Super mitochondria-enriched extracellular vesicles enable enhanced mitochondria transfer

Abstract

Mitochondria transfer is a spontaneous process that releases functional mitochondria to damaged cells via different mechanisms including extracellular vesicle containing mitochondria (EV-Mito) to restore mitochondrial functions. However, the limited EV-Mito yield makes it challenging to supply a sufficient quantity of functional mitochondria to damaged cells, hindering their application in mitochondrial diseases. Here, we show that the release of EV-Mito from mesenchymal stem cells (MSCs) is regulated by a calcium-dependent mechanism involving CD38 and IP3R signaling (CD38/IP3R/Ca²⁺ pathway). Activating this pathway through our non-viral gene engineering approach generates super donor MSCs which produce Super-EV-Mito with a threefold increase in yield compared to Ctrl-EV-Mito from normal MSCs. Leber’s hereditary optic neuropathy (LHON), a classic mitochondrial disease caused by mtDNA mutations, is used as a proof-of-concept model. Super-EV-Mito rescues mtDNA defects and alleviates LHON-associated symptoms in LHON male mice. This strategy offers a promising avenue for enhancing mitochondria transfer efficiency and advancing its clinical application in mitochondrial disorders.

Introduction

Mitochondrial dysfunction is the source and key hallmark of mitochondrial disorders, with a minimum prevalence of 1 in 5000 adults¹. However, effective therapeutic strategies are currently lacking for many mitochondrial diseases in clinical practice^2,3. Notably, Leber’s hereditary optic neuropathy (LHON) refers to a classical mitochondrial disorder caused by mutations in mitochondrial DNA (mtDNA) which encodes key proteins related to mitochondrial function^4,5. Although some antioxidant drugs and nutritional support can assist in the treatment of LHON, there is currently no effective therapeutic approach targeting the mitochondrial dysfunction underlying LHON⁶.

Mitochondria transplantation is an effective approach for restoring mitochondrial functions in patients with mitochondrial diseases because the dysfunctional mitochondria are replenished by exogenous functional mitochondria^7,8,9. This approach involves directly transplanting isolated free mitochondria into the lesion site. However, studies on mitochondria transplantation are still in their relatively early stages. Since the first reported application of mitochondria transplantation in treating mitochondrial disorders in 2009, only two clinical studies have been published (ClinicalTrials.gov: NCT02586298 and NCT04976140). Advancement of this technology is hindered by the lack of selective identification of dysfunctional cells and the rapid inactivation of isolated mitochondria outside the cell^10,11. Therefore, developing an effective method to deliver functional mitochondria specifically into damaged cells is a critical yet challenging objective.

Inspired by the natural intercellular transfer of mitochondria, healthy cells can deliver mitochondria to injured cells via extracellular vesicles (EVs)¹², intercellular nanotube connections¹³, and other mechanisms in an attempt to repair mitochondrial dysfunction¹⁴. Accumulating evidence indicated that the paracrine effect of stem cells is a key mechanism underlying their ability to repair damaged cells^15,16. This is primarily because stem cells transfer healthy mitochondria to injured recipient cells via extracellular vesicles containing mitochondria (EV-Mito), thereby compensating for the functional impairment of damaged mitochondria¹⁷. In this phenomenon, EV-Mito not only protects mitochondrial activity through the membrane structure of EVs but also selectively recognizes damaged recipient cells for functional mitochondrial delivery^18,19,20,21. This approach appears to address the main challenges of mitochondria transplantation technology: the lack of selectivity and the rapid deactivation of isolated mitochondria. Consequently, some studies are now focusing on collecting EV-Mito for the treatment of mitochondrial diseases. However, these studies have uncovered a challenge: the limited mitochondrial mass of stem cells as donor cells restricts the mitochondrial content of EV-Mito^22,23. Addressing how to create super donor cells capable of generating more high-quality EV-Mito would expand this platform’s scalability and enhance its feasibility for treating mitochondrial diseases.

In this study, we demonstrate that the release of EV-Mito from mesenchymal stem cells (MSCs) is regulated by a calcium-dependent mechanism involving CD38 and inositol trisphosphate receptor (IP3R) signaling (CD38/IP3R/Ca²⁺ pathway). Furthermore, we develop our non-viral gene delivery vectors called CAP (C: N,N-cystamine-bis-acrylamide, A: agmatine dihydrochloride, P: p-aminoethylbenzenesulfonic acid) to load a CD38 plasmid, forming CAP/pCD38 complexes^24,25. These CAP/pCD38 complexes are used to genetically engineer adipose-derived MSCs, thereby activating the CD38/IP3R/Ca²⁺ pathway in MSCs. Genetically engineered MSCs using CAP/pCD38 complexes release extracellular mitochondrial particles at 3 times the level of untreated MSCs, indicating that this approach produces super donor MSCs for more EV-Mito. To validate the therapeutic efficacy of super donor MSC-derived EV-Mito (Super-EV-Mito) in mitochondrial diseases, we introduce LHON cell and LHON-like male mouse models. We find that Super-EV-Mito significantly enhances the cellular uptake of exogenous mitochondria, alleviating LHON disease symptoms. Meanwhile, Super-EV-Mito demonstrates mitochondrial function restoration in both rotenone (Rot)-induced widespread mitochondrial damage models at the cellular and mouse levels, further expanding their potential applications in other mitochondrial diseases. As mitochondria serve as the primary therapeutic agent in mitochondria transplantation, their activity and quantity inevitably influence the efficiency of mitochondria transplantation. Our study suggests that genetically engineered stem cells can generate super-donor cells, thereby providing a greater quantity and higher quality mitochondrial material for mitochondria transplantation.

Results

The release of EV-Mito from MSCs mediated by the CD38/IP3R/Ca²⁺ pathway

Mitochondria transplantation is an effective treatment for mitochondrial disorders; however, it is restricted by the limited selectivity in recipient cells and loss of mitochondrial activity in isolated mitochondria. Fortunately, mitochondria transfer as a spontaneous process selectively delivers functional mitochondria to damaged cells, thereby overcoming the aforementioned limitations²⁶. Meanwhile, accumulating evidences suggest that the paracrine effects of stem cells, including the extracellular release of mitochondrial particles, constitute a key mechanism underlying their ability to repair damaged cells¹⁵. Therefore, stem cells are widely used as donor cells in studies of mitochondria transfer^27,28. We attempted to investigate the pathways regulating the secretion of EV-Mito by MSCs and explored whether enhancing these pathways could increase the yield of EV-Mito.

Previous studies have identified that the release of extracellular mitochondrial particles by donor cells, such as astrocytes and osteoblasts, is mediated by CD38 signaling^29,30. Therefore, we aim to elucidate how CD38-mediated pathways influence the release of EV-Mito. We employed the exogenous supplementation of CD38 plasmid DNA (pCD38) to upregulate CD38 expression in MSCs and observe changes in the genetically engineered MSCs. To construct genetically engineered MSCs, we introduced our previously developed polymeric material, named CAP [C: N,N’-cystamine-bis-acrylamide (CBA), A: agmatine dihydrochloride (Agm), P: 4-(2-aminoethyl) benzenesulfonamide (PAEBS)]²⁴. Before conducting formal experiments, we confirmed the structure of the synthesized CAP and characterized the CAP/pDNA complexes. CAP was synthesized following the procedure described in a previous study (Supplementary Fig. 1) and verified using ¹H NMR spectroscopy (Supplementary Fig. 2)²⁴. Subsequently, the pDNA binding capacity of CAP was detected by agarose gel electrophoresis. The results demonstrated that CAP could completely bind pDNA at the weight ratio of 5:1 (Supplementary Fig. 3). Additionally, nanoparticle sizes and zeta potentials of the CAP/pDNA complex were approximately 102.3 ± 1.8 nm and 20.9 ± 0.9 mV, respectively (Supplementary Fig. 4). CAP contains disulfide bonds within its linkers, allowing degradation in the intracellular reductive environment of MSCs and the rapid gene release. To simulate the intracellular reductive conditions and investigate the degradation of CAP, we used dithiothreitol (DTT) to mimic the intracellular reductive environment of MSCs. The pDNA was released after 10 mM DTT treatment (2 h), indicating that the vector material had degraded for gene release (Supplementary Fig. 5)^31,32. The transfection efficiency of CAP/pGFP complexes was significantly superior to that of PEI 25 K/pGFP complexes and Lipofectamine 3000 Transfection Reagent (Lipo3000)/pGFP complexes, where PEI 25 K and Lipo3000 are widely considered the gold standards for plasmid gene transfection (Supplementary Fig. 6). One of the purposes of gene delivery of stem cells is to engineer stem cells to have certain functions or secrete certain proteins to act on target sites³³. Low toxicity gene vectors are essential for gene transfection and future applications in vivo. We performed MTT assay to investigate the cellular viability of the CAP vector. It can be found that CAP/pDNA complexes had no obvious toxicity (Supplementary Fig. 7). These results indicated that CAP offered both higher transfection efficiency and better safety compared to Lipo3000, making it more suitable for engineering MSCs in the subsequent research applications.

We obtained genetically engineered MSCs and analyzed them using gene ontology (GO) analysis. As shown in Fig. 1A, these results between normal MSCs and CD38-enhanced MSCs indicated that CD38 upregulation affected calcium-dependent pathways, accompanied by alterations in the extracellular region, extracellular space, positive regulation of cytosolic Ca²⁺ concentration and positive regulation of mitochondrial fission. Therefore, we hypothesized that the mitochondrial functions and the release of EV-Mito from MSCs may be regulated by a mechanism involving CD38 and Ca²⁺ up-regulation.

