Protein biomarker
A tumor sheds its identity into blood years before a scan can see it. So does a dying neuron. The question researchers keep answering the hard way — through invasive biopsy, through imaging that only resolves disease after real damage — is whether that identity can be read non-invasively, early, and repeatedly. Extracellular vesicles (EVs) carry exactly this kind of information on their surface, not just inside them. Most biomarker programs still look in the wrong compartment.
Key Takeaways
- EV surface proteins reflect the identity and pathological state of the parent cell, and they’re accessible without lysing the vesicle — unlike luminal cargo (RNA, soluble proteins).
- Surface protein signatures have identified early-stage disease across multiple tissue types, including cancers and neurodegenerative conditions, often before symptomatic or bulk-fluid changes appear.
- MISEV2023 guidelines increasingly treat surface protein characterization as central to EV biomarker validity, not an optional add-on to cargo analysis.
- Multiplexed, no-lysis surface profiling preserves vesicle integrity for downstream use and cuts the hands-on time that lysis-dependent workflows require.
Why the EV Surface Matters as Much as What’s Inside
Every cell in the body sheds EVs continuously, and those vesicles carry a molecular fingerprint of the cell that made them [1]. Most EV biomarker work has historically focused on cargo — microRNAs, soluble proteins, DNA fragments packaged inside the vesicle. That focus made sense early in the field, when cargo analysis borrowed directly from established proteomic and genomic pipelines. But cargo comes at a cost: reading it requires lysing the vesicle, which destroys the particle and adds a processing step that’s hard to standardize across labs [2].
Surface proteins don’t have that problem. They sit on the outer face of the lipid bilayer, exposed and available for antibody-based detection without breaking anything open [2]. That single structural fact changes what’s practical. A surface marker panel can be run on intact vesicles, preserved for repeat testing, and multiplexed across dozens of targets in a single well. A cargo panel generally can’t.
The biological argument for surface proteins is just as strong as the practical one. Surface protein composition isn’t uniform across EVs — it varies by the cell type that produced the vesicle and by that cell’s pathological state at the moment of shedding [3]. A landmark proteomic survey of EVs and particles across more than 400 human tissue and plasma samples identified surface and surface-associated proteins that distinguished tumor tissue from normal tissue with over 90% sensitivity and specificity, and separated cancer types from one another using plasma alone [3]. That’s a strong existence proof: the surface of a circulating vesicle can carry tissue-of-origin and disease-state information specific enough to support a diagnostic classifier.
What “Early Indicator” Actually Requires
Calling something an early indicator is a specific claim, not a general compliment. It means the signal has to be detectable before the disease process is far enough along that intervention options have narrowed. That’s a high bar, and it’s where a lot of promising EV markers quietly fail — not because the biology is wrong, but because the marker only separates cases from controls once disease is already advanced.
Surface protein panels have cleared that bar in several disease areas where early detection is the entire clinical problem. In hepatocellular carcinoma, an EV-based surface protein assay was developed specifically to catch early-stage disease, when curative treatment is still possible and when conventional serum markers are least reliable [4]. In colorectal cancer, a systematic screen of EV surface proteins identified three markers — validated across a multicenter cohort of over 400 individuals — that distinguished cancer patients from healthy controls with area-under-the-curve performance above 0.88 in an independent test set [5]. Neither result required imaging. Neither required tissue.
Neurodegeneration poses the same problem in a different shape. By the time a clinical diagnosis is confirmed, substantial neuronal loss has typically already occurred, and the biological processes researchers most want to catch — early misfolding, early synaptic dysfunction, early inflammatory response — are happening in a compartment blood tests have historically struggled to reach cleanly. Extracellular vesicles that originate in the central nervous system and cross into peripheral blood carry a proteomic signature that reflects those brain-side processes, and platforms built to characterize CNS-derived EV surface markers in blood and other body fluids have demonstrated the technical feasibility of separating disease-relevant subpopulations from the broader circulating EV pool [6]. The clinical translation work is still active, but the core premise — that a peripheral, repeatable blood draw can carry CNS-specific surface signal — is no longer speculative [7].
A Closer Look at What “Cleared the Bar” Means in Practice
The colorectal cancer example is worth unpacking, because the process behind it is a useful template for how surface protein discovery is supposed to work. Researchers started with an unbiased proteomic screen of EVs isolated from patient serum, comparing surface protein expression between confirmed colorectal cancer cases and healthy controls using four-dimensional data-independent acquisition mass spectrometry — a method capable of resolving thousands of proteins per sample [5]. That screen surfaced seven candidate surface proteins showing significant differential expression. Only three survived the next step: validation on a purpose-built, extraction-free microarray run across a multicenter cohort exceeding 400 individuals, split into training and independent test sets [5].
