SCIENCE ADVANCES | RESEARCH ARTICLE DISEASES AND DISORDERS Copyright © 2021 The Authors, some Mitovesicles are a novel population of rights reserved; exclusive licensee extracellular vesicles of mitochondrial origin American Association for the Advancement altered in Down syndrome of Science. No claim to original U.S. Government Works. Distributed Pasquale D'Acunzo1,2, Rocío Pérez-González1,2*, Yohan Kim1,2, Tal Hargash1, Chelsea Miller1, under a Creative Melissa J. Alldred1,2, Hediye Erdjument-Bromage3,4, Sai C. Penikalapati1, Monika Pawlik1, Commons Attribution Mitsuo Saito2,5, Mariko Saito2,6, Stephen D. Ginsberg1,2,7,8, Thomas A. Neubert3,4, NonCommercial Chris N. Goulbourne1, Efrat Levy1,2,7,9† License 4.0 (CC BY-NC). Mitochondrial dysfunction is an established hallmark of aging and neurodegenerative disorders such as Down syndrome (DS) and Alzheimer’s disease (AD). Using a high-resolution density gradient separation of extracellular vesicles (EVs) isolated from murine and human DS and diploid control brains, we identify and characterize a pre- viously unknown population of double-membraned EVs containing multiple mitochondrial proteins distinct from previously described EV subtypes, including microvesicles and exosomes. We term these newly identified Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 mitochondria-derived EVs “mitovesicles.” We demonstrate that brain-derived mitovesicles contain a specific sub- set of mitochondrial constituents and that their levels and cargo are altered during pathophysiological processes where mitochondrial dysfunction occurs, including in DS. The development of a method for the selective isolation of mitovesicles paves the way for the characterization in vivo of biological processes connecting EV biology and mitochondria dynamics and for innovative therapeutic and diagnostic strategies. INTRODUCTION dictory reports were published on the nature of secreted mitochondrial Extracellular vesicles (EVs) are heterogeneous nanoscale vesicles se- material, ranging from a proposal that whole mitochondria, mitochondria-­ creted into the extracellular space by all cell types studied, including encapsulated microvesicles, or exosomes packed with mitochondrial brain cells (1). Microvesicles and exosomes are the main EV subtypes cargo can be secreted from cells in vitro (3–7). Notably, these studies identified to date. Microvesicles are 100- to 1000-nm EVs that bud relied on differential centrifugation-based methods to isolate EVs that from the plasma membrane and contain specific proteins, such as failed to differentiate among EV subpopulations and between vesicular annexin A1 and annexin A2 (2). Exosomes are 50- to 150-nm EVs and nonvesicular components (2). Thus, mechanisms underlying how that originate from the endocytic pathway upon fusion of the late mitochondrial constituents are secreted and their biological functions endosomes/multivesicular bodies (MVBs) with the plasma membrane remain unknown. Furthermore, in vivo release of mitochondrial and are enriched in endosomal sorting complex required for transport components into brain extracellular space has not been described. (ESCRT) proteins, such as the ALG-2-interacting protein X (Alix) and We previously developed a protocol to isolate small EVs from the the tumor susceptibility gene 101 protein (TSG101), and in tetra- brain, using a sucrose density gradient (10, 11). With this approach, spanins, such as CD63 and CD81 (2). we demonstrated that exosomes are key players in both Alzheimer’s Previous in vitro studies demonstrated that multiple cell types se- disease (AD) and Down syndrome (DS) pathogenesis, exerting com- crete mitochondrial components into the culture media (3–7). It was plex pleiotropic roles as neuroprotective and neuropathogenic agents suggested that entire mitochondria are transferred between cells and (1, 12–16). Here, we apply a novel high-resolution density gradient have a neuroprotective function (8). It was also demonstrated that on isolated brain EVs. We demonstrate separation of various brain mitochondrial stress stimulates the release of molecules, including EV subtypes and the existence of a previously unknown population mitochondrial DNA (mtDNA), and damage-associated molecular pat- of mitochondria-derived EVs. These EVs, which we term mitovesi- terns (DAMPs) that have strong proinflammatory activity (9). Contra- cles, differ morphologically, by size and content, from intracellular mitochondria and from canonical EV subtypes. We also show that 1 the levels and cargo of mitovesicles are altered in DS brains, which Center for Dementia Research, Nathan S. Kline Institute for Psychiatric Research, Orangeburg, NY 10962, USA. 2Department of Psychiatry, New York University School have been characterized by the concomitant presence of mitochon- of Medicine, New York, NY 10016, USA. 3Department of Cell Biology, New York Uni- drial and endosomal abnormalities (17). These data demonstrate that versity School of Medicine, New York, NY 10016, USA. 4Kimmel Center for Biology brain cells constitutively secrete mitochondrial vesicles and that this and Medicine at the Skirball Institute, New York University School of Medicine, pathway is altered under conditions in which mitochondrial dys- New York, NY 10016, USA. 5Division of Analytical Psychopharmacology, Nathan S. Kline Institute for Psychiatric Research, Orangeburg, NY 10962, USA. 6Division of function occurs, providing new insights into heterogeneous EV biol- Neurochemistry, Nathan S. Kline Institute for Psychiatric Research, Orangeburg, NY ogy and its relationship with mitochondrial dynamics. 10962, USA. 7NYU Neuroscience Institute, New York University School of Medicine, New York, NY 10016, USA. 8Department of Neuroscience and Physiology, New York University School of Medicine, New York, NY 10016, USA. 9Department of Biochemistry & Molecular Pharmacology, New York University School of Medicine, New York, NY RESULTS 10016, USA. Isolation of distinct subpopulations of brain EVs by a *Present address: Center for Networked Biomedical Research on Neurodegenerative high-resolution density gradient Diseases (CIBERNED), Department of Neurology, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain. To separate brain EV subpopulations and determine which EV sub- †Corresponding author. Email: efrat.levy@nki.rfmh.org types contain mitochondrial proteins, we modified our original method D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 1 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE of isolation and fractionated brain EVs with an OptiPrep-based step EVs found in LDFs were larger by cryoEM and TEM than other gradient density column. Briefly, murine right hemibrains were dis- EV subtypes (Fig. 2, B and C), consistent with NTA results. A small sected, digested in papain to loosen the extracellular matrix, filtrated, size difference between IDFs and Fr8 was also observed by NTA, and differentially centrifuged to eliminate cells, undigested materi- consistent with a similar trend observed by cryoEM and TEM (Fig. 2, als, and cellular debris. A crude EV pellet was subsequently washed B and C). In summary, our data show that the OptiPrep density and further purified through a high-resolution step gradient (table gradient identifies three brain EV subtypes based on total protein S1). Eight distinct fractions were obtained using this approach content, morphology, electron density, and size. (Fig. 1A). Brain EV fractions were initially characterized in wild- type mice. Nanoparticle tracking analysis (NTA) demonstrated that Protein analysis of EV components in LDFs, IDFs, and Fr8 the number of vesicles is higher in fractions 1, 2, 3, 4, and 5 (Fr1, Western blot analyses were performed to identify the type of vesicles Fr2, Fr3, F4, and Fr5) when compared with Fr6 and Fr7 and reached contained in each EV fraction. Antibodies to annexin A1, annexin a minimum in Fr8 (Fig. 1B). EV protein content did not parallel the A2, and annexin V, markers of microvesicles (2), identified these number of vesicles per fraction, suggesting differences in the amount plasma membrane–derived vesicles mainly in LDFs (Fig. 3, A and B). of total proteins per vesicle (Fig. 1C). Normalization of the quantityCaveolin-1, a protein found mainly in microvesicles (18), showed a of proteins to the number of EVs in each fraction revealed three similar distribution (Fig. 3E). Thus, brain microvesicles are concen- patterns, with low-density fractions (LDFs; corresponding to Fr1, trated in LDFs and likely correspond to the electron-lucent, single-­ Fr2, and Fr3) having low protein amount per vesicle, intermediate-­ membrane EVs observed by cryoEM. Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 density fractions (IDFs; corresponding to Fr4, Fr5, and Fr6) show- The ESCRT-associated proteins TSG101 and Alix, markers of ESCRT-­ ing a higher protein amount per vesicle, and Fr8 containing the dependent exosomes, were mostly found in IDFs with a peak in Fr5, highest protein amount per vesicle (Fig. 1D). paralleling the distribution of granular, single- membrane EVs (Fig. 3, EVs found in LDFs were larger as characterized by NTA than A and C). Enriched within IDFs were also other exosomal markers such other EV subtypes (Fig. 1, E and F). A small size difference between as the lysosome-associated membrane protein 2 (LAMP2), the heat IDFs and Fr8 was also observed, with IDF EVs being slightly smaller shock cognate 71 kDa protein (HSC70), Flotilin-1, the glyceraldehyde-­ than EVs found in Fr8 (Fig. 1, E and F). These data indicate that the 3-phosphate dehydrogenase (GAPDH), and poly-ubiquitylated pro- number, protein amount, and size of brain EVs differ substantially teins (2), as well as the endosomal proteins early endosome antigen 1 between OptiPrep fractions. (EEA1), Rab5B, Rab7A, Rab27A, Syntaxin-6 (Syx6) and the autophagy protein Microtubule-associated proteins 1A/1B light chain 3B (LC3B) Morphological analysis of brain EVs by cryogenic (Fig. 3, E and F), confirming the endosomal origin of these vesicles. electron microscopy The exosomal tetraspanins CD63 and CD81 (2), which are found on Imaging of unfixed, nonstained, hydrated brain EVs using cryogenic both ESCRT-dependent and ESCRT-independent exosomes (19), were electron microscopy (cryoEM; Fig. 2A) preserves native EV mor- detected in IDFs but showed a broader distribution across the frac- phology better than transmission electron microscopy (TEM; Fig. 2A, tions, enriched within Fr3 to Fr7 (Fig. 3, A, C, and E). insets with red borders). The eight EV fractions contained three Microvesicle, exosome, and endocytic markers were undetectable main types of EVs, each enriched in LDFs, IDFs, and Fr8 (Fig. 2A, in Fr8. This fraction, the only one that contains a high proportion of magnifications a, b, and c), as well as several evenly distributed double membrane, electron-dense EVs (Fig. 2), was uniquely enriched minor subtypes of EVs (Fig. 2, A to E). None of the samples visual- in mitochondrial proteins found in the mitochondrial outer membrane ized by either cryoEM or TEM contained debris, broken mem- (MOM), mitochondrial inner membrane (MIM), and mitochondrial branes, or protein aggregates, an indication of their purity. CryoEM matrix (mito-matrix), as assessed with antibodies directed against the revealed extensive heterogeneity of the brain EVs, including single-, voltage-dependent anion channel (VDAC), the cytochrome c oxidase double-, and triple-membrane–bound vesicles, with different subunit 4 (COX-IV), and the pyruvate dehydrogenase E1 compo- luminal contents. LDFs were particularly homogeneous and con- nent subunit alpha (PDH-E1), respectively (Fig. 3, A and D). tained mainly large (Fig. 2, B and C), single-membrane, electron-­ Proteins found abundantly in the cytosol (-actin) and the nu- lucent EVs (about 75% of the total; Fig. 2, D and E) and a lesser cleus (Lamin A) and markers of unrelated intracellular compartments number of large, single-membrane vesicles with an intraluminal (the 130 kDa cis-Golgi matrix protein, GM-130), were not observed in vesicle-like body (11 to 15%). Some of these EVs bore spiked mem- any of the fractions, confirming a lack of intracellular contaminants. brane microdomains (Fig. 