Fig. 1: CD38 signal increases mitochondrial calcium and enhances extracellular mitochondria.

A GO analysis based on the significant differentially expressed genes between untreated and CAP/pCD38-treated MSCs. B Gene expression of pCD38 in MSCs after different treatments. C Expression of IP3R in MSCs after different treatments, red: IP3R protein, blue: DAPI. D The determination of mitochondrial Ca²⁺. E Normalized fluorescence intensity of Mitotracker green (MTG) of extracellular mitochondria in a condition medium. CD38 expression (F), IP3R expression (G), content of mitochondrial Ca²⁺ (H), and normalized fluorescence intensity of MTG of extracellular mitochondria (I) with and without IP3R knockdown. J Oxygen consumption rate (OCR) from Seahorse cell mito stress test of MSCs after pCD38 transfection. K Maximum respiration of MSCs after pCD38 transfection. L MMP detection. M A schematic diagram illustrating the release of EV-Mito from MSCs regulated by the CD38/IP3R/Ca²⁺ pathway. M was created in BioRender. Wang, Y. (2025) https://BioRender.com/8fkjduj. Data are presented as mean ± SD. n = 3 biologically independent samples for each in (B, D–K), n = 5 biologically independent samples for each in (L). GO enrichment analysis was performed via two-sided hypergeometric test (A). Statistical significances were performed via one-way ANOVA with two-sided Tukey’s HSD post hoc test (B, E–I, K, L) or two-sided Games-Howell test (D).

We confirmed the expression of CD38 based on CAP. Lipo3000 was set as the positive control group. CAP-mediated plasmid gene delivery obviously promoted the expression of CD38 in MSCs compared with the untreated and Lipo3000 groups, suggesting that CAP enabled the upregulation of CD38 expression in MSCs (Fig. 1B). In addition, CD38 catalyzes the synthesis of the calcium-mobilizing second messenger cyclic ADP-ribose (cADPR)²⁹, resulting in a significant elevation of cytoplasmic calcium levels (Supplementary Fig. 8A). Mitochondria are major intracellular calcium stores^34,35. However, it is currently unclear whether CD38 regulates mitochondrial Ca²⁺ to control the release of EV-Mito in MSCs. Based on the fact that IP3R is the receptor responsible for calcium transfer from the endoplasmic reticulum to mitochondria and that most actively secreted cellular events involve calcium regulation, we examined IP3R/mitochondrial Ca²⁺ signaling as a possible mechanism driving extracellular mitochondria release from MSCs³⁶. Immunofluorescence and Western blot assays demonstrated that CAP-mediated upregulation of CD38 enhanced IP3R expression in MSCs (Fig. 1C and Supplementary Fig. 8B). Subsequently, CAP/pDNA enhanced mitochondrial Ca²⁺ level compared to the other groups (Fig. 1D). Furthermore, Mitotracker Green (MTG) was carried out to stain mitochondria of MSCs. After untreated, Lipo3000/pCD38, or CAP/pCD38 complex treatment, the normalized fluorescence intensity (FI) of MTG-labeled EV-Mito in condition medium was detected using the multifunctional microplate reader. As shown in Fig. 1E, CAP exhibited a higher normalized FI of MTG compared with the untreated and Lipo3000, revealing that the upregulation of CD38 led to an increased release of mitochondria in extracellular spaces by MSCs. These findings suggested that the upregulation of CD38 may promote IP3R expression and mitochondrial Ca²⁺ level in MSCs, enhancing an increased release of EV-Mito in the extracellular space.

Furthermore, we identified the IP3R siRNA to suppress IP3R expression. Our results showed that the CAP/pCD38 complex promoted CD38 expression, consistent with the result in Fig. 1B. However, IP3R knockdown (KD) did not affect CD38 expression (Fig. 1F). Notably, IP3R KD significantly inhibited CD38-mediated IP3R activation (Fig. 1G). These findings suggested that CD38 upregulation enhanced IP3R expression and the inhibition of IP3R did not exert a negative feedback effect on CD38 expression. In addition, the upregulation of CD38 increased mitochondrial Ca²⁺ concentration, while blocking IP3R expression significantly inhibited the CD38-induced rise in mitochondrial Ca²⁺ levels (Fig. 1H). Additionally, IP3R knockdown also reduced the release of mitochondria into the extracellular space by MSCs (Fig. 1I).

Previous studies have shown that elevated intracellular Ca²⁺ promotes vesicle-plasma membrane fusion, facilitating the release of EVs and their cargo^37,38. Pharmacological inhibition studies further demonstrated that the release of EV-Mito was markedly suppressed by actin polymerization inhibitors, whereas tubulin inhibitors had no obvious effect (Supplementary Fig. 8C, D), suggesting that EV-Mito release is actin-dependent but independent of tubulin dynamics³⁹. This mechanistic insight highlighted a calcium-sensitive, actin-dependent vesicle trafficking pathway as a key contributor to EV-Mito secretion.

To investigate whether the upregulation of CD38 gene expression mediated by the CAP/pCD38 complex negatively impacts mitochondrial function in MSCs, we assessed mitochondrial respiratory function in MSCs using the oxygen consumption rate (OCR) analysis. The OCR curve in CAP/pCD38-treated MSCs was higher than untreated normal MSCs (Fig. 1J, K), demonstrating that the upregulation of CD38 expression not only did not impair the mitochondrial respiratory chain function in MSCs but may actually enhance it. This could be due to the upregulation of CD38 activating processes such as cellular metabolic process (Fig. 1A), which in turn enhanced mitochondrial functions. MMP measurement was used to assess the effects of CAP/pCD38 treatment on mitochondrial integrity. As shown in Fig. 1L, the ratio of mitochondria with normal MMP to those with depolarized MMP in CAP/pCD38-treated MSCs was significantly higher than that in untreated MSCs, indicating that CD38-mediated Ca²⁺ elevation did not compromise MMP.

Therefore, upregulating CD38 expression in MSCs enhanced mitochondrial Ca²⁺ content through IP3R activation, leading to improved mitochondrial activity and augmented mitochondrial release as EV-Mito (Fig. 1M). At the same time, the upregulation of CD38 did not negatively affect mitochondrial respiratory function and MMP level in donor cells (MSCs).

Preparation and characterization of Super-EV-Mito

A series of experimental validations demonstrated that the release of EV-Mito from MSCs was regulated by the CD38/IP3R/Ca²⁺ pathway. Based on these findings, we genetically engineered MSCs using CAP/pCD38 complexes to create super donor MSCs. In brief, we utilized the gold-standard plasmid gene transfection reagent (Lipo3000) and our previously developed non-viral gene delivery vector (CAP) to load the pCD38, forming Lipo3000/pCD38 and CAP/pCD38 complexes. These complexes were used to genetically engineer adipose-derived MSCs, thereby activating the CD38/IP3R/Ca²⁺ pathway in MSCs. After treating MSCs with Lipo3000/pCD38 or CAP/pCD38 for 4 h, the medium was replaced with a fresh medium. MSCs were incubated for an additional 20 h or 44 h. The condition medium was centrifuged at 300 × g for 10 min to remove cell debris, followed by centrifugation at 2500 × g for 10 min to eliminate residual impurities. The supernatant was centrifuged at 18,000 × g for 30 min to obtain EV-Mito as a pellet. In subsequent experiments, EV-Mito collected from untreated MSCs was designated as Ctrl-EV-Mito, this from Lipo3000/pCD38-treated MSCs as Lipo-EV-Mito, and this from CAP/pCD38-treated MSCs as Super-EV-Mito (Fig. 2A).

Fig. 2: Preparation and characterizations of Super-EV-Mito.

A Extraction and purification of Super-EV-Mito. B Particle sizes of Super-EV-Mito. C Structure illumination microscopy (SIM) images of Super-EV-Mito, representative images from n = 3 independent experiments. D The proportion of released mitochondria encapsulated in Super-EV-Mito using nano-flow cytometry analysis. FITC: mEmerald-TOMM20 labeled mitochondria, PC5: CD81 in EVs. E The proportion of released mitochondria within EVs calculated by TEM images. F Western blots of EV-Mito and mitochondrial markers in different EV-Mito groups, representative images from n = 3 independent experiments. G The MFI and positive ratio of MTG of EV-Mito using flow cytometry analysis. H MitoSOX fluorescence intensity normalized by MTG intensity. I Mitochondrial basal respiration detection in different EV-Mito. i: Standard curve between mitochondrial protein and fluorescence intensity of Mitotracker Green probe; ii: OCR normalized by mitochondrial protein in different EV-Mito groups. J The stability of mitochondria in EV-Mito by monitoring changes in mitochondrial basal respiration. A was created in BioRender. Wang, Y. (2025) https://BioRender.com/1bhqw40. Data are presented as mean ± SD. n = 3 biologically independent samples for each in (E, G–J). Statistical significances were performed via one-way ANOVA with two-sided Tukey’s HSD post hoc test (G (ii), I (ii)), two-sided Games-Howell test (G (i), H), or two-sided t-test (J). Simple linear regression by GraphPad software (I (i)).

