ATP Synthase Abundance in Neuronal Extracellular Vesicles Reflects Changes in the Mitochondria of Parent Neurons

ABSTRACT

Mitochondrial proteins are found in extracellular vesicles (EVs) such as neuron‐derived EVs (NEVs). Yet whether and how NEV‐borne mitochondrial proteins relate to the state of mitochondria in the parent neurons is unclear. Studying the mitochondrial ATP synthase in primary hippocampal neurons and their released EVs, we discovered that the abundance of ATP synthase in NEVs echoes the catalytic activity level of ATP synthase in neurons. We also observed, unexpectedly, that within the neuron, the quantity of ATP synthase remains constant irrespective of the level of its activity. Using non‐canonical amino acid tagging coupled with proximity ligation assay, we found that the amount of nascent ATP synthase is linearly correlated to its activity, which may contribute to maintaining the overall quantity of ATP synthase in the neuron stable. Furthermore, we identified a sub‐population of mitochondria‐derived vesicles (MDVs) that carry ATP synthase and are not targeted to lysosomal degradation. Our findings suggest a strategy used by neurons in regulating and fine‐tuning mitochondrial ATP synthase through MDV and NEV generation. Further studies are needed to elucidate the relationship between ATP synthase–containing‐NEVs and ‐MDVs.

1. Introduction

Mitochondrial ATP synthase, also known as F₀F₁ ATP synthase, is a key enzyme for cell respiration through oxidative phosphorylation (reviewed in von Ballmoos et al. 2009; Junge and Nelson 2015; Weber and Senior 2000). In typical cells, oxidative phosphorylation in mitochondria generates 36 molecules of ATP from one molecule of glucose, whereas only two molecules of ATP are generated by glycolysis in the cytoplasm (Vives‐Bauza et al. 2007). The ATP synthase is comprised of two parts, F₀ and F ₁, and each part is composed of multiple subunits (von Ballmoos et al. 2009; Junge and Nelson 2015; Weber and Senior 2000). An observed—but not yet understood—biochemical feature of ATP synthase is its long half‐life (Dorrbaum et al. 2018; Fornasiero et al. 2018; Bomba‐Warczak et al. 2021; Krishna et al. 2021). Whereas the average half‐life of most cellular proteins is 5.4 days, the mean half‐life of the ATP synthase (for different subunits) is 11.2 days (Dorrbaum et al. 2018; Fornasiero et al. 2018). The long lifespan of the ATP synthase is somewhat unexpected and seems preposterous given high levels of reactive oxygen species (ROS) inside mitochondria (Cadenas and Davies 2000; Adam‐Vizi and Chinopoulos 2006; Murphy MP 2009; Rexroth et al. 2012; Willems et al. 2015). ROS‐elicited damage of mitochondrial proteins has been strongly implicated in the pathogenesis of neurodegenerative diseases (Raha and Robinson 2000; Balaban et al. 2005; Ebanks and Chakrabarti 2022). Yet our understanding of the strategies that cells and neurons implement to safeguard mitochondrial proteins and renew ROS‐damaged proteins is limited.

Extracellular vesicles (EVs) are nano‐sized membranous particles that contain a vast variety of cargoes (Couch et al. 2021; Nogueras‐Ortiz et al. 2024). Recent work has found that mitochondrial proteins, including ATP synthase, are among the cargoes of EVs and neuronal EVs (NEVs) (Crewe et al. 2021; Peruzzotti‐Jametti et al. 2021; Rabas et al. 2021; Todkar et al. 2021; Vasam et al. 2021; Vilcaes et al. 2021). Furthermore, the NEV mitochondrial protein cargoes are aberrant in individuals with Alzheimer's disease (Yao et al. 2021), Parkinson's disease (Picca et al. 2020), Down syndrome (D'Acunzo et al. 2021), multiple sclerosis (Ladakis et al. 2022) and Fragile X‐associated tremor/ataxia syndrome (Yao et al. 2024), suggesting that mitochondrial protein abnormalities may be a common feature across many brain diseases. EVs have been viewed by analogy as containing ‘message(s) in a bottle’, with their cargo reflecting abnormalities in their parent cells. It is unclear, however, how mitochondrial components in EVs/NEVs are related to mitochondria in their originating cells.

In this study, we investigated ATP synthase–containing NEVs and mitochondria in their parent cells, using primary hippocampal neurons as a model system. In response to various treatments, we observed concerted changes in the levels of ATP synthase in NEVs and the levels of ATP synthase catalytic activity in the neurons, with the amount of neuronal ATP synthase remaining unchanged. We also found that changes to ATP synthase activity lead to corresponding changes in the production of nascent ATP synthase in neurons.

2. Materials and Methods

2.1. Antibodies and Chemicals

The following antibodies (Abs) were used in this study. Dilutions of the antibodies used are provided in the respective section describing immunoblotting, immunofluorescence light microscopy, immunogold electron microscopy or FUNCAT‐PLA.

Primary antibodies are as follows: mouse anti‐VAMP2 Ab (104211, Synaptic System); rabbit anti‐VAMP2 Ab (104008, Synaptic System); mouse anti‐L1CAM Ab (554273, BD Biosciences), rabbit anti‐L1CAM Ab (ab208155, abcam), rabbit anti‐Alix Ab (NBP1‐90201, Novus), mouse anti‐OXPHOS Ab cocktail (ab110413, abcam), rabbit anti‐ATP synthase_α Ab (176569, abcam), mouse anti‐ATP synthase‐β Ab (MAB3494, Sigma), rabbit anti‐TOM20 Ab (ab186735, abcam), mouse anti‐PDH Ab (ab110333, abcam), guinea pig anti‐MAP2 Ab (188004, Synaptic Systems), mouse anti‐synaptophysin Ab (S5768, Sigma), guinea pig anti‐GAD Ab (198104, Synaptic Systems), rabbit anti‐cFos Ab (2250, Cell Signaling), rabbit anti‐biotin Ab (5597, Cell Signaling), mouse anti‐biotin Ab (B7653, Sigma), mouse or rabbit anti‐actin Ab (A5441 or A2066, Sigma).

Secondary antibodies are as follows: (1) immunofluorescence and flow cytometry: goat anti‐mouse IgG conjugated Alexa Fluor 488 or Fluor Alexa 568 or Fluor Alexa 647 (A11029 or A11031 or A31571, ThermoFisher Scientific), goat anti‐rabbit Alexa Fluor 488 or Alexa Fluro 568 (A11034 or A11036, ThermoFisher Scientific), goat anti‐guinea pig Alexa Fluor 568 or Alexa Fluor 633 (A11075 or A21105); (2) immunoblots: HRP linked, anti‐mouse IgG or anti‐rabbit IgG (7076 or 7074, Cell Signaling); (3) immunogold electron microscopy: goat anti‐mouse 20‐nm gold or goat anti‐rabbit 10‐nm gold (17121‐5 or 17013‐3, Ted Pella).

