Unless otherwise noted, all chemicals were from Sigma (St. Louis, MO, US). BIOPS biopreservation solution was made using the Veksler et al. formulation^[8].

Donor hearts included in this study were obtained in 2021 through the Institutional Review Board-approved University of Colorado Adult Cardiac Tissue Bank, approval number COMIRB 01-568. Explanted left ventricular tissue from nonfailing donor hearts from adults of all races, genders, and ethnic backgrounds was included, and all participants provided informed consent. Nonfailing left ventricular tissue was procured from donors with normal systolic ventricular function (ejection fraction > 50%) whose hearts were not used for transplant due to lack of an appropriate recipient. Small pieces of left ventricular heart tissue, about 5 mm per side, were immediately placed in BIOPS preservation solution at 4 °C and maintained in this solution until processing^[8]. Mitochondria were isolated from fresh tissue within 12 h of donor-heart excision, and supernatant samples were frozen at -80 °C until processing for EVs occurred once all donor hearts were collected.

Subpopulations of myocardial mitochondria, SSL, located beneath the plasma membrane, and IF, residing between the myofibrils, were isolated using differential centrifugation and trypsin digestion as previously described by Palmer et al. with some modifications for human tissue^[2]. Briefly, freshly explanted cardiac tissue pieces, no larger than 5 mm, were placed in ice-cold BIOPS solution. Tissue was trimmed of fat, placed into CP1 mitochondrial isolation buffer as described in Palmer et al., and minced using scissors^[2]. Low-speed centrifugations were performed at 1,500 × g and high-speed at 4,200 × g. One mg of trypsin was used per gram of tissue. Mitochondria were resuspended in potassium buffer, and protein was quantified by bicinchoninic acid (BCA) assay (Thermo Scientific Pierce, Waltham, MA, US).

Figure 1 shows a schematic of isolating the two major types of heart mitochondria, SSL and IF, and extracting the two associated vesicle types, SHEV and IHEV, respectively. After high-speed centrifugation (4,200 × g) supernatants were collected and pooled and then frozen until further processing. The collected supernatants were then thawed, and precipitation was performed using a ratio of 67 µL ExoQuick (cat # EXOQ-20A1) per 250 µL of supernatant. This solution was incubated on ice for 30 min, then centrifuged at 4 °C for 10 min at 1,000 × g. The EV pellets were resuspended in the same volume of PBS as the starting supernatant and stored at -80 °C until use.

Total EVs isolated from SHEV and IHEV samples were assessed for their sizes and concentrations using the Nanosight nanoparticle tracking analysis (NTA) instrument (NS300, Software Version: NTA 3.2 Dev Build 3.2.16; Malvern Panalytical, Malvern WR14 1XZ, UK). For each measurement, the EVs were freshly diluted 1:100 in ddH₂O before reading, and the reported EV concentration was adjusted accordingly. Average curves for SHEVs and IHEVs are shown in Supplementary Figures 1 and 2, respectively.

To determine the presence of EV-associated tetraspanins in the SHEV and IHEV preparations, Leprechaun analysis was performed. EVs were analyzed using Tetraspanin kits (Unchained Labs: 251-1044) according to the manufacturer’s instructions. In summary, these chips capture small EVs (sEVs) based on the three most commonly reported tetraspanins (CD63, CD81, and CD9) and also contain murine IgG spots which serve as controls. The captured EVs can subsequently be co-stained with fluorescently conjugated antibodies (custom or, in this work, tetraspanins, the latter being provided in kit 251-1044) to assess the number and size of the sEVs of each subtype (CD63, CD81, and CD9) that also contain the other tetraspanins (e.g., how many CD9-captured sEVs also have CD63 and CD81). For each co-staining antibody, fluorescently tagged CD63-647, CD9-488, and CD81-555 were utilized for staining.

EVs were prepared for mass spectrometry using S-Trap^TM microfilters (Protifi, Huntington, NY) according to the manufacturer’s protocol. Digested peptides were cleaned using Pierce^TM C18 Spin Tips (Thermo Scientific) according to the manufacturer’s protocol, dried in a vacuum centrifuge, and resuspended in 0.1% formic acid in mass spectrometry-grade water.