Characterization of the CAP/pCD38-treated EV-Mito (Super-EV-Mito) revealed the particle sizes ranging from 350 to 950 nm (Fig. 2B). The zeta potential of Super-EV-Mito was −1.76 ± 0.27 (Supplementary Table 1). Multimodality Structured Illumination Microscopy (multi-SIM) imaging system was used to observe Super-EV-Mito. Before multi-SIM imaging, MSC membranes and mitochondria were labeled with DiO and Mitotracker Red dyes, respectively. As shown in Fig. 2C, DiO-labeled membrane structures enveloped Mitotracker Red-labeled mitochondria. Mitotracker Red fluorescence signals were observed within individual Super-EV-Mito, confirming that Super-EV-Mito contained mitochondria. It has been reported that not all released mitochondria are encapsulated within EVs; a portion may indeed exist in the form of free mitochondria. We conducted nano-flow cytometry analysis to quantify the proportion of released mitochondria encapsulated in EVs. The result showed that approximately 24.87 ± 2.76% of mitochondrial particles were double-positive, indicating their encapsulation within EVs, while the remaining mitochondria were FITC-positive, suggesting that they existed as free mitochondria (Fig. 2D). However, the double-positive population in nano-flow cytometry only reflects the proportion of mitochondria associated with EVs and does not directly visualize how many mitochondria are present within individual EVs. To address this, we further quantified the mitochondrial content within EVs using transmission electron microscopy (TEM) analysis. Statistical evaluation of Super-EV-Mito TEM images revealed that 6 out of every 10 released mitochondria were encapsulated in EVs, while the remaining 4 existed in a free form (Fig. 2E and Supplementary Fig. 9). There is heterogeneity in EV-Mito structures.

Super-EV-Mito contains more and better functional mitochondria

To further investigate whether the quantity and quality of mitochondria contained within Super-EV-Mito are improved after activation of the CD38/IP3R/Ca²⁺ pathway, we used Western blot assay and flow cytometry for assessing the quantity of EVs collected from each group. Subsequently, we evaluated the mitochondrial quality within the EVs using mitochondrial ROS level and basal respiration detection.

Western blot assay was carried out to explore the expressions of key EV membrane markers (CD81) and mitochondrial proteins (MT-ND4 and COX IV) under equal protein loading conditions. The bands for the key EV membrane proteins and mitochondrial proteins in the Super-EV-Mito were stronger compared to those in the Ctrl-EV-Mito and Lipo-EV-Mito, indicating an increased release of EV-Mito in the Super-EV-Mito group (Fig. 2F). Flow cytometry was further employed to quantify mitochondrial content, based on MTG mean fluorescence intensity (MFI) and positivity ratios. As shown in Fig. 2G, the MTG positivity ratio was 17.7% in the Ctrl-EV-Mito group, while the MTG positivity ratios in the Lipo-EV-Mito and Super-EV-Mito groups were significantly higher, reaching 51.7% and 55.2%, respectively. The data confirmed that the activation of the CD38/IP3R/Ca²⁺ pathway led to the generation of super-donor MSCs, from which a greater quantity of mitochondria-containing EVs (Super-EV-Mito) could be collected.

Furthermore, we assessed the mitochondrial activity in the Super-EV-Mito by measuring mitochondrial ROS and basal respiration. MitoSOX probe was employed to assess mitochondrial ROS levels within EV-Mito, reflecting the functional activity of mitochondria in these EVs. The MitoSOX/MTG intensity in Super-EV-Mito was lower than the EV-Mito group, suggesting that CD38 activation enhanced the quality of mitochondria encapsulated within these EVs (Fig. 2H). In addition, we investigated whether Super-EV-Mito contains more functional mitochondria by assessing the basal respiration. We established a standard curve correlating Mitotracker Green fluorescence intensity with mitochondrial protein concentration (Fig. 2I (i)). Based on this calibration, we quantified the basal respiration per unit of mitochondrial protein using OCR assays. As shown in Fig. 2I (ii), the basal respiration per unit mitochondrial protein in Super-EV-Mito was significantly higher than that in Ctrl-EV-Mito and Lipo-EV-Mito.

We evaluated the mitochondrial function in EV-Mito at different time points to reflect the stability of mitochondria within EV-Mito. We isolated free mitochondria from CAP/pCD38 complex-treated MSCs as a control group (called isolated mitochondria from Super MSCs). As shown in Fig. 2J, the basal respiration of Super-EV-Mito, as measured by OCR, remained relatively stable for up to two days. In contrast, the mitochondrial function of isolated mitochondria from Super MSCs dropped significantly starting on the first day, showing lower activity compared to the Super-EV-Mito group. The finding indicated that mitochondria encapsulated within EVs exhibited superior functional stability, maintaining their activity for at least two days. We speculated that this stability was mainly attributed to the membrane structure of Super-EV-Mito, which protected mitochondrial activity. To investigate this hypothesis, we compared the mitochondrial activity in Super-EV-Mito and isolated mitochondria under conditions of high extracellular Ca²⁺. The rapid deactivation of isolated mitochondria in the extracellular environment is primarily attributed to the interference of high extracellular Ca²⁺ concentrations⁴⁰. The Ca²⁺ concentration gradient (0, 0.0001, 1.25, and 2.5 mM) was used to simulate the mitochondrial functionality of Super-EV-Mito under cytosolic, extracellular, and blood calcium conditions, respectively. Accordingly, we assessed changes in MMP using the MMP-dependent Mitotracker Red probe (MTR) to characterize mitochondrial functional alterations under high extracellular Ca²⁺ concentrations. We measured the FI of MTR in Super-EV-Mito and isolated mitochondria from Super MSCs without any treatment at 0 h using a multifunctional microplate reader, designated as FI_S and FI_I, respectively. After 24 h of treatment with varying concentrations of Ca²⁺, the FI of Super-EV-Mito and isolated mitochondria was measured. The normalized FI of MTR for Super-EV-Mito = FI after treatment/FI_S × 100%. Similarly, the normalized FI of MTR for isolated mitochondria from Super MSCs = FI after treatment/FI_I × 100%. As the concentrations of Ca²⁺ increased, the normalized FI of MTR in the Super-EV-Mito group remained above 70%, whereas the normalized FI of MTR in isolated mitochondria showed an obvious decrease compared with Super-EV-Mito. Our results demonstrated that the membrane structure of Super-EV-Mito protected mitochondria from Ca²⁺-induced MMP loss, suggesting that Super-EV-Mito was likely to preserve mitochondrial function under high Ca²⁺ conditions and resist external stress (Supplementary Fig. 10).

Therefore, the Super-EV-Mito collected by activating the CD38/IP3R/Ca²⁺ pathway in MSCs exhibited a higher quantity and mitochondrial activity compared to the Ctrl-EV-Mito collected from normal MSCs. This also indicated that the activation of the CD38/IP3R/Ca²⁺ pathway led to the generation of super-donor cells, thereby holding the potential to provide a greater quantity and higher quality mitochondrial material for mitochondria transfer therapy.

Super-EV-Mito repairs mitochondrial functions in LHON cell model

LHON is a classical ocular mitochondrial disease resulting from mutations in mtDNA that encode key proteins related to mitochondrial function. Abnormalities in mtDNA encoding three subunits of complex I in the mitochondrial respiratory chain are primary pathogenic factors in more than 95% of LHON cases. Notably, 90% of Asian LHON patients are closely associated with a homoplasmic nucleotide substitution of guanine to adenine at position 11778 in mtDNA⁴¹. The mutations at this site lead to a functional abnormality in the key protein ND4 of mitochondrial complex I, causing mitochondrial dysfunctions, including mitochondrial respiratory chain dysfunction, decreased ATP generation, and MMP depolarization⁴². To investigate the therapeutic effects of Super-EV-Mito in mitochondrial diseases, we utilized LHON model cells (GM10742 cells) and evaluated the mitochondrial functional restoration capacity of different formulations after administration⁴³.

The cellular uptake of Super-EV-Mito in GM10742 cells was explored by flow cytometry analysis. We transduced MSCs with a viral vector encoding mEmerald-TOMM20 to specifically label mitochondria within the MSCs. We then utilized mEmerald-TOMM20-labeled Ctrl-EV-Mito, Lipo-EV-Mito, and Super-EV-Mito to track uptake in recipient cells. Ctrl-EV-Mito, Lipo-EV-Mito, and Super-EV-Mito exhibited varying levels of mitochondrial uptake in recipient GM10742 cells. Among these, Super-EV-Mito increased the MFI of mEmerald-TOMM20 compared to the other EV-Mito, likely due to the higher content of mitochondria in Super-EV-Mito (Fig. 3A).

Fig. 3: Super-EV-Mito restores the mitochondrial functions of GM10742 cells.

A Cell uptake of different groups evaluated by flow cytometry analysis. B The expression of MT-ND4 and COX IV protein detected by Western blot assay, representative images from n = 3 independent experiments. The quantitative data (C) and CLSM images (D) of MMP detection. E OCR from Seahorse cell mito stress test of GM10742 cells in various treatment groups. Spare respiratory capacity (F) and maximum respiration (G) of GM10742 cells after varying treatment groups. H The ATP contents. I Detection of mitochondrial ROS level. J The mPTP opening detection. K Cell proliferation evaluated using EdU assay. L KEGG pathway enrichment analysis based on the significant differentially expressed genes between untreated GM10742 cells and Super-EV-Mito-treated GM10742 cells. Data are presented as mean ± SD. n = 3 biologically independent samples for each in (A, E–K). n = 8 biologically independent samples for each in (C). Statistical significances were performed via one-way ANOVA with two-sided Tukey’s HSD post hoc test (A, F, H, J, K) or two-sided Games-Howell test (C, G, I). KEGG pathway enrichment analysis was performed via two-sided hypergeometric test (L).