That two-stage structure — broad discovery screen, then narrowed validation on an independent cohort — is what separates a real early-indicator candidate from a marker that merely looks good in the discovery data. The final three-protein panel classified disease status with an area-under-the-curve of 0.882 in training and 0.937 in the independent test set, using a machine-learning model built on surface protein levels alone, with no imaging or tissue input [5]. The hepatocellular carcinoma example followed a similar logic: the assay was purpose-built around early-stage detection, the point in disease progression where conventional serum tumor markers are least reliable and where curative treatment options are still on the table [4].
Neither program treated a single marker as sufficient. Both treated the surface protein signature as a panel-level property, distributed across more than one target, validated against an independent population before anyone called it a biomarker rather than a candidate. That distinction — candidate versus validated biomarker — is where a lot of the field’s disappointment lives, and it’s a distinction surface protein programs have started taking as seriously as the cargo-analysis side of the field has for years [1].
Why Surface Signal Survives Where Bulk-Fluid Markers Struggle
A biomarker that circulates in bulk plasma, unattached to any particular structure, is diluted by every cell type that also sheds it or a version of it. That’s the recurring problem with several first-generation blood-based neurodegeneration markers: real signal, real biological relevance, but swamped by peripheral contributions that have nothing to do with the disease process under study.
Anchoring a marker to a specific vesicle population changes that math. If a surface protein is enriched on EVs from a particular tissue or cell type, measuring that protein in the context of vesicle-level detection — rather than as a free-floating plasma analyte — narrows the denominator. The signal is still diluted by the sheer volume of circulating EVs from every other tissue, but it’s no longer diluted by every free protein of the same identity circulating outside a vesicle at all. This is a large part of why EV-anchored surface protein panels have outperformed some bulk-fluid approaches for tissue-specific detection tasks in cancer, and why the same logic is actively being tested for CNS-derived signal [6].
| Detection approach | Requires lysis | Tissue-of-origin resolution | Multiplexing capacity |
|---|---|---|---|
| Bulk plasma protein (e.g., ELISA on whole plasma) | No | Low — reflects total circulating pool | Limited without separate assays |
| EV luminal cargo (RNA, soluble protein) | Yes | Moderate — cargo can be cell-type-specific | Constrained by lysis/extraction volume |
| EV surface protein (intact vesicle) | No | High — surface markers report cell of origin | High — antibody panels scale well |
Table 1. Conceptual comparison of detection approaches by tissue-of-origin resolution and workflow constraints. Illustrative, synthesized from characterization literature [2,3].
The Standardization Problem Nobody Gets to Skip
None of this works without rigor, and the EV field has been explicit about where it’s still falling short. The most recent Minimal Information for Studies of Extracellular Vesicles guidelines (MISEV2023) devote substantial attention to surface protein characterization specifically, reflecting a field-wide recognition that surface marker claims need the same standardization discipline that cargo analysis has been working toward for a decade [1]. A 2025 review of EV surface protein biosensing methods makes the underlying tension explicit: surface markers are attractive precisely because they avoid lysis, but the heterogeneity of surface protein expression from vesicle to vesicle — and the lack of a single consensus isolation method — means results can vary meaningfully between labs studying the same disease [2].
That heterogeneity isn’t a reason to abandon surface protein work. It’s a reason to standardize how it’s measured. A single-marker readout, run once, on an ultracentrifugation-isolated sample, is exactly the kind of result that MISEV2023 authors flag as hard to reproduce [1]. A multiplexed panel, run consistently across sample batches without an isolation step that itself introduces variability, is a meaningfully different proposition — not because multiplexing is a magic fix, but because it removes one of the largest sources of batch-to-batch inconsistency before the antibody panel even runs.
[FIGURE 2 — Illustrative: Surface marker heterogeneity across isolation methods] Figure 2. Conceptual bar chart showing relative surface marker signal variability across common isolation approaches (ultracentrifugation, size-exclusion chromatography, precipitation kits, no-isolation multiplex profiling), based on qualitative patterns described in EV characterization literature [1,2]. Labeled illustrative — no proprietary data.
Why Multiplexing Is a Standardization Tool, Not Just a Throughput Feature
It’s worth being precise about why multiplexing helps here, because the benefit isn’t simply “more markers per run.” A single-marker antibody assay run on separately isolated, separately processed samples accumulates variability at every step — isolation yield, isolation-method-dependent co-purification of non-vesicular protein, storage and freeze-thaw handling, and the antibody assay itself [2]. Each of those steps is a place where two labs, or even two technicians in the same lab, can diverge without either one making an error.
A multiplexed panel run on the same intact-vesicle sample, in the same well, collapses several of those variability sources into a single controlled step. Every marker in the panel is exposed to the same isolation history (or lack of one), the same storage conditions, and the same assay chemistry at the same time. That doesn’t eliminate biological heterogeneity between vesicles — surface marker expression genuinely does vary from vesicle to vesicle within the same sample, and no assay design changes that [2]. But it does mean that when a panel shows a difference between two markers in the same patient, researchers can be more confident the difference reflects biology rather than differential handling of two separate single-marker assays run days apart. This is a distinct claim from “multiplexing is more efficient,” and it’s the reason MISEV2023 treats measurement consistency, not just marker choice, as central to whether a surface protein finding is likely to replicate [1].
