Where are EV-based liquid biopsies actually headed?
Toward provenance rather than sensitivity. Detection limits are no longer the binding constraint on EV diagnostics. The next decade will be decided by three separate problems: resolving which tissue a vesicle came from, standardizing pre-analytical handling well enough to survive multi-site studies, and completing a regulatory path that no EV-derived measurement has yet walked end to end.
Key Takeaways
- Analytical sensitivity has improved faster than analytical specificity of origin. Bulk plasma EV measurements now detect signal reliably while still averaging across every secreting tissue in the body.
- In healthy donors, vesicles from haematopoietic and endothelial cells dominate the circulating pool; deconvolution estimates put the combined solid-tissue contribution as low as 0.2% [1].
- Pre-analytical handling remains the largest single source of irreproducibility, with more than forty documented variables across hundreds of protocols in active use [2].
- Regulatory throughput, not assay performance, is the rate-limiting step for biomarkers intended to support drug development decisions [3].
- The field is consolidating toward indication-specific, provenance-defined panels — not a single universal EV test.
Sensitivity stopped being the bottleneck
For most of the last decade, the central technical question in EV diagnostics was whether a signal could be detected at all. Plasma contains vesicles at concentrations that defeated conventional immunoassays, and the standard workaround — ultracentrifugation or polymer precipitation followed by a single-analyte readout — consumed sample volume and introduced its own variance.
That question has largely been answered. Digital and single-vesicle platforms now resolve individual molecules on individual vesicles, including protein, mRNA, and both double- and single-stranded DNA [4]. Bead-based multiplex immunoassays measure tetraspanin surface markers directly in neat plasma and in conditioned media without a prior isolation step; validation across fifteen cell lines showed sensitivity sufficient to quantify EV markers with no pre-concentration at all [5].
The consequence is easy to miss. When the analytical floor drops below the level at which biology varies, further gains in sensitivity buy nothing. What they do instead is make a different problem visible: the measurement is now precise enough to demonstrate, unambiguously, that the wrong quantity is being measured.
MISEV2023 anticipated this. The guidelines explicitly caution against assuming the origin of EVs — whether by cell type or by biogenesis pathway — and recommend characterizing heterogeneity using at least two complementary technologies [6]. That recommendation is not a reporting formality. It is a statement that the field’s central unknown has shifted from how much to from where.
Tissue of origin is the frontier that matters
Circulating EVs are not a sample of one tissue. They are a sample of every tissue that secretes into blood, weighted by secretion rate, clearance rate, and the accident of which cells sit closest to the vasculature. In healthy individuals, the majority of circulating vesicles are believed to arise from platelets, erythrocytes, leukocytes, and the endothelium contacting the blood [1].
The proportions are less favourable than most study designs assume. Secretion rates do not track parental cell abundance: reported EV-to-cell ratios span roughly 0.13 erythrocyte-derived vesicles per erythrocyte to about 1.9 × 10³ monocyte-derived vesicles per monocyte [1]. Computational deconvolution of tissue-specific transcriptomic signatures has placed the total solid-tissue contribution to the circulating pool as low as 0.2% [1]. And lipoproteins — biophysically similar in size and buoyant density — outnumber vesicles in plasma by several orders of magnitude [1].
Figure 1. Reported estimates of tissue contribution to the circulating EV pool
| Source | Reported contribution | Method basis |
|---|---|---|
| Haematopoietic and endothelial cells | Majority of circulating vesicles | Surface marker phenotyping; flow cytometry |
| All solid tissues combined | As low as 0.2% | Transcriptomic deconvolution |
| Hepatocytes | Up to 20% | Protein marker estimates |
| Placenta (pregnancy) | Up to 15% | Protein marker estimates |
| Skeletal muscle | 1–5% | Protein marker estimates |
| Colon | 1–4% | Protein marker estimates |
| Cardiomyocytes | ~3% (transgenic mouse model) | Fluorescent reporter |
Estimates collated in a 2025 review of tissue-specific EV detection and isolation [1]. Protein-marker and transcriptomic estimates conflict, and both depend on antibody cross-reactivity and on the method used to enumerate total vesicles in the denominator. Read as ranges, not values.