Subsequently, we systematically investigated whether the introduction of EV-Mito could effectively improve mitochondrial function in GM10742 cells with mtDNA mutations. The expression of mitochondrial ND4 protein (MT-ND4) in GM10742 cells was limited due to the mutations at position 11778 in mtDNA. Super-EV-Mito promoted the MT-ND4 expression in GM10742 cells (Fig. 3B and Supplementary Fig. 11A). Super-EV-Mito treatment increased COX IV expression compared with Ctrl-EV-Mito treatment (Fig. 3B and Supplementary Fig. 11B). This demonstrated that EV-Mito could increase the levels of mitochondrial proteins. The JC-1 is a fluorescent dye commonly used to assess MMP^44,45. In functional mitochondria with high MMP, JC-1 aggregates form and emit red fluorescence. JC-1 remains in its monomeric form when MMP is low. The green fluorescence of monomers in the untreated and Ctrl-EV-Mito-treated GM10742 cells was higher than that in the Lipo-EV-Mito and Super-EV-Mito-treated GM10742 cells. In contrast, the red fluorescence of aggregates in the untreated GM10742 cells was weaker than that in the Lipo-EV-Mito and Super-EV-Mito-treated GM10742 cells. The ratio of red to green fluorescence provides a quantitative measure of MMP, making it a reliable indicator of mitochondrial health and function. The value of red/green fluorescence intensity after Super-EV-Mito treatment was obviously higher than the untreated group (Fig. 3C, D). It suggested that untreated GM10742 cells exhibited abnormalities in MMP, which could be improved through the supplementation of exogenous mitochondria provided by Super-EV-Mito. Furthermore, the mitochondrial respiratory chain function of GM10742 cells after treatment with different formulations was evaluated using OCR assays. Super-EV-Mito significantly enhanced the spare respiratory capacity and maximum respiration of GM10742 cells, indicating that Super-EV-Mito treatment improved the mitochondrial respiratory chain function in mtDNA-mutant GM10742 cells (Fig. 3E–G). Additionally, intracellular ATP levels and mitochondrial ROS levels were measured, revealing that Super-EV-Mito effectively increased ATP production while reducing mitochondrial ROS levels in GM10742 cells (Fig. 3H, I). The opening of the mPTP is another critical indicator of mitochondrial functionality. Excessive opening of mPTP is often associated with mitochondrial dysfunction. The degree of mPTP opening was assessed using Calcein fluorescence. As shown in Fig. 3J, untreated GM10742 cells exhibited relatively low Calcein fluorescence. After treatment with Ctrl-EV-Mito, Lipo-EV-Mito, and Super-EV-Mito, Calcein fluorescence values progressively increased, with the highest levels observed in the Super-EV-Mito group, suggesting that Super-EV-Mito produced the best therapeutic effect on mitochondrial function. Furthermore, the EdU assay was applied to investigate cellular proliferation. Due to mitochondrial dysfunction caused by mtDNA mutations, untreated GM10742 cells displayed limited proliferative capacity. Super-EV-Mito provided “energy” to GM10742 cells by supplementing exogenous mitochondria, thereby enhancing their proliferative capacity (Fig. 3K). After being treated with Super-EV-Mito, GM10742 cells were collected for KEGG pathway enrichment analysis. The significantly differentially expressed genes between untreated GM10742 cells and Super-EV-Mito-treated GM10742 cells were enriched in mitochondrial metabolism, AMPK signaling pathway, and mitophagy (Fig. 3L).

We further investigated the long-term therapeutic potential of Super-EV-Mito by examining both their persistence in GM10742 cells and their sustained ability to restore mitochondrial function. Our results showed that mtDNA originating from donor rat mitochondria could persist in GM10742 cells for up to 5 days post-administration (Supplementary Fig. 12A). Furthermore, a single administration of Super-EV-Mito led to significantly improved mitochondrial function in GM10742 cells, which was maintained for up to 5 days (Supplementary Fig. 12B).

In summary, Super-EV-Mito effectively delivered exogenous mitochondria to LHON model cells, restoring mitochondrial functions, including MMP, ATP production, mitochondrial ROS levels, and mPTP opening. This restoration enhanced the proliferative capacity of LHON model cells. Furthermore, as Super-EV-Mito contain a greater quantity of mitochondria with superior functionality, the therapeutic efficacy of the Super-EV-Mito treatment group in vitro was superior to the Ctrl-EV-Mito treatment group.

Super-EV-Mito inhibits vision loss in LHON-like male mouse model

Given the exceptional mitochondrial function restoration achieved by Super-EV-Mito in GM10742 cells, we investigated the therapeutic potential of Super-EV-Mito for mitochondrial disease in the LHON male mouse model. The safety of Super-EV-Mito was assessed in vivo. Various doses of Super-EV-Mito were set and administrated in the normal C57BL/6J male mice, including 1-fold (Super-EV-Mito-1), 2-fold (Super-EV-Mito-2) and 4-fold (Super-EV-Mito-4) Super-EV-Mito. Different doses of Super-EV-Mito were administered again 7 days after the first administration. 14 days after the second administration, whole blood and serum samples were collected from mice in each group for routine hematological analysis and liver and kidney function tests (Supplementary Tables 2 and 3). As shown in Supplementary Fig. 13A, B, even at high doses, Super-EV-Mito-4 did not induce systemic toxicity in mice, nor did it cause significant damage to the retina or major organs. However, direct transplantation of MSCs into the eye appeared to negatively impact the structural integrity of the retinal layers and led to elevated expression of inflammation-related proteins (Supplementary Fig. 13C)^46,47. Therefore, Super-EV-Mito exhibited a high level of in vivo safety. Prior to conducting formal efficacy studies in vivo, we assessed the in vivo persistence of donor-derived mtDNA in ocular tissues to evaluate the long-term therapeutic potential of Super-EV-Mito. Following a single intravitreal administration, donor rat mtDNA remained detectable in the eye for up to 14 days (Supplementary Fig. 12C).

LHON model male mice were purchased from Aurora Bioscience Co., Ltd. (Suzhou, China). A mitochondria-targeted adeno-associated virus (MTS-AAV) containing the mutant human ND4 gene (mutND4), followed by mitochondrial-encoded mCherry, was microinjected into zygotes. Female founders (F0) exhibiting mCherry fluorescence on ophthalmoscopy were backcrossed with normal males for producing the first generation of offspring (F1): mutant ND4 C57BL/6 J male mice⁴⁸. The therapeutic effect of Super-EV-Mito was evaluated in LHON model male mice (Fig. 4A). Specifically, three-month-old LHON model male mice received an initial treatment, followed by a second treatment seven days later. After two weeks, the optomotor test and electroretinography (ERG) assessments were conducted on all groups of mice. Furthermore, the eyeballs were collected for intraocular ATP level measurement and retinal H&E sections. Additionally, we introduced Idebenone (Ide), the only drug currently approved for adjunctive treatment of LHON, as a positive control group⁴⁹. Visual recovery in each group was evaluated using the optomotor test and ERG assessments. For the optomotor test, a rotating drum with alternating black-and-white striped circular panels was constructed. The drum was rotated clockwise for 2 min, followed by 1 min of rest, and then rotated counterclockwise for another 2 min. Video recordings were used to count the number of head turns made by each mouse within the 2-minute rotation periods. The results showed that wild-type male mice (WT group) exhibited an average of 15.00 ± 1.41 head movements/2 min, while PBS-treated LHON model male mice showed only 6.71 ± 0.76 head movements/2 min, indicating severe visual impairment in the LHON male mice. Ide treatment via daily oral administration resulted in partial visual recovery, whereas mice treated with Super-EV-Mito demonstrated significantly better visual recovery compared to the Ide group (Fig. 4B). ERG, as an important diagnostic tool for detecting retinal and choroidal hereditary diseases, was utilized to evaluate retinal function in mice after different treatments. ERG assessments were conducted on mice following 24 h of dark adaptation. The ERG waveforms of WT male mice, PBS-treated, Ide-treated, and Super-EV-Mito-treated LHON model male mice were shown in Fig. 4C. Statistical analysis of the amplitude variations in a-wave and b-wave revealed that PBS-treated LHON model male mice showed a significant decrease in the absolute amplitudes of both the a-wave and b-wave relative to wild-type males. However, treatment with Super-EV-Mito markedly restored the absolute values of both waves, with the b-wave amplitude recovering to levels comparable to those of wild-type male mice (Fig. 4D, E). H&E staining sections were carried out to analyze the retinal structure (Fig. 4F). Retinal layer thickness and RGC counts were quantified from H&E-stained sections. Super-EV-Mito significantly increased retinal layer thickness and restored RGC numbers (Fig. 4G, H). Both parameters in the Super-EV-Mito-treated group showed no significant differences compared to WT mice, indicating that Super-EV-Mito effectively repaired retinal structures. Further investigation into the total expression of MT-ND4 protein in LHON model male mice after different treatments revealed that Super-EV-Mito treatment restored ND4 protein expression in the eyes by supplying exogenous functional mitochondria. Western blot analysis of mitochondrial COX IV protein showed that mitochondria transplantation mediated by Super-EV-Mito increased the number of exogenous mitochondria, thereby elevating COX IV protein expression in the eyes of the mice (Fig. 4I, J). ATP generation in the eyes of the mice was further assessed. The mitochondrial dysfunction caused by mtDNA mutations resulted in significantly lower ATP content in the eyes of PBS-treated LHON model male mice compared to wild-type mice. Super-EV-Mito treatment significantly increased ATP content in the eyes (Fig. 4K).

