The field of human extracellular vesicles (EVs) is challenged by understanding their form and function. Circulating EVs are increasingly recognized as important mediators of intercellular communication and promising sources of highly relevant, minimally invasive biomarkers. Yet, defining what constitutes an EV in complex biofluids such as human plasma remains a persistent challenge.
In a recent Nature Cell Biology study, a team at the Baker Heart and Diabetes Institute, Australia, and collaborating universities addressed this challenge by combining high-sensitivity mass spectrometry, proteomics, surfaceome profiling, lipidomics, and single-vesicle flow cytometry with machine-learning-based data integration. Through extensive biochemical, biophysical, and multi-omics characterization, combining patient samples across different cohorts, the work establishes a molecular reference framework for human circulating EVs.
We reached out to David Greening, Head of Molecular Proteomics at the Baker Heart and Diabetes Institute, and group leader Alin Rai, that both led this study, to learn more about the team’s motivations and challenges – and how integrated multi-omics could improve reproducibility, classification, and translational potential in EV research.
What initially motivated your team to rethink the molecular definition of human circulating extracellular vesicles?
We have known for many years what (proteins and lipids) really makes an EV that are released by cells in culture systems (cells grown in a dish; in vitro), and this knowledge can be largely attributed to mass spectrometry, a powerful technology that can rapidly identify thousands of molecules in an unbiased manner. And for a very long time researchers have tried to gain similar understanding in EVs from humans. Plasma, being highly accessible and most widely studied biofluid, is therefore a prime source of EVs relevant to human biology and clinical utility.
However, our “molecular definition” of human Evs unfortunately remained limited to several hundred proteins, many of which were abundant plasma proteins, and those proteins which we know are functionally relevant to EV biology (e.g., biogenesis, signaling, membrane curvature) at least in culture, remained nowhere to be seen (identified). So, our first question was whether human EVs similarly have a different complexity and architecture of in vitro EVs. What does this mean with respect to transferability and translatability in our EV knowledge between culture systems and humans? However, with our deep knowledge and expertise in mass spectrometry technology we were also fully aware of the limitations of using low purity biological samples if we were to truly dissect what human EVs are made up of.
Point in case, there are trillions of non EV particles in blood for every one million EVs – what this means is insufficient purity would give us a molecular profile that of plasma components and not truly EVs. This was our primary motivation – which is to obtain EVs from plasma of exceptionally high purity, with minimal contamination of non EV particles. This is why we hypothesized that the conventional EV molecular composition in humans needs a careful reassessment
EVs are master communicators of multicellular life – key in orchestrating the complex signaling networks that define physiology. Over the last several decades, EV trafficking and intercellular signaling have emerged as an important mechanism of cell communication. As new concepts have linked EVs to many physiological and pathological processes – and as their presence in blood plasma has become clear – the field of EV research has represented a new paradigm in exploring and translating EV-based therapies and their use as diagnostic, prognostic, and predictive biomarkers.
EVs are being translated presently as clinically relevant and effective tools in detection and monitoring in health and disease as minimally invasive tools from blood or other bodily fluids. Their complex molecular cargo reflects their cells and tissues of origin, and advances in detection sensitivity are increasingly enabling that cargo to be identified and monitored. As a new diagnostic strategy, circulating EVs offer the ability to decode their molecular “language” and uncover new patterns that could identify people with early signs of coronary heart disease – years before symptoms appear. The knowledge established today lays the groundwork for the targeting and detection strategies of tomorrow. By understanding the EV surfaceome (the surface “barcode” displayed), we can begin to engineer “designer vesicles” that mimic the body’s own communication system, and enable real-time, dynamic therapeutic monitoring.
Nevertheless, important questions remained – principally, how to reliably resolve EVs from the complexity of plasma. Conventional (cell-derived) EV markers are too inconsistent to reliably separate vesicles from other plasma components, including abundant proteins, aggregates, and lipoproteins. Furthermore, few studies have performed high-quality EV isolation across large, diverse patient groups while simultaneously analyzing and integrating both protein and lipid data. Using our strategy, these signatures were validated across multiple independent cohorts and EV subtypes, providing not only a technical solution but also a conceptual framework for the field.
Could you briefly explain how your multi-omics workflow was designed?
We employed technologies including high-sensitivity mass spectrometry for protein and lipid identification, surfaceome profiling to capture the surface membrane protein and corona network, and single-vesicle flow cytometry, alongside multiple machine-learning frameworks for comprehensive analysis, integration, and interrogation. As strategies to isolate and understand such circulating vesicles, we used density fractionation enrichment and lipid-based affinity capture, combined with extensive biophysical and biochemical characterization. Our discovery includes a conserved set of 182 proteins and 52 lipids intrinsic to circulating EVs, as well as a panel of 29 proteins and 114 lipids that are non-EV features in plasma (part of the lipoprotein network). Together, these serve as biological markers for EV research in humans.
