Platelet‐derived extracellular vesicles are increased in sera of Alzheimer's disease patients, as revealed by Tim4‐based assays

Alzheimer's disease (AD) is the most common form of dementia, characterized by the accumulation of β‐amyloid plaques and the formation of neurofibrillary tangles. Extracellular vesicles (EVs) are small vesicles surrounded by a lipid bilayer membrane, which may be involved in the progression of AD. Glycans are essential building blocks of EVs, and we hypothesized that EV glycans may reflect pathological conditions of various diseases. Here, we performed glycan profiling of EVs prepared from sera of three AD patients (APs) compared to three healthy donors (HDs) using lectin microarray. Distinct glycan profiles were observed. Mannose‐binding lectins exhibited significantly higher signals for AP‐derived EVs than HD‐derived EVs. Lectin blotting using mannose‐binding lectin (rPALa) showed a single protein band at ~ 80 kDa exclusively in AP‐derived EVs. LC‐MS/MS analysis identified a protein band precipitated by rPALa as CD61, a marker of platelet‐derived exosomes (P‐Exo). Sandwich assays using Tim4 with specificity for phosphatidylserine on EVs and antibodies against P‐Exo markers (CD61, CD41, CD63, and CD9) revealed that P‐Exo is significantly elevated in sera of APs (n = 16) relative to age‐ and sex‐matched HDs (n = 16). Tim4‐αCD63 showed the highest value for the area under the curve (0.957) for discriminating APs from HDs, which should lead to a better understanding of AD pathology and may facilitate the development of a novel diagnostic method for AD.

Alzheimer's disease (AD) is the most common form of dementia, characterized by the accumulation of b-amyloid plaques and the formation of neurofibrillary tangles. Extracellular vesicles (EVs) are small vesicles surrounded by a lipid bilayer membrane, which may be involved in the progression of AD. Glycans are essential building blocks of EVs, and we hypothesized that EV glycans may reflect pathological conditions of various diseases. Here, we performed glycan profiling of EVs prepared from sera of three AD patients (APs) compared to three healthy donors (HDs) using lectin microarray. Distinct glycan profiles were observed. Mannose-binding lectins exhibited significantly higher signals for AP-derived EVs than HDderived EVs. Lectin blotting using mannose-binding lectin (rPALa) showed a single protein band at~80 kDa exclusively in AP-derived EVs. LC-MS/ MS analysis identified a protein band precipitated by rPALa as CD61, a marker of platelet-derived exosomes (P-Exo). Sandwich assays using Tim4 with specificity for phosphatidylserine on EVs and antibodies against P-Exo markers (CD61, CD41, CD63, and CD9) revealed that P-Exo is significantly elevated in sera of APs (n = 16) relative to age-and sex-matched HDs (n = 16). Tim4-aCD63 showed the highest value for the area under the curve (0.957) for discriminating APs from HDs, which should lead to a better understanding of AD pathology and may facilitate the development of a novel diagnostic method for AD.
Alzheimer's disease (AD), the most prevalent cause of dementia, is a growing global health concern with enormous implications for individuals and society [1]. AD is a chronic progressive neurodegenerative disorder which is characterized by aggregation of b-amyloid (Ab) plaques and formation of neurofibrillary tangles (NFTs), derived from hyperphosphorylated tau proteins. These factors lead to increasing neural dysfunction. AD is most common cause of dementia in the elderly. Up to 26.6 million people worldwide suffer from AD, and its prevalence is estimated to quadruple by 2050 [2]. Extracellular vesicles (EVs) are small vesicles surrounded by a lipid bilayer membrane. EVs include exosomes of endocytic origin, microvesicles generated by plasma membrane budding, and apoptotic bodies [3]. EVs are important mediators of intercellular communication, as vehicles for proteins, lipids, and nucleic acids during transmission of biological signals between cells [4]. EVs may be involved in progression of AD and nerve cell injury [5]. In contrast, EVs might act on Ab to reduce injury in the central nervous system [6]. MicroRNA in EVs is also attractive targets as biomarkers for diagnosis of AD [7]. Biomarkers are necessary for improving diagnostic sensitivity and specificity, and for monitoring biological characteristics of AD before manifestation clinical symptoms.
Glycans are also essential building blocks of EVs. We previously analyzed EVs derived from human induced pluripotent stem cells (hiPSCs) for comparison with EVs from non-hiPSCs using a glycan profiling technology, called high-density lectin microarray [8]. EVs derived from hiPSCs retain the characteristic glycan signature of hiPSCs. Thus, we hypothesized that glycans of EVs could reflect pathological conditions of various diseases.
In this study, we initially analyzed glycans of EVs isolated from sera of three AD patients (APs) and compared them with EVs from three healthy donors (HDs) using high-density lectin microarray [9,10]. APderived EVs showed distinct glycan profiles from HDderived EVs. Mannose-binding lectins showed higher binding to AP-derived EVs than to HD-derived EVs. A single protein band at~80 kDa was exclusively detected by mannose-binding lectin (rPALa) in APderived EVs. The rPALa-positive protein band was identified as integrin b3, also known as CD61, that is specifically expressed on platelets. Western blotting indicated that aCD61 showed stronger binding to APderived EVs than HD-derived EVs. Further, antibodies against exosome markers (CD9 and CD63) also showed stronger binding in western blotting to APderived EVs than HD-derived EVs. Quantitative analysis using Tim4 with specificity for phosphatidylserine (PS) on EVs and antibodies against markers of platelet-derived exosomes (P-Exo) demonstrated that P-Exo is increased in the sera of APs. Overall, we demonstrate for the first time that P-Exo is increased in sera of APs, which should lead to a better understanding of the pathology of the disease and may allow the development of a novel diagnostic marker for AD.