The following chemicals and reagents were used in this study: poly‐L‐lysine (P2636, Sigma); Neurobasal medium (2100304) and B‐27 supplement (17504044) were from ThermoFisher Scientific; RIPA buffer (89900) and Halt Protease and Phosphatase inhibitor cocktail (78444) were from ThermoFisher Scientific; poly ethylene glycol (92897, Sigma); 16% paraformaldehyde solution (15710) and 8% glutaraldehyde (16020) were from EMS; ATP synthase microplate assay (ab109716, abcam); Shh was prepared exactly as described (Yao et al. 2015; Yao et al. 2017); SAG (Smoothened agonist; ALX‐270‐426 M001, ENZO Life Sciences); Humanin (330936‐69‐1, Peptides Sciences); Bicuculline (14340), Glutamine (G8540), MG‐132 (474791), Bafilomycin A (B1793), Antimycin A (A8674), and Anisomycin (A5862) were from Sigma; Methionine‐free NbActiv medium was custom‐made by TransnetYX; methionine (J61904.22), Click‐IT AHA (azidohomoalanine; C10102), Biotin Alkyne (B10185), TBTA [Tris(1‐benzyl‐1H‐1,2,3‐triazolylmethyl]amine (H66485), TCEP [Tris‐(2‐carboxyethyl)phosphine hydrochloride] (PG82080) were from ThermoFisher Scientific); CuSO₄ (PHR1477), Duolink in situ PLA probe anti‐rabbit PLUS (DUO92002), Duolink in situ PLA probe anti‐mouse MINUS (DUO92004), and Duolink in situ detection reagents Green (DUO92014) were from Sigma.

2.2. Primary Hippocampal Neuron Cultures

Timed pregnant female Sprague‐Dawley rats (Charles River) were used as the source of embryonic brain tissues for establishing cultures of hippocampal neurons. All animal procedures were approved by the NIA Animal Care and Use Committees and complied with the NIH Guide for Care and Use of Laboratory Animals.

Primary hippocampal neuron cultures were prepared from embryonic day 18 rats as previously described (Kaech and Banker 2006; Yao et al. 2017). Hippocampi were dissected, and dissociated cells were grown in Neurobasal medium supplemented with B27 either on poly‐L‐lysine (1 mg/mL) coated glass coverslips (#1.5), or on poly‐L‐lysine (0.1 mg/mL) coated plastic culture dishes. Neurons were used at 6–8 DIV (days in vitro) for most experiments and at 14–20 DIV for Bicuculline experiments.

2.3. NEV Isolation

NEVs were isolated from conditioned media of cultured hippocampal neurons by ultracentrifugation. Approximately, 24‐mL neuron‐conditioned medium was collected from two 6‐well plates of cultured neurons. All subsequent steps were carried out at 4°C. Cells and large cellular debris were removed by centrifugation at 2000 × g for 10 min. The supernatant was centrifuged at 120,000 × g for 2 h (twice) using a SW 32 Ti rotor (Beckman Coulter) in an Optima XE‐90 ultracentrifuge (Beckman Coulter). The NEV pellet was re‐suspended in 10–20 µL of 0.2 µm–filtered PBS or in RIPA buffer, depending on subsequent experiments.

2.4. Nanoparticle Tracking Analysis (NTA)

NEV diameter and concentration were estimated using NanoSight NS500, with a G532‐nm laser and NTA 3.1 nanoparticle tracking software. Before introducing to the instrument, NEV samples were diluted in 0.2 µm–filtered particle‐free PBS to achieve an optimal reading range of the instrument (∼100 particles per field of view). Readings were obtained according to the manufacturer's instructions. NTA data were exported to Excel, and particle concentrations for each band of diameter were calculated.

2.5. Immunoblotting

We used a standard protocol for immunoblotting. Cultured hippocampal neurons were scraped from plates, collected and lysed on ice in 50–100‐µL RIPA buffer containing Halt Protease and Phosphatase inhibitor cocktail. Purified NEVs were lysed in 10 µL of the same lysis buffer used for neurons. Protein concentrations in neurons and NEV samples were measured using the BCA assay. Protein samples (20 µg of proteins from neuronal lysates and 3–5 µg of proteins from NEV lysates) were heated at 95°C for 10 min, resolved in 4%–20% polyacrylamide gels, and transferred to nitrocellulose membranes. The membranes were incubated with antibodies (primary Abs for overnight at 4°C and secondary Abs for 1 h at RT). ATP synthase Ab, VAMP2 Ab, TOM20 Ab and synaptophysin Ab were used at 1:1000, OXPHOS Ab was used at 1:500, actin Ab was used at 1:5000; HRP‐linked secondary Abs were used at 1:2000. Bound antibodies were visualized by the enhanced chemiluminescence method.

2.6. Immunofluorescence Labelling of NEVs

NEVs were immunolabelled using a protocol described by Mondal et al. (2019) with minor modifications. Purified NEVs were permeabilized in 0.001% Triton X‐100 (in PBS, pH 7.4) for 5 min, and an equal volume of 20% polyethene glycol (PEG, 10,000) was added to achieve a final concentration of 10% PEG. The mixture was centrifuged at 3500 × g for 5 min at room temperature. After removing supernatant, the NEV‐containing pellet was suspended in 200 µL of PBS.

Primary antibodies were added to the NEV‐containing suspension. The antibodies used in this experiment were as follows: ATP synthase Abs, L1CAM Ab, VAMP2 Ab, CD81 Ab, CD63 Ab and Alix Ab. All primary antibodies were used at 1:50. The NEV‐antibody mixtures were incubated overnight at 4°C. After adding 200 µL of 20% PEG to the 200 µL of NEV‐antibody suspension, the mixture was centrifuged at 3500 × g for 5 min. The supernatant was removed, the pellet was re‐suspended in 200 µL of PBS, and 200 µL of 20% PEG was added, and the mixture was centrifuged at 3500 × g for 5 min. The PEG precipitation of NEVs was repeated for the third time, and the resultant pellet was then suspended in 200 µL of PBS.

Alexa Fluor 488 or 568 conjugated secondary antibodies were added to the primary antibody‐ NEV suspension at 1:200 dilution and incubated for 1 h. The mixture was then subjected to three rounds of PEG precipitation and wash (identical to the above steps for primary antibody), and the pellet containing immunolabelled NEVs was suspended in 10–20 µL of PBS. Approximately, 10 µL of labelled NEVs was placed on a microscope slide, mounting medium was added and covered with a glass coverslip (#1.5). Samples were imaged within 24 h after labelling.

2.7. Image Acquisition and Presentation of Immunofluorescence‐Labelled NEVs

Images of immunofluorescence‐labelled NEVs were acquired with a 100×/1.46‐numerical aperture oil objective lens on a Zeiss LSM 980 confocal microscope with Airyscan. The 488‐nm laser was used for Alexa Fluor 488, and the 561‐nm laser was used for Alexa Fluor 568. Lasers were used at 1% power or less. Detection gain was adjusted to cover the full dynamic range and to avoid saturated pixels. All images were acquired in SR configuration 8‐bit mode with 1224 × 1224‐pixel xy resolution. Raw czi image files were processed into de‐convoluted Airyscan images using the Zen Blue software, and tiff image files were exported for analysis and presentation.

2.8. Electron Microscopy and Immunoelectron Microscopy of NEVs

Electron microscopy and immunogold electron microscopy of NEVs were carried out following a protocol described by Thery et al. (2006). For electron microscopy, purified NEVs were fixed in 2% paraformaldehyde, deposited onto Formvar‐carbon coated EM grids, washed in PBS, fixed again in 1% glutaraldehyde and rinsed in distilled H₂O. NEV samples were then contrasted in uranyl‐oxalate solution (pH 7), embedded in 2% methyl cellulose and 4% uranyl acetate at a ratio of 9:1. After removing excess solution by blotting on filter paper, the grids with NEVs were air‐dried and stored in a grid storage box.