2A, magnification a), usually at two These data demonstrate that microvesicles are enriched in LDFs, opposite poles of the vesicle, which were occasionally found in while exosomes are distributed from F3 to Fr7, with ESCRT-dependent other EV types (Fig. 2D, IV and VIII), although never on electron-­ exosomes concentrated in IDFs with a peak in Fr5. Fr8 is distinct dense EVs (Fig. 2A, magnifications b and c). IDFs were more var- from the other fractions and contains nonmicrovesicular, non- ied, including a higher presence of smaller, granular, electron-dense exosomal double-membrane EVs containing abundant mitochondrial EVs, whose frequency peaked in Fr5 (46% of the total) with de- markers. We have named these newly identified EVs mitovesicles. creased frequency in Fr6 and Fr7. The luminal granularity was, in most vesicles, confined to the center of the vesicles and absent from Lipid analysis of LDF, IDF, and Fr8 brain EVs the area adjacent to the membrane (Fig. 2A, magnification b). Fr7 Cellular membranes, including plasma membrane, endosomes, and was particularly heterogeneous, while Fr8 was uniquely char- mitochondria, differ in their lipid composition, and lipid content acterized by a high percentage of double membrane, highly electron-­ differences are reflected in EVs derived from these respective mem- dense EVs [45% of the total; Fig. 2, A (magnification c), D, and E]. branes, with some unique characteristics of each EV type. We char- Infrequently (about 6%), these vesicles were surrounded by a single acterized the EV content in each fraction of lipids that account for membrane. >95% of lipids found in EVs (20), including phosphatidylserine D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 2 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 Fig. 1. Subpopulations of brain EVs with different morphometric features are separated through a high-resolution density gradient. (A) Schematic protocol of the brain EV isolation and a representative picture of the OptiPrep density gradient fractionation. (B) NTA, (C) protein content, and (D) quantification of protein content per vesicle in EV fractions isolated from the brain of wild-type mice, normalized to the brain weight. (E) Diameter analysis of brain EVs by NTA. The bell curves are normal- ized to the mode of the distribution. (F) Percentage of vesicles with a diameter that falls within a 25-nm bin, as estimated in (E). The numbers of LDFs (Fr1 to Fr3, green) and of IDFs (Fr4 to Fr6, blue) were combined as they showed similar results. Fr8 is in red. Statistical test: (B) to (D), one-way ANOVA with Tukey’s multiple comparisons test; (F), two-way ANOVA with Tukey’s multiple comparisons test. Number of independent isolations: (B), 7; (C), 15; (D) to (F), 6. Data shown in (B) to (F): Means ± SD. *P < 0.05, **P < 0.01, ***P < 10−3, #P < 10−4. (PS), phosphatidylcholine (PC), phosphatidylethanolamine (PE), plasma membrane of oligodendrocytes but also at low levels in neurons, phosphatidic acid (PA), cardiolipin (CL), sterols [cholesterol and where it colocalizes with Caveolin-1 (22). This is consistent with the cholesteryl esters (CE)], acylglycerols [mono- (MAG), di- (DAG), presence of GalCer mainly in Fr1. Levels of the other lipids analyzed and triacylglycerols (TAG)], lysobisphosphatidic acid (LBPA), and did not differ between fractions or were only marginally represented the sphingolipids ceramide, sphingomyelin (SM), galactocerebro- (Fig. 4, F to H). LBPA and MAG were below the level of detection. side (GalCer), GD1a, and GT1b. The amount of lipids per microg- Exclusion of LBPA from the list of the brain EV lipids is consistent am of protein was higher in LDFs compared with Fr8, with IDFs with the previous data on EVs isolated from cell media (20). having an intermediate quantity, consistent with their differences in Based on a comprehensive analysis of morphology, protein mark- density (Fig. 4A). When analyzed by molar percentage, the amount ers, and lipid composition, we show that brain microvesicles and of different classes of lipids in Fr1 to Fr5 (sterols: 39 to 43%, PC: 29 exosomes are enriched in distinct density fractions. Within these to 32%, sphingolipids: 8 to 11%, PS: 9 to 10%, PE: 4%, acylglycerols: two groups, some heterogeneity of vesicle subtypes is apparent, for 1%, and CL <1%) was similar to previous reports on small EVs iso- example, in the differences seen in ESCRT-dependent and overall lated from cell culture media of several cell lines, with PA being the exosome distribution. In addition, we have identified a type of brain only exception, as previously it was found only in trace amounts (20). EV isolated in the most-dense fraction: mitovesicles. Mitovesicles Fr8 displayed a different lipid composition (Fig. 4B). CL, a unique have morphological characteristics consistent with a mitochondrial tetra-acyl lipid almost exclusively found in the MIM, peaked in Fr8 origin (e.g., double membrane and electron density) and are uniquely (Fig. 4C), which also displayed a higher amount of PC (48%) and PE enriched in mitochondrial proteins while lacking microvesicle and (9%) and a lower content of cholesterol (18%) and SM (1.7%) com- exosomal protein markers and containing a lipid profile that is con- pared with the other fractions (Fig. 4, D to F). sistent with a mitochondrial origin. Fr8 was also poor in PA, which, in turn, was found at higher levels in IDFs, in agreement with the endosomal origin of IDF vesi- Mitovesicle proteomic and metabolic analyses cles. PA was reported to be enriched in MVBs, closely associated We undertook a qualitative liquid chromatography–mass spectrom- with the ESCRT complex, where it facilitates membrane budding and etry (LC-MS) analysis of Fr8 isolated from wild-type mouse brains fusion in cooperation with the Cbl-interacting protein of 85 kDa to determine the full range of detectable mitochondrial proteins. (CIN85) (21). Another lipid component showing a different pattern We found 279 mitochondria-specific proteins of 673 proteins between fractions was GalCer, which peaked in Fr1. GalCer is found identified in Fr8, which were classified according to known func- almost exclusively in the central nervous system, particularly in the tion (Fig. 5A). D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 3 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 Fig. 2. Cryogenic electron microscopy reveals enrichment of specific EV subtypes in each fraction. (A) Representative photomicrographs of brain EVs imaged by either cryoEM or TEM (insets with red borders). Unlike TEM, under cryoEM morphology was preserved. Dashed boxes: area magnified in a, b, and c, respectively. (B and C) Quantification of vesicle diameter in each EV fraction imaged under cryoEM or TEM. (D and E) Representative photomicrographs and relative percentage per fraction of morphologically different EV subtypes imaged under cryoEM. Statistical test in (B) and (C): Kruskal-Wallis with Dunn’s multiple comparisons test. Number of replicates in (B) and (C): 30 random photomicrographs per fraction from three independent isolations; (E): 30 random cryoEM photomicrographs per fraction from two independent isolations. Data shown in (B) and (C): box and whiskers plot (whiskers: 1st and 99th percentiles). Scale bars, 100 nm (A and D; main figure), 50 nm [(A) insets (a, b, and c magnifications and TEM photomicrographs)]. *P < 0.05, **P < 0.01, #P < 10−4. The mitovesicle proteome was enriched in catabolic pathways In contrast, mitovesicles lacked several biosynthetic and import (table S2). Electron transport chain (ETC) and the pyruvate dehydro- proteins. Proteins involved in the cytosol-to-mitochondria peptide genase complex subunits were almost fully represented, as well as import machinery (TOM/TIM23/PAM supercomplex), ETC assem- other energy production pathways, such as Krebs cycle, ketone body bly factors, iron-sulfur cluster formation and loading, ubiquinone catabolism, fatty acid -oxidation, neurotransmitter metabolism, biosynthesis, mitoribosome components, mitochondrial transfer RNA and membrane transporters of metabolites directly linked to energy (mt-tRNA) metabolism, and mtDNA replication, transcription and production. The mitochondrial intermembrane bridging (MIB) translation were only scarcely found in mitovesicles, although relatively complex, which tethers the MIM to the MOM and promotes MIM abundant in intracellular mitochondria. We also identified the pres- curvature, was also detected. ence of pro-fission proteins and a complete lack of pro-fusion factors. D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 4 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 Fig. 3. LDFs, IDFs, and Fr8 are enriched in microvesicles, exosomes, and mitovesicles, respectively. (A) Representative Western blot analysis and Coomassie Blue staining (loading control) of EVs isolated from the brain of wild-type mice and fractionated by an OptiPrep density gradient. The same volume of each fraction was loaded. (B) Densitometric quantifications of the microvesicle, (C) exosome, and (D) mitovesicle markers shown are normalized to the brain weight. (E) Representative Western blot analyses of EVs isolated from the brain of wild-type mice and fractionated by an OptiPrep density gradient. The same volume of each fraction was loaded. (F) Representative Western blot analyses of the endocytic proteins found in brain EVs. Brain homogenate (BH) was loaded as a reference. Statistical test in (B) to (D): one-way ANOVA with Tukey’s multiple comparisons test. Number of independent isolations: 10. Data shown in (B) to (D): Means ± SD. *P < 0.05, **P < 0.01, ***P < 10−3, #P < 10−4. D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 5 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 Fig. 4. Lipid analysis of brain EV fractions confirms the nature of LDF, IDF, and Fr8 EVs. (A) Quantification of the total lipid content per fraction. The lipids analyzed were phosphatidylserine (PS), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidic acid (PA), cardiolipin (CL), sterols [cholesterol and cholesteryl esters (CE)], acylglycerols [mono- (MAG), di- (DAG), and triacylglycerols (TAG)], and the sphingolipids ceramide, sphingomyelin (SM), galactocerebroside (GalCer), GD1a, and GT1b. The graph shows the total amount of these lipids normalized to the amount of proteins in each fraction as estimated by BCA. (B) Schematic lipid composition of the EVs found in each fraction expressed as a molar percentage. The lipids analyzed were described in (A) and grouped according to their families. (C) Cardiolipin molar percentage of brain EVs. (D) PC and cholesterol molar percentage of brain EVs. (E) GalCer, PA, and PE molar percentage of brain EVs. (F) Sphingomyelin, ceramide, GD1a, and GT1b content (expressed as a molar percentage over the total) of brain EVs. (G) Phosphatidylserine content (expressed as a molar percentage over the total) of brain EVs. No statistical differences were found. (H) Cholesteryl esters, DAG, and TAG content (expressed as a molar percentage over the total) of brain EVs. No statistical differ- ences were found. Number of independent isolations: (A), 5; (B), 5; (C), 4; (D), 5; (E), 4 to 6; (F), 4; (G), 5; and (H), 4. Statistical test in (A) to (H): one-way ANOVA with Tukey’s multiple comparisons test. Data shown in (A) to (H): Means ± SD. *P < 0.05, **P < 0.01, #P < 10−4. Recent evidence indicates neuronal and astrocytic mitochondria inantly expressed by neurons (23). Each of these proteins was identi- contain different molecular profiles and have differential enrichment fied in our Fr8 LC-MS analysis (table S2), suggesting that both of specific proteins (23). Specifically, Sideroflexin-5 (Sfxn5) and the astrocytes and neurons have the capability of secreting mitovesicles monoamine oxidase type B (MAO–B) are astrocyte specific, whereas into the extracellular milieu. the protein NipSnap homolog 1 (Nipsnap1), the OCIA domain-­ LC-MS findings were validated by Western blot analysis of Fr8. containing protein 2 (Ociad2), the mitochondial calcium uniporter The presence of the anion channel VDAC, neuroamine degrading (Mcu), and the monoamine oxidase type A (MAO-A) are predom- enzymes MAO-A, MAO-B, PROHIBITIN 1 and 2, the optic atrophy