Fig. 4: The therapeutic effect of Super-EV-Mito in the LHON male mouse model.

A Flowchart for Super-EV-Mito treatment in the LHON male mouse model. B The optomotor test (CW clockwise, CCW counterclockwise). C ERG waveforms in varying treatment group. Amplitude variations of a- (D) and b-wave (E) in ERG. F H&E sections of the different treatment groups. Thickness of general retina (G) and retinal ganglion cells (RGC) numbers (H) were summarized to evaluate the restoration of retinal structure. I The expression of MT-ND4 and COX IV protein detected by Western blot assay, representative images from n = 3 independent experiments. J The quantitative results of MT-ND4 and COX IV protein expression were obtained using ImageJ software. K The ATP contents after varying treatments. A and B were created in BioRender. Wang, Y. (2025) https://BioRender.com/vmapofh. Data are presented as mean ± SD. n = 3 eyes for each in (D, E, G, H, J, K). n = 7 independent process of optomotor tests for each in (B). Statistical significances were performed via one-way ANOVA with two-sided Tukey’s HSD post hoc test (B, D, E, G, H, J, K).

Expand the application of Super-EV-Mito in a generalized rotenone-induced mitochondrial disease model

Inspired by the therapeutic efficacy of Super-EV-Mito in the mtDNA mutant LHON male mice, we investigated their potential to supplement exogenous functional mitochondria for the treatment of a broader range of mitochondrial disorders, beyond primary mitochondrial disorder stemmed from mtDNA mutations. To this end, we utilized Rot, a chemical inhibitor of mitochondrial complex I, to establish a Rot-induced mitochondrial disease model. As noted in previous studies, Rot-induced cell and male mouse models have been widely used as models for various mitochondrial diseases, including LHON and Parkinson’s disease. The binding sites of Rot are located within complex I, specifically targeting regions such as decylubiquinone and the ND4 subunit, resulting in complex I dysfunction^50,51.

The Rot-damaged cell model was established by pre-treating HeLa cells with 0.1 µM Rot, and the cellular uptake of exogenous mitochondria and mitochondrial functional restoration effects of Super-EV-Mito were assessed in this model. Mitochondria from donor cells (MSCs) were labeled by mEmerald-TOMM20, and mitochondria-enriched vesicles were obtained following untreated, Lipo3000/pCD38, or CAP/pCD38 transfection. Rot-induced cells were treated with EV-Mito from different groups, and the MFI of mEmerald-TOMM20 in Rot-induced cells was analyzed using flow cytometry and confocal laser scanning microscopy (CLSM). The data exhibited that Rot-damaged cells treated with Super-EV-Mito exhibited the highest fluorescence intensity, indicating that Super-EV-Mito-mediated mitochondria transfer was feasible even in Rot-damaged cells (Fig. 5A, B). In addition, we conducted uptake inhibition experiments to investigate the mechanism by which mEmerald-TOMM20-labeled Super EV-Mito was internalized into the cytoplasm of recipient cells. As shown in the Supplementary Fig. 14, we found that Super-EV-Mito entered recipient cells via multiple endocytic pathways, with macropinocytosis and caveolae-mediated endocytosis being the predominant mechanisms.

Fig. 5: Protection effect of the Super-EV-Mito in Rot-induced cell and mouse models.

Cell uptake of mEmerald-TOMM20-labeled EV-Mito evaluated by flow cytometry analysis (A) and CLSM (B), green: mEmerald-TOMM20-labeled EV-Mito. CLSM images (C) and quantitative data (D) of MMP detection. E Western blots of MT-ND4 and COX IV protein expressions. F The ATP contents after varying treatments. G Mitochondrial ROS level detected by MitoSOX probe. H The detection of mPTP opening. I Cell proliferation evaluated using EdU assay. J Flowchart for Super-EV-Mito treatment in the Rot-induced LHON male mice. K Optomotor test. L Immunofluorescence staining of mitochondrial complex I after varying treatments (red: mitochondrial complex I; blue: nucleus), GCL ganglion cell layer, INL inner nuclear layer, ONL outer nuclear layer. M Representative H&E images in the varying treatment groups. N Thickness of the general retina after different treatments. O RGC numbers in different treatment groups. J was created in BioRender. Wang, Y. (2025) https://BioRender.com/xjzc6tt. Data are presented as mean ± SD. n = 3 biologically independent samples for each in (A, F–I, N, O). n = 7 biologically independent samples for each in (D), n = 7 independent process of optomotor tests for each in (K). Statistical significances were performed via one-way ANOVA with two-sided Tukey’s HSD post hoc test (A, D, G–I, N, O) or two-sided Games-Howell test (F, K).

In Rot-induced cells, mitochondrial damage led to reduced MMP, whereas Super-EV-Mito treatment restored MMP to levels comparable to the PBS group without Rot exposure (Fig. 5C, D). Given that Rot also damages the ND4 protein site within mitochondrial complex I, the Rot + untreated group exhibited lower MT-ND4 expression than the untreated group. Super-EV-Mito restored MT-ND4 expression by introducing exogenous mitochondria (Fig. 5E and Supplementary Fig. 15). Moreover, Ctrl-EV-Mito, Lipo-EV-Mito, and Super-EV-Mito all increased ATP content in Rot-damaged cells (Fig. 5F) and reduced mitochondrial ROS accumulation induced by Rot (Fig. 5G). Super-EV-Mito markedly reduced mPTP opening, as indicated by higher Calcein Green fluorescence (Fig. 5H). Most importantly, all vesicle formulations enhanced the proliferation capacity of Rot-induced cells to varying degrees, with Super-EV-Mito showing the most pronounced effect (Fig. 5I). These findings suggested that Super-EV-Mito not only restored mitochondrial functions in mtDNA-mutated LHON cells but also regulated mitochondrial function in Rot-induced cells with widespread mitochondrial damage, ultimately rescuing cell proliferation activity.

Based on our research and studies from other investigators, a Rot-induced male mouse model was established to simulate ocular mitochondrial diseases associated with widespread mitochondrial dysfunction by intravitreal injection of 2.5 mM Rot^42,52. The first treatment was administered on day 7 post-modeling, followed by a second treatment on day 14. On day 28, visual function in mice was evaluated (Fig. 5J). Optomotor response testing showed that Super-EV-Mito significantly restored visual impairment caused by Rot (Fig. 5K). Rot exposure led to a decline in complex I expression in the retinal layers, particularly in the retinal ganglion cell layer (GCL). Treatment with Super-EV-Mito promoted the upregulation of complex I expression (Fig. 5L). H&E staining revealed that Rot induced severe retinal structural damage, including a marked reduction in retinal layer thickness and RGC numbers. After Super-EV-Mito treatment, both retinal layer thickness and RGC counts were significantly improved (Fig. 5M–O).

A series of studies conducted on gene-mutated LHON cells and the LHON male mouse model demonstrated that Super-EV-Mito could effectively supplement exogenous mitochondria via efficient mitochondria transfer, thereby restoring mitochondrial function and achieving effective treatment of LHON (Fig. 6). Furthermore, Super-EV-Mito effectively mitigated disease progression by enhancing mitochondria transfer efficiency and promoting the mitochondrial functions in the Rot-induced pan-mitochondrial disease model.

Fig. 6: A schematic illustration of Super-EV-Mito for the treatment of LHON.

A The construction of CAP/pCD38 complexes. B The generation of super donor cells for EV-Mito and a schematic representation of Super-EV-Mito for mitochondrial repair and delaying the progression of LHON. A and B were created in BioRender. Wang, Y. (2025) https://BioRender.com/z7mnb7d. CAP a non-viral gene delivery vector, EV-Mito extracellular vesicle containing mitochondria, MSC mesenchymal stem cells.

Discussion

Mitochondrial diseases, characterized by mitochondrial dysfunction, affect a minimum prevalence of 1 in 5000 adults. However, the majority of mitochondrial diseases currently lack effective treatments. Mitochondria transfer is a natural process where functional mitochondria are released into damaged cells via EV-Mito to restore mitochondrial function under pathological conditions. To overcome the challenges of limited yield and quality of mitochondria-enriched EVs, which hinder their therapeutic potential for mitochondrial diseases, we employed our previously engineered non-viral transfection vector (CAP) to deliver the pCD38 into MSCs. This approach activates the CD38/IP3R/Ca²⁺ pathway, thereby generating super donor cells. These super donor cells produced EV-Mito (termed Super-EV-Mito) containing 3 times the mitochondrial content of normal MSC-derived EVs, with significantly enhanced mitochondrial quality. This strategy not only enabled the collection of a greater number of high-quality EV-Mito, reducing the production cost of Super-EV-Mito-mediated mitochondria transfer therapy but also utilized the membrane structure of EVs to preserve mitochondrial activity.

