Surface phenotyping and quantitative proteomics reveal differentially enriched proteins of brain-derived extracellular vesicles in Parkinson’s disease

Extracellular vesicles (EVs) are produced by all cell types and are found in all tissues and biofluids. EV proteins, nucleic acids, and lipids are a “nano-snapshot” of the parent cell that may be used for novel diagnostics of various diseases, including neurodegenerative disorders. Currently, diagnosis of the most common neurodegenerative movement disorder, Parkinson’s disease (PD), relies on manifestations of late-stage progression, which may furthermore associate with other neurodegenerative diseases such as progressive supranuclear palsy (PSP). Here, we profiled surface markers and other protein contents of brain-derived extracellular vesicles (bd-EVs) from PD (n= 24), PSP (n=25) and control (n=24). bdEVs displayed tetraspanins and certain microglia, astrocyte, and neuron markers, while quantitative proteomics revealed enrichment of several proteins in PD vs. control and/or PSP, including clathrin heavy chain 1 and 14-3-3 protein gamma. This characterization of EVs in the source tissue provides insights into local dynamics as well as biomarker candidates for investigation in peripheral fluids.


INTRODUCTION 37
Parkinson's disease (PD) is the most common neurodegenerative movement disorder, affecting 38 approximately 1% of individuals over 65 years of age (de Lau & Breteler, 2006). within the manufacturer's recommended particle count range (Arab et al., 2021), and events were 122 recorded for 1 min. Flow rate and side-scatter intensity were converted into particle concentration 123 and size distribution using calibration curves. 124 Transmission electron microscopy (TEM) 125 12 of 73 samples were randomly chosen and imaged by TEM as previously described (Arab et 126 al., 2021). Briefly, 10 µl of each sample was freshly thawed and adsorbed to glow-discharged 127 carbon-coated 400 mesh copper grids by flotation for 2 min. Three consecutive drops of 1× Tris-128 buffered saline were prepared on Parafilm. Grids were washed by moving from one drop to 129 another, with a flotation time of 1 min on each drop. The rinsed grids were then negatively stained 130 with 1% uranyl acetate (UAT) with tylose (1% UAT in deionized water (dIH2O), double-filtered 131 through a 0.22-μm filter). Grids were blotted, then excess UAT was aspirated, leaving a thin layer 132 of stain. Grids were imaged on a Hitachi 7600 TEM operating at 80 kV with an XR80 charge-133 coupled device (8 megapixels, AMT Imaging, Woburn, MA, USA). 134

Western blot (WB) 135
BH, BHC, 2K, 10K, and EV and protein SEC fractions were lysed in 1× radioimmunoprecipitation 136 assay buffer (RIPA) supplemented with protease inhibitor cocktail. A total of 20 μL of lysates were 137 resolved using a 4% to 15% Criterion TGX Stain-Free Precast gel, then transferred onto an 138 Immuno-Blot PVDF membrane. Antibodies to CD81, CD63, and CD9 were used to detect EV 139 membrane markers, and anti-GM130 antibody was used to detect Golgi intracellular 140 contamination. Antibodies were diluted in PBS-T containing 5% blotting-grade blocker (Bio-Rad, 141 #1706404). Membranes were incubated overnight (≈16 h). After several washes in PBS-T, rabbit 142 anti-mouse IgGk BP-HRP and mouse anti-rabbit IgGk BP-HRP secondary antibodies were diluted 143 in blocking buffer, and membranes were incubated for 1 h at room temperature (RT). Pierce™ 144 ECL Western Blotting Substrate (Thermo Fisher, 32106) was applied, and blots were visualized 145 using a Thermo Fisher iBright 1500 imaging system. 146

Single-particle interferometric reflectance imaging sensor (SP-IRIS) 147
EVs were phenotyped with EV-TETRA-C ExoView Tetraspanin kits and an ExoView TMR100 148 scanner (NanoView Biosciences, Boston, MA) according to the manufacturer's instructions and 149 as described previously (Arab et al., 2021). Concentrations as measured by NFCM, were adjusted 150 such that around 1e9 particles from each sample were mixed 1:1 with ExoView incubation buffer 151 (IB). 35µl of this mixture was placed onto the chip and incubated at RT for 16h (no shaking). Chips 152 were washed with IB and incubated 1h at RT with a fluorescently-labeled antibody cocktail of anti-153 human CD81 (JS-81, CF555), CD63 (H5C6, CF647), and CD9 (HI9a, CF488A) at dilutions of 154 1:1200 (v:v) in a 1:1 (v:v) mixture of IB and blocking buffer. All chips were scanned and imaged 155 with the ExoView scanner using both SP-IRIS Single Particle Interferometric Reflectance Imaging 156 Sensor and fluorescence detection. Data were analyzed using NanoViewer 2.8.10 Software. 157

