The search for reliable biomarkers is becoming increasingly central to biomedical research. In oncology and precision medicine, biomarkers can support earlier disease detection, patient stratification, prognosis, treatment selection, and longitudinal monitoring. Yet conventional approaches often depend on tissue biopsies, which provide valuable biological information but remain invasive and are not always suitable for repeated sampling.
This has accelerated interest in liquid biopsy strategies capable of detecting disease-associated molecular signals in accessible biological fluids. Among the most promising biological materials in this field are extracellular vesicles (EVs).
Released by virtually all cell types, EVs are membrane-enclosed particles capable of transporting proteins, lipids, metabolites, DNA, and multiple classes of RNA. Their molecular composition can reflect the physiological or pathological state of the cells from which they originate, while their lipid membrane protects part of this molecular cargo from extracellular degradation.
These characteristics make EVs particularly attractive as minimally invasive sources of biomarkers. At the same time, the rapid development of transcriptomics, proteomics, metabolomics, lipidomics, multi-omics integration, and artificial intelligence is providing increasingly sophisticated tools for decoding the molecular information carried by EVs.
For cell-derived EV research, however, this analytical sophistication also raises an important upstream question: how controlled and reproducible are the cellular conditions under which those vesicles are produced?
Extracellular Vesicles: More Than Cellular Waste
EVs were historically regarded largely as mechanisms through which cells eliminated unwanted material. This interpretation has changed substantially.
They are now recognized as active participants in intercellular communication, capable of transferring biologically active molecules between cells over both local and distant environments.
EVs comprise a heterogeneous family of particles. According to the biogenesis-oriented framework described in the review and aligned with MISEV2023, major categories include exosomes, microvesicles, and apoptotic vesicles. Because EV populations overlap substantially in size and definitive biogenesis can be difficult to establish experimentally, the umbrella term EVs remains preferable when vesicle origin has not been demonstrated.
Exosomes, generally associated with the endosomal pathway, form as intraluminal vesicles within multivesicular bodies before release following fusion with the plasma membrane. Their biogenesis involves a sophisticated molecular machinery including the ESCRT pathway, lipids such as ceramides, tetraspanins including CD9 and CD63, small GTPases such as Rab27a/b and Rab11, and accessory proteins involved in trafficking and cargo selection.
Microvesicles, in contrast, arise through direct outward budding of the plasma membrane, involving changes in phosphatidylserine distribution and cytoskeletal remodeling.
This complexity is important because EV production and composition are closely connected to cellular biology.
A Protected Molecular Fingerprint of the Producing Cell
The biological value of EVs lies largely in the diversity of the information they carry.
Their lipid bilayers contain characteristic combinations of cholesterol, sphingomyelin, phosphatidylserine, ceramide, and other lipid species. Their surfaces can display tetraspanins, integrins, MHC molecules, transporters, receptors, and proteins involved in EV biogenesis.
Their cargo extends far beyond membrane components.
EVs can contain signaling proteins, metabolic enzymes, transcription-related factors, microRNAs, messenger RNAs, long non-coding RNAs, tRNAs, DNA fragments, metabolites, and bioactive lipids.
This molecular diversity creates what can be considered an EV molecular fingerprint.
Importantly, the vesicular membrane can protect encapsulated molecules from extracellular proteases and nucleases. RNA is particularly relevant in this context because extracellular RNA is otherwise vulnerable to rapid enzymatic degradation.
Once EVs enter biological fluids, their molecular identity may become even more complex through formation of a biomolecular corona, with additional biomolecules adsorbing onto the vesicle surface and potentially influencing stability, biodistribution, and cellular interactions.
EV composition is therefore not static. It is influenced by the biological state of the producing cell and by environmental and physiological conditions. This characteristic is fundamental to their biomarker potential, but it is also one of the reasons why experimental standardization is critical.
Why EVs Are Attractive Biomarker Candidates
EVs have been identified in numerous biological fluids, including blood, urine, saliva, cerebrospinal fluid, and breast milk.
This accessibility creates an opportunity to obtain disease-associated molecular information through minimally invasive sampling.
In principle, EV analysis could therefore complement conventional tissue-based diagnostics while enabling repeated longitudinal measurements during disease progression or treatment.
Their potential extends beyond simple disease detection. Because EV cargo can reflect the molecular state of the cell of origin, EV profiling may contribute to patient stratification, prognosis, treatment monitoring, and precision medicine.