Study population
Sera from 29 HDs and 20 APs were obtained from BioIVT, which are collected following the written informed consent (Westbury, NY, USA). Participant's demographic data are presented in Table S1. In the quantitative analysis using Tim4-based sandwich assays, 29 HDs and 20 APs populations were examined as whole cohort. In addition, to correct the difference in age and sex between two groups, we performed 1 : 1 propensity score-matching (PSM) analysis. After PSM, 16 subjects from each group were selected and examined as PSM cohort. All experiments were designed and performed in accordance with the standards set by the Declaration of Helsinki and guidelines of the National Institute of Advanced Industrial Science and Technology (AIST) ethics committee assigning authors.

Preparation of EVs
Extracellular vesicles were prepared from serum using MagCapture TM Exosome Isolation Kit PS (Wako Pure Chemical Industries, Ltd., Osaka, Japan) following the manufacturer's instructions [11]. Serum (1 mL) was pre-cleaned by sequential centrifugation at 1200 g for 20 min and 10 000 g for 30 min to remove cells, cellular debris, and large EVs. Supernatants were incubated with Tim4-immobilized magnetic beads and 1/500 volume of exosome binding enhancer at 4°C overnight. Subsequently, magnetic beads were washed with washing buffer (1 mL, supplemented with 1/500 volume of exosome binding enhancer) three times. EVs were then eluted with exosome elution buffer (50 µL). Elution step was repeated twice, and totally, 100 µL of EV samples was collected. Protein concentration was quantified with the bicinchoninic acid assay (Thermo Fisher Scientific K.K., Tokyo, Japan). The size distribution and particle number of the prepared EVs were analyzed by nanoparticle tracking analysis (NTA). EVs were diluted to a concentration of 4-8 9 10 8 particles per mL with PBS and analyzed in triplicates using a Nanosight LM10 system (Marvern Instruments Ltd., Worcestershire, UK) equipped with a blue laser. Experimental conditions were as follows: Measurement Time: 60 s; Blur: Auto; Detection Threshold: 4-5; Min Track Length: Auto; Min Expected Size: Auto. Representative data are shown of three independent experiments. Morphology of EVs was examined using an HT7700-transmission electron microscopy (TEM). EVs (5 µL) were mixed with 4% paraformaldehyde (5 µL), incubated with Formvar film for TEM (PVF-C10 STEM; Okenshoji Co., Ltd., Tokyo, Japan) for 10 min, and negatively stained with 2% phosphotungstic acid for 20 s. After dry, EVs were observed by TEM. Representative data are shown.

