Comparative and Integrated Analysis of Plasma Extracellular Vesicles Isolations Methods in Healthy Volunteers and Patients Following Myocardial Infarction

in Healthy Volunteers and Patients Following Myocardial Infarction 2 Short title: Integrated Analysis of Plasma Extracellular Isolation Methods 3 Daan Paget 1,2, Antonio Checa 3, Benedikt Zöhrer 4,5, Raphael Heilig 6, Mayooran 4 Shanmuganathan1,7, Raman Dhaliwal 8, Errin Johnson 8, Maléne Møller Jørgensen 9,10, Rikke 5 Bæk9, Oxford Acute Myocardial Infarction Study (OxAMI)1,7, Craig E. Wheelock3,5,11, Keith M. 6 Channon1,7, Roman Fischer 6, Daniel C. Anthony 2, Robin P. Choudhury 1,7 and Naveed 7 Akbar1. 8 1 Division of Cardiovascular Medicine, Radc liffe Department of Medicine, University of 9 Oxford, Oxford, United Kingdom. 10 2 Department of Pharmacology, University of Oxford, Oxford, United Kingdom. 11 3 Unit of Integrative Metabolomics, Institute of Environmental Medicine, Karolinska Institute, 12 Stockholm, Sweden. 13 4 Respiratory Medicine Unit, Department of Medicine Solna and Center for Molecular 14 Medicine, Karolinska Institutet, Stockholm, Sweden. 15 5 Department of Respiratory Medicine and Allergy, Karolinska University Hospital, 16 Stockholm, Sweden. 17 6 Target Discovery Institute, Centre for Medicines Discovery, Nuffield Department of 18 Medicine, University of Oxford, Oxford, United Kingdom. 19 7 Acute Vascular Imaging Centre, Radcliffe Department of Medicine, University of 20 Oxford, United Kingdom. 21 8 Sir William Dunn School of Pathology, University of Oxford, Oxford, United Kingdom. 22 9 Department of Clinical Immunology, Aalborg University Hospital, Aalborg, Denmark. 23 10 Department of Clinical Medicine, Aalborg University, Aalborg, Denmark. 24 11 Gunma University Initiative for Advanced Research (GIAR), Gunma University, Showa-25 machi, Maebashi, Gunma, Japan. 26 27 Word Count: 11,258 28 Number of Figures: 8 29 *Corresponding Author: Dr Naveed Akbar 30 E-mail: [email protected] 31 Telephone Number: 01865 234902 32 Address: Division of Cardiovascular Medicine, Radcliffe Department of Medicine, John 33 Radcliffe Hospital, Level 6, West Wing, OX3 9DU, United Kingdom. 34 35 36 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2

37 Plasma extracellular vesicle (EV) number and composition are altered following myocardial 38 infarction (MI), but to properly understand the signi ficance of these changes it is essential to 39 appreciate how the different isolation methods affect EV characteristics, proteome and 40 sphingolipidome. Here, we compared plasma EV isolated from platelet-poor plasma from 41 four healthy donors and six MI patients at presentation and 1-month post-MI using 42 ultracentrifugation, polyethylene glycol precip itation, acoustic trapping, size-exclusion 43 chromatography (SEC) or immunoaffinity c apture. The isolated EV were evaluated by 44 Nanoparticle Tracking Analysis, Western blot, transmission electron microscopy, an EV-45 protein array, untargeted proteomics (LC-MS/MS) and targeted sphingolipidomics (LC-46 MS/MS). The application of the five different plasma EV isolation methods in patients 47 presenting with MI showed that the choice of plasma EV isolation method influenced the 48 ability to distinguish elevations in plasma EV concentration following MI, enrichment of EV-49 cargo (EV-proteins and sphingolipidomics) and associations with the size of the infarct 50 determined by cardiac magnetic resonance imaging 6 months-post-MI. Despite the selection 51 bias imposed by each method, a core of EV associated proteins and lipids was detectable 52 using all approaches. However, this study highlights how each isolation method comes with 53 its own idiosyncrasies and makes the comparison of data acquired by different techniques in 54 clinical studies problematic. 55 56 Key words 57 Ultracentrifugation, size exclusion chromatography, precipitation, acoustic trapping, 58 immunoaffinity capture, plasma, omics, human. 59 60 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 3

61 Plasma extracellular vesicles (EV) are increas ed in number and carry altered protein, lipid 62 and RNA cargo in the peripheral blood in many pathologies [1-6]. Analysis of plasma EV by 63 omics approaches may provide unparalleled insight into multiple disease mechanisms and 64 disease monitoring in patients for personalized medicine. However, it is unclear how different 65 plasma EV isolation methods influence the plas ma EV-profile or the so-called ‘EV-signature’ 66 in patients. 67 68 Current methods for the isolation of heter ogeneous plasma EV include: ultracentrifugation 69 (UC), density ultracentrifugation, field-flow fr actionation, size-exclusion chromatography 70 (SEC), precipitation, acoustic trapping [7 ] and immunoaffinity capture [8]. However, 71 unsurprisingly, laborious protoc ols that yield pure EV from plasma are less well favoured in 72 large cohorts [9, 10] than protocols that are easier, lower cost, and more convenient. These 73 methodological predilections are further impacted by the availability of stored biobank 74 plasma, which often carry contaminating erythrocytes, immune cells, and platelets [11]. 75 Irrespective of pre-storage processing, all plasma is a rich source of lipoproteins 76 (apolipoprotein A and B), albumin, globulins and fibrinogens, which can co-isolate with 77 plasma EV. The proportion of these cell-derived and non-cellular contaminants in the purified 78 sample is method-dependent [12] and they may obscure EV associated cargo [13, 14]. 79 80 Comparative isolation studies for plasma EV often assess EV size and concentration by 81 Nanoparticle Tracking Analysis (NTA) and morphol ogy by transmission electron microscopy 82 (TEM). EV markers are usually evaluated using western blot or flow cytometry [15-18] . 83 These standard analyses are often driven by t he requirements of scientific bodies and 84 consensus statements, which might be consider ed too prescriptive and refractory to change 85 as our understanding of EV biology evolves [14]. However, additional assessments of how 86 isolation methods influence the plasma EV pr eparation have been explored with a range of 87 techniques, including proteomics [16, 18-21], profiling of cytokines [22], RNA integrity [23, 88 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 4 24], lipidomics [25], flow cytometry [15, 26] or a combination of these methods [16]. 89 However, the wide range of plasma EV isolation and characterization techniques has 90 evolved largely in the absence of systematic method characterization comparisons. As a 91 consequence, the absence of an understanding of t he impact of isolation techniques has led 92 to plenty of uncertainty in relation to the interpretation of EV discoveries in clinical samples. 93 94 Here, we used platelet-free plasma from the same healthy volunteers to compare five 95 different plasma EV isolation methods including : UC, precipitation, acoustic trapping, SEC 96 and immunoaffinity capture using tetraspanins CD9, CD63 and CD81 ( Figure 1). We then 97 sought to determine how plasma EV isolation methods influence EV characteristics in a set 98 of clinically well characterised individuals. Ac ute myocardial infarction (MI) is an important 99 pathology; it is also an example of sterile inflammation where plasma EV number and 100 composition are acutely altered [5, 6]. Plasma EV from MI patients at two different time 101 points were isolated using the five different EV isolation methods (Figure 1 ). Integrated 102 unsupervised analysis of all the acquired characterization data from the five different 103

was used to identify the similarities and differences between each method and to 104 highlight how each technique might influence the protein and sphingolipid composition. 105 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 5