However, the same biological diversity that makes EVs informative also makes them challenging to study.
EV populations are heterogeneous, and their measured molecular composition can be influenced by donor characteristics, physiological state, sample collection, processing, storage, isolation methodology, and downstream analytical procedures.
EVs as Active Participants in Cancer Biology
The relationship between EVs and cancer is particularly important because tumor-derived EVs are not merely passive reflections of malignant cells.
Tumor-derived extracellular vesicles (TDEVs) can actively participate in remodeling the tumor microenvironment.
Their molecular cargo has been associated with angiogenesis, stromal activation, immune evasion, metastatic progression, pre-metastatic niche formation, and therapy resistance.
Tumor-derived EVs can, for example, transport immune-regulatory molecules such as PD-L1, immunosuppressive cytokines, and microRNAs capable of influencing antitumor immune responses.
Surface molecules can also influence EV targeting. Integrins and other membrane proteins may contribute to preferential interactions with particular recipient cells and tissues, potentially participating in the organ-specific dissemination of tumor-associated signals.
EVs can additionally transfer molecules associated with therapeutic resistance between tumor cells, allowing adaptive characteristics to propagate within heterogeneous cancer cell populations.
This creates an intriguing duality.
The same vesicles that participate in cancer progression can potentially become molecular reporters of that progression.
From Liquid Biopsy to Molecular Profiling
Tumor-associated EVs may contain tumor DNA fragments, oncogenic receptors such as EGFR and HER2, disease-associated RNAs, proteins, metabolites, and lipids.
Analyzing this cargo therefore provides an opportunity to investigate cancer through several complementary molecular dimensions.
The development of high-throughput technologies has accelerated this field dramatically. Modern sequencing, mass spectrometry, and computational approaches can measure thousands of molecular features from biological samples.
This has given rise to four particularly relevant EV profiling strategies:
Transcriptomics investigates RNA cargo and disease-associated expression patterns.
Proteomics examines proteins, signaling components, and post-translational modifications.
Metabolomics captures metabolic changes associated with cellular state and disease.
Lipidomics characterizes membrane and signaling lipids that may reflect alterations in cellular metabolism and EV biogenesis.
Each provides valuable information. But each also represents only one layer of a much larger biological system.
EV Transcriptomics: Capturing Protected RNA Signatures
One of the major advantages of EV-based transcriptomics is the relative protection provided to RNA molecules by the vesicular membrane.
EVs contain diverse RNA classes, including mRNAs, miRNAs, lncRNAs, tRNAs, piRNAs, and other non-coding RNAs. Their distribution can vary between EV populations and according to cellular or pathological state.
The biomarker potential is already visible in exploratory oncology studies.
In breast cancer, for example, EV transcriptomic profiling identified dynamic RNA changes between diagnosis and seven days after surgery. A combined eight-RNA signature achieved an AUC of 0.90 under leave-one-out cross-validation, and several components were associated with estrogen receptor and HER2 status.
These observations demonstrate an important principle: multimolecular signatures may provide greater discriminatory power than isolated biomarkers.
However, promising predictive performance does not automatically imply clinical readiness. Independent cohort validation remains insufficient for many proposed EV-RNA signatures, leaving important risks of overfitting and limited generalizability.
Proteomics: Reading Functional Changes in EV Cargo
Proteins represent another major dimension of EV biology.
Using highly sensitive approaches such as liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS), researchers can identify and quantify large numbers of EV-associated proteins.
This is particularly valuable because proteomics can capture not only protein abundance but also post-translational modifications (PTMs) such as phosphorylation, glycosylation, ubiquitination, acetylation, and methylation.
In non-small-cell lung cancer, quantitative EV phosphoproteomics described in the source detected more than 2,000 phosphorylation sites across approximately 1,500 proteins, including numerous kinases.
Such molecular resolution can potentially reveal disease-associated signaling alterations that would remain invisible through measurements of protein abundance alone.
Metabolomics and Lipidomics: Capturing Disease-Associated Biochemistry
EVs also transport metabolites, including amino acids, carbohydrates, nucleotides, cofactors, and other small molecules.
This creates an opportunity to investigate disease-associated metabolism through a vesicle-enriched biological fraction.
In one lung cancer study described in the review, metabolomic profiling of urinary EVs generated a random forest-based diagnostic panel with an AUC of up to 0.96 in testing and 84% accuracy in a validation cohort.
EV lipidomics provides another complementary perspective.