For immunogold electron microscopy, purified NEVs were fixed in 2% paraformaldehyde, deposited onto Formvar carbon–coated grids, washed twice in PBS, three times in 50 mM glycine in PBS and blocked in 5% BSA. NEVs were then incubated with VAMP2 Ab (1:50), L1CAM Ab (1:25) or ATP synthase Ab (1:25) in blocking solution for 2 h. Control samples were incubated in blocking solution alone. After washing in wash buffer (0.1% BSA in PBS), NEV samples were incubated with goat anti‐mouse 20‐nm gold or goat anti‐rabbit 10‐nm gold, at 1:20 in blocking solution for 30 min. Labelled NEVs were washed in PBS, post‐fixed in 1% glutaraldehyde, rinsed in distilled H₂O, then contrasted and embedded as described above.

NEV samples were examined and imaged in a JEOL JEM‐2100 transmission electron microscope equipped with an Orius SC1000B camera (Gatan) and DigitalMicrograph software.

2.9. Flow Cytometry Analysis (FCA)

Flow cytometry was performed using our previously described nanoscale multiplex flow cytometry methodology for single‐NEV analysis (Nogueras‐Ortiz et al. 2024). NEVs were isolated from ∼50 mL of conditioned media and re‐suspended in 150–200 µL of PBS (1X). Half of this volume was adjusted to 250 µL and labelled with the total EV marker blue succinimidyl ester (BSE; CellTrace Blue, Thermo Fisher Scientific, #C34574) by incubating with an equal volume of 40‐µM BSE in PBS‐1X for 4 h at 37°C in the dark. Excess unbound dye was removed via three rounds of buffer exchange using 100‐kDa cut‐off spin filters (Amicon Ultra‐0.5, Millipore Sigma, #UFC510008) with PBS‐1X, resulting in a final sample volume of ∼200 µL. The BSE‐labelled NEV suspension was then split in half: one half was fixed by incubation with an equal volume of 16% paraformaldehyde (Electron Microscopy Sciences, #15710) for 5 min at room temperature, followed by four rounds of buffer exchange with PBS‐1X. The other half remained unfixed for comparative analysis. The final NEV volume was brought to 1600 µL in PBS‐1X and divided into 200‐µL aliquots for antibody labelling of ATP synthase. Each aliquot was incubated with the following antibodies at a final concentration of 0.2 ng/µL. The information for the primary and secondary antibodies is provided above (Antibodies and Chemicals). Primary antibody incubation was performed overnight at 4°C with gentle rotation in the dark after adding 10 µL of each antibody stock. For partial membrane permeabilization, PFA‐fixed NEVs were co‐incubated with 10 µL of a Tween‐20 stock solution to achieve a final concentration of 0.05%. Following primary labelling, 10 µL of each secondary antibody was added and incubated for 2 h at room temperature. Flow cytometry acquisition was carried out on a CytoFLEX LX instrument (Beckman Coulter) using acquisition parameters previously described.

2.10. ATP Synthase Assay

ATP synthase in NEVs and cultured hippocampal neurons was measured using a microplate assay (abcam). The assay was carried out according to the manufacturer's instructions with slight modifications. For purified NEV samples, after setting 5‐µL aliquots aside for NTA, the remaining NEVs were lysed with assay buffer containing 10% detergent (both of which were provided in the assay kit) on ice for 10 min before loading in the assay plate. For neuronal samples, neurons were homogenized in the assay buffer, the detergent was then added (10% final concentration) and incubated on ice for 10 min. Following centrifugation, supernatants were collected, protein concentration was measured using the BCA assay and lysate samples containing 20‐µg proteins were loaded into the assay plate. The plate was incubated at 4°C overnight, and activity and quantity of ATP synthase were measured. The ratio of the activity and quantity represents the specific activity of ATP synthase. All samples were run in duplicate.

2.11. Metabolic Labelling With AHA and Proximity Ligation Assay (PLA)

We detected newly synthesized proteins in hippocampal neurons using the FUNCAT (fluorescent noncanonical amino acid tagging) method as described (Dieterich et al. 2010; Tom Dieck et al. 2012). Following pharmacological treatments in Neurobasal medium with B27 for 18 h, neurons were incubated with methionine‐free NbActiv medium for 30 min. Neurons were then incubated with 4 mM AHA (azidohomoalanine) in methionine‐free NbActiv medium containing the original pharmacological treatments for 2 h at 37°C. Neurons were washed with PBS‐MC (PBS, pH 7.4, 1 mM MgCl₂, 0.1 mM CaCl₂), fixed with paraformaldehyde (4% paraformaldehyde and 4% sucrose in PBS‐MC) for 20 min, washed in PBS pH 7.4, permeabilized with 0.1% Triton X‐100 for 15 min and incubated in blocking solution (4% goat serum in PBS) for 1 h.

To identify incorporated AHA, a biotin tag was added by click chemistry (Dieterich et al. 2010; Tom Dieck et al. 2012). A click reaction mixture (in PBS, pH 7.8) composed of (1) 2 µM biotin alkyne tag, (2) 500 µM TCEP [Tris‐(2‐carboxyethyl)phosphine hydrochloride], (3) 200 µM TBTA [Tris(1‐benzyl‐1H‐1,2,3‐triazol‐4‐yl)methyl amine] and (4) 200 µM CuSO4 was applied to the neurons and incubated at RT overnight with gentle rocking and protected from light.

After washes with FUNCAT‐wash buffer (PBS pH 7.8, 0.1‐mM EDTA, 1% Tween‐20) and with PBS pH 7.4, the neurons were incubated with a pair of primary antibodies diluted in blocking solution at 4°C overnight. For detecting newly synthesized ATP synthase, mouse ATP synthase Ab (1:200) and rabbit biotin Ab (1:200) were used. For detecting newly synthesized PDH (pyruvate dehydrogenase [PDH]), mouse anti‐PDH Ab (1:200) and rabbit biotin Ab (1:200) were used. The coincidence detection of the antibody pairs (newly synthesized protein of interest) was visualized by using PLA as described (Tom Dieck et al. 2015) and per the manufacturer's instructions (Sigma). Briefly, neurons were incubated with respective secondary antibodies conjugated to different oligonucleotides (for example, rabbit PLA‐plus and mouse PLA‐minus), linker oligonucleotides were amplified and fluorescent probe was added. Neurons were post‐fixed for 5 min and labelled with guinea pig anti‐MAP2 Ab.

The control experiments included the following: AHA (4 mM) incubation in the presence of protein synthesis inhibitor Anisomycin (40 µM), replacing AHA with methionine (4 mM), or omitting biotin Ab.

We imaged neurons from each experiment within 24 h after the PLA procedure on a Zeiss LSM 980 confocal microscope. We selected neurons to image based on MAP2 and DAPI staining regardless of the intensity of the PLA signals (newly synthesized proteins). Images were acquired with a 40×/1.30‐numerical aperture oil objective lens in 8‐bit mode with 2048 × 2048 pixel xy resolution, and detector gain in each channel adjusted to cover the full dynamic range. Imaging conditions were kept the same within an experiment.