Multiplexed ELISA 158
Prototype S-PLEX® ultrasensitive assays (Meso Scale Diagnostics, Rockville, MD) were used for 159 intact EVs. Five multiplexed assay panels were assembled in this fashion. According to the 160 manufacturer's recommendations, samples were diluted up to 30-fold in "diluent 52," added to the 161 plates, and incubated at RT with continuous shaking. Panel 1, comprising antibodies targeting 162 relatively abundant surface markers, was incubated for 1 hour, while the remaining panels, 163 targeting lower-abundance markers, were incubated for 4 hours to improve sensitivity. EVs 164 captured by each antibody were detected using MSD's S-PLEX® ultrasensitive assay methods 165 with a cocktail of detection antibodies targeting CD63, CD81, and CD9. Assay plates were read 166 with MSD GOLD™ Read buffer B on an MSD® SECTOR instrument. bdEVs from the 73 subjects, 167 as well as additional positive and negative controls, were assayed with each panel. 168

c-g). 233
For all cohort samples, particle counts and size distribution per 100 mg of tissue were measured 234 using NFCM. There were no significant differences in particle counts ( and ELISA surface phenotyping is presented below. 241 bdEV, tetraspanin, and cell marker phenotyping suggest disease-associated differences 242 To assess the relative contribution of each brain cell type to total bdEVs, multiplexed ELISAs were 243 used to assay 36 proteins, including markers of one or more specific cell types: microglia, 244 neurons, astrocytes, and endothelial cells (Fig. 3a). 245 Tetraspanins. CD63 signal was significantly greater in PD compared with the control group. 246 CD81 and CD9 signals were significantly higher for PD and PSP than controls (Fig. 3b), but there 247 were no differences between the disease groups. Several cellular origin surface markers were 248 increased in bdEVs from PD patients compared with control (Fig. 3). CD11b, CD18, and MER proto-oncogene, tyrosine kinase (MERTK) were below the limit of 258 detection (data not shown) ITGB5. 259 Neuronal markers. CD24, nerve growth factor receptor (CD271), and neuronal cell adhesion 260 molecules NRCAM and NCAM were significantly different for disease groups compared with 261 controls, but PD did not differ from PSP. Thy-1 cell surface antigen (CD90/Thy1) and CD166 262 differed between PD and control but not between PSP and control. Only low levels of signal for 263 L1 cell adhesion molecule (L1CAM) were detected, with no significant differences between groups 264 (Fig. 3d). 265 Astrocytic markers. Ganglioside G2 (GD2) and CD44 were detected abundantly compared with 266 ganglioside GD1a (GD1a) and gap junction alpha-1 protein (GJA1, Fig. 3e). These markers show 267 significant differences between the pathological groups compared with control. groups from control (Fig. 3 i). 280 Differentially expressed proteins in brain-derived EVs as revealed by quantitative 281 proteomics 282 Relative quantitative mass spectrometry was performed for bdEVs from five individuals in each 283 group (PD, PSP, and control). A total of 1369 proteins were detected in at least one sample. 284 Amongst these proteins, the gene ontology (GO) term "transport" was the most represented 285 (Supp Fig. 3a). 306 master proteins were successfully tagged for quantitative comparison. After 286 removing duplicates, per calculated ratios PD vs. control, PD vs. PSP, and PSP vs. control, 26 287 proteins were identified as significantly differentially abundant (Fig. 4a, Table 3). In this list, there 288 are 17 proteins less abundant by Log 2-fold-change (Log2 fc) less than or equal to -0.32 and p-289 value <0.05, and 7 proteins more abundant by Log2 fc greater than or equal to 0.32 and p-value 290 <0.05 (Supp Fig. 3b-d). Ankyrin-1 (ANK1) and stomatin (STOM) were more abundant in PSP vs. 291 control and less abundant in PD vs. PSP. 292 Plasma membrane calcium-transporting ATPase 2 (ATP2B2), guanine nucleotide-binding protein 293 G subunit beta-2 (GNB2), and guanine nucleotide-binding protein G subunit gamma-3 (GNG3) 294 differed only between PD vs. control, with lower abundance in PD (Fig. 4b). Five proteins differed 295 only between PD and PSP (Fig. 4c). Clathrin heavy chain 1 (CLTC), F-actin-capping protein 296 subunit alpha-1 (CAPZA1), ADP-ribosylation factor 5 (ARF5), synaptic vesicle glycoprotein 2A 297 (SV2A) were less abundant in PD, whereas cathepsin D (CTSD) was more abundant. 298 Comparison between PSP and control returned three unique proteins more abundant in PSP; 299 namely excitatory amino acid transporter 1 (SLC1A3), erythrocyte membrane protein band 4.2 300 (EPB42), and solute carrier family 2 (SLC2A1) (Fig. 4d). 301 Eleven proteins overlapped between PD vs. control and PD vs. PSP (Fig. 4e). Calcium-302 transporting ATPase 1 (ATP2B1), glial fibrillary acidic protein (GFAP), 14-3-3 GAMMA (YWHAG), 303 sodium/calcium exchanger 2 (SLC8A2), brevican core protein (BCAN), ribosomal protein S9 304 (RPS9), ribosomal protein S18 (RPS18), ribosomal protein L15 (RPL15), actin-related protein 2 305 (ACTR2) were less abundant in PD, whereas Ras-related protein Rap-1b (Rap1b) and hyaluronan 306 and proteoglycan link protein 2 (HAPLN2) were more abundant in PD. Two proteins overlapped 307 between PD vs. control and PSP vs. control (Fig. 4f), and two proteins overlapped between PD 308 vs. PSP and PSP vs. control (Fig. 4g). Although our mass spectrometry protocol was not directed 309 towards membrane proteins, several membrane markers measured by ELISA were also detected 310 by mass spectrometry (Table 4), including astrocyte marker CD44 and neuronal marker NCAM1. 311 EV marker CD81 was detected but not TMT labeled. 312