Lipids influence EV membrane architecture, biogenesis, stability, signaling, and interactions with recipient cells. Cancer-derived EVs have demonstrated alterations in sphingolipids, ceramides, phospholipids, and glycerophospholipids.
Plasma EV lipid profiles have even been used experimentally to distinguish women with primary or metastatic breast cancer from healthy controls with approximately 90% classification accuracy under leave-one-out cross-validation.
These results remain exploratory, but they illustrate how different molecular layers can independently contain clinically relevant information.
The next logical step is to bring these layers together.
Multi-Omics: Reconstructing the Biological System
Cancer cannot be adequately represented as a single molecular abnormality.
Genomic alterations influence transcription. Transcriptional changes alter protein expression and signaling. Protein activity reshapes metabolism. Metabolic changes influence lipid composition and cellular behavior.
These processes form interconnected biological networks.
Multi-omics integration attempts to capture this complexity by combining information from several molecular layers rather than studying each independently.
For EV research, this approach is particularly compelling because EVs naturally contain several categories of biomolecules within the same biological material.
Different integration strategies can be applied depending on experimental design.
Early integration combines molecular features before modeling.
Late integration analyzes each omics layer separately and subsequently combines the results.
Intermediate integration uses statistical or machine-learning models to identify relationships across omics layers during analysis.
There is no universal optimal strategy. The appropriate approach depends on sample structure, missing data, feature distributions, batch effects, study design, and the biological or clinical objective.
Artificial Intelligence and the Next Generation of EV Biomarkers
As multi-omics datasets expand, their dimensionality creates a new analytical challenge.
This is where machine learning and artificial intelligence are becoming increasingly relevant.
AI-based approaches can analyze thousands of molecular variables simultaneously and identify multidimensional relationships that may remain difficult to capture using conventional statistical methods.
Eventually, EV-derived transcriptomic, proteomic, metabolomic, and lipidomic features could be integrated with clinical variables to support more sophisticated models for disease detection, patient stratification, progression prediction, or therapeutic response.
But AI cannot compensate for poor experimental design.
EV multi-omics datasets frequently combine high dimensionality with relatively small cohorts, creating substantial risks of overfitting. Pre-analytical heterogeneity, missing data, batch effects, and annotation biases can further influence model performance.
Consequently, AI-assisted EV biomarkers currently require rigorous external validation, careful regularization, reproducible computational workflows, and explainability analyses before meaningful clinical translation can occur.
Standardization Begins Before Omics Analysis
Much of the discussion surrounding EV reproducibility understandably focuses on isolation and downstream analysis.
However, the broader biological logic extends further upstream.
EV composition is influenced by cellular state and environmental conditions. The source specifically describes EV cargo as responsive to physiological and environmental cues including hypoxia, metabolic stress, exercise, temperature changes, and therapeutic interventions.
For EVs obtained directly from patients, controlling these variables is inherently difficult and requires careful metadata collection and sample standardization.
For EVs generated from in vitro cell cultures, however, an additional variable becomes experimentally controllable: the cell culture environment itself.
This distinction is important.
If cells are used as an experimental or production source of EVs, variation introduced during culture could potentially contribute to variation in the EV populations subsequently isolated and analyzed. Standardizing downstream EV isolation while allowing substantial variability in upstream cellular conditions would therefore leave part of the experimental system uncontrolled.
This provides a direct conceptual bridge between EV research and advanced cell culture engineering.
From Controlled Cell Culture to Reproducible EV Research
For cell-derived EV workflows, reproducibility should be considered as an end-to-end process:
Cell culture → EV production → EV isolation → molecular extraction → omics profiling → data integration → biomarker identification.
Every stage can introduce variability.
As EV research moves toward increasingly sensitive molecular analyses, the requirements placed on upstream culture systems are therefore likely to become more stringent.
A controlled cell culture platform should ideally provide reproducible culture conditions, homogeneous distribution within the culture environment, limited uncontrolled mechanical stress, and a predictable route toward larger experimental volumes.
This becomes especially relevant when EV research involves sensitive cell populations or when researchers need to transition from exploratory experiments toward larger and more standardized production workflows.
Where SoftXS™ Fits into This Scientific Landscape
This upstream challenge is directly relevant to the technological principles behind Cellura’s SoftXS™ platform.
SoftXS™ is a bladeless bioreactor technology designed to generate homogeneous culture conditions through controlled vessel motion rather than conventional impeller-based mechanical agitation.
