Rapid and high-yield recovery of plasma-derived extracellular vesicles using modified chromatography with soluble protein depletion for biomarker discovery

Abstract

Extracellular vesicles (EVs) are critical mediators of intercellular communication by transferring proteins, lipid and nucleic acids between cells. EVs in biofluids, particularly blood, have gathered significant interest as potential biomarkers for disease diagnosis. However, isolating EVs from blood poses a challenge due to the high concentration of plasma proteins, which obscure the detection of low abundant EV-associated proteins. Here, we optimized a simplified and efficient method for isolating plasma-derived EVs by combining size exclusion chromatography (SEC) with flow-through chromatography using Capto Core 700 beads. A brief incubation of SEC-derived EV fractions with Capto Core beads (qEV + CC) enabled us to isolate intact, high-purity EVs with reduced soluble plasma protein contamination. As a comparison, MagReSyn-based method was not compatible with elution of intact EVs after the purification and showed significant contamination of soluble plasma proteins. Data-independent acquisition-based liquid chromatography-mass spectrometry of isolated plasma-EVs using the qEV + CC approach identified over 1,000 EV-associated proteins, including an increased presence of brain derived proteins and markers linked to neurodegenerative diseases, such as amyloid precursor protein and apolipoprotein E. These findings were further validated by super-resolution microscopy at a single EV resolution. Bioinformatic pathway and network analyses revealed enrichment of pathways involved in RNA processing, cell adhesion and synaptic function, highlighting the potential of EV molecules for broad disease biomarker discovery. Our findings present an optimized method for efficient purification of plasma-derived EVs, providing a valuable tool for advancing EV-based biomarker development.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12964-025-02263-3.

Introduction

Extracellular vesicles (EVs) are small membranous vesicles with the diameter ranging from 30 − 1,000 nm [1, 2]. EVs have recently gained increased attention as they are significantly involved in intercellular communication by transferring diverse cargo molecules, including proteins, lipids and nucleic acids under physiological or pathological conditions [2, 3]. For example, amyloid precursor protein (APP) and apolipoprotein E (APOE), the key proteins implicated in the pathogenesis of neurodegenerative diseases, are reported to selectively sorted into EVs, revealing the potential roles of EVs in disease propagation and intercellular signaling [4]. Previous studies showed EV cargo proteins are useful biomarkers for cancer [5, 6] and cardiovascular disease [7] to predict disease conditions. Furthermore, accumulating evidence suggests that EVs can penetrate blood brain barrier (BBB) [8–10], especially under the inflammatory condition when the permeability and integrity of BBB is perturbed [10]. Therefore, EV-associated molecules in the periphery can be promising biomarkers to understand the physiological brain state and progression of neurodegenerative diseases.

Blood is the most used biofluid and promising source for studying EV-based biomarkers due to the dynamic and low-invasive accessibility [11, 12]. For example, Muraoka et al. reported the proteomic profiling of EVs separated from the plasma samples of American football players at risk for chronic traumatic encephalopathy (CTE) and found the combination of COL6A3, RELN and COL6A1 in plasma EVs can differentiate these players from the control group [13]. Notably, a most recent study with a large cohort of disease samples identified plasma extracellular vesicle 3R/4R tau ratio and TDP-43 level as diagnostic biomarkers in frontotemporal dementia and amyotrophic lateral sclerosis, providing a potential target engagement in clinical trials [14]. Although there are several EV separation methods, including ultracentrifugation (UC), density gradient centrifugation, size exclusion chromatography (SEC), immuno-capture and polymeric precipitation [15], the characterization of plasma-derived EVs faces challenges due to the high abundance of plasma proteins such as albumin, immunoglobulins, and lipoproteins, which are present at concentrations several orders of magnitude higher than EV-associated proteins [16, 17]. Particularly, in bottom-up liquid chromatography tandem mass spectrometry (LC-MS/MS), these high-abundance proteins dominate ionization and fragmentation during data acquisition, thereby suppressing signals from lower-abundance peptides and reducing overall proteome coverage. This dynamic range issue, combined with the naturally low concentration of EVs in biological fluids, limits the detection of rare or disease-relevant EV proteins [16, 17]. To overcome these limitations, data-independent acquisition mass spectrometry (DIA-MS) provides a more comprehensive and reproducible analysis by systematically sampling all detectable ions, improving quantification and detection of low-abundance proteins despite complex backgrounds [18].

In this study, we combined SEC using a qEVorignial 70 nm column with flow-through chromatography (Capto Core 700) to remove abundant soluble plasma proteins and purify plasma-derived EVs by maximizing the depth of EV protein coverage. The Capto Core 700 resin bead (CC), designed for the intermediate purification of substantial biomolecules such as viruses, functions efficiently in a flow-through mode with an exclusion limit of 700 kDa [19, 20]. Previous studies have shown its ability to bind and trap free proteins, leaving pure EVs in solution after incubation with the biofluid [21, 22]. Here, we provide a comprehensive comparison of the physical and biological properties of EVs isolated from human plasma using five different methods: qEV alone, 16-hr incubation with CC at 4 ℃ followed by qEV (CC + qEV), qEV followed by 1-hr incubation with CC at room temperature (qEV + CCL), qEV followed by 5-min incubation with CC (qEV + CCS) at room temperature and recently reported Mag-Net protocol (MagReSyn^® SAX, a hyper-porous magnetic bead technique that captures membrane-bound particles) [23]. Our findings indicate a significant increase in purity of EVs isolated by SEC followed with Capto Core beads, whereas Mag-net based purification still has a significant contamination of soluble plasma proteins. Additionally, a greater number of plasma-EV proteins was identified, particularly in the qEV + CCS group, by DIA-MS. Proteomic profiling of plasma-EVs revealed a marked reduction in some of abundant plasma proteins (e.g., ALB, IGG1 and APOA1) and a higher enrichment of common EV markers (e.g., CD9, CD81 and CD63) in the qEV + CCS method. Notably, cell-type enrichment analysis of the plasma-EV proteome revealed a higher presence of brain-specific proteins, including those strongly associated with neurodegenerative diseases, in the qEV + CCS group. Pathway and protein network analysis further revealed that the proteins, uniquely identified in qEV + CCS group compared to other groups, are associated with RNA binding, cell adhesion, metabolism and synaptic function. Our study establishes a robust methodology for the rapid purification of plasma-derived EVs and provides a foundation to support the development of plasma-EV-based biomarkers for potential disease diagnosis.

Results

Purification of plasma-derived EVs by mixed chromatography

Previous studies reported that SEC-based isolation minimally alters EV characteristics with less plasma protein contamination compared to other conventional isolation approaches such as ultracentrifugation and precipitation [24, 25]. Here, we performed a comprehensive comparison of the protein cargos within EVs isolated from human plasma, employing five distinct purification strategies that combine SEC with flow-through chromatography. Initially, the 500 µL aliquot of human plasma from one healthy individual underwent serial centrifugation (1500 x g and 5000 x g) to remove cells and debris. The resulting supernatant was subsequently used for different EV isolation method: SEC (qEVoriginal 70 nm column); qEV + CCS (SEC followed by Capto Core 700 5-min short incubation); qEV + CCL (SEC followed by Capto Core 700 1-hour long incubation); CC + qEV (Capto Core 700 16-hour pre-treatment followed by SEC); MagReSyn (magnetic ReSyn beads). The final EV volume is adjusted to the equal amount of 50 µL for all groups. The EVs obtained from each purification method were subjected to comprehensive downstream analysis to assess EV properties and protein profiles.

