Plasma-Derived Extracellular Vesicle Proteomics

Abstract

Extracellular vesicles (EVs) are nanometer-scale lipid bilayer-enclosed particles released by cells under physiological and pathological conditions. Their molecular cargos including proteins can reflect the chemical composition and physiological state of the parent cells, carrying signatures of health and disease. As such, EVs are valuable tools for biomarker discovery and mechanistic studies. Among them, plasma-derived EVs (pEVs) are particularly promising, as sampling plasma allows to capture EVs from virtually all the tissues and organs. The minimally invasive nature of plasma collection further enhances the diagnostic and therapeutic potential of pEVs. Proteomic profiling of pEVs enables the identification of disease specific EV-biomarkers. However, the complexity of plasma, with high levels of abundant proteins and a large EV heterogeneity, presents challenges for pEV proteomics. Mass spectrometry (MS) has emerged as the preferred state-of-the-art analytical tool for pEV studies due to its non-biased ability to characterize thousands of proteins in an experiment and its ability to identify low-abundant EV proteins. Here, a comprehensive overview of the advancements in MS-based pEV proteomics in recent 5 years are presented with a focus on three key areas: sample preparation methodologies, MS-based approaches for protein identification and quantification, and description of pEV studies for basic and disease research. Technical advancements enable greater proteomic details from pEVs, enhancing biomarker discovery, elucidating disease mechanisms, and advancing an understanding of EVs’ biological roles.

Graphical abstract

1. Introduction

Extracellular vesicles (EVs) are nanometer-sized, lipid membrane-enclosed particles released by many cells under both physiological and pathological states.^1–3 EVs display a diverse range of sizes (large EVs are defined as being >200 nm and small EVs <200 nm) and biogenesis (exosomes from the endosomal pathway, microvesicles from direct plasma membrane budding and several other processes).⁴ EVs play significant roles in intercellular communication by transporting proteins, metabolites, and nucleic acids to recipient cells, thereby influencing a large variety of physiological processes such as immune responses, tissue repair, inflammation, and disease progression.^5–13 Importantly, the protein cargo of EVs reflects their cell of origin and cell state, making them attractive candidates for biomarker discovery, mechanism studies, and also understanding the roles of EVs in biological functions in health and disease.⁶

EVs are secreted into extracellular space by most cell types and are found in most biological fluids and conditioned cell culture media^15–17. Among various fluids, plasma-derived EVs (pEVs) are particularly notable. pEVs, reported in in 1967 by Peter Wolf as “platelet dust”¹⁸, are a heterogeneous mixture of vesicles originating from endothelial cells, circulating blood cells, megakaryocytes, and a wide range of tissue cells that release EVs into the bloodstream. Due to the systemic circulation of plasma, sampling plasma allows the capture of EVs from virtually all tissues and organs. This provides a snapshot of an individual’s physiological and pathological states. The accessibility, abundance, and stability of plasma-derived EVs further underscore their potential as a rich source for biomarker discovery and disease mechanism elucidation^19–26. The widespread interest in pEVs is reflected in the literature, where studies on pEVs account for approximately 15% of annually published EV proteomics research (Figure 1).

Figure 1.

Publications in EV proteomics annually since 2010 retrieved from a PubMed search in January 2025 using the terms “EV+Proteomics” (blue), and “Plasma+EV+Proteomics” (pink).

Despite their significant promise, analyzing EV proteins from plasma presents challenges due to the complex nature of plasma and the heterogeneity of the pEV populations. Blood contains a wide array of soluble macromolecules (such as lipoproteins), intact and fragmented cells, and non-vesicular nucleic acids that closely resemble EVs in terms of size, density, and composition. These similarities make it difficult to fully separate EVs from non-EV particles using current isolation techniques, including ultracentrifugation, size-exclusion chromatography, and density gradient centrifugation.²⁷ These isolation challenges are compounded by the extraordinarily wide dynamic range of protein concentrations in plasma, where just a few high-abundance proteins—such as albumin, immunoglobulins, and fibrinogen—account for more than 75% of the total protein content.^28,29 These dominant proteins can suppress the detection of low-abundance EV proteins and introduce significant analytical interference.

Just as is true with other applications of proteomics, mass spectrometry (MS), with its high sensitivity, dynamic range, and multiplexing capabilities, has emerged as the principal analytical platform for EV proteomics.³⁰ To improve proteome depth beyond what is achievable with standard workflows, recent efforts have also emphasized the value of prefractionation, enrichment strategies and MS acquisition strategies to reduce sample complexity and enhance detection of low-abundance proteins with the presence of co-isolated non-EV contaminants.