Lectin microarray analysis
Extracellular vesicles (0.4 µg) were labeled with Cy3-N-hydroxysuccinimide ester (GE Healthcare Ltd., Tokyo, Japan), diluted with probing buffer [25 mM Tris/HCl (pH 7.5), 140 mM NaCl, 2.7 mM KCl, 1 mM CaCl 2 , 1 mM MnCl 2 , and 1% Triton X-100] to 0.5 µgÁmL À1 , and incubated in high-density lectin microarray at 20°C overnight. After washing with probing buffer, fluorescence images were captured using a Bio-Rex scan 200 evanescent-field activated fluorescence scanner (Rexxam Co. Ltd., Kagawa, Japan). Lectin signals of triplicate spots were averaged for each protein sample and normalized relative to the mean value of all 96 lectins. An unpaired Student's t-test was performed for comparison of the lectin signal in two groups. Having derived P-value for each lectin, q-values were calculated using the Storey approach and statistical significance was set at q-value < 0.05 [12]. Unsupervised clustering was performed by employing the average linkage method using CLUSTER 3.0 software [13]. The heat map with clustering was acquired using Java Treeview. Principal component analysis (PCA) was performed by spss statistics 19 (IBM Japan, Tokyo, Japan).

Lectin-precipitation and LC-MS/MS
Biotinylated rPALa lectin was immobilized on Dynabeads M280 Streptavidin (Thermo Fisher Scientific K.K.) and incubated with protein samples extracted from EVs at 4°C overnight. After washing with PBST (PBS containing 1% Triton X-100), bound proteins were eluted with 0.2% SDS solution at 95°C for 5 min and cooled on ice. The elution fraction was separated by SDS/PAGE followed by silver staining as described above. Protein bands were diced into small pieces, washed with diluted water and acetonitrile, and dried under reduced pressure. Subsequently, gel pieces were incubated with trypsin at 37°C for 16 h. Tryptic peptides were extracted with 5% formic acid and 50% acetonitrile buffer containing 0.1% trifluoroacetic acid, followed by desalting with C18 spin columns and dried under reduced pressure. Dried peptides were diluted in 3% acetonitrile and 0.1% formic acid, and analyzed by nano LC-MS/MS system composed of Q Exactive HF-X (Thermo Fisher Scientific K.K.) and UltiMate 3000 RSLCnano LC System (Thermo Fisher Scientific K.K.). Acquired MS/MS data were searched using PEAKS Studio X+ (Bioinformatics Solutions Inc., Waterloo, Canada) against amino acid sequences in the reviewed Uni-ProtKB/Swiss-Prot database (Homo sapiens). Search parameters were as follows: Digestion Enzyme, Trypsin; Precursor tolerance, 10 p.p.m.; Fragment tolerance, 0.015 Da; Maximum missed cleavage, 2; Variable Modifications, Oxidation (M). Peptide matching criteria were set at the number of unique peptides per protein ≥ 2. We excluded the desmosomal proteins including keratin family, desmoglein, and plakoglobin, from the analysis, since these proteins were potentially originated from the contaminated human skin and hair [14]. Glycoproteins were determined by searching in the UniProt database (https://www.uniprot.org/).

Statistical analysis
Wilcoxon-Mann-Whitney test, Chi-squared test, Spearman's rank correlation test, PSM analysis, and receiving operator curve (ROC) analysis were performed using EZR on R commander version 1.41 [15]. Statistical significance was set at P-value < 0.05.