We first performed biophysiological characterization of plasma-derived EVs isolated by different methods. Manually fractionated Fraction 7–10 from SEC isolation were determined and pooled as the EV-enriched fractions with the lower amount of soluble proteins according to protein quantification assay and our previous publication [13] (Supplemental Fig. 1). The cryogenic electron microscopy (Cryo-EM) analysis revealed that EV isolated by the distinct methods exhibited the small, spherical structure with a clearly defined lipid bilayer membrane [26], indicating the preservation of vesicle integrity (Fig. 1A). Then, nanoparticle tracking analysis (NTA) and nano-flow cytometry (NanoFCM) was performed to evaluate the size distribution and particle concentration of EVs from each isolation method. NTA showed that EVs isolated by Capto Core 700 beads after qEV (qEV + CCL, qEV + CCS) had clearer peak of particle distributions (Fig. 1A) when compared to other two methods (qEV, CC + qEV). For NanoFCM analysis, we calibrated with the S16M-Exo calibration bead for small particle (40–200 nm) analysis and used MemGlow 488 lyophilic dye to stain membrane-bound particles to distinguish EVs (P1: MemGlow 488⁺) from other non-vesicular particles such as exomeres (P2: MemGlow 488⁻) as shown in Fig. 1B and S2A [27]. Notably, the qEV + CCL method showed a significant decrease in EV concentration (Fig. 1B and C), suggesting that prolonged incubation with Capto Core 700 beads may lead to higher EV loss, likely due to increased non-specific binding or trapping of EVs to CC beads. Our data revealed the concentration of total particles and P1 population was reduced in qEV + CCL (mean ± SD, Total: 6.12E9 ± 2.72E9 / mL, P1: 4.44E9 ± 2.38E9 / mL) and qEV + CCS (mean ± SD, Total: 7.01E9 ± 1.53E9 / mL, P1: 4.48E9 ± 1.24E9 / mL) compared to qEV (mean ± SD, Total: 2.16E10 ± 7.96E9 / mL, P1: 1.21E10 ± 1.06E10 / mL) and CC + qEV group (mean ± SD, Total: 1.45E10 ± 6.26E9 / mL, P1: 1.95E10 ± 4.88E9 / mL; Fig. 1C and S2B), which is consistent with NTA data (mean ± SD, qEV: 1.10E11 ± 4.07E10, CC + qEV: 1.06E11 ± 3.93E10, qEV + CCL: 2.88E10 ± 1.41E10, qEV + CCS: 4.80E10 ± 5.20E9). Interestingly, the qEV + CCS method (mean ± SD, Total: 76.47 ± 1.90 nm, P1: 79.80 ± 1.97 nm) shifted the size distribution of P1 population toward smaller size of particles compared to qEV (mean ± SD, Total: 82.17 ± 9.39 nm, P1: 92.47 ± 7.84 nm), CC + qEV (mean ± SD, Total: 86.97 ± 2.75 nm, P1: 93.37 ± 3.87 nm) and qEV + CCL (mean ± SD, Total: 79.17 ± 0.76 nm, P1: 82.80 ± 1.49 nm), indicating a potential enrichment of small vesicles (Fig. 1B and C). In addition, no difference of total particle size and percentage of P1 population was observed across different groups (Figure S2B).

Fig. 1.

Characterization of plasma-derived EVs (pEVs) isolated from different purification methods. A, Nanoparticle tracking analysis (NTA) of size distribution and Cryo-EM images of pEVs isolated by distinct methods. Dilution factor was 300 for qEV + CCL and qEV + CCS group. Scale bar, 200 nm. B, Representative size distribution histograms of pEVs by nano-flow cytometry (NanoFCM). The percentage, size and concentration of different particle subpopulations were shown. P1, MemGlow_488 positive particles; P2, MemGlow_488 negative particles; Total, overall particles. C, Quantitative analysis of EV concentration and size measured by NTA and NanoFCM among different isolation methods. D, Silver staining of proteins from neat plasma and pEVs purified from five methods. 0.01 µL plasma was used as a positive control. 1 µL EVs (total 50 µL) from 500 µL of plasma separated by qEV, CC + qEV, qEV + CCL, qEV + CCS and MagReSyn were loaded. E, Quantification of total pEV proteins before (left panel) and after lysis (right panel) for each isolation method measured by microBCA assay. F, EV purity revealed by ratio of the number of particles to the total proteins after lysis was compared among different purification methods. Data are shown as the mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons test was applied. ns, no significance. *p < 0.05. Three biological replicates were used

Next, we assessed the protein yield in neat plasma and plasma-derived EVs isolated from distinct methods. The Mag-Net method, a rapid enrichment strategy of plasma-EV via hyper-porous strong-anion exchange magnetic microparticles (MagReSyn beads) [23], was introduced for the comparison. Silver staining revealed that the most abundant plasma proteins including albumin (ALB) and globulins [11], ranging in size from 50 to 100 kDa, were observed in the neat plasma, qEV, and CC + qEV groups (Fig. 1D). Intermediate levels of plasma proteins were also detected in the MagReSyn group, while the least amount was present in the qEV + CCL and qEV + CCS groups, suggesting that Capto Core 700 beads are more effective at removing plasma proteins (Fig. 1D). The micro bicinchoninic acid (micro-BCA) protein assay further demonstrated more enriched EV cargo proteins in the qEV + CCL and qEV + CCS compared to the qEV and CC + qEV groups after RIPA lysis (Fig. 1E). Since MagReSyn is not compatible with the release of intact EV from the beads, all the EV samples were lysed after MagReSyn-based isolation. The EV particle-to-protein ratio indicated that the purity of EVs was highest in the qEV + CCS samples (Fig. 1F), showing that the qEV + CCS method resulted in a more efficient and purified isolation of EVs with reduced protein contamination compared to other methods. These findings suggest that the Capto Core beads, particularly used in the qEV + CCS isolation approach, provide the optimal yield and purity of EVs, with a higher concentration of membrane-bound vesicles and minimal protein contamination, making it ideal for applications requiring high-quality EVs.