In this review, we summarize recent advances in pEV proteomics, with an emphasis on MS-based approaches published in the past five years. We begin by discussing critical aspects of sample preparation, including EV isolation—previously addressed in MISEV 2018²⁷ and MISEV 2023⁴ guidelines but revisited here specifically in the context of downstream MS proteomic workflows—along with plasma protein depletion and prefractionation strategies, followed by protein extraction and digestion methods. We then describe MS acquisition and quantification approaches relevant to pEV analysis and conclude by highlighting recent applications of pEV proteomics in disease research.

2. Sample preparation methodologies

Blood is a highly complex biofluid, containing over 10,000 distinct proteins with concentrations spanning a dynamic range of at least 12 orders of magnitude. This vast dynamic range poses a significant challenge for the downstream analysis, particularly when targeting low-abundant EV proteins. To address these challenges, careful optimization of sample preparation workflows is essential. Key steps include improving EV isolation, protein preconcentration, and maximizing the efficiency of protein extraction and enzymatic digestion.

2.1. EV isolation

The selection of EV isolation technique is critical, as it affects both the specificity and recovery of pEVs, thereby influencing downstream MS analysis. Common EV isolation methods summarized in MISEV2023⁴ include precipitation, differential ultracentrifugation (UC), size-exclusion chromatography (SEC), filter concentration, density gradients, and immunoprecipitation. Each approach has distinct advantages and limitations, and combining multiple techniques often yields higher-quality EV preparations by leveraging the strengths of each method. Figure 2 illustrates the tradeoffs between recovery and specificity among commonly used isolation techniques.

Figure 2.

Position of EV separation and concentration methods on a recovery (yield) versus specificity grid. Adapted from Welsh et al.⁴ under the Creative Commons Attribution License (CC BY). License details: https://creativecommons.org/licenses/by/4.0/.

High-yield methods like UC are widely used due to their simplicity, accessibility, and ability to recover large quantities of EVs. However, these methods often co-isolate non-vesicular contaminants, including serum proteins, protein aggregates, and apoptotic vesicles.^33,34 In contrast, high-specificity methods, such as density gradient centrifugation, reduce background proteins but may exclude low-abundant proteins due to lower recovery rates. UC and SEC are two of the most used techniques for isolating EVs, each with its own advantages and drawbacks. Baranyai et al.³⁵ demonstrated that SEC isolation yields pEVs with high purity due to reduced albumin contamination. However, Takov et al.³⁶ challenged this finding by conducting a detailed comparison of isolation techniques. They concluded that SEC typically yields higher particle-to-protein ratios compared to UC, but is more susceptible to contamination from soluble factors, including lipoproteins. Furthermore, the author noted that other isolation techniques such as density gradient centrifugation are not capable of completely separating EVs from lipoproteins³⁷, particularly those containing apolipoprotein B. This suggests that achieving complete separation of EVs from contaminants in plasma samples remains a difficult and complex task due to the inherent nature of the plasma matrix. In addition to the specificity of the EV isolation methods, the choice of technique is often influenced by the available sample volume. SEC is particularly well-suited for smaller volumes (500–1000 µL)³³, whereas UC is more effective for processing larger volumes. When studying pEV proteomics via MS, common purification strategies include UC alone, SEC alone, a combination of UC with ultrafiltration, or SEC followed by ultrafiltration. While combining isolation techniques can increase EV purity, it may come at the cost of reduced recovery rates, increased variability in results and longer separation procedures. The strategy to utilize combination of EV isolation techniques are detailed in Turner et al³³. Notably, hemolysis, or the rupture of red blood cells during sample collection and processing, introduces hemoglobin into the plasma, contributing additional contaminants that may interfere with downstream analyses. Proper pre-processing steps such as centrifugation to remove blood cells are crucial to minimize hemolysis and prevent the formation of EV-like particles during repeated freeze-thaw cycles.