D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 6 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 Fig. 5. Mitovesicles are metabolically active EVs with a specific subset of mitochondrial proteins. (A) EVs were isolated from brains of 12-month-old wild-type mice and fractionated through an OptiPrep step gradient. Fr8 EV lysates were analyzed by LC-MS, and the mitochondrial proteins found were clustered according to their known functions. Mitochondrial pathways that were highly underrepresented are indicated as “high” and “low,” respectively. (B) Representative Western blot analysis of brain EVs with antibodies to mitochondrial proteins. (C) Representative Western blot analyses of brain EVs using antibodies to DRP1 and PINK1. (D) Relative number of MTDR-positive (MTDR+) vesicles found in Fr8 EVs as quantified by NTA and expressed as percentage over the total number of EVs. As a control, samples were treated with the uncoupler FCCP that dissipates mitochondrial potential. (E) Analysis of ATP levels in brain Fr8 using a luciferase-based assay. As a control, samples were treated with oligomycin + antimycin-A (OA). Each experiment was run in triplicate. Chemiluminescence was quantified every 1.5 min. (F) MAO activity was measured in Fr8 brain EVs using a commercial MAO enzymatic activity assay. As controls, samples were treated with clorgyline (MAO-A inhibitor), pargyline (MAO-B inhibitor), or both. Fluorescence was read every 5 min. For additional data see table S2. Brain homogenate (BH) was loaded as a reference. Statistical test in (D): Student’s t test; (E) and (F): two-way ANOVA with Tukey’s multiple comparisons test. Number of independent isolations: (A), 3; (B) and (C), 10; (D), 6; (E) and (F), 4. Data shown in (D) to (F): Means ± SD. *P < 0.05, **P < 0.01, ***P < 10−3, and #P < 10−4. protein 1 homolog (OPA1), ETC complex subunits NADH dehydro- and PDH-E1 were confirmed. Moreover, Tomm20, subunit of the genase beta subcomplex subunit 8 (NDUFB8), Succinate dehydro- TOM complex, and mitofusin-2, the main pro-fusion protein in the genase subunit B (SDH-B), ubiquinol-cytochrome-c reductase murine brain, were undetectable by both LC-MS and Western complex core protein 2 (UQCRC2), COX-IV, and ATP synthase sub- blot analysis (Fig. 5B). The shift toward pro-fission pathways in mito­ unit alpha (ATP5A), chaperone heat shock protein 60 KDa (HSP60), vesicles as indicated by LC-MS was also confirmed by the analysis of D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 7 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE OPA1 isoforms. The indication of a fused network, revealed in respectively (tables S3 and S5). The quantitative analysis of the pro- brain homogenates by a stronger signal for OPA1-L (long isoforms) teome revealed that 11 proteins were less abundant in Ts2 EVs than for OPA1-S (short isoforms), was not observed in Fr8. compared with 2N (P < 0.05; fig. S1A and table S4). Among them The dynamin-related protein 1 (DRP1), master regulator of mito- was cystatin C (P = 0.023), consistent with our previous observa- chondrial dynamics, was not found in mitovesicles, confirming LC-MS tions that a moderate higher cystatin C expression can mitigate Ts2 data (Fig. 5C). However, both monomeric and tetrameric DRP1 were phenotypes in vivo (29). By contrast, 149 proteins were up-regulated present in exosome-enriched fractions, as reported for the full-length in Ts2 EVs compared with 2N (P < 0.05; fig. S1A and table S6). As isoform of the PTEN-induced kinase 1 PINK1 (fl-PINK1), a protein expected, several of these are due to higher expression of genes linked that can be found both in the cytosol and on the MOM (24). When to the segmental trisomy of Ts2 mice. For example, Sod1, which is imported into the mito-­matrix, PINK1 is cleaved into a ~53-kDa triplicated in Ts2 mice (28), was higher in Ts2 EVs compared with fragment (PINK1-1) or a ~44-kDa peptide (PINK1-2), the latter 2N levels (P = 0.001). The majority of identified proteins not from being a mitophagy-independent isoform (25). Mitovesicles lacked triplicated regions in Ts2 EVs are mitochondrial (fig. S1B). A total fl-PINK1 and PINK1-1 but retained PINK1-2. of 187 mitochondria-specific proteins were identified in the EVs (fig. LC-MS analysis of Fr8 showed the presence of the core machin- S1C). Among them, a large cohort was greater in Ts2 compared ery to produce adenosine 5′-triphosphate (ATP) in mitovesicles with 2N EVs (73 proteins; P < 0.05), and 81 were exclusively found (table S2). Accordingly, we posited that mitovesicles could retain a in Ts2 EVs (tables S5 and S6). Among the 50 most differentially pre­ functional intermembrane H+ potential and synthetize ATP. To test sent proteins by P value, 38 (76%) were mitochondrial. In addition, Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 this hypothesis, we isolated EVs from wild-type mouse hemibrains the four proteins with the highest fold change were mitochondrial and incubated Fr8 with pyruvate, malate, and adenosine 5′-diphosphate (fig. S1D). Gene set enrichment analysis revealed that several mito- (ADP) to boost oxidative phosphorylation and with MitoTracker Deep chondrial pathways were up-regulated in Ts2 EVs (table S7). We Red (MTDR), a dye that is incorporated more efficiently into polarized conclude that the mitochondrial proteome found in brain EVs is and actively respiring mitochondria (26). We measured by NTA the enriched in Ts2 mice compared with 2N littermates and represents number of vesicles positive (i.e., above the detection limit) for MTDR. a prominent pathway altered in vivo within DS brain EVs. As a negative control, samples were treated with carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP), an uncoupler that Brain mitovesicle levels in DS dissipates mitochondrial potential. Approximately 50% of identified To determine whether the changes in EVs isolated from the brains vesicles in Fr8 were stained with MTDR, and FCCP treatment de- of Ts2 mice as compared with 2N littermates are due to a modula- creased the number of positive vesicles to ~17% (Fig. 5D), provid- tion of mitovesicle secretion or cargo loading, we further character- ing evidence for the presence of an active potential. To determine ized Ts2 and 2N mitovesicles. Brain-derived Fr8 showed a higher whether mitovesicles can synthetize ATP, we incubated EVs at 37°C protein content (Fig. 6A) and higher mitovesicle numbers in Ts2 and measured ATP using a commercially available ATP assay kit, mice compared with 2N littermates as estimated by NTA (Fig. 6B). resulting in a twofold increase in ATP. As negative controls, treat- In agreement with analysis of total protein content, NTA, and LC- ment of the mitovesicles with oligomycin + antimycin-A (OA), spe- MS data, levels of mitochondrial proteins were also higher in Ts2 EVs cific blockers of the ETC complexes V and III, respectively, did not compared with 2N littermates (Fig. 6, C and D) but did not change affect ATP production (Fig. 5E), demonstrating that mitovesicles are in the respective brain homogenates (Fig. 6, E to G), suggesting an metabolically active and able to autonomously produce ATP. increase in mitovesicle levels in the brain parenchyma. As we have MAO enzymatic activity of mitovesicles was also analyzed because previously reported (1, 12), the number of exosomes is higher in MAO-A and MAO-B were detected in Fr8. As negative controls, Ts2 brain compared with 2N littermates, consistent with the eleva- irreversible inhibitors of MAO-A (clorgyline), MAO-B (pargyline), tion of exosome markers and the increase in protein content and or a combination of both were used, demonstrating that MAO ac- vesicle numbers in IDFs. Although annexin A2 levels were higher in tivity is maintained in mitovesicles, with MAO-B inhibition having the Ts2 LDFs (Fig. 6, C and D), microvesicle numbers were unaf- a stronger effect than MAO-A inhibition (Fig. 5F). Concomitant pres- fected (Fig. 6, A and B). Consequently, annexin A2 alterations were ence of both proteins and activities suggests that both neurons and likely related to an altered cargo loading of this protein in the Ts2 astrocytes secrete enzymatically active mitovesicles into the murine mice rather than an alteration of microvesicle secretion, consistent brain extracellular space. with the up-regulation of annexin A2 in Ts2 brain homogenates (Fig. 6, E and F).Analysis of EVs obtained from human postmortem Proteomic analysis of EVs isolated from brains of a DS brains of individuals with DS and age-matched cognitively normal mouse model controls devoid of neuropathology (table S8) confirmed our find- DS brains are characterized by extensive mitochondrial abnor- ings in the murine model (Fig. 6, H and I). These data demonstrate malities originating in fetal stages and persisting throughout life a higher number of mitovesicles in DS brain parenchyma, both in span of individuals with DS (27). Accordingly, we hypothesized that murine and human postmortem brains. mitovesicles are aberrant in DS and DS models. To test our hypoth- esis, we isolated EVs from brains of the trisomic mouse model Brain mitovesicle cargo in DS Ts[Rb(12.1716)]2Cje (hereafter Ts2) (28) and diploid (2N) littermates, We sought to determine whether mitochondrial proteins, mtDNA, combining all EV fractions for each genotype. Ten micrograms of and mtRNA are altered in their abundance per mitovesicle by Ts2 total EVs per mouse was analyzed by label-free LC-MS–based pro- trisomy. Equal numbers of mitovesicles as determined by NTA were teomics (tables S3 to S7). Approximately 1600 proteins were identi- analyzed, revealing that UQCRC2, VDAC, and SDH-B levels were fied in total, as detected in at least one mouse per genotype. A total lower in Ts2 mitovesicles compared with 2N littermates, while of 16 and 81 proteins were exclusively found in 2N and Ts2 samples, COX-IV, NDUFB8, and HSP60 were unchanged (Fig. 7, A and B). D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 8 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 Fig. 6. Brain mitovesicle levels are altered by neuropathology. (A) Protein levels and (B) number of EVs normalized to the brain weight from 12-month-old 2N and Ts2 mice fractionated by an OptiPrep gradient. LDFs, IDFs, and Fr7/Fr8 were combined. Fr8 alone is also shown. (C) Representative Western blot analyses and (D) densitometric quantification of microvesicle, exosome, and mitovesicle markers in murine brain EVs. Quantifications were performed combining the signal from all the fractions normal- ized to the brain weight. (E to G) Representative Western blots and densitometric analyses of microvesicle, exosome, and mitovesicle markers in brain homogenates of 12-month-old Ts2 and 2N mice. -actin was used as a loading control. APP, triplicated in Ts2 mice, was used as a positive control. Two experiments shown in the blot (I and II). Graphs are plotted as a fold change over the 2N controls. (H) Representative Western blot analyses and (I) densitometric quantification of microvesicle, exosome, and mitovesicle markers in human postmortem brain EVs. Quantifications were performed combining the signal from all the fractions normalized to the brain weight. For additional specifications. see table S8. Statistical test in (A) and (B): two-way ANOVA with Tukey’s multiple comparisons test; (D), (F), (G), and (I): Student’s t test. Number of independent isolations: (A) and (D), 8; (B) and (I), 4; (F) and (G), 3. Data shown in (A), (B), (D), (F), (G), and (I): Means ± SD. *P < 0.05, **P < 0.01, ***P < 10−3, #P < 10−4. D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 9 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 Fig. 7. Brain mitovesicle cargo is altered in DS. (A and B) Western blot analyses of EVs isolated from the brain of Ts2 and 2N control mice and fractionated through an OptiPrep gradient. Equal numbers of Fr8 EVs as estimated by NTA were loaded for each isolation and tested for mitochondrial markers. Two isolations are shown in the blot (I and II), the densitometric analysis is plotted as a fold change over the 2N controls. (C) Quantification of mtDNA (normalized to the number of vesicles and plotted as a fold change of trisomic versus control mice) by qPCR in murine brain Fr8 