We explored the mechanism by which CD38 signaling promotes the release of EV-Mito: (1) CD38 catalyzes the synthesis of the calcium-mobilizing second messenger cADPR, resulting in a significant elevation of cytoplasmic calcium levels; (2) Mitochondria are important intracellular calcium stores. We observed that upregulation of CD38 expression led to increased expression of IP3Rs on the ER. These IP3Rs rapidly mediate Ca²⁺ transfer to mitochondria. (3) Previous studies have shown that elevated intracellular Ca²⁺ levels can trigger the fusion of vesicles with the plasma membrane, driving vesicle-plasma membrane fusion and facilitating EV release. As shown in Fig. 2G, the MTG positivity ratio was 17.7% in the Ctrl-EV-Mito group, while the MTG positivity ratios in the Lipo-EV-Mito and Super-EV-Mito groups were significantly higher, reaching 51.7% and 55.2%, respectively.

To further determine whether activation of the CD38/IP3R/Ca²⁺ signaling pathway enhances both the quantity and quality of mitochondria encapsulated within Super-EV-Mito, we conducted a series of assays including Western blot, flow cytometry, and TEM to quantify EVs derived from different groups. In addition, we assessed mitochondrial quality by mitochondrial ROS level and basal respiration detection. These results collectively demonstrated that activation of the CD38/IP3R/Ca²⁺ axis induces the generation of super donor MSCs, which produce EVs with both increased mitochondrial payload and improved mitochondrial functionality. This strategy may thus serve as a promising approach for enhancing the therapeutic efficacy of mitochondria transfer.

In both a typical ocular mitochondrial disease model of LHON and a Rot-induced mitochondrial disease model, Super-EV-Mito alleviated mitochondrial dysfunction-associated vision loss in mice. In mitochondrial diseases, pathogenic mutations in mtDNA often coexist with wild-type mtDNA in a state known as heteroplasmy. The severity of such diseases is closely linked to the proportion of mutant mtDNA, with a well-recognized threshold effect-clinical symptoms typically emerge when the proportion of mutant mtDNA exceeds ~60%. Consequently, strategies aimed at reducing the mutant load below this pathological threshold are considered promising therapeutic avenues. Mitochondria transfer represents one such strategy by introducing exogenous, healthy mitochondria into affected cells. This approach can restore mitochondrial function through two complementary mechanisms: (1) providing wild-type mtDNA to dilute the mutant load and shift the heteroplasmic ratio, and (2) stimulating mitophagy to selectively eliminate dysfunctional mitochondria enriched in mutant mtDNA. Importantly, the therapeutic benefit of mitochondria transfer does not rely on directly editing or repairing mutant mtDNA, but rather on reestablishing mitochondrial homeostasis through the combined effects of mitochondrial supplementation and mitophagy activation.

Although our results demonstrated that CD38 overexpression significantly enhanced mitochondrial activity and functionality in both MSCs and their derived Super-EV-Mito, it is important to note that such alterations in mitochondrial function may also influence MSC metabolism and differentiation potential. To minimize these potential risks, we limited the transfection procedure to a single round per MSC batch and collected EVs only once from each batch. This precautionary measure was taken to avoid repeated use of the same MSC population for EV production. In future studies, we aim to further optimize MSC engineering and EV harvesting strategies to maximize the proportion of mitochondria encapsulated within EVs while minimizing production costs, thereby enhancing the translational feasibility of this approach.

Therefore, this study suggests that engineering super donor cells is an advanced source for the scalable production of EV-Mito to treat mitochondrial disorders. This approach offers a promising strategy for treating primary mitochondrial diseases such as LHON caused by mtDNA mutations, while also opening a avenue for addressing secondary mitochondrial dysfunction associated with broader disease contexts.

Methods

Materials

Cystamine dihydrochloride, acryloyl chloride, agmatine dihydrochloride (Agm), and anhydrous DMSO from J&K Chemical (Beijing, China). 4-(2-aminoethyl) benzenesulfonamide (ABS) from Shanghai Macklin Biochemical Co., Ltd. (Shanghai, China). Mitotracker Red or Green probe, MitoSOX probe, Mitochondria Isolation Kit, Lipofectamine 3000 (Lipo3000), LysoTracker Deep Red probe, SDS-PAGE and agarose from Thermo Fisher Scientific (Waltham, USA). Bicinchoninic acid (BCA) and Dulbecco’s Modified Eagle Medium (DMEM) from KeyGEN Biotech (Nanjing, China). MMP Kit of JC-1, 4,6-diamino2-phenylindole (DAPI), Hoechst 33342, Calcein-AM, Enhanced ATP Assay Kit (s0027), HRP-Goat Anti-Mouse IgG (A0216), HRP-Goat Anti-Rabbit IgG (A0208), Alexa Fluor 647-Goat Anti-Rabbit IgG (A0468) and BeyoClick™ EdU Cell Proliferation Kit with TMB (C0088S) from Beyotime Biotechnology (Shanghai, China). β-actin Rabbit mAb (AC026), COX IV Rabbit pAb (A6564), IP3R Rabbit mAb (A23760), GFAP Rabbit pAb (A14673), CD81 Rabbit mAb (A22528) and MT-ND4 Rabbit pAb (A17970) from ABclonal Technology (Wuhan, China). Anti-NDUFB8 Recombinant Rabbit mAb (ET7108-25) from HuaBio (Hangzhou, China). Anti-Vimentin mAb (ab92547), Anti-Complex I mAb (ab109798) and Rhod-2 AM (ab142780) were purchased from Abcam (Cambridge, UK). The siRNA targeting IP3R (siIP3R) from Riobio Company Limited (Supplementary Sequence 1, Guangzhou, China). Idebenone (Ide) from Tokyo chemical industry (Tokyo, Japan). Rotenone (Rot) from Aladdin Bio-Chem Technology (Shanghai, China). NaCl, NaOH, and HCl were purchased from Yuanye Bio-Technology (Shanghai, China).

Synthesis of CAP

CBA is one of the monomers of CAP and obtained on the basis of the previous synthetic procedure²⁴. Firstly, cystamine dihydrochloride was dissolved in water at a concentration of 50 mM. Then the solution was transferred at a volume of 50 mL into a three-necked flask. Acryloyl chloride (0.1 M) was dissolved in dichloromethane at a volume of 10 mL. After that, NaOH solution (100 mM, 10 mL) was prepared. 10 mL of acryloyl chloride (0.05 M), as well as NaOH solution, were transferred in two separate dripping funnels. Then the two components were added drop by drop in a three-necked flask at the same time. The mixing process lasted for 60 min at 0 °C while stirring. Another 6 h was spent to perform the reaction at room temperature (r.t.). Finally, dichloromethane was used to dissolve the mixture. The purification of CBA was performed by crystallization using ethyl acetate.

CAP was synthesized by Michael’s addition reaction in which the N, N-cystamine-bis-acrylamide (C), agmatine dihydrochloride (A), and p-aminoethylbenzenesulfonic acid (P) were used as monomers. The three monomers were dissolved in methanol by the molar ratios as follows: n (A):n (P):n (C) = 1.2:0.8:2. The reaction should avoid the light and be protected by argon for 24 h at 90 °C after adding triethylamine. Then, the pH of the product was adjusted by 100 mM HCl until four. Impurities were removed from CAP to ultrapure water by a 24-hour-long dialysis procedure with a Dialysis membrane (MWCO = 500), and then CAP was obtained by lyophilization. ¹H Nuclear magnetic resonance (¹H NMR) was performed to confirm the structures.

The plasmid construction

The CD38 plasmid DNA (pCD38, Supplementary Sequence 2) and plasmid GFP (pGFP) was constructed. The pDNA used in this study was amplified in E. coli strain DH5α, and then an Endo-free Plasmid Maxi Kits were performed to extract pDNA.

Preparation and characterization of CAP/pDNA complexes

The pDNA (pCD38 or pGFP) was dissolved in H₂O. Anhydrous dimethyl sulfoxide (DMSO) was used to dissolve and store CAP (200 mg/mL). The CAP mother solution was diluted with H₂O into various concentrations. Then pDNA solution was added drop by drop to CAP solution in vortex condition. Then the incubation of the mixture (the weight ratio of pDNA and CAP = 1:30) was performed for 0.5 h at 4 °C to acquire CAP/pDNA complexes. Besides, the reagent protocol of Lipo3000 was referred to prepare the Lipo3000/pDNA complexes. Dynamic light scattering analyzer (DLS) performed by Nano ZS90 zetasizer was used to detect the particle sizes and zeta potentials of NPs.

Isolation of MSCs

The enzymatic digestion method was used to isolate MSCs from the epididymal adipose tissue²⁵. Adipose tissues from Sprague-Dawley rats were collected and digested by type I collagenase. Then, DMEM with 10% FBS was used to culture MSCs. After the 3-day culture, unadhered impure cells were removed by fresh medium. Cells were cultured when 90% confluence was achieved.

Cell culture

HeLa cells (human breast cancer cell line) were obtained from the Shanghai Institute of Cell Bank, Chinese Academy of Sciences (Shanghai, China). GM10742 cells were purchased from the Coriell Institute for Medical Research. MSCs and HeLa cells were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin, while GM10742 cells were maintained in RPMI-1640 containing 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. All cells were incubated at 37 °C in a humidified atmosphere with 5% CO₂. For all cell experiments in this study, the untreated group served as the blank control.

Rot-induced HeLa cells

HeLa cells were constantly exposed to either an FBS-free medium or a Rot-contained FBS-free medium at the concentration of 0.1 μM for 12 h. Finally, the medium was replaced for the subsequent studies.