Preparation of heart samples was performed as previously described^[9]. Briefly, lyophilized samples were processed by stepwise extraction with CHAPS and high-salt guanidine hydrochloride, followed by chemical digestion with hydroxylamine hydrochloride in guanidine hydrochloride, generating cellular, soluble extracellular matrix (ECM), and insoluble ECM fractions for each sample, respectively. The protein concentration of each fraction for each sample was measured using the A660 Protein Assay (Pierce^TM). 30 mg of protein resulting from each fraction was subjected to proteolytic digestion using a filter-aided sample preparation protocol with 10 kDa molecular weight cutoff filters (Sartorius Vivacon 500 #VN01H02). Samples were reduced with 5 mM tris(2-carboxyethyl)phosphine, alkylated with 50 mM 2-chloroacetamide, and digested overnight with trypsin (enzyme-to-substrate ratio of 1:100) at 37 °C. Peptides were recovered from the filter using successive washes with 0.2% formic acid.

Digested peptides were loaded onto individual Evotips following the manufacturer’s protocol and separated on an Evosep One chromatography system (Evosep, Odense, Denmark) using a Pepsep column (150 μm inner diameter, 15 cm) packed with ReproSil C18 1.9 μm, 120 Å resin. The system was coupled to a timsTOF Pro mass spectrometer (Bruker Daltonics, Bremen, Germany) via the nano-electrospray ion source (Captive Spray, Bruker Daltonics). The mass spectrometer was operated in Parallel Accumulation-Serial Fragmentation (PASEF) mode. We set the ramp time to 100 ms and acquired 10 PASEF tandem mass spectrometry (MS/MS) scans per acquisition cycle. MS and MS/MS spectra were recorded from m/z 100 to 1,700. The ion mobility was scanned from 0.7 to 1.50 Vs/cm². Precursors for data-dependent acquisition were isolated within ±1 mass-to-charge (m/z) unit and fragmented with an ion mobility-dependent collision energy, which increased linearly from 20 to 59 eV in positive mode. Low-abundance precursor ions with an intensity above a threshold of 500 counts but below a target value of 20,000 counts were repeatedly scheduled and otherwise dynamically excluded for 0.4 min.

Fragmentation spectra were searched against the UniProt human proteome database using the MSFragger-based FragPipe computational platform^[10]. Contaminants and reverse decoys were automatically added to the database. The precursor-ion mass tolerance and fragment-ion mass tolerance were set to 15 and 20 ppm, respectively. EVs were searched with fixed modifications set to carbamidomethylation, and variable modifications set to oxidation. Tissue samples were searched with fixed modifications set as carbamidomethyl, and variable modifications were set as oxidation, hydroxyproline oxidation, Gln->pyro-Glu (N-terminus), deamidation, and acetyl (Protein N-terminus). Two missed tryptic cleavages were allowed, and the protein-level false discovery rate was ≤ 1%.

The 3,758 proteins in the proteomics output were categorized using Claude Science, beta version, and spot-checked manually. After normalizing proteins so each category totaled 100%, Claude first matched the proteins against MitoCarta version 3.0, which identified 757 mitochondrial proteins. It was then instructed to create groupings based on the protein content using UniProt Gene Ontology (GO) cellular-component process keyword rules with a curated override list for cardiac paralog families. Five major categories were used: Mitochondria & Glycolysis, Cytoskeleton & Sarcomere, Vesicle Trafficking, ER Golgi & Secretory, and Other Non-mitochondrial. The Other Non-Mitochondrial category was further investigated for SHEV and IHEV against human protein records in Vesiclepedia 2024: an EVs and extracellular particles repository^[11], to determine levels of known EV proteins contained within this category.