EVs. (D) Quantification of the total RNA (normalized to the number of vesicles) extracted from murine brain Fr5 and Fr8 EVs. (E) qPCR cycles of mt-RNR1 for equal amounts of Fr8 cDNA in Ts2 and 2N brain-derived EVs. (F) qPCR analysis of five mitochondria-encoded genes in Fr8 EV complementary DNAs, normalized to mt-RNR1. Statistical test in (D): two-way ANOVA with Tukey’s multiple comparisons test; (B), (C), and (E): Student’s t test; and (F): mixed effects models. Number of independent isolations: (B), 4; (C) to (F), 5. Data shown in (B) to (F): Means ± SD. *P < 0.05, **P < 0.01, ***P < 10−3, #P < 10−4. Quantification of mtDNA in Fr8 by quantitative polymerase chain tion of 5 of the 13 genes found in mtDNA (ND1, ND4L, Cyb, COX-II, reaction (qPCR) demonstrated that Ts2 mitovesicles had more and ATP8) revealed they were all significantly down-regulated in mtDNA per vesicle than 2N littermate mitovesicles (Fig. 7C). No Ts2 mitovesicles compared with 2N littermates (Fig. 7F). difference in the bulk amount of RNA between the genotypes was These findings show that the protein and nucleic acid composi- observed (Fig. 7D). By qPCR, the most abundant mitochondria-­ tion of mitovesicles is affected by DS genotype, with more mtDNA encoded RNA species in mitovesicles was ribosomal RNA, which did and lower mRNA levels per vesicle. not differ between genotypes (Fig. 7E). Thus, the mitochondria-­ To directly link mitovesicle biology to mitochondrial integrity, encoded 12S RNA (mt-RNR1) was used as a housekeeping gene in we used an in vitro model of mitochondrial damage and analyzed further analysis of mitochondria-encoded mRNA levels. Investiga- the EVs released into the cell media. Human primary fibroblasts were D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 10 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE treated with antimycin-A (anti-A), a specific inhibitor of the ETC cles were interpreted as exosomes bearing mitochondrial proteins, complex III that causes mitochondrial reactive oxygen species (ROS) lipids, and mtDNA. Although we cannot rule out that some mito- accumulation without inducing mitophagy (30). Mitochondrial stress chondrial constituents can be found in exosomes (as the case for following anti-A treatment was confirmed in our experimental con- full-length PINK1; Fig. 5C), this interpretation may be due to the ditions by Western blot analysis of cell lysates. OPA1 cleavage, a marker fact that mitovesicles and exosomes copellet at 100,000g and previ- of mitochondrial fragmentation, was increased, and the levels of ous studies that used low-resolution gradients failed to separate COX-IV decreased relative to controls, indicating oxidative phos- these two populations. Other independent studies have reported the phorylation impairment (fig. S2, A to C). Mito-matrix proteins existence of particularly large (>1 m) vesicles containing intact mito­ PDH-E1 and HSP60, preferentially degraded through mitophagy chondria (3, 8). A limitation of our study is that we apply a 0.2-m (31), were unchanged (fig. S2, A to C). Western blot analysis of EVs filtration step to further purify EVs from intracellular contaminants, isolated from the cell media show that the mitovesicle markers and with this step, we lose large EVs that may contain entire mito- COX-IV, PDH-E1, and HSP60 were up-regulated in EVs secreted chondria (or even mitochondrial clusters). In our experience, the by anti-A–treated cells, while exosome and microvesicle proteins 0.2-m filtration step is unavoidable for clean, debris-free prepara- were unaffected (fig. S2, D and E). These data indicate that mito- tions from solid tissues, and therefore, the question remains open chondrial damage directly perturbs mitovesicle biology, either by on whether these structures exist in vivo and what their specific stimulating the release of these vesicles or by up-regulating mito­ contribution is to brain homeostasis and neuronal pathological pro- vesicle cargo loading in single EV. However, COX-IV and HSP60 cesses when compared with mitovesicles. Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 are among the proteins that were unchanged in their content per Mitovesicles differ from intracellular mitochondria. They are up EV in Ts2 brain mitovesicles (Fig. 7, A and B), suggesting cargo to 10-fold smaller, have a narrower intermembrane space [~6 nm, loading for these proteins is not affected by mitochondrial dysfunc- compared to 20 nm of native mitochondria (32)], and lack several tion in vivo. Moreover, PDH-E1 and COX-IV are among the mi- mitochondrial structures, including cristae, mitochondrial ribosomes, tochondrial proteins that successfully predicted a higher release of and proteins constitutively found in mitochondria such as Tomm20 mitovesicles in Ts2 brains (Fig. 6C, fig. S1D, and table S6). The pre- (23). On the other hand, these same features are shared with a spe- ponderance of evidence supports the use of COX-IV, PDH-E1, cific subset of intracellular vesicles known as mitochondria-derived and HSP60 as bona fide markers for mitovesicle levels in total iso- vesicles (MDVs). MDVs bud from the mitochondrial surface, may lated EVs and reinforces the idea that mitochondrial damage directly include the MOM, MIM, and mito-matrix, and are targeted to MVBs stimulates mitovesicle secretion in a mitophagy-independent fash- (33). This suggests MDVs might be released into the extracellular ion, although we cannot exclude the less likely hypothesis of a coor- space as mitovesicles upon fusion of MVBs with the plasma mem- dinated perturbation in cargo loading for all these markers at the brane (34), although the mechanisms responsible for mitovesicle same time. formation from the mitochondria and their secretion into the extra- Overall, our data show that mitovesicle levels and composition cellular space need to be established. Nevertheless, MDVs are het- reflect mitochondrial alterations within the cell of origin and may erogeneous, and some characteristics of these intracellular vesicles serve as a biomarker to assess brain mitochondrial dysfunctions in are not shared with mitovesicles. Several subtypes of MDVs have neurological disorders. been described in previous studies, including mitochondria-anchored protein ligase (MAPL)–containing MDVs, as well as Tomm20 or vacuolar protein sorting-associated protein 35 (Vps35)–positive MDVs DISCUSSION (33). These proteins are not found in mitovesicle preparations by We have established and validated a new experimental approach to LC-MS, and Tomm20 negativity was validated by Western blot analysis. isolate and separate EV subpopulations in vivo from the brain ex- This would point to a model by which several subsets of MDVs with tracellular matrix. We demonstrate that brain EV fractions diverge specific functions exist. One of these subsets would be secreted into in morphometrical properties, morphology, protein, and lipid com- the extracellular space (PDH-E1+ Tomm20− MDVs), possibly exerting position, as well as in in microvesicle and exosome markers. Other specific extracellular functions other than mere carriers of mito- density gradients, mainly sucrose based, have been applied to purify chondrial debris outside the cell. Supportive of this contention, our EVs from the brain but have not been able to generate an unequiv- data demonstrate that mitovesicles are not encapsulated by an ex- ocal separation between EV subtypes (10). This report is novel in ternal membrane, enabling extracellularly the enzymatic activity of that it implements a high-resolution gradient and shows that EV MOM-anchored proteins such as MAO-A and MAO-B and the subpopulations can be fractionated from solid tissues in vivo. Our ability to synthetize ATP at direct contact with the relative substrates data confirm that EV subtypes are not efficiently separated by a dif- (Fig. 5). This evidence would highlight a differential function for ferential centrifugation approach alone, as they largely overlap in mitovesicles relative to an entire mitochondria encapsulated into terms of dimension. However, EV subtypes can be reproducibly microvesicles that are likely shielded from the extracellular space by and rigorously separated by density, as previously demonstrated for the presence of the microvesicular membrane and that have been EVs isolated from cell media, within density ranges that are highly described to serve mainly as energy supply for receiving cells (8). consistent with what was used in vitro to achieve this goal (2). An unanswered question remains on the role that mitovesicles Our high-resolution approach led to the identification of a pre- have during the neuropathological process, such as evidenced in DS. viously unknown third pool of dense, metabolically competent EVs We show that the number of brain mitovesicles is higher (Fig. 6) of mitochondrial origin, which we term mitovesicles, and reinforces and that their composition is altered (Fig. 7) in the brains of Ts2 mice the importance of high-resolution density step gradients in EV and individuals with DS as compared with age-matched diploid con- preparations. A large body of evidence demonstrates the presence trols. DS is characterized by mitochondrial dysfunction and ROS of mitochondrial material in 100,000g EV pellets (3–7). These vesi- accumulation, as was previously demonstrated either in cell lines D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 11 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE (17) or in DS brains (27), as well as in the brains of related murine found in our LC-MS data was Bax, and this overlaps with the human models, independently of the amyloid  precursor protein (APP) plasma dataset. Thus, the analysis of mitovesicles isolated from bio- and the superoxide dismutase isoform 1 (SOD1) triplication (35). fluids can represent an innovative, yet unexplored biomarker approach In addition, it was shown that DS cells are mitophagy deficient (17), to test mitochondrial status in vivo in a noninvasive way, a procedure suggesting that brain cells potentially accumulate toxic mitochondria that we are currently exploring in neurological diseases such as AD. over time in the absence of a prosurvival mitochondria quality control In summary, similar to more extensively defined EV populations mechanism, such as mitophagy, which is able to digest damaged mito- such as exosomes (16), mitovesicles may reflect and modulate chondria (36, 37). Mitovesicle secretion into the brain extracellular pathological changes in the intracellular organelle from which they are space of individuals with mitochondrial dysfunction may contribute derived as well as play important roles in cell-to-cell communication to the disease or act to eliminate detrimental mitochondrial material or in the remodeling of the extracellular environment. A rich appreci- from the cell and mitigate oxidative stress in a homeostatic feedback ation has developed in the last decade of the multiple functions and loop, although additional support is necessary to validate this notion. clinical relevance of other EV subtypes. The mitovesicles defined and In addition, mitovesicle release may be a mechanism to transfer described herein are likely to contribute in similar ways to our under- bioenergetically competent packages from one cell to the other. A standing of the biology of mitochondrial regulation and dysfunction. similar cooperative model has been proposed to explain the transfer of mitochondria encapsulated into microvesicles from astrocytes to neurons in vitro, a phenomenon up-regulated in the presence of ox- MATERIALS AND METHODS Murine models Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 idative stress and that may be considered a neuroprotective mecha- nism (8). In DS, the up-regulation of mitovesicle secretion follows A breeding colony of B6EiC3Sn-Rb(12.Ts171665Dn)2Cje/CjeDnJ this pattern; however, the alteration in the composition (and partic- (abbreviated in Ts2, RRID:IMSR_JAX:004850) mice (28) was main- ularly, the down-regulation in EVs of proteins involved in ATP pro- tained on a B6EiC3SnF1/J (abbreviated in 2N, RRID:IMSR_ duction such as UQCRC2 and SDH-B; Fig. 7, A and B) would be JAX:001875) background. Only Rd.