We used imageJ to quantify PLA signals in neurons. Images (tiff) exported from Zeiss Zen software were opened in Fiji and split into single‐channel images (MAP2, PLA and DAPI). To quantify PLA signals in soma, the soma area in the MAP2 image was manually outlined, and the area was saved as a region of interest (ROI). This ROI was then applied to the PLA image, and the integrated density of PLA signals within the ROI was measured. To quantify PLA signals in dendrites, single MAP2‐stained dendrites were manually traced starting from the soma, and then straightened with the Straighten Plugin in Fiji. The puncta of PLA signals along the dendrites were scored in 20‐µm segments.

2.12. Immunofluorescence Labelling of MDVs

For Antimycin A (AA) and Bafilomycin (BAF) experiments, neurons (6–8 DIV) were treated with AA (5 nM) or BAF (250 nM) or both for 5 h. For Shh and SAG experiments, neurons were treated with Shh (5%) or SAG (400 nM) for 18 h. After washing once in PBS (pH 7.4), neurons were fixed in pre‐warmed 4% paraformaldehyde containing 4% sucrose for 15 min. Fixed neurons were washed three times with PBS, permeabilized with 0.05% Triton X‐100 for 10 min and blocked with 10% BSA. Neurons were then incubated with the following primary antibodies in 3% BSA: mouse anti‐ATP synthase Ab (1:250), rabbit anti‐TOM20 Ab (1:250) and guinea pig anti‐MAP2 Ab (1:2500) overnight at 4°C. Following three washes in PBS, neurons were incubated with appropriate cross‐absorbed secondary antibodies (1:1000): goat anti‐mouse IgG Alexa Fluor 488, goat anti‐rabbit Alexa Fluor 568 and goat anti‐guinea pig Alexa Fluor 633 for 1 h. Labelled neurons were washed five times with PBS and mounted with mounting media containing DAPI.

2.13. Image Acquisition, Analysis and Presentation of Immunolabelled MDVs

Neurons co‐stained with ATP synthase, TOM20 and MAP2 were imaged on a Zeiss LSM 980 microscope with Airyscan (100×/1.46‐numerical aperture objective). Images were acquired in SR configuration at a 1224 × 1224‐pixel resolution in 8‐bit mode and the z‐stack was set to cover the entire volume of a neuron with optical slice thickness set to optimal (step size of 0.12 µm). Laser power and detector gain in each channel were adjusted to cover the full dynamic range but to avoid saturated pixels. The image acquisition settings were kept the same between different groups within an experiment. Maximum intensity projections were created in the Zen Blue software.

We analysed images using Fiji. We focused on examining mitochondria and MDVs in dendrites. ROIs—segments of MAP2‐labelled dendrites—were selected and straightened. For quantification of MDVs, images were split into single channels, and after thresholding, the number of ATP synthase‐MDVs or TOM20‐MDVs was counted using the Analyse Particles in Fiji with the particle size set at 5–100 pixel units. For mitochondrial size measurement, individual mitochondria on binary images were manually outlined, and the areas were measured. The number of ATP synthase‐MDVs or TOM20‐MDVs was normalized over the mitochondrial area (µm²) in the same image.

Same as for imaging methods described for other experiments, the brightness and contrast levels of all the images for MDVs were adjusted evenly across the image for each channel in Adobe Photoshop. No additional digital image processing was performed.

2.14. Statistical Analysis

Statistical analysis was performed using KaleidaGraph (5.0). Groups were compared using an unpaired two‐tailed Student's t test. p < 0.05 was considered statistically significant. The values represent the mean ± SEM. All experiments were performed three or more times. Sample number and number of replicates are stated in each figure legend.

3. Results

3.1. Characterization of Extracellular Vesicles Released From Primary Hippocampal Neurons

We isolated NEVs from the conditioned medium of cultured rat hippocampal neurons (Figure 1A) and characterized them using multiple methods. Nanoparticle tracking analysis showed a mixed population of smaller and larger NEVs (with ∼70% of particles at <200 nm, ∼30% at >200 nm) (Figure 1B). Electron microscopy revealed mostly single‐membranous NEVs, varying in size and containing small intraluminal vesicular structures and electron‐dense materials (Figure 1C). Immunofluorescence confocal microscopy with Airyscan showed NEVs co‐labelled with L1CAM and CD81, L1CAM and Alix (Figure 1D,E) and L1CAM and CD63 (Figure S1A). We previously demonstrated by immunoblots the presence of L1CAM, Alix and Flot‐1, and the absence of GM130 in NEV lysates (Nogueras‐Ortiz et a. 2024). In the present study, we expanded this evidence by conducting immunoblots of NEV lysates for VAMP2, a synaptic protein and established marker for NEVs (Vilcaes et al. 2021), finding it readily detectable (Figures 1F and S1B,C). However, the synaptic protein synaptophysin (SYP) and the mitochondrial protein TOM20 were not detected by immunoblots (Figure 1F), consistent with observations made by others (D'Acunzo et al. 2021; Vilcaes et al. 2021). Immunofluorescence microscopy showed NEVs co‐labelled with L1CAM and VAMP2 (Figure 1G). Finally, immunogold electron microscopy confirmed the presence of L1CAM and VAMP2 in NEVs (Figure 1H).

FIGURE 1.

Characterization of NEVs from cultured hippocampal neurons. (A) Immunofluorescence image of a hippocampal neuron co‐labelled with L1CAM (green) and Tuj1 (neuron‐specific class III β‐tubulin) (red). Scale bar, 20 µm. (B) Nanoparticle tracking analysis showing size distribution of NEVs. (C) Electron micrographs showing size and ultrastructure of NEVs. Scale bars, 200 nm. (D, E) Immunofluorescence Airyscan images showing examples of NEVs co‐labelled with L1CAM (green) and CD81 (red) or Alix (red). Scale bars, 200 nm. (F) Immunoblots of NEVs showing readily detectable VAMP2 (a synaptic protein and known NEV marker) in NEV lysates (two different NEV samples). Note that synaptophysin (SYP) and TOM20 are undetectable in NEV lysates by immunoblotting. Lysates of cultured hippocampal neurons and glia were used as references. Uncropped blots are shown in Figure S1B. (G) Immunofluorescence Airyscan images showing NEVs co‐labelled with L1CAM (green) and VAMP2 (red). Scale bars, 200 nm. (H) Electron micrographs of immunogold showing examples of NEVs labelled with L1CAM (10‐nm gold) or VAMP2 (20‐nm gold). Scale bars, 200 nm.

3.2. Determination of ATP Synthase in NEVs

A variety of mitochondrial constituents have been found in EVs and NEVs (Picca et al. 2020; Crewe et al. 2021; D'Acunzo et al. 2021; Peruzzotti‐Jametti et al. 2021; Rabas et al. 2021; Todkar et al. 2021; Vasam et al. 2021; Vilcaes et al. 2021; Yao et al. 2021; Ladakis et al. 2022, 2024). We focused on assessing mitochondrial oxidative phosphorylation proteins in NEVs. We analysed NEV lysates using a cocktail of five antibodies that are specifically against mitochondrial complexes I–IV and ATP synthase (Materials and Methods). In cellular lysates of neurons, as expected, the antibody cocktail detected all four complexes and ATP synthase (Figures 2A and S2A). In lysates of NEVs, the protein band corresponding to ATP synthase was clearly visible (Figures 2A and S2A), whereas the other four complexes were not evident even in blots with prolonged exposure (Figure S2A).