Proteomic analysis of plasma-derived EVs by different isolation methods

Neat plasma and plasma EVs were further evaluated by quantitative DIA-MS to explore how the distinct methods affected EV proteomic characteristics. Proteins identified in at least one out of three repeats in each group were selected for analysis. The uniform manifold approximation and projection (UMAP) analysis of these identified proteins exhibited a clear separation between neat plasma and EVs separated by the distinct methods (Fig. 2A). A total of 1240 proteins with at least one valid value were identified (FDR < 1%), of which 778, 870, 908, 925, 941 and 1153 proteins, including 589 in common were found in neat plasma, qEV, CC + qEV, qEV + CCL, MagReSyn and qEV + CCS, respectively (Fig. 2B; Supplementary Table 1). We next examined the expression levels of top 20 classic EV markers listed in the Vesiclepedia dataset [28], along with major plasma proteins across the different groups. The heatmap with label-free quantification (LFQ) value revealed that all these EV markers were identified in the qEV + CCL and, particularly, the qEV + CCS group, with higher expression levels of markers such as CD9, CD63, CD81, Annexin A5 (ANXA5), TSG101, Annexin A2 (ANXA2), and SDCBP, compared to the other methods (Fig. 2C). In contrast, several major plasma proteins, including transferrin (TF), apolipoprotein A1 (APOA1), ALB and immunoglobulin G (IgG1), were significantly reduced in the qEV + CCL and qEV + CCS groups (Fig. 2D), suggesting the effectiveness of Capto Core beads in depleting some high-abundance plasma proteins from the EV fractions. However, other plasma proteins such as APOB, APOC1-4, IGM, and C1Q were more concentrated in the qEV + CCL and qEV + CCS groups, likely reflecting the selective removal of high density lipoparticle (HDL)-associated proteins by Capto Core beads, while (very) low density lipoparticle ((V)LDL)-associated components and some complements were retained. Quantitative analysis of ALB, apolipoprotein A1 (APOA1) and IGG1 from DIA-MS showed significantly lower protein levels in the qEV + CCS group (Fig. 2E). These results suggest that EVs are enriched, and several contaminating proteins are depleted by the SEC method combined with Capto Core beads (Supplementary Table 1). Although co-isolation of LDL/VLDL particles is observed in the qEV + CCS group, we still achieved better coverage of EV-specific markers, suggesting that these lipoparticles may have a lower impact on masking EV-specific signals compared to highly abundant soluble proteins such as albumin, IGG1, and APOA1.

Fig. 2.

Proteomics analysis of neat plasma and plasma-derived EVs purified by distinct methods. A, UMAP of proteome profile from six group samples showing the good separation across neat plasma and different pEVs. B, The number of identified proteins among six groups by DIA-MS. The proteins identified in at least one out of three replicates were included and the number of their identifications were shown. C, The Heatmap showing the top enriched common EV proteins across neat plasma and pEVs isolated by qEV, CC + qEV, qEV + CCL, qEV + CCS and MagReSyn based on their protein intensity as determined from the proteome data. D, The Heatmap showing the protein intensity of common plasma proteins in neat plasma and pEVs isolated by qEV, CC + qEV, qEV + CCL, qEV + CCS and MagReSyn. E, Quantitative analysis of the expression of plasma proteins including albumin (ALB), apolipoprotein A1 (APOA1) and immunoglobulin (IGG1) in neat plasma and different pEVs. Data are shown as the mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons test was applied. *p < 0.05. **p < 0.01, ****p < 0.0001. Three biological replicates were used

Dynamic range of plasma-EV proteome across distinct methods

We next ranked the protein abundance to evaluate the effect of each isolation method on the dynamic range of EV proteome compared to neat plasma (Fig. 3). The signal plotted for the protein abundance across different groups is indicated as Log10 protein value. The expression levels of the classic EV marker proteins and plasma proteins are indicated in blue and red, respectively. As expected, the plasma proteins including APOA1, ALB, IgG1 and A2M (alpha 2 Macroglobulin proteins) were the most abundant in neat plasma, while only a few of EV marker proteins including PDCD6IP (Alix), ANXA2, HSP90B1 and TSG101 were identified (Fig. 3A). Then, we compared the dynamic range of EV proteome in plasma. Despite EV purification methods were applied, those plasma proteins are still some of the most abundant proteins in EV samples, which is analogous to what has been reported in other studies [23, 29, 30]. In the qEV method, we observed the increased number of detectable EV-specific markers, such as CD9 and CD81 (Fig. 3B), suggesting that qEV can partially separate EVs from plasma proteins. Although the addition of Capto Core beads before qEV isolation (CC + qEV) improves protein separation by reducing some plasma proteins, the increase in the ranks of EV markers is not substantial (Fig. 3C), indicating that the method still faces limitations in enriching EV-specific contents. In contrast, the addition of Capto Core beads after qEV isolation showed marked improvement in reducing plasma proteins and enriching EV-specific markers (Fig. 3D and E). To be noted, all those classic EV markers such as SDCBP, CD9, CD63, CD81, TSG101, ANXA2, Flotillin-1 (FLOT1), and PDCD6IP, are identified in qEV + CCS method with the intermediate abundance ranks (Fig. 3E). Similarly, the MagReSyn method revealed comparable results to CC + qEV, with moderate enrichment of EV proteins (Fig. 3F). Quantitative analysis of the common EV proteins including PDCD6IP, FLOT1, ANXA2 and CD9 from DIA-MS revealed the greater enrichment of these EV markers in the qEV + CCL and qEV + CCS group (Fig. 3G). Consistently, a higher enrichment of those EV proteins in the qEV + CCS group was validated by immunoblotting of equal amounts of plasma and EV proteins across all groups (Fig. 3H). We also confirmed the relatively high expression of APOB in the qEV + CCS group, suggesting the potential co-isolation of APOB⁺ lipoprotein particles within EVs (Fig. 3H). However, the qEV + CCS group exhibited a lower abundance of the non-EV protein including Albumin and APOA1 compared to neat plasma, qEV and MagReSyn (Figs. 2E and 3H), suggesting effectiveness of this method in isolating EV-specific proteins while decreasing some major plasma protein contamination.

Fig. 3.

Dynamic range of the proteome from neat plasma and plasma-derived EVs purified by distinct methods. A to F represents neat plasma, qEV, CC + qEV, qEV + CCL, qEV + CCS and MagReSyn, respectively. The protein expression value was measured using the average of triplicate measurements by DIA-MS. The signal plotted for the protein abundance in all samples was indicated as Log10 value. The common EV marker proteins are shown in blue and the classic abundant plasma proteins in red. G, Quantitative analysis of the expression of common EV proteins including Alix (PDCD6IP), Flotillin-1 (FLOT1), Annexin A2 (ANXA2) and tetraspanin 9 (CD9) in neat plasma and different plasma-EVs. Data are shown as the mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons test was applied. *p < 0.05. **p < 0.01, ***p < 0.01, ****p < 0.0001. Three biological replicates were used. N.D., non-detected. H, Immunoblotting of EV marker proteins including Alix, FLOT1, ANXA2 and CD9, and non-EV proteins including APOB, Albumin and APOA1 in neat plasma and plasma-derived EV samples

Cell type-specificity and disease-association of plasma-EV proteome across distinct methods