2.2. EV enrichment

Despite the implementation of common EV isolation strategies, the inherent complexity of plasma continues to pose significant challenges, as numerous plasma constituents can co-elute with EVs and interfere with downstream analyses. These include cellular contaminants such as blood cells, platelets, lipoproteins, as well as high-abundance soluble proteins such as serum albumin, immunoglobulins, fibrinogen, and ribonucleoprotein aggregates.^4,31,38 For example, lipoproteins are particularly challenging to remove due to their size overlap with EVs. While size- and density-based methods like SEC and density gradient centrifugation are partially effective, lipoproteins often co-purify with EVs. Additional strategies, such as immunoaffinity-based depletion targeting apolipoproteins or combining ultracentrifugation with chemical precipitation, have shown promise in selectively removing lipoproteins, thereby improving EV purity.^39,40 Furthermore, differences in input volume present challenges between species. While human plasma EV isolation typically begins with 0.5–1 mL of plasma, mouse studies are often limited to 100–200 µL per animal, necessitating the use of efficient and low-input enrichment techniques. In response to these limitations, several newer enrichment strategies have been developed to improve the selectivity, speed, and reproducibility of pEV capture.

One such approach is the Tim4-phosphatidylserine (PS) affinity method developed by the Hanayama group.^41,42 Coupling Tim4 to magnetic beads or to automated liquid-handling systems allows for targeted EV capture from plasma with high specificity. Compared to antibody-based methods, Tim4-PS affinity enrichment offers broader coverage across EV subtypes without the bias toward surface protein expression. It also avoids detergent-sensitive epitopes, making it more suitable for proteomics. EVTRAP (Extracellular Vesicles Total Recovery and Purification) developed by the Tao group is also.⁴³ EVTRAP utilizes functionalized magnetic beads modified with a combination of hydrophilic and lipophilic groups, which exhibit unique affinity toward lipid-coated EVs. This method enables fast and reproducible EV isolation from complex body fluids.^43–48 However, due to its lipid-targeting mechanism, EVTRAP may co-enrich lipoproteins and protein aggregates through nonspecific interactions.

Recently, the development of aptamer technology has seen impacts on the enrichment of EVs.^49,50 Unlike antibodies, aptamers are single-stranded oligonucleotides that fold into defined 3D structures and can be engineered to reversibly bind EV surface markers under physiological conditions. This unique property allows aptamer-based methods to isolate, concentrate, and release EVs gently and efficiently—overcoming limitations associated with harsh elution conditions and cost in antibody-based affinity isolation. For example, Zhang et al. demonstrated reversible capture of CD63+ EVs using a DNA aptamer-streptavidin magnetic bead system, enabling the release of intact EVs by structure-disrupting antisense oligonucleotides.⁵¹

Another innovative method is the Mag-Net approach⁵², which employs hyper-porous strong anion exchange magnetic microparticles to selectively capture membrane-bound vesicles based on surface charge. The particles create a sieving effect in plasma, allowing for the enrichment of EVs while reducing soluble protein contaminants. The Mag-Net method is reproducible, inexpensive, and requires <100 μL plasma input, making it suitable for high-throughput clinical applications. In addition to these biochemical affinity- and charge-based methods, microfluidic separation is emerging as a novel and increasingly adopted platform for EV isolation.^53,54 These devices can exploit physical characteristics such as size, charge, deformability, and immunoaffinity to achieve precise EV separation with minimal sample volume and high reproducibility. As emerging technologies, these affinity-, charge-, and microfluidics-based enrichment methods hold significant promise but require further validation and standardization to ensure their robustness across diverse biological samples and applications.

Emerging studies have also begun to define core molecular features of circulating plasma EVs that may enable the development of future antibody-based capture strategies. For example, Rai et al.⁵⁵ performed deep proteomic and lipidomic profiling of over 140 human plasma samples and identified a conserved panel of 182 EV-enriched proteins and 52 hallmark lipids. Notably, they characterized the EV surfaceome, mapping 151 surface proteins and highlighting ADAM10 and PS (36:1) as robust markers that distinguish EVs from non-EV plasma components. These findings offer valuable targets for selective EV capture and a foundational reference for advancing EV-based diagnostics. Building on this concept, the ability to enrich tissue-derived EVs using tissue-specific surface markers is particularly attractive, as it enables direct, noninvasive monitoring of tissue-specific changes. For clinical biomarker applications, several researchers have reported the feasibility of detecting and isolating EVs using different antibodies from the tissues, including the brain, liver, prostate, and other tissues in body fluids.^56–58

2.3. Plasma protein depletion

Soluble proteins are generally smaller and denser than EVs, allowing for their separation from EVs by SEC, density gradient, centrifugation, or combinations thereof.⁴ However, even after isolating EVs, abundant plasma proteins can often persist and dominate pEV proteomic signals, obscuring the detection of the EV-specific proteins. To address this, abundant protein depletion strategies are frequently employed to further enhance the purity and analytical depth of plasma-derived EV preparations, ensuring more accurate and comprehensive pEV analyses. Abundant protein depletion is typically performed using immunoaffinity columns, which can efficiently remove 3–14 dominant plasma proteins, as demonstrated in several studies.^59–61