FIGURE 2.

ATP synthase in NEVs. (A) Immunoblots showing detectable ATP synthase in NEV lysates. Lysates of neurons were used as references. An antibody cocktail against the five OXPHOS complexes was used, in which the ATP synthase antibody is specific for its α subunit (Materials and Methods). Arrows point at the ATP synthase α subunit, and complex II (cIII), complex II (cII) and complex I (cI). (B) Immunoblots showing ATP synthase β subunit in NEV lysates. Lysates of neurons and heart mitochondria were used as references. The ATP synthase antibody is specific for the β subunit (Materials and Methods). Arrow points at the ATP synthase β subunit. Uncropped blots are shown in Figure S2B. (C) Electron micrograph of immunogold showing an example of NEV labelled with ATP synthase (β subunit). (D, E) Immunofluorescence Airyscan images showing NEVs co‐labelled with ATP synthase (green) with either VAMP2 (red in D) or L1CAM (red in E). (F) Immunofluorescence Airyscan images showing two examples of NEVs co‐labelled with ATP synthase subunits α and β. (G) PLA (proximity ligation assay) analysis confirming co‐existence of ATP synthase α and β subunits (green) on NEVs (VAMP2, red). Scale bars, 200 nm (C–G). (H) Dot plot shows the fluorescent signal of ATP synthase α versus β subunit in BSE‐positive events with a quadrant gate enclosing NEVs not carrying ATP synthase (blue events), single‐positive for either ATP synthase α or β subunit (yellow and red events, respectively), and double‐positive for both subunits (green events). (I) A pie chart depicts the average proportion (±SEM) of NEVs harbouring ATP synthase that are single‐positive for either the α or β subunit, or double‐positive for both subunits. A total of two biological replicates were analysed across two independent experiments. (J) Bar graph displaying the fold change in the percentage of ATP synthase‐positive NEVs following partial membrane permeabilization relative to intact NEVs. A total of three biological replicates were analysed across three independent experiments (p** = 0.0079; unpaired t‐test). Positive FCA events were defined as those with vSSC versus fluorescent signals distinguishable from negative controls (Figure S2F).

We investigated ATP synthase in NEVs further. The ATP synthase antibody from the antibody cocktail is specific for the α subunit of ATP synthase. We repeated immunoblots using a β subunit‐specific antibody. Along with VAMP2 in the same NEV lysate sample, ATP synthase β subunit was readily detectable (Figures 2B and S2B). Immunogold electron microscopy provided the direct visual evidence of ATP synthase (β subunit) residing in NEVs (Figure 2C). Immunofluorescence microscopic analysis demonstrated co‐staining of ATP synthase with the NEV‐specific marker VAMP2 (Figure 2D) and L1CAM (Figure 2E). Furthermore, co‐immunolabelling of ATP synthase α‐ and β‐subunit (Figure 2F), an analysis using proximity ligation assay (PLA; Fredriksson et al. 2002; Leuchowius et al. 2011) (Figures 2G and S2C), and flow cytometry analysis (FCA; Figure 2H,I) collectively confirmed the coexistence of these two ATP synthase subunits in single NEVs. Finally, FCA showed detectable ATP synthase labelling on NEVs only after permeablization (Figure 2J), suggesting that ATP synthase molecules likely reside within rather than on the surface of NEVs.

From the above experimental approaches, we concluded that ATP synthase is a true NEV cargo and proceeded to a quantitative assessment of ATP synthase in NEVs. We used a microplate assay, which measures ATP synthase activity and quantity in tandem (Materials and Methods). Given small amounts of proteins in NEVs (as compared to cellular proteins), we evaluated the sensitivity of this assay. The assay reliably measured the level (quantity) of ATP synthase using as low as 1 µg of total protein. To measure the activity of ATP synthase, however, the assay required at least five‐fold greater amount of total proteins (>5 µg) (Figure S2D,E). Our NEV preparations typically yielded 2–5 µg of proteins (per sample). Despite attempts to scale up NEV preparations, we found that measurements of ATP synthase activity in NEV samples were inconsistent and unreliable, even with greater amounts of protein. (A plausible explanation for this finding is that the g‐force generated from ultracentrifugation during NEVs isolation may disrupt ATP synthase structure, thus impacting its activity.) Therefore, in subsequent experiments, we measured both ATP synthase activity and quantity in neuronal samples, but only its quantity in NEV samples.

3.3. ATP Synthase in NEVs Reflects ATP Synthase in the Neuron

We previously found that ATP synthase in NEVs is altered in patients with Alzheimer's disease (Yao et al. 2021) and Fragile X‐associated tremor/ataxia syndrome (Yao et al. 2024). To determine whether NEV‐borne ATP synthase quantity reflects ATP synthase quantity and activity in the parent neurons, we investigated ATP synthase in primary hippocampal neurons and the EVs they release. We measured ATP synthase activity and quantity in neurons, and its quantity in NEVs (Figure 3A).

FIGURE 3.

ATP synthase in NEVs and neurons. (A) Scheme depicting the experimental setup for analysis of ATP synthase in cultured hippocampal neurons and their EVs. The duration of all treatments was 18 h. (B) Comparison of ATP synthase in NEVs and neurons after treatment with Shh (Sonic hedgehog; 5%) or SAG (Shh agonist; 400 nM). Data from the treated group were normalized to the untreated control within each experiment. Each symbol represents one experiment (n = 4 independent experiments). The activation of the Shh pathway in the cultured hippocampal neurons in response to exogenously applied Shh is shown in Figure S3A. (C) Comparison of ATP synthase in NEVs and neurons treated with HN (Humanin; 1 µM). Data were normalized to control within each experiment. Each symbol represents one experiment (n = 4 independent experiments). (D) Comparison of ATP synthase in NEVs and neurons treated with Bic (Bicuculine; 40 µM). Each symbol represents one experiment (n = 6 independent experiments). The presence of GABAergic neurons in the neuronal culture and their response to Bic are shown in Figure S3D,E. (E) Comparison of ATP synthase in NEVs and neurons treated with Glut (Glutamate; 50 µM). Each symbol represents one experiment (n = 6 independent experiments). Data are mean ± SEM. Unpaired two‐tailed t‐test was used to calculate significance. *p < 0.05; **p < 0.01; ns, not significant.

Upregulation of Sonic hedgehog (Shh) signalling pathway activity in neurons by exposing them to Shh leads to an increase in mitochondrial respiration (Yao et al. 2017; Chung et al. 2022). We treated neurons with Shh (5%) or a Shh agonist SAG (Shh agonist; 400 nM) for 18 h (Figures 3A and S3A,B) and measured ATP synthase in neurons and in NEVs purified from the conditioned medium. In NEVs, Shh and SAG increased ATP synthase quantity (Figure 3B; Shh vs. control, p = 0.043; SAG vs. control: p = 0.047; n = 4 independent experiments). It is unlikely that this increased ATP synthase in NEVs is due to an increase in NEVs, as their numbers were similar between the treated and untreated control (Figure S3C). In neurons, Shh and SAG increased ATP synthase activity (Figure 3B; Shh vs. control, p = 0.029; SAG vs. control, p = 0.049, n = 4); however, the quantity of ATP synthase was unchanged (Figure 3B; Shh vs. control, p = 0.47; SAG vs. control, p = 0.18), which is in contrast to the increased ATP synthase quantity observed for NEVs.