Given our strong interest in leveraging circulating plasma EVs as biomarkers for brain diseases, we investigated the overlap between proteins found in plasma-derived EVs and those in brain-derived EVs. By comparing our plasma-derived EV proteome to a previously published brain EV protein dataset in which brain-EVs were isolated using sucrose gradient ultracentrifugation [27], we found that the qEV + CCS method identified the highest number of overlapping proteins between brain-EVs and plasma-EVs (368 proteins), followed by qEV + CCL (300 proteins), MagReSyn (274 proteins), CC + qEV (260 proteins), and qEV alone (242 proteins) (Fig. 4A). In contrast, only 205 overlapping proteins were identified in neat plasma. We further assessed brain cell type-specific molecules within the plasma-EV proteome across these methods by referencing a published brain cell types database [31]. The qEV + CCS method yielded a higher percentage (qEV + CCS, 37.8%; neat plasma, 25.9%; qEV, 29.7%; CC + qEV, 30.3%; qEV + CCL, 30.7%; MagReSyn, 33.5%) of central nervous system (CNS) cell-type markers, with a notable enrichment of endothelial and glial proteins (Fig. 4B). Additionally, using our previously published cell type-specific EV proteomic dataset in which EVs were isolated using sequential centrifugation and size exclusion chromatography [32] from iPSC-derived neurons, astrocytes, microglia, and oligodendrocytes, we observed that the qEV + CCS method consistently contained a higher percentage of CNS-derived EV proteins (qEV + CCS, 223%; neat plasma, 150%; qEV, 194%; CC + qEV, 169%; qEV + CCL, 213%; MagReSyn, 199%; Fig. 4C). These results suggest that the qEV + CCS method is particularly effective for isolating EVs reflective of brain-derived cargos, including CNS-specific biomarkers.

Fig. 4.

Brain specific cell-type enrichment analysis of proteome from neat plasma and plasma-derived EVs. A, The number of overlapped proteins in plasma and pEVs with brain-derived EVs. The dataset of brain-EVs was from our previous publication (Zhang et al., 2023). B, The bar plot showed the percentage of brain cell type-specific markers identified in neat plasma and pEVs. The top 1000 cell specific markers were selected according to dataset (Sharma et al., 2015) and the percentage of each cell type was calculated as 100* no. of detected cell type proteins / 1000. End, endothelial cell; Ast, astrocyte; Mic, microglia; Oli, oligodengrocyte; Neu, neuron. C, The bar plot showed the percentage of brain cell type-specific EV markers identified in neat plasma and pEVs. The cell specific EV markers were selected from our previous cell-type specific EV proteomics study (You et al., 2022) and the percentage of each cell type was calculated as 100* no. of detected cell type proteins / total cell type EV proteins in background dataset (iAst-EV, 240; iMic-EV, 51; iOli-EV, 10; iNeu-EV, 21). D, The heatmap showed the expression level of proteins associated with neurodegenerative diseases in neat plasma and pEVs purified by different methods. E, Cell-type-Specific Enrichment Analysis (CSEA) results of the proteome of pEVs isolated by qEV and qEV + CCS in 12 human organ systems (left panel) and top 20 general cell types (right panel). The dashed red line indicates the significant threshold (p = 3.69 × 10^-5) corrected with 1355 tissue-cell types. The solid grey line indicates the nominal significance (p = 1 × 10^-3). In each category of tissues, dots represent all the tissue-cell types and descend by the order of most significant tissue-cell type in this tissue. Y-axis is the–log10(combined p-value) from the CSEA result. The results highlighted cardiovascular system were the most related organ system and other top significant tissue types are digestive system and endocrine system. In each category of general cell types, dots represent all tissue-cell types annotated in this general cell type and descend by the order of most significant tissue-cell type in this tissue. Y-axis is the–log10(combined p-value) from the CSEA result. The three most significant general cell types were annotated as epithelial cell, monocyte, and endocrine cells, respectively. The number of identified pEV proteins above the significance threshold was higher in qEV + CCS group than that in qEV group, suggesting improved coverage of EV proteins following purification from Capto Core beads

To further investigate proteins associated with neurodegenerative diseases, such as Alzheimer’s and Parkinson’s disease, we referenced the plasma EV proteins isolated by different methods against genome-wide association study (GWAS) datasets for neurodegenerative diseases from GeneSigDB and DsigDB (https://maayanlab.cloud/Enrichr; Supplementary Table 2). Key proteins including APOE, APP, CLU, CTSB, and HLA-DRB1, known for their critical roles in neurodegenerative processes [33], were more abundantly enriched by the qEV + CCS, qEV + CCL, and MagReSyn methods compared to others (Fig. 4D). Notably, the qEV + CCS method identified the highest levels of these neurodegenerative disease-associated proteins, further underscoring its suitability for studies focused on EV biomarkers relevant to neurodegenerative conditions. Finally, we conducted cell type-specific enrichment analysis (CSEA) of plasma EV proteins for various human tissues and cell types using the online WebCSEA tool (https://bioinfo.uth.edu/webcsea/index.php). In each category of organ systems (Fig. 4E, left panel), dots represent all the tissue-cell types stratified by the adult organ system. In each category of general cell types (Fig. 4E, right panel), dots represent all tissue-cell types annotated in this general cell type and descend by the order of most significant tissue-cell type. Our analysis revealed that both qEV and qEV + CCS groups showed enrichment in the digestive and lymphatic systems; however, the qEV + CCS group exhibited a higher enrichment of proteins with statistically significant p-values (Fig. 4E). In terms of cell types, the qEV + CCS group was more enriched for proteins associated with endothelial, myeloid, and endocardial cells, while the qEV group showed higher enrichment for neutrophil and enterocyte proteins (Fig. 4E). These findings suggest that adding Capto Core beads to SEC effectively enhances the coverage of plasma EV proteome potentially linked to a broader range of organ and cell origins.

Pathway and protein network analysis revealed enriched protein clusters in plasma-EVs isolated from the qEV + CCS method

Venn diagrams were then used to examine the common and unique plasma-EV proteins with at least one valid value identified in the qEV + CCS group compared to other groups. Specifically, 442 proteins were identified exclusively in the qEV + CCS group over neat plasma, 345 over qEV alone, 308 over CC + qEV, 231 over qEV + CCL, and 267 over MagReSyn (Fig. 5A). We first conducted pairwise comparisons of these common proteins identified in at least two out of three replicates between the qEV + CCS method and each of the other tested methods, as well as neat plasma. The significantly upregulated and downregulated proteins were selected with FDR corrected p value < 0.05 and|fold change| > 2 as shown in the volcano plots (Figure S3). Specifically, we observed an enrichment of proteins associated with extracellular matrix and cell adhesion (e.g., ITGB1, ICAM2, VCAN), and some specific apolipoprotein family members (e.g., APOC1-4, APOB, APOE) increased in qEV + CCS compared to qEV, CC + qEV and MagReSyn. By contrast, the complements (e.g., C2, C3, C8B, C9) and IGG proteins were significantly reduced in qEV + CCS. In addition, no significantly altered proteins were identified between qEV + CCS and qEV + CCL.

Fig. 5.