Note that the protein-based and transcriptome-based estimates do not agree with each other. That disagreement is itself the finding: the field does not yet have a reference standard for what circulating EV composition actually is under native, dynamic conditions.
Immunoaffinity capture on origin-restricted surface antigens is the mechanism that addresses this, and it confers two distinct benefits rather than one. It removes signal for an analyte of interest contributed by non-target tissues, which improves specificity. It also depletes the abundant contaminants that suppress signal for that analyte, which improves sensitivity [1]. The second benefit is routinely overlooked. Origin enrichment is not a purity exercise performed at the cost of yield; done well, it raises the detectable signal from the population that matters.
The CNS literature offers the clearest demonstration, because brain is the hardest tissue to sample and therefore the one where provenance is least substitutable. Enrichment methods that isolate cell-specific vesicles from biofluids have made it possible to monitor difficult-to-access organs without disturbing them, and the contents of neuron-derived vesicles in blood track alterations that occur during neurodegenerative pathogenesis [7]. In a cohort of early Alzheimer’s and control donors, proBDNF measured in origin-enriched neuron-derived vesicles was significantly lower in the disease group, while the same analyte measured in unprocessed plasma from the same donors showed no difference at all [8]. The signal was present in both samples. Only one of them could see it.
The same principle holds for intervention response. In a randomized controlled trial of exercise in mild-to-moderate Alzheimer’s disease, neuroprotective protein levels in plasma neuron-derived vesicles rose in the exercise arm across 16 weeks and remained unchanged in controls, with the effect most pronounced in APOE ε4 carriers [9]. That is a pharmacodynamic readout from a blood draw, in a population where the alternative was serial lumbar puncture.
Standardization will decide which assays survive multi-site studies
An assay that performs well in one lab, on one operator’s samples, collected under one protocol, tells you very little about what happens when the same assay is deployed across eight clinical sites. This is where most EV biomarker programs quietly fail, and it is not primarily an assay problem.
The ISEV Blood EV Task Force has documented hundreds of pre-analytical protocols in active use across more than forty distinct variables — collection tube and anticoagulant, needle gauge, handling time, centrifugation speed and duration, temperature, freeze-thaw history, residual platelet content [2]. Several studies have reported significant differences in molecular composition between plasma- and serum-derived EV preparations drawn from the same donors [2]. Practical guidance on blood collection and plasma preparation now exists in accessible form [10], and MISEV2023 devotes specific attention to the complexity of blood as a matrix, including the cells, lipoproteins and soluble proteins retained in most EV preparations [6].
What matters for study design is the ordering. When pre-analytical variance exceeds the biological effect size, no amount of downstream analytical precision recovers the signal. The variance was introduced before the sample reached the assay.
Figure 2. Where variance enters an EV biomarker measurement
COLLECTION PREPARATION ENRICHMENT DETECTION ───────────────────── ──────────────────── ───────────────────── ───────────────────── Tube / anticoagulant Centrifugation Capture antigen Assay format Needle gauge Residual platelets Antibody specificity Multiplex vs. simplex Handling time Plasma vs. serum Recovery efficiency Calibration Storage temperature Freeze-thaw cycles Cross-reactivity Operator Donor state (diet, Lipoprotein carry- Depletion of non- Plate and run effects exercise, circadian, over target vesicles medication) ▲ Largest documented source ▲ Most heavily optimized of irreproducibility over the last decade (>40 variables, [2])
Conceptual and illustrative. Variable inventory drawn from the ISEV Blood EV Task Force reporting framework [2] and MISEV2023 [6]. The asymmetry is the point: field investment has concentrated at the right-hand end of the workflow while the dominant variance sits at the left.
The corollary is that reported precision figures should be read as compound quantities. When an enrichment-plus-detection workflow reports coefficients of variation between 8.0% and 22.7% across five independent experiments on repeatedly processed plasma samples, that range combines variance from both the isolation step and the measurement step [8]. In the same evaluation, variability between two proficient operators was lower than variability between donors, and the workflow transferred successfully to an independent laboratory [8]. Those are the specific claims worth demanding from a platform, and they are distinct from a detection limit.