2.4. Sample prefractionation

Sample prefractionation enhances proteome coverage by reducing sample complexity prior to MS analysis. Prefractionation can be performed either before or after protein digestion, depending on the method. For example, sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) separates intact proteins by molecular weight and is typically applied at the protein level. In contrast, strong cation exchange (SCX) chromatography, high- or low-pH reversed-phase chromatography, and high-resolution isoelectric focusing (HiRIEF) are peptide-level fractionation techniques applied after digestion, separating peptides based on charge, hydrophobicity, or isoelectric point, respectively. For instance, Sharma et al. employed HiRIEF-LC-MS/MS to perform deep proteome profiling of pEVs and identified 2,076 proteins in pEV fractions from 12 human plasma samples, demonstrating the effectiveness of peptide-level isoelectric focusing for expanding the detectable proteome depth of low-abundance EV samples.⁵⁹

2.5. Protein digestion

Following EV isolation and protein depletion, efficient protein extraction and digestion are critical steps to ensure robust bottom-up proteomic analysis. EVs are encapsulated by a lipid bilayer, which can pose a barrier to efficient protein digestion, as the membrane shields many proteins from protease access. Disrupting this lipid bilayer is therefore a critical step in EV protein extraction, enabling proteases to efficiently digest encapsulated and membrane-associated proteins. Urea-based buffers, commonly used in EV proteomics, play a dual role by denaturing proteins and weakening lipid-lipid and protein-lipid interactions, thus facilitating partial disruption of the bilayer. However, urea alone may not completely disrupt the membrane, and it is often combined with mechanical methods, such as sonication, or with detergents like SDS to ensure complete lysis^33,59. Additionally, Guan et al.⁶² optimized and evaluated the extraction condition for EV proteins and found that the intensity of human serum albumin (HSA) decreased with different degrees when using different concentrations of acetonitrile or methanol. They found 50% acetonitrile or 50% methanol were the optimum solvent to deplete HSA. Eventually, they identified 1290 and 1963 protein groups from human plasma-derived exosomes and microvesicles from eight human plasma samples and showed a 79% overlap between the exosome database and their dataset. Once proteins are effectively extracted, efficient enzymatic digestion is essential to maximize proteome coverage. A two-step digestion strategy has proven advantageous for overcoming the limitations of protease activity under high-denaturant conditions.^59,63,64 Initially, proteins are pre-treated with Lys-C, which remains active in high concentrations of urea (6–8 M) and can cleave proteins into smaller, more accessible fragments. After this initial digestion, the urea concentration is diluted to below 1 M—often with calcium ions added to stabilize enzymatic activity—followed by an overnight digestion with trypsin. This sequential approach ensures efficient and complete digestion while minimizing missed cleavages.

2.6. Post-translational modification enrichment

During EV biogenesis and their involvement in biological processes, proteins undergo diverse post-translational modifications (PTMs), including phosphorylation, glycosylation, and ubiquitination. These PTMs play essential roles in regulating EV formation, signaling functions, and intercellular communication.^65.66Several strategies have been developed for enriching modified peptides from complex EV digests. Antibody-based enrichment methods offer a straightforward and scalable approach. In these workflows, modification-specific antibodies (e.g., anti-phosphotyrosine or anti-acetyl-lysine) are immobilized on solid supports such as magnetic or sepharose beads, allowing selective capture of modified peptides. These methods are flexible across sample volumes and formats, making them attractive for EV-based studies. However, their utility can be limited by moderate selectivity, cross-reactivity, and the availability of high-affinity antibodies for certain PTMs.^66,67

Alternatively, chemical-based enrichment strategies provide reproducible, cost-effective, and often more selective approaches for PTM-specific capture.⁶⁸ For example, immobilized metal affinity chromatography (IMAC) and titanium dioxide (TiO₂)-based materials are widely used for phosphopeptide enrichment due to their robust interaction with phosphate groups.^69–72 These approaches have been successfully applied to EV proteomes to identify phosphosite-specific signaling events and pathways. In addition, emerging chemoselective methods are being developed for less abundant or more labile PTMs such as glycosylation and ubiquitination, although their application to EVs remains more limited.