Comparative proteome analysis of plasma-derived EVs separated by the qEV + CCS method. A, Venn diagram of identified proteins in qEV + CCS group compared to neat plasma, qEV, CC + qEV, qEV + CCL and MagReSyn, respectively. B, Gene Ontology (GO) and KEGG pathway analysis of combined proteins uniquely identified in qEV + CCS group over neat plasma, qEV, CC + qEV, qEV + CCL and MagReSyn. Total 565 proteins were included for the analysis. BP, Biological Process; CC, Cellular Component; MF, Molecular Function. C, STRING network analysis of 565 unique proteins identified in qEV + CCS group. The analysis focused on high-confidence interactions and identify their top GO categories. The top five clusters of protein nodes were calculated by k‐means depend on their proximity and frequent occurrences, and then these clusters were labelled with their respective prominent pathways. Colored solid lines between proteins represented the interaction evidence. Cluster 1: Cytosolic ribosome; Cluster 2: Cell matrix adhesion; Cluster 3: Metabolic pathways; Cluster 4: Spliceosome; Cluster 5: Synapse

Next, by consolidating these unique proteins detected by qEV + CCS from each comparison, we obtained a total of 565 proteins, which were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes (KEGG) pathway analysis. Our results indicate that these enriched plasma-EV proteins combined from each comparison in the qEV + CCS group are significantly associated with GO and KEGG terms including cytoplasmic translation, cell-matrix adhesion, extracellular exosomes, synapses, ribosome, and RNA binding (Fig. 5B; Supplementary Table 3). This result indicates a marked advantage of the qEV + CCS approach for plasma-EV protein purification.

Furthermore, STRING network analysis of the 565 unique proteins identified five distinct functional clusters (red, yellow, green, cyan, and blue clusters, Fig. 5C; Supplementary Table 4). The red cluster includes cytosolic ribosome proteins (e.g., RPS9, RPL8, ATP6V1A, ATP5F1A), which are associated with translation and ribosome assembly [34] as well as ATP synthesis and binding [35]. The yellow cluster consists of cell-matrix adhesion proteins (e.g., ITGAM, PECAM1, CD47), which are relevant to neurodegeneration and neuroinflammation processes [36, 37]. The green cluster highlights metabolic pathway proteins, such as members of the RAB GTPase family (e.g., RAB5A, RAB5B), essential for endocytosis and metabolism [38]. The cyan cluster includes spliceosome-related proteins (e.g., HNRNPA2B1, HNRNPM, HNRNPK), which are critical for RNA splicing and processing [39, 40]. Lastly, the blue cluster features synapse-related proteins (e.g., APP, ADAM10), which are integral to synaptic function and neurotransmission within the central nervous system [41, 42]. These findings collectively highlight the capability of qEV + CCS method to enrich plasma-EV proteins involved in important and diverse biological functions, providing a robust framework for exploring EV-based blood biomarkers for various diseases.

Single particle analysis showing distinct EV subtypes in plasma-EVs isolated from different methods

To further investigate the distinct EV subtypes in plasma-EVs isolated using different methods, we employed direct stochastic optical reconstruction microscopy (dSTORM)-based imaging (Nanoimager, ONI Oxford) to analyze the enrichment and distribution of key EV proteins at the single-particle level. The MagReSyn group is excluded for single-EV analysis because this method is not compatible with the elution of intact EVs. The proteins selected for this analysis, including tetraspanins (CD9, CD81, and CD63), APP, and APOE, were identified in our proteomic analysis as highly relevant to neurodegenerative diseases. Fluorescence-conjugated antibodies against CD9, CD81 and CD63 were used to stain single EVs (Fig. 6A). The clustering analysis revealed that the qEV + CCS method resulted in a significantly greater proportion of CD63-positive particles compared to qEV alone (Fig. 6B). Additionally, the percentage of CD81⁺CD63⁺ double-positive EVs was markedly higher in the qEV + CCS group (Fig. 6B), suggesting the method’s superior ability to enrich CD63⁺ EV subpopulations.

Fig. 6.

Single-EV analysis of tetraspanins, APP, and APOE in plasma-EVs across distinct methods using super-resolution microscopy. A, Representative images showing the fluorescent staining of tetraspanins (CD81, CD9, CD63) in the single plasma-EV isolated across different methods (qEV, CC + qEV, qEV + CCL, and qEV + CCS). Scale bar: 200 nm. B, Quantitative analysis of the tetraspanins-positive particles across four isolation methods. Three biological replicates were used. C, Representative images showing the fluorescent staining of APP and pan-tetraspanins (CD81, CD9, CD63) in the single plasma-EV isolated across different methods. Scale bar: 200 nm. D, Percentage of APP⁺ pan-tetraspanins⁺ particles (normalized to total pan-tetraspanins⁺ particles) in plasma-EVs among different groups. E, Representative images showing the fluorescent staining of APOE and pan-tetraspanins (CD81, CD9, CD63) in the single plasma-EV isolated across different methods. Scale bar: 200 nm. F, Percentage of APOE⁺ pan-tetraspanins⁺ particles (normalized to total pan-tetraspanins⁺ particles) in plasma-EVs among different groups. Data are shown as the mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons test was applied. *p < 0.05. **p < 0.01. Each dot in D and F represents a single image captured from three biological replicates by Nanoimager (ONI). At least 10 images were captured for each group

Finally, we performed super-resolution imaging of single plasma-EV by staining with fluorescent antibodies against pan-tetraspanins (CD9/CD63/CD81) and AD-related markers (APP and APOE). Our analysis demonstrated that the proportion of APP⁺ (purple) and pan‐tetraspanins⁺ (cyan) double‐positive EVs was higher in the qEV + CCS group (24.63%) compared to qEV alone (15.02%) (Fig. 6C-D). Similarly, a statistically significant increase in the APOE⁺ (purple) and pan‐tetraspanins⁺ (cyan) double‐positive EV subpopulation was observed in qEV + CCS compared to both qEV and CC + qEV (Fig. 6E-F). Overall, our data confirmed the improved enrichment of different EV subpopulations by the qEV + CCS method, supporting its potential as a valuable tool for EV-based blood biomarker development.

Discussion

The isolation of EVs from blood is crucial for advancing our understanding of the biological functions of circulating EVs and developing them as reliable disease biomarkers. However, isolating EVs from blood is challenging due to their overlapping physical and biochemical characteristics with lipoproteins and protein complexes, which complicates their separation [43, 44]. Currently, the commonly used isolation methods, such as UC [45], ultrafiltration [46], and Sect. [47], often lead to significant contamination of EV samples with plasma proteins. Therefore, several new strategies have been developed to improve EV purity. For example, the combination of asymmetric flow field-flow fractionation (AF4) with density cushion ultracentrifugation has been shown to effectively isolate EVs from human blood with reduced contaminants [16]. A Superose 6-based SEC approach has demonstrated the capability to separate small EVs from biofluids with minimal soluble protein and lipoprotein contamination [48]. These methods may require extensive labor and technology, whereas our approach uses readily available, standardized reagents, potentially increasing accessibility and ease of implementation. In this study, we provided a simplified and efficient method for isolating plasma-derived EVs by combining SEC with Capto Core 700 beads. Our findings reveal the potential of utilizing Capto Core beads to effectively remove abundant plasma proteins, significantly enhance EV purity and enable comprehensive coverage of the plasma-EV proteome (Table 1). This optimized method provides a practical and scalable approach for isolating high-purity EVs, supporting their application in biomarker discovery research.