We next tested the effect of humanin, a small mitochondrial peptide known for its ability to protect neurons (Hashimoto et al. 2001; Gong et al. 2018; Jung et al. 2020), possibly by stimulating mitochondrial functions (Jung et al. 2020; Liang et al. 2022). We incubated hippocampal neurons with humanin (1 µM) for 18 h and measured ATP synthase. In NEVs, we detected a slight yet consistent increase in ATP synthase quantity with humanin (Figure 3C ; p = 0.046, n = 4), without a detectable change in the number of NEVs (Figure S3C). In neurons, we detected increased ATP synthase activity in response to humanin (Figure 3C; p = 0.005, n = 4), whereas ATP synthase quantity was not changed (Figure 3C; p = 0.73).

We also assessed whether neuronal activity induces any changes in ATP synthase in neurons and their EVs. Cultured hippocampal neurons mimic many physiological aspects of neurons in vivo. The hippocampal neuronal cultures contain a mixture of neuronal subtypes including γ‐aminobutyric acid (GABA)–releasing (GABAergic) neurons (Figure S3D). We used the GABA_A receptor antagonist Bicuculline (Bic) to enhance neuronal activity in cultured cells (Turrgiano et al. 1998; Rivell et al. 2019; Dorrbaum et al. 2020). Exposing to Bic (40 µM, 18 h) robustly induced expression of cFos in neurons (Figure S3E), indicating enhanced neuronal activity (Bullitt 1990). In NEVs, ATP synthase quantity was lower in Bic‐treated neurons (Figure 3D; p = 0.0037, n = 6 experiments), while the total number of NEVs was unaffected (Figure S3F). In neurons, ATP synthase activity was also significantly lower in response to Bic (Figure 3D; p = 0.014, n = 6), a pattern consistent with the reduced oxygen consumption rate of these neurons (Figure S3G). In contrast, ATP synthase quantity in neurons was not significantly different between Bic‐treated and untreated control cultures (Figure 3D; p = 0.106).

To assess ATP synthase in neurons under stress, we exposed neurons to Glutamate (Glut; 50 µM for 18 h), an experimental paradigm known to increase neuronal activity to the point of inducing excitotoxicity (Mattson et al. 1989) and impairing mitochondrial respiration in cultured neurons (Yao et al. 2017; Figure S3H). NEVs’ ATP synthase quantity varied widely among six independent experiments, although their average was not different from that of controls (Figure 3E; p = 0.82, n = 6 experiments). In neurons, ATP synthase activity was reduced significantly with Glut treatment (Figure 3E; p = 0.011, n = 6), while the overall ATP synthase quantity remained unchanged (Figure 3E; p = 0.199).

Taken together, neuronal ATP synthase activity dynamically responds to stimuli: going up in neurons treated with mitochondria‐stimulating factors Shh and humanin, but going down in neurons upon Bic‐enhanced or Glut‐elicited neuronal activity. Except for the Glut‐treated neurons, which may have experienced a combination of increased neuronal activity and excitotoxicity, the quantity of ATP synthase in NEVs seems to mirror changes in ATP synthase activity in the neuron.

3.4. Newly Synthesized ATP Synthase in Hippocampal Neurons

Our data showed that the abundance of ATP synthase in neurons stays constant regardless of activity level. We hypothesized that elevated neuronal ATP synthase activity may trigger its turnover, with upregulated protein synthesis of ‘new’ ATP synthase accompanying the removal of ‘old’ ATP synthase (at least partially) through NEVs; as a result, the overall abundance of ATP synthase is maintained. We thus investigated the synthesis rate of ATP synthase in neurons. Considering our previous observation that the mRNAs encoded for multiple subunits of ATP synthase were elevated (≥2 folds) in neurons upon exposure to Shh (Yao et al. 2017), we focused our analyses on the effect of Shh on new ATP synthase protein synthesis. We treated neurons with Shh or SAG and assessed nascent ATP synthase using a method that couples fluorescent noncanonical amino acid tagging (FUNCAT) with PLA (Dieterich et al. 2010; Tom Dieck et al. 2015). In FUNCAT, azidohomoalanine (AHA) enters cells and incorporates into nascent protein during translation; next, with click chemistry, a biotin‐based tag is added to AHA. The subsequent PLA uses an antibody to tag biotin and another antibody to tag the protein of interest, enabling the visualization of specific newly synthesized proteins within cells (Tom Dieck et al. 2015). We used FUNCAT in combination with PLA, implementing a biotin antibody and an ATP synthase antibody to detect newly synthesized ATP synthase in neurons (Figures 4A and S4A,B). In the untreated control neurons, fluorescence signals reflecting nascent ATP synthase were readily visualized in the soma as well as proximal segments of dendrites (Figure 4B,C). In Shh‐ and SAG‐treated neurons, fluorescence signal intensity was noticeably higher and more spread out along proximal dendrites (Figure 4B,C). Quantification of the fluorescence signal intensity revealed significantly higher nascent ATP synthase in the soma and the initial 100 µm of proximal dendrites of Shh‐ and SAG‐treated neurons compared to controls (Figure 4D,E). We also tested the effect of proteasome inhibitor MG132. Notably, co‐treatment of Shh or SAG with MG132 (10 µM) did not change Shh‐ or SAG‐elicited synthesis of ATP synthase (Figures 4D,E and S4C), supporting the notion that higher levels of new ATP synthase likely result from increased protein synthesis rate rather than reduced protein degradation. We also analysed the mitochondrial matrix protein PDH and its synthesis rate. While PDH protein was present in mitochondria throughout neurons (Figure S4A), the pool of newly synthesized PDH was primarily detected in the soma of neurons (Figure 4F). Noticeably, the level of new PDH was not different between the untreated control and Shh‐treated neurons (Figure 4G).

FIGURE 4.

Newly synthesized ATP synthase in hippocampal neurons. (A) Experimental workflow for visualizing newly synthesized proteins in cultured neurons. (B) Representative fluorescence images showing newly synthesized ATP synthase (green) in neurons (red, MAP2) untreated (Ctr) and treated with Shh or SAG. (C) Representative straightened dendrites (red) with newly synthesized ATP synthase (green). Fluorescence images of dendrites were traced, straightened and aligned to produce the image montages. (D) Comparison of the mean signal intensity of newly synthesized ATP in the soma of untreated neurons and neurons treated with Shh or SAG, or with Shh or SAG in the presence of proteasome inhibitor MG132 (10 µM). The intensity values are normalized to the soma size (µm²) of the corresponding neurons. Each symbol represents an individual neuron. A total of 176 neurons from five experiments were analysed. (E) Comparison of newly synthesized ATP (green, puncta number) in dendrites in 20‐µm bins. A total of 1225 dendritic segments from five experiments were analysed. Note, fluorescence images of Shh/MG132‐ or SAG/MG132‐treated neurons and their dendrites are shown in Figure S4C. (F) Representative fluorescence images showing newly synthesized PDH (green) in neurons (red, MAP2) untreated or treated with Shh. A total of 57 neurons from three experiments were analysed. (G) Quantification of the mean signal intensity of newly synthesized PDH in the soma of neurons. The intensity values are normalized to the soma size (µm²) of the corresponding neurons. A total of 58 neurons from three independent experiments were analysed. Data are mean ± SEM. Unpaired two‐tailed t‐test was used to calculate significance. *p < 0.05, **p < 0.01; ***p < 0.001; ns, not significant. Scale bars, 10 µm (B, C, F).