106 Healthy Volunteers and Acute Myocardial Infarction Patients. 107 All human investigations were conducted in accordance with the Declaration of Helsinki. The 108 Oxfordshire Research Ethics Committee (Ref: 08/H0603/41 and 11/SC/0397) approved the 109 human clinical protocols. All healthy volunt eers and myocardial Infarction (MI) patients 110 provided informed written consent for inclusion in the study. 111 112 Generation of Platelet Poor Plasma 113 Platelet-poor plasma was generated from healthy volunteers (N=4), patients presenting with 114 MI (N=6) and from the same patients 1-month post-MI. 10-20 mL of whole blood was 115 collected in EDTA (Greiner Bio-One, Stonehouse, United Kingdom) coated tubes and 116 centrifuged at 1,000 x g for 25 minutes. The plasma was collected and centrifuged again for 117 10 minutes at 5,000 x g to produce platelet-p oor plasma. The platelet-poor plasma was 118 stored in 500 µL aliquots at -80 °C for future use. 119 120 Isolation of plasma EV using UC 121 Plasma aliquots were thawed at room temperature and EV were isolated by UC by 122 transferring 500 µL of platelet-poor plasma from healthy volunteers or 100 µL from platelet-123 poor plasma from MI patients to a 13.2 mL QuickSeal tube (Beckman Coulter, California, 124 United States). The tubes were filled with a 16 G hypodermic needles (Microlance, VWR, 125 Pennsylvania, United States) fitted to a 10 mL syringe (VWR, Pennsylvania, United States) 126 and after plasma was injected into the tube, 13 mL of phosphate buffered saline (PBS, 127 Thermo Fischer Scientific, Massachusetts , United States) was added. The tubes were 128 sealed using a soldering iron (Zacro, 60 W) and centrifuged using an Optima MAX-XP 129 ultracentrifuge (Beckman Coulter, California, United States) at 120,000 x g for 120 minutes 130 at 4 °C with a MLA55 fixed-angle rotor (Beckman Coulter, California, United States) [5, 6]. 131 The pelleted plasma EV were resuspended in 100 µL PBS or RIPA buffer (Thermo Fischer 132 Scientific, Massachusetts, United States) for subsequent analysis. 133 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 6 134 Isolation of plasma EV using precipitation 135 Plasma EV were isolated by precipitation by using the Total Exosome Isolation Kit 136 (Invitrogen, Massachusetts, United States). 500 µL of platelet-poor plasma from healthy 137 volunteers or 100 µL of platelet-poor plasma from MI patients was thawed at room 138 temperature. After vortexing, 20-100 μ L of the Exosome Precipitation Reagent (Total 139 Exosome Isolation Kit (for plasma), ThermoFis her Scientific, Massachusetts, United States) 140 was added and mixed by repeated pipetting. Samples were incubated for 30 minutes on ice, 141 the mixture was centrifuged at 10,000 × g for 10 minutes by a Haraeus Fresco 17 benchtop 142 centrifuge (ThermoFisher Scientific, ibid) at room temperature. The supernatant was 143 removed and samples were centrifuged again at 2,000 x g for 2 minutes to remove residual 144 supernatant. The pelleted plasma EV were resuspended in 100 µL PBS or RIPA buffer for 145 downstream analysis by repeated pipetting. 146 147 Isolation of plasma EV using acoustic trapping 148 Acoustic trapping of plasma for t he isolation of EV was achieved by diluting plasma 1:1 with 149 PBS as previously described [18]. Briefly, sa mples were loaded onto a Costar 96-well plate 150 (Corning, New York, United States) and inserted in the AcouSort device (Version 2.0) 151 (AcouSort AB, Lund, Sweden). Acoustic waves were produced by a waveform generator 152 (Keysight 33210A, Keysight, California, United States) with a 9.2 V output and were directed 153 into a borosilicate capillary acoustic trapping unit (AcouSort AB, Lund, Sweden). Following 154 activation of the waveform generator and t he acoustic trapping unit was initialised, 50 μ L of 155 12 μ m polystyrene beads (AcouSort AB, Lund, Sweden) were loaded into the acoustic 156 trapping unit using a syringe pump (Tricontinent C2400 (Tric ontinent, Fürstenfeldbruck, 157 Germany) set to 50 μ L/min. Each run consisted of 100 μ L of 1:1 diluted plasma and were 158 loaded with a syringe pump speed of 20 μ L/min. After acoustic trapping, the samples were 159 washed by aspirating 15 μ L of PBS at 20 μ L/min and dispensing 50 μ L at 20 μ L/min. After 160 the PBS wash, the samples were eluted in 50 μ L of PBS and used for downstream analysis. 161 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 7 162 Isolation of plasma EV using size exclusion chromatography 163 Size exclusion chromatography (SEC) isolation of plasma EV was achieved by using the 164 Exo-spin™ 96 (Cell Guidance Systems, Missouri, United States) [27]. Columns were 165 equilibrated at room temperature for 15 minutes prior use. Afterwards, the columns were 166 washed twice with 250 μ L of PBS. Plasma was thawed at room temperature and 100 μ L of 167 plasma from healthy volunteers or MI patients was loaded into each column. Plasma EV 168 were eluted by adding 200 μ L of PBS to the top of the column eluted under gravity. Plasma 169 EV were collected and stored for subsequent analysis. 170 171 Isolation of plasma EV using immunoaffinity capture 172 Prior to isolation, the plasma was passed through a SEC method as described above. 173 Exosome-Human CD9 Isolation Reagent, Exosome-Human CD63 Isolation/Detection 174 Reagent and Exosome-Human CD81 Isolation Reagent (all Invitrogen, Massachusetts, 175 United States) were used to capture plasma EV. The mixture was created by resuspending 176 each bead solution and mixing by repeated pipetting. Control IgG-isotype (10400C, Thermo 177 Fischer Scientific, Massachusetts, United States) matched beads were conjugated according 178 to the instructions of the Dynabeads Antibody Coupling Kit (Invitrogen, Massachusetts, 179 United States). After SEC pre-isolation, the samples were incubated with 80 μ L of the 180 combined CD9, CD63 and CD81 beads (end concentration 1 x 10 7 / mL, equal quantity of 181 each bead mixture) or equally concentrated IgG control beads at 4 °C for 18 hours under 182 continual rotation by a vertical rotor (Grant Bio, Essex, United Kingdom). After incubation, 183 the beads were pelleted for 5 minutes at room temperature using a Dynal Magnet 184 (Invitrogen, Massachusetts, United States ) and the supernatant was collected for 185 subsequent analysis. Following pelleting, the samples were washed three times with 200 μ L 186 of PBS and magnetic beads were pelleted as described above. Following the washes, the 187 beads were resuspended in RIPA buffer and supernatant was separated by using a 2,000 x 188 g spin and collected for subsequent analysis. 189 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 8 190 191 Nanoparticle Tracking Analysis 192 Plasma EV size distribution and concentration were determined by Nanoparticle Tracking 193 Analysis (NTA) using a Zetaview device (Particle Metrix, Inning am Ammersee, Germany) as 194 previously described [5, 6]. Prior to injection into the sample chamber, samples were diluted 195 in PBS. The Zetaview measured the sample chamber from 11 positions in 2 cycles. The 196 settings were set at sensitivity 80, fa me 30 and shutter speed 100. Silica 100-nm 197 microspheres (Polysciences Inc., Philadelphia, United States) were used to quality check the 198 instrument performance routinely. The particl e concentration per method was calculated as 199 the change ( Δ) compared to the the control sample, in which PBS replaced the plasma 200 sample. 201 202 Protein concentration 203 Protein concentration was determined by bici nchoninic assay (BCA) (Thermo Fischer 204 Scientific, Massachusetts, United States). A standard curve with Bovine Serum Albumin 205 (Thermo Fischer Scientific, Massachusetts, Unit ed States) was used to calculate the protein 206 concentration. Isolated EV were diluted 1:2 or 1:6 with RIPA buffer and needle sonicated by 207 a SonoPuls HD2070 (Bandelin, Berlin, Germany) at 40% power for 10 seconds. 25 μ L of 208 sonicated sample or standard was incubated in duplicate with 175 μ L of a 25:1 ratio between 209 Reagent A and Reagent B (Thermo Fischer Scientific, Massachusetts, United States) and 210 incubated at 37 °C for 30 minutes. After inc ubation, absorbance was measured at 562 nm 211 using a plate reader (FLUOstar Omega plate reader, BMG Labtech, Aylesbury, United 212 Kingdom). 213 214 Transmission Electron Microscopy 215 Transmission electron microscopy (TEM) of the isolated EV was conduc ted as previously 216 described [6]. Briefly, grids (300 mesh Cu carbon film) were glow discharged for 20 seconds 217 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 9 at 15 mA (Leica EM ACE 200). The isolated plasma EV samples were added to the grid for 2 218 minutes, blotted, stained with 2% uranyl acet ate for 20 seconds, blotted and allowed to air 219 dry. Images were acquired on a 120 kV Tecnai 12 TEM (Thermo Fischer Scientific, 220 Massachusetts, United States) equipped with a OneView digital camera (Gatan, California, 221 United States). TEM images of control samples, in which the isolation method was run with 222 PBS in the place of the plasma, were obtained for each method. Immunoaffinity-based 223 isolated EV were fixed with 1.6% glutaraldehyde in PBS for 1 hour at 4 °C. Following 224 fixation, the beads were fixed in 4% agarose with PBS. The agarose was cut into small 225 cubes (1-2 mm3). Sections were collected onto 200 mesh Cu grids and imaged using a 226 Gatan OneView camera with a FEI Tecnai 12 TEM at 120kV. 227 228 Proteomics 229 Isolated plasma EV were processed for proteomics as previously described [28, 29]. In 230 short, the samples were reduced in 5 mM dithioth reitol for 30 minutes at room temperature. 231 Subsequently, the samples were alkylated with 20 mM iodacetamide for 30 minutes and 232 precipitation using chloroform-methanol prec ipitation. Quantified protein groups with /i2≥ 2 233 unique peptides were included in the comparison of proteomes. The LFQ values for each 234 protein group was deducted by the LFQ from cont rol samples per isolation method. Protein 235 abundance was normalised by Log-transformation. Gene Ontology (GO) analysis of the 236 protein groups [30] was conducted by using Genontology.org [30] and the data were 237 extracted on 19-07-2021. P-values were correc ted for false discovery rate (FDR) and 238 significance was set at p<0.05. Fisher exact tests were conducted using R on published 239 databases EVpedia [31], Vesiclepedia [32] and Exocarta [33]. 240 241 EV-Array 242 Isolated plasma EV were analysed by a targeted EV-Array, which has been described 243 previously [6, 34]. In short, a protein micr oarray plate was generated with the following 244 antibodies: CD146 (P1H12), Flotillin-1, TS G101 (Abnova, Taiwan), CD9, CD81 (Ancell 245 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 10 corporation, Minnesota, United States); CD16 (3G8, BD Biosciences, California, United 246 States); Alix (3A9), VEGFR2 (7D4-6; Biolegend, California, United States); CD63 (Bio-Rad, 247 California, United States); ICAM-1 (R6.5, eBioscience, California, United States); Endoglin 248 (LSbio, Washington, United States); Tissue factor (323,514), VCAM-1 (HAE-2Z), 249 Thrombomodulin (501733), CD31 (AF806, R&D Systems, Minnesota, United States), VE-250 Cadherin (AF938, R&D Systems, Minnesota, United States). After blocking with the blocking 251 buffer (50 mM ethanolamine, 100 mM Tris, 0.1% SDS, pH 9.0) for 30 minutes, the wells 252 were emptied, and the plate was dried for 5 hours and sealed. The samples were incubated 253 in the antibody coated microarray plate overnight at 2-8 °C. Following a wash, each well was 254 incubated with a 100 µL of a detection antibody cocktail (biotinylated anti-human-CD9, -255 CD63 and -CD81 (Ancell, Minnesota, United States). After another wash, 100 µL 256 streptavidin-Cy3 Life Technologies, Massac husetts, United States) diluted 1:3,000 was 257 added to each well and incubated for 30 minutes. The plate was scanned using a 258 sciREADER FL2 microarray scanner (Scienion AG, Berlin, Germany), at 535 nm and an 259 exposure time of 2,000 milliseconds. For each protein the control, PBS sample value was 260 subtracted from the result. 261 262 Western blot 263 Isolated plasma EV or controls were lysed using RIPA buffer with protease and phosphatase 264 inhibitors PhosSTOP (Roche, Basel, Switzerland) and cOmplete (Roche, Basel, Switzerland) 265 and were needle sonicated by a SonoPuls HD2070 (Bandelin, Berlin, Germany) at 40% 266 power for 10 seconds as previously described [5, 6]. Following sonication, samples were 267 incubated at 95 °C for 5 minutes to reduce. 8 µg of protein was combined with NuPage LDS 268 sample buffer (4x) agent (Invitrogen, Massachusetts, United States). The samples were 269 loaded onto a 4-12% bis-tris gradient gel (Nu PAGE 4-12% Bis-Tris Protein Gel; 1.5 mm 270 (ThermoFisher Scientific, Massachusetts, United States) with Amersham ECL Full Range 271 ladder (Cytiva Life Sciences, Massachusetts , United States). Separated samples were 272 transferred to a nitrocellulose membrane (Amersham Proton 0.2 μ m, GE Healthcare, Illinois, 273 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 11 United States) and blocked for non-specific bi nding in 5 % skimmed milk powder (Marvel 274 Original, New York, United States) in 0.5% PBS-tween 20 (Sigma-Aldrich, Missouri, United 275 States) for 1 hour. Membranes were incubated with primary antibodies overnight: ALIX 276 (ab117600, Abcam, Cambridge, United Kingdom) (1/1,000 dilution), CD63 (EXOAB-KIT-1, 277 System Biosciences, California, United States) (1/1,000 dilution), ApoB (Ab139401, Abcam, 278 Cambridge, United Kingdom) (1/20,000 dilution), albumin (MAB1455, R&D Systems, 279 Minneapolis, Canada) (1/8,000 dilution), ApoA-I (Mab36641, R&D systems, Minneapolis, 280 Canada) (1/8,000 dilution) and H3 (D1H2, Cell Signalling Technology, Massachusetts, 281 United States) (1/1,000 dilution) 5 % milk in PBS-tween (PBS-T). Membranes were washed 282 three times with PBS-T and incubated with secondary-horse radi sh peroxidase (HRP) 283 conjugated antibodies (1/20,000 α -mouse W402B or 1/50,000 α -rabbit W401B, Promega, 284 Wisconsin, United States) for 1 hour. The membranes were washed once again with PBS-T 285 before incubating them with enhance chemiluminescence substrate (Pierce ECL, 286 ThermoFisher Scientific, Massachusetts, United States) for imaging (Bio-Rad ChemiDoc MP 287 Imaging system, California, United States). 288 289 Sphingolipidomics 290 Sphingolipids were determined as previously described [35]. 25 μ L of each sample were 291 combined with 10 μ L solution containing labelled sphingo lipid internal standards in LC-MS 292 Methanol (Honeywell, North Carolina, United States) was added to each sample. This was 293 followed by the addition of 100 μ L of methanol to each sample. Samples were then vortexed 294 for 30 seconds and sonicated for 15 minutes in a Fisherbrand Ultrasound bath S60 (Fischer-295 Scientific, Massachusetts, United States) with ice. Next, samples were centrifuged at 12,000 296 x g during 15 minutes at 6 /i2 C. Finally, 80 μ L of the supernatant were transferred to an LC-297 MS amber vial (Waters, Wilmslow, United Kingdom) equipped with a 150 μ L insert. Samples 298 were randomized by time point of the individual within an extraction method. Samples were 299 analyzed on an Acquity UPLC coupled to a Xevo TQ-S Mass Spectrometer (Waters, 300 Wilmslow, United Kingdom) as previously described. 301 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 12 302 Bioinformatics and Integrated analysis 303 Integrated analysis was conducted by combining data from the NTA, protein concentration, 304 EV-Array, proteomic and sphingolipidomic analysis. The data was normalized per row and 305 log transformed. Following data normalization, a principal component analysis was 306 conducted in R 4.0.0 [37] using the factoextra [38] and FactoMineR [39] packages. For 307 visualization ggplot2 [40] was used. Each of the principal components were then used to 308 create a heatmap by extracting the values from the eigenvalues for each of the principal 309 components and processing them in pheatmap [41]. The data was clustered by using double 310 hierarchical clustering using the pheatmap R package. 311 312 Data availability 313 All data produced in the present study are available upon reasonable request to the 314 Corresponding Author. 315 316 Statistical analysis 317 Data was plotted as mean with standard deviation. Normality of the data was confirmed 318 using QQ plot and D'Agostino-Pearson normality test. For paired analysis, at the two time 319 points, or for two independent groups, pair ed or unpaired Students T-test were used 320 respectively (GraphPad Prism 9). Correlation analysis was carried out using linear Pearson 321 regression analysis (Graphpad Prism 9). One-way ANOVA with Bonferroni correction post-322 hoc tests were used for analyses with >3 independent groups. The proteomic and lipidomic 323 data was analyzed using a Kruskal-Wallis statistica l test with Bonferroni post-hoc tests. p 324 values <0.05 were considered significant. 325 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 13