Table 1.

Summary of plasma EV isolation methods using modified SEC and Capto core 700

The biophysiological characterization of EVs isolated from qEV + CCS approach in our study highlights a substantial reduction in soluble protein contaminants with a marked enrichment of EVs. This aligns with findings from other recent studies, which have demonstrated the advantages of employing Capto Core beads to remove soluble proteins from EV preparations [20, 21]. Notably, we observed that a brief 5-min incubation of SEC-derived EV fractions with Capto Core beads was sufficient to achieve substantial enrichment of EV proteins and depletion of soluble proteins. However, our data also indicate that extended incubation with Capto Core 700 beads leads to a marked decrease in EV yield, despite effective removal of contaminants. This suggests that prolonged incubation times may compromise EV recovery, emphasizing the need for optimized incubation durations to balance EV purity and yield when employing Capto Core beads for EV isolation.

Additionally, we comprehensively compared the plasma-EV proteome from different isolation methods by DIA-MS. The proteomics analysis demonstrated that the qEV + CCS protocol consistently outperforms other tested methods in the number of proteins identified from human plasma-EVs, despite some variability due to the limited number of samples herein. Consistently, we observed the qEV + CCS approach not only enriches EV-specific markers such as tetraspanins and TSG101 but also expands the diversity of EV-associated proteins, while reducing levels of abundant plasma proteins (e.g., ALB, IGG, and APOA1). Despite the classic plasma proteins being depleted by over 90% using the qEV + CCS method, a limitation remains as certain plasma proteins persist as among the most abundant in our samples. This reflects the inherent challenge in enriching plasma EVs, as lipoprotein particles, which share similar size and density characteristics with extracellular vesicles, frequently co-enrich with EV fractions [49]. For example, APOB100, an apolipoprotein predominantly associated with low-density lipoprotein (LDL) and very low-density lipoprotein (VLDL) particles, which have diameters comparable to extracellular vesicles, remains enriched in samples processed with the qEV + CCS method. Further improvement of this method is necessary to enhance the specificity and purity of plasma EV isolation.

Furthermore, our findings indicate that EVs isolated by the qEV + CCS method are enriched with proteins associated with a range of brain cell types, such as endothelial and glial cells, as well as proteins associated with neurodegenerative diseases, including APP [50], APOE [51], and CLU [52]. The increased representation of APP⁺ and APOE⁺ plasma-EVs in the qEV + CCS group was validated by dSTORM-based single EV analysis. To be noted, APOE and β-amyloid protein are known to interact with EVs, as demonstrated in the study by Van Niel et al. [4], and could be selectively associated with EV membranes. However, the co-enrichment of APOE alongside APOC1-4 and APOB may also suggest partial retention of LDL particles during isolation. The pathway and protein network analyses of enriched proteins isolated by the qEV + CCS method further support its superiority over other isolation methods, revealing distinct protein clusters which are involved in ribosome assembly, metabolic pathway, cell-matrix adhesion, spliceosome and synapse. This increased proteomic coverage highlights the potential of the qEV + CCS method for high-throughput analyses of the plasma EV proteome. With growing interest in utilizing plasma EVs as non-invasive biomarkers for brain disorders [14, 53], our study reinforces the utility of the qEV + CCS method in accelerating biomarker discovery.

Conclusion

In summary, our study establishes the qEV + CCS method, which combines SEC with Capto Core 700 beads, as a rapid and effective approach for isolating high-purity plasma-derived EVs. This method significantly enhances the detection of EV-specific and disease-relevant markers, minimizes plasma protein contamination, and captures a broader range of EV subtypes. Our findings suggest the qEV + CCS method as a promising tool for EV-based biomarker research, particularly in neurodegenerative diseases and other conditions where reliable circulating biomarkers are urgently needed.

Materials and methods

Plasma sample acquisition and Preparation

Plasma from heathy individuals (n = 3) were collected in EDTA tubes at Mayo Clinic center. For each preparation, a starting volume of 500 µL of plasma was thawed on ice, followed by thawing for 4 min at 37 °C, and treating with 5 µL protease and phosphatase inhibitor mixture (Cat# 78443, ThermoFisher Scientific, MA, USA). The prepared plasma sample was first centrifuged at 1,500 × g for 10 min at 4 °C. The supernatant was then centrifuged at 5,000 × g for 20 min at 4 °C. The supernatant was collected as neat plasma and subjected to EV separation.

Plasma-EV isolation by SEC (qEV)

qEVoriginal 70 nm columns (IZON Science, Christchurch, New Zealand) were used to separate enriched EV fractions from plasma as previously described [13]. Briefly, 500 µL prepared plasma sample was loaded to the column, followed by elution with double-filtered PBS (dfPBS, filtered using a 0.22 μm pore-size filter [#SLGP033RS, EMD Millipore, USA]). Fractions 7, 8, 9 and 10 were pooled as EV-enriched fractions based on the high particle number measurements and relatively low soluble protein content (refer to Supplementary Fig. 1). The combined fractions were concentrated with Amicon Ultra centrifugal 100 kDa filters (Cat# UFC510008, Millipore Sigma) to a volume of 50 µL.

Plasma-EV isolation by CC + qEV

In the CC + qEV (Capto Core 700 beads followed by qEV column) method, 500 µL neat plasma was mixed with 100 µL of pre-washed Capto Core 700 (Cat# 17548101, Cytiva) resin beads overnight (16 h) at 4 °C. The resulting sample was transferred to a Corning Costar Spin-X 0.45 μm centrifuge tube filter (Cat# CLS8163, Sigma-Aldrich) and centrifuged at 12,000 x g for 5 min at 4 °C to remove the remained resin beads. The plasma sample was further subjected to EV isolation by qEV columns as described above, with fractions 7–10 collected. The pooled fractions were concentrated with Amicon Ultra centrifugal 100 kDa filters to a final volume of 50 µL.

Plasma-EV isolation by qEV + CCL

In the qEV + CCL (qEV followed by 1 h incubation with Capto Core 700 beads) method, the plasma-EVs were first isolated by qEV column as describe above. Next, the 500 µL EV fractions were mixed with 100 µL of pre-washed Capto Core 700 resin beads for 1 h at room temperature to remove the soluble proteins. The sample was then transferred to Corning Costar Spin-X 0.45 μm centrifuge tube filter and centrifuged at 12,000 x g for 5 min at 4 °C. The flowthrough of the sample was concentrated with Amicon Ultra centrifugal 100 kDa filters to a final volume of 50 µL.

