water maze test paradigm includes a navigation task and spatial exploration session. During the navigation task, the rat was first placed on the platform for 30 s of acclimation. Then, the rat was released randomly from 1 quadrant with the head facing the wall. The time for animals to search and climb onto the platform was recorded. A maximum cut-off time of 120 s was used for rats that could not locate the platform. Four repeated sessions were performed for each rat on 4 consecutive days, with the location of the platform being fixed. The average time latency was recorded to evaluate the spatial learning ability. One day after the end of the navigation task, the hidden platform was removed, and the animal was released from each quadrant. The movement traces during 120 s were recorded. The time spent in the target quadrant, which is the original place of the hidden platform, was analyzed in addition to the total distance in the target quadrant, and crossing times were also recorded to evaluate the spatial memory of rats. PBMC isolation and co-cultured with SAOs Peripheral blood was collected from healthy donors (male, aged 18–20 yr), with approval from the Ethics Committee of the First Affiliated Hospital of Dalian Medical University (LCKY2016-60). Mononuclear cells were isolated by centrifugation using a Ficoll-Hypaque density gradient (Lymphoprep^TM, Stem cell technology). The peripheral blood mononuclear cells (PBMCs) were then washed twice with PBS and suspended in RPMI media containing 10% serum at a concentration of 1 × 10^6 cells/ml. Subsequently, PBMCs (1 × 10^5 cells) were cultured either with 0.4 μg PHA (11249738001, Sigma-Aldrich) or 5 μg SAOs in 96-well culture plates with 0.2 ml medium (RPMI with 10% FBS). Cell proliferation was assessed at 1, 3, and 7 days of culture using the CCK-8 assay. HE staining The rats were anaesthetized (10% isoflurane) and transcardially perfused with 200 ml of 5 mM sodium phosphate buffered 0.9% (w/v) saline (PBS, pH 7.2–7.4) followed by 500 ml of 4% paraformaldehyde in phosphate buffer. Then, the brains were rapidly removed. After being dehydrated in ethanol and defatted in xylene, the brain was embedded in paraffin, and 4 μm thick horizontal slices were obtained by using a rotary microtome (Leica Biosystems). For HE staining, the sections were dipped in haematoxylin for 3 min, washed in running tap water and destained in hydrochloric acid alcohol for several seconds. The sections were washed again and then dipped in eosin for 15 s. They were subsequently dehydrated in an alcohol gradient, cleared in xylene and sealed, and the plates were observed under a microscope. Routine blood tests The experimental animals were deeply anaesthetized (10% isoflurane) and fixed in the supine position on the operating table. The chest and abdomen were wiped with alcohol, and the abdominal aorta was separated by cutting the abdominal cavity along the ventral median line with surgical scissors. First, the blood vessel was fixed to avoid vessel displacement. Then, the needle was inserted into the abdominal aorta almost parallel (less than 30°) towards the end of the heart to draw the required amount of blood. After blood was collected, the needle eye was gently pressed with a cotton ball, and the needle was quickly pulled out. The collected arterial blood was tested by an automatic hematology analyzer for routine whole blood indicators (including white blood cells, red blood cells, hemoglobin, platelets, neutrophils, eosinophils, basophils, etc. Blood biochemistry The protocol was the same as that for routine blood collection, and the collected arterial blood was tested by an automatic hematology analyzer for serum biochemical indexes (including alanine aminotransferase, aspartate aminotransferase, glucose, urea nitrogen, creatinine, total cholesterol, triglyceride, total protein, albumin, etc. Statistical analysis Statistical analysis was performed using GraphPad Prism (v.9.0.2) The number of replicates and tests used are detailed in figure legends for each analysis. Reporting summary Further information on research design is available in the [248]Nature Portfolio Reporting Summary linked to this article. Supplementary information [249]Supplementary Information^ (74.8MB, pdf) [250]Peer Review File^ (6MB, pdf) [251]Reporting Summary^ (112.3KB, pdf) Source data [252]Source Data^ (85.8MB, xlsx) Acknowledgements