Multiplex detection contributes to this differently than it is usually marketed. Measuring several analytes from one capture event on one aliquot removes an entire class of between-run variance, because those analytes share a sample, a plate, and a set of handling conditions. The LuminEV Research Kit was validated against exactly this specification — simplex-versus-multiplex agreement, linearity, spike-in recovery, inter- and intra-assay precision, and reproducibility between operators [5]. Whichever platform a program selects, that is the shape of the validation package that survives a multi-site deployment.
The regulatory path is the gating item few have walked
Assume a program solves provenance and standardization. It then meets a bottleneck that has nothing to do with vesicle biology.
A 2025 systematic review of EV biomarker translation catalogued the regulatory layer explicitly, spanning laboratory-developed tests, in-house in vitro diagnostics, CLIA requirements, ISO standards and full IVD approval. It identified the discrepancy between established clinical laboratory standards and experimental EV technology as a distinct roadblock, separate from assay development [11]. An earlier assessment written from clinical laboratory practice put it more bluntly: despite viable isolation methods and a wide analyte space accessible by standard detection technologies, the attrition rate from published academic biomarker reports to deployable clinical assays is very high [12].
For biomarkers intended to support drug development decisions rather than clinical diagnosis, the relevant pathway is formal qualification, and its throughput is sobering. Across eight years of the FDA Biomarker Qualification Program, 61 projects had been accepted as of July 2025 and eight biomarkers had been qualified. Half of accepted projects remained at the initial Letter of Intent stage. For projects that reached the Qualification Plan stage, plan development took a median of 32 months — 47 months for those pursuing surrogate endpoints [3].
Figure 3. Validation stages and current field position for EV-based measurements
| Stage | What it establishes | Status for EV-derived measurements |
|---|---|---|
| Analytical validation | Linearity, precision, recovery, operator and site transfer | Demonstrated for multiple platforms and analytes [5][8] |
| Pre-analytical control | Protocol-level reproducibility across collection sites | Reporting frameworks published; adoption uneven [2][6] |
| Clinical validation | Association with a defined clinical or pathological state | Extensive in cohorts; sparse in prospective multi-site designs [11] |
| Clinical utility | Measurement changes a decision and improves an outcome | Largely unaddressed |
| Regulatory qualification | Accepted context of use across drug development programs | No EV-derived measurement has completed the pathway [3][11] |
Stage definitions follow standard biomarker development practice; the status column is synthesized from the cited sources. Illustrative framing rather than a formal regulatory schema.
The strategic implication is that the first EV-derived measurement to complete qualification will define the evidentiary bar for everything behind it. Programs generating data now are, whether or not they intend to be, generating that precedent. This argues for building the pre-analytical and multi-site reproducibility package earlier than a discovery-stage program would naturally choose to.
What comes next, and what probably does not
The universal EV test is not coming
Forecasts of this field have a recurring failure mode: they predict a universal EV test. A single panel, on a single platform, reporting across oncology, neurodegeneration, cardiology and transplant rejection.
That is not where the biology points. The value of an EV measurement comes from restricting it to a defined cellular origin, and origin restriction is inherently indication-specific. The antigen that isolates hepatocyte-derived vesicles is not the antigen that isolates neuron-derived vesicles, the validated marker inventory differs enormously in maturity across tissues [1], and cross-reactivity behaves differently in each case. A panel general enough to serve every indication would, by construction, be measuring the bulk plasma pool again.
Two other expectations deserve tempering. Single-analyte EV tests are unlikely to carry high-stakes decisions, because the interpretive weight sits in the pattern across analytes from a common origin rather than in any one concentration. And retrospective biobank material will not substitute for prospectively collected samples in the validation studies that matter, because the pre-analytical variables that dominate EV variance were mostly not recorded when those biobanks were assembled [2].