Eight phospholipid classes were quantified using liquid chromatography coupled to electrospray ionization mass spectrometry in an API 4000 mass spectrometer (Sciex, Framingham, MA). 50 µg of protein (BCA protein assay, Pierce) from isolated mitochondria, vesicles, or tissue (homogenized in PBS) from two different hearts was extracted according to previously published methods with 100 nmol tetramyristalcardiolipin as an internal standard (Avanti Polar Lipids, Alabaster, AL, US). These samples were analyzed using a full scan, allowing separation of phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylinositol (PI), phosphatidylserine (PS), sphingomyelin (SM), and cardiolipin (CL). The quantification of these phospholipids was performed using standard curves generated using the SPLASH LipidoMIX as a reference standard, along with tetraoleoylcardiolipin as a reference standard for cardiolipin and tetramyristalcardiolipin as an internal standard (all from Avanti Polar Lipids). A curve was generated for each individual phospholipid, which was then used to quantify its amount [Supplementary Figure 3]. The percentage of each phospholipid was calculated by combining the individual phospholipid species and determining the percentage that each contributed to each separate phospholipid. Phospholipid molecular species are listed by their m/z in the Supplementary Table 1. The sum of all phospholipids was generated by combining the individual species.

The University of Colorado Institutional Animal Care and Use Committee approved all animal experiments (Approval 01480), and they were conducted in accordance with National Institutes of Health guidelines. Neonatal rat ventricular myocytes (NRVMs) were isolated from ventricles of 2-day-old Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA, USA) by enzymatic digestion as described previously^[12]. Briefly, neonatal rats were euthanized using rapid decapitation, and cells were isolated by trypsin digestion of their ventricles. Because NRVMs were freshly isolated for each experiment and maintained for less than 6 days without serial passage, routine mycoplasma testing was not performed. Cells were plated at 50,000 cells/well in 0.1% gelatin-coated (Sigma-Aldrich) 96-well Seahorse plates in growth medium containing MEM-Hanks (Thermo Fisher Scientific, cat. #11575032) and 5% bovine calf serum (Gemini BioProducts, cat. #100-506-500) with a minimum of 5 wells per condition. After 24 h, medium was changed to serum-free MEM-Hanks containing insulin (Sigma-Aldrich, cat. #I1882), transferrin (Sigma-Aldrich, cat. #T1283), bovine serum albumin (Sigma-Aldrich, cat. #A4919-25G), vitamin B12 (Sigma-Aldrich, cat. #V6629), and penicillin G (Sigma-Aldrich, cat. #V7794). All media were buffered with HEPES (Gibco/Thermo Fisher Scientific, cat. #15630080; 20 mM final concentration, pH 7.5). EVs in PBS were added to each well for 48 h before the Seahorse assay. NRVMs were treated with a standardized dose of 5 × 10⁸ particles/well, corresponding to approximately 1 × 10⁴ particles per plated cell. This dosage falls within the range of previously published studies on cardiomyocytes^[13-15]. Following Seahorse analysis, CyQUANT Direct was used to quantify viable-cell abundance and normalize bioenergetic measurements between wells. Untreated wells received an equivalent volume of PBS only, while some wells had no cells in order to test the intrinsic mitochondrial function of the EVs used in this work. Two independent NRVM isolations were used for the functional studies, with each isolation prepared from ventricles pooled from 53 neonatal rats (106 animals total).

The Agilent Cell Mito Stress Test was used to quantify oxygen consumption rates (OCR) and extracellular acidification rates (ECAR). Briefly, a day before the assay, a calibration cartridge was hydrated with sterile Seahorse calibrant solution at 37 °C in a non-CO₂ incubator overnight. On the day of the assay, the cells were gently washed and supplied with warmed Seahorse media [Dulbecco’s Modified Eagle Medium (DMEM) plus substrates at final concentrations of 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose, with the pH adjusted to 7.4 using NaOH] and were incubated for 45 min - 1 h at 37 °C in a non-CO₂ incubator to de-gas. Freshly diluted inhibitors were preloaded into the calibration cartridge in three separate ports to achieve final well concentrations of 1.5 µM oligomycin, 1.0 µM carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (FCCP), and 0.9 µM each of rotenone/antimycin A. After the Seahorse assay, we normalized data between wells using CyQUANT^TM Direct (Thermo Fisher Scientific, Waltham, MA, USA) to quantify live cells per well. Cells were incubated for an hour in CyQUANT dye and imaged with a fluorescent plate reader with excitation set to 485/20 nm and emission 528/20 nm. All parameters of mitochondrial bioenergetics were calculated using Agilent Seahorse Wave Software 2.0.6.