In vitro gene transfection

MSCs were seeded on 24-well culture plates. Medium was replaced by different formulations at a concentration of 1 μg pDNA/well diluted by DMEM without FBS. The Lipo3000 group served as the positive control.

For transfection assay, pGFP was used for detection. The different formulations were incubated with the cells for 4 h. Medium was replaced with 1 mL fresh DMEM. Another 20 h cultivation was performed before detection. Finally, the positive rate (%) and mean fluorescence intensity were assessed by flow cytometry.

Extraction and purification of extracellular vesicles

MSCs were cultured in tissue culture flasks with a 75 cm² growth area (T75). After 48 h, MSCs were transfected by the different formulations (10 μg pCD38/flask).

The culture medium was replaced with EV-depleted medium and collected after 48 h. The medium was sequentially centrifuged at 300 × g for 10 min, 2500 × g for 10 min, and finally at 18,000 × g for 30 min at 4 °C to isolate EV-Mito from the supernatant. EVs marker expression of CD81 and the mitochondrial marker expression of COX IV and MT-ND4 were measured by Western blot.

Characterization of EV-Mito

Ctrl-EV-Mito, Lipo-EV-Mito and Super-EV-Mito were extracted from MSCs treated with PBS, Lipo3000/pCD38 and CAP/pCD38. EV-Mito was resuspended at 0.1 mg/mL (protein concentration) in PBS for DLS and zeta potential detections.

For nanoparticle tracking analysis (NTA) detection, EV-Mito was resuspended in 40 μL PBS and measured using a ZetaView Particle Metrix (Particle Metrix, PMX-120, Germany). Polystyrene beads at 100 nm (Thermo Fisher Scientific, Fremont, CA) were used to calibrate the instrument. NTA software was carried out to identify the concentrations of nanoparticles.

For multi-SIM imaging system (NanoInsights-Tech Co., Ltd.), EV-Mito was resuspended at 0.1 mg/mL (protein concentration). DiO and Mitotracker Red were used to label the EV-membrane and mitochondria. The multi-SIM images were taken using single slice mode with 50 mW laser power and 30 ms exposure time. Images were then reconstructed using the SIM Imaging Analyser software⁵³.

For nano-flow detection, MSCs were transfected with mEmerald-TOMM20 using recombinant adenovirus (Ad-mEmerald-TOMM20) at a virus titer of 2.0 × 10⁸ PFU/mL for 24 h. Then, MSCs were transfected with CAP/pCD38 complexes. The EV-Mito was extracted and purified. CD81 rabbit mAb was used as the label of EV-membrane. ABflo^® 647-conjugated Goat anti-Rabbit IgG (H + L) was set as the secondary antibody. The proportion of released mitochondria encapsulated in EVs were analyzed using the Flow NanoAnalyzer U30E (NanoFCM, Xiamen, China).

For transmission electron microscopy (TEM) imaging, EV-Mito was pipetted onto a 150-mesh copper and incubated for 3 min. Mesh copper was stained with 2% uranyl acetate for 8 min, and washed with distilled water. Then it was stained with 2.6% lead citrate. The images were recorded by TEM (HT7800, HITACHI, Japan). The TEM image analysis was performed using the software ImageJ.

For flow cytometry analysis, MSCs were cultured in T75 flasks. After 48 h, MSCs were treated with the different formulations (10 μg pCD38/flask). Then the medium was replaced with 100 nM of Mitotracker Green (MTG) for 45 min in serum-free medium. The EV-Mito was extracted and washed with PBS for flow cytometry analysis.

For Ca²⁺ protection, Super-EV-Mito was obtained and mixed with a series of varying concentrations of Ca²⁺ for 24 h at 4 °C. The fluorescence intensity of Mitotracker Red was detected by a multifunctional microplate reader.

For pharmacological inhibition studies, MSCs were seeded on 24-well culture plates, transfected with CAP/pCD38 complex, and stained with MTG. Then the medium was replaced by an EV-depleted medium with different inhibitors (cytochalasin B (20 μM), Paclitaxel (20 nM)) and collected after 24 h. Then, EV-Mito was extracted and purified for the flow cytometry analysis.

CD38 and IP3R expressions

MSCs were seeded on 24-well culture plates and transfected with pCD38 at a concentration of 1 μg pCD38/well. The untreated group was the negative group and the Lipo3000 group served as the positive control. The PE-labeled Anti-CD38 antibody was used for immunofluorescence detection. In addition, the Anti-IP3R antibody and Alexa Fluor 647-Goat Anti-Rabbit IgG antibody were used for immunofluorescence detection. The fluorescence intensity was assessed by flow cytometry.

Mitochondrial calcium detection

MSCs were seeded on 24-well culture plates and transfected with CAP/pCD38. The untreated group was the negative group and the Lipo3000 group served as the positive control. The rhod-2-AM (2 μM) was stained for 30 min at 37 °C. The MFI of rhod-2 was assessed by flow cytometry.

Cytoplasmic calcium detection

MSCs were seeded on 24-well culture plates and transfected with CAP/pCD38 complexes. The fluo-4-AM (2 μM) was used to stain the cytoplasmic calcium for 30 min at 37 °C. The MFI of fluo-4 was assessed by flow cytometry.

Extracellular mitochondria detection

MSCs were seeded on 10 cm culture dishes and transfected with pCD38. MTG was carried out to stain mitochondria in MSCs. After 48 h of Lipo3000/pCD38 or CAP/pCD38 complex treatment, condition medium was collected and the normalized FI of MTG-labeled EV-Mito in condition medium was detected using a multifunctional microplate reader.

In vitro pharmacodynamic studies

Cell uptake

MSCs were transfected with mEmerald-TOMM20 using recombinant adenovirus (Ad-mEmerald-TOMM20) at a virus titer of 2.0 × 10⁸ PFU/mL for 24 h. Then, MSCs were transfected with different treatment groups. The EVs were extracted and purified. The Rot-induced HeLa cells were plated in the 35 mm glass bottom dishes (8 × 10⁴ cells/well). The GM10742 cells were plated in the 24-well plate at 1 × 10⁵ cells/well. Then cells were treated at 6 μg EV protein/well in complete growth medium for 24 h. After washing by PBS, cell uptake was detected by CLSM (Zeiss LSM800, Germany) and flow cytometry.

Single-dose expression duration of Super-EV-Mito

The GM10742 cells were plated in the 24-well plate at 1 × 10⁵ cells/well and cultured with different EV-Mito at 6 μg EV-Mito protein/well. After different time points (1, 2, 3, 5 days), cells were collected for qPCR detection. The primers of rat MT-ND4 DNA were as follows: Forward 5’-CGATCCATTATCCACCCCAC-3’, Reverse 5’- TGTTGGGATTAGAGTGGCTTCG-3’, and the primers of human GAPDH DNA were: Forward 5’- GGTCTGAGGTTAAATATAGCTGCTG-3’, Reverse 5’-TTGATTTGCCAAGTTGCCTG-3’. The relative copy numbers of rat MT-ND4 were quantified through 2^{−(∆∆Ct)} by real-time PCR analysis (CFX connect, Bio-Rad, USA).

MMP detection

The HeLa cells or Rot-induced HeLa cells were plated in the 35 mm glass bottom dishes (8 × 10⁴ cells/well). Cells were cultured without Mito-EVs (untreated group) or with Ctrl-EV-Mito, Lipo-EV-Mito and Super-EV-Mito (6 μg EV-Mito protein/well). After 48 h of incubation, the cells were covered by 1 mL 1 × JC-1 dye for 15 min. Then, the cells were observed with CLSM at 488 nm (JC-1 monomers) and 543 nm (JC-1 aggregates) excitation.

Adenosine triphosphate (ATP) measurements

GM10742 cells, HeLa cells, or Rot-induced HeLa cells were plated at 24-well plates overnight and then treated with different groups at 6 μg EV-Mito protein/well for 48 h. Then the ATP Bioluminescent Assay Kits were used to assess the ATP content in different cells. The luminescence was recorded by chemiluminometer (Luminoskan Ascent, Thermo Fisher Scientific, USA).

Seahorse analysis

The oxygen consumption rate (OCR) of MSCs and GM10742 cells was determined with a Seahorse XFe96 Analyzer (Agilent, USA). MSCs were plated onto a cell culture plate at 3 × 10⁴ cells per well for 24 h, then transfected with Lipo3000/pCD38 or CAP/pCD38 complexes for 12 h before Seahorse analysis. In this test, the medium was replaced with the assay medium, the medium (pH 7.4) containing glucose (10 mM) (Agilent, USA), L-glutamine (2 mM) and sodium pyruvate (1 mM) (Gibco, USA). After equilibration, the OCR was measured steadily with the treatment of oligomycin (2 μM), FCCP (5 μM), rotenone and antimycin A (1 μM). GM10742 cells were plated onto a 24-well plate and treated with Ctrl-EV-Mito, Lipo-EV-Mito and Super-EV-Mito (6 μg EV-Mito protein/well) for 48 h. On the test day, the GM10742 cells were collected and resuspended with the assay medium, then seeded onto the cell culture plate at 1 × 10⁴ cells per well. After equilibration, the OCR was measured.