3.5. ATP Synthase–Containing Mitochondrial Vesicles in Hippocampal Neurons

Having established that NEV‐borne ATP synthase reflects ATP synthase in neurons, we next set out to identify intracellular sources or carriers of mitochondrial ATP synthase. We focused on mitochondria‐derived vesicles (MDVs)—small membranous structures (60–150 nm) directly released from mitochondria (Neuspiel et al. 2008; Yao et al. 2020; Konig and McBride 2024). There are distinct subsets of MDVs that contain specific cargoes ranging from outer or inner mitochondrial membrane constituents to mitochondrial matrix contents (Konig and McBride 2024 and references within). We examined ATP synthase in MDVs of hippocampal neurons. We co‐immunolabelled neurons with the neuronal marker MAP2, the known MDV cargo protein TOM20 (Neuspiel et al. 2008; Yao et al. 2020) and ATP synthase (Figure 5A). Densely packed mitochondria in neuronal somas prevented unambiguous identification of MDVs due to poor separation of small dim MDV puncta in the vicinity of bright fluorescently labelled mitochondria (Figure 5A); therefore, we focused on analysing the dendrites where individual mitochondria were easily visualized and analysed (Figure 5B). We saw TOM20‐labelled MDVs protruding from or adjacent to mitochondria (Figure 5B). We also observed ATP synthase–labelled‐MDVs that were similar in size and shape to TOM20‐MDVs, although a majority of ATP synthase‐MDVs did not have TOM20 labelling (Figure 5B). Among 1458 mitochondria analysed (through three separate experiments), we observed 3285 MDVs containing TOM20 only, 1016 MDVs containing ATP synthase only and just 25 MDVs containing both TOM20 and ATP synthase. Given its paucity, the subset of MDVs that contained both TOM20 and ATP synthase was not considered further in this study.

FIGURE 5.

ATP synthase in MDVs. (A) Schematic of primary hippocampal neurons co‐immunolabelled for ATP synthase (green), TOM20 (red) and dendritic marker MAP2 (blue). Note, crowded mitochondria in the cell body of the neuron. Scale bar, 10 µm. (B) An example of Airyscan fluorescence image showing ATP synthase‐(green) and TOM20‐(red) containing MDVs in a segment of straighten dendrite (blue). Enlarged views of Boxes 1–3 illustrating ATP synthase‐MDVs (green arrowheads) and TOM20‐MDVs (red arrowheads) distinct from each other. Intensity distributions along scan lines (white dashed) are plotted below. Note for mitochondria, spatial overlap between green and red profiles as expected with inner membrane ATP synthase and outer membrane TOM20. For MDVs, most ATP synthase‐MDV and TOM20‐MDV profiles do not overlap except few cases (*). Scale bars, 1 µm for images; 200 nm for profiles. (C) Examples of Airyscan fluorescence images showing noticeably greater number of TOM20‐MDVs (red arrowheads) when neurons were treated with a low dose of Antimycin A (AA; 5 nM) or Bafilomycin A (BAF; 250 nM) or both (AA/BAF). Images are representative of 3–5 experiments examined for each condition. Scale bars, 1 µm. (D) Quantification of ATP synthase‐MDVs and TOM20‐MDVs. Each symbol represents data from one dendritic segment, similar to the top panel in (B). A total of 356 dendritic segments encompassing 2706 mitochondria from 3 to 5 independent experiments were examined (n = 3 experiments for ATP synthase‐MDVs, n = 5 experiments for TOM20‐MDVs). (E) Quantification of ATP synthase‐MDVs and TOM20‐MDVs in neurons treated with Shh or SAG. A total of 110 dendritic segments encompassing 1063 mitochondria from three independent experiments were examined. Examples of Airyscan fluorescence images are shown in Figure S5. For both (D) and (E), MDV counts were normalized to the total size (µm²) of mitochondria in the same dendritic segment. Data are mean ± SEM. Unpaired two‐tailed t‐test was used to calculate significance. **p < 0.01; ***p < 0.001; ns, not significant.

Subjecting neurons to mild stress by exposing them to a low dose of Antimycin A (AA; 5 nM for 5 h) triggers mitochondria to produce more TOM20‐containing MDVs (Lin et al. 2017; Yao et al. 2020). We tested whether low‐dose AA treatment also induces the formation of ATP synthase–containing MDVs in neurons. As expected, the number of TOM20‐MDVs increased from 2.2 ± 0.22 before AA to 3.5 ± 0.31 after AA treatment (Figure 5C; 5D right; p = 0.001). However, AA did not change the number of ATP synthase–containing MDVs (Figure 5C,D left; 1.6 ± 0.29 before AA vs. 1.9 ± 0.55 after AA, p = 0.68). Given that MDVs deliver some of their cargoes, including TOM20, to lysosomes (Sugiura et al. 2014; Konig and McBride 2024), we compared the number of TOM20‐MDVs and ATP synthase‐MDVs in neurons following inhibition of lysosomal acidification with Bafilomycin (BAF). The number of TOM20‐MDV was substantially higher in neurons treated with BAF than the untreated control, and even higher with BAF plus AA (Figure 5C; 5D right; for both BAF and BAF/AA, p < 0.001), suggesting their lysosomal destination. In contrast, the numbers of ATP synthase‐MDVs were not different between untreated neurons and treated with either BAF or BAF/AA, arguing against a similar lysosomal destination (Figure 5C,D left; p = 0.06 for BAF, p = 0.14 for BAF/AA).

We asked whether Shh signalling activity affects the production of these two populations of MDVs. Notably, the number of TOM20‐MDVs was not different between the untreated control and Shh or SAG‐treated neurons (Figure 5E right and Figure S5; 2.5 ± 0.23 for untreated, 2.3 ± 0.31 for Shh‐treated, p = 0.57, 2.3 ± 0.28 for SAG‐treated, p = 0.54). However, the number of ATP synthase‐MDVs increased substantially (Figure 5E left and Figure S5; 1.2 ± 0.21 for untreated, 2.4 ± 0.31 for Shh‐treated, p = 0.0012, 2.6 ± 0.31 for SAG‐treated, p = 0.0002).

Thus, our data suggest that the production of TOM20‐MDVs and ATP synthase‐MDVs from mitochondria is governed by different cellular signals and that these two sub‐populations of MDVs seem to follow different pathways and have distinct fates. The putative mechanistic connections between intracellular ATP synthase–containing MDVs and extracellular ATP synthase–containing NEVs will be an important topic for future studies.

4. Discussion

This study confirmed that mitochondrial ATP synthase is a cargo of NEVs. Experimental manipulations that either increased or decreased ATP synthase activity in neurons led to corresponding changes in the amount of ATP synthase in released NEVs, the prevalence of ATP synthase–containing MDVs intracellularly, as well as the production of nascent ATP synthase. These findings suggest a novel mechanism that neurons use to maintain the quantity, and possibly the quality, of ATP synthase in mitochondria.