326 Plasma EV number and size is influenced by the isolation method 327 The concentration of isolated EV (expre ssed as delta over a matched PBS or and IgG 328 control sample) determined by NTA differed per method (UC 9.5 x 10 9 ± 1.8 x 10 9 EV / mL, 329 precipitation 6.1 x 1011 ± 2.7 x 1011 EV / mL, acoustic trapping 6.4 x 10 9 ± 2.8 x 109 EV / mL, 330 SEC 5.5 x 109 ± 1.9 x 109 EV / mL and immunoaffinity 2.8 x 1010 ± 7.1 x 109 EV / mL, Figure 331 2A). In agreement with previously published stud ies, precipitation yielded significantly higher 332 numbers of particles / mL compared to UC [15, 36] and SEC [15, 37] (both p<0.01). Plasma 333 EV number, isolated by precipitation, were al so higher than the concentration acquired by 334 acoustic trapping or by immunoaffinity capture (both p<0.01) ( Figure 2A ). The size and 335 concentration distribution for each biological replicate exhibited uniformity within each 336 isolation method and the different isolation methods did give a similar size distribution profile 337 overall, which ranged from 15 nm ( Figure 2B). However, the mean size of EV isolated by 338 UC was significantly higher compared to the other methods (UC 143.7 ± 3.4 nm, versus 339 precipitation 94.2 ± 3.9 nm, acoustic trapping 81.2 ± 3.9 and SEC 87.2 ± 1.5, p<0.01 all) 340 (Table 1). 341 342 Plasma EV protein concentration 343 The protein concentration (expressed as the change over a matched PBS or and IgG 344 control) also differed across the EV isolation methods: UC 982 ± 113 µg / mL, precipitation 345 9,478 ± 3,174 µg / mL, acoustic trapping 1,211 ± 141 µg / mL, SEC 382 ± 51 µg / mL and 346 immunoaffinity 33 ± 12 µg / mL ( Figure 2C) . EV generated by precipitation had a 347 significantly higher protein concentration compared to UC, acoustic trapping, SEC or 348 immunoaffinity (all p<0.01, Figure 2C ). The purity of EV isolation can be estimated by 349 calculating a ratio of EV number (EV / mL) to protein concentration (µg / mL) [44]. The EV 350 purity ratio differed between isolation method: UC 9.0 x 10 6 ± 1.4 x 10 6; precipitation 4.7 x 351 108 ± 1.5 x 10 8; acoustic trapping 2.1 x 10 7 ± 7.6 x 10 6; SEC 6.8 x 10 6 ± 4.0 x 10 6 and 352 immunoaffinity capture 3.1 x 10 8 ± 1.0 x 10 8. Both the immunoaffinity and precipitation 353 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 14

achieved a significantly higher EV purity ratio when compared to UC, SEC or 354 acoustic trapping (p<0.01 compared to all methods) (Supplemental Figure 1). 355 356 Plasma EV protein characterisation by Western blot 357 As NTA is unable to distinguish between plasma EV and similarly sized protein aggregates 358 and lipoproteins, we analysed the isolated pl asma EV obtained from each method for a 359 number of markers by western blotting. ALIX and CD63 were included as the EV markers. 360 For the plasma lipoprotein contaminants, we measured ApoB and ApoA-I. Histone H3 for 361 cellular contaminant by western blot and albumin was also included in the set. Each isolation 362

showed the presence of EV markers ALIX and CD63 (Figure 2D). Apolipoproteins 363 (ApoA-I and ApoB) were also present in all isolation methods. All methods were negative for 364 markers of cellular contamination by histone H3. Surprisingly, the IgG isotype control for 365 immunoaffinity capture using CD9, CD63 and CD81 also showed the presence of EV 366 markers ALIX, CD63 and ApoB, albumin, ApoA-I but histone H3 was absent (Figure 2D). 367 368 Plasma EV morphology by TEM 369 To determine the morphology of isolated plasma EV, we undertook TEM for the five different 370 plasma EV isolation methods versus a matched PBS control or an IgG control for 371 immunoaffinity capture beads. TEM analysis show ed EV-like particles for each isolation 372

and an absence of EV-like particles in their respective PBS isolation controls 373 (Figure 2E-H ). Immunoaffinity beads CD9, CD63 and CD81 were embedded in agar and 374 sectioned to visualise the bead surface for EV-like structures versus the IgG control. IgG 375 control showed no EV-like particles present ( Figure 2I). Immunoaffinity capture using CD9, 376 CD63 and CD81 beads showed intact EV-like particles captured on the bead surface 377 (Figure 2J). 378 379 Plasma EV compositional analysis using a high throughput protein EV-Array 380 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 15 The results detailed above provide evidence for the relative success of each method to yield 381 plasma EV and assessment of relative plasma EV purity [39]. However, these techniques do 382 not readily distinguish the abundance of s pecific plasma EV populations carrying, for 383 instance, specific cell-associated markers, or easily allow quantitative assessment of the 384 abundance of contaminating lipoproteins. Thus, we probed the composition of the plasma 385 EV isolated from each method for general EV markers CD9, CD63, CD81, ALIX, TSG101, 386 Flotillin 1, Annexin V and a panel of 18 cell associated markers, which may distinguishes EV 387 from platelets, endothelial cells, immune cells, muscle and lipoprotein contaminants 388 apolipoprotein E (ApoE) and apolipoprotein H (A poH) using a validated high throughput EV-389 protein antibody array (Figure 3 ) [34]. We utilised a matched PBS control for UC, 390 precipitation, acoustic trapping and SEC and an IgG control for immunoaffinity capture. For 391 the EV-associated proteins, CD9 and CD81, they were significantly higher in plasma EV 392 isolated by UC compared to the other methods (p<0.01 for both). Annexin V was significantly 393 higher in the acoustic trapping samples co mpared to SEC and immunoaffinity capture 394 (p<0.01 and p<0.05, respectively). 395 396 Hierarchical clustering of the EV-Array acquired data for the different isolation methods 397 indicates that there are significant method-dependent differences for cell associated EV 398 markers (Figure 3). CD31 was significantly higher in precipitation isolated plasma EV versus 399 acoustic trapping (p<0.05). CD41 was significantly higher in UC isolated plasma EV 400 compared to precipitation isolated EV (p<0.05) and CD16 content was significantly higher in 401 the precipitation isolated plasma EV samples compared to SEC (p<0.01). There were no 402 differences between immunoaffinity capture using CD9, CD63 and CD81 and IgG controls 403 (Figure 3). 404 405 406 Unbiased proteomic analysis of plasma EV 407 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 16 We next determined the proteomic profile of each isolation method using unbiased LC-408 MS/MS versus their respective PBS or IgG cont rols for each method. The proteomic profile 409 following plasma EV isolation showed a significantly higher number of quantified protein 410 groups compared to their respective controls for all isolation methods, except immunoaffinity 411 capture, which displayed similar results to the IgG control beads (UC p<0.01, precipitation 412 p<0.05, acoustic trapping p<0.01 and SEC p<0.01, Figure 4A). The choice of plasma EV 413 isolation method influenced the overall EV-proteome, but nine protein groups were common 414 across all methods ( Supplemental Figure 2 ). These were: immunoglobulin heavy constant 415 gamma 1, alpha-2-macroglobulin, haptoglobin, immunoglobulin heavy constant µ, 416 immunoglobulin kappa constant, serpin family A member 1, albumin, fibrinogen alpha chain 417 and Apo-A1. 418 419 We used hierarchical clustering of the top 30 quantified protein groups across the different 420 plasma EV isolation methods and found a distinct separation between isolation methods. UC 421 and SEC isolated plasma EV clustered together and precipitation, acoustic trapping and 422 immunoaffinity capture formed a distinct separate cluster (Figure 4B). Hierarchical clustering 423 indicates that these differences between UC /SEC and precipitation, acoustic trapping and 424 immunoaffinity capture were driven by an abundance of protein groups with common plasma 425 proteins (ALB, A2M, ApoB, C3, TF, HP, Apo-A1, SERPINA1, CP and ITIH2), fibrinogens 426 (FGG and FGB) and immunoglobulins (IGHG1, IGKC, IGHG3, IGHM, IGK, IGHG2). 427 428 To better understand the nature of the proteomic profile for each isolation method, we 429 conducted unbiased Gene Ontology (GO) pathway analysis of the protein groups associated 430 with each isolation method. All methods showed a significant association with EV pathways: 431 Blood Microparticle GO: 0072562 and Extracellular Exosome GO:0070062 (all p<0.01, 432 Supplemental Figure 3A). However, there was no clear separation between the different 433 isolation methods using this pathway analysis approach. To further scrutinize the EV 434 proteomic profile obtained per plasma EV isolation method we undertook a statistical 435 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 17 comparison using a Fisher’s exact test with published EV-databases EVpedia [31], 436 Vesiclepedia [32] and Exocarta [33]. This determined the similarity between plasma EV 437 proteomic profiles obtained from the five diffe rent isolations methods to those published 438 previously by showing the size of the inters ect. There was a significant overlap between the 439 five different isolation methods and the archived EV databases (p<0.05, all methods, 440 Supplemental Figure 3B ). UC, SEC and immunoaffinity capture showed the greatest 441 similarity with published databases. Whereas precipitation and acoustic trapping showed 442 less similarity. Furthermore, we determined whether our plasma EV acquired proteomic 443 profiles from the five different methods were similar to previously published plasma EV 444 proteomic data for Exospin, SEC, ExoQuick, IZON35, IZON70, Optiprep and Exo-easy [16, 445 17] and found a significant overlap for all five plasma EV isolation methods (p<0.01) 446 (Supplemental Figure 4). 447 448 Targeted sphingolipidomic of plasma EV 449 EV membranes are largely composed of lipi ds [46] and lipoproteins are a predominant 450 contaminant in plasma EV samples [41], whic h are influenced by the choice of isolation 451