Prism version 11 (GraphPad Software, La Jolla, California, USA) was used for statistical analysis and data visualization. The donor heart was considered the biological unit for analyses of human-derived samples; repeated measurements or assay wells from the same donor-derived preparation were treated as technical replicates and not as independent biological observations. Normality was assessed using the Shapiro-Wilk test for datasets containing more than two biological observations. Given the limited biological sample sizes, results of normality testing were interpreted cautiously. Normality was not assessed for analyses limited to two donor hearts. Treatment effects were analyzed using an ordinary one-way analysis of variance (ANOVA) with Tukey’s test correction for multiple comparisons, with P < 0.05 being significant (comparators for each experiment are listed in the figure legends). Error bars indicate the standard error of the mean (SEM). Due to the limited sample size of the phospholipid analysis (n = 2), we interpreted the P-value cautiously, ensuring a large fold change and low variability to justify the statistical analysis in such a small sample size. For the functional studies, measurements from approximately five technical Seahorse wells treated with each donor-derived EV preparation were averaged to generate a single donor-level value. Treatment effects were calculated relative to the mean untreated response from the same NRVM preparation (Δ = EV-treated donor mean - untreated mean). Donor-level Δ values were compared with the untreated reference (Δ = 0) using two-sided one-sample t-tests. Comparisons between SHEV and IHEV donor-level Δ values were performed using Welch’s unpaired t-test because complete matched SHEV/IHEV preparations were not available for all donor hearts. Data analyzed by Claude Science in Supplementary Figure 4 used Benjamini-Hochberg correction with q < 0.1 reported, since the experimental group (n = 4) was small for this type of analysis.

Demographics for the donors of the hearts used in this study are shown in Table 1.

Table 2 lists the concentration, mean size, and mode of the EV particles from the four hearts. Supplementary Figure 1 shows the size profiles measured by Nanosight for SHEVs, and Figure 2 for IHEVs.

The Leprechaun (Unchained Labs) and Tetraspanin kits (Unchained Labs: 251-1044) were used to assess the presence of EV-associated tetraspanins based on size and expression. Figure 2 shows the size of the particles containing CD9, CD81, and CD63. Supplementary Figure 5 also shows that, of the seven EVs tested, all four SHEV samples and three IHEV samples (from hearts 1-3) contained CD9, CD63, and CD81. These particle sizes are consistent with their classification as small EVs.

Mass spectrometry-based proteomics was performed on samples from EVs (SHEV and IHEV) from four separate hearts, and the SSL and IF mitochondria and whole left ventricular (tissue) homogenate from two of these hearts based on limited sample availability. Figure 3 displays an overlap in the amount of protein detected in these samples with mitochondrial proteins identified using the Mitocarta database^[16]. Because the mitochondrial samples are not highly purified and heart mitochondria are closely associated with endoplasmic reticulum (ER) and structural proteins, their proteomic profiles also contained non-mitochondrial proteins. The SSL and IF mitochondria contained 71.2 ± 2.8 and 82.6 ± 0.3 percent mitochondrial proteins, respectively, whereas whole heart tissue contained 14.8 ± 0.2 percent mitochondrial proteins. The SHEV and IHEV preparations contained even fewer mitochondrial proteins: 5.5 ± 0.9 and 3.8 ± 0.3 percent, respectively. The hierarchical clustering of these five cellular fractions based on protein content for these samples is shown in Figure 4. As expected, the two mitochondrial subtypes cluster together, with SHEV and IHEV more closely related to mitochondria than to whole tissue.

To determine the functional profile of proteins in these vesicles, we analyzed the proteins contained within the different cellular fractions [Figure 5]. These proteins were categorized into four different functional groups plus an “other” category for proteins that did not fit into a discrete functional category. The five categories are Mitochondrial and Glycolytic (including all mitochondrial proteins); Sarcomere and Cytoskeleton; ER, Golgi, and Secretory; Vesicle Trafficking; and Other Non-Mitochondrial. The top five proteins in each category for SHEV and IHEV are listed in descending order in Table 3.