3. MS approaches for pEV protein identification

3.1. Data acquisition

After isolation, digestion and sample clean-up, the next step is to comprehensively characterize the protein cargos of pEVs. To maximize protein detection, nanoscale liquid chromatograph (nanoLC) is typically preferred, as it enhances the sensitivity for low-abundant EV proteins. On the MS side, advancements in next-generation Orbitraps and quadrupole time-of-flight (QTOF) analyzers have played a pivotal role in enhancing EV proteomics. A more recent development involves ion mobility spectrometry (IMS), particularly trapped ion mobility spectrometry (TIMS) coupled with QTOF instruments. A notable application of TIMS is parallel accumulation-serial fragmentation (PASEF), increases sensitivity and throughput by accumulating precursor ions in the TIMS funnel and sequentially releasing them based on their ion mobility for fragmentation. The additional ion mobility separation dimension effectively preconcentrates ions prior to mass analysis, enabling the detection of low-abundant EV proteins and improve the proteome coverage.

In terms of the MS data acquisition approaches, data dependent acquisition (DDA) and data independent acquisition (DIA) are both widely employed, each offering distinct characteristics. In DDA, the most intense precursor ions are sequentially selected for fragmentation, generating high-resolution MS/MS spectra that enable accurate peptide identifications. The data produced by DDA is relatively straightforward to analyze and less computationally intensive, benefiting from well-established methodologies and a wide range of available protocols and analytical tools. DDA is also well suited to uncover unexpected proteins without relying on predefined spectral libraries. DDA approach has been widely used in pEV proteome studies.^{33,61,62,73,74} Despite its strengths, DDA has limitations in detecting low-abundant EV cargo proteins due to its tendency to favor abundant ions. To address this and improve proteome coverage, strategies such as prefractionation⁵⁹, long-gradient separations, and the integration of additional separation dimensions (such as TIMS) are often employed. For example, by synchronizing precursor ion accumulation with rapid sequential fragmentation, PASEF-DDA enhances sensitivity, acquisition speed, and proteome depth while maintaining high resolution, demonstrated by Garza et al.⁷⁵ on a timsTOF Pro instrument.

On the other hand, DIA captures detectable precursor ions within a specific mass range without bias toward high-abundant proteins. This approach generates comprehensive MS/MS spectra, enhancing peptide and protein identification thus enabling deeper and more consistent proteome coverage. However, the computational demands for DIA data analyses remain high, and it relies on robust spectral libraries for optimal peptide identification. To address this, running a few DDA experiments on representative samples, one can construct a spectral library that serves as a reference database of peptide fragmentation patterns for subsequent DIA analyses. By matching DIA acquired spectra to the pre-built spectral library, the combined strategy enables high reproducibility and enhanced sensitivity for both high- and low-abundant pEV proteins, making it an increasingly popular approach in pEV studies.^76,77 More recently, researchers have explored multiplexed DIA (mDIA), which integrates DIA with isobaric or metabolic labeling techniques to further expand throughput and detection depth.^78–80

Several targeted acquisition strategies have been applied to EV proteomics, demonstrating significant potential for precise and robust quantification. For instance, the Tao group has published multiple studies on salivary EV proteomics using techniques such as multiple reaction monitoring (MRM)⁸¹ and parallel reaction monitoring combined with PASEF (PRM-PASEF)⁸². MRM is a well-established targeted approach that offers excellent specificity and sensitivity for quantifying the tryptic peptides. However, the approach is lower throughput as it only monitors a predefined set of transitions per run, perhaps restricting it use for broader proteomic analyses. Additionally, developing the required MRM assays requires extensive optimization, including careful selection of transitions and collision energies, which can be labor-intensive for larger-scale studies. In contrast, PRM-PASEF offers a versatile and high-throughput solution. By first generating a spectral library via several DDA runs, PRM-PASEF enables precise targeting of specific proteins of interest while allowing for the simultaneous detection of hundreds of peptides in a single run. However, PRM-PASEF requires access to advanced instrumentation like the timsTOF and is computationally more demanding due to the need for spectral library generation and data processing. A notable example by the Tao group utilized PRM-PASEF to study the salivary EV proteome for assessing therapeutic outcomes in oral squamous cell carcinoma (OSCC).⁸² They initially performed DDA-PASEF to compare the salivary EV proteomes of healthy individuals and OSCC patients, identifying biomarkers and assessing differentially regulated proteins. These proteins were then used to construct a targeted library for PRM-PASEF analysis, enabling precise quantification of key protein targets in pre-operative and post-operative samples. While this study focused on salivary EVs, the approach is applicable to pEV quantification studies as well, where the complexity of plasma and its abundant proteins often necessitate targeted strategies for consistent and robust protein detection.