Results
Glycan profiling of EVs derived from sera of APs In order to isolate small EVs, we used Tim4-affinity method, which has been reported to isolate EVs with high purity and low protein contaminants [11]. EVs were captured from sera of APs and HDs using magnetic beads immobilized with mouse Tim4, which binds specifically to PS exposed on EVs in a Ca 2+ -dependent manner [11]. Lipid bilayer vesicles in purified EV fractions from APs and HDs were examined by TEM (Fig. 1A). NTA revealed that average diameters of AP-derived EVs and HD-derived EVs were 197 AE 10.  (Table S1).  (Table S1). respectively (Fig. 1B). The number of particles at 1 µgÁmL À1 was 6 9 10 8 (AP-derived EVs) and 3 9 10 8 (HD-derived EVs) particles per mL. Small EVs could be purified from sera of both APs and HDs. The purified EVs were fluorescently labeled with Cy3-N-hydroxysuccinimide ester and incubated in a high-density lectin microarray containing 96 lectins (Table S2). Fluorescence signals were normalized by the average signals of 96 lectins and assessed by principal component and cluster analysis (Figs 1C and S1). EVs derived from APs and HDs were clearly separated. By Student's t-test, 13 lectins were selected that show significantly different signals between AP-and HD-derived EVs (q < 0.05, Table 1). Importantly, 8 of 13 lectins showed binding specificity to mannosylated glycans, suggesting higher expression of mannosylated glycans in AP-derived EVs than HD-derived EVs (Table 1).

Lectin blotting and identification of carrier proteins
We searched for carrier proteins of mannosylated glycans in AP-derived EVs. Lectin blotting using mannose-binding lectin (rPALa) showed a single protein band at 80 kDa in EVs purified from sera of patients of AD, but not HD-derived EVs (Fig. 1D). In order to identify the 80 kDa protein, protein lysates of AP-and HD-derived EVs were incubated with rPALaimmobilized beads followed by SDS/PAGE and silver staining. The protein band at 80 kDa was then analyzed by LC-MS/MS. Eleven glycoproteins with higher counts in AP than HD were detected by LC-MS/MS (Table S3). Among them, integrin b3, also known as CD61 and GPIIIa, was the sole protein with the average molecular weight around 80 kDa [16]. Higher count of integrin b3 was detected in AP-derived EVs (spectral count: 9) than HD-derived EVs (spectral count: 13) (Table S3). CD61 is a glycoprotein containing six N-glycosylation sites, specifically expressed on platelets and megakaryocytes [17].

Qualitative analysis by western blotting
In order to confirm whether CD61 shows higher expression in AP-derived EVs than HD-derived EVs, we then performed western blotting to confirm the results of Fig. 1. The equal number of particles or the equal protein amounts were run on SDS/PAGE and blotted with antibodies (Figs 2 and S2). Consistently, aCD61 showed stronger staining to AP-derived EVs than HD-derived EVs at 80 kDa in western blots (Figs 2 and S2). CD61 forms heterodimeric complexes with CD41 (integrin aII) as well as CD51 (integrin aV). CD61/CD41 complex is known to be expressed exclusively on platelets and megakaryocytes, whereas CD61/CD51 complex is expressed on various cell types such as endothelium, fibroblasts, smooth muscle, osteoclasts, and leukocytes [18]. Anti-CD41 pAb showed higher intensity to AP-derived EVs than HDderived EVs (Fig. 2B). In contrast, no obvious difference of aCD51 blot was observed between HD-and AP-derived EVs (Fig. 2C). Thus, we hypothesized that platelet-derived EVs (P-EVs) were enriched in APderived EVs. Heijnen et al. [19] classified P-EVs into two types: exosomes (P-Exos) and microvesicles (P-MVs). P-Exos are enriched in CD9, CD63, and CD41/ CD61, whereas factor X and prothrombin are restricted to P-MVs [20]. Antibodies against conventional exosome markers, CD9 and CD63, also showed greater binding to AP-derived EVs than HD-derived EVs (Figs 2D,E and Fig. S2). aCD63 showed broad protein bands ranging from 30 to 50 kDa, since CD63 is glycosylated [21]. Consistently, the 27 kDa protein band corresponding to CD9 was much stronger in APderived EVs than HD-derived EVs revealed in silver staining (Fig. 1D). These results suggest that P-Exos positive with CD9, CD63, CD41, and CD61, might be increased in APs relative to HDs.