In the study, we used high-resolution density gradient separation (DGS) to isolate a major EV subtype, known as small EVs, from human plasma. We verified the enrichment strategy and EV identity using a range of biochemical and biophysical characterization methods, ensuring separation of EVs from non-EV particles in plasma. We then generated detailed proteome and lipidome maps, defining EV hallmark features and, in parallel, identifying markers that distinguish non-EV particles. These markers – ADAM10 and PS(36:1) in particular – enable precise differentiation between EV and non-EV particles using machine learning across different human cohorts and enrichment strategies.
Our findings highlight how unbiased multi-omics profiling can identify novel, biologically relevant molecular features in human circulating EVs. By offering biologically grounded criteria for defining EV identity, the work addresses the longstanding challenge of EV purity in plasma isolations. Rather than relying on depletion of known contaminants or inconsistent operational definitions, we integrated multi-omics profiling with machine learning to identify conserved markers with robust classification and translational potential.
Did any of the findings particularly surprise you?
What surprised us the most is the convergence of EV knowledge from culture systems (accrued over the decades) that could be potentially translated to human EVs – including many biogenesis proteins and lipids, membrane curvature, lipid raft and membrane receptors, important cargo loading molecular features, and surface proteins that guide EVs to different cells/organs etc. to be able to confidently identify these proteins and lipids (many of which known to function together) gives us confidence that similar EV biology, pathways, networks, and processes are operational in humans too. This is a huge step in understanding the form and function directly in humans. Such insight revealed that hallmark features such as ADAM10 and PS(36:1) are selectively present on only a subset of EVs, underscoring the heterogeneity of the EV population. This selective distribution highlights the need to better understand how cargo is loaded during biogenesis, and what functional distinctions exist among EV subtypes. Future single-vesicle analyses that combine nanoscale ‘-omics’ with advanced statistical and AI methods will be crucial for resolving the full spectrum of EV diversity.
While single-vesicle resolution would offer even deeper insight, bulk EV analysis can still uncover biologically meaningful and source-representative features which contribute to systemic intercellular communication. Regarding source attribution, protein signatures associated with diverse cell types were represented in plasma EVs including endothelial cells, fibroblasts, hepatocytes, cardiomyocytes, kidney cells and haematopoietic cells (such as platelets).
As a resource for the field, we generated a freely-accessible molecular reference map – termed EVMap. This searchable atlas provides a platform to resolve and understand what is an EV from circulation, including the network of proteins displayed on their membrane surface, and key information on their likely cell, tissue, and organ origins.
How do you see this changing the way we study extracellular vesicles and their roles in human health and disease?
We anticipate that this knowledge could completely revolutionize biomarker discoveries in human health and disease, by resolving true EV proteins and lipids and disease specific features. Our core set of proteins and lipids gives us a way to standardize population-based studies against which disease specific changes can be monitored, potentially even temporally.
By offering biologically grounded criteria for defining EV identity, this study addresses the longstanding issue of the purity of EVs isolated from plasma – more than lists and universal EV subset – but a benchmark molecular criterion in their definition. This high-resolution blueprint of proteins and lipids in human EVs – creating a framework for practical foundation across basic, translational and clinical applications. As the field advances, these hallmark features will enhance the specificity and reproducibility of EV-related studies, enabling the development of scalable diagnostic platforms and targeted therapeutic strategies.
This refined, highly selective multi-molecular marker set is compatible with targeted, scalable assays such as enzyme-linked immunosorbent assay or targeted mass spectrometry, making it highly practical for clinical translation. Furthermore, the ranked feature list generated by our machine learning framework provides a valuable resource for the research community, enabling prioritization of alternative markers based on available reagents or disease-specific applications.
Looking ahead, what are the next steps for this work?
The EVMap opens new frontiers of exploration across a range of fields; from bioengineering technology to biomarker discovery and early disease detection, clinical management, and population health. By decoding these EV messages, we’re beginning to truly read the body’s own health reports. Through diverse collaborations, our team is now expanding this work to integrate EV multi-omics data from larger cohorts, to evolve into liquid biopsy tools with real clinical impact.
Challenges and active areas of research remain, including advances in capture and detection technologies, improving reproducibility, and the continued development of standardization and robust assessment workflows. Another major focus is deciphering EV heterogeneity – how diverse EV subtypes relate to disease drivers, and how these insights can be translated clinically.
The immediate next steps include focused investigations on molecular stability across diverse disease states, to establish the use of these core EV features as a reliable foundation for human EV studies in large population-based cohorts. This will enable consistent characterization and cross-study comparability moving forward.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