3.2. Quantitative analysis

Quantitative analysis is often required for pEV proteomics, especially in the context of biomarker discovery, where the measurement of protein abundance can be related to disease state. Advancements in label-free quantification (LFQ) and stable isotope-labeled relative quantification have significantly enhanced the utility of MS-based EV studies. LFQ quantifies proteins based on the intensity or peak area of their MS signals without the need for labeling, offering a cost-effective, unbiased, and versatile approach for comparative proteomics. By allowing the direct comparison of protein levels between samples, LFQ facilitates the identification of differentially expressed proteins and provides insights into the biological functions and roles of EVs. However, missing values is an issue when using DDA based LFQ method. This issue arises primarily from the stochastic nature of DDA, where only the most abundant precursor ions are selected for fragmentation, potentially missing low-abundance peptides in highly dynamic and complex samples like pEVs. To address this limitation, data imputation methods have been introduced to estimate and replace missing values with predicted ones using statistical or machine learning models⁸³, such as Bayesian Ridge regression, which leverages probabilistic linear regression for accurate predictions. Another approach to overcome missing values involves using PRM-PASEF, which reduces the occurrence of missing values by guaranteeing the acquisition of comprehensive MS/MS spectra for each targeted peptide, thereby ensuring reliable and consistent detection of selected proteins across multiple samples.

Besides LFQ approaches, chemical labeling with isobaric tandem mass tags, such as isobaric tags for relative and absolute quantification reagents (iTRAQ) and tandem mass tag (TMT) reagents, have also been popular methods of proteomics.^76,84 Using a 10-plex TMT-based LC-MS/MS approach, 672 pEV proteins were identified in Alzheimer’s disease (AD) patients, with 473 overlapping the ExoCarta database, and 335 proteins quantified consistently across datasets (n = 40).⁸⁴ The labeling approach is particularly advantageous in workflows involving sample prefractionation prior to LC-MS analysis, as it allows multiplexed fractions to be run together—substantially improving throughput without compromising proteome coverage, as demonstrated by Sharma et al⁵⁹. However, co-isolation interference during precursor selection may compromise quantification accuracy of TMT experiments, particularly in complex samples like plasma, where high-abundance proteins dominate. This often leads to ratio compression and reduced dynamic range, masking low-abundant EV proteins. Additionally, incomplete labeling, higher sample input requirements, and the high cost of TMT reagents can pose challenges, especially given the small volumes of pEVs. Optimizing workflows with high-resolution instruments and robust depletion strategies is critical to mitigate these issues and improve data quality.

3.3. Post-acquisition analysis

3.3.1. Contaminant identification and removal

Part of the data analysis process must deal with the removal of information from non-EV contaminants. Even if best practices are followed during sample preparation, contaminants are expected in the resulting list of identified proteins. These can originate from the workflow itself (e.g., keratins from skin and hair), from the EV isolation method (e.g., lipoproteins from density gradients), or from the biological matrix (e.g., albumin or immunoglobulins in plasma). While some contaminants may form part of the protein corona and loosely associate with vesicles, their inclusion in downstream analyses can lead to misinterpretation.

To address this ambiguity, standardization of contaminant reporting and removal remains an ongoing need in the field. Different data analysis platforms vary in how contaminants are flagged—MaxQuant automatically annotates common contaminants, whereas Proteome Discoverer requires manual configuration of contaminant filters within the consensus workflow. Tools such as cRAP (common Repository of Adventitious Proteins) and contaminant databases reported by Frankenfield et al.⁸⁵ can be used to systematically exclude background proteins. However, these lists may not be updated regularly and may not fully capture contaminants specific to pEV workflows.

A commonly used approach is to cross-reference identified proteins against curated EV-specific databases such as Vesiclepedia, ExoCarta, etc. Proteins that are consistently identified in EVs across multiple high-quality studies—but absent in common contaminant lists—can strengthen confidence in biological relevance. Moreover, the inclusion of spatial annotations (e.g., lumenal, membrane-bound, corona-associated) when available can help differentiate true vesicle components from loosely associated proteins.