What the next five years probably do look like
The plausible trajectory is unglamorous and fairly specific. Marker inventories for tissue-restricted capture will be systematically validated rather than opportunistically adopted, which means published characterization of cross-reactivity and tissue expression for every candidate antigen. Pre-analytical reporting will move from recommended to expected, and studies that cannot document their collection conditions will stop being publishable in the journals that matter. Multi-site precision data will become a standard element of platform documentation rather than a competitive differentiator. And one program, somewhere, will carry an origin-defined EV measurement through a formal qualification pathway and publish what the agency asked for.
None of that is a breakthrough. It is the work that converts a measurement into evidence. For teams designing biomarker strategy now, the useful question is not whether EV liquid biopsy reaches the clinic, but which stage of that path their own program is actually resourced to complete. Selecting a platform that has published its precision, its recovery and its operator-transfer data is a reasonable place to start; for surface-protein profiling in plasma, the LuminEV Research Kit documents that package in the peer-reviewed literature [5].
References
- Newman L, Rowland A. Detection and isolation of tissue-specific extracellular vesicles from the blood. J Extracell Biol. 2025;4(6):e70059. https://doi.org/10.1002/jex2.70059
- Lucien F, Gustafson D, Lenassi M, et al. MIBlood-EV: minimal information to enhance the quality and reproducibility of blood extracellular vesicle research. J Extracell Vesicles. 2023;12(12):e12385. https://doi.org/10.1002/jev2.12385
- Collins G, Allen JD, Andrews HS, Navarro-Serer B, Stewart MD. Hurry up and wait: timelines and takeaways from the Biomarker Qualification Program. Ther Innov Regul Sci. Published online October 26, 2025. https://doi.org/10.1007/s43441-025-00889-6
- Zhang Y, Meng X, Greening DW, et al. Unveiling heterogeneity: innovations and challenges in single-vesicle analysis for clinical translation. J Extracell Vesicles. 2025;14(12):e70209. https://doi.org/10.1002/jev2.70209
- Tordoff E, Allen J, Elgart K, et al. A novel multiplexed immunoassay for surface-exposed proteins in plasma extracellular vesicles. J Extracell Vesicles. 2024;13(11):e70007. https://doi.org/10.1002/jev2.70007
- Welsh JA, Goberdhan DCI, O’Driscoll L, et al. Minimal information for studies of extracellular vesicles (MISEV2023): from basic to advanced approaches. J Extracell Vesicles. 2024;13(2):e12404. https://doi.org/10.1002/jev2.12404
- Cleary JA, Kumar A, Craft S, Deep G. Neuron-derived extracellular vesicles as a liquid biopsy for brain insulin dysregulation in Alzheimer’s disease and related disorders. Alzheimers Dement. 2025;21(2):e14497. https://doi.org/10.1002/alz.14497
- Eitan E, Thornton-Wells T, Elgart K, et al. Synaptic proteins in neuron-derived extracellular vesicles as biomarkers for Alzheimer’s disease: novel methodology and clinical proof of concept. Extracell Vesicles Circ Nucl Acids. 2023;4(1):133–150. https://doi.org/10.20517/evcna.2023.13
- Delgado-Peraza F, Nogueras-Ortiz C, Simonsen AH, et al. Neuron-derived extracellular vesicles in blood reveal effects of exercise in Alzheimer’s disease. Alzheimers Res Ther. 2023;15(1):156. https://doi.org/10.1186/s13195-023-01303-9
- Nieuwland R, Siljander PR-M. A beginner’s guide to study extracellular vesicles in human blood plasma and serum. J Extracell Vesicles. 2024;13(1):e12400. https://doi.org/10.1002/jev2.12400
- Droste M, Puhka M, van Royen ME, et al. Roadblocks of urinary EV biomarkers: moving toward the clinic. J Extracell Vesicles. 2025;14(7):e70120. https://doi.org/10.1002/jev2.70120
- Enderle D, Noerholm M. Are extracellular vesicles ready for the clinical laboratory? J Lab Med. 2022;46(4):273–282. https://doi.org/10.1515/labmed-2022-0064