The OCR of different EV-Mito and isolated mitochondria was performed by the similar method as previous study⁵⁴. In brief, EV-Mito and isolated mitochondria were suspended in mitochondrial assay solution (MAS: 220 mM mannitol, 70 mM sucrose, 10 mM KH₂PO₄, 5 mM MgCl₂, 2 mM HEPES, 1 mM EGTA and BSA 0.2% (w/v)) and seeded to XFe96 microplate at a volume of 20 μL/well. The microplate was centrifuged at 2000 × g for 20 min at 4 °C. After centrifugation, 180 μL of 37 °C-preheated MAS containing substrate (10 mM pyruvate, 2 mM malate, 4 μM FCCP) was added to each well. The OCR was measured by XFe96 analyzer.

Evaluation of mPTP opening

The mPTP opening was measured with Calcein-AM and CoCl₂. In brief, the cells in each group were treated, collected and cultured with Calcein-AM (2 μM) and CoCl₂ (1 mM) at 37 °C for 30 min. Finally, cells were washed with 1 mM CoCl₂ on a shaker incubator for 25 min. The MFI of Calcein was assessed by flow cytometry.

Determination of mitochondrial ROS

GM10742 cells, HeLa cells, or Rot-induced HeLa cells were plated at 24-well plates overnight and then treated with Ctrl-EV-Mito, Lipo-EV-Mito and Super-EV-Mito (6 μg EV-Mito protein/well) for 48 h. Then the cells were incubated with 500 nM of MitoSOX (M36007, Thermo Fisher Scientific, USA) at 37 °C for 30 min. The MFI of MitoSOX was assessed by flow cytometry.

Cell proliferation assay

GM10742 cells, HeLa cells or Rot-induced HeLa cells were treated with Ctrl-EV-Mito, Lipo-EV-Mito and Super-EV-Mito (6 μg EV-Mito protein/well) for 48 h. Cell proliferation was detected with a BeyoClick™ EdU Cell Proliferation Kit with TMB (C0088S, Beyotime Biotechnology, China). The absorbance was investigated by the multifunctional microplate reader at 370 nm (Thermo Fisher Scientific Multiskan GO, USA).

RNA sequencing

Total RNA was extracted using Total RNA Kit (R6834, Omega, USA). Conduct quality control using NanoDrop One spectrophotometer (NanoDrop Technologies, Wilmington, DE), Qubit 3.0 Fluorometer (Life Technologies, Carlsbad, USA), and electrophoresis. RNA purity was checked using the kaiao K5500^® Spectrophotometer (Kaiao, Beijing, China). RNA integrity and concentration was assessed using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). The library preparation using VAHTS Universal V6 RNA-seq Library Prep Kit for MGI (Vazyme, Nanjing, China). RNA sequencing was performed using DNBSEQ T7 Sequencer (Complete Genomics, USA).

Raw sequencing data were processed with FastQC for quality control, and clean reads were aligned to the rat or human reference genome using HISAT. Transcript abundance was quantified with RSEM and expressed as FPKM. Differentially expressed genes (DEGs) were identified by comparing treatment and control groups (fold change ≥ 2.0, posterior probability ≥ 0.8, FDR-adjusted P < 0.05). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed with clusterProfiler (v3.14.3), with terms considered significant at adjusted P < 0.05.

Western blot analysis

The cells were collected for extracting proteins which were quantized by the BCA Protein Kit. The samples were detected by SDS-PAGE and the protein bands were transferred into the NC membranes. After incubation for 2 h with 5% skim milk, the membranes were put with primary antibody (MT-ND4, GAPDH and COX IV antibody) at 4 °C all night. Next, the membrane was cultured with the secondary antibodies (2 h) and the results were evaluated using electrogenerated chemiluminescence (ECL, Tanon, China).

Eyes were collected for the extraction of protein. Then the proteins were detected by using the same method as cells. MT-ND4 and COX IV antibodies were used for analysis, and GAPDH was used as a loading control.

Animals

C57BL/6J male mice (4-6 weeks) were purchased from East China Normal University Laboratory Animal Technology Co. Ltd. (Shanghai, China) and mutant mtND4R340H mitochondria transgenesis (mtTg) male mice (3 months of age) were purchased from Aurora Bioscience Co. Ltd. (Suzhou, China). All mice were housed under a 12 h light-dark cycle at 25 °C with 50% relative humidity. Mice were randomly assigned to pharmacodynamic studies.

In vivo pharmacodynamic study

Three-month-old mutant mtND4R340H mtTg LHON male mice were randomly divided into 3 groups. Mice receiving an intravitreal injection of 2 μL PBS or Super-EV-Mito (1 μg EV-Mito protein/eye) were designated as the PBS group and the Super-EV-Mito group, respectively. Mice administered Ide orally at a daily dose of 60 mg/kg were designated as the Ide group. Age-matched wild-type male mice were included as the WT group. After 2 weeks of the second time of administration, the optomotor test and ERG were performed. The eyes were harvested for further analysis. Each group contained 5 mice.

To establish the Rot-induced LHON male mouse model, the mice were injected with 3 μL of 2.5 mM Rot (dissolved in trilaurin) into the vitreous chamber⁵. After model establishment, male Rot-induced LHON mice were randomly divided into five groups. Mice receiving an intravitreal injection of 2 μL PBS were designated as the Rot + PBS group. Mice treated intravitreally with Ctrl-EV-Mito, Lipo-EV-Mito, or Super-EV-Mito (1 μg EV-Mito protein/eye) were defined as the Rot + Ctrl-EV-Mito group, Rot + Lipo-EV-Mito group, and Rot + Super-EV-Mito group, respectively. Mice administered Ide orally at a daily dose of 60 mg/kg were designated as the Rot + Ide group. Age-matched male mice without Rot were included as the PBS group. Each group contained 5 mice.

For the optomotor test, a circular platform (6 cm radius) in black alternating with a white electric roller was settled to put the mouse. Mice were put to habituate the experiment for 5 min. There was an interval of 30 s between the clockwise (2 min) and counterclockwise (2 min). The movement times of the head were counted. A movement was defined as qualified if the mouse’s head moved with the roller while the body stayed still.

For ERG measurement, mice were dark-adapted overnight, anesthetized, and placed on the recording platform. Corneal ring electrodes and subcutaneous needle electrodes in the skin and tail were positioned, and dark-adapted 0.01 ERG responses (flash intensity = 10 cd·s/m²) were recorded using the Espion Visual Electrophysiology System (Diagnosys LLC, USA).

The mitochondrial complex I expression was analyzed using immunofluorescence.

The eyes of mice were collected for H&E staining. The thickness of the general retina layer and the number of retinal ganglion cells (RGC) were analyzed.

Safety evaluation in vivo

Each group contained 5 mice. After 2 weeks of the second time of administration, the eyes, heart, kidney, liver, lung and spleen of mice were collected for H&E staining. During the dosing period, the mice were weighed every two days until the 28^th day. Then the mice were sacrificed, and the blood was taken for further experiments. After blood was obtained, the blood was placed in an EP tube at r.t. for 1 h and centrifuged at 900 × g for 10 min to collect the serum. The serum samples were taken for ALT, AST, and BUN using the relevant assay kits.

Single-dose expression duration of Super-EV-Mito in vivo

The two eyes of all groups’ mice were injected with Super-EV-Mito (1 μg EV-Mito protein/eye). After different days (1, 2, 3, 5, 7, 14 days) of treatment, eyes were collected. The qPCR was performed to measure the content of rat MT-ND4 DNA. The primers of rat MT-ND4 DNA were as follows: Forward 5’-CGATCCATTATCCACCCCAC-3’, Reverse 5’-TGTTGGGATTAGAGTGGCTTCG-3’, and the primers of mouse Actin DNA were: Forward 5’-GATCACTCAGAACGGACACCAT-3’, Reverse 5’-GGCTCATCAAATGCCCACA-3’. The relative copy numbers of rat MT-ND4 were quantified through 2^{−(∆∆Ct)} by real-time PCR analysis (CFX connect, Bio-Rad, USA).

Statistics and reproducibility

In Fig. 1E, normalized FI of MTG = FI of MTG/protein content. In Fig. 2H, MitoSOX/MTG intensity = FI of MitoSOX/FI of MTG. In Figs. 3H and 4K, ATP content = ATP concentration of cells/protein concentration. In Fig. 5F, data was normalized by cell numbers. Quantitative data in these experiments were presented as mean ± standard deviation (SD) from sample numbers (n). And the comparison of the mean values between multiple groups adopted the one-way analysis of variance (ANOVA) test, the pairwise comparisons adopted Tukey’s HSD post hoc test when the data satisfied the homogeneity of variance, and the Games-Howell test was used for the comparison between multiple groups if the above conditions were not met the homogeneity of variance. And the pairwise comparisons adopted t-test. All tests were two-sided, the P > 0.05 presented not significant (NS), P < 0.05 presented significant, and P < 0.01, P < 0.001 and even P < 0.0001 presented highly significant. Comparisons of all groups were analyzed using the SPSS 23.0. Statistical tests utilized for each experiment and the reproducibility of experiments were specified in the legends of figures. No data were excluded from the analysis. The ImageJ software (National Institutes of Health, USA) was used for quantitative analysis in fluorescence intensity for confocal images and Western blot.

Ethical statement

All the animal protocols and procedures were performed under the guidelines for human and responsible use of animals in research approved by the regional ethics committee of China Pharmaceutical University (2024-07-036) and JOINN Laboratories Co., Ltd. (Suzhou, China) (S-ACU24-0975).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.