[12, 42]. We compared the lipidomic prof ile of the different plasma EV isolations 452

by undertaking targeted sphingolipidom ic analysis. The sphingolipidomic analysis 453 showed a significantly higher number of sphingo lipids compared to the respective controls 454 for all isolation methods except immunoaffinity capture (UC p<0.01, precipitation p<0.05, 455 acoustic trapping p<0.01 and SEC p<0.01, Figure 5A ). However, there were distinct 456 sphingolipidomic differences between the other methods. Eleven of the quantified 457 sphingolipids were common to all methods, which were DhCer(d18:0/24:0), Cer(d18:1/22:0), 458 Cer(d18:1/24:1), Cer(d18:1/24_:0), SM(d18:1/ 16:0), SM(d18:1/18:0), SM(d18:1/24:1), 459 SM(d18:1/24:0), HexCer(d18:1/16:0), He xCer(d18:1/24:1) and LacCer(d18:1/24:1) 460 (Supplemental Figure 5). Hierarchical clustering analysis of the sphingolipidomic profile for 461 each of the isolation methods showed that there are three distinct clusters ( Figure 5B ). 462 Immunoaffinity capture and precipitation formed two separate individual clusters, whilst UC, 463 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 18 acoustic trapping and SEC clustered together. Hi erarchical clustering indicates that these 464 group differences were driven by an abundance of sphingomyelins (16:0, 18:0, 24:0 and 465 24:1), ceramides (22:0, 24:0 and 24:1), hexos ylceramides (24:1) and lactosylceramides 466 (16:0) in the precipitation group and a lack of these sphingolipids in the immunoaffinity 467 capture group. 468 469 Integrated comparison of plasma EV isolation methods 470 These data shows that plasma EV isolation methods have divergent impact on the yield of 471 EV, their purity, the type of EV isolated from plasma that are associated with a particular cell 472 source and the overall proteomic and sphi ngolipidomic profile. To better understand how 473 these individual method associated differences influenced the plasma profile, we undertook 474 integrated analysis of all the different plasma EV isolation methods with all of the acquired 475 data. This included the EV particle concentrati on, protein concentration, EV-protein array 476 data, proteomic and sphingolipidomic data. To condense the multiple different variables, we 477 converted the data into principal component s. Principal component analysis with two 478 components (PCA1 and PCA2) accounted for 47.2% of the variance and showed clear 479 separation between the different isolation methods ( Figure 6A/B). UC, SEC and 480 precipitation form distinct clusters, wher eas acoustic trapping and i mmunoaffinity capture 481 clustered together. Specific principal component analysis indicated that the PCA1 is driven 482 by proteomics acquired data and PCA2 is driven by sphingolipidomic data ( Supplementary 483 Figure 6). These integrated data show that the EV isolation method influences the omic and 484 integrated-based plasma EV-profile, which ma y influence interpretation of EV-acquired data 485 for clinical biomarker discovery and precision diagnostics. 486 487 Plasma EV isolation methods influence the diagnostic potential of plasma EV from patients 488 following MI 489 We next determined whether the choice of plasma EV isolation method impacts the ability to 490 detect changes in EV-profile in a disease state. We obtained plasma at time of presentation 491 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 19 with MI, but prior to percutaneous coronary intervention (PCI), and a matched control plasma 492 sample was obtained at 1-month post-MI from the same patients. The clinical patient 493 characteristics are detailed in Table 2. 494 495 496 497 Plasma EV concentration in MI is influenced by the isolation method 498 To mitigate any potential variability in bio-banked plasma samples we used three technical 499 plasma replicates per patient, per time point and per method. There were significantly more 500 plasma EV at time of presentation with MI versus the 1 month control follow up when plasma 501 EV were isolated by UC, precipitation and acoustic trapping (p<0.05 for all) ( Figure 7A), but 502 not by SEC and immunoaffinity capture (UC presentation: 2.8 x 10 9 EV / mL vs. UC follow-503 up: 2.1 x 10 9 EV / mL; precipitation presentation: 1.5 x 10 12 EV / mL vs. precipitation follow-504 up: 7.9 x 10 11 EV / mL; acoustic trapping presentation: 4.3 x 10 10 EV / mL vs. acoustic 505 trapping follow-up: 1.9 x 1010 EV / mL; SEC presentation: 2.5 x 10 11 EV / mL vs. SEC follow-506 up: 9.2 x 10 10 EV / mL and immunoaffinity presentation: 2.0 x 10 11 EV / mL vs. 507 immunoaffinity capture follow-up: 9.8 x 10 10 EV / mL, Figure 7A). The technical variance 508 between the three independent isolations from each patient, and at each time, point showed 509 that SEC gave significantly less variance compared to UC (10.0 ± 3.1 % vs. 22.7 ± 13.0 %, 510 p<0.01) ( Supplemental Figure 7A ). The mean plasma EV size was similar between the 511 time of presentation and 1 month follow up control for the five different plasma EV isolation 512

( Supplemental Figure 7B ) and the size and concentration distribution profile 513 showed no distinct differences between time points or dependent on the plasma EV isolation 514

(Supplemental Figure 7C). 515 516 The plasma EV isolation method influences plasma EV concentration association with infarct 517 size 518 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 20 We have previously reported that the total c oncentration of plasma EV in the peripheral 519 blood isolated by UC at the time of presentation correlates with the size of myocardial injury 520 and scar, determined by late gadolinium enhanced (LGE) MRI 6 months post-MI [5, 6]. 521 Therefore, we determined whether plasma EV c oncentrations from the five different plasma 522 EV isolation methods influenced the ability to determine this important clinical association. 523 Infarct size at 6-months post-MI significantly correlated with the plasma EV concentration at 524 time of presentation for UC samples (R 2 = 0.89, p=0.02), but not for precipitation, acoustic 525 trapping, SEC or immunoaffinity capture (R 2=0.10 p=0.60, R2=0.16 p=0.50, R 2=0.05 p=0.71 526 and R2=0.10, p=0.78, respectively) (Figure 7B). 527 528 Cell associated plasma EV are not influenced by the isolation method in MI 529 We determined whether the isolation methods influenced cell associated plasma EV-530 markers in MI patients at presentation versus the respective 1 month follow up sample. The 531 high through put EV-Array data was expressed as fold change over the matched follow-up 532 control samples per patient. Plasma EV markers were increased at time of presentation vs. 533 follow-up control ( Figure 7C). However, the pattern of change in plasma EV-markers were 534 not consistent across the di fferent isolation methods ( Figure 7C). Hierarchical clustering of 535 the plasma EV marker response at the time of presentation with MI showed method-536 dependent clustering. Similarly, cell associated markers on plasma EV were differentially 537 enriched following MI ( Figure 7C ). Once again, unbiased clustering showed method-538 dependent groupings. Together these data suggest that changes in plasma-EV protein are 539 influenced by the choice of plasma EV isolation method following MI in isolation and when 540 clustered together. 541 542 Plasma EV sphingolipidomic profiles are altered following MI and influenced by the isolation 543

544 Next, we profiled EV-sphingolipids in the acute phase following MI (calculated as fold over 545 matched follow-up control samples) using the different plasma EV isolation methods. Plasma 546 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 21 EV-sphingolipidomic profiles for samples generat ed by precipitation, acoustic trapping and 547 SEC contained significantly more quantified lipid groups at time of presentation with MI 548 versus samples obtained by UC and immunoaffinity capture (p<0.01, all). We ranked plasma 549 EV-lipid profiles based on their overall abundanc e of lipid groups and found that ceramides 550 were significantly higher in the precipitation isolated plasma EV samples, followed by SEC, 551 acoustic trapping, ultracentrifugation / im munoaffinity capture (p<0.01). Whereas the 552 sphingomyelins were significantly higher in t he precipitation group versus all other isolation 553

(p<0.001) ( Figure 7D). The clustered plasma EV-sphingolipid profile following MI 554 was influenced by the choice of isolation method. Several ceramides and sphingomyelins 555 were significantly higher in the plasma EV isolated by precipitation or acoustic trapping 556 compared to UC (Supplemental Figure 8). In addition, the fold increase of Cer(d18:1/22:0), 557 SM(d18:1/18:1) and sphinganine (d18:0) in plasma EV isolated by precipitation at time of 558 presentation vs. 1-month follow up significantly correlated with the infarct size of the patients 559 at 6-months post-MI (R2 = 0.91 p=0.01, R2 = 0.78 p=0.05 and R2 = 0.78 p=0.05, respectively) 560 (Supplemental Figure 9 ). This result further confirmed that plasma EV isolation methods, 561 using targeted sphingolipidomic methods, are associated with unique method dependent 562 profiles in patients following MI. 563 564 Integrated-omics comparison of plasma EV isolation methods in MI patients 565 Finally, we integrated all the acquired plasma EV data from the MI patients in the acute 566 presentation and follow-up time points. A principal component analysis was constructed to 567 assess if the plasma EV profile at time of presentation differed from follow-up using a 95% 568 confidence level. Integrated analysis show ed that plasma EV profiles were only 569 distinguishable at time of presentation with MI for precipitation and acoustic trapping, but not 570 for those plasma EV isolated by UC, SEC and immunoaffinity capture ( Figure 8 ). These 571 findings indicate that different plasma EV isolation methods influence the diagnostic potential 572 of plasma liberated EV following MI when mult iple datasets are integrated to determine 573 plasma EV profiles for potential panel biomarker discovery. 574 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 22

575 Plasma EV-cargo may provide unparalleled insight into tissue homeostasis and pathological 576 processes to facilitate identification of pati ents for focused therapies, but the methods to 577 isolate plasma EV may impact the EV characteri stics. Here, we found that; (I) the choice of 578 plasma EV isolation method affected the plasma EV concentration, sphingolipid and 579 proteomic profile, (II) but the five different plasma EV isolation methods shared a common 580 EV protein and sphingolipid profile. (III) Plasma EV isolation by immunoaffinity capture using 581 anti-CD9, -CD63 and -CD81 coated antibody beads with the current protocol yields a similar 582 profile to IgG control beads. (IV) The isolation method affected the ability to detect 583 alterations in plasma EV sphingolipids and prot eins in MI patients and (V) their association 584 with infarct size determined by cardiac MRI 6 months post-MI. For example, the plasma EV 585 concentration at time of presentation obtained us ing UC provides prognostic information, but 586 this method is less suitable as a tool to generate EV with higher purity for use in mechanistic 587 studies. 588 589 Previous studies have sought to determine t he influence of plasma EV isolation methods, 590 but have often focused on comparing the net influence on the EV-proteome [3, 16, 17, 43] or 591 RNA profiles [16] after isolation. These approaches can neglect the increased dimensionality 592 of EV components, which carry numerous biologically active molecules such as proteins, 593 lipids, RNA/DNA and metabolites. Furthermore, there is varied assessment of 594 contaminations in EV preparations for use in ‘omics’ studies, which may mask or skew the 595 interpretation of datasets. The approach employed here utilizes the same independent 596 biological replicates across multiple methods for plasma EV isolation and characterization, in 597 conjunction with a control PBS/vehicle sample or an IgG for immunoaffinity capture beads. 598 Here, the repeated analysis of the same plasma samples across multiple methods has 599 enabled individual biological variability and potential plasma EV isolation irregularities to be 600 explored. By using individual and integrated analysis, we showed that each method of 601 plasma EV isolation was internally consistent; in so much that the samples clustered 602 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 23 according to their isolation method in unsupervised hierarchical clustering and principal 603 component analysis of all the acquired data. 604 605 Consistent with previous reports, our pl asma EV-proteome was dependent on the plasma 606 EV isolation method [16, 17]. However, our data shows that there are common plasma 607 proteins, such as albumin and apolipoproteins ApoA-I and ApoB, present in all five isolation 608