The experimental design of this study was to manipulate neuronal mitochondria and analyse ATP synthase in neurons and NEVs. Initially, we attempted to challenge neuronal mitochondria using classic mitochondrial inhibitors, such as Oligomycin; however, we found that the viability of the primary hippocampal neurons was severely and irreversibly compromised by Oligomycin despite low doses and short duration of exposure (data not shown), preventing any further experiments. We, therefore, chose to use Shh‐treated neurons as an experimental paradigm, given the ability of Shh to stimulate mitochondrial activity (Yao et al. 2017; Chung et al. 2022) while supporting and protecting neurons (Bambakidis et al. 2012; Huang et al. 2013; Chechneva and Chen 2015; Yao et al. 2015, Yao et al. 2017).

Because our focus was on the neuron and because neuronal activity is closely linked to mitochondrial function (Bindokas et al. 1998; Kann and Kovacs 2007), we asked how neuronal activity may change mitochondrial ATP synthase in neurons and whether their released NEVs may reflect any cellular changes. Bic‐enhanced neural activity (Turrgiano et al. 1998; Rivell et al. 2019; Dorrbaum et al. 2020) consistently—though somewhat unexpectedly—decreased mitochondrial respiration (Figure S3G), along with reduced neuronal ATP synthase activity in cells and lower ATP synthase in NEVs (Figure 3D). Using Glut to stimulate neurons and induce excitotoxicity (Mattson et al. 1989), we observed that neurons also exhibited decreased ATP synthase activity (Figure 3H) and lower mitochondrial respiration (Figure S3H). Unlike Bic‐treated neurons, however, NEV‐associated ATP synthase was not demonstrably altered by Glut due to high variance in its effects. This variation could be due to NEVs from Glut‐treated neurons being contaminated by mitochondrial components‐containing cellular debris, as a result of Glut‐elicited reduction in cellular viability (Yao et al. 2017).

Recent investigations using a variety of methodologies have found that mitochondrial proteins are significantly longer‐lived than the average lifetime of most cellular proteins (Dorrbaum et al. 2018; Fornasiero et al. 2018; Bomba‐Warczak et al. 2021; Krishna et al. 2021). Among the long‐lived mitochondrial proteins are the proteins involved in mitochondrial oxidative phosphorylation—the electron transport chain complexes and the ATP synthase (Dorrbaum et al. 2018; Fornasiero et al. 2018). The long lifespan of these mitochondrial proteins poses a challenge owing to their inevitable exposure to high levels of ROS—the incidental byproducts of oxidative phosphorylation (Adam‐Vizi and Chinopoulos 2006; Murphy 2009; Willems et al. 2015). Accordingly, cells—particularly neurons—need to possess and deploy mechanisms to renew ATP synthase by replacing ROS‐damaged proteins. The positive correlation between neuronal ATP synthase activity level and NEV‐carried ATP synthase may reflect ROS‐damaged ATP synthase being expelled from neurons by way of NEVs, although this possibility would need to be demonstrated directly. Future investigations addressing whether NEVs‐carried ATP synthase is in fact oxidatively modified (Ebanks and Chakrabarti 2022) will be necessary. In the meantime, we speculate that oxidative stress by ROS inflicts damage on the ATP synthase that simultaneously triggers the removal of impaired proteins via NEVs and the synthesis of new proteins; in conjunction, these two mechanisms keep a constant pool of functional ATP synthase to support neuronal activity.

The mitochondrial content–containing EVs dubbed as ‘mitovesicles’ (D'Acunzo et al. 2021) are different from MDVs—the latter are intracellular vesicles released directly from mitochondria (Neuspiel et al. 2008; Yao et al. 2020; Konig and McBride 2024). While there still are gaps in our knowledge about how these two types of vesicles are related, the overlapping proteomes—albeit partially—of MDVs and mitovesicles (Vasam et al. 2021) imply that they may be linked. Several important conclusions regarding MDVs may be drawn from this study. First, ATP synthase is a cargo of MDVs, consistent with the findings of others (Roberts et al. 2021; Vasam et al. 2021; Hazan Ben‐Menachem et al. 2023). Second, ATP synthase and TOM20 inhabit separate sub‐populations of MDVs. Third, unlike TOM20‐containing MDVs, AA (Antimycin A)‐induced stress does not affect the abundance of ATP synthase‐MDVs. Robert et al. (2021) observed that AA did not enrich ATP synthase‐MDVs in the brain, whereas Vasam et al. (2021) reported AA influencing ATP synthase‐MDVs in heart cells. These seemingly opposite findings could be due to differences in cell types. Fourth, also unlike TOM20‐MDVs, ATP synthase‐MDVs are not destined to lysosomes for degradation, hinting that ATP synthase–containing MDVs could instead travel to multivesicular bodies, where they may subsequently transition into ATP synthase–containing NEVs (or have their cargo recycled into NEVs).

NEVs may remove ATP synthase damaged by oxidate stress from neurons. Alternatively, NEVs could contain functionally competent ATP synthase. Using cell‐free MDVs isolated from bovine heart, Soubannier et al. (2012) found MDVs harbouring oxidized mitochondrial protein cargos, but ATP synthase was not among them. Furthermore, a recent study demonstrated that MDVs from Saccharomyces cerevisiae contain functional ATP synthase (Hazan Ben‐Menachem et al. 2023). Likewise, EVs have been found to contain respiration‐competent mitochondria (Crewe et al. 2021; Peruzzotti‐Jametti et al. 2021). Therefore, if NEV‐borne ATP synthase can generate ATP, these vesicular carriers may deliver small ‘energy factories’ between cells of the brain. We are just beginning to understand the functions of NEVs in the brain beyond their utility as a source of biomarkers. This study helps bridge this knowledge gap and opens the way to studies investigating the functional significance of MDV and NEV release vis‐a‐vis neuronal energetics and function, for emitting as well as, potentially, receiving organelles and cells.

4.1. Limitations of the Study

The aim of this study has been to understand how the NEV‐borne mitochondrial ATP synthase relates to ATP synthase in the parent mitochondria of NEV‐originating neurons. Although we demonstrated the presence of ATP synthase in a population of NEVs, the identity of the ATP synthase–carrying sub‐population of NEVs (i.e., whether these are exosomes, microvesicles or both) remains to be identified. It would also be of great interest to characterize the functional state of ATP synthase carried by NEVs, something that we attempted but were unable to determine in this study. Moreover, understanding the biogenesis and intracellular trajectory of ATP synthase–containing NEVs, their link to MDVs, as well as their functional impacts on neighbouring cells extracellularly, are research questions warranting further investigations.

Author Contributions

Pamela J. Yao: conceptualization, investigation, writing–original draft, methodology, validation, visualization, writing–review and editing, formal analysis, project administration, data curation, supervision. Carlos Nogueras‐ortiz: investigation, writing–review and editing, methodology, data curation, formal analysis, validation. Krishna Ananthu Pucha: methodology, data curation, writing–review and editing. Dimitrios Kapogiannis: funding acquisition, writing–review and editing, resources, supervision, formal analysis, project administration.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Material: jev270140‐sup‐0001‐SuppMat.docx

Data Availability Statement

All data needed to evaluate the conclusions in the paper are present in the paper and the Supporting Materials.