Quantitative analysis by Tim4-based sandwich assays
The increase of P-Exos in AP-derived EVs was validated and quantified with sandwich assays using Tim4-immobilized microplates and antibodies against CD61, CD41, CD9, and CD63. Antibody against another exosome marker, CD81, was also used as a detection probe. Five sandwich assays were constructed and applied for the analysis of sera of 29 HDs and 20 APs. Tim4-coated microplates were incubated directly with sera of HDs and APs, and detected with antibodies against CD61, CD9, CD63, CD41, and CD81. Participant's demographic data are presented in Table S1. Consistent with western blotting data, values of Tim4-aCD61, Tim4-aCD9, Tim4-aCD63, and Tim4-aCD41 were significantly elevated in AP-derived sera (Fig. S3a-d). Tim4-aCD81 sandwich assay showed only a slight increment for AP sera (Fig. S3e). We found that there is a significant difference in age between two groups in whole cohort, indicating that the effect of age cannot be ignored (Table 2). Indeed, values of Tim4-aCD9 and Tim4-aCD81 sandwich assays showed a weak but significant correlation with age (P < 0.05) in the HD group (Table S4). Therefore, we performed 1 : 1 propensity score matching (PSM) to correct the difference in age and sex. After PSM, 16 samples were selected in each group, with no significant differences in age and sex ( Table 2). Even in PSM cohort, the values of Tim4-aCD61, Tim4-aCD41, Tim4-aCD9, and Tim4-aCD63 were significantly higher in the AP group (Fig. 3A-D).