tested, with significant similarities to archived EV-databases [31-33] and published 609 studies [16, 17, 31]. In particular, the plasma EV-proteome indicated that UC/SEC cluster 610 together based on the top 30 protein groups quantified, due to higher quantities of 611 fibrinogens and immunoglobulins compared to the other methods. This is, potentially, a 612 methodological constraint of employing untargeted mass spectrometry on plasma EV, which 613 favors the most abundant peptides. 614 615 To better determine the protein profile of known plasma EV markers and cell associated 616 markers, we utilized a high throughput EV-Array [50] following plasma EV isolation from the 617 five different methods. We found that CD9 and CD81 were the most abundant in plasma EV 618 derived by UC versus precipitation and SEC in healthy volunteers. Additionally, we found 619 that cell-associated markers such as CD41 (platelets) and CD16 (immune cells) were 620 enriched in UC and precipitation compared to ot her methods only in healthy volunteers. This 621 high throughput antibody EV-Array circumvents issues of non-EV associated protein and 622 lipid contaminants in plasma EV preparations. However, the five different plasma EV 623 isolation methods used here may have influenced the EV-protein corona. Proteins such as 624 albumin and apolipoproteins (ApoA-I, ApoB, ApoC-III and ApoE) show an association with 625 the surface of EV [45], which may influence antibody mediated binding to the array or 626 subsequent detection of EV-associated proteins in this sandwich ELISA like technique. 627 Protein coronas found on EV are affected by the isolation method [45] and our data shows 628 that the plasma EV-proteome and EV-Array acquired data are also influenced by the 629 isolation method. Indeed, our western blot and proteomic data reported considerable 630 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 24 variation in the amount of albumin across the me thods, which is a predominant protein in the 631 EV protein corona. In particular, precipitat ion isolated preparations was associated with 632 lower albumin, which could be due to the pres ence of PEG and high salt concentrations in 633 the isolation buffer that might be expected to influence albumin associations with the EV 634 protein corona. The temporal dynamics of the EV-protein and possibly EV-lipid corona are 635 poorly understood, and further elucidation of these interactions will be essential to 636 understand the data from these preparations. 637 638 Immunoaffinity based methods, such as magnetic and polystyrene bead conjugations to 639 specific antibodies, such as tetraspanins (CD9, CD63 and CD81), have been heralded as an 640 important advancement in the capture of highly pure EV populations from the plasma. 641 Immunoaffinity based methods could circum vent issues of lipoprotein and protein 642 contamination in plasma EV yields, which ar e common in methods that use EV physical 643 characteristics such as size and density for isolation such as UC and SEC [16, 46]. To the 644 best of our knowledge this is the first study to compare magnetic beads coated with 645 antibodies for tetraspanins (CD9, CD63 and CD81) with a matched IgG control for plasma 646 EV isolation, using detailed integrated analyses. The plasma EV-profile of proteins and lipids 647 for the immunoaffinity beads and matched IgG control was similar by Western blot, EV-648 protein-array, the number of peptides groups in proteomic profiles, GO pathway analysis and 649 targeted sphingolipidomics. TEM imaging of IgG beads showed no visible EV-like particles 650 or membranous structures, indicative of whole or sheared EV-particles, captured to the bead 651 surface. However, we hypothesize that soluble non-EV-associated proteins (such as C3, 652 A2M, ALB and fibrinogens A and B) and lipids (such as ApoB, ApoE and ApoA-I) are 653 interacting with IgG beads, which have become associated with EV-profiles in omics studies 654 and in archived databases [31-33]. Studies using less complicated matrices, like conditioned 655 cell culture media, show a more robust distinction between anti-CD9, CD63 and CD81 beads 656 versus IgG control beads [8, 47], indicating t hat the similarity between tetraspanin and IgG 657 antibody beads here might be matrix specific. Future plasma EV studies using antibody-658 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 25 bead isolation should include a matched IgG cont rol per characterisation when using omic-659 approaches 660 661 Having determined plasma EV isolations met hod similarities and differences on the EV-662 characteristics and cargo in healthy volunteers, we assessed the influence of plasma EV 663 isolation methods in a clinical situation where plasma EV number and composition are 664 altered. We have previously shown that plasma EV are altered following MI to mediate long 665 range signalling and induce the mobilization of splenic-neutrophils, splenic-monocytes and 666 orchestrate their transcriptional programming [5, 6]. The choice of plasma EV isolation 667

determined whether there was a higher concentration of EV immediately following 668 MI compared to a 1-month follow-up control sample from the same patients. The new data 669 here support our previous observations that th e concentration of plasma EV isolated by UC 670 from the time of presentation with MI correlates with the infarct size [5]. However, plasma EV 671 concentrations derived from the other isolat ion methods in the same patients did not 672 associate with infarct size. One possible explanation is the modest sample size utilised in 673 this multi-method comparison (N=6), which is smaller than those utilised by us in previous 674 publications (N=15-22) [5, 6]. 675 676 Plasma EV sphingolipids are predictive of MI [3]. However, the influence of plasma EV 677 isolation methods on the plasma EV-sphingolipidomic profile were not reported. We found a 678 distinct sphingolipid profile between the plasma EV isolation methods. Precipitation-based 679 EV isolation yields a significantly higher lipi d content, principally due to higher proportions of 680 sphingomyelins and ceramides. Sphingomyelins are found in cell membranes and 681 associated with high- and low-density lipoproteins [48, 49]. The high presence of 682 sphingomyelins and ceramide in the plasma EV profile of the precipitation-based method is 683 likely due to the co-isolation of lipoproteins. PEG-based solutions co-isolate lipoproteins [56]. 684 Lipoproteins contain ApoA-I and ApoB, which were higher in plasma EV derived by 685 precipitation-based isolation when compared to the other methods by Western blot. 686 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 26 Lipoproteins can also masquerade as EV-like particles in dynamic light scattering in 687 techniques such as NTA [12]. Precipitation isolated plasma EV had the highest 688 concentration of particles / mL when compar ed to other methods. EV Cer(d18:1/20:0) only 689 showed elevation in precipitation-based is olation compared other methods. Similarly, 690 precipitation isolated EV showed sphinganine(d18:0) correlated with infarct size at 6-months, 691 but this was not the case for the other methods. Analysis of plasma EV-sphingolipid remains 692 challenging due to constraints due to lipoparticl e remnants [57]; however we show here for 693 the first time there are common sphingolipids fo r different plasma EV isolation methods. In 694 particular, plasma EV-protein and EV-sphingolipids clustered uniquely based on the isolation 695 method, which may impact plasma EV di agnostics where a panel of proteins and 696 sphingolipids are used for differentiation of clinical disease or outcome. 697 698 In summary, our data show that the choice of plasma EV isolation method influences the 699 concentration of plasma EV, the EV-proteome and EV-sphingolipid profile in healthy 700 volunteers and MI patients, where methodological differences determined associations with 701 infarct size. Precipitation based plasma EV isol ation gives the highest particle concentration 702 and enriches for more sphingolipids, but co-isolate large quantities of apolipoproteins. In 703 addition, immunoaffinity capture by using antibodies against tetraspanins yields similar EV-704 protein and EV-sphingolipid profiles to IgG control beads. Unbiased integrated analysis 705 shows that plasma EV isolation methods cluster independently, but despite the selection 706 predispositions imposed by each method, a co re of EV associated proteins and lipids was 707 detectable using all the approaches. 708 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 27 Acknowledgments 709 DP acknowledges his funding from the Department of Pharmacology, the Alison Brading 710 Memorial Fund Lady Margaret Hall, the Clarendon Fund provided by the University of Oxford 711 and the Medical Research Council (MR/N013468/1). NA and RC acknowledge support by 712 research grants from the British Heart Foundat ion (BHF) Centre of Research Excellence, 713 Oxford (NA and RC: RE/13/1/30181 and RE/18/3/34214); British Heart Foundation Project 714 Grant (NA and RC: PG/18/53/33895); the Tripartite Immunometabolism Consortium, Novo 715 Nordisk Foundation (RC: NNF15CC0018486); Oxford Biomedical Research Centre (BRC); 716 Nuffield Benefaction for Medicine and the Wellcome Institutional Strategic Support Fund 717 (ISSF) (NA) and a Health Research Bridging Salary Scheme (HRBSS) to N.A. BZ work was 718 supported by funding from the European Union’s Horizon 2020 Research and Innovation 719 Program under Marie Sklodowska-Curie Grant Agreement 812890, ArthritisHeal. The views 720 expressed are those of the author(s) and not necessarily those of the National Health 721 Service, the National Institutes of Health Research or the Department of Health 722 723 Declaration statement 724 The Author(s) declare(s) that there is no conflict of interest. 725 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 28

726 1. de Miguel Pérez, D., et al., Extracellular vesicle-miRNAs as liquid biopsy biomarkers 727 for disease identification and prognosis in metast atic colorectal cancer patients. Sci Rep, 728 2020. 10(1): p. 3974. 729 2. Dong, L., et al., Circulating Long RNAs in Serum Extracellular Vesicles: Their 730 Characterization and Potential Application as Biomarkers for Diagnosis of Colorectal Cancer. 731 Cancer Epidemiol Biomarkers Prev, 2016. 25(7): p. 1158-66. 732 3. Burrello, J., et al., Sphingolipid compositio n of circulating extracellular vesicles after 733 myocardial ischemia. Sci Rep, 2020. 10(1): p. 16182. 734 4. Kalani, M.Y.S., et al., Extracellular microRNAs in blood differentiate between 735 ischaemic and haemorrhagic stroke subtypes. J Extracell Vesicles, 2020. 9(1): p. 1713540. 736 5. Akbar, N., et al., Endothelium-derived extracellular vesicles promote splenic 737 monocyte mobilization in myocardial infarction. JCI Insight, 2017. 2(17). 738 6. Akbar, N., et al., Rapid neutrophil mobili sation by VCAM-1+ endothelial extracellular 739 vesicles. Cardiovasc Res, 2022. 740 7. Evander, M., et al., Non-contact acoustic capture of microparticles from small plasma 741 volumes. Lab Chip, 2015. 15(12): p. 2588-96. 742 8. Kowal, J., et al., Proteomic comparison defines novel markers to characterize 743 heterogeneous populations of extracellular vesicl e subtypes. Proc Natl Acad Sci U S A, 744 2016. 113(8): p. E968-77. 745 9. Onódi, Z., et al., Isolation of High-Purity Extracellular Vesicles by the Combination of 746 Iodixanol Density Gradient Ultracentrifugati on and Bind-Elute Chromatography From Blood 747 Plasma. Front Physiol, 2018. 9: p. 1479. 748 10. Nordin, J.Z., et al., Ultrafiltration with size-exclusion liquid chromatography for high 749 yield isolation of extracellular vesicles preserving intact biophysical and functional properties. 750 Nanomedicine, 2015. 11(4): p. 879-83. 751 11. Geyer, P.E., et al., Plasma Proteome Profiling to detect and avoid sample-related 752 biases in biomarker studies. EMBO Mol Med, 2019. 11(11): p. e10427. 753 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 29 12. Brennan, K., et al., A comparison of methods for the isolation and separation of 754 extracellular vesicles from protein and lipid particles in human serum. Scientific Reports, 755 2020. 10(1): p. 1039. 756 13. Van Deun, J., et al., The impact of dis parate isolation methods for extracellular 757 vesicles on downstream RNA profiling. J Extracell Vesicles, 2014. 3. 758 14. Thery, C., et al., Minimal information for studies of extracellular vesicles 2018 759 (MISEV2018): a position statement of the Internat ional Society for Extracellular Vesicles and 760 update of the MISEV2014 guidelines. J Extracell Vesicles, 2018. 7(1): p. 1535750. 761 15. Tian, Y., et al., Quality and efficiency assessment of six extracellular vesicle isolation 762