Plasma-EV isolation by qEV + CCS

In the qEV + CCS (qEV followed by 5 min incubation with Capto Core 700 beads) method, the 500 µL EV fractions resulting from qEV isolation were mixed with 100 µL of pre-washed Capto Core 700 resin beads for 5 min at room temperature to remove the soluble proteins. The sample was then transferred to Corning Costar Spin-X 0.45 μm centrifuge tube filter and centrifuged at 12,000 x g for 5 min at 4 °C. The flowthrough of the sample was concentrated with Amicon Ultra centrifugal 100 kDa filters to a final volume of 50 µL.

Plasma-EV isolation by mag-net protocol (MagReSyn)

In the Mag-Net method, MagReSyn SAX beads (Cat# MR-SAX002, Resyn Biosciences) were prepared and equilibrated according to the manufacturer’s instructions [23]. 500 µL neat plasma was mixed with 25 µL equilibrated beads for 1 h at room temperature and the membrane-bound particles were lysed using 50 µL cold RIPA buffer (Cat# 89900, Thermo Fisher Scientific) supplemented with a protease and phosphatase inhibitor mixture.

Protein extraction and concentration

Neat plasma and plasma-derived EVs isolated by different methods were lysed in cold RIPA buffer supplemented with a protease and phosphatase inhibitor mixture and sonicated for 2 min. Protein concentration was determined by using Micro Bicinchoninic acid (BCA) assay (Cat# 23235, Thermo Fisher Scientific). All assays were incubated at 60^◦C for 30 min. The absorbance of the protein was read using a Biotek Synergy Mx plate reader at 562 nm.

Cryogenic electron microscopy (Cryo-EM)

The morphology of isolated EVs was analyzed using cryogenic electron microscopy (Cryo-EM). Briefly, 5 µL of each EV-enriched sample was applied onto an EM carbon grid (Quantifoil), which was subsequently blotted for 3s at 88% humidity and flash-frozen into precooled liquid ethane using a Vitrobot Mark IV (FEI, USA). This procedure embedded the samples in a thin layer of amorphous ice, preserving them in their native state and providing protection against radiation damage. The samples were examined using an FEI Titan Krios electron microscope equipped with a Gatan Bioquantum K3 direct electron detector (Thermo Fisher Scientific) with the Leginon software for automatic data acquisition in a counting mode [54]. Images were captured using the Falcon II DED at magnifications of 18,000x and 37,000x, with a defocus value in the range of − 2 to − 5 μm. The accumulated total dose per image did not exceed approximately 50 e−/Ų.

Nanoparticle tracking analysis (NTA)

The size distribution and concentration of EVs were analyzed by nanoparticle tracking analysis using a NanoSight NS300 instrument (Malvern, Worcestershire, UK) as previously described [3]. Plasma-derived EVs from was prediluted in filtered PBS to achieve a concentration within the 107 to 108 range for optimal analysis. For each sample, 700-µl diluted EVs were injected into the sample-carrier cell, and the particles were automatically tracked and sized using Brownian motion and diffusion coefficients to record four videos of 30 s. The parameters were set at 22.5° ± 0.5 °C for detection level 5 and camera levels 13 to 15. The size mode (nanometers), mean size (nanometers), and concentration (particle number per milliliter) of the EVs were calculated by combining the data from the four records.

Nano-flow cytometry (NanoFCM)

Nano-flow cytometry (Flow Nanoanalyzer; NanoFCM Inc, Xiamen, China) was applied to analyze fluorescently labeled single particle. This Flow Nanoanalyzer provides high measures of fluorescence, size, and concentration by using the concentration standard/quality control (QC) beads (S16M-Exo), a size standard, and a blank as described in the protocol [55]. EV samples were first diluted to a concentration of 1 × 10¹⁰ particles/mL in dfPBS. 0.5µL 20nM Mem-Glow 488 fluorogenic membrane dye (Cat#MG01-02, Cytoskeleton) was incubated along with 4.5 µL of EV sample. The EV sample was mixed and vortexed for a duration of 5 s and incubated at 4^oC in the dark overnight. Following incubation, 1 µL of labeled sample was diluted at 1:100 in dfPBS. The sample was loaded and boosting pressure was applied for 45 s before switching to sampling pressure and recording for 2 min. The EV subpopulations were analyzed by using the auto-threshold function.

Super-resolution microscopy (nanoimager)

Super-resolution images of EVs were acquired using dSTORM from ONI (Nanoimager S, Oxford Nanoimaging, UK) equipped with a 100×, 1.4 numerical aperture oil immersion objective. 1 µL of EV sample was diluted in 3 µL dfPBS and incubated overnight at 4^oC with pan-tetraspanin antibodies cocktail consisting of 1 µL each of anti-CD9‐CF 488, CD63-CF 568 and CD81‐CF 647 (all from EV‐MAN‐1.0, ONI) to assess the positivity of different tetraspanins, or fluorescein isothiocyanate (FITC) anti-human CD9 (HI9a, catalog no. 312104), FITC anti-human CD81 (5A6, catalog no. 349504), FITC anti-human CD63 (H5C6, catalog no. 353006), Alexa647 anti–β amyloid (catalog no. 800718, all from BioLegend) and Alexa647 anti-APOE antibody (#ab196194, Abcam). Labeled EV samples were immobilized on the house-made TIM4⁺ cover-glasses according to our published protocol [27]. Three thousand five hundred total images were recorded for localization mapping using 40, 50, and 40% power on the 488-, 561-, and 640-nm lasers, respectively. Colocalization analysis of the different molecules in a single EV was proceeded with the CODI online analysis platform (https://alto.codi.bio/). Localization clusters with more than 20 individual molecules were regarded as EVs. EV clusters with Skew ranging from 1 to 2.9 and Circ ranging from 0.43 to 1 were selected for further analysis. A marker was considered as positive in a single particle when more than three individual molecules were detected in the same channel within a radius of 150 nm around the center of a cluster. Five fields of view were acquired from each sample for quantitative analysis.

Immunoblotting

Equivalent amounts (2 µg) of EVs were measured by the microBCA kit and denatured in Laemlli sample buffer (Cat# 1610737, BioRad) and all the samples were loaded on 4–20% TGX stain-free SDS-PAGE gels (Cat# 4568096, BioRad) or normal SDS-PAGE gels (Cat# 4561096, Bio-Rad) and then electro-transferred to 0.45 μm nitrocellulose membranes (Cat# 1620115, Bio-Rad). The membranes were blocked with 5% non-fat milk (Cat# 9999 S, Cell Signaling Technology) and incubated with primary antibodies overnight at 4℃. The following primary antibodies were utilized for immunoblotting: anti-CD9 (1:500, #CBL162, Millipore), anti-Alix (1:400, #Ab76608, Abcam), anti-FLOT1(1:500, #610820, BD Biosciences), anti-Annexin A2 (ANXA2, 1:200, #ab178677, Abcam) and anti-apolipoprotein A1 (APOA1, 1:100, #sc-69755, Santa Cruz). The membrane was further incubated with appropriate horseradish peroxidase (HRP)-labelled secondary antibodies (anti-mouse IgG, HRP-linked antibody, #7076S; anti-rabbit IgG, HRP-linked antibody, #7074S, Cell Signaling Technology) for 1 h, and immunoreactivity was captured using enhanced chemiluminescence solutions (Millipore, WBKLS0100).