Phospholipids, especially CL, are essential for proper mitochondrial function. Furthermore, since CL, unique among phospholipids, is found only in mitochondria, its presence serves as a marker of mitochondrial-derived membranes^[17]. Therefore, we determined the profile of 8 different phospholipids, including CL, from 2 of the hearts for which we had EV, mitochondria, and whole homogenate samples. The absolute amounts of phospholipids are shown in Figure 6, and relative amounts are expressed as a percentage of the total phospholipid content in Figure 7. IHEVs had a much greater phospholipid content (expressed per mg of protein) than SHEVs, often higher even than mitochondrial samples [Figure 6A-I]. The relative amounts of different phospholipids in the two vesicle types were similar, with a higher percentage of SM than in mitochondria or tissue and higher PA and PG than in tissue. SHEVs had a higher percentage of PE than SSL mitochondria and a lower percentage of CL [Figure 7]. CL molecular species are defined as having a unique number of carbons and double bonds on their four fatty acyl side chains. CL molecular species are often altered in response to stress and disease, but it is unknown whether EVs would be enriched for certain CL species. Figure 8 shows the percent distribution of the nine most abundant CL species in mitochondria, EVs, and tissue. Most species were not significantly different between EVs and mitochondria or tissue, except for two higher-molecular-weight species [Figure 8G and H]. These data indicate that there is a conversion between the CL species having side chains (carbons: double bonds) of 76:12 to 74:9, resulting in a loss of 2 carbons and 3 double bonds on CL.

To determine whether SHEVs or IHEVs affect the function of cardiomyocyte mitochondria or have an effect on their own, a standardized dose of 5 × 10⁸ EV particles was added to primary NRVMs (1 × 10⁴ particles per plated cell) for 48 h, a standard length of time in our laboratory to produce an effect on mitochondria, and the mitochondrial function of NRVMs was assessed using the Seahorse Bioanalyzer. SHEV treatment increased ATP production and basal respiration relative to untreated NRVMs, whereas IHEV treatment decreased maximal respiration [Figure 9]. SHEV- and IHEV-associated responses also differed significantly in ATP production, basal respiration, and maximal respiration. Although maximal respiration tended to increase following SHEV treatment, this change did not reach statistical significance. Importantly, SHEV and IHEV preparations added to Seahorse wells in the absence of NRVMs exhibited no detectable oxygen consumption, indicating that the EV preparations did not contribute measurable intrinsic respiratory activity under these assay conditions [Supplementary Figure 6]. The principal bioenergetic responses were reproduced in a second independent NRVM isolation using the same donor-derived EV preparations [Supplementary Figure 7]. CyQUANT measurements obtained following 48 h EV treatment showed no significant reduction in viable-cell abundance in either SHEV- or IHEV-treated NRVMs compared with untreated controls [Supplementary Figure 8].

The primary limitation of this study is the limited availability of fresh nonfailing human donor hearts, which restricted sample size for several analyses, particularly the proteomic and lipidomic comparisons. Accordingly, these findings should be interpreted as descriptive, establishing regionally distinct EV populations in the human heart and their associated molecular and functional differences. Larger studies will be required to confirm these observations across a broader donor population. In addition, tissue samples were obtained from the free wall of the left ventricle after removal of epicardial fat. Because intact ventricular tissue was processed rather than anatomically separated myocardial layers or purified cell populations, we cannot determine the relative contributions of epicardial vs. endocardial tissue or definitively assign the cellular origin of the EVs. Finally, EVs were enriched using a precipitation-based separation method to maximize recovery from the limited amount of available human tissue. While this approach enabled comprehensive downstream molecular and functional analyses, it may also enrich non-EV particles or other extracellular components. Future studies using orthogonal isolation methods, together with additional characterization and diseased tissue, will further refine these observations. Despite these limitations, the molecular profiles and functional responses were highly consistent across independently derived donor samples, supporting the reproducibility of the observed regional differences.

This study identifies regionally distinct EV populations in the nonfailing human heart that differ in protein and phospholipid composition and in their effects on cardiomyocyte mitochondrial bioenergetics. SHEVs enhanced cardiomyocyte bioenergetic function, whereas IHEVs reduced maximal respiratory capacity, even though neither EV population showed intrinsic respiratory activity in the absence of cardiomyocytes. Together, these findings support a previously unrecognized role for region-specific EV-mediated communication in the human heart and provide a foundation for determining how these EV populations and their functions are altered in cardiac disease.