by nano-flow cytometry. J Extracell Vesicles, 2020. 9(1): p. 1697028. 763 16. Veerman, R.E., et al., Molecular evaluation of five different isolation methods for 764 extracellular vesicles reveals different clinic al applicability and subcellular origin. Journal of 765 Extracellular Vesicles, 2021. 10(9): p. e12128. 766 17. de Menezes-Neto, A., et al., Size-exclusion chromatography as a stand-alone 767 methodology identifies novel markers in mass spectrometry analyses of plasma-derived 768 vesicles from healthy individuals. J Extracell Vesicles, 2015. 4: p. 27378. 769 18. Gidlöf, O., et al., Proteomic profiling of extracellular vesicles reveals additional 770 diagnostic biomarkers for myocardial infarcti on compared to plasma alone. Sci Rep, 2019. 771 9(1): p. 8991. 772 19. Askeland, A., et al., Mass-Spectrometry Based Proteome Comparison of 773 Extracellular Vesicle Isolation Methods: Comparison of ME-kit, Size-Exclusion 774 Chromatography, and High-Speed Centrifugation. Biomedicines, 2020. 8(8): p. 246. 775 20. Palviainen, M., et al., Extracellular vesicles from human plasma and serum are 776 carriers of extravesicular cargo-Implications for biomarker discovery. PLoS One, 2020. 777 15(8): p. e0236439. 778 21. Cao, F., et al., Proteomics comparison of exosomes from serum and plasma 779 between ultracentrifugation and polymer-based prec ipitation kit methods. Electrophoresis, 780 2019. 40(23-24): p. 3092-3098. 781 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 30 22. Jung, H.H., et al., Cytokine profiling in serum-derived exosomes isolated by different 782 methods. Scientific Reports, 2020. 10(1): p. 14069. 783 23. Stranska, R., et al., Comparison of membrane affinity-based method with size-784 exclusion chromatography for isolation of exosome-like vesicles from human plasma. 785 Journal of Translational Medicine, 2018. 16(1): p. 1. 786 24. Gutiérrez García, G., et al., Analysis of RNA yield in extracellular vesicles isolated by 787 membrane affinity column and differential ul tracentrifugation. PLoS One, 2020. 15(11): p. 788 e0238545. 789 25. Peterka, O., et al., Lipidomic characterization of exosomes isolated from human 790 plasma using various mass spectrometry techniques. Biochim Biophys Acta Mol Cell Biol 791 Lipids, 2020. 1865(5): p. 158634. 792 26. Shtam, T., et al., Evaluation of immune and chemical precipitation methods for 793 plasma exosome isolation. PLoS One, 2020. 15(11): p. e0242732. 794 27. Welton, J.L., et al., Ready-made chromatography columns for extracellular vesicle 795 isolation from plasma. J Extracell Vesicles, 2015. 4: p. 27269. 796 28. Thompson, A.G., et al., CSF extracellular vesicle proteomics demonstrates altered 797 protein homeostasis in amyotrophic lateral sclerosis. Clin Proteomics, 2020. 17: p. 31. 798 29. Subedi, P., et al., Comparison of methods to isolate proteins from extracellular 799 vesicles for mass spectrometry-based proteomic analyses. Anal Biochem, 2019. 584: p. 800 113390. 801 30. Mi, H., et al., PANTHER version 14: more genomes, a new PANTHER GO-slim and 802 improvements in enrichment analysis tools. Nucleic Acids Res, 2019. 47(D1): p. D419-D426. 803 31. Kim, D.K., et al., EVpedia: an integrated database of high-throughput data for 804 systemic analyses of extracellular vesicles. J Extracell Vesicles, 2013. 2. 805 32. Kalra, H., et al., Vesiclepedia: a compendium for extracellular vesicles with 806 continuous community annotation. PLoS Biol, 2012. 10(12): p. e1001450. 807 33. Simpson, R.J., H. Kalra, and S. Mathiv anan, ExoCarta as a resource for exosomal 808 research. J Extracell Vesicles, 2012. 1. 809 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 31 34. Jorgensen, M., et al., Extracellular Vesi cle (EV) Array: microarray capturing of 810 exosomes and other extracellular vesicles for multiplexed phenotyping. J Extracell Vesicles, 811 2013. 2. 812 35. Akawi, N., et al., Fat-Secreted Ceramides Regulate Vascular Redox State and 813 Influence Outcomes in Patients With Cardiovascular Disease. Journal of the American 814 College of Cardiology, 2021. 77(20): p. 2494-2513. 815 36. Serrano-Pertierra, E., et al., Characte rization of Plasma-Derived Extracellular 816 Vesicles Isolated by Different Methods: A Comparison Study. Bioengineering (Basel), 2019. 817 6(1). 818 37. Gámez-Valero, A., et al., Size-Exclusi on Chromatography-based isolation minimally 819 alters Extracellular Vesicles' characteristics compared to precipitating agents. Sci Rep, 2016. 820 6: p. 33641. 821 38. Webber, J. and A. Clayton, How pure are your vesicles? J Extracell Vesicles, 2013. 822 2. 823 39. Théry, C., et al., Minimal information for studies of extracellular vesicles 2018 824 (MISEV2018): a position statement of the Internat ional Society for Extracellular Vesicles and 825 update of the MISEV2014 guidelines. J Extracell Vesicles, 2018. 7(1): p. 1535750. 826 40. Skotland, T., et al., An emerging focus on lipids in extracellular vesicles. Adv Drug 827 Deliv Rev, 2020. 159: p. 308-321. 828 41. Karimi, N., et al., Detailed analysis of t he plasma extracellular vesicle proteome after 829 separation from lipoproteins. Cell Mol Life Sci, 2018. 75(15): p. 2873-2886. 830 42. Yuana, Y., et al., Co-isolation of extracellular vesicles and high-density lipoproteins 831 using density gradient ultracentrifugation. J Extracell Vesicles, 2014. 3. 832 43. Gidlof, O., et al., Proteomic profiling of extracellular vesicles reveals additional 833 diagnostic biomarkers for myocardial infarcti on compared to plasma alone. Sci Rep, 2019. 834 9(1): p. 8991. 835 44. Bæk, R. and M.M. Jørgensen, Multiplexed Phenotyping of Small Extracellular 836 Vesicles Using Protein Microarray (EV Array). Methods Mol Biol, 2017. 1545: p. 117-127. 837 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 32 45. Tóth, E., et al., Formation of a protein corona on the surface of extracellular vesicles 838 in blood plasma. J Extracell Vesicles, 2021. 10(11): p. e12140. 839 46. Tian, B.M., et al., Human platelet lysate supports the formation of robust human 840 periodontal ligament cell sheets. J Tissue Eng Regen Med, 2018. 12(4): p. 961-972. 841 47. Mathieu, M., et al., Specificities of exosome versus small ectosome secretion 842 revealed by live intracellular tracking of CD63 and CD9. Nat Commun, 2021. 12(1): p. 4389. 843 48. Martínez-Beamonte, R., et al., Sphingomye lin in high-density lipoproteins: structural 844 role and biological function. Int J Mol Sci, 2013. 14(4): p. 7716-41. 845 49. Sódar, B.W., et al., Low-density lipoprotei n mimics blood plasma-derived exosomes 846 and microvesicles during isolation and detection. Sci Rep, 2016. 6: p. 24316. 847 50. Izzo, C., F. Grillo, and E. Murador, Improved method for determination of high-848 density-lipoprotein cholesterol I. Isolation of high-density lipoproteins by use of polyethylene 849 glycol 6000. Clin Chem, 1981. 27(3): p. 371-4. 850 51. Simonsen, J.B., What Are We Looking At? Extracellular Vesicles, Lipoproteins, or 851 Both? Circ Res, 2017. 121(8): p. 920-922. 852 853 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 33 Figures 854 Figure 1: A methodological overview. Platelet poor plasma was obtained from healthy 855 volunteers (n=4) and patients presenting with myocardial infarction (MI) (n=6) (and from the 856 same patients 1-month post-MI) and plasma extracellular vesicles (EV) isolated using five 857 different methods: ultracentrifugation (UC), precipitation, acoustic trapping, size exclusion 858 chromatography (SEC) and immunoaffinity capture with a matched vehicle phosphate 859 buffered saline (PBS) or IgG control. The plasma EV were analyzed using Nanoparticle 860 Tracking Analysis, protein concentration, Western blot, transmission electron microscopy, a 861 targeted EV-protein array for EV-markers CD9, CD63, CD81, ALIX, TSG101, flotillin, 862 Annexin V and 18 other cell associated markers, untargeted proteomics (LC-MS/MS) and 863 targeted sphingolipidomics (LC-MS/MS). The data were analyzed in insolation and following 864 integrated hierarchical clustering and principal component analysis. 865 866 Figure 2: Plasma EV characterization using different isolation methods. (A) Total 867 plasma extracellular vesicles (EV) / mL concentrations and (B) size and concentration 868 distribution profiles were obtained by N anoparticle Tracking Analysis (NTA) using 869 ultracentrifugation (UC), precipitation, acous tic trapping, size exclusion chromatography 870 (SEC) and immunoaffinity capture (n=4). Values are presented as a delta compared to a 871 vehicle control or IgG control. Scale is logarithmic. (C) Protein concentration of plasma EV 872 using UC, precipitation, acoustic trapping, SE C and immunoaffinity capture (n=4). Values 873 are presented as a delta compared to a vehicle control or IgG control. (D) Western blot of 874 plasma EV derived from UC, precipitation, acoustic trapping, SEC and immunoaffinity 875 capture versus controls using EV marker s ALIX and CD63, lipoprotein contaminants 876 apolipoprotein B (ApoB), and apolipoprotein A-I (ApoA-I), plasma contaminant albumin and 877 cellular contaminant histone H3. Endothelial cell EV and Peripheral blood mononuclear cells 878 (PBMCs) were used for H3 positive controls. (E-J). Transmission electron microscopy (TEM) 879 images of isolated plasma EV from UC, precipitation, acoustic trapping, SEC and 880 immunoaffinity capture versus controls. Each sub panel contains a zoomed-in image (left 881 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 34 image), an overview image (top right) and a control vehicle image (bottom right). For the 882 immunoaffinity bead capture images, the red arrows indicate EV particles. The scale bar is 883 200 nm for the zoomed images surrounded by a dashed line and 1000 nm for the overview 884 images and J. Values in A and C are group average ± standard deviation (SD). Data are 885 group average ± standard deviation (SD) (n=6). 886 Data was analyzed by one-way ANOVA with post-hoc Bonferroni correction. ***p<0.001. 887 888 Figure 3: Heatmap of plasma EV derived from different isolation methods using the 889 EV- protein-Array. The heatmap contains extracellular vesicles (EV) markers CD9, CD81, 890 CD63, ALIX, TSG101, Flotillin 1 and Annexin V for ultracentrifugation (UC), precipitation, 891 acoustic trapping, size exclusion chromatography (SEC) and immunoaffinity capture, lipid 892 contaminants apolipoprotein H (ApoH) and apolipoprotein E (ApoE) and cell associated 893 markers. Values are presented as a delta com pared to a vehicle control or IgG control and 894 are log normalized (n=4 per isolation method). Data was analyzed by Krusalski-Wallis test 895 with post-hoc Bonferroni correction. *p<0.05, ***p<0.001. 896 897 Figure 4: Proteomic comparison of plasma EV isolation methods. (A) The number of 898 protein groups quantified by unbiased proteomics for plasma extracellular vesicles (EV) 899 derived by: ultracentrifugation (UC), precipit ation, acoustic trapping, size exclusion 900 chromatography (SEC) and immunoaffinity capture, versus control vehicle (PBS) or an IgG 901 control. Protein groups were only included as quantified if they had ≥ 2 unique peptides. 902 (n=3-4). One-way ANOVA with post-hoc Bonferroni correction. Data are group averages ± 903 standard deviation (SD). *p<0.05, ***p<0.001. (B) A heatmap of the top 30 quantified protein 904 groups across all isolation methods. Values from control samples were subtracted to 905 account for background and the values were log normalised. Hierarchical clustering of the 906 isolation methods was conducted using a complete clustering method. (n=3-4). 907 908 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 35 Figure 5: Sphingolipidomic analysis of plasma EV from different isolation methods. 909 (A) Number of sphingolipids quantified in plasma extracellular vesicles (EV) isolated by 910 ultracentrifugation (UC), precipitation, acoustic trapping, size-exclusion chromatography 911 (SEC) and immunoaffinity capture and subjected to targeted sphingolipid analysis versus 912 control vehicle (PBS) or an IgG control. Da ta are group averages ± standard deviation (SD) 913 and were analysed by One-way ANOVA with po st-hoc Bonferroni co rrection. ***p<0.001. 914 (n=4) (B) Heat map of plasma EV sphingolipids. Values from control samples were 915 subtracted to account for background and the values were log normalised. Hierarchical 916 clustering of the isolation methods was conducted using a complete clustering method. (n=3-917 4). Data were analysed by Krusalski-Wallis test with post-hoc Bonferroni correction. 918 919 Figure 6: Principal component analysis of plasma EV characteristics following data 920 integration. (A) A principal component (PC) analysis of plasma extracellular vesicles (EV) 921 isolated by ultracentrifugation (UC), precipitation, acoustic trapping, size-exclusion 922 chromatography (SEC) and immunoaffinity capture: including plasma EV concentration, 923 protein concentration, EV-protein-Array, proteomics and sphingolipidomics. The integrated 924 data was condensed to PC1 and PC2. (B) A heatmap with the various principal components 925 to compare the different isolation plasma EV isolation methods. Hierarchical clustering of the 926 isolation methods was conducted using a complete clustering method. (n=3-4). 927 928 Figure 7: Plasma EV analysis using different isolation methods in patients following 929 presentation with myocardial (MI) infarction compared to samples from the same 930 patients after a 1-month follow up. (A) Comparison of concentration and size-distribution 931 profile by Nanoparticle Tracking Analysis of plas ma extracellular vesicles (EV) from patients 932 following presentation with MI and after a 1-month follow up using: ultracentrifugation (UC), 933 precipitation, acoustic trapping, size exclusion chromatography (SEC) and immunoaffinity 934 capture (n=6 per timepoint). Data are group average ± standard deviation (SD). Paired T-935 test analysis *p<0.05. (B) A Pearson correlation analysis of plasma EV / mL derived from 936 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 36 each isolation method at time of presentation vs. the infarct size determined by cardiac MRI 937 using late gadolinium enhancement 6-months post-infarct (n=5). (C) Heatmap of plasma EV-938 Array analysis. The top section of the heatmap contains the different EV markers CD9, 939 CD81, CD63, ALIX, TSG101, Flotillin 1 and Annexin V. Values are presented as fold over 940 the respective matched follow up control sample. (D) Heatmap showing the plasma EV 941 sphingolipidomic profiles in MI patents at presentation versus a 1-month follow up for UC, 942 precipitation, acoustic trapping, SEC and immuno affinity capture. Values are presented as 943 fold over the respective matched follow up. 944 945 Figure 8: Principal component analysis of integrated plasma EV characterisation at 946 time of presentation with myocardial infarction (MI) vs. 1-moth follow-up in the same 947 patients using different plasma EV isolation methods. A principal component (PC) 948 analysis of plasma EV isolated by ultracentrif ugation (UC), precipitation, acoustic trapping, 949 size-exclusion chromatography (SEC) and immunoaffinity capture from patients presenting 950 with MI versus a 1-month follow up control. Plasma EV characteristics include including 951 plasma EV concentration, protein concentrati on, EV-protein-Array and sphingolipidomics. 952 The integrated data was condensed to PC1 and PC2. (n=5-6). The eclipses indicate the 953 95% confidence interval. 954 955 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 37 Tables 956 Table 1: Nanoparticle Tracking Analysis showing the mean and median size of 957 plasma- EV isolated by ultracentrifugation (UC), pr ecipitation, acoustic trapping, size-958 exclusion chromatography (SEC). Immunoaffinity capture can not be acquired. Data are 959 group averages ± standard deviation (SD) and were analyzed by One-way ANOVA with 960 post-hoc Bonferroni correction. ***p<0.001. (n=4). 961 962 Table 2: Clinical characteristics of the MI patients. Age, sex (M/F), glucose, white blood 963 cells counts, troponin peak, cholesterol, diabetes status, smoker status, infarct size 964 determined by late gadolinium enhancement MRI 6 months post-AMI and left ventricle 965 ejection fraction (LVEF %) 6 moths post-MI . Data are group averages ± standard deviation 966 (SD) (n=6). 967 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 38 Supplemental Figures 968 Supplemental Figure 1: The ratio of plasma extracellular vesicles (EV) to protein 969 concentration per isolation method: ultracentrifugation (UC), precipitation, acoustic 970 trapping, size-exclusion chromatography (SEC ) and immunoaffinity capture. EV are 971 expressed as per mL and protein concentration is expressed µg. Values are expressed as 972 delta over phosphate buffer solution (PBS) control or IgG control. Data are group averages ± 973 standard deviation (SD) (n=4). One-way ANOVA with post-hoc Bonferroni correction. 974 **p<0.01, ***p<0.001. 975 976 Supplemental Figure 2: Venn diagram of qua ntified protein groups in different plasma 977 extracellular vesicle (EV) isolation methods: ultracentrifugation (UC), precipitation, 978 acoustic trapping, size-exclusion chromatogr aphy (SEC) and immunoaffinity capture. 979 Quantified protein groups within a method were pooled after subtraction of phosphate buffer 980 solution (PBS) control or IgG control. Protein groups that had more than 1 repeat within a 981