3.3.2. Tissue sources of pEVs

It is possible to infer the tissue origins of pEVs post-acquisition via the detected proteins, thereby enhancing their clinical utility for biomarker discovery. Muraoka et al. introduced a computational strategy combining tissue-specific protein annotated from the Human Protein Atlas (HPA) with protein coregulation analysis across individuals.⁴² Proteins identified in their deep pEV proteomics datasets (>3,000 proteins) were cross-referenced with HPA-defined tissue-enriched proteins, defined as those showing at least four-fold higher expression in a specific tissue relative to others. Using this approach, they identified 120 liver-specific proteins and 34 brain-specific proteins in pEVs. Notably, neuron- and oligodendrocyte-specific proteins such as STXBP1, GPM6A, and GDI1 were detected, which are associated with key neurodegenerative disease pathways. While HPA data are transcript-based and may not perfectly reflect protein abundance in EVs, this post-acquisition strategy offers a promising framework for deconvoluting the tissue origins of EVs and facilitating the discovery of pEV-derived biomarkers, particularly as more comprehensive proteomic references become available. Such advances in characterizing distinct EV subpopulations from blood can significantly expanded their utility in both therapeutic assessment and diagnostics.

4. Plasma-derived EV proteomics for disease study

Building on these methodological advancements, the growing recognition of pEVs as systemic molecular conveyors has fueled extensive research into their potential for biomarker discovery and disease mechanism elucidation. For instance, pEV proteomics is becoming widespread for cardiovascular diseases (CVDs), which involve systemic processes such as inflammation and endothelial dysfunction.⁸⁶ pEVs have been associated with various forms of CVDs,^55,87,88 including myocardial infarction (MI), where they are linked to cardiovascular risk factors and structural changes in atherosclerotic plaques.⁸⁹ Studies by Gidlöf²³ and Rezeli⁹⁰ revealed that the pEV proteome after MI offers additional diagnostic value compared to plasma alone, identifying proteins like chymotrypsin C, proto-oncogene tyrosine-protein kinase SRC, and C-C motif chemokine ligand 17 that were altered in pEVs but not in the plasma. Cheow et al.²⁵ also identified six dysregulated pEV proteins reflecting post-infarct pathways such as complement activation, lipoprotein metabolism and platelet activation, suggesting a role for EV-mediated processes in myocardial injury and healing.

Beyond CVDs, pEV proteomics have been employed to investigate infectious diseases and their systemic effects. For instance, pEVs have been shown to capture systemic features and retain long-term traces of diseases, such as the residual effects of COVID-19²², highlighting their role in coagulation activity and inflammation. Additionally, a study by Jeannin et al.²⁴ revealed significant alterations in the protein content of pEVs in patients infected with human T-lymphotropic virus type 1 (HTLV-1), demonstrating that pEVs from HTLV-1-infected donors contain markers of metabolic and mitochondrial stress.

In cancer research, pEVs have offered unique insights into tumor biology, facilitating the identification of biomarkers for early detection and therapeutic targets for cancer study.^74,76,77 Vykoukal et al.⁷⁴ conducted an in-depth proteomic characterization of pEVs from lung adenocarcinoma patients and identified 108 significant differential proteins between lung adenocarcinoma cases and controls. Beyond global proteomics, PTMs detect in EV proteins—particularly phosphorylation and glycosylation—have also been explored.^8,67,91,92 For instance, Chen et al.⁸ employed label-free quantitative phosphoproteomics and identified 144 phosphoproteins in pEVs that were significantly elevated in eighteen patients with breast cancer compared to six healthy controls. These findings underscore the potential of phosphoproteins in pEVs as promising biomarkers for cancer screening and disease monitoring.

In neurodegenerative research, pEVs have shown great promise as sources of biomarkers for diagnosing complex disorders like AD and dementia with Lewy bodies (DLB). Gamez-Valero et al.²⁶ characterized the pEV proteome from DLB patients compared to age-matched healthy controls, and identified gelsolin and butyrylcholinesterase as differentially expressed proteins between DLB and controls. The study suggested gelsolin may serve as a potential biomarker for distinguishing between DLB and AD. In addition, Zhang et al.⁸⁴ identified 12 differentially expressed proteins in pEVs from AD patients (n = 40) compared to healthy controls (n=40), including a marked decrease in S100A8. Functional assays using Aβ₁–₄₂-induced SH-SY5Y cell models showed that knockdown of S100A8 reduced Aβ aggregation, especially within pEVs, supporting its role in disease pathology. These findings emphasize the utility of pEV proteomics in advancing the diagnosis of neurodegenerative diseases.