What This Means for a Translational Biomarker Program
None of the disease-area examples above were CNS-exclusive, and that’s the point. Surface protein biology doesn’t respect a single therapeutic area — a marker panel logic validated in colorectal or liver cancer surface protein work applies structurally to a neurodegeneration program looking for its own early, tissue-anchored signal [4,5]. The specifics of which markers matter will differ by disease. The argument for why surface, not cargo, is where researchers should be looking first does not.
For a biopharma team building a translational biomarker strategy — whether the target indication is oncology, a CNS disorder, or something else entirely — the practical takeaway is narrower than “EVs are promising.” It’s: know what compartment of the EV you’re actually measuring, and know whether your current assay requires lysis it doesn’t need to require. Surface protein profiling that skips isolation and lysis steps entirely — assessing intact EVs directly from unprocessed plasma — is one way research teams are addressing this at the workflow level, and it’s the design principle behind LuminEV’s multiplex EV panel, which characterizes surface protein signatures across sample cohorts without an upstream isolation step.
What Regulators Will Want Before Any of This Reaches a Trial Endpoint
Discovery-stage performance numbers, however strong, are not the same thing as a biomarker a regulator will accept as a trial endpoint or a companion diagnostic input. Analytical validation — establishing that an assay measures the same analyte the same way across runs, operators, and sites — sits between a promising discovery cohort result and any qualification submission, and it applies with equal force to surface protein panels as it does to any other blood-based biomarker modality [1]. The MISEV2023 update is explicit that reproducibility documentation, not just performance metrics from a single study, is what separates a research-grade finding from something that can support a regulatory conversation [1].
That has direct implications for how a surface protein program should be sequenced. A discovery cohort — even one as large as the 400-plus-patient colorectal cancer study — establishes that a signal exists and roughly how strong it is [5]. It does not establish that the assay measuring that signal will produce the same result when run six months later, in a different lab, on a different instrument. Bridging that gap requires the kind of analytical validation work — precision, reproducibility, stability, and cutoff-setting studies — that biopharma teams already run for conventional protein biomarkers, applied to a surface protein readout instead of a bulk plasma one. Programs that treat this validation step as a late-stage formality rather than a parallel workstream tend to be the ones that stall between an encouraging publication and an actual qualified endpoint.
Where the Field Goes From Here
The next phase of this work isn’t proving that EV surface proteins carry disease signal — that case is increasingly well established across multiple disease areas. It’s building the standardized, multiplexed measurement infrastructure that lets surface protein findings replicate across labs and translate into assays sponsors can actually run at trial scale. Expect continued convergence between EV characterization guidelines and clinical assay validation standards over the next several years, as more surface protein candidates move from discovery cohorts into companion diagnostic and pharmacodynamic contexts.
References
[1] Welsh JA, Arkesteijn GJA, Bremer M, et al. Minimal information for studies of extracellular vesicles (MISEV2023). J Extracell Vesicles. 2024;13(5):e12451.
[2] Pham TTH, Sakamoto H, Hasegawa T, Sakamoto C, Suye SI, Chuang HS. Small extracellular vesicle-associated surface protein biomarkers: emerging roles, opportunities, and challenges in diagnostics. Front Bioeng Biotechnol. 2025;13:1714972.
[3] Hoshino A, Kim HS, Bojmar L, et al. Extracellular Vesicle and Particle Biomarkers Define Multiple Human Cancers. Cell. 2020;182(4):1044-1061.e18.
[4] Sun N, Zhang C, Lee YT, et al. HCC EV ECG score: An extracellular vesicle-based protein assay for detection of early-stage hepatocellular carcinoma. Hepatology. 2023;77(3):774-788.
[5] Huang Z, Deng C, Ma C, He G, Tao J, Zhang L, et al. Identification and validation of the surface proteins FIBG, PDGF-β, and TGF-β on serum extracellular vesicles for non-invasive detection of colorectal cancer: experimental study. Int J Surg. 2024;110(8):4672-4687.
[6] Brahmer A, Geiß C, Lygeraki A, et al. Assessment of technical and clinical utility of a bead-based flow cytometry platform for multiparametric phenotyping of CNS-derived extracellular vesicles. Cell Commun Signal. 2023;21(1):276.
[7] Croese T, Furlan R. Extracellular vesicles in neurodegenerative diseases. Mol Aspects Med. 2018;60:52-61.
[8] Suthar J, Taub M, Carney RP, Williams GR, Guldin S. Recent developments in biosensing methods for extracellular vesicle protein characterization. WIREs Nanomed Nanobiotechnol. 2023;15(1):e1839.