were included in the Venn diagram. 982 983 Supplemental Figure 3: Gene Ontology analysis of plasma extracellular vesicle (EV) 984 isolation method proteomics: ultracentrifugation (UC), precipitation, acoustic trapping, 985 size-exclusion chromatography (SEC) and immunoaffinity capture. (A) Quantified protein 986 groups were expressed as delta over phosphate buffer solution (PBS) control or IgG control. 987 Quantified protein groups that appeared more than once per method were included in the 988 Gene Ontology (GO) analysis. The GO cellular components associated with each method 989 were ranked based on the false discovery rate (FDR) and the top five were included in the 990 tables. In addition, the GO enrichment scores were plotted against the false discovery rate p-991 values in the scatterplot. (B) The overlap between quantified protein groups per plasma EV 992 isolation method and published EV proteomics databases EVpedia, Exocarta and 993 Vesiclepedia. The table contains the percent age of overlapping protein groups vs. total 994 protein groups per isolation method, with a p-value calculated by fisher-exact test. 995 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 39 996 Supplemental Figure 4: The overlap between the number of quantified protein groups 997 per plasma extracellular vesicle ( EV) isolation method and published plasma EV 998 proteomic datasets. Quantified protein groups within a method were pooled after 999 subtraction of phosphate buffer solution (PBS) control or IgG control. Protein groups that had 1000 more than 1 repeat within a method were included. Plasma EV-proteomes determined by 1001 ultracentrifugation (UC), precipitation, acoustic trapping, size-exclusion chromatography 1002 (SEC) and immunoaffinity capture were compared with published EV proteomic datasets. 1003 The overlap of protein groups between each method were compared and listed out of the 1004 total protein groups quantified. 1005 1006 Supplemental Figure 5: Venn diagram showing the overlap between quantified 1007 sphingolipids in different plasma extracellular vesicles (EV) isolation methods: 1008 ultracentrifugation (UC), precipitation, acoustic trapping, size-exclusion chromatography 1009 (SEC) and immunoaffinity capture. Sphingolipids within a method were pooled after 1010 subtraction of phosphate buffer so lution (PBS) control or IgG control. Sphingolipids that had 1011 more than 1 repeat within a method were included in the Venn diagram. 1012 1013 Supplemental Figure 6: Scree plots of each principal component following the 1014 integrated data analysis of plasma extracellular vesicle (EV) isolated with the different 1015

ultracentrifugation (UC), precipitati on, acoustic trapping, size-exclusion 1016 chromatography (SEC) and immunoaffinity capture. Data from the acquired unbiased 1017 proteomics, targeted sphingolipidomics, EV-Arr ay analysis and concentration/size analysis 1018 was pooled for each isolation method. Principal components were generated with the R-1019 package Factoextra and FactoMineR. Each scree plot indicates the percentage of variance 1020 explained for the principal component and the top 30 drivers for the principal component. 1021 Data was labelled based on the origin; Lipid indicates data from the targeted sphingolipid 1022 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 40 data, EV indicates data from the EV-protein-array and unlabeled was selected for data from 1023 the proteomics. 1024 1025 Supplemental Figure 7: Plasma extracellular vesicle (EV) concentration and size 1026 characteristics after isolation from myocardial infarction (MI) patients at time of 1027 presentation and one month follow up. (A) The variance of plas ma EV concentration 1028 isolated by each method measured by Nanoparti cle Tracking Analysis (NTA). The variance 1029 was calculated by relative standard deviation ov er three different measurements. Values are 1030 expressed as delta over phosphate buffer so lution (PBS) control or IgG control. (B) The 1031 average size of the isolated plasma EV for each method at time of presentation vs. follow up 1032 measured by NTA. Immunoaffinity capture can not be acquired. (C) Size distribution profiles 1033 of the plasma EV from MI patients at time of presentation and one month follow up. 1034 Immunoaffinity capture can not be acquired. Data are group averages ± standard deviation 1035 (SD) (n=12 for A, n=6 for B, n=6 per time point for C). One-way ANOVA with post-hoc 1036 Bonferroni correction. *p<0.05 for A and B. 1037 1038 Supplemental Figure 8: Sphingolipid concentrations of plasma extracellular vesicles 1039 (EV) isolated by different methods: ultracentrifugation (UC), precipitation, acoustic 1040 trapping, size-exclusion chromatography (SEC) and immunoaffinity capture. The fold change 1041 in sphingolipid concentration in plasma EV isolated at time of presentation with myocardial 1042 infarction (MI) versus the concentration in plasma EV isolated 1-month post-MI per isolation 1043 method. Data are group averages ± standard deviation (SD) and were analysed by one-way 1044 ANOVA with post-hoc Bonferroni correction. *p<0.05 and **p<0.01. (n=5-6) 1045 1046 Supplemental Figure 9: Plasma extracellular vesicle (EV) sphingolipid concentration 1047 correlation with infarct size. A Pearson correlation analysis was conducted with the the 1048 fold change in sphingolipid ( A: Cer(d18:0/22:0) ; B: sphinganine (d18:0) ; C: SM(d18:1/18:1) 1049 concentration in plasma EV isolated at time of presentation wi th myocardial infarction (MI) 1050 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint 41 versus the concentration in plas ma EV isolated 1-month post-MI by the precipitation method 1051 versus the infarct size determined by cardiac MRI using late gadolinium enhancement 6-1052 months post-infarct (n=5). 1053 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Figure 1 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint A B C D E Figure 2 Precipitation Widefield UC SEC Acoustic trapping F G H 0 100 200 300 400 500 0 2×105 4×105 6×105 8×105 1×106 UC Size (nm) Extracellular vesicles / mL 0 100 200 300 400 500 0 5×106 1×107 1.5×107 SEC Size (nm) Extracellular vesicles / mL 0 100 200 300 400 500 0 1×106 2×106 3×106 Acoustic Trapping Size (nm) Extracellular vesicles / mL 0 100 200 300 400 500 0 2×107 4×107 6×107 8×107 Precipitation Size (nm) Extracellular vesicles / mL UC Precipitation Acoustic trapping SEC Immunoaffinity 109 1010 1011 1012 1013 Extracellular vesicles / mL ✱✱✱ ✱✱✱ ✱✱✱ ✱✱✱ UC Precipitation Acoustic trapping SEC Immunoaffinity 0 5,000 10,000 15,000 Protein concentration (μg / mL) ✱✱✱ ✱✱✱ ✱✱✱ ✱✱✱ I J Immunoaffinity IgG control All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Figure 3 log 2 1.5 1 0.5 0 -0.5 -1 **CD9 *CD81 CD63 Alix TSG101 Flotillin 1 Annexin V ApoE CD142 VCAM1 ICAM1 *CD31 EpCAM VE-Cadherin Myosin *CD41 CD105 CD146 CD45 VEGFR2 *CD16 CD14 N-Cadherin CD141 ApoH All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Figure 4 UC control UC Precipitation control Precipitation Acoustic trapping control Acoustic trapping SEC control SEC IgG Immunoaffinity 0 50 100 150 Number of protein groups quantified ✱✱✱ ✱ ✱✱✱ ✱✱✱ HP IGKC A2M IGHG1 ApoB C3 IGK IGHG2 ALB FGA FGG FGB ApoA-I TF IGHG3 IGHM C4A ITIH1 C1QA TTR CFB SERPINC1 ITIH2 C5 IGHA1 SERPINA1 FN1 CP PRSS1 IGHM2 SEC I UC I UC II UC III SEC III SEC II Immunoaffinity IV Immunoaffinity II Immunoaffinity III SEC IV Acoustic trapping I Acoustic trapping IV Acoustic trapping II Acoustic trapping III Immunoaffinity I Precipitation I Precipitation IV Precipitation II Precipitation III 8 7 6 5 4