derived extracellular vesicles from Parkinson's subjects 314
Next, we compared our proteomic data set to the ExoCarta "Top 100" list of putative EV-315 associated proteins. 47 proteins were present in our samples, with 37 successfully labeled with 316 TMT tags and 10 not labeled (Table 5). Some have already been listed as distinguishing bdEV 317 from the tested groups, e.g., YWHAG, STOM, GNB2, RAP1B, and CLTC. More EV markers were 318 identified but without statistically significant differences between tested groups, e.g., CD81, 319 flotillin-1 (Flot-1), RAB proteins (RAB1A, RAB5c), annexins (ANXA1, ANXA11, ANXA2, ANXA4, 320 ANXA5, ANXA6), enolase-1 (Eno1), and 14-3-3 proteins (YWHAE, YWHAH, YWHAQ, YWHAZ). 321 In addition, flotillin-2 (Flot-2) and syntenin-1 EV markers are present in our data set but not in the 322 top-100 ExoCarta list. 323 We followed a similar approach and compared a list of 1952 known PD genes downloaded from 324 DisGeNET with our data set (Supp Fig. 4a). We found 80 to overlap with identified proteins, 325 including 58 that were successfully labeled. Interestingly, several genes overlap with differentially 326 abundant proteins listed above, including YWHAG, RAP1B, HAPLN2, ANK1, and CTSD. 327 Although not all of the 58 labeled proteins were selected after the applied thresholds (Log 2-fold- controls. Our results suggest the possibility that HAPLN2 overexpression is also reflected in EV 367 association. We speculate that this protein might be involved in spread of PD pathology. NRCAM, and CD90 (Fig 3d). Like L1CAM, these proteins are also found on non-neuronal cells 385 and outside the CNS. 386 In conclusion, multiple apparent PD markers were identified in this study, offering opportunities 387 for follow-up. A strength of our study was the use of complementary profiling approaches. While 388 the relatively unbiased proteomics approach was not specifically focused on membrane proteins 389 (Hu et al., 2018), it was complemented by a targeted multiplexed ELISA approach to measure 390 brain cell type-specific markers including membrane-associated proteins. Despite a relatively 391 modest group size, significant differences were detected across categories. These differences 392 are thus even more likely to be detected in larger validation cohorts. In sum, our study identified 393 several brain cell type-specific markers on the surface of bdEVs that may be exploited for 394 detection in the periphery, as well as markers that distinguish disease and control groups.          Flotillin-1 + GAPDH Glyceraldehyde-3-phosphate dehydrogenase +* GDI2 Rab GDP dissociation inhibitor beta -GNAI2 Guanine nucleotide-binding protein G(i) subunit alpha-2 +* GNAS Guanine nucleotide-binding protein G(s) subunit alpha isoforms short -GNB1 Guanine nucleotide-binding protein G(I)/G(S)/G(T) subunit beta-1 -GNB2 Guanine nucleotide-binding protein +* HIST1H4A Histone H4 -