For silver staining, the protein gels were fixed and stained according to the manufacturer’s instructions (Cat# 24612, Thermo Fisher Scientific). All the images were detected in the Bio-Rad imaging station.

Data-independent-acquisition mass spectrometry (DIA-MS)

The equal amount of proteins (5 µg) from neat plasma and isolated plasma-derived EVs were lysed in cold RIPA buffer supplemented with a protease and phosphatase inhibitor mixture. The protein samples were fresh frozen on dry ice and subsequently dried using a Speed-Vac. Next, the proteins were trypsin-digested using the S-Trap™ Micro Spin protocol (PROTIFI). Briefly, following lysis (10% SDS in 100 mM triethylammonium bicarbonate (TEAB)), the proteins underwent reduction by 10mM TCEP and alkylation by 20mM Iodoacetamide. Following an overnight digestion at 37 °C, 40 µL of 50 mM TEAB elution buffer was added and spun at 4000 rpm for 1 min. This was followed by a 1-minute spin with 40 µL of 0.2% Formic acid (FA) and, lastly, 40 µL of 50% acetonitrile (ACN). The elution was then pooled together and dried using a Speed-Vac. The digested samples were reconstituted with 20 µL of 0.1% FA in 5% ACN, followed by centrifugation at 16,000 rcf for 15 min. Subsequently, 18 µL of the sample was transferred to non-binding HPLC vials.

The samples (4 µL injection) were analyzed on a TimsTOF Pro2 (Bruker) mass spectrometer, which was coupled to a nanoElute LC system from Bruker. For all analyses, a Bruker Fifteen Pepsep column (150 mm x 75 μm id, 1.9 μm, 120 Å pore size) was utilized. A 60-minute gradient was employed for data-dependent-acquisition (DDA)-PASEF, and a 30-minute gradient used for DIA-PASEF analysis, with a flow rate set at 500 nL/min. The captive nano-electrospray voltage was maintained at 1600 V, using one column configuration (no trap) and a solvent composition of Solvent A: 0.1% FA in water and Solvent B: 0.1% FA in ACN.

DDA-PASEF method: The method consists of 10 MS/MS PASEF scans per topN acquisition cycle, with ramp and accumulation times of 100 ms, covering an m/z range from 100 to 1700 and an ion mobility range (1/K0) from 0.70 to 1.30 V s/cm^2. Collision energy settings followed a linear function of ion mobility, ranging from 20 eV at 0.6 V s/cm^2 to 59 eV at 1.6 V s/cm^2, using default parameters. Calibration of the instrument was performed using three ions from the ESI-L Tuning Mix (Agilent) (m/z 622, 922, 1222).

DIA-PASEF method: Samples were acquired with a method consisting of 22 mass width windows (40 Da width, from 275 to 1155 Da) with 1 mobility windows each covering the ion mobility range (1/K0) from 0.70 to 1.30 V s/cm2. These windows were optimized with the Window Editor utility from the instrument control software (timsControl, Bruker) using one DDA-PASEF run acquired of the analyzed samples. Collision energy setting and calibration were the same as DDA-PASEF analysis.

Proteomics analysis

The direct DIA + workflow in Spectronaut v18.2 (Biognosys) was used for analyzing the DIA-PASEF dataset with no need to build a library from the DDA runs. The raw data files from Bruker were converted to HTRMS files automatically right after acquisition via Spectronaut HTRMS Converter program. Data were searched against a reviewed Human (Swiss-Prot) FASTA database (downloaded on 08/30/2023). The default factory settings were used for the Pulsar search and DIA analysis (including Trypsin/P as enzyme, 7–52 peptide length range; Oxidation of Me and acetylation of Protein N-terminal as variable modifications, carbamidomethyl of C as fixed modification, and 1% FDR for PSM, peptide and protein identification). Automatic cross run normalization and Protein LFQ method were used for protein quantification, and quantity was determined at MS2 level using the area of extracted chromatogram traces.

Proteins identified in at least one out of three biological repeats were selected for further analysis. The visualization of UMAP, heatmaps and pathway bubble plots were generated using the online tool (Sangerbox 3.0) [56]. The protein expression rank and bar plots were generated using Graphpad prism 9. Functional pathway analyses including Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes (KEGG) were performed using the DAVID online resources [57, 58] with false discovery rate (FDR) corrected P value < 0.05. Cell-type enrichment analysis of EV proteins was performed using the WebCSEA online dataset (https://bioinfo.uth.edu/webcsea/index.php) [59]. Neurodegenerative disease enrichment analysis of plasma-EV proteins was performed using the Enrichr online tool by selecting the corresponding GWAS library (https://maayanlab.cloud/Enrichr/).

The enriched proteins in qEV + CCS group were mapped to GO terms using the STRING software. This mapping was performed considering high confidence neighbours and the proteins were ranked based on fold change values [60]. The top clusters of protein nodes were conducted by k-means depend on their close and frequent and labelled by their prominent pathways [61].

Statistical analysis

All data are presented as the mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism (version 9.0). Datasets were analyzed for significance using ordinary one-way ANOVA with Tukey’s post hoc test or Kruskal-Wallis nonparametric ANOVA with Dunn’s post hoc test if the data failed to pass normality tests of Shapiro-Wilk. Statistical significance was defined as p value < 0.05. P values in functional enrichment analyses were adjusted for multiple comparisons using FDR correction with the Benjamini-Hochberg method where indicated.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Y.Y., Z.Z. and T.I. designed research; Y.Y., Z.Z., S.C., S.N., P.V., B.M, S.M. and N.R. carried out experiments; Y.Y., Z.Z., S.C., S.N. and B.M. analyzed data; Y.Y., S.I. and T.I. provided the oversight and suggestion on data interpretation; M.G.P. provided advice on mass spectrometry; M.B. and P.M. provide human samples and interpret data; Y.Y., Z.Z. and S.C. wrote the manuscript with input from co-authors; Y.Y. and T.I. supervised the study and contributed to manuscript preparation and editing. All authors have read and approved the final version of the manuscript.

Funding

This work is supported by CurePSP Research Grant-691-2024-01-Pipeline (YY) and funded in part by Florida Department of Health Ed and Ethel Moore Alzheimer’s Disease Research Program-22A04 (YY), Alzheimer’s Association AARF-22-918114 (YY), Cure Alzheimer’s Fund (TI, SI), NIH R01 AG054199 (TI), R01 AG054672 (TI), R01 AG066429 (TI), R01 AG067763 (TI), R01 AG072719 (TI), RF1 AG082704 (SI, TI), and R01 AG079859 (SI).

Data availability

The data presented in this study is available in this article and supplementary material.

Declarations

Ethics approval and consent to participate

This study was exempt from human subjects as assessed by the Mayo Clinic Internal Review Board. Biosafety protocols were approved by the Mayo Clinic Institutional Biosafety Committee.

Consent for publication

Not applicable.

Informed consent

Not applicable.

Competing interests

The authors declare no competing interests.