Method UC Precipitation Acoustic trapping SEC Immunoaffinity log A B All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Figure 5 A B UC Control Ultracentrifugation EV Precipitation control Precipitation EV Acoustic trapping control Acoustic trapping EV SEC control SEC IgG Immunoaffinity EV 0 10 20 30 Number of sphingolipids ✱✱✱ ✱✱✱ ✱✱✱ ✱✱✱ SEC I UC I UC II UC III SEC III SEC II Immunoaffinity IV Immunoaffinity II Immunoaffinity III SEC IV Acoustic trapping I Acoustic trapping IV Acoustic trapping II Acoustic trapping III Immunoaffinity I Precipitation I Precipitation IV Precipitation II Precipitation III UC IV HexCer(d18:1/18:0) LacCer(d18:1/24:0) Sphingosine(d18:1) Sphinganine (d18:0) S1P(d18:1) Spa1P(d18:0) DhCer(d18:0/16:0) DhCer(d18:0/24:0) Cer(d18:1/12:0) Cer(d18:1/14:0) Cer(d18:1/16:0) Cer(d18:1/18:1) Cer(d18:1/18:0) Cer(d18:1/20:0) Cer(d18:1/22:0) Cer(d18:1/24:1) Cer(d18:1/24:0) SM(d18:1/12:0) SM(d18:1/16:0) SM(d18:1/18:1) SM(d18:1/18:0) SM(d18:1/24:1) SM(d18:1/24:0) HexCer(d18:1/16:0) HexCer(d18:1/24:1) LacCer(d18:1/16:0) LacCer(d18:1/24:1)

Method UC Precipitation Acoustic trapping SEC Immunoaffinity log 4 3 2 1 0 -1 -2 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Figure 6

UC Precipitation Acoustic trapping SEC Immunoaffinity Precipitation UC SEC Immunoaffinity Acoustic trapping -10 -5 0 5 10 PC1 (30.4%) PC2 (16.8%) -5 0 5 10 A B Principal component 1 Principal component 2 Principal component 3

UC Precipitation Acoustic trapping SEC Immunoaffinity 15 10 5 -5 -10 0 SEC I UC I UC II UC III SEC III SEC II Immunoaffinity IV Immunoaffinity II Immunoaffinity III SEC IV Acoustic trapping I Acoustic trapping IV Acoustic trapping II Acoustic trapping III Immunoaffinity I Precipitation I Precipitation IV Precipitation II Precipitation III UC IV All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint A Figure 7 B UC Precipitation TSG101 Flotillin-1 Annexin V CD63 Alix CD9 CD81 5 4 3 2 1 C Precipitation Acoustic trapping SEC D ApoE CD45 Myosin EpCAM CD146 VE-Cadherin CD105 ApoH CD41 VCAM1 CD16 CD31 ICAM1 CD141 N-Cadherin CD142 CD14 Patient I Patient II Patient III Patient IV Patient V Patient VI TSG101 Flotillin-1 Annexin V CD63 Alix CD9 CD81 5 4 3 2 1 ApoE CD45 Myosin EpCAM CD146 VE-Cadherin CD105 ApoH CD41 VCAM1 CD16 CD31 ICAM1 CD141 N-Cadherin CD142 CD14 Patient I Patient II Patient III Patient IV Patient V Patient VI Acoustic trapping TSG101 Flotillin-1 Annexin V CD63 Alix CD9 CD81 2.5 2 1.5 1 0.5 Patient I Patient II Patient III Patient IV Patient V Patient VI ApoE CD45 Myosin EpCAM CD146 VE-Cadherin CD105 ApoH CD41 VCAM1 CD16 CD31 ICAM1 CD141 N-Cadherin CD142 CD14 SEC TSG101 Flotillin-1 Annexin V CD63 Alix CD9 CD81 5 4 3 2 1 ApoE CD45 Myosin EpCAM CD146 VE-Cadherin CD105 ApoH CD41 VCAM1 CD16 CD31 ICAM1 CD141 N-Cadherin CD142 CD14 Patient I Patient II Patient III Patient IV Patient V Patient VI UC Precipitation Acoustic trapping SEC Immunoaffinity Immunoaffinity TSG101 Flotillin-1 Annexin V CD63 Alix CD9 CD81 ApoE CD45 Myosin EpCAM CD146 VE-Cadherin CD105 ApoH CD41 VCAM1 CD16 CD31 ICAM1 CD141 N-Cadherin CD142 CD14 5 4 3 2 1 Patient I Patient II Patient III Patient IV Patient V Patient VI UC DhCer(d18:1/16:0) Cer(d18:1/14:0) HexCer(d18:1/18:0) LacCer(d18:1/16:0) Cer(d18:1/24:1) HexCer(d18:1/16:0) HexCer(d18:1/24:1) S1P(d18:1) Spinganine (d18:0) Sphingosine (d18:1) HexCer(d18:1/18:1) Cer(d18:1/18:1) SM(d18:1/12:0) SM(d18:1/18:0) SM(d18:1/24:1) SM(d18:1/24:0) Cer(d18:1/16:0) LacCer(d18:1/24:1) Cer (d18:1/22:0) Cer(d18:1/24:0) Cer(d18:1/18:0) SM(d18:1/18:1) Spa1P(d18:0) SM(d18:1/16:0) Cer(d18:1/20:0) Patient II Patient III Patient IV Patient V Patient VI 5 4 3 2 1 0 5 10 15 20 25 0 1×109 2×109 3×109 4×109 Infarct (%) Extracellular vesicles / mL R2 = 0.89 p = 0.02 5 10 15 20 25 -2×10 12 0 2×1012 4×1012 6×1012 Infarct (%) Extracellular vesicles / mL R2 = 0.10 p = 0.60 5 10 15 20 25 -5×10 10 0 5×1010 1×1011 1.5×10 11 2×1011 Infarct (%) Extracellular vesicles / mL R2 = 0.16 p = 0.50 5 10 15 20 25 -1×10 12 -5×10 11 0 5×1011 1×1012 Infarct (%) Extracellular vesicles / mL R2 = 0.05 p = 0.71 5 10 15 20 25 -4×10 11 -2×10 11 0 2×1011 4×1011 6×1011 8×1011 Infarct (%) Extracellular vesicles / mL R2 = 0.03 p = 0.78 UC Precipitation Acoustic trapping SEC Immunoaffinity DhCer(d18:0/16:0) HexCer(d18:1/24:1) Spinganine(d18:0) Sphingosine (d18:0) Cer(d18:1/18:1) Cer(d18:1/16:0) HexCer(d18:1/18:1) Spa1P(d18:0) SM(d18:1/16:0) 10 8 6 4 2 Patient I Patient II Patient III Patient IV Patient V Patient VI Cer(d18:1/24:1) Cer(d18:1/20:0) Cer(d18:1/22:0) LacCer(d18:1/24:1) LacCer(d18:1/16:0) SM(d18:1/24:0) SM(d18:1/24:1) SM(d18:1/18:0) S1P(d18:1) HexCer(d18:1/16:0) Cer(d18:1/24:0) SM(d18:1/18:1) Cer(d18:1/18:0) SM(d18:1/12:0) Cer(d18:1/14:0) HexCer(d18:1/18:0) Sphingosine(d18:0) SM(d18:1/12:0) SM(d18:1/18:0) SM(d18:1/24:1) SM(d18:1/24:0) SM(d18:1/18:1) SM(d18:1/16:0) 8 6 4 2 Patient I Patient II Patient III Patient IV Patient V Patient VI Cer(d18:1/14:0) Spa1P(d18:0) S1P(d18:1) HexCer(d18:1/24:1) Spinganine(d18:0) HexCer(d18:1/18:1) Cer(d18:1/18:1) HexCer(d18:1/16:0) Cer(d18:1/24:0) Cer(d18:1/22:0) Cer(d18:1/18:0) Cer(d18:1/24:1) LacCer(d18:1/24:1) Cer(d18:1/20:0) HexCer(d18:1/18:0) Cer(d18:1/16:0) DhCer(d18:0/16:0) LacCer(d18:1/16:0) 5 4 3 2 1 Sphingosine(d18:0) SM(d18:1/12:0) SM(d18:1/18:0) SM(d18:1/24:1) SM(d18:1/24:0) SM(d18:1/18:1) SM(d18:1/16:0) Patient I Patient II Patient III Patient IV Patient V Patient VI Cer(d18:1/14:0) Spa1P(d18:0) S1P(d18:1) HexCer(d18:1/24:1) Spinganine(d18:0) HexCer(d18:1/18:1) Cer(d18:1/18:1) HexCer(d18:1/16:0) Cer(d18:1/24:0) Cer(d18:1/22:0) Cer(d18:1/18:0) Cer(d18:1/24:1) LacCer(d18:1/24:1) Cer(d18:1/20:0) HexCer(d18:1/18:0) Cer(d18:1/16:0) DhCer(d18:0/16:0) LacCer(d18:1/16:0) Immunoaffinity 8 6 4 2 Cer(d18:1/14:0) HexCer(d18:1/18:0) LacCer(d18:1/16:0) Cer(d18:1/24:1) HexCer(d18:1/16:0) HexCer(d18:1/24:1) S1P(d18:1) Spinganine(d18:0) Sphingosine (d18:0) Cer(d18:1/18:1) SM(d18:1/12:0) SM(d18:1/18:0) SM(d18:1/24:1) SM(d18:1/24:0) Cer(d18:1/16:0) HexCer(d18:1/18:1) Cer(d18:1/22:0) Cer(d18:1/24:0) Cer(d18:1/18:0) SM(d18:1/18:1) Spa1P(d18:0) SM(d18:1/16:0) Cer(d18:1/20:0) Lac Cer(d18:1/24:1) Patient I Patient II Patient III Patient IV Patient V Patient VI DhCer(d18:0/16:0) All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Figure 8 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Table 1 ***Ultracentrifugation Precipitation Acoustic trapping SEC Mean size (nm) Median size (nm) 143.7 ± 3.4 125.0 ± 2.9 94.2 ± 3.9 84.0 ± 3.7 81.2 ± 3.9 70.7 ± 4.6 87.2 ± 1.5 75.0 ± 1.4 - -Immunoaffinity All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint Table 2 Average ± SD Age 67.0 ± 11.4 Sex (male/female) 6/0 Glucose (mmol/L) 5.7 ± 0.5 (N=3) White blood cell count (x10^9) 7.6 ± 2.2 Troponin (peak ng/L) at presentation 358.7 ± 401.2 Cholesterol (mmol/L) 4.8 ± 1.2 Diabetic status (diabetes / non) 0/6 Smoker status (smoker / non) 1/5 Infarct size at 6-months 15 ± 7.8 LVEF at follow up (%) 47 ± 5.0 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 16, 2022. ; https://doi.org/10.1101/2022.04.12.22273619doi: medRxiv preprint