−/− mice (not carrying the reti- responsible of the failure in restoring normal ATP levels. nal degeneration mutation found in some Ts2 colonies) have Mitovesicle-related MAO activity is preserved. Accordingly, it is been used for our studies. Trisomic and 2N mice were examined at reasonable to conclude that mitovesicles modulate levels of neuro­ 12 months of age in all our experiments, because this is the age at transmitters in the extracellular space, potentially linking mito- which Ts2 mice start to show an EV-related phenotype (1, 12). chondrial damage and neurotransmission in a different system. The C57BL/6J mice (indicated as wild-type mice in the text, RRID:IMSR_ increase in mitovesicle levels in DS brain may be related to an eleva- JAX:000664) were assessed at 12 months of age for the characteriza- tion of extracellular MAO activity, which likely leads to a decline in tion of the protocol of isolation of brain EVs. Both females and the levels of monoamines. Notably, DS brains are characterized by a males were used in all experiments. All animal procedures were per- specific decrease in monoamines such as serotonin and dopamine, formed following the National Institutes of Health guidelines with but not glutamate or glycine (38). Further studies are needed to elu- approval from the Institutional Animal Care and Use Committee at cidate whether mitovesicles are a driver of this decline. the Nathan S. Kline Institute for Psychiatric Research. Last, we hypothesize a role for mitovesicles in the neuroinflam- mation process. Mitovesicles may activate microglia as proinflam- Cell lines matory agents given the presence of mitochondrial material and Foreskin male untransformed primary fibroblasts from a 2-year-old DAMPs in single vesicles. Support for this comes from Parkin– apparently healthy subject (abbreviated in fibroblasts, Coriell, cata- knockout (KO) mice, in which mitochondrial stress causes an increase log no. AG07095, RRID:CVCL_0N66) were grown at 37°C in in mtDNA levels in plasma. This drives a sustained up-regulation of Dulbecco’s modified Eagle’s medium (DMEM), supplemented with neuroinflammation (9). How non-necrotic cells extrude DAMPs 10% (v/v) fetal bovine serum, penicillin (100 U/ml) + streptomycin and mtDNA into the extracellular space has yet to be demonstrated, (100 g/ml), and 2 mM GlutaMAX (all reagents from Thermo Fisher although mitovesicles would appear to be appropriate for this sig- Scientific, Waltham, MA, USA) at 5% CO2 in a humidified incuba- naling activity. Although not demonstrated in the present report, tor. The EV-depleted cell medium (conditioned cell medium) was the presence of mitovesicles in the blood, similarly to other EVs, is obtained by overnight ultracentrifugation of the supplemented me- likely. A strong support for this assumption comes from public re- dia at 100,000g (36,000 rpm) at 4°C in a fixed-angled rotor (k-factor: positories of whole proteomes detected in human plasma by LC-MS, 207.5, 45Ti rotor type, Beckman Coulter, Brea, CA, USA). To avoid such as the Human Protein Atlas (39–41), available at https:// cellular senescence, experiments were performed using cells at pas- proteinatlas.org/humanproteome/blood/proteins+detected+ sage 14 or lower (starting from frozen stocks at passage 8). Myco- in+ms. These high-throughput data not only show that several plasma contamination was routinely checked under a microscope mitochondrial proteins are present in human plasma but also that through 4′,6-diamidino-2-phenylindole (DAPI) staining of an ali- these mitochondrial datasets are highly selective and largely overlap quot of the cells in use by ensuring the absence of intracellular, ex- with the mitovesicle proteome (table S2). For instance, mitochon- tranuclear DAPI signal. drial proteins reported in human plasma comprise ETC complex subunits, Krebs cycle enzyme, HSP60, and MAO-B. However, no Human subject details mitochondrial ribosomal proteins, mitofusins, TOM subunits, Postmortem frozen brain tissues (Brodmann area 9) from human Coenzyme Q10 biosynthesis protein (COQ) members, or mtDNA DS and control subjects (table S8) were provided by J. Wegiel, head replication/transcription regulators [e.g., polymerase gamma (POLG) of the Brain and Tissue Bank Department, Department of Develop- or the mitochondrial transcription factor A (TFAM)] are reported in mental Neurobiology, New York State Institute for Basic Research the list. Of 14 members of the Bcl2 family, the only reported protein in Developmental Disabilities, Staten Island, NY. D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 12 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Isolation of EVs from brain parenchyma EV fractions were further diluted in degassed, particle-free, 0.2-m- Murine EVs were isolated from the right hemiforebrains (hereafter filtered PBS to reach the appropriate concentrations of particles to “hemibrain”) as previously described (10, 11). Briefly, hemibrains fall into the ideal reading range of the machine (around 100 parti- were minced and incubated with papain (20 U/ml) (Worthington, cles per field of view). Empirically, we found that the most adequate Lakewood, NJ, USA) in Hibernate A (BrainBits, Springfield, IL, concentrations were 1:50,000 to 1:25,000 for Fr1 to Fr4, 1:20,000 to USA) for 15 min at 37°C. The enzymatic digestion was stopped by 1:10,000 for Fr5, 1:10,000 for Fr6 and Fr7, and 1:3,000 to 1:1,000 for the addition of ice-cold leupeptin (5 g/ml), antipain dihydrochlo- Fr8. Readings were performed as per the manufacturer’s instruction. ride (5 g/ml), pepstatin A (5 g/ml), 1 mM phenylmethanesulfonyl The parameters set were as follows: sensitivity, 85; minimum size, 10; fluoride, 1 M E64 (all reagents from Sigma-Aldrich, St. Louis, MO, maximum size, 1000; shutter, 100; frame rate, 30; trace length, 15; USA) in Hibernate A. The solution was gently disassociated by pi- bin size, 5 nm; and positions per single, reading 11. The temperature petting and centrifuged at 300g for 10 min at 4°C to eliminate undi- of the machine never exceeded room temperature during acquisitions. gested tissue and intact cells. The supernatant was subsequently For the MTDR analysis, Fr8 EVs were resuspended in buffer filtered twice, first through a 40-m cell strainer (Fisher Scientific, B with 1 mM ADP, 1 mM pyruvate, 1 mM malate (Sigma-Aldrich), Pittsburgh, PA, USA) and then through a 0.2-m surfactant-free and 500 nM MTDR (Thermo Fisher Scientific) and incubated at cellulose acetate syringe filter (Corning Incorporated, Corning, NY, 37°C for 15 min with or without FCCP 20 M (Sigma-Aldrich). USA). The cleared mixture was centrifuged at 2000g for 10 min at Samples were eventually read using the ZetaView PMX-220 twin at 4°C to discard large debris and apoptotic bodies, and the superna- a concentration of 1:10,000 to ensure proper dilution of the dye and Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 tant was centrifuged at 10,000g for 30 min at 4°C to discard smaller minimization of the background. debris. The supernatant was then ultracentrifuged at 100,000g for 70 min at 4°C. The pellet contained the crude EV preparation. The BCA protein assay pellet was washed once in phosphate-buffered saline (PBS) (pH 7.4) Pierce BCA (bicinchoninic acid) Protein Assay Kit (Thermo Fisher (Corning Incorporated), recentrifuged at 100,000g for 70 min at Scientific) was used as per the manufacturer’s instructions with 4°C, and resuspended in a 40% (v/v) OptiPrep (iodixanol) solution, minimal adjustments. Two microliters of EVs diluted in PBS (in containing 10 mM tris-HCl (pH7.4), 0.25 M sucrose, and 40% iodixanol triplicate) were incubated at 65°C for 30 min in a polypropylene (all reagents from Sigma-Aldrich). An OptiPrep density step gradient microplate (Corning Incorporated) with the BCA Reagent Solution was set up by carefully layering on the top of the 40% OptiPrep-­ (50:1 reagent A to reagent B, provided in the kit). A bovine serum equilibrated EVs a decreasing scale of OptiPrep solutions (20, 15, 13, albumin (BSA; provided in the kit) standard curve was loaded in 11, 9, 7, and 5%). The column was centrifuged overnight at 4°C in a parallel for each plate analyzed. The colorimetric reaction was swinging bucket rotor at 200,000g (k-factor: 137.3, SW40Ti rotor, quantified as the absorbance at 562 nm using the VersaMax micro- Beckman Coulter). Afterward, 1.5-ml fractions (eight fractions in plate reader (Molecular Devices, San Jose, CA, USA). total), corresponding to the different interphases, were collected, washed in PBS, and centrifuged at 100,000g for 70 min at 4°C in a fixed-angle ATP assay rotor (k-factor: 61.2, MLA80 rotor, Beckman Coulter). Pellets were Fr8 brain EVs were resuspended in buffer B, and protein content resuspended in PBS, buffer B (225 mM sucrose, 44 mM KH2PO4, was quantified by the BCA assay as described above. An ATP stan- 12.5 mM magnesium acetate, and 6 mM EDTA, all reagents from dard curve and 5 g of these preparations were plated in triplicate in Sigma-Aldrich) or lysed in radioimmunoprecipitation assay (RIPA) white, opaque, MicroFluor1 plates (Thermo Fisher Scientific) and buffer with protease inhibitors for further analysis, as described below. incubated in the presence or absence of 2 M oligomycin + 4 M antimycin-A (OA) in an ATP assay MasterMix [1 mM ADP, 1 mM pyr­ EV isolation from cell media uvate, 1 mM malate (reagents from Sigma-Aldrich), 1 mM DTT, Fibroblasts at 80% confluence in 150-mm plates (Corning Incorpo- 100 M EDTA, 5 mM MgSO4, 25 mM tricine buffer (pH 7.8), 0.5 mM rated) were treated for 24 hours with 4 M anti-A (Sigma-Aldrich) d-luciferin, and recombinant firefly luciferase (1.25 g/ml) (reagents in DMSO (Sigma-Aldrich) or with DMSO only in EV-depleted cell provided in the ATP Determination Kit, catalog no. A22066, from medium. The medium was collected and centrifuged at 300g for Invitrogen, Thermo Fisher Scientific)]. Chemiluminescence was read 10 min at 4°C. The supernatant was serially centrifuged at 2000g for in a SpectraMax microplate reader (Molecular Devices) at 30°C and 10 min at 4°C to discard large debris and at 10,000g for 30 min at after reaching 37°C every 1.5 min. 4°C for large debris. The supernatant was then ultracentrifuged at 100,000g (36,000 rpm) for 70 min at 4°C in a fixed-angled rotor Monoamine oxidase assay (45Ti rotor type, Beckman Coulter), and the pellet was washed in The MAO enzymatic assay was performed using the MAO Assay Kit 1 ml of PBS and centrifuged again at 100,000g (50,000 rpm) for (catalog no. MAK136) from Sigma-Aldrich following the manufac- 70 min at 4°C in a fixed-angled rotor (k-factor: 79.9, TLA-55 rotor turer’s instructions. Briefly, 2 g of Fr8 brain EVs were incubated in type, Beckman Coulter). The washed EV pellet was then resuspended the presence or absence of 1 M clorgyline (irreversible MAO-A in RIPA buffer and prepared for Western blot analysis. inhibitor), 1 M pargyline (irreversible MAO-B inhibitor), or both together (provided with the kit) in triplicate in a fluorescence-­ Nanotrack analysis optimized, dark microplate (Corning Incorporated) for 10 min at Fresh, unfrozen brain EV OptiPrep fractions from wild-type, 2N, room temperature before the start of the assay. A MAO assay mix and Ts2 mice were resuspended in PBS and analyzed by NTA (provided with the kit) was then added to the samples, and the plate (ZetaView model PMX-120 basic or PMX-220 twin for fluores- was read in a SpectraMax microplate reader (Molecular Device) at cence analysis, Particle Metrix, Meerbusch, Germany) to estimate room temperature with excitation/emission wavelengths of 530/585 nm, the amount and the dimension of the vesicles. Before the acquisition, respectively. Fluorescence was read every 5 min for 2 hours. D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 13 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE Negative stain of EVs and TEM Liquid chromatography–tandem mass spectrometry Brain EVs were fixed in a paraformaldehyde (PFA) solution [contain- 2N and Ts2 fractions of brain EVs were combined for the analysis of ing 2% (w/v) PFA in a 100 mM sodium cacodylate buffer, reagents the levels of mitochondrial proteins in Ts2 EVs (fig. S1). Ten micro- from Electron Microscopy Sciences, Hatfield, PA, USA], while grams of EV