The platform has been developed around several principles relevant to advanced cellular workflows: low hydrodynamic stress, homogeneous suspension, reproducibility, compatibility with fragile biological materials, and scalability. Cellura’s existing work has demonstrated the platform with multiple cellular systems, including iPSCs, NK cells, organoids, and microcarrier-based cultures.
This creates an interesting technological perspective for EV research.
A controlled, low-shear culture environment could provide a platform in which EV-producing cells are maintained under reproducible upstream conditions before vesicle isolation and molecular characterization.
However, the distinction between technological relevance and demonstrated EV performance is essential.
The EV review does not evaluate SoftXS™, and the Cellura material provided here does not establish that SoftXS™ increases EV yield, modifies EV cargo, improves biomarker sensitivity, or produces more homogeneous EV populations. Those questions require dedicated experimental studies.
The scientifically relevant opportunity is therefore to investigate them.
Low-Shear Cell Culture as an Emerging EV Research Question
The relationship between culture hydrodynamics and EV biology represents a particularly interesting area for future investigation.
Because EV composition can respond to cellular and environmental states, understanding whether different culture conditions influence EV secretion, heterogeneity, molecular cargo, and downstream omics signatures could become important for both fundamental research and bioprocess development.
For Cellura, this suggests several experimentally testable questions:
- How does the culture environment influence EV concentration and size distribution?
- Are transcriptomic, proteomic, metabolomic, or lipidomic EV signatures preserved across scale?
- Does hydrodynamic stress influence the molecular cargo of secreted EV populations?
- Can a standardized culture platform facilitate the transition from exploratory EV research toward larger-scale EV production and characterization?
These should currently be considered research questions rather than established SoftXS™ claims.
Answering them experimentally could help connect two rapidly developing fields: advanced cell bioprocessing and extracellular-vesicle multi-omics.
From Bulk EV Analysis to Single-EV Resolution
The importance of upstream reproducibility may become even greater as EV analytics move toward single-vesicle resolution.
Traditional bulk analysis averages molecular information across heterogeneous EV populations. This can obscure rare subpopulations carrying potentially important biological information.
Emerging single-EV technologies aim to characterize individual vesicles or defined subpopulations with much greater molecular resolution.
At this level of analytical sensitivity, subtle variations introduced during sample generation, culture, isolation, or processing could become increasingly visible.
The transition from bulk EV analysis to single-EV profiling therefore strengthens the case for controlling not only downstream analytical workflows but the entire experimental chain.
Toward an End-to-End Framework for EV Precision Medicine
The future of EV biomarkers will likely depend on the convergence of several technologies rather than a single breakthrough.
Advanced cell culture systems can address upstream biological reproducibility.
Standardized EV isolation and characterization can improve sample comparability.
Transcriptomics, proteomics, metabolomics, and lipidomics can reveal complementary molecular dimensions.
Multi-omics integration can reconstruct relationships across these layers.
Artificial intelligence can help identify multidimensional signatures and build predictive models.
And increasingly sensitive single-EV technologies may reveal molecular heterogeneity that bulk measurements cannot resolve.
Together, these developments point toward an end-to-end approach in which the biological system and analytical system are treated as parts of the same workflow.
Looking Ahead
Extracellular vesicles are emerging as far more than microscopic carriers released by cells. They represent dynamic molecular interfaces between cellular biology, disease mechanisms, biomarker discovery, and precision medicine.
Their ability to transport protected proteins, RNAs, metabolites, and lipids makes them particularly attractive for minimally invasive molecular profiling. Multi-omics approaches are now beginning to reveal the full complexity of this cargo, while AI offers increasingly powerful strategies for identifying relationships across large molecular datasets.
Yet greater analytical sophistication also increases the importance of experimental control.
For cell-derived EV research, standardization should not begin at EV isolation. It should begin with the cells themselves.
This creates a relevant scientific frontier for advanced bioprocess technologies such as SoftXS™. Rather than assuming that low-shear culture improves EV production, the next step is to determine experimentally how controlled hydrodynamic environments influence EV secretion, molecular composition, heterogeneity, and reproducibility across scale.
As EV research progresses from single biomarkers toward multi-omics signatures, single-vesicle analysis, and AI-assisted precision medicine, connecting controlled cell production with rigorous downstream molecular characterization could become an important component of the next generation of extracellular-vesicle research.
Scientific background: Unlocking the Power of Extracellular Vesicles: Multi-Omics Integration for Cancer Biomarker Discovery. 2026.