pEVs have also been studied in conditions like obesity. Pereira et al.⁹³ found reduced protein diversity in obese pEVs, while control EVs were enriched in protein-folding proteins. They identified four proteins unique to the control state across plasma and gut EVs, suggesting pEVs as potential non-invasive markers of gut health. A shift from glycation to acetylation in chromatin-related proteins in obese gut EVs also points to altered transcriptional regulation during obesity. Previous work from our team demonstrated the potential of EVs in understanding skeletal muscle recovery, particularly through a cell-free strategy utilizing pericyte-derived EVs to enhance antioxidant pathways and extracellular matrix remodeling after disuse and aging.⁹⁴ These findings revealed the therapeutic relevance of EVs in tissue repair, presenting a cellular strategy to recover skeletal muscle after disuse. Building on this foundation, ongoing work in our laboratory and others has advanced the proteomic analysis strategy to study the changes of pEVs in different physiological status including endurance exercise and aging models. Recent studies suggest that pEVs from exercised individuals carry proteomic signatures indicative of enhanced antioxidant defenses and systemic adaptations⁹⁵, while aging-related pEV profiles may reflect the ability for pEVs to induce senescence⁹⁶. This growing body of research underscores the versatility of pEVs in capturing dynamic physiological states and their expanding significance in the fields of aging and exercise biology.

5. Future perspectives

A major challenge in pEV proteomics is the intrinsic heterogeneity of the EV population. Current MS-based workflows typically analyze bulk EV isolates, obscuring whether specific proteins are concentrated within distinct EV subpopulations or more diffusely distributed across the entire pool. This lack of resolution hampers our ability to dissect the functional diversity of EVs and their disease relevance—an issue particularly pronounced in pEVs due to their diverse tissue origins. Although curated databases such as ExoCarta and Vesiclepedia provide insight into potential tissue sources, much about EV heterogeneity remains unclear. Addressing this will require integration of multi-omics strategies that combine proteomics with complementary datasets, such as transcriptomics, lipidomics, and metabolomics, from the same pEV samples. These integrative approaches can reveal subpopulation-specific molecular signatures and link them to disease-relevant biological functions. As high-throughput technologies continue to accelerate data generation, there is a growing need for advanced computational frameworks to handle, interpret, and integrate multi-dimensional EV datasets. Correlation-based, network-based, and machine learning–driven methods are increasingly employed to enable system-level insights that reflect the regulatory interplay across molecular layers.^97,98

Perhaps one of the most exciting emerging frontiers in EV proteomics is the application of single-object MS analysis. Recent developments in single-vesicle analysis can be enabled by innovations in flow cytometry and live and high-resolution microscopy techniques, creating new opportunities for probe EV heterogeneity at the individual EV level and to characterize their physical and molecular properties at nanometer resolution.⁹⁹ After all, the last decade has seen amazing advances for single-cell MS which enables cell heterogeneity to be studied.^100–102 Now examples of single organelle MS are appearing. Of course, a major challenge relates to their small size, low protein content of individual vesicles, surface adsorption and sample loss during transfer. Technologies such as single ion mass spectrometry^103,104 as well as pore-based sequencing^105–107 offer possibilities for large enhancements in protein detection which could become enabling for individual pEV measurements. Lessons from single-cell proteomics are also highly informative. Techniques like ultra-miniaturized sample handling platforms (e.g., nanoPOTS^108–110, proteoCHIP^111,112, nPOP^113,114), ultralow-flow LC, and strategies to minimize surface adsorption have enabled identification of hundreds to thousands of proteins from single cells by reducing sample loss and increasing MS ionization efficiency.¹¹⁵ These systems reduce sample preparation volumes to the picoliter scale, thereby limiting nonspecific adsorption, and use one-pot workflows to avoid losses during transfers. Detergents such as n-dodecyl-β-D-maltoside (DDM) have also been widely adopted to prevent protein loss during processing and are compatible with downstream MS analysis.¹¹⁶ Together, the optimizations including volume minimization, adsorption reduction, and one-pot automation will be essential for extending MS workflows to single-EV analyses.

Figure 3.

A representative study design and workflow for in-depth proteome profiling of pEVs. The EVs were separated from plasma using size exclusion chromatography (SEC) and characterized using Coomassie blue staining, nanoparticle tracking analysis (NTA), bead-based flow cytometry analysis, transmission electron microscopy (TEM) and Western blot. High-resolution iso-electric focusing (HiRIEF) was used to perfectionate the tryptic peptides before LC-MS. Adapted from Sharma et al.⁵⁹ under the Creative Commons Attribution License (CC BY). License details: https://creativecommons.org/licenses/by/4.0/.