lysates were analyzed from three mice for each geno- Formvar-covered carbon-coated copper, 200 mesh grids (Ted Pella type. For the qualitative analysis of the mitovesicle proteome Inc., Redding, CA, USA) were glow discharged in a PELCO easiGlow (Fig. 5A), Fr8 EVs were isolated from three wild-type mice and frac- (Ted Pella Inc.). Three microliters of fixed EVs were added to the tionated through an OptiPrep-based step gradient. EVs were run grid, held by self-clamping tweezers, and left for 20 s. The solution briefly into a 15% SDS–polyacrylamide gel electrophoresis gel until was then wicked off with a Whatman filter paper and negatively all proteins focused into a single 1-cm band. Gels were washed three stained in 1% (w/v) uranyl acetate (Electron Microscopy Sciences) times in ddH2O for 15 min each and visualized by overnight stain- before being immediately wicked off using filter paper. This process ing with EZ-Run Protein Gel Staining Solution (Thermo Fisher Sci- was repeated three times before a final incubation of 5 min. The entific). Stacked and stained protein gel regions were excised and excess solution was wicked off, and the grid was allowed to air dry. destained, using standard, published protocols (43). In-gel digestion Imaging took place on a Talos L120C transmission electron micro- was performed overnight, with MS-grade trypsin (Trypsin Gold, scope operating at 120 kV (Thermo Fisher Scientific). Promega, Madison, WI, USA) at a concentration of 5 ng/l in a 50 mM NH4HCO3 digest buffer. After acidification with 10% (v/v) Cryo–electron microscopy formic acid (final concentration of 0.5 to 1% formic acid), resulting Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 R 2/2 copper 200 mesh Quantifoil grids (Ted Pella Inc.) were glow peptides were desalted using hand-packed, reversed-phase Empore discharged in a PELCO easiGlow (Ted Pella Inc.). Three microliters C18 Extraction Disks (3M, St. Paul, MN, USA). of a 1:10 (v/v) dilution of EVs in PBS were applied to the grid before Desalted peptides were concentrated to a very small droplet by being blotted with filter paper and then rapidly frozen in liquid eth- vacuum centrifugation and reconstituted in 10 l of 0.1% (v/v) for- ane using a Vitrobot Mark IV (Thermo Fisher Scientific). The grids mic acid in water. Around 90% of the peptide material was used for were then transferred to a 914 high tilt model cryo-tomography LC followed by data-dependent-acquisition (DDA) mode tandem holder (Gatan, AMETEK, Berwyn, PA, USA) and imaged using MS (LC-MS/MS). A Q Exactive HF mass spectrometer was coupled low-dose conditions on a Talos L120C transmission electron micro­ directly to an EASY-nLC 1000 (Thermo Fisher Scientific) equipped scope operating at 120 kV (Thermo Fisher Scientific). with a self-packed 75 m × 20 cm reverse-phase column (ReproSil-­ Pur C18, 3 M, Dr. Maisch GmbH, Germany) for peptide separa- Western blot analysis tion. Analytical column temperature was maintained at 50°C by a Samples examined by Western blot were supplemented with a 6× column oven (Sonation GmBH, Germany). Peptides were eluted Laemmli sample buffer [375 mM tris-HCl (pH 7.4), 9% SDS, 50% with a 3 to 40% acetonitrile gradient over 110 min at a flow rate of glycerol, 9% -mercaptoethanol (-ME), 0.03% bromophenol blue, 250 nl/min. The mass spectrometer was operated in DDA mode all reagents from Sigma-Aldrich], boiled for 5 min at 95°C, and with survey scans acquired at a resolution of 120,000 [at mass/ loaded on a 4 to 20% gradient precast polyacrylamide gel (Bio-Rad, charge ratio (m/z) 200] over a scan range of 300 to 1750 m/z. Up to Hercules, CA, USA) to run at 120 V for 2 hours. With the exception 15 most abundant precursors from the survey scan were selected of the blots in Fig. 7 (A and B), equal volumes of brain EVs were with an isolation window of 1.6 Th for fragmentation by higher-­ loaded. For DRP1 immunoblots, gels were run in nonreducing con- energy collisional dissociation with normalized collision energy of ditions (no -ME was added to the Laemmli buffer) to detect DRP1 27. The maximum injection time for the survey and MS/MS scans tetramers. Gels were then either stained for 30 min at room tem- was 60 ms, and the ion target value (automatic gain control, AGC) perature in a Coomassie solution [0.25% (w/v) Coomassie Brilliant for both scan modes was set to 3 × 106. Blue G-250 (Sigma-Aldrich) in 40% (v/v) methanol and 10% acetic The mass spectra files were processed using the MaxQuant acid (v/v) (Fisher Scientific)] or transferred overnight at 100 mA onto proteomics data analysis workflow (version 1.6.0.1) with the polyvinylidene difluoride membranes (Immobilon, EMD Millipore, Andromeda search engine (44). Raw MS files were used to extract Billerica, MA). When needed, as indicated in the figures, membranes peak lists that were searched with the Andromeda search engine were stained using SYPRO Ruby Protein Blot Stain (Lonza, Basilea, against human or mouse proteomes and a file containing contami- Switzerland) following the manufacturer’s instructions before the nants such as human keratins. Trypsin was specified allowing up to blocking step. Membranes were subsequently blocked for 1 hour in two missed cleavages with the minimum required peptide length set 5% (w/v) blotting-grade nonfat dry milk (Bio-Rad) or in 5% (w/v) to be seven amino acids. N-acetylation of protein N termini, oxida- BSA (Sigma-Aldrich), depending on the conditions suggested by the tion of methionines, and deamidation of asparagines and gluta- manufacturers for the primary antibody applied. A comprehensive mines were set as variable modifications. For the initial main search, list of the primary antibodies used is provided in table S9. The parent peptide masses were allowed mass deviation of up to 20 parts monoclonal antibody against APP (clone C1/6.1) (42) was a gift from per million. Peptide spectral matches and protein identifications P. M. Mathews. Horseradish peroxidase–conjugated secondary anti- were filtered using a target-decoy approach at a false discovery rate bodies were purchased from Jackson ImmunoResearch (West Grove, of 1%. We used the raw MS1 intensity for protein quantitation. PA, USA) and used as per the manufacturer’s instructions. Protein bands were visualized either on reflection autoradiography films or Lipid analysis of brain EVs with the iBright FL1500 imaging system (Thermo Fisher Scientific) Each fraction obtained from the OptiPrep density gradient and sus- after incubation of the blots for 5 min with enhanced chemilumi- pended in 20 l of PBS was lyophilized, and the lyophilized powder nesence (ECL) (Pierce, Rockford, IL, USA) or femto ECL (Pierce) was sonicated in 200 l of the mixture of methyl-tert-butyl ether when the signal was too faint. and methanol (1:1, v/v). The resultant fine powder suspension was D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 14 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE centrifuged at 10,000g for 5 min, and the precipitate was reextracted relative gene-level differences between study groups (47, 51–53). with 200 l of the above solvent mixture. The supernatants were Negative controls consisted of the reaction mixture without in- combined to obtain the total lipid extracts. The total lipid extracts put RNA. were partitioned according to the method of Matyash et al. (45). The lower aqueous-methanol phase containing gangliosides was mtDNA extraction from Fr8 EVs and qPCR evaporated to dryness, applied to high-performance thin-layer To extract the DNA found in Fr8 brain EVs without RNA contam- chromatography (HPTLC) plates, and developed with chloroform/ ination, equal amounts of Fr8 EVs were incubated with DNase and methanol/0.25% KCl (5:4:1). The plates including various concen- exonuclease III (Thermo Fisher Scientific) for 30 min at 37°C and trations of GM2 standards were stained with an orcinol reagent, then inactivated at 65°C for 10 min. Protein kinase (Thermo Fisher scanned with the Gel Logic molecular imaging system (Carestream Scientific) digestion was performed for 1 hour at 50°C and heat in- Health, Rochester, NY, USA) and analyzed by Multi Gauge version activated at 95°C for 5 min. DNA was then purified using AMPure 2.0 (Fujifilm USA Medical Systems, Stamford, CT, USA). The con- XP magnetic beads (Beckman Coulter) and resuspended in nuclease-­ centration of each ganglioside was calculated using GM2 standards. free water. DNA was measured by NanoDrop 2000c (Thermo Fisher The upper organic methyl-tert-butyl ether phase obtained by the Scientific) and directly used for qPCR for the mitochondrially en- partition described above was evaporated to dryness and separated coded 12S RNA (mt-RNR1, Mm04260177_s1) gene. into neutral and acidic lipids using DEAE-Sephadex columns. The neutral lipid fraction (including sterols, acylglycerols, ceramide, Quantification and statistical analysis Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 GalCer, PE, PC, and SM) and various concentrations of each lipid Statistical analyses and graphs were performed using the software standard were applied on HPTLC plates and developed first with GraphPad Prism (version 6.01, GraphPad Holdings, San Diego, CA, hexane/ethyl acetate/water/acetic acid (3:4:5:2) until the solvent front USA). All data are shown as means ± SD. For clarity and transpar- ascended to 4 cm from origin, second with hexane/methyl-tert-­ ency, each replicate is shown as a superimposed dot plot and rep- butyl ether/acetic acid (65:35:2) until 6.5 cm, and third with hexane/ resents a different isolation. None of the replicates were excluded methyl-tert-butyl ether (98/2/1) until 9 cm above origin. The acidic from the analysis. When two groups were compared, a two-tailed, lipid fraction (including PA, CL, and PS) and various concentrations unpaired, Student’s t test was performed. When more than two of each lipid were developed on HPTLC with acetone/benzene/ groups were taken into consideration and one variable was analyzed acetic acid/water (20:30:4:1). After development plates were first (e.g., differences between the fractions), one-way analysis of vari- dipped in 20% methanol and then stained in 0.0001% primuline. ance (ANOVA) with Tukey’s multiple comparisons test was used. Fluorescent lipid bands were scanned and analyzed as described When two variables were analyzed simultaneously (e.g., changing above for ganglioside analyses. of the enzymatic activity over time compared with a control condi- tion), a two-way ANOVA with Tukey’s multiple comparisons was EV RNA purification and qPCR performed. Significance was set to a P value of <0.05. Variables con- Fr5 and Fr8 brain EVs were resuspended in PBS, and the RNA sidered, normalization procedures, exact value of n, and other spe- was purified from 16 l of the suspension (46) from each fraction cific statistical information are provided for each experiment in the using the miRNeasy Mini Kit per the manufacturer’s specifi- respective figure legend or in the graph shown in the figure, when cations (217004, Qiagen, Germantown, MD, USA). The deoxyribo- possible. nuclease (DNase) digestion step was performed twice sequentially Protein bands were quantified using ImageJ (54) and analyzed before proceeding with final washes and purification. This was and graphed through GraphPad Prism. EV proteins were routinely performed to ensure no mtDNA contamination in the following normalized to the net weight of brain tissue from which they were qPCRs as mtDNA does not contain introns and could amplify isolated. Densitometric analyses of EVs isolated from conditioned with the TaqMan primers used for qPCR. RNA quantity and puri- cell media were normalized to the SyproRuby total protein staining, fication quality were measured using an Agilent 2100 Bioanalyzer as indicated in the respective figure legends. The level of the pro- (Pico 6000 kit, Agilent, Santa Clara, CA, USA). Reverse tran- teins found in the brain and cell homogenate were normalized over scription of 40 ng of input RNA per sample was performed using a housekeeping loading control, as indicated in the respective figure random hexamers as primers (2.5 M final concentration) and legends. Superscript IV per the manufacturer’s specifications in a 50-l reac- The