Discussion
In the last two decades, biological fluid-derived EVs have been intensively investigated as biomarkers for diagnosis and therapy of various diseases, such as cancer, cardiovascular disease, and neurodegenerative disease [23][24][25]. We performed glycan profiling of EVs in sera of APs using high-density lectin microarray. Mannose-binding lectins were enriched and displayed increased signals in APs compared to HDs. The mannose-binding lectin rPALa showed a single specific band at 80 kDa, which was identified as CD61. Since membrane proteins are typically modified with complex N-glycans, mannosylation of CD61 might reflect unique functions for EVs, such as cellular uptake and tissue specificity. P-EVs were first discovered in 1967 as 'platelet dusts', tiny membrane fragments, released from platelets with the same procoagulant activity as platelets [26]. P-EVs ranged from 30 to 1000 nm in diameter and are characterized by expression of platelet markers (e.g., CD41, CD61) [20]. P-EVs are reported to be produced from platelets upon activation by various stimuli, including exposure to platelet agonists (thrombin and collagen), complement proteins, shear stress, and adhesion to vessel walls [27]. P-EVs are categorized into two groups, P-Exos and P-MVs. We demonstrate that P-Exos, relatively small P-EVs, ranging from 30 to 200 nm in diameter, are increased in sera of APs. Several lines of evidence show increased platelet activation in APs [28]. Sevush et al. [29] showed increased levels of platelet aggregates, expression of CD62p on platelets, and leukocyte-platelet complexes in AP groups. Ciabattori et al. reported increased levels of urinary 11-dehydro-thromboxane B 2 (platelet activation marker), 8-iso-prostaglandin (lipid oxidation marker), and decreased vitamin E (antioxidant) in APs, suggesting that the platelet activation is accompanied by oxidative stress and a decrease in antioxidants [30]. Bermejo et al. [31] showed augmented platelet levels of COX2 and plasma IL-6 level in AP groups, suggesting that activation of platelets with peripheral inflammatory responses. Prodan et al demonstrated that coated platelets, which are collagen and thrombin activated platelets, were significantly elevated in APs compared to age-matched controls. The number of coated platelets is positively correlated with MMSE score in APs [32], similar to our results obtained by Tim4-aCD41 and Tim4-aCD61 sandwich assays. Thus, increased levels of P-Exos in sera of APs might be due to platelet activation.
An increase in circulating P-EVs has also been reported in a variety of diseases including cancer, infections, rheumatoid arthritis, and stroke [27,33]. P-EVs display a variety of bioactive substrates, such as growth factors, cytokines, and lipids that affect the pathophysiology of disease. For example, P-EVs promote cancer cell proliferation and invasion by stimulating MAPK signaling and increasing matrix metalloproteinases [34]. P-EVs also enhance the formation of new blood vessels during tumor growth via concerted action of FGF-2, VEGF, and a lipid factor [27,35]. P-EVs could trigger neurogenesis and angiogenesis by stimulating ERK and PI3K/Akt signaling in endogenous neural stem/progenitor cells after a stroke [36]. Conversely, P-EVs exacerbate inflammation in rheumatoid arthritis through IL-1-induced secretion of pro-inflammatory cytokines in synoviocytes [37]. Although the pathological role of P-EVs on APs is unclear, these studies suggest that P-EVs may counteract neurodegeneration in the brain of APs by enhancing angiogenesis and neurogenesis. Alternatively, P-EVs may exacerbate AD pathology by amplifying inflammation. The pathological role of increased P-EVs in APs will need to be addressed in future studies.
In this study, we developed Tim4-based sandwich assays to characterize EVs in sera of APs. Values of Tim4-aCD61, Tim4-aCD9, Tim4-aCD63, and Tim4-aCD41 were increased in sera from APs, whereas Tim4-aCD81 showed comparable signals between APs and HDs, suggesting that P-Exos positive for PS, CD61, CD41, CD9, and CD63, are exposed on exosomes released from platelets. In contrast, CD81 might be expressed on EVs of different origins. Since there is no specific marker for each type of EVs: exosome, microvesicle, apoptotic body, it is difficult to know which types of EVs were specifically detected by Tim4based sandwich assays. The Tim4-aCD63 sandwich assay showed the best performance for predicting AD (AUC: 0.957) in ROC curve analysis, indicating that CD63-positive EVs are increased in sera of AP. Tim4-aCD41 and Tim4-aCD61 might be suitable for specific detection of P-EVs. However, it should be noted that the elevation of P-EVs might not be AD-specific pathology as mentioned above, and Tim4-aCD63 sandwich assay may also respond to sera of patients with other diseases. Further studies are essential to apply this finding to clinical settings.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. List of sera from HDs and APs. Table S2. Lectins used for lectin microarray. Table S3. Identification of rPALa-precipitated glycoproteins at 80 kDa by LC-MS/MS. Table S4. Spearman's rank correlation coefficient between the values of sandwich assays and ages in HDs and APs. Table S5. Spearman's rank correlation coefficient among Tim4-based sandwich assays. Fig. S1. Cluster analysis of lectin microarray data of EVs purified from sera of HDs and APs. Lectin microarray data of EVs purified from sera of APs (n = 3) and HDs (n = 3) were normalized, log-transformed, and analyzed by Cluster 3.0 with average linkage methods. The zero value of lectin signal was converted to 1. Sample No.4, 5, 6 for HDs and No.30, 31, 32 for APs were used in Table S1. yellow: high, blue: low.  Fig. S3. Quantitative analysis of HD-and AP-derived EVs using Tim4-based sandwich assays in whole cohort. Box-whisker plots of the data of whole cohort (HDs: n = 29, APs: n = 20) analyzed by sandwich assays using immobilized Tim4 and overlay antibodies against CD61 (a), CD9 (b), CD63 (c), CD41 (d), and CD81 (e). OD: optical density. Sera used in this study are listed in Table S1. P-values obtained by Wilcoxon-Mann-Whitney Test are indicated in the figure.