concentration of each lipid is provided as lipid weight per tion volume (18090050, Thermo Fisher Scientific). qPCR was per- microgram of proteins in each fraction. To estimate the lipid com- formed using TaqMan qPCR primers (Life Technologies, Grand position of each EV, we transformed the mass in moles and calcu- Island, NY, USA) for the following genes: mitochondrially encoded lated the molar percentage of a specific lipid population over the 12S RNA (mt-RNR1, Mm04260177_s1), ATP synthase F0 subunit 8 total. The approximated molecular masses considered in our analy- (Atp8; Mm04225236_g1), cytochrome b (Cytb; Mm04225271_g1), sis were 649 g/mol for cholesteryl esters, 387 g/mol for cholesterol, cytochrome c oxidase subunit II (COX-II; Mm03294838_g1), 786 g/mol for PC, 743 g/mol for PE, 787 g/mol for PS, 682 g/mol for NADH (reduced form of nicotinamide adenine dinucleotide) de- PA, 587 g/mol for DAG, 843 g/mol for TAG, 1893 g/mol for GD1a, hydrogenase subunit 4 (ND4; Mm04225294_s1), and NADH de- 1989 g/mol for GT1b, 1466 g/mol for CL, 538 g/mol for ceramide, hydrogenase subunit 1 ND1 (ND1; Mm04225274_s1). Samples 737 g/mol for SM, and 718 for GalCer. were assayed in triplicate on a real-time qPCR cycler (ABI 7900HT, For calculating the relative frequency of different subtypes of Thermo Fisher Scientific) in 96-well optical plates with cover- EVs in the OptiPrep fractions (Fig. 2E), 30 random cryoEM photo- film as described previously (47–50). Standard curves and cycle micrographs per fraction from two independent isolations were ac- threshold (Ct) were generated using standards obtained from quired. Given the paucity of vesicles found in some of the fractions, total mouse brain RNA. The Ct method was used to determine especially for Fr8, data were pooled together, resulting in 192 vesicles D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 15 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE counted in total for Fr1, 220 for Fr2, 269 for Fr3, 237 for Fr4, 193 for same mouse (59). Significance was judged at the level  = 0.01, two Fr5, 170 for Fr6, 169 for Fr7, and 93 for Fr8. sided; false discovery rate based on an empirical null distribu- For the EV diameter estimation, we used two different approaches: tion due to strong correlation between genes was controlled at electron microscopy (EM) and NTA. Thirty random cryoEM or TEM level  = 0.1. photomicrographs per fraction from three independent isolations were acquired. All the EVs in the images were analyzed except for SUPPLEMENTARY MATERIALS the vesicles found at the border that were not fully included in the Supplementary material for this article is available at http://advances.sciencemag.org/cgi/ content/full/7/7/eabe5085/DC1 field of view (and so the diameter could not be estimated accurately) and pooled for each fraction. Given that many larger EVs, especially View/request a protocol for this paper from Bio-protocol. by TEM, were not exactly circular but tended to have an ovoid or an irregular shape (see Fig. 2A), we did not measure the diameter di- REFERENCES AND NOTES 1. S. A. Gauthier, R. Pérez-González, A. Sharma, F.-K. Huang, M. J. Alldred, M. Pawlik, G. Kaur, rectly, as it would have been slightly different in each position. Ac- S. D. Ginsberg, T. A. Neubert, E. Levy, Enhanced exosome secretion in Down syndrome cordingly, we quantified the area of each vesicle through the software brain - A protective mechanism to alleviate neuronal endosomal abnormalities. ImageJ and calculated post hoc the diameter assuming a circular Acta Neuropathol. Commun. 5, 65 (2017). shape, by the formula d = 2√(A/), where d is the diameter and A is 2. D. K. Jeppesen, A. M. Fenix, J. L. Franklin, J. N. Higginbotham, Q. Zhang, L. J. Zimmerman, D. C. Liebler, J. Ping, Q. Liu, R. Evans, W. H. Fissell, J. G. Patton, L. H. Rome, D. T. Burnette, the area. The resulting box and whiskers plot clearly showed that R. J. Coffey, Reassessment of Exosome Composition. Cell 177, 428–445.e18 (2019). the distribution of this parameter was not normal but positively 3. F. Puhm, T. Afonyushkin, U. Resch, G. Obermayer, M. Rohde, T. Penz, M. Schuster, Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 skewed for all the fractions, as the median did not divide the dis- G. Wagner, A. F. Rendeiro, I. Melki, C. Kaun, J. Wojta, C. Bock, B. Jilma, N. Mackman, tribution in two symmetrical halves (Fig. 2, B and C). For this E. Boilard, C. J. Binder, Mitochondria are a subset of extracellular vesicles released by reason, we used the Kruskal-Wallis H test by ranks (with Dunn’s activated monocytes and induce type I IFN and TNF responses in endothelial cells. Circ. Res. 125, 43–52 (2019). multiple comparisons test), which is the nonparametric equivalent of 4. S. C. Jang, R. Crescitelli, A. Cvjetkovic, V. Belgrano, R. Olofsson Bagge, K. Sundfeldt, the one-way ANOVA (calculated through the software GraphPad T. Ochiya, R. Kalluri, J. Lötvall, Mitochondrial protein enriched extracellular vesicles Prism). The positive skewness of the curve was largely expected discovered in human melanoma tissues can be detected in patient plasma. J. Extracell. because of a well-­established systematic error, called “fishnet artifact,” Vesicles 8, 1635420 (2019). 5. M. Guescini, S. Genedani, V. Stocchi, L. F. Agnati, Astrocytes and Glioblastoma cells by which very small EVs tend to be more easily lost during EM sample release exosomes carrying mtDNA. J. Neural Transm. (Vienna) 117, 1 (2010). preparation. Ultimately, this causes an overestimation of the propor- 6. D. G. Phinney, M. D. Giuseppe, J. Njah, E. Sala, S. Shiva, C. M. St Croix, D. B. Stolz, tion of the large EVs over the small ones (55). S. C. Watkins, Y. P. Di, G. D. Leikauf, J. Kolls, D. W. H. Riches, G. Deiuliis, N. Kaminski, A similar slight positive skewness could be observed by NTA S. V. Boregowda, D. H. McKenna, L. A. Ortiz, Mesenchymal stem cells use extracellular (Fig. 1E). This was also predictable. In the density range of bio- vesicles to outsource mitophagy and shuttle microRNAs. Nat. Commun. 6, 8472 (2015). 7. R. A. Haraszti, R. A. Haraszti, M.-C. Didiot, E. Sapp, J. Leszyk, S. A. Shaffer, H. E. Rockwell, logical samples, NTA cannot detect vesicles that are smaller than F. Gao, N. R. Narain, M. D. Figlia, M. A. Kiebish, N. Aronin, A. Khvorova, High-resolution 50 nm without catching noise signal, while no limitations are pro- proteomic and lipidomic analysis of exosomes and microvesicles from different cell vided in the upper limit (at least for the dimension range of the sources. J. Extracell. Vesicles 5, 32570 (2016). EVs), as per the manufacturer’s guidelines. To circumvent the 8. K. Hayakawa, E. Esposito, X. Wang, Y. Terasaki, Y. Liu, C. Xing, X. Ji, E. H. Lo, Transfer of mitochondria from astrocytes to neurons after stroke. Nature 535, 551–555 (2016). non-normality of the distribution, we performed the analysis using 9. D. A. Sliter, J. Martinez, L. Hao, X. Chen, N. Sun, T. D. Fischer, J. L. Burman, Y. Li, Z. Zhang, a binning system, as shown in the figure, and as successfully applied D. P. Narendra, H. Cai, M. Borsche, C. Klein, R. J. Youle, Parkin and PINK1 mitigate elsewhere (12). STING-induced inflammation. Nature 561, 258–262 (2018). With respect to the analysis of LC-MS data, identified proteins 10. R. Perez-Gonzalez, S. A. Gauthier, A. Kumar, E. Levy, The exosome secretory pathway within EVs were quantified using 2N mice as the reference to cal- transports amyloid precursor protein carboxyl-terminal fragments from the cell into the brain extracellular space. J. Biol. Chem. 287, 43108–43115 (2012). culate log-fold differences. Proteins with at least two reported in- 11. R. Pérez-González, S. A. Gauthier, A. Kumar, M. Saito, M. Saito, E. Levy, A method for tensities per group (Ts2 and 2N) were included in the analysis and isolation of extracellular vesicles and characterization of exosomes from brain compared using a Student’s t test with a P value cutoff (P < 0.05) extracellular space. Methods Mol. Biol. 1545, 139–151 (2017). deemed significant. Differentially regulated proteins were depicted 12. P. D'Acunzo, T. Hargash, M. Pawlik, C. N. Goulbourne, R. Pérez-González, E. Levy, Enhanced generation of intraluminal vesicles in neuronal late endosomes in the brain using ggplot and pheatmap packages in R. Proteins were identified of a Down syndrome mouse model with endosomal dysfunction. Dev. Neurobiol. 79, as “mitochondrial” if verified in mitochondria in at least two of 656–663 (2019). these three databases: MitoCarta 2.0 (56), UniProt (available at 13. R. Pérez-González, S. Sahoo, S. A. Gauthier, Y. Kim, M. Li, A. Kumar, M. Pawlik, L. Benussi, https://uniprot.org) (57), and the Human Protein Atlas (available at R. Ghidoni, E. Levy, Neuroprotection mediated by cystatin C-loaded extracellular vesicles. http://proteinatlas.org) (39, 41). Gene set enrichment analysis Sci. Rep. 9, 11104 (2019). 14. K. Y. Peng, R. Pérez-González, M. J. Alldred, C. N. Goulbourne, J. Morales-Corraliza, (GSEA) of gene sets was performed using fgsea package (58) in R M. Saito, M. Saito, S. D. Ginsberg, P. M. Mathews, E. Levy, Apolipoprotein E4 genotype (table S5). Gene sets were determined to be significant if Padj < 0.05 compromises brain exosome production. Brain 142, 163–175 (2019). with 100,000 permutations and a minimum gene set size of 10. 15. R. Pérez-González, S. A. Gauthier, A. Sharma, C. Miller, M. Pawlik, G. Kaur, Y. Kim, E. Levy, Fgsea was used on Go gene sets from the Msigdb database available A pleiotropic role for exosomes loaded with the amyloid  precursor protein carboxyl- terminal fragments in the brain of Down syndrome patients. Neurobiol. Aging 84, 26–32 from the Broad Institute, available at https://cran.r-project.org/ (2019). web/packages/msigdbr/index.html. Murine brain tissue expression 16. P. M. Mathews, E. Levy, Exosome production is key to neuronal endosomal pathway and gene function indicated in table S2 were derived from UniProt integrity in neurodegenerative diseases. Front. Neurosci. 13, 1347 (2019). and Human Protein Atlas (40) annotations. 17. M. Bordi, S. Darji, Y. Sato, M. Mellén, M. J. Berg, A. Kumar, Y. Jiang, R. A. Nixon, Last, for qPCR analysis of mitochondrial mRNA, the signal mTOR hyperactivation in Down syndrome underlies deficits in autophagy induction, autophagosome formation, and mitophagy. Cell Death Dis. 10, 563 (2019). intensity ratio of each gene was modeled as a function of mouse 18. H. Lee, C. Li, Y. Zhang, D. Zhang, L. E. Otterbein, Y. Jin, Caveolin-1 selectively regulates study group, using mixed effects models with random mouse ef- microRNA sorting into microvesicles after noxious stimuli. J. Exp. Med. 216, 2202–2220 fect to account for the correlation between repeated assays on the (2019). D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 16 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE 19. G. van Niel, S. Charrin, S. Simoes, M. Romao, L. Rochin, P. Saftig, M. S. Marks, E. Rubinstein, S. Navani, C. A.-K. Szigyarto, J. Odeberg, D. Djureinovic, J. O. Takanen, S. Hober, T. Alm, G. Raposo, The tetraspanin CD63 regulates ESCRT-independent and -dependent P.-H. Edqvist, H. Berling, H. Tegel, J. Mulder, J. Rockberg, P. Nilsson, J. M. Schwenk, endosomal sorting during melanogenesis. Dev. Cell 21, 708–721 (2011). M. Hamsten, K. von Feilitzen, M. Forsberg, L. Persson, F. Johansson, M. Zwahlen, 20. T. Skotland, N. P. Hessvik, K. Sandvig, A. Llorente, Exosomal lipid composition and the role G. von Heijne, J. Nielsen, F. Pontén, Proteomics. Tissue-based map of the human of ether lipids and phosphoinositides in exosome biology. J. Lipid Res. 60, 9–18 (2019). proteome. Science 347, 1260419 (2015). 21. J. Zhang, X. Zheng, X. Yang, K. Liao, CIN85 associates with endosomal membrane and 41. M. Uhlen, P. Oksvold, L. Fagerberg, E. Lundberg, K. Jonasson, M. Forsberg, M. Zwahlen, binds phosphatidic acid. Cell Res. 19, 733–746 (2009). C. Kampf, K. Wester, S. Hober, H. Wernerus, L. Björling, F. Ponten, Towards a knowledge- 22. R. D. Singh, V. Puri, J. T. Valiyaveettil, D. L. Marks, R. Bittman, R. E. Pagano, Selective based human protein atlas. Nat. Biotechnol. 28, 1248–1250 (2010). caveolin-1-dependent endocytosis of glycosphingolipids. Mol. Biol. Cell 14, 3254–3265 42. P. M. Mathews, Y. Jiang, S. D. Schmidt, O. M. Grbovic, M. Mercken, R. A. Nixon, Calpain (2003). activity regulates the cell surface distribution of amyloid precursor protein. Inhibition 23. C. Fecher, L. Trovò, S. A. Müller, N. Snaidero, J. Wettmarshausen, S. Heink, O. Ortiz, of calpains enhances endosomal generation of -cleaved C-terminal APP fragments. I. Wagner, R. Kühn, J. Hartmann, R. M. Karl, A. Konnerth, T. Korn, W. Wurst, D. Merkler, J. Biol. Chem. 277, 36415–36424 (2002). S. F. Lichtenthaler, F. Perocchi, T. Misgeld, Cell-type-specific profiling of brain 43. S. Kishinevsky, T. Wang, A. Rodina, S. Y. Chung, C. Xu, J. Philip, T. Taldone, S. Joshi, mitochondria reveals functional and molecular diversity. Nat. Neurosci. 22, 1731–1742 M. L. Alpaugh, A. Bolaender, S. Gutbier, D. Sandhu, F. Fattahi, B. Zimmer, S. K. Shah, (2019). E. Chang, C. Inda, J. Koren III, N. G. Saurat, M. Leist, S. S. Gross, V. E. Seshan, C. Klein, 24. E. K. Steer, M. K. Dail, C. T. Chu, Beyond mitophagy: Cytosolic PINK1 as a messenger of M. J. Tomishima, H. Erdjument-Bromage, T. A. Neubert, R. C. Henrickson, G. Chiosis, mitochondrial health. Antioxid. Redox Signal. 22, 1047–1059 (2015). L. Studer, HSP90-incorporating chaperome networks as biosensor for disease-related 25. W. Lin, U. J. Kang, Characterization of PINK1 processing, stability, and subcellular pathways in patient-specific midbrain dopamine neurons. Nat. Commun. 9, 4345 (2018). localization. J. Neurochem. 106, 464–474 (2008). 44. J. Cox, N. Neuhauser, A. Michalski, R. A. Scheltema, J. V. Olsen, M. Mann, Andromeda: 26. R. Zhou, A. S. Yazdi, P. Menu, J. Tschopp, A role for mitochondria in NLRP3 inflammasome A peptide search engine integrated into the MaxQuant environment. J. Proteome Res. 10, activation. Nature 469, 221–225 (2011). 1794–1805 (2011). Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 27. A. Izzo, N. Mollo, M. Nitti, S. Paladino, G. Calì, R. Genesio, F. Bonfiglio, R. Cicatiello, 45. V. Matyash, G. Liebisch, T. V. Kurzchalia, A. Shevchenko, D. Schwudke, Lipid extraction M. Barbato, V. Sarnataro, A. Conti, L. Nitsch, Mitochondrial dysfunction in down by methyl-tert-butyl ether for high-throughput lipidomics. J. Lipid Res. 49, 1137–1146 syndrome: Molecular mechanisms and therapeutic targets. Mol. Med. 24, 2 (2018). (2008). 46. M. Eldh, J. Lötvall, C. Malmhäll, K. Ekström, Importance of RNA isolation methods 28. A. J. Villar, P. V. Belichenko, A. M. Gillespie, H. M. Kozy, W. C. Mobley, C. J. Epstein, for analysis of exosomal RNA: Evaluation of different methods. Mol. Immunol. 50, 278–286 Identification and characterization of a new Down syndrome model, Ts[Rb(12.1716)]2Cje, (2012). resulting from a spontaneous Robertsonian fusion between T(1716)65Dn and mouse 47. M. J. Alldred, H. M. Chao, S. H. Lee, J. Beilin, B. E. Powers, E. Petkova, B. J. Strupp, chromosome 12. Mamm. Genome 16, 79–90 (2005). S. D. Ginsberg, CA1 pyramidal neuron gene expression mosaics in the Ts65Dn murine 29. G. Kaur, S. A. Gauthier, R. Perez-Gonzalez, M. Pawlik, A. B. Singh, B. Cosby, P. S. Mohan, model of Down syndrome and Alzheimer's disease following maternal choline J. F. Smiley, E. Levy, Cystatin C prevents neuronal loss and behavioral deficits via the supplementation. Hippocampus 28, 251–268 (2018). endosomal pathway in a mouse model of down syndrome. Neurobiol. Dis. 120, 165–173 48. M. J. Alldred, S. H. Lee, E. Petkova, S. D. Ginsberg, Expression profile analysis of vulnerable (2018). CA1 pyramidal neurons in young-middle-aged Ts65Dn mice. J. Comp. Neurol. 523, 61–74 30. C. Vives-Bauza, C. Zhou, Y. Huang, M. Cui, R. L. A. de Vries, J. Kim, J. May, M. A. Tocilescu, (2015). W. Liu, H. S. Ko, J. Magrané, D. J. Moore, V. L. Dawson, R. Grailhe, T. M. Dawson, C. Li, 49. M. J. Alldred, S. H. Lee, E. Petkova, S. D. Ginsberg, Expression profile analysis of K. Tieu, S. Przedborski, PINK1-dependent recruitment of Parkin to mitochondria in hippocampal CA1 pyramidal neurons in aged Ts65Dn mice, a model of Down syndrome mitophagy. Proc. Natl. Acad. Sci. U.S.A. 107, 378–383 (2010). (DS) and Alzheimer's disease (AD). Brain Struct. Funct. 220, 2983–2996 (2015). 31. P. D'Acunzo, F. Strappazzon, I. Caruana, G. Meneghetti, A. D. Rita, L. Simula, G. Weber, 50. M. J. Alldred, K. E. Duff, S. D. Ginsberg, Microarray analysis of CA1 pyramidal neurons F. D. Bufalo, L. D. Valle, S. Campello, F. Locatelli, F. Cecconi, Reversible induction of in a mouse model of tauopathy reveals progressive synaptic dysfunction. Neurobiol. Dis. mitophagy by an optogenetic bimodular system. Nat. Commun. 10, 1533 (2019). 45, 751–762 (2012). 32. W. Kühlbrandt, Structure and function of mitochondrial membrane protein complexes. 51. S. D. Ginsberg, M. J. Alldred, S. E. Counts, A. M. Cataldo, R. L. Neve, Y. Jiang, J. Wuu, BMC Biol. 13, 89 (2015). M. V. Chao, E. J. Mufson, R. A. Nixon, S. Che, Microarray analysis of hippocampal CA1 33. A. Sugiura, G. L. McLelland, E. A. Fon, H. M. McBride, A new pathway for mitochondrial neurons implicates early endosomal dysfunction during Alzheimer's disease progression. quality control: Mitochondrial-derived vesicles. EMBO J. 33, 2142–2156 (2014). Biol. Psychiatry 68, 885–893 (2010). 34. A. Picca, F. Guerra, R. Calvani, C. Bucci, M. R. L. Monaco, A. R. Bentivoglio, F. Landi, 52. Y. Jiang, K. A. Mullaney, C. M. Peterhoff, S. Che, S. D. Schmidt, A. Boyer-Boiteau, R. Bernabei, E. Marzetti, Mitochondrial-derived vesicles as candidate biomarkers in S. D. Ginsberg, A. M. Cataldo, P. M. Mathews, R. A. Nixon, Alzheimer's-related endosome parkinson's disease: Rationale, design and methods of the exosomes in PArkinson disease dysfunction in Down syndrome is A-independent but requires APP and is reversed by (EXPAND) study. Int. J. Mol. Sci. 20, (2019). BACE-1 inhibition. Proc. Natl. Acad. Sci. U.S.A. 107, 1630–1635 (2010). 35. E. A. Shukkur, A. Shimohata, T. Akagi, W. Yu, M. Yamaguchi, M. Murayama, D. Chui, 53. M. J. Alldred, H. M. Chao, S. H. Lee, J. Beilin, B. E. Powers, E. Petkova, B. J. Strupp, T. Takeuchi, K. Amano, K. H. Subramhanya, T. Hashikawa, H. Sago, C. J. Epstein, S. D. Ginsberg, Long-term effects of maternal choline supplementation on CA1 pyramidal A. Takashima, K. Yamakawa, Mitochondrial dysfunction and tau hyperphosphorylation neuron gene expression in the Ts65Dn mouse model of Down syndrome and Alzheimer's in Ts1Cje, a mouse model for Down syndrome. Hum. Mol. Genet. 15, 2752–2762 (2006). disease. FASEB J. 33, 9871–9884 (2019). 36. A. Di Rita, P. D’Acunzo, L. Simula, S. Campello, F. Strappazzon, F. Cecconi, 54. C. A. Schneider, W. S. Rasband, K. W. Eliceiri, NIH Image to ImageJ: 25 years of image AMBRA1-mediated mitophagy counteracts oxidative stress and apoptosis induced by analysis. Nat. Methods 9, 671–675 (2012). neurotoxicity in human neuroblastoma SH-SY5Y cells. Front. Cell. Neurosci. 12, 92 (2018). 55. R. Linares, S. Tan, C. Gounou, A. R. Brisson, Imaging and quantification of extracellular 37. A. Di Rita, A. Peschiaroli, P. D’Acunzo, D. Strobbe, Z. Hu, J. Gruber, M. Nygaard, vesicles by transmission electron microscopy. Methods Mol. Biol. 1545, 43–54 (2017). M. Lambrughi, G. Melino, E. Papaleo, J. Dengjel, S. E. Alaoui, M. Campanella, V. Dötsch, 56. S. E. Calvo, K. R. Clauser, V. K. Mootha, MitoCarta2.0: An updated inventory of mammalian V. V. Rogov, F. Strappazzon, F. Cecconi, HUWE1 E3 ligase promotes PINK1/PARKIN- mitochondrial proteins. Nucleic Acids Res. 44, D1251–D1257 (2016). independent mitophagy by regulating AMBRA1 activation via IKK. Nat. Commun. 9, 57. The UniProt Consortium, UniProt: A worldwide hub of protein knowledge. Nucleic Acids 3755 (2018). Res. 47, D506–D515 (2019). 38. N. Whittle, S. B. Sartori, M. Dierssen, G. Lubec, N. Singewald, Fetal Down syndrome brains 58. G. Korotkevich, V. Sukhov, A. Sergushichev, Fast gene set enrichment analysis. bioRxiv, exhibit aberrant levels of neurotransmitters critical for normal brain development. 060012 (2019). Pediatrics 120, e1465–e1471 (2007). 59. C. McCulloch, S. Searle, J. Neuhaus, Generalized, Linear, and Mixed Models 39. P. J. Thul, L. Åkesson, M. Wiking, D. Mahdessian, A. Geladaki, H. A. Blal, T. Alm, A. Asplund, (John Wiley & Sons, 2011). L. Björk, L. M. Breckels, A. Bäckström, F. Danielsson, L. Fagerberg, J. Fall, L. Gatto, C. Gnann, S. Hober, M. Hjelmare, F. Johansson, S. Lee, C. Lindskog, J. Mulder, C. M. Mulvey, P. Nilsson, P. Oksvold, J. Rockberg, R. Schutten, J. M. Schwenk, Å. Sivertsson, E. Sjöstedt, Acknowledgments: We dedicate this work to the memory of Mitsuo Saito, our colleague and M. Skogs, C. Stadler, D. P. Sullivan, H. Tegel, C. Winsnes, C. Zhang, M. Zwahlen, friend, whose contribution to understanding the role of lipids in neurotransmission was A. Mardinoglu, F. Pontén, K. von Feilitzen, K. S. Lilley, M. Uhlén, E. Lundberg, A subcellular unsurpassed. His insight, knowledge, and passion for science will be missed by many. Mitsuo map of the human proteome. Science 356, eaal3321 (2017). Saito passed away on November 20, 2020. We thank P. M. Mathews and L. Rachmany for 40. M. Uhlén, L. Fagerberg, B. M. Hallström, C. Lindskog, P. Oksvold, A. Mardinoglu, discussions; J. Wegiel for providing human samples; B. Barreto, M. Kurz, and S. Bhamdeo for Å. Sivertsson, C. Kampf, E. Sjöstedt, A. Asplund, I. M. Olsson, K. Edlund, E. Lundberg, technical support; S. Das, S. DeRosa, and N. Patel for animal husbandry: S. H. Lee for statistical D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 17 of 18 SCIENCE ADVANCES | RESEARCH ARTICLE analysis; F. Huang for producing MS data shown in fig. S1; and G. Ferrari for coordinating and paper are present in the paper and/or the Supplementary Materials. Additional data related to managing our laboratory. Funding: This work was supported by the National Institute on this paper may be requested from the authors. Aging (AG017617, AG056732, and AG057517) and the National Institute on Drug Abuse (DA044489) to E.L., and NIH Shared Instrumentation Grant (S10RR027990) to T.A.N. Author Submitted 26 August 2020 contributions: Conceptualization: P.D. and E.L.; methodology: P.D. and R.P.-G.; formal analysis: Accepted 23 December 2020 P.D. and S.C.P.; investigation: P.D., R.P.-G., T.H., H.E-B., Y.K., C.M., M.J.A., and C.N.G.; resources: Published 12 February 2021 M.P.; writing original and revised drafts: P.D. and E.L.; visualization: P.D. and E.L.; supervision: 10.1126/sciadv.abe5085 E.L., Mariko Saito, Mitsuo Saito, S.D.G., and T.A.N.; project administration: E.L. and P.D.; funding acquisition: E.L. and T.A.N. Competing interests: The authors declare that they have no Citation: P. D'Acunzo, R. Pérez-González, Y. Kim, T. Hargash, C. Miller, M. J. Alldred, competing interests. Data and materials availability: The raw MS data generated during this H. Erdjument-Bromage, S. C. Penikalapati, M. Pawlik, M. Saito, M. Saito, S. D. Ginsberg, study are available at MassIVE (UCSD, https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp) T. A. Neubert, C. N. Goulbourne, E. Levy, Mitovesicles are a novel population of extracellular with accession number MSV000085447. All data needed to evaluate the conclusions in the vesicles of mitochondrial origin altered in Down syndrome. Sci. Adv. 7, eabe5085 (2021). Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 D’Acunzo et al., Sci. Adv. 2021; 7 : eabe5085 12 February 2021 18 of 18 Mitovesicles are a novel population of extracellular vesicles of mitochondrial origin altered in Down syndrome Pasquale DAcunzo, Roco Prez-Gonzlez, Yohan Kim, Tal Hargash, Chelsea Miller, Melissa J. Alldred, Hediye Erdjument- Bromage, Sai C. Penikalapati, Monika Pawlik, Mitsuo Saito, Mariko Saito, Stephen D. Ginsberg, Thomas A. Neubert, Chris N. Goulbourne, and Efrat Levy Sci. Adv., 7 (7), eabe5085. DOI: 10.1126/sciadv.abe5085 Downloaded from https://www.science.org at Hospital Sant Pau on March 28, 2023 View the article online https://www.science.org/doi/10.1126/sciadv.abe5085 Permissions https://www.science.org/help/reprints-and-permissions Use of this article is subject to the Terms of service Science Advances (ISSN 2375-2548) is published by the American Association for the Advancement of Science. 1200 New York Avenue NW, Washington, DC 20005. The title Science Advances is a registered trademark of AAAS. Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).